diff --git a/config/application.yml.default b/config/application.yml.default
index bc7af40f6cf..157004e3980 100644
--- a/config/application.yml.default
+++ b/config/application.yml.default
@@ -157,8 +157,8 @@ identity_pki_local_dev: false
 idv_account_verified_email_campaign_id: '20241028'
 idv_acuant_sdk_upgrade_a_b_testing_enabled: false
 idv_acuant_sdk_upgrade_a_b_testing_percent: 50
-idv_acuant_sdk_version_alternate: '11.9.2'
-idv_acuant_sdk_version_default: '11.9.3'
+idv_acuant_sdk_version_alternate: '11.9.3'
+idv_acuant_sdk_version_default: '11.9.3.508'
 idv_attempt_window_in_hours: 6
 idv_available: true
 idv_contact_phone_number: (844) 555-5555
diff --git a/public/acuant/11.9.2/AcuantCamera.min.js b/public/acuant/11.9.2/AcuantCamera.min.js
deleted file mode 100644
index ab037366c8e..00000000000
--- a/public/acuant/11.9.2/AcuantCamera.min.js
+++ /dev/null
@@ -1 +0,0 @@
-var AcuantCameraUI=function(){"use strict";let e=null,t=null,i=null,a=null,n={start:function(e,i){s=e.onError,i&&(g=i,g.text.hasOwnProperty("BIG_DOCUMENT")||(g.text.BIG_DOCUMENT="TOO CLOSE"));AcuantCamera.isCameraSupported?r||(r=!0,v(),function(e){let i=0,n=(new Date).getTime();a=document.getElementById("acuant-camera"),a&&a.addEventListener("acuantcameracreated",E);AcuantCamera.start((a=>{!function(e,t){if(t>=3)return!0;{let t=(new Date).getTime()-e;return t<w&&(w=t),w<p}}(n,i)?(i+=1,n=(new Date).getTime(),AcuantCamera.setRepeatFrameProcessor()):w<p?function(e,t){if(e){if(o&&-1===o.state)return;t.onFrameAvailable&&t.onFrameAvailable(e),o=e,o.state===AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT?null===l&&(c=(new Date).getTime(),function(e){null===l&&(l=setTimeout((function(){o.state=h,T(e,"AUTO")}),f))}(t)):(C(),c=null)}else C(),c=null}(a,e):function(e){o={state:m},t.addEventListener("click",b,!1),t.callback=e}(e)}),e,A)}(e)):A("Camera not supported.",AcuantJavascriptWebSdk.START_FAIL_CODE)},end:y},r=!1,o=null,c=null,d=null,s=null,l=null,u=null;const h=-1,m=-2;let g={text:{NONE:"ALIGN",SMALL_DOCUMENT:"MOVE CLOSER",GOOD_DOCUMENT:null,BIG_DOCUMENT:"TOO CLOSE",CAPTURING:"CAPTURING",TAP_TO_CAPTURE:"TAP TO CAPTURE"}};const f=3e3,p=400;let w=Number.MAX_VALUE;function v(){o=null,c=null,d=null}function y(){v(),r=!1,l&&(clearTimeout(l),l=null),u&&(clearTimeout(u),u=null),e&&(e.removeEventListener("play",x,0),e=null),a&&a.removeEventListener("acuantcameracreated",E),t&&(t.removeEventListener("click",b),t.getContext("2d").clearRect(0,0,t.width,t.height)),AcuantCamera.end()}function C(){l&&(clearTimeout(l),l=null)}function b(e){T(e.currentTarget.callback,"TAP")}function E(){e=document.getElementById("acuant-player"),t=document.getElementById("acuant-ui-canvas"),i=t.getContext("2d"),t.setAttribute("role","img"),e.addEventListener("play",x,0)}function T(e,t){AcuantCamera.triggerCapture((i=>{y(),e.onCaptured(i),AcuantCamera.evaluateImage(i.data,i.width,i.height,i.isPortraitOrientation,t,(t=>{e.onCropped(t)}))}))}function A(e,t){y(),s&&s(e,t),s=null}function x(){!function a(){e&&!e.paused&&!e.ended&&r&&(!function(){if(i.clearRect(0,0,t.width,t.height),o)if(o.state===h)I("#00ff00"),D("rgba(0, 255, 0, 0.2)"),O(g.text.CAPTURING,.05,"#00ff00",!1);else if(o.state===m)I("#000000"),O(g.text.TAP_TO_CAPTURE);else if(o.state===AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT)if(I("#ffff00"),D("rgba(255, 255, 0, 0.2)"),g.text.GOOD_DOCUMENT)O(g.text.GOOD_DOCUMENT,.09,"#ff0000",!1);else{let e=Math.ceil((f-((new Date).getTime()-c))/1e3);e<=0&&(e=1),O(e+"...",.09,"#ff0000",!1)}else o.state===AcuantCamera.DOCUMENT_STATE.SMALL_DOCUMENT?(I("#ff0000"),O(g.text.SMALL_DOCUMENT)):o.state===AcuantCamera.DOCUMENT_STATE.BIG_DOCUMENT?(I("#ff0000"),O(g.text.BIG_DOCUMENT)):(I("#000000"),O(g.text.NONE));else I("#000000"),O(g.text.NONE)}(),u=setTimeout(a,100))}()}function O(e,t=.04,a="#ffffff",n=!0){let r=k(),o=window.orientation,c=i.measureText(e),d=.01*Math.max(r.width,r.height),s=.02*Math.max(r.width,r.height),l=(r.height-s-c.width)/2,u=-(r.width/2-d),h=90;0!==o&&(h=0,l=(r.width-d-c.width)/2,u=r.height/2-s+.04*Math.max(r.width,r.height)),i.rotate(h*Math.PI/180),n&&(i.fillStyle="rgba(0, 0, 0, 0.5)",i.fillRect(Math.round(l-d),Math.round(u+d),Math.round(c.width+s),-Math.round(.05*Math.max(r.width,r.height)))),i.font=(Math.ceil(Math.max(r.width,r.height)*t)||0)+"px Sans-serif",i.fillStyle=a,i.fillText(e,l,u),S(e),i.rotate(-h*Math.PI/180)}const S=e=>{d||(d=document.createElement("p"),d.id="doc-state-text",d.style.height="1px",d.style.width="1px",d.style.margin="-1px",d.style.overflow="hidden",d.style.position="absolute",d.style.whiteSpace="nowrap",d.setAttribute("role","alert"),d.setAttribute("aria-live","assertive"),t.parentNode.insertBefore(d,t)),d.innerHTML!=e&&(d.innerHTML=e)};function k(){return{height:t.height,width:t.width}}function M(e,t){let a=window.orientation,n=k(),r=.08*n.width,o=.07*n.height;switch(0!==a&&(r=.07*n.width,o=.08*n.height),t.toString()){case"1":r=-r;break;case"2":r=-r,o=-o;break;case"3":o=-o}!function(e,t,a){i.beginPath();const n=Math.round(e.x),r=Math.round(e.y);i.moveTo(n,r),i.lineTo(Math.round(n+t),r),i.moveTo(n,r),i.lineTo(n,Math.round(r+a)),i.stroke()}(e,r,o)}function D(e){if(o&&o.points&&4===o.points.length){i.beginPath(),i.moveTo(Math.round(o.points[0].x),Math.round(o.points[0].y));for(let e=1;e<o.points.length;e++)i.lineTo(Math.round(o.points[e].x),Math.round(o.points[e].y));i.fillStyle=e,i.strokeStyle="rgba(0, 0, 0, 0)",i.stroke(),i.fill()}}function I(e){let t=k();if(i.lineWidth=Math.ceil(.0025*Math.max(t.width,t.height))||1,i.strokeStyle=e,o&&o.points&&(-1===o.state||o.state===AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT))for(let e in o.points)M(o.points[e],e);else{let e,i,a,n,r={x:t.width/2,y:t.height/2};t.width>t.height?(a=.85*t.width,n=.85*t.width/1.5887,n>.85*t.height&&(a=a/n*.85*t.height,n=.85*t.height)):(a=.85*t.height/1.5887,n=.85*t.height,a>.85*t.width&&(n=n/a*.85*t.width,a=.85*t.width)),e=a/2,i=n/2,[{x:r.x-e,y:r.y-i},{x:r.x+e,y:r.y-i},{x:r.x+e,y:r.y+i},{x:r.x-e,y:r.y+i}].forEach(((e,t)=>{M(e,t)}))}}return n}(),AcuantCamera=(()=>{"use strict";let e=null,t=null,i=null,a=null,n=null,r=null;const o={NO_DOCUMENT:0,SMALL_DOCUMENT:1,BIG_DOCUMENT:2,GOOD_DOCUMENT:3},c={NONE:0,ID:1,PASSPORT:2},d=700,s=1920;let l,u,h=null,m=null,g=null,f=!1,p=!1,w=null,v={start:S,startManualCapture:k,triggerCapture:function(t){let i,a;try{if(0==e.videoWidth)throw"width 0";n.width=e.videoWidth,n.height=e.videoHeight,r.drawImage(e,0,0,n.width,n.height),i=r.getImageData(0,0,n.width,n.height),r.clearRect(0,0,n.width,n.height),a=window.matchMedia("(orientation: portrait)").matches}catch(e){return void ie()}t({data:i,width:n.width,height:n.height,isPortraitOrientation:a})},end:W,DOCUMENT_STATE:o,ACUANT_DOCUMENT_TYPE:c,isCameraSupported:"mediaDevices"in navigator&&function(){let e=!1;return function(t){(/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino|android|playbook|silk/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4)))&&(e=!0)}(navigator.userAgent||navigator.vendor||window.opera),(e||C())&&!y()}(),isIOSWebview:function(){const e=window.navigator.standalone,t=window.navigator.userAgent.toLowerCase(),i=/safari/.test(t);return/iphone|ipod|ipad/.test(t)&&!i&&!e}(),isIOS:C,setRepeatFrameProcessor:Z,evaluateImage:G};function y(){return navigator.userAgent.toLowerCase().indexOf("firefox")>-1}function C(){return/iPad|iPhone|iPod/.test(navigator.platform)&&_()[0]>=13||"MacIntel"===navigator.platform&&navigator.maxTouchPoints>1}function b(){let e;e=u||navigator.userAgent;const t=e.match(/SM-[N|G|S]\d{3}/);if(!t)return!1;const i=parseInt(t[0].match(/\d{3}/)[0],10),a=e.match(/SM-S\d{3}/)?900:970;return!isNaN(i)&&i>=a}const E=function(){let e={frameScale:1,primaryConstraints:{video:{facingMode:{exact:"environment"},aspectRatio:4/3,resizeMode:"none"}},fixedHeight:null,fixedWidth:null};C()?L()?(e.primaryConstraints.video.aspectRatio=1*Math.max(window.innerWidth,window.innerHeight)/Math.min(window.innerWidth,window.innerHeight),e.primaryConstraints.video.height={min:1440,ideal:2880}):N()?e.primaryConstraints.video.width={min:s,ideal:s}:e.primaryConstraints.video.height={min:1440,ideal:1440}:e.primaryConstraints.video.height={min:1440,ideal:1440};return e}();function T(t){j().then((()=>{f=!0,e.srcObject=t,function(){b()&&document.addEventListener("visibilitychange",z);window.addEventListener("resize",R),e&&(e.addEventListener("play",J),e.addEventListener("loadedmetadata",F))}(),e.play()}))}function A(e,t){document.cookie="AcuantCameraHasFailed="+t,W(),m&&"function"==typeof m?m(e,t):(console.error("No error callback set. Review implementation."),console.error(e,t))}function x(){return new Promise((e=>{navigator.mediaDevices.enumerateDevices().then((t=>{const i={suffix:void 0,device:void 0},a=t.find((e=>"Back Dual Wide Camera"===e.label));(function(){let e=_();return e&&-1!=e&&e.length>=1&&e[0]>=16&&e[1]>=4})()&&a?i.device=a:t.filter((e=>"videoinput"===e.kind)).forEach((e=>{if(t=e.label,["rear","back","rück","arrière","trasera","trás","traseira","posteriore","后面","後面","背面","задней","الخلفية","후","arka","achterzijde","หลัง","baksidan","bagside","sau","bak","tylny","takakamera","belakang","אחורית","πίσω","spate","hátsó","zadní","darrere","zadná","задня","stražnja","belakang","बैक"].some((e=>t.includes(e)))){let t=e.label.split(","),a=parseInt(t[0][t[0].length-1]);(a||0===a)&&(void 0===i.suffix||i.suffix>a)&&(i.suffix=a,i.device=e)}var t})),e(i.device)})).catch((()=>{e()}))}))}function O(e,t=0){0===t&&l.dispatchEvent(new Event("acuantcameracreated"));const i=Boolean(e.video.deviceId);navigator.mediaDevices.getUserMedia(e).then((e=>{!i&&t<2?x().then((i=>{i&&i.deviceId!==e.getVideoTracks()[0].getSettings().deviceId?(E.primaryConstraints.video.deviceId=i.deviceId,U(e),O(E.primaryConstraints,t++)):T(e)})):T(e)})).catch((e=>{A(e,AcuantJavascriptWebSdk.START_FAIL_CODE)}))}function S(a,o,c){if(c&&(m=c),function(){let e="AcuantCameraHasFailed=";return decodeURIComponent(document.cookie).includes(e)}())return c("Live capture has previously failed and was called again. User was sent to manual capture.",AcuantJavascriptWebSdk.REPEAT_FAIL_CODE),void k(o);l=document.getElementById("acuant-camera"),l?(l.style.position="relative",l.innerHTML='<video id="acuant-player" autoplay playsinline style="display:block; image-orientation:none; object-fit: fill;"></video><canvas id="acuant-ui-canvas" style="display:block; position:absolute; top:0; left:0;"></canvas>',e=document.getElementById("acuant-player"),n=document.createElement("canvas"),r=n.getContext("2d",{willReadFrequently:!C()}),t=document.getElementById("acuant-ui-canvas"),f?A("already started.",AcuantJavascriptWebSdk.START_FAIL_CODE):e&&t?(i=t.getContext("2d"),a&&(h=a),navigator.userAgentData&&navigator.userAgentData.getHighEntropyValues?navigator.userAgentData.getHighEntropyValues(["model"]).then((e=>{"string"==typeof e?u=e:"string"==typeof e.model&&(u=e.model)})).finally((()=>{b()?E.primaryConstraints.video.zoom=2:P()&&(E.primaryConstraints.video.zoom=1.6),O(E.primaryConstraints)})):(b()?E.primaryConstraints.video.zoom=2:P()&&(E.primaryConstraints.video.zoom=1.6),O(E.primaryConstraints))):A("Missing HTML elements.",AcuantJavascriptWebSdk.START_FAIL_CODE)):A("Expected div with 'acuant-camera' id",AcuantJavascriptWebSdk.START_FAIL_CODE)}function k(e){g=e,a||(a=document.createElement("input"),a.type="file",a.capture="environment",a.accept="image/*",a.onclick=function(e){e&&e.target&&(e.target.value="")}),a.onchange=D,a.click()}let M=-1;function D(e){n=document.createElement("canvas"),r=n.getContext("2d"),r.mozImageSmoothingEnabled=!1,r.webkitImageSmoothingEnabled=!1,r.msImageSmoothingEnabled=!1,r.imageSmoothingEnabled=!1;let t=e.target,i=new FileReader;const a=e.target.files[0]&&e.target.files[0].name&&e.target.files[0].name.toLowerCase().endsWith(".heic");i.onload=a?e=>{var t;(t=e.target.result,new Promise(((e,i)=>{const a=window["magick-wasm"];a?a.initializeImageMagick().then((()=>{a.ImageMagick.read(new Uint8Array(t),(t=>{const{width:i,height:a}=I(t.width,t.height);n.width=i,n.height=a,t.resize(i,a),t.writeToCanvas(n);const o=r.getImageData(0,0,i,a);e({data:o,width:i,height:a})}))})):i({error:"HEIC image processing failed. Please make sure Image Magick scripts were integrated as expected.",code:AcuantJavascriptWebSdk.HEIC_NOT_SUPPORTED_CODE})}))).then((e=>{M=6,g.onCaptured(e),G(e.data,e.width,e.height,!1,"MANUAL",g.onCropped)})).catch((e=>g.onError(e.error,e.code)))}:e=>{M=function(e){const t=new DataView(e.target.result);if(65496!=t.getUint16(0,!1))return-2;const i=t.byteLength;let a=2;for(;a<i;){if(t.getUint16(a+2,!1)<=8)return-1;const e=t.getUint16(a,!1);if(a+=2,65505==e){if(1165519206!=t.getUint32(a+=2,!1))return-1;const e=18761==t.getUint16(a+=6,!1);a+=t.getUint32(a+4,e);const i=t.getUint16(a,e);a+=2;for(let n=0;n<i;n++)if(274==t.getUint16(a+12*n,e))return t.getUint16(a+12*n+8,e)}else{if(65280!=(65280&e))break;a+=t.getUint16(a,!1)}}return-1}(e);let t=document.createElement("img");t.onload=()=>{const{width:e,height:i}=I(t.width,t.height);n.width=e,n.height=i,r.drawImage(t,0,0,e,i);const a=r.getImageData(0,0,e,i);r.clearRect(0,0,e,i),t.remove(),g.onCaptured({data:a,width:e,height:i}),G(a,e,i,!1,"MANUAL",g.onCropped)},t.src="data:image/jpeg;base64,"+ae(e.target.result)},t&&t.files[0]&&i.readAsArrayBuffer(t.files[0])}function I(e,t){let i=2560,a=1920;N()&&(i=s,a=Math.floor(1440));if((e>t?e:t)>i)if(e<t){a=i,i=a/(t/e)}else{a=i/(e/t)}else i=e,a=t;return{width:i,height:a}}function U(e){e.getTracks().forEach((e=>{e.stop()}))}function W(){f=!1,p=!1,M=-1,w&&(clearTimeout(w),w=null),function(){b()&&document.removeEventListener("visibilitychange",z);window.removeEventListener("resize",R),e&&(e.removeEventListener("play",J),e.removeEventListener("loadedmetadata",F))}(),e&&(e.pause(),e.srcObject&&U(e.srcObject),e=null),l&&(l.innerHTML=""),a&&(a.remove(),a=null)}function _(){if(/iP(hone|od|ad)/.test(navigator.platform))try{const e=navigator.appVersion.match(/OS (\d+)_(\d+)_?(\d+)?/);return[parseInt(e[1],10),parseInt(e[2],10),parseInt(e[3]||0,10)]}catch(e){return-1}return-1}function N(){let e=_();return e&&-1!=e&&e.length>=1&&15==e[0]}function P(){if(function(){let e=_();return e&&-1!=e&&e.length>=1&&e[0]>=17}()){let e=[screen.width,screen.height],t=Math.max(...e),i=Math.min(...e);if(852==t&&393==i)return!0;if(844==t&&390==i)return!0}return!1}function L(){let e=decodeURIComponent(document.cookie);if(e.includes("AcuantForceRegularCapture=true"))return!1;if(e.includes("AcuantForceDistantCapture=true"))return!0;if(function(){let e=_();return e&&-1!=e&&e.length>=1&&16==e[0]&&e[1]<4}()){let e=[screen.width,screen.height],t=Math.max(...e),i=Math.min(...e);if(852==t&&393==i)return!0;if(932==t&&430==i)return!0;if(844==t&&390==i)return!0;if(926==t&&428==i)return!0}return!1}function R(){i.clearRect(0,0,t.width,t.height),e&&(C()&&(function(){const e=window.navigator.userAgent,t=e.indexOf("WebKit")>-1,i=e.indexOf("CriOS")>-1;return t&&i&&C()}()||function(){let e=_();return e&&-1!=e&&e.length>=2&&14==e[0]&&e[1]>=4}())?(W(),S()):H())}function H(){let i,a;if(e.videoWidth<e.videoHeight?E.canvasScale=e.videoHeight/e.videoWidth:E.canvasScale=e.videoWidth/e.videoHeight,i=window.innerWidth,a=window.innerHeight,E.fixedHeight||(E.fixedHeight=l.clientHeight),E.fixedWidth||(E.fixedWidth=l.clientWidth),E.fixedHeight&&E.fixedWidth&&E.fixedWidth>E.fixedHeight&&(E.fixedHeight=l.clientWidth,E.fixedWidth=l.clientHeight),window.matchMedia("(orientation: portrait)").matches){if(E.fixedWidth)E.fixedWidth<i&&(i=E.fixedWidth),a=i*E.canvasScale;else if(E.fixedHeight){let e=E.fixedHeight/E.canvasScale;e<i&&(a=E.fixedHeight,i=e)}}else if(E.fixedWidth)E.fixedWidth<a&&(a=E.fixedWidth),i=a*E.canvasScale;else if(E.fixedHeight){let e=E.fixedHeight/E.canvasScale;e<a&&(i=E.fixedHeight,a=e)}e.width=t.width=i,e.height=t.height=a}function z(){!a&&f&&"visible"===document.visibilityState&&(W(),S())}function B(){e&&!e.paused&&(Z(),w=setTimeout(B,100))}function F(){e.videoWidth+e.videoHeight<1e3?(W(),A("Camera not supported",AcuantJavascriptWebSdk.START_FAIL_CODE)):(H(),B())}function J(){w&&(clearTimeout(w),p=!1,B())}function G(e,t,i,a,n,r){AcuantJavascriptWebSdk.singleWorkerModel?async function(e,t,i,a,n){let r={isPortraitOrientation:n};await j();const o=await q(e,t,i);if(!o)return null;r={...r,...o},V(),await new Promise((e=>{AcuantJavascriptWebSdk.startMetricsWorker(e)}));const c=await Q(e,t,i),d=await K(r.image);r={...r,...c,...d},AcuantJavascriptWebSdk.endMetricsWorker();const{imageBase64:s,imageBytes:l}=$(r,a);r.image.bytes=l;const u=await Y(s);r.image.barcodeText=u;const h=await ee(r,a,s);await j();const m=await X(h);return r.image.data=m,V(),r}(e,t,i,n,a).then(r):async function(e,t,i,a,n){let r={isPortraitOrientation:n};const[o,c]=await Promise.all([q(e,t,i),Q(e,t,i)]);if(!o)return null;r={...r,...o};const d=await K(r.image);r={...r,...c,...d};const{imageBase64:s,imageBytes:l}=$(r,a);r.image.bytes=l;const u=await Y(s);r.image.barcodeText=u;const h=await ee(r,a,s),m=await X(h);return r.image.data=m,r}(e,t,i,n,a).then(r)}function j(){return new Promise((e=>{AcuantJavascriptWebSdk.startImageWorker(e)}))}function V(){AcuantJavascriptWebSdk.endImageWorker()}function q(e,t,i){return new Promise((a=>{AcuantJavascriptWebSdk.crop(e,t,i,{onSuccess:({image:e,dpi:t,cardType:i})=>a({image:e,dpi:t,cardType:i}),onFail:a})}))}function Q(e,t,i){return new Promise((a=>{AcuantJavascriptWebSdk.moire(e,t,i,{onSuccess:(e,t)=>a({moire:e,moireraw:t}),onFail:()=>a({moire:-1,moireraw:-1})})}))}function K(e){return new Promise((t=>{AcuantJavascriptWebSdk.metrics(e,e.width,e.height,{onSuccess:(e,i)=>t({sharpness:e,glare:i}),onFail:()=>t({sharpness:-1,glare:-1})})}))}function X(e){return new Promise((t=>{const i=function(e){const t=window.atob(e.split("base64,")[1]),i=t.length,a=new Uint8Array(new ArrayBuffer(i));for(let e=0;e<i;e++)a[e]=t.charCodeAt(e);return a}(e);AcuantJavascriptWebSdk.sign(i,{onSuccess:e=>t("data:image/jpeg;base64,"+ae(e)),onFail:t})}))}async function Y(e){if(!document.getElementById(AcuantJavascriptWebSdk.BARCODE_READER_ID))return null;try{return await function(e){let t=e.split(","),i=t[0].match(/:(.*?);/)[1],a=atob(t[1]),n=a.length,r=new Uint8Array(n);for(;n--;)r[n]=a.charCodeAt(n);const o=new File([r],"imageFile",{type:i});return new Html5Qrcode(AcuantJavascriptWebSdk.BARCODE_READER_ID,{formatsToSupport:[Html5QrcodeSupportedFormats.PDF_417]}).scanFile(o,!1)}(e)}catch{return null}}function Z(){if(!f||p)return;if(0==e.videoWidth)return void ie();p=!0;let t=Math.max(e.videoWidth,e.videoHeight),i=Math.min(e.videoWidth,e.videoHeight),a=0,s=0;if(t>d&&i>500?e.videoWidth>=e.videoHeight?(E.frameScale=d/e.videoWidth,s=d,a=e.videoHeight*E.frameScale):(E.frameScale=d/e.videoHeight,s=e.videoWidth*E.frameScale,a=d):(E.frameScale=1,s=e.videoWidth,a=e.videoHeight),s==n.width&&a==n.height||(n.width=s,n.height=a),f){let t;try{r.drawImage(e,0,0,e.videoWidth,e.videoHeight,0,0,n.width,n.height),t=r.getImageData(0,0,n.width,n.height),r.clearRect(0,0,n.width,n.height)}catch(e){return void ie()}!function(t,i,a){AcuantJavascriptWebSdk.detect(t,i,a,{onSuccess:function(t){if(!n||!e||e.paused||e.ended)return;t.points.forEach((t=>{void 0!==t.x&&void 0!==t.y&&(t.x=t.x/E.frameScale*e.width/e.videoWidth,t.y=t.y/E.frameScale*e.height/e.videoHeight)}));const i=Math.min(t.dimensions.width,t.dimensions.height)/Math.min(n.width,n.height),a=Math.max(t.dimensions.width,t.dimensions.height)/Math.max(n.width,n.height),r=2==t.type;let d=.8,s=.85,l=.6,u=.65;r&&(d=.9,s=.95),C()&&(l=.65,u=.7,L()?r?(d=.72,s=.77,l=.22,u=.28):(d=.41,s=.45,l=.22,u=.28):r&&(d=.95,s=1,l=.7,u=.75));const m=!t.isCorrectAspectRatio||i<l&&a<u,g=i>=d||a>=s;t.type===c.NONE?t.state=o.NO_DOCUMENT:t.state=g?o.BIG_DOCUMENT:m?o.SMALL_DOCUMENT:o.GOOD_DOCUMENT,h(t),p=!1},onFail:function(){if(!n||!e||e.paused||e.ended)return;let t={};t.state=o.NO_DOCUMENT,h(t),p=!1}})}(t,n.width,n.height)}}function $({image:e,cardType:t,isPortraitOrientation:i},a){n&&r||(n=document.createElement("canvas"),r=n.getContext("2d")),n.width=e.width,n.height=e.height;let o=r.createImageData(e.width,e.height);!function(e,t){for(let i=0;i<e.length;++i)e[i]=t[i]}(o.data,e.data),r.putImageData(o,0,0);let c="AUTO"!=a&&"TAP"!=a;2!==t&&(!c&&i||c&&6===M)&&(r.rotate(Math.PI),r.drawImage(n,0,0,-n.width,-n.height));const d=r.getImageData(0,0,n.width,n.height).data,s="undefined"!=typeof acuantConfig&&"number"==typeof acuantConfig.jpegQuality?acuantConfig.jpegQuality:1,l=n.toDataURL("image/jpeg",s);return r.clearRect(0,0,n.width,n.height),n&&(r=null,n.remove(),n=null),{imageBase64:l,imageBytes:d}}async function ee({dpi:e,cardType:t,sharpness:i,glare:a,moire:n,moireraw:r},o,c){const d=await new Promise((e=>{AcuantJavascriptWebSdk.getCvmlVersion({onSuccess:t=>{e(t)},onFail:()=>{e("unknown")}})})),s=JSON.stringify({cvml:{cropping:{iscropped:!0,dpi:e,idsize:2===t?"ID3":"ID1",elapsed:-1},sharpness:{normalized:i,elapsed:-1},moire:{normalized:n,raw:r,elapsed:-1},glare:{normalized:a,elapsed:-1},version:d},device:{version:te(),capturetype:o}});return AcuantJavascriptWebSdk.addMetadata(c,{imageDescription:s,dateTimeOriginal:(new Date).toUTCString()})}function te(){const e=navigator.userAgent;let t,i=e.match(/(opera|chrome|safari|firefox|msie|trident(?=\/))\/?\s*(\d+)/i)||[];return/trident/i.test(i[1])?(t=/\brv[ :]+(\d+)/g.exec(e)||[],"IE "+(t[1]||"")):"Chrome"===i[1]&&(t=e.match(/\b(OPR|Edge)\/(\d+)/),null!=t)?t.slice(1).join(" ").replace("OPR","Opera"):(i=i[2]?[i[1],i[2]]:[navigator.appName,navigator.appVersion,"-?"],null!=(t=e.match(/version\/(\d+)/i))&&i.splice(1,1,t[1]),i.join(" "))}function ie(){N()||navigator.maxTouchPoints&&navigator.maxTouchPoints>=2&&/MacIntel/.test(navigator.platform)?A("Camera capture failed due to unexpected sequence break. This usually indicates the camera closed or froze unexpectedly. In iOS 15+ this is intermittently occurs due to a GPU Highwater failure. Swap to manual capture until the user fully reloads the browser. Attempting to continue to use live capture can lead to further Highwater errors and can cause to OS to cut off the webpage.",AcuantJavascriptWebSdk.SEQUENCE_BREAK_CODE):A("Camera capture failed due to unexpected sequence break. This usually indicates the camera closed or froze unexpectedly. Swap to manual capture until the user fully reloads the browser.",AcuantJavascriptWebSdk.SEQUENCE_BREAK_CODE)}function ae(e){let t="";const i=new Uint8Array(e),a=i.byteLength;for(let e=0;e<a;e++)t+=String.fromCharCode(i[e]);return window.btoa(t)}return v})();
\ No newline at end of file
diff --git a/public/acuant/11.9.2/AcuantPassiveLiveness.min.js b/public/acuant/11.9.2/AcuantPassiveLiveness.min.js
deleted file mode 100644
index 0b5e602e3ec..00000000000
--- a/public/acuant/11.9.2/AcuantPassiveLiveness.min.js
+++ /dev/null
@@ -1,261 +0,0 @@
-!function(e){function t(t){for(var n,a,s=t[0],i=t[1],o=0,l=[];o<s.length;o++)a=s[o],Object.prototype.hasOwnProperty.call(r,a)&&r[a]&&l.push(r[a][0]),r[a]=0;for(n in i)Object.prototype.hasOwnProperty.call(i,n)&&(e[n]=i[n]);for(u&&u(t);l.length;)l.shift()()}var n={},r={0:0};function a(t){if(n[t])return n[t].exports;var r=n[t]={i:t,l:!1,exports:{}};return e[t].call(r.exports,r,r.exports,a),r.l=!0,r.exports}a.e=function(){return Promise.resolve()},a.m=e,a.c=n,a.d=function(e,t,n){a.o(e,t)||Object.defineProperty(e,t,{enumerable:!0,get:n})},a.r=function(e){"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},a.t=function(e,t){if(1&t&&(e=a(e)),8&t)return e;if(4&t&&"object"==typeof e&&e&&e.__esModule)return e;var n=Object.create(null);if(a.r(n),Object.defineProperty(n,"default",{enumerable:!0,value:e}),2&t&&"string"!=typeof e)for(var r in e)a.d(n,r,function(t){return e[t]}.bind(null,r));return n},a.n=function(e){var t=e&&e.__esModule?function(){return e.default}:function(){return e};return a.d(t,"a",t),t},a.o=function(e,t){return Object.prototype.hasOwnProperty.call(e,t)},a.p="",a.oe=function(e){throw console.error(e),e};var s=window.webpackJsonp=window.webpackJsonp||[],i=s.push.bind(s);s.push=t,s=s.slice();for(var o=0;o<s.length;o++)t(s[o]);var u=i;a(a.s=126)}([function(e,t,n){"use strict";n.d(t,"a",(function(){return Z})),n.d(t,"b",(function(){return P})),n.d(t,"c",(function(){return L})),n.d(t,"d",(function(){return v})),n.d(t,"e",(function(){return l})),n.d(t,"f",(function(){return te}));let r,a,s=!1;const i="undefined"!=typeof window?window:{},o=i.document||{head:{}},u={$flags$:0,$resourcesUrl$:"",jmp:e=>e(),raf:e=>requestAnimationFrame(e),ael:(e,t,n,r)=>e.addEventListener(t,n,r),rel:(e,t,n,r)=>e.removeEventListener(t,n,r),ce:(e,t)=>new CustomEvent(e,t)},l=e=>Promise.resolve(e),c=(()=>{try{return new CSSStyleSheet,"function"==typeof(new CSSStyleSheet).replace}catch(e){}return!1})(),h=(e,t,n,r)=>{n&&n.map(([n,r,a])=>{const s=e,i=p(t,a),o=d(n);u.ael(s,r,i,o),(t.$rmListeners$=t.$rmListeners$||[]).push(()=>u.rel(s,r,i,o))})},p=(e,t)=>n=>{try{256&e.$flags$?e.$lazyInstance$[t](n):(e.$queuedListeners$=e.$queuedListeners$||[]).push([t,n])}catch(e){ae(e)}},d=e=>0!=(2&e),f=new WeakMap,m=e=>{const t=e.$cmpMeta$,n=e.$hostElement$,r=t.$flags$,a=(t.$tagName$,()=>{}),s=((e,t,n,r)=>{let a=g(t),s=oe.get(a);if(e=11===e.nodeType?e:o,s)if("string"==typeof s){e=e.head||e;let t,n=f.get(e);n||f.set(e,n=new Set),n.has(a)||(t=o.createElement("style"),t.innerHTML=s,e.insertBefore(t,e.querySelector("link")),n&&n.add(a))}else e.adoptedStyleSheets.includes(s)||(e.adoptedStyleSheets=[...e.adoptedStyleSheets,s]);return a})(n.shadowRoot?n.shadowRoot:n.getRootNode(),t);10&r&&(n["s-sc"]=s,n.classList.add(s+"-h")),a()},g=(e,t)=>"sc-"+e.$tagName$,y={},b=e=>"object"===(e=typeof e)||"function"===e,v=(e,t,...n)=>{let r=null,a=!1,s=!1,i=[];const o=t=>{for(let n=0;n<t.length;n++)r=t[n],Array.isArray(r)?o(r):null!=r&&"boolean"!=typeof r&&((a="function"!=typeof e&&!b(r))&&(r=String(r)),a&&s?i[i.length-1].$text$+=r:i.push(a?x(null,r):r),s=a)};if(o(n),t){const e=t.className||t.class;e&&(t.class="object"!=typeof e?e:Object.keys(e).filter(t=>e[t]).join(" "))}if("function"==typeof e)return e(null===t?{}:t,i,k);const u=x(e,null);return u.$attrs$=t,i.length>0&&(u.$children$=i),u},x=(e,t)=>{const n={$flags$:0,$tag$:e,$text$:t,$elm$:null,$children$:null,$attrs$:null};return n},w={},k={forEach:(e,t)=>e.map(S).forEach(t),map:(e,t)=>e.map(S).map(t).map(I)},S=e=>({vattrs:e.$attrs$,vchildren:e.$children$,vkey:e.$key$,vname:e.$name$,vtag:e.$tag$,vtext:e.$text$}),I=e=>{if("function"==typeof e.vtag){const t=Object.assign({},e.vattrs);return e.vkey&&(t.key=e.vkey),e.vname&&(t.name=e.vname),v(e.vtag,t,...e.vchildren||[])}const t=x(e.vtag,e.vtext);return t.$attrs$=e.vattrs,t.$children$=e.vchildren,t.$key$=e.vkey,t.$name$=e.vname,t},A=(e,t,n,r,a,s)=>{if(n!==r){let o=re(e,t),l=t.toLowerCase();if("class"===t){const t=e.classList,a=N(n),s=N(r);t.remove(...a.filter(e=>e&&!s.includes(e))),t.add(...s.filter(e=>e&&!a.includes(e)))}else if("style"===t){for(const t in n)r&&null!=r[t]||(t.includes("-")?e.style.removeProperty(t):e.style[t]="");for(const t in r)n&&r[t]===n[t]||(t.includes("-")?e.style.setProperty(t,r[t]):e.style[t]=r[t])}else if("ref"===t)r&&r(e);else if(o||"o"!==t[0]||"n"!==t[1]){const i=b(r);if((o||i&&null!==r)&&!a)try{if(e.tagName.includes("-"))e[t]=r;else{let a=null==r?"":r;"list"===t?o=!1:null!=n&&e[t]==a||(e[t]=a)}}catch(e){}null==r||!1===r?!1===r&&""!==e.getAttribute(t)||e.removeAttribute(t):(!o||4&s||a)&&!i&&(r=!0===r?"":r,e.setAttribute(t,r))}else t="-"===t[2]?t.slice(3):re(i,l)?l.slice(2):l[2]+t.slice(3),n&&u.rel(e,t,n,!1),r&&u.ael(e,t,r,!1)}},E=/\s/,N=e=>e?e.split(E):[],C=(e,t,n,r)=>{const a=11===t.$elm$.nodeType&&t.$elm$.host?t.$elm$.host:t.$elm$,s=e&&e.$attrs$||y,i=t.$attrs$||y;for(r in s)r in i||A(a,r,s[r],void 0,n,t.$flags$);for(r in i)A(a,r,s[r],i[r],n,t.$flags$)},T=(e,t,n,a)=>{let s,i,u=t.$children$[n],l=0;if(null!==u.$text$)s=u.$elm$=o.createTextNode(u.$text$);else if(s=u.$elm$=o.createElement(u.$tag$),C(null,u,!1),null!=r&&s["s-si"]!==r&&s.classList.add(s["s-si"]=r),u.$children$)for(l=0;l<u.$children$.length;++l)i=T(e,u,l),i&&s.appendChild(i);return s},R=(e,t,n,r,s,i)=>{let o,u=e;for(u.shadowRoot&&u.tagName===a&&(u=u.shadowRoot);s<=i;++s)r[s]&&(o=T(null,n,s),o&&(r[s].$elm$=o,u.insertBefore(o,t)))},_=(e,t,n,r,a)=>{for(;t<=n;++t)(r=e[t])&&(a=r.$elm$,D(r),a.remove())},F=(e,t)=>e.$tag$===t.$tag$,M=(e,t)=>{const n=t.$elm$=e.$elm$,r=e.$children$,a=t.$children$,s=t.$text$;null===s?(C(e,t,!1),null!==r&&null!==a?((e,t,n,r)=>{let a,s=0,i=0,o=t.length-1,u=t[0],l=t[o],c=r.length-1,h=r[0],p=r[c];for(;s<=o&&i<=c;)null==u?u=t[++s]:null==l?l=t[--o]:null==h?h=r[++i]:null==p?p=r[--c]:F(u,h)?(M(u,h),u=t[++s],h=r[++i]):F(l,p)?(M(l,p),l=t[--o],p=r[--c]):F(u,p)?(M(u,p),e.insertBefore(u.$elm$,l.$elm$.nextSibling),u=t[++s],p=r[--c]):F(l,h)?(M(l,h),e.insertBefore(l.$elm$,u.$elm$),l=t[--o],h=r[++i]):(a=T(t&&t[i],n,i),h=r[++i],a&&u.$elm$.parentNode.insertBefore(a,u.$elm$));s>o?R(e,null==r[c+1]?null:r[c+1].$elm$,n,r,i,c):i>c&&_(t,s,o)})(n,r,t,a):null!==a?(null!==e.$text$&&(n.textContent=""),R(n,null,t,a,0,a.length-1)):null!==r&&_(r,0,r.length-1)):e.$text$!==s&&(n.data=s)},D=e=>{e.$attrs$&&e.$attrs$.ref&&e.$attrs$.ref(null),e.$children$&&e.$children$.map(D)},O=(e,t)=>{const n=e.$hostElement$,s=e.$vnode$||x(null,null),i=(o=t)&&o.$tag$===w?t:v(null,null,t);var o;a=n.tagName,i.$tag$=null,i.$flags$|=4,e.$vnode$=i,i.$elm$=s.$elm$=n.shadowRoot||n,r=n["s-sc"],M(s,i)},L=e=>ee(e).$hostElement$,P=(e,t,n)=>{const r=L(e);return{emit:e=>$(r,t,{bubbles:!!(4&n),composed:!!(2&n),cancelable:!!(1&n),detail:e})}},$=(e,t,n)=>{const r=u.ce(t,n);return e.dispatchEvent(r),r},B=(e,t)=>{t&&!e.$onRenderResolve$&&t["s-p"]&&t["s-p"].push(new Promise(t=>e.$onRenderResolve$=t))},z=(e,t)=>{if(e.$flags$|=16,4&e.$flags$)return void(e.$flags$|=512);B(e,e.$ancestorComponent$);return fe(()=>U(e,t))},U=(e,t)=>{const n=(e.$cmpMeta$.$tagName$,()=>{}),r=e.$lazyInstance$;let a;return t&&(e.$flags$|=256,e.$queuedListeners$&&(e.$queuedListeners$.map(([e,t])=>H(r,e,t)),e.$queuedListeners$=null),a=H(r,"componentWillLoad")),n(),q(a,()=>W(e,r,t))},W=async(e,t,n)=>{const r=e.$hostElement$,a=(e.$cmpMeta$.$tagName$,()=>{}),s=r["s-rc"];n&&m(e);const i=(e.$cmpMeta$.$tagName$,()=>{});V(e,t),s&&(s.map(e=>e()),r["s-rc"]=void 0),i(),a();{const t=r["s-p"],n=()=>j(e);0===t.length?n():(Promise.all(t).then(n),e.$flags$|=4,t.length=0)}},V=(e,t,n)=>{try{t=t.render(),e.$flags$&=-17,e.$flags$|=2,O(e,t)}catch(t){ae(t,e.$hostElement$)}return null},j=e=>{e.$cmpMeta$.$tagName$;const t=e.$hostElement$,n=()=>{},r=e.$lazyInstance$,a=e.$ancestorComponent$;64&e.$flags$?(H(r,"componentDidUpdate"),n()):(e.$flags$|=64,K(t),H(r,"componentDidLoad"),n(),e.$onReadyResolve$(t),a||G()),e.$onRenderResolve$&&(e.$onRenderResolve$(),e.$onRenderResolve$=void 0),512&e.$flags$&&de(()=>z(e,!1)),e.$flags$&=-517},G=e=>{K(o.documentElement),de(()=>$(i,"appload",{detail:{namespace:"fas-web-ui-component-camera"}}))},H=(e,t,n)=>{if(e&&e[t])try{return e[t](n)}catch(e){ae(e)}},q=(e,t)=>e&&e.then?e.then(t):t(),K=e=>e.classList.add("hydrated"),Y=(e,t,n,r)=>{const a=ee(e),s=a.$instanceValues$.get(t),i=a.$flags$,o=a.$lazyInstance$;var u,l;u=n,l=r.$members$[t][0],n=null==u||b(u)?u:4&l?"false"!==u&&(""===u||!!u):2&l?parseFloat(u):1&l?String(u):u,8&i&&void 0!==s||n===s||(a.$instanceValues$.set(t,n),o&&2==(18&i)&&z(a,!1))},X=(e,t,n)=>{if(t.$members$){const r=Object.entries(t.$members$),a=e.prototype;if(r.map(([e,[r]])=>{(31&r||2&n&&32&r)&&Object.defineProperty(a,e,{get(){return t=e,ee(this).$instanceValues$.get(t);var t},set(n){Y(this,e,n,t)},configurable:!0,enumerable:!0})}),1&n){const t=new Map;a.attributeChangedCallback=function(e,n,r){u.jmp(()=>{const n=t.get(e);if(this.hasOwnProperty(n))r=this[n],delete this[n];else if(a.hasOwnProperty(n)&&"number"==typeof this[n]&&this[n]==r)return;this[n]=(null!==r||"boolean"!=typeof this[n])&&r})},e.observedAttributes=r.filter(([e,t])=>15&t[0]).map(([e,n])=>{const r=n[1]||e;return t.set(r,e),r})}}return e},Q=async(e,t,n,r,a)=>{if(0==(32&t.$flags$)){{if(t.$flags$|=32,(a=ie(n)).then){const e=()=>{};a=await a,e()}a.isProxied||(X(a,n,2),a.isProxied=!0);const e=(n.$tagName$,()=>{});t.$flags$|=8;try{new a(t)}catch(e){ae(e)}t.$flags$&=-9,e()}if(a.style){let e=a.style;const t=g(n);if(!oe.has(t)){const r=(n.$tagName$,()=>{});((e,t,n)=>{let r=oe.get(e);c&&n?(r=r||new CSSStyleSheet,r.replace(t)):r=t,oe.set(e,r)})(t,e,!!(1&n.$flags$)),r()}}}const s=t.$ancestorComponent$,i=()=>z(t,!0);s&&s["s-rc"]?s["s-rc"].push(i):i()},Z=(e,t={})=>{const n=()=>{},r=[],a=t.exclude||[],s=i.customElements,l=o.head,c=l.querySelector("meta[charset]"),p=o.createElement("style"),d=[];let f,m=!0;Object.assign(u,t),u.$resourcesUrl$=new URL(t.resourcesUrl||"./",o.baseURI).href,e.map(e=>{e[1].map(t=>{const n={$flags$:t[0],$tagName$:t[1],$members$:t[2],$listeners$:t[3]};n.$members$=t[2],n.$listeners$=t[3];const i=n.$tagName$,o=class extends HTMLElement{constructor(e){super(e),ne(e=this,n),1&n.$flags$&&e.attachShadow({mode:"open"})}connectedCallback(){f&&(clearTimeout(f),f=null),m?d.push(this):u.jmp(()=>(e=>{if(0==(1&u.$flags$)){const t=ee(e),n=t.$cmpMeta$,r=(n.$tagName$,()=>{});if(1&t.$flags$)h(e,t,n.$listeners$);else{t.$flags$|=1;{let n=e;for(;n=n.parentNode||n.host;)if(n["s-p"]){B(t,t.$ancestorComponent$=n);break}}n.$members$&&Object.entries(n.$members$).map(([t,[n]])=>{if(31&n&&e.hasOwnProperty(t)){const n=e[t];delete e[t],e[t]=n}}),Q(0,t,n)}r()}})(this))}disconnectedCallback(){u.jmp(()=>(e=>{if(0==(1&u.$flags$)){const t=ee(e);t.$rmListeners$&&(t.$rmListeners$.map(e=>e()),t.$rmListeners$=void 0)}})(this))}componentOnReady(){return ee(this).$onReadyPromise$}};n.$lazyBundleId$=e[0],a.includes(i)||s.get(i)||(r.push(i),s.define(i,X(o,n,1)))})}),p.innerHTML=r+"{visibility:hidden}.hydrated{visibility:inherit}",p.setAttribute("data-styles",""),l.insertBefore(p,c?c.nextSibling:l.firstChild),m=!1,d.length?d.map(e=>e.connectedCallback()):u.jmp(()=>f=setTimeout(G,30)),n()},J=new WeakMap,ee=e=>J.get(e),te=(e,t)=>J.set(t.$lazyInstance$=e,t),ne=(e,t)=>{const n={$flags$:0,$hostElement$:e,$cmpMeta$:t,$instanceValues$:new Map};return n.$onReadyPromise$=new Promise(e=>n.$onReadyResolve$=e),e["s-p"]=[],e["s-rc"]=[],h(e,n,t.$listeners$),J.set(e,n)},re=(e,t)=>t in e,ae=(e,t)=>(0,console.error)(e,t),se=new Map,ie=(e,t,r)=>{const a=e.$tagName$.replace(/-/g,"_"),s=e.$lazyBundleId$,i=se.get(s);return i?i[a]:n(236)(`./${s}.entry.js`).then(e=>(se.set(s,e),e[a]),ae)},oe=new Map,ue=[],le=[],ce=(e,t)=>n=>{e.push(n),s||(s=!0,t&&4&u.$flags$?de(pe):u.raf(pe))},he=e=>{for(let t=0;t<e.length;t++)try{e[t](performance.now())}catch(e){ae(e)}e.length=0},pe=()=>{he(ue),he(le),(s=ue.length>0)&&u.raf(pe)},de=e=>l().then(e),fe=ce(le,!0)},function(e,t,n){(function(t){var n=function(e){return e&&e.Math==Math&&e};e.exports=n("object"==typeof globalThis&&globalThis)||n("object"==typeof window&&window)||n("object"==typeof self&&self)||n("object"==typeof t&&t)||function(){return this}()||Function("return this")()}).call(this,n(33))},function(e,t){e.exports=function(e){try{return!!e()}catch(e){return!0}}},function(e,t,n){var r=n(52),a=Function.prototype,s=a.call,i=r&&a.bind.bind(s,s);e.exports=r?i:function(e){return function(){return s.apply(e,arguments)}}},function(e,t,n){var r=n(1),a=n(42).f,s=n(23),i=n(17),o=n(69),u=n(93),l=n(95);e.exports=function(e,t){var n,c,h,p,d,f=e.target,m=e.global,g=e.stat;if(n=m?r:g?r[f]||o(f,{}):(r[f]||{}).prototype)for(c in t){if(p=t[c],h=e.dontCallGetSet?(d=a(n,c))&&d.value:n[c],!l(m?c:f+(g?".":"#")+c,e.forced)&&void 0!==h){if(typeof p==typeof h)continue;u(p,h)}(e.sham||h&&h.sham)&&s(p,"sham",!0),i(n,c,p,e)}}},function(e,t,n){"use strict";var r,a,s,i=n(112),o=n(9),u=n(1),l=n(6),c=n(12),h=n(10),p=n(24),d=n(28),f=n(23),m=n(17),g=n(11).f,y=n(27),b=n(51),v=n(31),x=n(7),w=n(54),k=n(21),S=k.enforce,I=k.get,A=u.Int8Array,E=A&&A.prototype,N=u.Uint8ClampedArray,C=N&&N.prototype,T=A&&b(A),R=E&&b(E),_=Object.prototype,F=u.TypeError,M=x("toStringTag"),D=w("TYPED_ARRAY_TAG"),O=i&&!!v&&"Opera"!==p(u.opera),L=!1,P={Int8Array:1,Uint8Array:1,Uint8ClampedArray:1,Int16Array:2,Uint16Array:2,Int32Array:4,Uint32Array:4,Float32Array:4,Float64Array:8},$={BigInt64Array:8,BigUint64Array:8},B=function(e){var t=b(e);if(c(t)){var n=I(t);return n&&h(n,"TypedArrayConstructor")?n.TypedArrayConstructor:B(t)}},z=function(e){if(!c(e))return!1;var t=p(e);return h(P,t)||h($,t)};for(r in P)(s=(a=u[r])&&a.prototype)?S(s).TypedArrayConstructor=a:O=!1;for(r in $)(s=(a=u[r])&&a.prototype)&&(S(s).TypedArrayConstructor=a);if((!O||!l(T)||T===Function.prototype)&&(T=function(){throw F("Incorrect invocation")},O))for(r in P)u[r]&&v(u[r],T);if((!O||!R||R===_)&&(R=T.prototype,O))for(r in P)u[r]&&v(u[r].prototype,R);if(O&&b(C)!==R&&v(C,R),o&&!h(R,M))for(r in L=!0,g(R,M,{get:function(){return c(this)?this[D]:void 0}}),P)u[r]&&f(u[r],D,r);e.exports={NATIVE_ARRAY_BUFFER_VIEWS:O,TYPED_ARRAY_TAG:L&&D,aTypedArray:function(e){if(z(e))return e;throw F("Target is not a typed array")},aTypedArrayConstructor:function(e){if(l(e)&&(!v||y(T,e)))return e;throw F(d(e)+" is not a typed array constructor")},exportTypedArrayMethod:function(e,t,n,r){if(o){if(n)for(var a in P){var s=u[a];if(s&&h(s.prototype,e))try{delete s.prototype[e]}catch(n){try{s.prototype[e]=t}catch(e){}}}R[e]&&!n||m(R,e,n?t:O&&E[e]||t,r)}},exportTypedArrayStaticMethod:function(e,t,n){var r,a;if(o){if(v){if(n)for(r in P)if((a=u[r])&&h(a,e))try{delete a[e]}catch(e){}if(T[e]&&!n)return;try{return m(T,e,n?t:O&&T[e]||t)}catch(e){}}for(r in P)!(a=u[r])||a[e]&&!n||m(a,e,t)}},getTypedArrayConstructor:B,isView:function(e){if(!c(e))return!1;var t=p(e);return"DataView"===t||h(P,t)||h($,t)},isTypedArray:z,TypedArray:T,TypedArrayPrototype:R}},function(e,t,n){var r=n(87),a=r.all;e.exports=r.IS_HTMLDDA?function(e){return"function"==typeof e||e===a}:function(e){return"function"==typeof e}},function(e,t,n){var r=n(1),a=n(34),s=n(10),i=n(54),o=n(35),u=n(88),l=a("wks"),c=r.Symbol,h=c&&c.for,p=u?c:c&&c.withoutSetter||i;e.exports=function(e){if(!s(l,e)||!o&&"string"!=typeof l[e]){var t="Symbol."+e;o&&s(c,e)?l[e]=c[e]:l[e]=u&&h?h(t):p(t)}return l[e]}},function(e,t,n){var r=n(52),a=Function.prototype.call;e.exports=r?a.bind(a):function(){return a.apply(a,arguments)}},function(e,t,n){var r=n(2);e.exports=!r((function(){return 7!=Object.defineProperty({},1,{get:function(){return 7}})[1]}))},function(e,t,n){var r=n(3),a=n(16),s=r({}.hasOwnProperty);e.exports=Object.hasOwn||function(e,t){return s(a(e),t)}},function(e,t,n){var r=n(9),a=n(89),s=n(90),i=n(13),o=n(40),u=TypeError,l=Object.defineProperty,c=Object.getOwnPropertyDescriptor;t.f=r?s?function(e,t,n){if(i(e),t=o(t),i(n),"function"==typeof e&&"prototype"===t&&"value"in n&&"writable"in n&&!n.writable){var r=c(e,t);r&&r.writable&&(e[t]=n.value,n={configurable:"configurable"in n?n.configurable:r.configurable,enumerable:"enumerable"in n?n.enumerable:r.enumerable,writable:!1})}return l(e,t,n)}:l:function(e,t,n){if(i(e),t=o(t),i(n),a)try{return l(e,t,n)}catch(e){}if("get"in n||"set"in n)throw u("Accessors not supported");return"value"in n&&(e[t]=n.value),e}},function(e,t,n){var r=n(6),a=n(87),s=a.all;e.exports=a.IS_HTMLDDA?function(e){return"object"==typeof e?null!==e:r(e)||e===s}:function(e){return"object"==typeof e?null!==e:r(e)}},function(e,t,n){var r=n(12),a=String,s=TypeError;e.exports=function(e){if(r(e))return e;throw s(a(e)+" is not an object")}},function(e,t,n){var r=n(1),a=n(6),s=function(e){return a(e)?e:void 0};e.exports=function(e,t){return arguments.length<2?s(r[e]):r[e]&&r[e][t]}},function(e,t,n){var r=n(38);e.exports=function(e){return r(e.length)}},function(e,t,n){var r=n(70),a=Object;e.exports=function(e){return a(r(e))}},function(e,t,n){var r=n(6),a=n(11),s=n(129),i=n(69);e.exports=function(e,t,n,o){o||(o={});var u=o.enumerable,l=void 0!==o.name?o.name:t;if(r(n)&&s(n,l,o),o.global)u?e[t]=n:i(t,n);else{try{o.unsafe?e[t]&&(u=!0):delete e[t]}catch(e){}u?e[t]=n:a.f(e,t,{value:n,enumerable:!1,configurable:!o.nonConfigurable,writable:!o.nonWritable})}return e}},function(e,t,n){var r=n(45),a=n(3),s=n(73),i=n(16),o=n(15),u=n(159),l=a([].push),c=function(e){var t=1==e,n=2==e,a=3==e,c=4==e,h=6==e,p=7==e,d=5==e||h;return function(f,m,g,y){for(var b,v,x=i(f),w=s(x),k=r(m,g),S=o(w),I=0,A=y||u,E=t?A(f,S):n||p?A(f,0):void 0;S>I;I++)if((d||I in w)&&(v=k(b=w[I],I,x),e))if(t)E[I]=v;else if(v)switch(e){case 3:return!0;case 5:return b;case 6:return I;case 2:l(E,b)}else switch(e){case 4:return!1;case 7:l(E,b)}return h?-1:a||c?c:E}};e.exports={forEach:c(0),map:c(1),filter:c(2),some:c(3),every:c(4),find:c(5),findIndex:c(6),filterReject:c(7)}},function(e,t){e.exports=!1},function(e,t,n){var r=n(6),a=n(28),s=TypeError;e.exports=function(e){if(r(e))return e;throw s(a(e)+" is not a function")}},function(e,t,n){var r,a,s,i=n(130),o=n(1),u=n(12),l=n(23),c=n(10),h=n(68),p=n(58),d=n(59),f=o.TypeError,m=o.WeakMap;if(i||h.state){var g=h.state||(h.state=new m);g.get=g.get,g.has=g.has,g.set=g.set,r=function(e,t){if(g.has(e))throw f("Object already initialized");return t.facade=e,g.set(e,t),t},a=function(e){return g.get(e)||{}},s=function(e){return g.has(e)}}else{var y=p("state");d[y]=!0,r=function(e,t){if(c(e,y))throw f("Object already initialized");return t.facade=e,l(e,y,t),t},a=function(e){return c(e,y)?e[y]:{}},s=function(e){return c(e,y)}}e.exports={set:r,get:a,has:s,enforce:function(e){return s(e)?a(e):r(e,{})},getterFor:function(e){return function(t){var n;if(!u(t)||(n=a(t)).type!==e)throw f("Incompatible receiver, "+e+" required");return n}}}},function(e,t,n){var r=n(73),a=n(70);e.exports=function(e){return r(a(e))}},function(e,t,n){var r=n(9),a=n(11),s=n(36);e.exports=r?function(e,t,n){return a.f(e,t,s(1,n))}:function(e,t,n){return e[t]=n,e}},function(e,t,n){var r=n(67),a=n(6),s=n(37),i=n(7)("toStringTag"),o=Object,u="Arguments"==s(function(){return arguments}());e.exports=r?s:function(e){var t,n,r;return void 0===e?"Undefined":null===e?"Null":"string"==typeof(n=function(e,t){try{return e[t]}catch(e){}}(t=o(e),i))?n:u?s(t):"Object"==(r=s(t))&&a(t.callee)?"Arguments":r}},function(e,t,n){var r=n(11).f,a=n(10),s=n(7)("toStringTag");e.exports=function(e,t,n){e&&!n&&(e=e.prototype),e&&!a(e,s)&&r(e,s,{configurable:!0,value:t})}},function(e,t,n){var r=n(14);e.exports=r("navigator","userAgent")||""},function(e,t,n){var r=n(3);e.exports=r({}.isPrototypeOf)},function(e,t){var n=String;e.exports=function(e){try{return n(e)}catch(e){return"Object"}}},function(e,t,n){var r=n(30),a=Math.max,s=Math.min;e.exports=function(e,t){var n=r(e);return n<0?a(n+t,0):s(n,t)}},function(e,t,n){var r=n(135);e.exports=function(e){var t=+e;return t!=t||0===t?0:r(t)}},function(e,t,n){var r=n(3),a=n(13),s=n(136);e.exports=Object.setPrototypeOf||("__proto__"in{}?function(){var e,t=!1,n={};try{(e=r(Object.getOwnPropertyDescriptor(Object.prototype,"__proto__").set))(n,[]),t=n instanceof Array}catch(e){}return function(n,r){return a(n),s(r),t?e(n,r):n.__proto__=r,n}}():void 0)},function(e,t,n){var r,a=n(13),s=n(106),i=n(74),o=n(59),u=n(98),l=n(56),c=n(58),h=c("IE_PROTO"),p=function(){},d=function(e){return"<script>"+e+"<\/script>"},f=function(e){e.write(d("")),e.close();var t=e.parentWindow.Object;return e=null,t},m=function(){try{r=new ActiveXObject("htmlfile")}catch(e){}var e,t;m="undefined"!=typeof document?document.domain&&r?f(r):((t=l("iframe")).style.display="none",u.appendChild(t),t.src=String("javascript:"),(e=t.contentWindow.document).open(),e.write(d("document.F=Object")),e.close(),e.F):f(r);for(var n=i.length;n--;)delete m.prototype[i[n]];return m()};o[h]=!0,e.exports=Object.create||function(e,t){var n;return null!==e?(p.prototype=a(e),n=new p,p.prototype=null,n[h]=e):n=m(),void 0===t?n:s.f(n,t)}},function(e,t){var n;n=function(){return this}();try{n=n||new Function("return this")()}catch(e){"object"==typeof window&&(n=window)}e.exports=n},function(e,t,n){var r=n(19),a=n(68);(e.exports=function(e,t){return a[e]||(a[e]=void 0!==t?t:{})})("versions",[]).push({version:"3.26.1",mode:r?"pure":"global",copyright:"© 2014-2022 Denis Pushkarev (zloirock.ru)",license:"https://github.com/zloirock/core-js/blob/v3.26.1/LICENSE",source:"https://github.com/zloirock/core-js"})},function(e,t,n){var r=n(55),a=n(2);e.exports=!!Object.getOwnPropertySymbols&&!a((function(){var e=Symbol();return!String(e)||!(Object(e)instanceof Symbol)||!Symbol.sham&&r&&r<41}))},function(e,t){e.exports=function(e,t){return{enumerable:!(1&e),configurable:!(2&e),writable:!(4&e),value:t}}},function(e,t,n){var r=n(3),a=r({}.toString),s=r("".slice);e.exports=function(e){return s(a(e),8,-1)}},function(e,t,n){var r=n(30),a=Math.min;e.exports=function(e){return e>0?a(r(e),9007199254740991):0}},function(e,t,n){var r=n(24),a=String;e.exports=function(e){if("Symbol"===r(e))throw TypeError("Cannot convert a Symbol value to a string");return a(e)}},function(e,t,n){var r=n(91),a=n(41);e.exports=function(e){var t=r(e,"string");return a(t)?t:t+""}},function(e,t,n){var r=n(14),a=n(6),s=n(27),i=n(88),o=Object;e.exports=i?function(e){return"symbol"==typeof e}:function(e){var t=r("Symbol");return a(t)&&s(t.prototype,o(e))}},function(e,t,n){var r=n(9),a=n(8),s=n(92),i=n(36),o=n(22),u=n(40),l=n(10),c=n(89),h=Object.getOwnPropertyDescriptor;t.f=r?h:function(e,t){if(e=o(e),t=u(t),c)try{return h(e,t)}catch(e){}if(l(e,t))return i(!a(s.f,e,t),e[t])}},function(e,t,n){var r=n(94),a=n(74).concat("length","prototype");t.f=Object.getOwnPropertyNames||function(e){return r(e,a)}},function(e,t,n){var r=n(52),a=Function.prototype,s=a.apply,i=a.call;e.exports="object"==typeof Reflect&&Reflect.apply||(r?i.bind(s):function(){return i.apply(s,arguments)})},function(e,t,n){var r=n(62),a=n(20),s=n(52),i=r(r.bind);e.exports=function(e,t){return a(e),void 0===t?e:s?i(e,t):function(){return e.apply(t,arguments)}}},function(e,t,n){var r=n(3);e.exports=r([].slice)},function(e,t,n){var r=n(1);e.exports=r.Promise},function(e,t,n){var r=n(1),a=n(47),s=n(6),i=n(95),o=n(72),u=n(7),l=n(143),c=n(100),h=n(19),p=n(55),d=a&&a.prototype,f=u("species"),m=!1,g=s(r.PromiseRejectionEvent),y=i("Promise",(function(){var e=o(a),t=e!==String(a);if(!t&&66===p)return!0;if(h&&(!d.catch||!d.finally))return!0;if(!p||p<51||!/native code/.test(e)){var n=new a((function(e){e(1)})),r=function(e){e((function(){}),(function(){}))};if((n.constructor={})[f]=r,!(m=n.then((function(){}))instanceof r))return!0}return!t&&(l||c)&&!g}));e.exports={CONSTRUCTOR:y,REJECTION_EVENT:g,SUBCLASSING:m}},function(e,t,n){"use strict";var r=n(20),a=TypeError,s=function(e){var t,n;this.promise=new e((function(e,r){if(void 0!==t||void 0!==n)throw a("Bad Promise constructor");t=e,n=r})),this.resolve=r(t),this.reject=r(n)};e.exports.f=function(e){return new s(e)}},function(e,t){e.exports={}},function(e,t,n){var r=n(10),a=n(6),s=n(16),i=n(58),o=n(110),u=i("IE_PROTO"),l=Object,c=l.prototype;e.exports=o?l.getPrototypeOf:function(e){var t=s(e);if(r(t,u))return t[u];var n=t.constructor;return a(n)&&t instanceof n?n.prototype:t instanceof l?c:null}},function(e,t,n){var r=n(2);e.exports=!r((function(){var e=function(){}.bind();return"function"!=typeof e||e.hasOwnProperty("prototype")}))},function(e,t){e.exports=function(e){return null==e}},function(e,t,n){var r=n(3),a=0,s=Math.random(),i=r(1..toString);e.exports=function(e){return"Symbol("+(void 0===e?"":e)+")_"+i(++a+s,36)}},function(e,t,n){var r,a,s=n(1),i=n(26),o=s.process,u=s.Deno,l=o&&o.versions||u&&u.version,c=l&&l.v8;c&&(a=(r=c.split("."))[0]>0&&r[0]<4?1:+(r[0]+r[1])),!a&&i&&(!(r=i.match(/Edge\/(\d+)/))||r[1]>=74)&&(r=i.match(/Chrome\/(\d+)/))&&(a=+r[1]),e.exports=a},function(e,t,n){var r=n(1),a=n(12),s=r.document,i=a(s)&&a(s.createElement);e.exports=function(e){return i?s.createElement(e):{}}},function(e,t,n){var r=n(9),a=n(10),s=Function.prototype,i=r&&Object.getOwnPropertyDescriptor,o=a(s,"name"),u=o&&"something"===function(){}.name,l=o&&(!r||r&&i(s,"name").configurable);e.exports={EXISTS:o,PROPER:u,CONFIGURABLE:l}},function(e,t,n){var r=n(34),a=n(54),s=r("keys");e.exports=function(e){return s[e]||(s[e]=a(e))}},function(e,t){e.exports={}},function(e,t,n){var r=n(22),a=n(29),s=n(15),i=function(e){return function(t,n,i){var o,u=r(t),l=s(u),c=a(i,l);if(e&&n!=n){for(;l>c;)if((o=u[c++])!=o)return!0}else for(;l>c;c++)if((e||c in u)&&u[c]===n)return e||c||0;return!e&&-1}};e.exports={includes:i(!0),indexOf:i(!1)}},function(e,t,n){var r=n(37),a=n(1);e.exports="process"==r(a.process)},function(e,t,n){var r=n(37),a=n(3);e.exports=function(e){if("Function"===r(e))return a(e)}},function(e,t,n){var r=n(37);e.exports=Array.isArray||function(e){return"Array"==r(e)}},function(e,t,n){var r=n(5),a=n(78),s=r.aTypedArrayConstructor,i=r.getTypedArrayConstructor;e.exports=function(e){return s(a(e,i(e)))}},function(e,t,n){var r=n(212),a=n(10),s=n(119),i=n(11).f;e.exports=function(e){var t=r.Symbol||(r.Symbol={});a(t,e)||i(t,e,{value:s.f(e)})}},function(e,t){(function(t){e.exports=t}).call(this,{})},function(e,t,n){var r={};r[n(7)("toStringTag")]="z",e.exports="[object z]"===String(r)},function(e,t,n){var r=n(1),a=n(69),s=r["__core-js_shared__"]||a("__core-js_shared__",{});e.exports=s},function(e,t,n){var r=n(1),a=Object.defineProperty;e.exports=function(e,t){try{a(r,e,{value:t,configurable:!0,writable:!0})}catch(n){r[e]=t}return t}},function(e,t,n){var r=n(53),a=TypeError;e.exports=function(e){if(r(e))throw a("Can't call method on "+e);return e}},function(e,t,n){var r=n(20),a=n(53);e.exports=function(e,t){var n=e[t];return a(n)?void 0:r(n)}},function(e,t,n){var r=n(3),a=n(6),s=n(68),i=r(Function.toString);a(s.inspectSource)||(s.inspectSource=function(e){return i(e)}),e.exports=s.inspectSource},function(e,t,n){var r=n(3),a=n(2),s=n(37),i=Object,o=r("".split);e.exports=a((function(){return!i("z").propertyIsEnumerable(0)}))?function(e){return"String"==s(e)?o(e,""):i(e)}:i},function(e,t){e.exports=["constructor","hasOwnProperty","isPrototypeOf","propertyIsEnumerable","toLocaleString","toString","valueOf"]},function(e,t){t.f=Object.getOwnPropertySymbols},function(e,t,n){"use strict";var r=n(14),a=n(11),s=n(7),i=n(9),o=s("species");e.exports=function(e){var t=r(e),n=a.f;i&&t&&!t[o]&&n(t,o,{configurable:!0,get:function(){return this}})}},function(e,t,n){var r=n(27),a=TypeError;e.exports=function(e,t){if(r(t,e))return e;throw a("Incorrect invocation")}},function(e,t,n){var r=n(13),a=n(96),s=n(53),i=n(7)("species");e.exports=function(e,t){var n,o=r(e).constructor;return void 0===o||s(n=r(o)[i])?t:a(n)}},function(e,t,n){var r=n(3),a=n(2),s=n(6),i=n(24),o=n(14),u=n(72),l=function(){},c=[],h=o("Reflect","construct"),p=/^\s*(?:class|function)\b/,d=r(p.exec),f=!p.exec(l),m=function(e){if(!s(e))return!1;try{return h(l,c,e),!0}catch(e){return!1}},g=function(e){if(!s(e))return!1;switch(i(e)){case"AsyncFunction":case"GeneratorFunction":case"AsyncGeneratorFunction":return!1}try{return f||!!d(p,u(e))}catch(e){return!0}};g.sham=!0,e.exports=!h||a((function(){var e;return m(m.call)||!m(Object)||!m((function(){e=!0}))||e}))?g:m},function(e,t){e.exports=function(e){try{return{error:!1,value:e()}}catch(e){return{error:!0,value:e}}}},function(e,t,n){var r=n(24),a=n(71),s=n(53),i=n(50),o=n(7)("iterator");e.exports=function(e){if(!s(e))return a(e,o)||a(e,"@@iterator")||i[r(e)]}},function(e,t,n){var r=n(94),a=n(74);e.exports=Object.keys||function(e){return r(e,a)}},function(e,t,n){"use strict";var r=n(2);e.exports=function(e,t){var n=[][e];return!!n&&r((function(){n.call(null,t||function(){return 1},1)}))}},function(e,t,n){"use strict";var r=n(22),a=n(107),s=n(50),i=n(21),o=n(11).f,u=n(108),l=n(111),c=n(19),h=n(9),p=i.set,d=i.getterFor("Array Iterator");e.exports=u(Array,"Array",(function(e,t){p(this,{type:"Array Iterator",target:r(e),index:0,kind:t})}),(function(){var e=d(this),t=e.target,n=e.kind,r=e.index++;return!t||r>=t.length?(e.target=void 0,l(void 0,!0)):l("keys"==n?r:"values"==n?t[r]:[r,t[r]],!1)}),"values");var f=s.Arguments=s.Array;if(a("keys"),a("values"),a("entries"),!c&&h&&"values"!==f.name)try{o(f,"name",{value:"values"})}catch(e){}},function(e,t,n){"use strict";var r=n(1),a=n(3),s=n(9),i=n(112),o=n(57),u=n(23),l=n(165),c=n(2),h=n(77),p=n(30),d=n(38),f=n(113),m=n(166),g=n(51),y=n(31),b=n(43).f,v=n(11).f,x=n(114),w=n(86),k=n(25),S=n(21),I=o.PROPER,A=o.CONFIGURABLE,E=S.get,N=S.set,C=r.ArrayBuffer,T=C,R=T&&T.prototype,_=r.DataView,F=_&&_.prototype,M=Object.prototype,D=r.Array,O=r.RangeError,L=a(x),P=a([].reverse),$=m.pack,B=m.unpack,z=function(e){return[255&e]},U=function(e){return[255&e,e>>8&255]},W=function(e){return[255&e,e>>8&255,e>>16&255,e>>24&255]},V=function(e){return e[3]<<24|e[2]<<16|e[1]<<8|e[0]},j=function(e){return $(e,23,4)},G=function(e){return $(e,52,8)},H=function(e,t){v(e.prototype,t,{get:function(){return E(this)[t]}})},q=function(e,t,n,r){var a=f(n),s=E(e);if(a+t>s.byteLength)throw O("Wrong index");var i=E(s.buffer).bytes,o=a+s.byteOffset,u=w(i,o,o+t);return r?u:P(u)},K=function(e,t,n,r,a,s){var i=f(n),o=E(e);if(i+t>o.byteLength)throw O("Wrong index");for(var u=E(o.buffer).bytes,l=i+o.byteOffset,c=r(+a),h=0;h<t;h++)u[l+h]=c[s?h:t-h-1]};if(i){var Y=I&&"ArrayBuffer"!==C.name;if(c((function(){C(1)}))&&c((function(){new C(-1)}))&&!c((function(){return new C,new C(1.5),new C(NaN),1!=C.length||Y&&!A})))Y&&A&&u(C,"name","ArrayBuffer");else{(T=function(e){return h(this,R),new C(f(e))}).prototype=R;for(var X,Q=b(C),Z=0;Q.length>Z;)(X=Q[Z++])in T||u(T,X,C[X]);R.constructor=T}y&&g(F)!==M&&y(F,M);var J=new _(new T(2)),ee=a(F.setInt8);J.setInt8(0,2147483648),J.setInt8(1,2147483649),!J.getInt8(0)&&J.getInt8(1)||l(F,{setInt8:function(e,t){ee(this,e,t<<24>>24)},setUint8:function(e,t){ee(this,e,t<<24>>24)}},{unsafe:!0})}else R=(T=function(e){h(this,R);var t=f(e);N(this,{bytes:L(D(t),0),byteLength:t}),s||(this.byteLength=t)}).prototype,F=(_=function(e,t,n){h(this,F),h(e,R);var r=E(e).byteLength,a=p(t);if(a<0||a>r)throw O("Wrong offset");if(a+(n=void 0===n?r-a:d(n))>r)throw O("Wrong length");N(this,{buffer:e,byteLength:n,byteOffset:a}),s||(this.buffer=e,this.byteLength=n,this.byteOffset=a)}).prototype,s&&(H(T,"byteLength"),H(_,"buffer"),H(_,"byteLength"),H(_,"byteOffset")),l(F,{getInt8:function(e){return q(this,1,e)[0]<<24>>24},getUint8:function(e){return q(this,1,e)[0]},getInt16:function(e){var t=q(this,2,e,arguments.length>1?arguments[1]:void 0);return(t[1]<<8|t[0])<<16>>16},getUint16:function(e){var t=q(this,2,e,arguments.length>1?arguments[1]:void 0);return t[1]<<8|t[0]},getInt32:function(e){return V(q(this,4,e,arguments.length>1?arguments[1]:void 0))},getUint32:function(e){return V(q(this,4,e,arguments.length>1?arguments[1]:void 0))>>>0},getFloat32:function(e){return B(q(this,4,e,arguments.length>1?arguments[1]:void 0),23)},getFloat64:function(e){return B(q(this,8,e,arguments.length>1?arguments[1]:void 0),52)},setInt8:function(e,t){K(this,1,e,z,t)},setUint8:function(e,t){K(this,1,e,z,t)},setInt16:function(e,t){K(this,2,e,U,t,arguments.length>2?arguments[2]:void 0)},setUint16:function(e,t){K(this,2,e,U,t,arguments.length>2?arguments[2]:void 0)},setInt32:function(e,t){K(this,4,e,W,t,arguments.length>2?arguments[2]:void 0)},setUint32:function(e,t){K(this,4,e,W,t,arguments.length>2?arguments[2]:void 0)},setFloat32:function(e,t){K(this,4,e,j,t,arguments.length>2?arguments[2]:void 0)},setFloat64:function(e,t){K(this,8,e,G,t,arguments.length>2?arguments[2]:void 0)}});k(T,"ArrayBuffer"),k(_,"DataView"),e.exports={ArrayBuffer:T,DataView:_}},function(e,t,n){var r=n(29),a=n(15),s=n(115),i=Array,o=Math.max;e.exports=function(e,t,n){for(var u=a(e),l=r(t,u),c=r(void 0===n?u:n,u),h=i(o(c-l,0)),p=0;l<c;l++,p++)s(h,p,e[l]);return h.length=p,h}},function(e,t){var n="object"==typeof document&&document.all,r=void 0===n&&void 0!==n;e.exports={all:n,IS_HTMLDDA:r}},function(e,t,n){var r=n(35);e.exports=r&&!Symbol.sham&&"symbol"==typeof Symbol.iterator},function(e,t,n){var r=n(9),a=n(2),s=n(56);e.exports=!r&&!a((function(){return 7!=Object.defineProperty(s("div"),"a",{get:function(){return 7}}).a}))},function(e,t,n){var r=n(9),a=n(2);e.exports=r&&a((function(){return 42!=Object.defineProperty((function(){}),"prototype",{value:42,writable:!1}).prototype}))},function(e,t,n){var r=n(8),a=n(12),s=n(41),i=n(71),o=n(128),u=n(7),l=TypeError,c=u("toPrimitive");e.exports=function(e,t){if(!a(e)||s(e))return e;var n,u=i(e,c);if(u){if(void 0===t&&(t="default"),n=r(u,e,t),!a(n)||s(n))return n;throw l("Can't convert object to primitive value")}return void 0===t&&(t="number"),o(e,t)}},function(e,t,n){"use strict";var r={}.propertyIsEnumerable,a=Object.getOwnPropertyDescriptor,s=a&&!r.call({1:2},1);t.f=s?function(e){var t=a(this,e);return!!t&&t.enumerable}:r},function(e,t,n){var r=n(10),a=n(134),s=n(42),i=n(11);e.exports=function(e,t,n){for(var o=a(t),u=i.f,l=s.f,c=0;c<o.length;c++){var h=o[c];r(e,h)||n&&r(n,h)||u(e,h,l(t,h))}}},function(e,t,n){var r=n(3),a=n(10),s=n(22),i=n(60).indexOf,o=n(59),u=r([].push);e.exports=function(e,t){var n,r=s(e),l=0,c=[];for(n in r)!a(o,n)&&a(r,n)&&u(c,n);for(;t.length>l;)a(r,n=t[l++])&&(~i(c,n)||u(c,n));return c}},function(e,t,n){var r=n(2),a=n(6),s=/#|\.prototype\./,i=function(e,t){var n=u[o(e)];return n==c||n!=l&&(a(t)?r(t):!!t)},o=i.normalize=function(e){return String(e).replace(s,".").toLowerCase()},u=i.data={},l=i.NATIVE="N",c=i.POLYFILL="P";e.exports=i},function(e,t,n){var r=n(79),a=n(28),s=TypeError;e.exports=function(e){if(r(e))return e;throw s(a(e)+" is not a constructor")}},function(e,t,n){var r,a,s,i,o=n(1),u=n(44),l=n(45),c=n(6),h=n(10),p=n(2),d=n(98),f=n(46),m=n(56),g=n(137),y=n(99),b=n(61),v=o.setImmediate,x=o.clearImmediate,w=o.process,k=o.Dispatch,S=o.Function,I=o.MessageChannel,A=o.String,E=0,N={};try{r=o.location}catch(e){}var C=function(e){if(h(N,e)){var t=N[e];delete N[e],t()}},T=function(e){return function(){C(e)}},R=function(e){C(e.data)},_=function(e){o.postMessage(A(e),r.protocol+"//"+r.host)};v&&x||(v=function(e){g(arguments.length,1);var t=c(e)?e:S(e),n=f(arguments,1);return N[++E]=function(){u(t,void 0,n)},a(E),E},x=function(e){delete N[e]},b?a=function(e){w.nextTick(T(e))}:k&&k.now?a=function(e){k.now(T(e))}:I&&!y?(i=(s=new I).port2,s.port1.onmessage=R,a=l(i.postMessage,i)):o.addEventListener&&c(o.postMessage)&&!o.importScripts&&r&&"file:"!==r.protocol&&!p(_)?(a=_,o.addEventListener("message",R,!1)):a="onreadystatechange"in m("script")?function(e){d.appendChild(m("script")).onreadystatechange=function(){d.removeChild(this),C(e)}}:function(e){setTimeout(T(e),0)}),e.exports={set:v,clear:x}},function(e,t,n){var r=n(14);e.exports=r("document","documentElement")},function(e,t,n){var r=n(26);e.exports=/(?:ipad|iphone|ipod).*applewebkit/i.test(r)},function(e,t){e.exports="object"==typeof Deno&&Deno&&"object"==typeof Deno.version},function(e,t,n){var r=n(45),a=n(8),s=n(13),i=n(28),o=n(102),u=n(15),l=n(27),c=n(103),h=n(81),p=n(145),d=TypeError,f=function(e,t){this.stopped=e,this.result=t},m=f.prototype;e.exports=function(e,t,n){var g,y,b,v,x,w,k,S=n&&n.that,I=!(!n||!n.AS_ENTRIES),A=!(!n||!n.IS_RECORD),E=!(!n||!n.IS_ITERATOR),N=!(!n||!n.INTERRUPTED),C=r(t,S),T=function(e){return g&&p(g,"normal",e),new f(!0,e)},R=function(e){return I?(s(e),N?C(e[0],e[1],T):C(e[0],e[1])):N?C(e,T):C(e)};if(A)g=e.iterator;else if(E)g=e;else{if(!(y=h(e)))throw d(i(e)+" is not iterable");if(o(y)){for(b=0,v=u(e);v>b;b++)if((x=R(e[b]))&&l(m,x))return x;return new f(!1)}g=c(e,y)}for(w=A?e.next:g.next;!(k=a(w,g)).done;){try{x=R(k.value)}catch(e){p(g,"throw",e)}if("object"==typeof x&&x&&l(m,x))return x}return new f(!1)}},function(e,t,n){var r=n(7),a=n(50),s=r("iterator"),i=Array.prototype;e.exports=function(e){return void 0!==e&&(a.Array===e||i[s]===e)}},function(e,t,n){var r=n(8),a=n(20),s=n(13),i=n(28),o=n(81),u=TypeError;e.exports=function(e,t){var n=arguments.length<2?o(e):t;if(a(n))return s(r(n,e));throw u(i(e)+" is not iterable")}},function(e,t,n){var r=n(47),a=n(105),s=n(48).CONSTRUCTOR;e.exports=s||!a((function(e){r.all(e).then(void 0,(function(){}))}))},function(e,t,n){var r=n(7)("iterator"),a=!1;try{var s=0,i={next:function(){return{done:!!s++}},return:function(){a=!0}};i[r]=function(){return this},Array.from(i,(function(){throw 2}))}catch(e){}e.exports=function(e,t){if(!t&&!a)return!1;var n=!1;try{var s={};s[r]=function(){return{next:function(){return{done:n=!0}}}},e(s)}catch(e){}return n}},function(e,t,n){var r=n(9),a=n(90),s=n(11),i=n(13),o=n(22),u=n(82);t.f=r&&!a?Object.defineProperties:function(e,t){i(e);for(var n,r=o(t),a=u(t),l=a.length,c=0;l>c;)s.f(e,n=a[c++],r[n]);return e}},function(e,t,n){var r=n(7),a=n(32),s=n(11).f,i=r("unscopables"),o=Array.prototype;null==o[i]&&s(o,i,{configurable:!0,value:a(null)}),e.exports=function(e){o[i][e]=!0}},function(e,t,n){"use strict";var r=n(4),a=n(8),s=n(19),i=n(57),o=n(6),u=n(163),l=n(51),c=n(31),h=n(25),p=n(23),d=n(17),f=n(7),m=n(50),g=n(109),y=i.PROPER,b=i.CONFIGURABLE,v=g.IteratorPrototype,x=g.BUGGY_SAFARI_ITERATORS,w=f("iterator"),k=function(){return this};e.exports=function(e,t,n,i,f,g,S){u(n,t,i);var I,A,E,N=function(e){if(e===f&&F)return F;if(!x&&e in R)return R[e];switch(e){case"keys":case"values":case"entries":return function(){return new n(this,e)}}return function(){return new n(this)}},C=t+" Iterator",T=!1,R=e.prototype,_=R[w]||R["@@iterator"]||f&&R[f],F=!x&&_||N(f),M="Array"==t&&R.entries||_;if(M&&(I=l(M.call(new e)))!==Object.prototype&&I.next&&(s||l(I)===v||(c?c(I,v):o(I[w])||d(I,w,k)),h(I,C,!0,!0),s&&(m[C]=k)),y&&"values"==f&&_&&"values"!==_.name&&(!s&&b?p(R,"name","values"):(T=!0,F=function(){return a(_,this)})),f)if(A={values:N("values"),keys:g?F:N("keys"),entries:N("entries")},S)for(E in A)(x||T||!(E in R))&&d(R,E,A[E]);else r({target:t,proto:!0,forced:x||T},A);return s&&!S||R[w]===F||d(R,w,F,{name:f}),m[t]=F,A}},function(e,t,n){"use strict";var r,a,s,i=n(2),o=n(6),u=n(12),l=n(32),c=n(51),h=n(17),p=n(7),d=n(19),f=p("iterator"),m=!1;[].keys&&("next"in(s=[].keys())?(a=c(c(s)))!==Object.prototype&&(r=a):m=!0),!u(r)||i((function(){var e={};return r[f].call(e)!==e}))?r={}:d&&(r=l(r)),o(r[f])||h(r,f,(function(){return this})),e.exports={IteratorPrototype:r,BUGGY_SAFARI_ITERATORS:m}},function(e,t,n){var r=n(2);e.exports=!r((function(){function e(){}return e.prototype.constructor=null,Object.getPrototypeOf(new e)!==e.prototype}))},function(e,t){e.exports=function(e,t){return{value:e,done:t}}},function(e,t){e.exports="undefined"!=typeof ArrayBuffer&&"undefined"!=typeof DataView},function(e,t,n){var r=n(30),a=n(38),s=RangeError;e.exports=function(e){if(void 0===e)return 0;var t=r(e),n=a(t);if(t!==n)throw s("Wrong length or index");return n}},function(e,t,n){"use strict";var r=n(16),a=n(29),s=n(15);e.exports=function(e){for(var t=r(this),n=s(t),i=arguments.length,o=a(i>1?arguments[1]:void 0,n),u=i>2?arguments[2]:void 0,l=void 0===u?n:a(u,n);l>o;)t[o++]=e;return t}},function(e,t,n){"use strict";var r=n(40),a=n(11),s=n(36);e.exports=function(e,t,n){var i=r(t);i in e?a.f(e,i,s(0,n)):e[i]=n}},function(e,t,n){var r=n(171),a=RangeError;e.exports=function(e,t){var n=r(e);if(n%t)throw a("Wrong offset");return n}},function(e,t,n){var r=n(91),a=TypeError;e.exports=function(e){var t=r(e,"number");if("number"==typeof t)throw a("Can't convert number to bigint");return BigInt(t)}},function(e,t,n){var r=n(20),a=n(16),s=n(73),i=n(15),o=TypeError,u=function(e){return function(t,n,u,l){r(n);var c=a(t),h=s(c),p=i(c),d=e?p-1:0,f=e?-1:1;if(u<2)for(;;){if(d in h){l=h[d],d+=f;break}if(d+=f,e?d<0:p<=d)throw o("Reduce of empty array with no initial value")}for(;e?d>=0:p>d;d+=f)d in h&&(l=n(l,h[d],d,c));return l}};e.exports={left:u(!1),right:u(!0)}},function(e,t,n){var r=n(7);t.f=r},function(e,t,n){var r=n(35);e.exports=r&&!!Symbol.for&&!!Symbol.keyFor},function(e,t){e.exports={CSSRuleList:0,CSSStyleDeclaration:0,CSSValueList:0,ClientRectList:0,DOMRectList:0,DOMStringList:0,DOMTokenList:1,DataTransferItemList:0,FileList:0,HTMLAllCollection:0,HTMLCollection:0,HTMLFormElement:0,HTMLSelectElement:0,MediaList:0,MimeTypeArray:0,NamedNodeMap:0,NodeList:1,PaintRequestList:0,Plugin:0,PluginArray:0,SVGLengthList:0,SVGNumberList:0,SVGPathSegList:0,SVGPointList:0,SVGStringList:0,SVGTransformList:0,SourceBufferList:0,StyleSheetList:0,TextTrackCueList:0,TextTrackList:0,TouchList:0}},function(e,t,n){var r=n(56)("span").classList,a=r&&r.constructor&&r.constructor.prototype;e.exports=a===Object.prototype?void 0:a},function(e,t,n){"use strict";var r=n(18).forEach,a=n(83)("forEach");e.exports=a?[].forEach:function(e){return r(this,e,arguments.length>1?arguments[1]:void 0)}},function(e,t){var n,r,a=e.exports={};function s(){throw new Error("setTimeout has not been defined")}function i(){throw new Error("clearTimeout has not been defined")}function o(e){if(n===setTimeout)return setTimeout(e,0);if((n===s||!n)&&setTimeout)return n=setTimeout,setTimeout(e,0);try{return n(e,0)}catch(t){try{return n.call(null,e,0)}catch(t){return n.call(this,e,0)}}}!function(){try{n="function"==typeof setTimeout?setTimeout:s}catch(e){n=s}try{r="function"==typeof clearTimeout?clearTimeout:i}catch(e){r=i}}();var u,l=[],c=!1,h=-1;function p(){c&&u&&(c=!1,u.length?l=u.concat(l):h=-1,l.length&&d())}function d(){if(!c){var e=o(p);c=!0;for(var t=l.length;t;){for(u=l,l=[];++h<t;)u&&u[h].run();h=-1,t=l.length}u=null,c=!1,function(e){if(r===clearTimeout)return clearTimeout(e);if((r===i||!r)&&clearTimeout)return r=clearTimeout,clearTimeout(e);try{r(e)}catch(t){try{return r.call(null,e)}catch(t){return r.call(this,e)}}}(e)}}function f(e,t){this.fun=e,this.array=t}function m(){}a.nextTick=function(e){var t=new Array(arguments.length-1);if(arguments.length>1)for(var n=1;n<arguments.length;n++)t[n-1]=arguments[n];l.push(new f(e,t)),1!==l.length||c||o(d)},f.prototype.run=function(){this.fun.apply(null,this.array)},a.title="browser",a.browser=!0,a.env={},a.argv=[],a.version="",a.versions={},a.on=m,a.addListener=m,a.once=m,a.off=m,a.removeListener=m,a.removeAllListeners=m,a.emit=m,a.prependListener=m,a.prependOnceListener=m,a.listeners=function(e){return[]},a.binding=function(e){throw new Error("process.binding is not supported")},a.cwd=function(){return"/"},a.chdir=function(e){throw new Error("process.chdir is not supported")},a.umask=function(){return 0}},function(e,t){function n(e){var t=new Error("Cannot find module '"+e+"'");throw t.code="MODULE_NOT_FOUND",t}n.keys=function(){return[]},n.resolve=n,e.exports=n,n.id=125},function(e,t,n){e.exports=n(244)},function(e,t,n){var r=n(67),a=n(17),s=n(131);r||a(Object.prototype,"toString",s,{unsafe:!0})},function(e,t,n){var r=n(8),a=n(6),s=n(12),i=TypeError;e.exports=function(e,t){var n,o;if("string"===t&&a(n=e.toString)&&!s(o=r(n,e)))return o;if(a(n=e.valueOf)&&!s(o=r(n,e)))return o;if("string"!==t&&a(n=e.toString)&&!s(o=r(n,e)))return o;throw i("Can't convert object to primitive value")}},function(e,t,n){var r=n(2),a=n(6),s=n(10),i=n(9),o=n(57).CONFIGURABLE,u=n(72),l=n(21),c=l.enforce,h=l.get,p=Object.defineProperty,d=i&&!r((function(){return 8!==p((function(){}),"length",{value:8}).length})),f=String(String).split("String"),m=e.exports=function(e,t,n){"Symbol("===String(t).slice(0,7)&&(t="["+String(t).replace(/^Symbol\(([^)]*)\)/,"$1")+"]"),n&&n.getter&&(t="get "+t),n&&n.setter&&(t="set "+t),(!s(e,"name")||o&&e.name!==t)&&(i?p(e,"name",{value:t,configurable:!0}):e.name=t),d&&n&&s(n,"arity")&&e.length!==n.arity&&p(e,"length",{value:n.arity});try{n&&s(n,"constructor")&&n.constructor?i&&p(e,"prototype",{writable:!1}):e.prototype&&(e.prototype=void 0)}catch(e){}var r=c(e);return s(r,"source")||(r.source=f.join("string"==typeof t?t:"")),e};Function.prototype.toString=m((function(){return a(this)&&h(this).source||u(this)}),"toString")},function(e,t,n){var r=n(1),a=n(6),s=r.WeakMap;e.exports=a(s)&&/native code/.test(String(s))},function(e,t,n){"use strict";var r=n(67),a=n(24);e.exports=r?{}.toString:function(){return"[object "+a(this)+"]"}},function(e,t,n){n(133),n(144),n(146),n(147),n(148),n(149)},function(e,t,n){"use strict";var r,a,s,i=n(4),o=n(19),u=n(61),l=n(1),c=n(8),h=n(17),p=n(31),d=n(25),f=n(76),m=n(20),g=n(6),y=n(12),b=n(77),v=n(78),x=n(97).set,w=n(138),k=n(141),S=n(80),I=n(142),A=n(21),E=n(47),N=n(48),C=n(49),T=N.CONSTRUCTOR,R=N.REJECTION_EVENT,_=N.SUBCLASSING,F=A.getterFor("Promise"),M=A.set,D=E&&E.prototype,O=E,L=D,P=l.TypeError,$=l.document,B=l.process,z=C.f,U=z,W=!!($&&$.createEvent&&l.dispatchEvent),V=function(e){var t;return!(!y(e)||!g(t=e.then))&&t},j=function(e,t){var n,r,a,s=t.value,i=1==t.state,o=i?e.ok:e.fail,u=e.resolve,l=e.reject,h=e.domain;try{o?(i||(2===t.rejection&&Y(t),t.rejection=1),!0===o?n=s:(h&&h.enter(),n=o(s),h&&(h.exit(),a=!0)),n===e.promise?l(P("Promise-chain cycle")):(r=V(n))?c(r,n,u,l):u(n)):l(s)}catch(e){h&&!a&&h.exit(),l(e)}},G=function(e,t){e.notified||(e.notified=!0,w((function(){for(var n,r=e.reactions;n=r.get();)j(n,e);e.notified=!1,t&&!e.rejection&&q(e)})))},H=function(e,t,n){var r,a;W?((r=$.createEvent("Event")).promise=t,r.reason=n,r.initEvent(e,!1,!0),l.dispatchEvent(r)):r={promise:t,reason:n},!R&&(a=l["on"+e])?a(r):"unhandledrejection"===e&&k("Unhandled promise rejection",n)},q=function(e){c(x,l,(function(){var t,n=e.facade,r=e.value;if(K(e)&&(t=S((function(){u?B.emit("unhandledRejection",r,n):H("unhandledrejection",n,r)})),e.rejection=u||K(e)?2:1,t.error))throw t.value}))},K=function(e){return 1!==e.rejection&&!e.parent},Y=function(e){c(x,l,(function(){var t=e.facade;u?B.emit("rejectionHandled",t):H("rejectionhandled",t,e.value)}))},X=function(e,t,n){return function(r){e(t,r,n)}},Q=function(e,t,n){e.done||(e.done=!0,n&&(e=n),e.value=t,e.state=2,G(e,!0))},Z=function(e,t,n){if(!e.done){e.done=!0,n&&(e=n);try{if(e.facade===t)throw P("Promise can't be resolved itself");var r=V(t);r?w((function(){var n={done:!1};try{c(r,t,X(Z,n,e),X(Q,n,e))}catch(t){Q(n,t,e)}})):(e.value=t,e.state=1,G(e,!1))}catch(t){Q({done:!1},t,e)}}};if(T&&(L=(O=function(e){b(this,L),m(e),c(r,this);var t=F(this);try{e(X(Z,t),X(Q,t))}catch(e){Q(t,e)}}).prototype,(r=function(e){M(this,{type:"Promise",done:!1,notified:!1,parent:!1,reactions:new I,rejection:!1,state:0,value:void 0})}).prototype=h(L,"then",(function(e,t){var n=F(this),r=z(v(this,O));return n.parent=!0,r.ok=!g(e)||e,r.fail=g(t)&&t,r.domain=u?B.domain:void 0,0==n.state?n.reactions.add(r):w((function(){j(r,n)})),r.promise})),a=function(){var e=new r,t=F(e);this.promise=e,this.resolve=X(Z,t),this.reject=X(Q,t)},C.f=z=function(e){return e===O||void 0===e?new a(e):U(e)},!o&&g(E)&&D!==Object.prototype)){s=D.then,_||h(D,"then",(function(e,t){var n=this;return new O((function(e,t){c(s,n,e,t)})).then(e,t)}),{unsafe:!0});try{delete D.constructor}catch(e){}p&&p(D,L)}i({global:!0,constructor:!0,wrap:!0,forced:T},{Promise:O}),d(O,"Promise",!1,!0),f("Promise")},function(e,t,n){var r=n(14),a=n(3),s=n(43),i=n(75),o=n(13),u=a([].concat);e.exports=r("Reflect","ownKeys")||function(e){var t=s.f(o(e)),n=i.f;return n?u(t,n(e)):t}},function(e,t){var n=Math.ceil,r=Math.floor;e.exports=Math.trunc||function(e){var t=+e;return(t>0?r:n)(t)}},function(e,t,n){var r=n(6),a=String,s=TypeError;e.exports=function(e){if("object"==typeof e||r(e))return e;throw s("Can't set "+a(e)+" as a prototype")}},function(e,t){var n=TypeError;e.exports=function(e,t){if(e<t)throw n("Not enough arguments");return e}},function(e,t,n){var r,a,s,i,o,u,l,c,h=n(1),p=n(45),d=n(42).f,f=n(97).set,m=n(99),g=n(139),y=n(140),b=n(61),v=h.MutationObserver||h.WebKitMutationObserver,x=h.document,w=h.process,k=h.Promise,S=d(h,"queueMicrotask"),I=S&&S.value;I||(r=function(){var e,t;for(b&&(e=w.domain)&&e.exit();a;){t=a.fn,a=a.next;try{t()}catch(e){throw a?i():s=void 0,e}}s=void 0,e&&e.enter()},m||b||y||!v||!x?!g&&k&&k.resolve?((l=k.resolve(void 0)).constructor=k,c=p(l.then,l),i=function(){c(r)}):b?i=function(){w.nextTick(r)}:(f=p(f,h),i=function(){f(r)}):(o=!0,u=x.createTextNode(""),new v(r).observe(u,{characterData:!0}),i=function(){u.data=o=!o})),e.exports=I||function(e){var t={fn:e,next:void 0};s&&(s.next=t),a||(a=t,i()),s=t}},function(e,t,n){var r=n(26),a=n(1);e.exports=/ipad|iphone|ipod/i.test(r)&&void 0!==a.Pebble},function(e,t,n){var r=n(26);e.exports=/web0s(?!.*chrome)/i.test(r)},function(e,t,n){var r=n(1);e.exports=function(e,t){var n=r.console;n&&n.error&&(1==arguments.length?n.error(e):n.error(e,t))}},function(e,t){var n=function(){this.head=null,this.tail=null};n.prototype={add:function(e){var t={item:e,next:null};this.head?this.tail.next=t:this.head=t,this.tail=t},get:function(){var e=this.head;if(e)return this.head=e.next,this.tail===e&&(this.tail=null),e.item}},e.exports=n},function(e,t,n){var r=n(100),a=n(61);e.exports=!r&&!a&&"object"==typeof window&&"object"==typeof document},function(e,t,n){"use strict";var r=n(4),a=n(8),s=n(20),i=n(49),o=n(80),u=n(101);r({target:"Promise",stat:!0,forced:n(104)},{all:function(e){var t=this,n=i.f(t),r=n.resolve,l=n.reject,c=o((function(){var n=s(t.resolve),i=[],o=0,c=1;u(e,(function(e){var s=o++,u=!1;c++,a(n,t,e).then((function(e){u||(u=!0,i[s]=e,--c||r(i))}),l)})),--c||r(i)}));return c.error&&l(c.value),n.promise}})},function(e,t,n){var r=n(8),a=n(13),s=n(71);e.exports=function(e,t,n){var i,o;a(e);try{if(!(i=s(e,"return"))){if("throw"===t)throw n;return n}i=r(i,e)}catch(e){o=!0,i=e}if("throw"===t)throw n;if(o)throw i;return a(i),n}},function(e,t,n){"use strict";var r=n(4),a=n(19),s=n(48).CONSTRUCTOR,i=n(47),o=n(14),u=n(6),l=n(17),c=i&&i.prototype;if(r({target:"Promise",proto:!0,forced:s,real:!0},{catch:function(e){return this.then(void 0,e)}}),!a&&u(i)){var h=o("Promise").prototype.catch;c.catch!==h&&l(c,"catch",h,{unsafe:!0})}},function(e,t,n){"use strict";var r=n(4),a=n(8),s=n(20),i=n(49),o=n(80),u=n(101);r({target:"Promise",stat:!0,forced:n(104)},{race:function(e){var t=this,n=i.f(t),r=n.reject,l=o((function(){var i=s(t.resolve);u(e,(function(e){a(i,t,e).then(n.resolve,r)}))}));return l.error&&r(l.value),n.promise}})},function(e,t,n){"use strict";var r=n(4),a=n(8),s=n(49);r({target:"Promise",stat:!0,forced:n(48).CONSTRUCTOR},{reject:function(e){var t=s.f(this);return a(t.reject,void 0,e),t.promise}})},function(e,t,n){"use strict";var r=n(4),a=n(14),s=n(19),i=n(47),o=n(48).CONSTRUCTOR,u=n(150),l=a("Promise"),c=s&&!o;r({target:"Promise",stat:!0,forced:s||o},{resolve:function(e){return u(c&&this===l?i:this,e)}})},function(e,t,n){var r=n(13),a=n(12),s=n(49);e.exports=function(e,t){if(r(e),a(t)&&t.constructor===e)return t;var n=s.f(e);return(0,n.resolve)(t),n.promise}},function(e,t,n){"use strict";var r=n(4),a=n(152);r({target:"RegExp",proto:!0,forced:/./.exec!==a},{exec:a})},function(e,t,n){"use strict";var r,a,s=n(8),i=n(3),o=n(39),u=n(153),l=n(154),c=n(34),h=n(32),p=n(21).get,d=n(155),f=n(156),m=c("native-string-replace",String.prototype.replace),g=RegExp.prototype.exec,y=g,b=i("".charAt),v=i("".indexOf),x=i("".replace),w=i("".slice),k=(a=/b*/g,s(g,r=/a/,"a"),s(g,a,"a"),0!==r.lastIndex||0!==a.lastIndex),S=l.BROKEN_CARET,I=void 0!==/()??/.exec("")[1];(k||I||S||d||f)&&(y=function(e){var t,n,r,a,i,l,c,d=this,f=p(d),A=o(e),E=f.raw;if(E)return E.lastIndex=d.lastIndex,t=s(y,E,A),d.lastIndex=E.lastIndex,t;var N=f.groups,C=S&&d.sticky,T=s(u,d),R=d.source,_=0,F=A;if(C&&(T=x(T,"y",""),-1===v(T,"g")&&(T+="g"),F=w(A,d.lastIndex),d.lastIndex>0&&(!d.multiline||d.multiline&&"\n"!==b(A,d.lastIndex-1))&&(R="(?: "+R+")",F=" "+F,_++),n=new RegExp("^(?:"+R+")",T)),I&&(n=new RegExp("^"+R+"$(?!\\s)",T)),k&&(r=d.lastIndex),a=s(g,C?n:d,F),C?a?(a.input=w(a.input,_),a[0]=w(a[0],_),a.index=d.lastIndex,d.lastIndex+=a[0].length):d.lastIndex=0:k&&a&&(d.lastIndex=d.global?a.index+a[0].length:r),I&&a&&a.length>1&&s(m,a[0],n,(function(){for(i=1;i<arguments.length-2;i++)void 0===arguments[i]&&(a[i]=void 0)})),a&&N)for(a.groups=l=h(null),i=0;i<N.length;i++)l[(c=N[i])[0]]=a[c[1]];return a}),e.exports=y},function(e,t,n){"use strict";var r=n(13);e.exports=function(){var e=r(this),t="";return e.hasIndices&&(t+="d"),e.global&&(t+="g"),e.ignoreCase&&(t+="i"),e.multiline&&(t+="m"),e.dotAll&&(t+="s"),e.unicode&&(t+="u"),e.unicodeSets&&(t+="v"),e.sticky&&(t+="y"),t}},function(e,t,n){var r=n(2),a=n(1).RegExp,s=r((function(){var e=a("a","y");return e.lastIndex=2,null!=e.exec("abcd")})),i=s||r((function(){return!a("a","y").sticky})),o=s||r((function(){var e=a("^r","gy");return e.lastIndex=2,null!=e.exec("str")}));e.exports={BROKEN_CARET:o,MISSED_STICKY:i,UNSUPPORTED_Y:s}},function(e,t,n){var r=n(2),a=n(1).RegExp;e.exports=r((function(){var e=a(".","s");return!(e.dotAll&&e.exec("\n")&&"s"===e.flags)}))},function(e,t,n){var r=n(2),a=n(1).RegExp;e.exports=r((function(){var e=a("(?<a>b)","g");return"b"!==e.exec("b").groups.a||"bc"!=="b".replace(e,"$<a>c")}))},function(e,t,n){"use strict";var r=n(4),a=n(62),s=n(60).indexOf,i=n(83),o=a([].indexOf),u=!!o&&1/o([1],1,-0)<0,l=i("indexOf");r({target:"Array",proto:!0,forced:u||!l},{indexOf:function(e){var t=arguments.length>1?arguments[1]:void 0;return u?o(this,e,t)||0:s(this,e,t)}})},function(e,t,n){"use strict";var r=n(4),a=n(18).find,s=n(107),i=!0;"find"in[]&&Array(1).find((function(){i=!1})),r({target:"Array",proto:!0,forced:i},{find:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}}),s("find")},function(e,t,n){var r=n(160);e.exports=function(e,t){return new(r(e))(0===t?0:t)}},function(e,t,n){var r=n(63),a=n(79),s=n(12),i=n(7)("species"),o=Array;e.exports=function(e){var t;return r(e)&&(t=e.constructor,(a(t)&&(t===o||r(t.prototype))||s(t)&&null===(t=t[i]))&&(t=void 0)),void 0===t?o:t}},function(e,t,n){var r=n(4),a=n(16),s=n(82);r({target:"Object",stat:!0,forced:n(2)((function(){s(1)}))},{keys:function(e){return s(a(e))}})},function(e,t,n){var r=n(3),a=n(17),s=Date.prototype,i=r(s.toString),o=r(s.getTime);"Invalid Date"!=String(new Date(NaN))&&a(s,"toString",(function(){var e=o(this);return e==e?i(this):"Invalid Date"}))},function(e,t,n){"use strict";var r=n(109).IteratorPrototype,a=n(32),s=n(36),i=n(25),o=n(50),u=function(){return this};e.exports=function(e,t,n,l){var c=t+" Iterator";return e.prototype=a(r,{next:s(+!l,n)}),i(e,c,!1,!0),o[c]=u,e}},function(e,t,n){"use strict";var r=n(4),a=n(62),s=n(2),i=n(85),o=n(13),u=n(29),l=n(38),c=n(78),h=i.ArrayBuffer,p=i.DataView,d=p.prototype,f=a(h.prototype.slice),m=a(d.getUint8),g=a(d.setUint8);r({target:"ArrayBuffer",proto:!0,unsafe:!0,forced:s((function(){return!new h(2).slice(1,void 0).byteLength}))},{slice:function(e,t){if(f&&void 0===t)return f(o(this),e);for(var n=o(this).byteLength,r=u(e,n),a=u(void 0===t?n:t,n),s=new(c(this,h))(l(a-r)),i=new p(this),d=new p(s),y=0;r<a;)g(d,y++,m(i,r++));return s}})},function(e,t,n){var r=n(17);e.exports=function(e,t,n){for(var a in t)r(e,a,t[a],n);return e}},function(e,t){var n=Array,r=Math.abs,a=Math.pow,s=Math.floor,i=Math.log,o=Math.LN2;e.exports={pack:function(e,t,u){var l,c,h,p=n(u),d=8*u-t-1,f=(1<<d)-1,m=f>>1,g=23===t?a(2,-24)-a(2,-77):0,y=e<0||0===e&&1/e<0?1:0,b=0;for((e=r(e))!=e||e===1/0?(c=e!=e?1:0,l=f):(l=s(i(e)/o),e*(h=a(2,-l))<1&&(l--,h*=2),(e+=l+m>=1?g/h:g*a(2,1-m))*h>=2&&(l++,h/=2),l+m>=f?(c=0,l=f):l+m>=1?(c=(e*h-1)*a(2,t),l+=m):(c=e*a(2,m-1)*a(2,t),l=0));t>=8;)p[b++]=255&c,c/=256,t-=8;for(l=l<<t|c,d+=t;d>0;)p[b++]=255&l,l/=256,d-=8;return p[--b]|=128*y,p},unpack:function(e,t){var n,r=e.length,s=8*r-t-1,i=(1<<s)-1,o=i>>1,u=s-7,l=r-1,c=e[l--],h=127&c;for(c>>=7;u>0;)h=256*h+e[l--],u-=8;for(n=h&(1<<-u)-1,h>>=-u,u+=t;u>0;)n=256*n+e[l--],u-=8;if(0===h)h=1-o;else{if(h===i)return n?NaN:c?-1/0:1/0;n+=a(2,t),h-=o}return(c?-1:1)*n*a(2,h-t)}}},function(e,t,n){n(168)("Uint8",(function(e){return function(t,n,r){return e(this,t,n,r)}}))},function(e,t,n){"use strict";var r=n(4),a=n(1),s=n(8),i=n(9),o=n(169),u=n(5),l=n(85),c=n(77),h=n(36),p=n(23),d=n(170),f=n(38),m=n(113),g=n(116),y=n(40),b=n(10),v=n(24),x=n(12),w=n(41),k=n(32),S=n(27),I=n(31),A=n(43).f,E=n(172),N=n(18).forEach,C=n(76),T=n(11),R=n(42),_=n(21),F=n(174),M=_.get,D=_.set,O=_.enforce,L=T.f,P=R.f,$=Math.round,B=a.RangeError,z=l.ArrayBuffer,U=z.prototype,W=l.DataView,V=u.NATIVE_ARRAY_BUFFER_VIEWS,j=u.TYPED_ARRAY_TAG,G=u.TypedArray,H=u.TypedArrayPrototype,q=u.aTypedArrayConstructor,K=u.isTypedArray,Y=function(e,t){q(e);for(var n=0,r=t.length,a=new e(r);r>n;)a[n]=t[n++];return a},X=function(e,t){L(e,t,{get:function(){return M(this)[t]}})},Q=function(e){var t;return S(U,e)||"ArrayBuffer"==(t=v(e))||"SharedArrayBuffer"==t},Z=function(e,t){return K(e)&&!w(t)&&t in e&&d(+t)&&t>=0},J=function(e,t){return t=y(t),Z(e,t)?h(2,e[t]):P(e,t)},ee=function(e,t,n){return t=y(t),!(Z(e,t)&&x(n)&&b(n,"value"))||b(n,"get")||b(n,"set")||n.configurable||b(n,"writable")&&!n.writable||b(n,"enumerable")&&!n.enumerable?L(e,t,n):(e[t]=n.value,e)};i?(V||(R.f=J,T.f=ee,X(H,"buffer"),X(H,"byteOffset"),X(H,"byteLength"),X(H,"length")),r({target:"Object",stat:!0,forced:!V},{getOwnPropertyDescriptor:J,defineProperty:ee}),e.exports=function(e,t,n){var i=e.match(/\d+$/)[0]/8,u=e+(n?"Clamped":"")+"Array",l="get"+e,h="set"+e,d=a[u],y=d,b=y&&y.prototype,v={},w=function(e,t){L(e,t,{get:function(){return function(e,t){var n=M(e);return n.view[l](t*i+n.byteOffset,!0)}(this,t)},set:function(e){return function(e,t,r){var a=M(e);n&&(r=(r=$(r))<0?0:r>255?255:255&r),a.view[h](t*i+a.byteOffset,r,!0)}(this,t,e)},enumerable:!0})};V?o&&(y=t((function(e,t,n,r){return c(e,b),F(x(t)?Q(t)?void 0!==r?new d(t,g(n,i),r):void 0!==n?new d(t,g(n,i)):new d(t):K(t)?Y(y,t):s(E,y,t):new d(m(t)),e,y)})),I&&I(y,G),N(A(d),(function(e){e in y||p(y,e,d[e])})),y.prototype=b):(y=t((function(e,t,n,r){c(e,b);var a,o,u,l=0,h=0;if(x(t)){if(!Q(t))return K(t)?Y(y,t):s(E,y,t);a=t,h=g(n,i);var p=t.byteLength;if(void 0===r){if(p%i)throw B("Wrong length");if((o=p-h)<0)throw B("Wrong length")}else if((o=f(r)*i)+h>p)throw B("Wrong length");u=o/i}else u=m(t),a=new z(o=u*i);for(D(e,{buffer:a,byteOffset:h,byteLength:o,length:u,view:new W(a)});l<u;)w(e,l++)})),I&&I(y,G),b=y.prototype=k(H)),b.constructor!==y&&p(b,"constructor",y),O(b).TypedArrayConstructor=y,j&&p(b,j,u);var S=y!=d;v[u]=y,r({global:!0,constructor:!0,forced:S,sham:!V},v),"BYTES_PER_ELEMENT"in y||p(y,"BYTES_PER_ELEMENT",i),"BYTES_PER_ELEMENT"in b||p(b,"BYTES_PER_ELEMENT",i),C(u)}):e.exports=function(){}},function(e,t,n){var r=n(1),a=n(2),s=n(105),i=n(5).NATIVE_ARRAY_BUFFER_VIEWS,o=r.ArrayBuffer,u=r.Int8Array;e.exports=!i||!a((function(){u(1)}))||!a((function(){new u(-1)}))||!s((function(e){new u,new u(null),new u(1.5),new u(e)}),!0)||a((function(){return 1!==new u(new o(2),1,void 0).length}))},function(e,t,n){var r=n(12),a=Math.floor;e.exports=Number.isInteger||function(e){return!r(e)&&isFinite(e)&&a(e)===e}},function(e,t,n){var r=n(30),a=RangeError;e.exports=function(e){var t=r(e);if(t<0)throw a("The argument can't be less than 0");return t}},function(e,t,n){var r=n(45),a=n(8),s=n(96),i=n(16),o=n(15),u=n(103),l=n(81),c=n(102),h=n(173),p=n(5).aTypedArrayConstructor,d=n(117);e.exports=function(e){var t,n,f,m,g,y,b,v,x=s(this),w=i(e),k=arguments.length,S=k>1?arguments[1]:void 0,I=void 0!==S,A=l(w);if(A&&!c(A))for(v=(b=u(w,A)).next,w=[];!(y=a(v,b)).done;)w.push(y.value);for(I&&k>2&&(S=r(S,arguments[2])),n=o(w),f=new(p(x))(n),m=h(f),t=0;n>t;t++)g=I?S(w[t],t):w[t],f[t]=m?d(g):+g;return f}},function(e,t,n){var r=n(24),a=n(3)("".slice);e.exports=function(e){return"Big"===a(r(e),0,3)}},function(e,t,n){var r=n(6),a=n(12),s=n(31);e.exports=function(e,t,n){var i,o;return s&&r(i=t.constructor)&&i!==n&&a(o=i.prototype)&&o!==n.prototype&&s(e,o),e}},function(e,t,n){"use strict";var r=n(3),a=n(5),s=r(n(176)),i=a.aTypedArray;(0,a.exportTypedArrayMethod)("copyWithin",(function(e,t){return s(i(this),e,t,arguments.length>2?arguments[2]:void 0)}))},function(e,t,n){"use strict";var r=n(16),a=n(29),s=n(15),i=n(177),o=Math.min;e.exports=[].copyWithin||function(e,t){var n=r(this),u=s(n),l=a(e,u),c=a(t,u),h=arguments.length>2?arguments[2]:void 0,p=o((void 0===h?u:a(h,u))-c,u-l),d=1;for(c<l&&l<c+p&&(d=-1,c+=p-1,l+=p-1);p-- >0;)c in n?n[l]=n[c]:i(n,l),l+=d,c+=d;return n}},function(e,t,n){"use strict";var r=n(28),a=TypeError;e.exports=function(e,t){if(!delete e[t])throw a("Cannot delete property "+r(t)+" of "+r(e))}},function(e,t,n){"use strict";var r=n(5),a=n(18).every,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("every",(function(e){return a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=n(114),s=n(117),i=n(24),o=n(8),u=n(3),l=n(2),c=r.aTypedArray,h=r.exportTypedArrayMethod,p=u("".slice);h("fill",(function(e){var t=arguments.length;c(this);var n="Big"===p(i(this),0,3)?s(e):+e;return o(a,this,n,t>1?arguments[1]:void 0,t>2?arguments[2]:void 0)}),l((function(){var e=0;return new Int8Array(2).fill({valueOf:function(){return e++}}),1!==e})))},function(e,t,n){"use strict";var r=n(5),a=n(18).filter,s=n(181),i=r.aTypedArray;(0,r.exportTypedArrayMethod)("filter",(function(e){var t=a(i(this),e,arguments.length>1?arguments[1]:void 0);return s(this,t)}))},function(e,t,n){var r=n(182),a=n(64);e.exports=function(e,t){return r(a(e),t)}},function(e,t,n){var r=n(15);e.exports=function(e,t){for(var n=0,a=r(t),s=new e(a);a>n;)s[n]=t[n++];return s}},function(e,t,n){"use strict";var r=n(5),a=n(18).find,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("find",(function(e){return a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=n(18).findIndex,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("findIndex",(function(e){return a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=n(18).forEach,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("forEach",(function(e){a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=n(60).includes,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("includes",(function(e){return a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=n(60).indexOf,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("indexOf",(function(e){return a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(1),a=n(2),s=n(3),i=n(5),o=n(84),u=n(7)("iterator"),l=r.Uint8Array,c=s(o.values),h=s(o.keys),p=s(o.entries),d=i.aTypedArray,f=i.exportTypedArrayMethod,m=l&&l.prototype,g=!a((function(){m[u].call([1])})),y=!!m&&m.values&&m[u]===m.values&&"values"===m.values.name,b=function(){return c(d(this))};f("entries",(function(){return p(d(this))}),g),f("keys",(function(){return h(d(this))}),g),f("values",b,g||!y,{name:"values"}),f(u,b,g||!y,{name:"values"})},function(e,t,n){"use strict";var r=n(5),a=n(3),s=r.aTypedArray,i=r.exportTypedArrayMethod,o=a([].join);i("join",(function(e){return o(s(this),e)}))},function(e,t,n){"use strict";var r=n(5),a=n(44),s=n(191),i=r.aTypedArray;(0,r.exportTypedArrayMethod)("lastIndexOf",(function(e){var t=arguments.length;return a(s,i(this),t>1?[e,arguments[1]]:[e])}))},function(e,t,n){"use strict";var r=n(44),a=n(22),s=n(30),i=n(15),o=n(83),u=Math.min,l=[].lastIndexOf,c=!!l&&1/[1].lastIndexOf(1,-0)<0,h=o("lastIndexOf"),p=c||!h;e.exports=p?function(e){if(c)return r(l,this,arguments)||0;var t=a(this),n=i(t),o=n-1;for(arguments.length>1&&(o=u(o,s(arguments[1]))),o<0&&(o=n+o);o>=0;o--)if(o in t&&t[o]===e)return o||0;return-1}:l},function(e,t,n){"use strict";var r=n(5),a=n(18).map,s=n(64),i=r.aTypedArray;(0,r.exportTypedArrayMethod)("map",(function(e){return a(i(this),e,arguments.length>1?arguments[1]:void 0,(function(e,t){return new(s(e))(t)}))}))},function(e,t,n){"use strict";var r=n(5),a=n(118).left,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("reduce",(function(e){var t=arguments.length;return a(s(this),e,t,t>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=n(118).right,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("reduceRight",(function(e){var t=arguments.length;return a(s(this),e,t,t>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(5),a=r.aTypedArray,s=r.exportTypedArrayMethod,i=Math.floor;s("reverse",(function(){for(var e,t=a(this).length,n=i(t/2),r=0;r<n;)e=this[r],this[r++]=this[--t],this[t]=e;return this}))},function(e,t,n){"use strict";var r=n(1),a=n(8),s=n(5),i=n(15),o=n(116),u=n(16),l=n(2),c=r.RangeError,h=r.Int8Array,p=h&&h.prototype,d=p&&p.set,f=s.aTypedArray,m=s.exportTypedArrayMethod,g=!l((function(){var e=new Uint8ClampedArray(2);return a(d,e,{length:1,0:3},1),3!==e[1]})),y=g&&s.NATIVE_ARRAY_BUFFER_VIEWS&&l((function(){var e=new h(2);return e.set(1),e.set("2",1),0!==e[0]||2!==e[1]}));m("set",(function(e){f(this);var t=o(arguments.length>1?arguments[1]:void 0,1),n=u(e);if(g)return a(d,this,n,t);var r=this.length,s=i(n),l=0;if(s+t>r)throw c("Wrong length");for(;l<s;)this[t+l]=n[l++]}),!g||y)},function(e,t,n){"use strict";var r=n(5),a=n(64),s=n(2),i=n(46),o=r.aTypedArray;(0,r.exportTypedArrayMethod)("slice",(function(e,t){for(var n=i(o(this),e,t),r=a(this),s=0,u=n.length,l=new r(u);u>s;)l[s]=n[s++];return l}),s((function(){new Int8Array(1).slice()})))},function(e,t,n){"use strict";var r=n(5),a=n(18).some,s=r.aTypedArray;(0,r.exportTypedArrayMethod)("some",(function(e){return a(s(this),e,arguments.length>1?arguments[1]:void 0)}))},function(e,t,n){"use strict";var r=n(1),a=n(62),s=n(2),i=n(20),o=n(200),u=n(5),l=n(201),c=n(202),h=n(55),p=n(203),d=u.aTypedArray,f=u.exportTypedArrayMethod,m=r.Uint16Array,g=m&&a(m.prototype.sort),y=!(!g||s((function(){g(new m(2),null)}))&&s((function(){g(new m(2),{})}))),b=!!g&&!s((function(){if(h)return h<74;if(l)return l<67;if(c)return!0;if(p)return p<602;var e,t,n=new m(516),r=Array(516);for(e=0;e<516;e++)t=e%4,n[e]=515-e,r[e]=e-2*t+3;for(g(n,(function(e,t){return(e/4|0)-(t/4|0)})),e=0;e<516;e++)if(n[e]!==r[e])return!0}));f("sort",(function(e){return void 0!==e&&i(e),b?g(this,e):o(d(this),function(e){return function(t,n){return void 0!==e?+e(t,n)||0:n!=n?-1:t!=t?1:0===t&&0===n?1/t>0&&1/n<0?1:-1:t>n}}(e))}),!b||y)},function(e,t,n){var r=n(86),a=Math.floor,s=function(e,t){var n=e.length,u=a(n/2);return n<8?i(e,t):o(e,s(r(e,0,u),t),s(r(e,u),t),t)},i=function(e,t){for(var n,r,a=e.length,s=1;s<a;){for(r=s,n=e[s];r&&t(e[r-1],n)>0;)e[r]=e[--r];r!==s++&&(e[r]=n)}return e},o=function(e,t,n,r){for(var a=t.length,s=n.length,i=0,o=0;i<a||o<s;)e[i+o]=i<a&&o<s?r(t[i],n[o])<=0?t[i++]:n[o++]:i<a?t[i++]:n[o++];return e};e.exports=s},function(e,t,n){var r=n(26).match(/firefox\/(\d+)/i);e.exports=!!r&&+r[1]},function(e,t,n){var r=n(26);e.exports=/MSIE|Trident/.test(r)},function(e,t,n){var r=n(26).match(/AppleWebKit\/(\d+)\./);e.exports=!!r&&+r[1]},function(e,t,n){"use strict";var r=n(5),a=n(38),s=n(29),i=n(64),o=r.aTypedArray;(0,r.exportTypedArrayMethod)("subarray",(function(e,t){var n=o(this),r=n.length,u=s(e,r);return new(i(n))(n.buffer,n.byteOffset+u*n.BYTES_PER_ELEMENT,a((void 0===t?r:s(t,r))-u))}))},function(e,t,n){"use strict";var r=n(1),a=n(44),s=n(5),i=n(2),o=n(46),u=r.Int8Array,l=s.aTypedArray,c=s.exportTypedArrayMethod,h=[].toLocaleString,p=!!u&&i((function(){h.call(new u(1))}));c("toLocaleString",(function(){return a(h,p?o(l(this)):l(this),o(arguments))}),i((function(){return[1,2].toLocaleString()!=new u([1,2]).toLocaleString()}))||!i((function(){u.prototype.toLocaleString.call([1,2])})))},function(e,t,n){"use strict";var r=n(5).exportTypedArrayMethod,a=n(2),s=n(1),i=n(3),o=s.Uint8Array,u=o&&o.prototype||{},l=[].toString,c=i([].join);a((function(){l.call({})}))&&(l=function(){return c(this)});var h=u.toString!=l;r("toString",l,h)},function(e,t,n){"use strict";var r=n(4),a=n(1),s=n(85),i=n(76),o=s.ArrayBuffer;r({global:!0,constructor:!0,forced:a.ArrayBuffer!==o},{ArrayBuffer:o}),i("ArrayBuffer")},function(e,t,n){var r=n(4),a=n(9),s=n(11).f;r({target:"Object",stat:!0,forced:Object.defineProperty!==s,sham:!a},{defineProperty:s})},function(e,t,n){n(210),n(214),n(215),n(216),n(217)},function(e,t,n){"use strict";var r=n(4),a=n(1),s=n(8),i=n(3),o=n(19),u=n(9),l=n(35),c=n(2),h=n(10),p=n(27),d=n(13),f=n(22),m=n(40),g=n(39),y=n(36),b=n(32),v=n(82),x=n(43),w=n(211),k=n(75),S=n(42),I=n(11),A=n(106),E=n(92),N=n(17),C=n(34),T=n(58),R=n(59),_=n(54),F=n(7),M=n(119),D=n(65),O=n(213),L=n(25),P=n(21),$=n(18).forEach,B=T("hidden"),z=P.set,U=P.getterFor("Symbol"),W=Object.prototype,V=a.Symbol,j=V&&V.prototype,G=a.TypeError,H=a.QObject,q=S.f,K=I.f,Y=w.f,X=E.f,Q=i([].push),Z=C("symbols"),J=C("op-symbols"),ee=C("wks"),te=!H||!H.prototype||!H.prototype.findChild,ne=u&&c((function(){return 7!=b(K({},"a",{get:function(){return K(this,"a",{value:7}).a}})).a}))?function(e,t,n){var r=q(W,t);r&&delete W[t],K(e,t,n),r&&e!==W&&K(W,t,r)}:K,re=function(e,t){var n=Z[e]=b(j);return z(n,{type:"Symbol",tag:e,description:t}),u||(n.description=t),n},ae=function(e,t,n){e===W&&ae(J,t,n),d(e);var r=m(t);return d(n),h(Z,r)?(n.enumerable?(h(e,B)&&e[B][r]&&(e[B][r]=!1),n=b(n,{enumerable:y(0,!1)})):(h(e,B)||K(e,B,y(1,{})),e[B][r]=!0),ne(e,r,n)):K(e,r,n)},se=function(e,t){d(e);var n=f(t),r=v(n).concat(le(n));return $(r,(function(t){u&&!s(ie,n,t)||ae(e,t,n[t])})),e},ie=function(e){var t=m(e),n=s(X,this,t);return!(this===W&&h(Z,t)&&!h(J,t))&&(!(n||!h(this,t)||!h(Z,t)||h(this,B)&&this[B][t])||n)},oe=function(e,t){var n=f(e),r=m(t);if(n!==W||!h(Z,r)||h(J,r)){var a=q(n,r);return!a||!h(Z,r)||h(n,B)&&n[B][r]||(a.enumerable=!0),a}},ue=function(e){var t=Y(f(e)),n=[];return $(t,(function(e){h(Z,e)||h(R,e)||Q(n,e)})),n},le=function(e){var t=e===W,n=Y(t?J:f(e)),r=[];return $(n,(function(e){!h(Z,e)||t&&!h(W,e)||Q(r,Z[e])})),r};l||(N(j=(V=function(){if(p(j,this))throw G("Symbol is not a constructor");var e=arguments.length&&void 0!==arguments[0]?g(arguments[0]):void 0,t=_(e),n=function(e){this===W&&s(n,J,e),h(this,B)&&h(this[B],t)&&(this[B][t]=!1),ne(this,t,y(1,e))};return u&&te&&ne(W,t,{configurable:!0,set:n}),re(t,e)}).prototype,"toString",(function(){return U(this).tag})),N(V,"withoutSetter",(function(e){return re(_(e),e)})),E.f=ie,I.f=ae,A.f=se,S.f=oe,x.f=w.f=ue,k.f=le,M.f=function(e){return re(F(e),e)},u&&(K(j,"description",{configurable:!0,get:function(){return U(this).description}}),o||N(W,"propertyIsEnumerable",ie,{unsafe:!0}))),r({global:!0,constructor:!0,wrap:!0,forced:!l,sham:!l},{Symbol:V}),$(v(ee),(function(e){D(e)})),r({target:"Symbol",stat:!0,forced:!l},{useSetter:function(){te=!0},useSimple:function(){te=!1}}),r({target:"Object",stat:!0,forced:!l,sham:!u},{create:function(e,t){return void 0===t?b(e):se(b(e),t)},defineProperty:ae,defineProperties:se,getOwnPropertyDescriptor:oe}),r({target:"Object",stat:!0,forced:!l},{getOwnPropertyNames:ue}),O(),L(V,"Symbol"),R[B]=!0},function(e,t,n){var r=n(37),a=n(22),s=n(43).f,i=n(86),o="object"==typeof window&&window&&Object.getOwnPropertyNames?Object.getOwnPropertyNames(window):[];e.exports.f=function(e){return o&&"Window"==r(e)?function(e){try{return s(e)}catch(e){return i(o)}}(e):s(a(e))}},function(e,t,n){var r=n(1);e.exports=r},function(e,t,n){var r=n(8),a=n(14),s=n(7),i=n(17);e.exports=function(){var e=a("Symbol"),t=e&&e.prototype,n=t&&t.valueOf,o=s("toPrimitive");t&&!t[o]&&i(t,o,(function(e){return r(n,this)}),{arity:1})}},function(e,t,n){var r=n(4),a=n(14),s=n(10),i=n(39),o=n(34),u=n(120),l=o("string-to-symbol-registry"),c=o("symbol-to-string-registry");r({target:"Symbol",stat:!0,forced:!u},{for:function(e){var t=i(e);if(s(l,t))return l[t];var n=a("Symbol")(t);return l[t]=n,c[n]=t,n}})},function(e,t,n){var r=n(4),a=n(10),s=n(41),i=n(28),o=n(34),u=n(120),l=o("symbol-to-string-registry");r({target:"Symbol",stat:!0,forced:!u},{keyFor:function(e){if(!s(e))throw TypeError(i(e)+" is not a symbol");if(a(l,e))return l[e]}})},function(e,t,n){var r=n(4),a=n(14),s=n(44),i=n(8),o=n(3),u=n(2),l=n(63),c=n(6),h=n(12),p=n(41),d=n(46),f=n(35),m=a("JSON","stringify"),g=o(/./.exec),y=o("".charAt),b=o("".charCodeAt),v=o("".replace),x=o(1..toString),w=/[\uD800-\uDFFF]/g,k=/^[\uD800-\uDBFF]$/,S=/^[\uDC00-\uDFFF]$/,I=!f||u((function(){var e=a("Symbol")();return"[null]"!=m([e])||"{}"!=m({a:e})||"{}"!=m(Object(e))})),A=u((function(){return'"\\udf06\\ud834"'!==m("\udf06\ud834")||'"\\udead"'!==m("\udead")})),E=function(e,t){var n=d(arguments),r=t;if((h(t)||void 0!==e)&&!p(e))return l(t)||(t=function(e,t){if(c(r)&&(t=i(r,this,e,t)),!p(t))return t}),n[1]=t,s(m,null,n)},N=function(e,t,n){var r=y(n,t-1),a=y(n,t+1);return g(k,e)&&!g(S,a)||g(S,e)&&!g(k,r)?"\\u"+x(b(e,0),16):e};m&&r({target:"JSON",stat:!0,arity:3,forced:I||A},{stringify:function(e,t,n){var r=d(arguments),a=s(I?E:m,null,r);return A&&"string"==typeof a?v(a,w,N):a}})},function(e,t,n){var r=n(4),a=n(35),s=n(2),i=n(75),o=n(16);r({target:"Object",stat:!0,forced:!a||s((function(){i.f(1)}))},{getOwnPropertySymbols:function(e){var t=i.f;return t?t(o(e)):[]}})},function(e,t,n){"use strict";var r=n(4),a=n(9),s=n(1),i=n(3),o=n(10),u=n(6),l=n(27),c=n(39),h=n(11).f,p=n(93),d=s.Symbol,f=d&&d.prototype;if(a&&u(d)&&(!("description"in f)||void 0!==d().description)){var m={},g=function(){var e=arguments.length<1||void 0===arguments[0]?void 0:c(arguments[0]),t=l(f,this)?new d(e):void 0===e?d():d(e);return""===e&&(m[t]=!0),t};p(g,d),g.prototype=f,f.constructor=g;var y="Symbol(test)"==String(d("test")),b=i(f.valueOf),v=i(f.toString),x=/^Symbol\((.*)\)[^)]+$/,w=i("".replace),k=i("".slice);h(f,"description",{configurable:!0,get:function(){var e=b(this);if(o(m,e))return"";var t=v(e),n=y?k(t,7,-1):w(t,x,"$1");return""===n?void 0:n}}),r({global:!0,constructor:!0,forced:!0},{Symbol:g})}},function(e,t,n){n(65)("iterator")},function(e,t,n){"use strict";var r=n(221).charAt,a=n(39),s=n(21),i=n(108),o=n(111),u=s.set,l=s.getterFor("String Iterator");i(String,"String",(function(e){u(this,{type:"String Iterator",string:a(e),index:0})}),(function(){var e,t=l(this),n=t.string,a=t.index;return a>=n.length?o(void 0,!0):(e=r(n,a),t.index+=e.length,o(e,!1))}))},function(e,t,n){var r=n(3),a=n(30),s=n(39),i=n(70),o=r("".charAt),u=r("".charCodeAt),l=r("".slice),c=function(e){return function(t,n){var r,c,h=s(i(t)),p=a(n),d=h.length;return p<0||p>=d?e?"":void 0:(r=u(h,p))<55296||r>56319||p+1===d||(c=u(h,p+1))<56320||c>57343?e?o(h,p):r:e?l(h,p,p+2):c-56320+(r-55296<<10)+65536}};e.exports={codeAt:c(!1),charAt:c(!0)}},function(e,t,n){var r=n(1),a=n(121),s=n(122),i=n(84),o=n(23),u=n(7),l=u("iterator"),c=u("toStringTag"),h=i.values,p=function(e,t){if(e){if(e[l]!==h)try{o(e,l,h)}catch(t){e[l]=h}if(e[c]||o(e,c,t),a[t])for(var n in i)if(e[n]!==i[n])try{o(e,n,i[n])}catch(t){e[n]=i[n]}}};for(var d in a)p(r[d]&&r[d].prototype,d);p(s,"DOMTokenList")},function(e,t,n){n(65)("asyncIterator")},function(e,t,n){var r=n(14),a=n(65),s=n(25);a("toStringTag"),s(r("Symbol"),"Symbol")},function(e,t,n){var r=n(1);n(25)(r.JSON,"JSON",!0)},function(e,t,n){n(25)(Math,"Math",!0)},function(e,t,n){n(4)({target:"Object",stat:!0,sham:!n(9)},{create:n(32)})},function(e,t,n){var r=n(4),a=n(2),s=n(16),i=n(51),o=n(110);r({target:"Object",stat:!0,forced:a((function(){i(1)})),sham:!o},{getPrototypeOf:function(e){return i(s(e))}})},function(e,t,n){"use strict";var r=n(4),a=n(123);r({target:"Array",proto:!0,forced:[].forEach!=a},{forEach:a})},function(e,t,n){var r=n(1),a=n(121),s=n(122),i=n(123),o=n(23),u=function(e){if(e&&e.forEach!==i)try{o(e,"forEach",i)}catch(t){e.forEach=i}};for(var l in a)a[l]&&u(r[l]&&r[l].prototype);u(s)},function(e,t,n){var r=n(9),a=n(57).EXISTS,s=n(3),i=n(11).f,o=Function.prototype,u=s(o.toString),l=/function\b(?:\s|\/\*[\S\s]*?\*\/|\/\/[^\n\r]*[\n\r]+)*([^\s(/]*)/,c=s(l.exec);r&&!a&&i(o,"name",{configurable:!0,get:function(){try{return c(l,u(this))[1]}catch(e){return""}}})},function(e,t,n){n(4)({target:"Object",stat:!0},{setPrototypeOf:n(31)})},function(e,t,n){"use strict";var r=n(4),a=n(3),s=n(63),i=a([].reverse),o=[1,2];r({target:"Array",proto:!0,forced:String(o)===String(o.reverse())},{reverse:function(){return s(this)&&(this.length=this.length),i(this)}})},function(e,t,n){"use strict";var r=n(4),a=n(63),s=n(79),i=n(12),o=n(29),u=n(15),l=n(22),c=n(115),h=n(7),p=n(235),d=n(46),f=p("slice"),m=h("species"),g=Array,y=Math.max;r({target:"Array",proto:!0,forced:!f},{slice:function(e,t){var n,r,h,p=l(this),f=u(p),b=o(e,f),v=o(void 0===t?f:t,f);if(a(p)&&(n=p.constructor,(s(n)&&(n===g||a(n.prototype))||i(n)&&null===(n=n[m]))&&(n=void 0),n===g||void 0===n))return d(p,b,v);for(r=new(void 0===n?g:n)(y(v-b,0)),h=0;b<v;b++,h++)b in p&&c(r,h,p[b]);return r.length=h,r}})},function(e,t,n){var r=n(2),a=n(7),s=n(55),i=a("species");e.exports=function(e){return s>=51||!r((function(){var t=[];return(t.constructor={})[i]=function(){return{foo:1}},1!==t[e](Boolean).foo}))}},function(e,t,n){var r={"./camera-comp_3.entry.js":[237,0]};function a(e){if(!n.o(r,e))return Promise.resolve().then((function(){var t=new Error("Cannot find module '"+e+"'");throw t.code="MODULE_NOT_FOUND",t}));var t=r[e],a=t[0];return n.e(t[1]).then((function(){return n(a)}))}a.keys=function(){return Object.keys(r)},a.id=236,e.exports=a},function(e,t,n){"use strict";n.r(t),function(e,r,a,s,i,o){n.d(t,"camera_comp",(function(){return Cq})),n.d(t,"camera_component",(function(){return Tq})),n.d(t,"control_panel",(function(){return _q}));var u=n(0);class l{static init(e){this.cameraModule=e}static captureEvent(e){this.cameraModule.dispatchEvent(new CustomEvent("capture",{detail:{photos:e},bubbles:!0,cancelable:!0,composed:!0}))}static closeEvent(){this.cameraModule.dispatchEvent(new CustomEvent("close",{bubbles:!1,cancelable:!0,composed:!0}))}static cameraReadyEvent(){this.cameraModule.dispatchEvent(new CustomEvent("cameraReady",{bubbles:!1,cancelable:!0,composed:!0}))}static errorEvent(e){this.cameraModule.dispatchEvent(new CustomEvent("error",{detail:{error:e},bubbles:!0,cancelable:!0,composed:!0}))}static detectionError(e){this.cameraModule.dispatchEvent(new CustomEvent("detectionError",{detail:{error:e},bubbles:!0,cancelable:!0,composed:!0}))}static loadedModels(e){this.cameraModule.dispatchEvent(new CustomEvent("loadedModels",{detail:{loadedModels:e},bubbles:!0,cancelable:!0,composed:!0}))}static detectorInitialized(e){this.cameraModule.dispatchEvent(new CustomEvent("detectorInitialized",{detail:{initializationTime:e},bubbles:!0,cancelable:!0,composed:!0}))}}var c=Object.defineProperty,h=(e=>n(125))(),p=(e,t)=>{for(var n in t)c(e,n,{get:t[n],enumerable:!0})},d={};p(d,{Abs:()=>et,Acos:()=>tt,Acosh:()=>nt,AdadeltaOptimizer:()=>Md,AdagradOptimizer:()=>Dd,AdamOptimizer:()=>Od,AdamaxOptimizer:()=>Ld,Add:()=>rt,AddN:()=>at,All:()=>st,Any:()=>it,ArgMax:()=>ot,ArgMin:()=>ut,Asin:()=>lt,Asinh:()=>ct,Atan:()=>ht,Atan2:()=>dt,Atanh:()=>pt,AvgPool:()=>ft,AvgPool3D:()=>gt,AvgPool3DGrad:()=>yt,AvgPoolGrad:()=>mt,BackendWasm:()=>vj,BatchMatMul:()=>bt,BatchToSpaceND:()=>vt,Bincount:()=>xt,BroadcastArgs:()=>kt,BroadcastTo:()=>wt,Callback:()=>BS,CallbackList:()=>Fb,Cast:()=>St,Ceil:()=>It,ClipByValue:()=>At,Complex:()=>Et,ComplexAbs:()=>Nt,Concat:()=>Ct,Conv2D:()=>Tt,Conv2DBackpropFilter:()=>Rt,Conv2DBackpropInput:()=>_t,Conv3D:()=>Ft,Conv3DBackpropFilterV2:()=>Mt,Conv3DBackpropInputV2:()=>Dt,Cos:()=>Ot,Cosh:()=>Lt,CropAndResize:()=>$t,Cumsum:()=>Pt,CustomCallback:()=>Ob,DataStorage:()=>Q,DenseBincount:()=>Bt,DepthToSpace:()=>zt,DepthwiseConv2dNative:()=>Ut,DepthwiseConv2dNativeBackpropFilter:()=>Wt,DepthwiseConv2dNativeBackpropInput:()=>Vt,Diag:()=>jt,Dilation2D:()=>Gt,Dilation2DBackpropFilter:()=>qt,Dilation2DBackpropInput:()=>Ht,ENV:()=>Qe,EarlyStopping:()=>WS,Einsum:()=>Yt,Elu:()=>Xt,EluGrad:()=>Qt,Environment:()=>qe,Equal:()=>Jt,Erf:()=>Zt,Exp:()=>en,ExpandDims:()=>tn,Expm1:()=>nn,FFT:()=>rn,Fill:()=>an,FlipLeftRight:()=>sn,Floor:()=>on,FloorDiv:()=>un,FromPixels:()=>ta,FusedBatchNorm:()=>ln,FusedConv2D:()=>aa,FusedDepthwiseConv2D:()=>sa,GPGPUContext:()=>WD,GatherNd:()=>hn,GatherV2:()=>cn,GraphModel:()=>IA,Greater:()=>pn,GreaterEqual:()=>dn,History:()=>Db,IFFT:()=>mn,Identity:()=>fn,Imag:()=>gn,InputSpec:()=>vb,IsFinite:()=>yn,IsInf:()=>bn,IsNan:()=>vn,KernelBackend:()=>Z,LRN:()=>Rn,LRNGrad:()=>_n,LayerVariable:()=>gb,LayersModel:()=>Wv,LeakyRelu:()=>xn,Less:()=>wn,LessEqual:()=>kn,LinSpace:()=>Sn,Log:()=>In,Log1p:()=>An,LogSoftmax:()=>Tn,LogicalAnd:()=>En,LogicalNot:()=>Nn,LogicalOr:()=>Cn,MathBackendWebGL:()=>jO,Max:()=>Fn,MaxPool:()=>Dn,MaxPool3D:()=>Ln,MaxPool3DGrad:()=>Pn,MaxPoolGrad:()=>On,MaxPoolWithArgmax:()=>$n,Maximum:()=>Mn,Mean:()=>Bn,Min:()=>zn,Minimum:()=>Un,MirrorPad:()=>Wn,Mod:()=>Vn,MomentumOptimizer:()=>$d,Multinomial:()=>jn,Multiply:()=>Gn,Neg:()=>Hn,NonMaxSuppressionV3:()=>Kn,NonMaxSuppressionV4:()=>Yn,NonMaxSuppressionV5:()=>Xn,NotEqual:()=>qn,OP_SCOPE_SUFFIX:()=>Fs,OneHot:()=>Zn,OnesLike:()=>Qn,Optimizer:()=>Fd,OptimizerConstructors:()=>zd,Pack:()=>Jn,PadV2:()=>er,Pool:()=>tr,Pow:()=>nr,Prelu:()=>rr,Prod:()=>ar,RMSPropOptimizer:()=>Bd,RNN:()=>Yx,Range:()=>sr,Rank:()=>Za,Real:()=>ir,RealDiv:()=>Kt,Reciprocal:()=>or,Reduction:()=>id,Relu:()=>ur,Relu6:()=>fr,Reshape:()=>lr,ResizeBilinear:()=>pr,ResizeBilinearGrad:()=>dr,ResizeNearestNeighbor:()=>cr,ResizeNearestNeighborGrad:()=>hr,Reverse:()=>mr,RotateWithOffset:()=>na,Round:()=>gr,Rsqrt:()=>yr,SGDOptimizer:()=>Pd,ScatterNd:()=>br,Select:()=>vr,Selu:()=>xr,Sequential:()=>Hv,Sigmoid:()=>Ar,Sign:()=>Ir,Sin:()=>kr,Sinh:()=>Sr,Slice:()=>wr,Softmax:()=>_r,Softplus:()=>Er,SpaceToBatchND:()=>Tr,SparseFillEmptyRows:()=>Fr,SparseReshape:()=>Mr,SparseSegmentMean:()=>Dr,SparseSegmentSum:()=>Or,SparseToDense:()=>Lr,SplitV:()=>Rr,Sqrt:()=>Nr,Square:()=>$r,SquaredDifference:()=>Pr,Step:()=>ea,StridedSlice:()=>Br,StringNGrams:()=>zr,StringSplit:()=>Ur,StringToHashBucketFast:()=>Wr,Sub:()=>Vr,Sum:()=>Cr,SymbolicTensor:()=>xb,Tan:()=>jr,Tanh:()=>Gr,Tensor:()=>Ya,TensorBuffer:()=>Ha,Tile:()=>Hr,TopK:()=>qr,Transform:()=>Kr,Transpose:()=>Yr,Unique:()=>Xr,Unpack:()=>Qr,UnsortedSegmentSum:()=>Zr,Variable:()=>Qa,ZerosLike:()=>Jr,_FusedMatMul:()=>ra,abs:()=>xu,acos:()=>wu,acosh:()=>ku,add:()=>gu,addN:()=>Su,all:()=>Iu,any:()=>Au,argMax:()=>Eu,argMin:()=>Nu,asin:()=>Cu,asinh:()=>Tu,atan:()=>Ru,atan2:()=>_u,atanh:()=>Fu,avgPool:()=>Ku,avgPool3d:()=>Yu,backend:()=>fu,backend_util:()=>jd,basicLSTMCell:()=>el,batchNorm:()=>nl,batchNorm2d:()=>rl,batchNorm3d:()=>al,batchNorm4d:()=>sl,batchToSpaceND:()=>tl,bincount:()=>il,booleanMaskAsync:()=>up,broadcastArgs:()=>ol,broadcastTo:()=>ul,broadcast_util:()=>Yi,browser:()=>eo,buffer:()=>Ii,callbacks:()=>KS,cast:()=>Ai,ceil:()=>ll,clipByValue:()=>cl,clone:()=>Ei,complex:()=>Ds,concat:()=>Xu,concat1d:()=>hl,concat2d:()=>pl,concat3d:()=>dl,concat4d:()=>fl,constraints:()=>rg,conv1d:()=>gl,conv2d:()=>ml,conv2dTranspose:()=>bl,conv3d:()=>vl,conv3dTranspose:()=>wl,copyRegisteredKernels:()=>ya,cos:()=>kl,cosh:()=>Sl,cosineWindow:()=>gp,cumsum:()=>Il,customGrad:()=>cc,data:()=>NA,denseBincount:()=>Al,deprecationWarn:()=>Zo,depthToSpace:()=>El,depthwiseConv2d:()=>Nl,deregisterOp:()=>ZS,device_util:()=>Ss,diag:()=>Cl,dilation2d:()=>Tl,disableDeprecationWarnings:()=>Qo,dispose:()=>au,disposeVariables:()=>Jo,div:()=>bu,divNoNan:()=>Ml,dot:()=>Dl,dropout:()=>fp,einsum:()=>Ol,elu:()=>Ll,enableDebugMode:()=>Xo,enableProdMode:()=>Yo,enclosingPowerOfTwo:()=>mp,engine:()=>eu,env:()=>Ye,equal:()=>Rl,erf:()=>Pl,exp:()=>$l,expandDims:()=>Bl,expm1:()=>zl,eye:()=>Wl,fft:()=>Ph,fill:()=>Vl,findBackend:()=>hu,findBackendFactory:()=>pu,floor:()=>jl,floorDiv:()=>yu,forceHalfFloat:()=>HO,fused:()=>bp,gather:()=>Gl,gatherND:()=>dp,gather_util:()=>oo,getBackend:()=>lu,getGradient:()=>ha,getKernel:()=>ca,getKernelsForBackend:()=>pa,getThreadsCount:()=>Mj,gpgpu_util:()=>vD,grad:()=>sc,grads:()=>ic,greater:()=>Hl,greaterEqual:()=>ql,ifft:()=>$h,imag:()=>Kl,image:()=>Nd,inTopKAsync:()=>yp,initializers:()=>Ug,input:()=>Xv,io:()=>Ci,irfft:()=>Bh,isFinite:()=>Yl,isInf:()=>Xl,isNaN:()=>Ql,keep:()=>su,kernel_impls:()=>jf,layers:()=>ib,leakyRelu:()=>Zl,less:()=>Jl,lessEqual:()=>ec,linalg:()=>Cd,linspace:()=>tc,loadGraphModel:()=>AA,loadLayersModel:()=>Yv,localResponseNormalization:()=>nc,log:()=>rc,log1p:()=>ac,logSigmoid:()=>fc,logSoftmax:()=>bc,logSumExp:()=>Nc,logicalAnd:()=>Cc,logicalNot:()=>Tc,logicalOr:()=>Rc,logicalXor:()=>_c,losses:()=>Td,matMul:()=>Gi,math:()=>ji,max:()=>mc,maxPool:()=>Fc,maxPool3d:()=>Mc,maxPoolWithArgmax:()=>Dc,maximum:()=>Oc,mean:()=>Lc,memory:()=>tu,meshgrid:()=>Bc,metrics:()=>bS,min:()=>zc,minimum:()=>Uc,mirrorPad:()=>Wc,mod:()=>Vc,model:()=>qv,models:()=>DS,moments:()=>Gc,movingAverage:()=>cp,mul:()=>vu,multiRNNCell:()=>Hc,multinomial:()=>qc,neg:()=>pc,nextFrame:()=>Vd,norm:()=>lp,notEqual:()=>Kc,oneHot:()=>Hi,ones:()=>$c,onesLike:()=>Yc,op:()=>Ms,outerProduct:()=>Xc,pad:()=>Qc,pad1d:()=>Zc,pad2d:()=>Jc,pad3d:()=>eh,pad4d:()=>th,pool:()=>rh,pow:()=>ah,prelu:()=>sh,print:()=>Ni,prod:()=>ih,profile:()=>nu,rand:()=>oh,randomGamma:()=>ch,randomNormal:()=>hh,randomUniform:()=>ph,range:()=>dh,ready:()=>uu,real:()=>fh,reciprocal:()=>mh,registerBackend:()=>du,registerCallbackConstructor:()=>Qv,registerGradient:()=>fa,registerKernel:()=>da,registerOp:()=>XS,regularizers:()=>OS,relu:()=>gh,relu6:()=>yh,removeBackend:()=>cu,reshape:()=>qu,reverse:()=>bh,reverse1d:()=>vh,reverse2d:()=>xh,reverse3d:()=>wh,reverse4d:()=>kh,rfft:()=>Uh,round:()=>Sh,rsqrt:()=>Ih,scalar:()=>Ah,scatterND:()=>hp,scatter_util:()=>lo,selu:()=>Eh,separableConv2d:()=>Nh,sequential:()=>Kv,serialization:()=>Fo,setBackend:()=>ou,setPlatform:()=>mu,setThreadsCount:()=>Fj,setWasmPath:()=>Cj,setWasmPaths:()=>Tj,setWebGLContext:()=>KF,setdiff1dAsync:()=>Ch,sigmoid:()=>Qu,sign:()=>Th,signal:()=>Ed,sin:()=>Rh,sinh:()=>_h,slice:()=>Zu,slice1d:()=>Fh,slice2d:()=>Mh,slice3d:()=>Dh,slice4d:()=>Oh,slice_util:()=>fo,softmax:()=>Lh,softplus:()=>dc,spaceToBatchND:()=>nh,sparse:()=>Rd,sparseToDense:()=>pp,spectral:()=>Ad,split:()=>zh,sqrt:()=>Wh,square:()=>jc,squaredDifference:()=>Vh,squeeze:()=>jh,stack:()=>Gh,step:()=>Hh,stridedSlice:()=>qh,string:()=>_d,sub:()=>gc,sum:()=>yc,sumOutType:()=>hs,tan:()=>Kh,tanh:()=>Ju,tensor:()=>Ls,tensor1d:()=>Yh,tensor2d:()=>Xh,tensor3d:()=>to,tensor4d:()=>Qh,tensor5d:()=>Zh,tensor6d:()=>Jh,tensor_util:()=>us,test_util:()=>Lo,tidy:()=>ru,tile:()=>Ul,time:()=>iu,topk:()=>ep,train:()=>Ud,transpose:()=>qi,truncatedNormal:()=>tp,unique:()=>np,unregisterGradient:()=>ga,unregisterKernel:()=>ma,unsortedSegmentSum:()=>rp,unstack:()=>ap,upcastType:()=>cs,util:()=>va,valueAndGrad:()=>oc,valueAndGrads:()=>uc,variable:()=>sp,variableGrads:()=>lc,version:()=>Oj,version_converter:()=>EA,version_core:()=>Ko,version_layers:()=>mv,version_wasm:()=>Dj,version_webgl:()=>GO,webgl:()=>qO,webgl_util:()=>BF,where:()=>_l,whereAsync:()=>op,zeros:()=>Pc,zerosLike:()=>Fl});var f=Object.create,m=Object.defineProperty,g=Object.getOwnPropertyDescriptor,y=Object.getOwnPropertyNames,b=Object.getPrototypeOf,v=Object.prototype.hasOwnProperty,x=(e,t)=>function(){return t||(0,e[y(e)[0]])((t={exports:{}}).exports,t),t.exports},w=(e,t)=>{for(var n in t)m(e,n,{get:t[n],enumerable:!0})},k=(e,t)=>{return((e,t,n,r)=>{if(t&&"object"==typeof t||"function"==typeof t)for(let a of y(t))v.call(e,a)||!n&&"default"===a||m(e,a,{get:()=>t[a],enumerable:!(r=g(t,a))||r.enumerable});return e})((n=m(null!=e?f(b(e)):{},"default",!t&&e&&e.__esModule?{get:()=>e.default,enumerable:!0}:{value:e,enumerable:!0}),m(n,"__esModule",{value:!0})),e);var n},S=x({"node_modules/.pnpm/long@4.0.0/node_modules/long/src/long.js"(e,t){t.exports=r;var n=null;try{n=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(e){}function r(e,t,n){this.low=0|e,this.high=0|t,this.unsigned=!!n}function a(e){return!0===(e&&e.__isLong__)}Object.defineProperty(r.prototype,"__isLong__",{value:!0}),r.isLong=a;var s={},i={};function o(e,t){var n,r,a;return t?(a=0<=(e>>>=0)&&e<256)&&(r=i[e])?r:(n=l(e,(0|e)<0?-1:0,!0),a&&(i[e]=n),n):(a=-128<=(e|=0)&&e<128)&&(r=s[e])?r:(n=l(e,e<0?-1:0,!1),a&&(s[e]=n),n)}function u(e,t){if(isNaN(e))return t?b:y;if(t){if(e<0)return b;if(e>=f)return S}else{if(e<=-m)return I;if(e+1>=m)return k}return e<0?u(-e,t).neg():l(e%d|0,e/d|0,t)}function l(e,t,n){return new r(e,t,n)}r.fromInt=o,r.fromNumber=u,r.fromBits=l;var c=Math.pow;function h(e,t,n){if(0===e.length)throw Error("empty string");if("NaN"===e||"Infinity"===e||"+Infinity"===e||"-Infinity"===e)return y;if("number"==typeof t?(n=t,t=!1):t=!!t,(n=n||10)<2||36<n)throw RangeError("radix");var r;if((r=e.indexOf("-"))>0)throw Error("interior hyphen");if(0===r)return h(e.substring(1),t,n).neg();for(var a=u(c(n,8)),s=y,i=0;i<e.length;i+=8){var o=Math.min(8,e.length-i),l=parseInt(e.substring(i,i+o),n);if(o<8){var p=u(c(n,o));s=s.mul(p).add(u(l))}else s=(s=s.mul(a)).add(u(l))}return s.unsigned=t,s}function p(e,t){return"number"==typeof e?u(e,t):"string"==typeof e?h(e,t):l(e.low,e.high,"boolean"==typeof t?t:e.unsigned)}r.fromString=h,r.fromValue=p;var d=4294967296,f=d*d,m=f/2,g=o(1<<24),y=o(0);r.ZERO=y;var b=o(0,!0);r.UZERO=b;var v=o(1);r.ONE=v;var x=o(1,!0);r.UONE=x;var w=o(-1);r.NEG_ONE=w;var k=l(-1,2147483647,!1);r.MAX_VALUE=k;var S=l(-1,-1,!0);r.MAX_UNSIGNED_VALUE=S;var I=l(0,-2147483648,!1);r.MIN_VALUE=I;var A=r.prototype;A.toInt=function(){return this.unsigned?this.low>>>0:this.low},A.toNumber=function(){return this.unsigned?(this.high>>>0)*d+(this.low>>>0):this.high*d+(this.low>>>0)},A.toString=function(e){if((e=e||10)<2||36<e)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative()){if(this.eq(I)){var t=u(e),n=this.div(t),r=n.mul(t).sub(this);return n.toString(e)+r.toInt().toString(e)}return"-"+this.neg().toString(e)}for(var a=u(c(e,6),this.unsigned),s=this,i="";;){var o=s.div(a),l=(s.sub(o.mul(a)).toInt()>>>0).toString(e);if((s=o).isZero())return l+i;for(;l.length<6;)l="0"+l;i=""+l+i}},A.getHighBits=function(){return this.high},A.getHighBitsUnsigned=function(){return this.high>>>0},A.getLowBits=function(){return this.low},A.getLowBitsUnsigned=function(){return this.low>>>0},A.getNumBitsAbs=function(){if(this.isNegative())return this.eq(I)?64:this.neg().getNumBitsAbs();for(var e=0!=this.high?this.high:this.low,t=31;t>0&&0==(e&1<<t);t--);return 0!=this.high?t+33:t+1},A.isZero=function(){return 0===this.high&&0===this.low},A.eqz=A.isZero,A.isNegative=function(){return!this.unsigned&&this.high<0},A.isPositive=function(){return this.unsigned||this.high>=0},A.isOdd=function(){return 1==(1&this.low)},A.isEven=function(){return 0==(1&this.low)},A.equals=function(e){return a(e)||(e=p(e)),(this.unsigned===e.unsigned||this.high>>>31!=1||e.high>>>31!=1)&&(this.high===e.high&&this.low===e.low)},A.eq=A.equals,A.notEquals=function(e){return!this.eq(e)},A.neq=A.notEquals,A.ne=A.notEquals,A.lessThan=function(e){return this.comp(e)<0},A.lt=A.lessThan,A.lessThanOrEqual=function(e){return this.comp(e)<=0},A.lte=A.lessThanOrEqual,A.le=A.lessThanOrEqual,A.greaterThan=function(e){return this.comp(e)>0},A.gt=A.greaterThan,A.greaterThanOrEqual=function(e){return this.comp(e)>=0},A.gte=A.greaterThanOrEqual,A.ge=A.greaterThanOrEqual,A.compare=function(e){if(a(e)||(e=p(e)),this.eq(e))return 0;var t=this.isNegative(),n=e.isNegative();return t&&!n?-1:!t&&n?1:this.unsigned?e.high>>>0>this.high>>>0||e.high===this.high&&e.low>>>0>this.low>>>0?-1:1:this.sub(e).isNegative()?-1:1},A.comp=A.compare,A.negate=function(){return!this.unsigned&&this.eq(I)?I:this.not().add(v)},A.neg=A.negate,A.add=function(e){a(e)||(e=p(e));var t=this.high>>>16,n=65535&this.high,r=this.low>>>16,s=65535&this.low,i=e.high>>>16,o=65535&e.high,u=e.low>>>16,c=0,h=0,d=0,f=0;return d+=(f+=s+(65535&e.low))>>>16,h+=(d+=r+u)>>>16,c+=(h+=n+o)>>>16,c+=t+i,l((d&=65535)<<16|(f&=65535),(c&=65535)<<16|(h&=65535),this.unsigned)},A.subtract=function(e){return a(e)||(e=p(e)),this.add(e.neg())},A.sub=A.subtract,A.multiply=function(e){if(this.isZero())return y;if(a(e)||(e=p(e)),n)return l(n.mul(this.low,this.high,e.low,e.high),n.get_high(),this.unsigned);if(e.isZero())return y;if(this.eq(I))return e.isOdd()?I:y;if(e.eq(I))return this.isOdd()?I:y;if(this.isNegative())return e.isNegative()?this.neg().mul(e.neg()):this.neg().mul(e).neg();if(e.isNegative())return this.mul(e.neg()).neg();if(this.lt(g)&&e.lt(g))return u(this.toNumber()*e.toNumber(),this.unsigned);var t=this.high>>>16,r=65535&this.high,s=this.low>>>16,i=65535&this.low,o=e.high>>>16,c=65535&e.high,h=e.low>>>16,d=65535&e.low,f=0,m=0,b=0,v=0;return b+=(v+=i*d)>>>16,m+=(b+=s*d)>>>16,b&=65535,m+=(b+=i*h)>>>16,f+=(m+=r*d)>>>16,m&=65535,f+=(m+=s*h)>>>16,m&=65535,f+=(m+=i*c)>>>16,f+=t*d+r*h+s*c+i*o,l((b&=65535)<<16|(v&=65535),(f&=65535)<<16|(m&=65535),this.unsigned)},A.mul=A.multiply,A.divide=function(e){if(a(e)||(e=p(e)),e.isZero())throw Error("division by zero");var t,r,s;if(n)return this.unsigned||-2147483648!==this.high||-1!==e.low||-1!==e.high?l((this.unsigned?n.div_u:n.div_s)(this.low,this.high,e.low,e.high),n.get_high(),this.unsigned):this;if(this.isZero())return this.unsigned?b:y;if(this.unsigned){if(e.unsigned||(e=e.toUnsigned()),e.gt(this))return b;if(e.gt(this.shru(1)))return x;s=b}else{if(this.eq(I))return e.eq(v)||e.eq(w)?I:e.eq(I)?v:(t=this.shr(1).div(e).shl(1)).eq(y)?e.isNegative()?v:w:(r=this.sub(e.mul(t)),s=t.add(r.div(e)));if(e.eq(I))return this.unsigned?b:y;if(this.isNegative())return e.isNegative()?this.neg().div(e.neg()):this.neg().div(e).neg();if(e.isNegative())return this.div(e.neg()).neg();s=y}for(r=this;r.gte(e);){t=Math.max(1,Math.floor(r.toNumber()/e.toNumber()));for(var i=Math.ceil(Math.log(t)/Math.LN2),o=i<=48?1:c(2,i-48),h=u(t),d=h.mul(e);d.isNegative()||d.gt(r);)d=(h=u(t-=o,this.unsigned)).mul(e);h.isZero()&&(h=v),s=s.add(h),r=r.sub(d)}return s},A.div=A.divide,A.modulo=function(e){return a(e)||(e=p(e)),n?l((this.unsigned?n.rem_u:n.rem_s)(this.low,this.high,e.low,e.high),n.get_high(),this.unsigned):this.sub(this.div(e).mul(e))},A.mod=A.modulo,A.rem=A.modulo,A.not=function(){return l(~this.low,~this.high,this.unsigned)},A.and=function(e){return a(e)||(e=p(e)),l(this.low&e.low,this.high&e.high,this.unsigned)},A.or=function(e){return a(e)||(e=p(e)),l(this.low|e.low,this.high|e.high,this.unsigned)},A.xor=function(e){return a(e)||(e=p(e)),l(this.low^e.low,this.high^e.high,this.unsigned)},A.shiftLeft=function(e){return a(e)&&(e=e.toInt()),0==(e&=63)?this:e<32?l(this.low<<e,this.high<<e|this.low>>>32-e,this.unsigned):l(0,this.low<<e-32,this.unsigned)},A.shl=A.shiftLeft,A.shiftRight=function(e){return a(e)&&(e=e.toInt()),0==(e&=63)?this:e<32?l(this.low>>>e|this.high<<32-e,this.high>>e,this.unsigned):l(this.high>>e-32,this.high>=0?0:-1,this.unsigned)},A.shr=A.shiftRight,A.shiftRightUnsigned=function(e){if(a(e)&&(e=e.toInt()),0===(e&=63))return this;var t=this.high;return e<32?l(this.low>>>e|t<<32-e,t>>>e,this.unsigned):l(32===e?t:t>>>e-32,0,this.unsigned)},A.shru=A.shiftRightUnsigned,A.shr_u=A.shiftRightUnsigned,A.toSigned=function(){return this.unsigned?l(this.low,this.high,!1):this},A.toUnsigned=function(){return this.unsigned?this:l(this.low,this.high,!0)},A.toBytes=function(e){return e?this.toBytesLE():this.toBytesBE()},A.toBytesLE=function(){var e=this.high,t=this.low;return[255&t,t>>>8&255,t>>>16&255,t>>>24,255&e,e>>>8&255,e>>>16&255,e>>>24]},A.toBytesBE=function(){var e=this.high,t=this.low;return[e>>>24,e>>>16&255,e>>>8&255,255&e,t>>>24,t>>>16&255,t>>>8&255,255&t]},r.fromBytes=function(e,t,n){return n?r.fromBytesLE(e,t):r.fromBytesBE(e,t)},r.fromBytesLE=function(e,t){return new r(e[0]|e[1]<<8|e[2]<<16|e[3]<<24,e[4]|e[5]<<8|e[6]<<16|e[7]<<24,t)},r.fromBytesBE=function(e,t){return new r(e[4]<<24|e[5]<<16|e[6]<<8|e[7],e[0]<<24|e[1]<<16|e[2]<<8|e[3],t)}}}),I=x({"(disabled):node_modules/.pnpm/node-fetch@2.6.6/node_modules/node-fetch/browser.js"(){}}),A=x({"(disabled):util"(){}}),E=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/alea.js"(e,t){!function(e,t,n){function r(e){var t,n=this,r=(t=4022871197,function(e){e=e.toString();for(var n=0;n<e.length;n++){var r=.02519603282416938*(t+=e.charCodeAt(n));r-=t=r>>>0,t=(r*=t)>>>0,t+=4294967296*(r-=t)}return 2.3283064365386963e-10*(t>>>0)});n.next=function(){var e=2091639*n.s0+2.3283064365386963e-10*n.c;return n.s0=n.s1,n.s1=n.s2,n.s2=e-(n.c=0|e)},n.c=1,n.s0=r(" "),n.s1=r(" "),n.s2=r(" "),n.s0-=r(e),n.s0<0&&(n.s0+=1),n.s1-=r(e),n.s1<0&&(n.s1+=1),n.s2-=r(e),n.s2<0&&(n.s2+=1),r=null}function a(e,t){return t.c=e.c,t.s0=e.s0,t.s1=e.s1,t.s2=e.s2,t}function s(e,t){var n=new r(e),s=t&&t.state,i=n.next;return i.int32=function(){return 4294967296*n.next()|0},i.double=function(){return i()+11102230246251565e-32*(2097152*i()|0)},i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.alea=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),N=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xor128.js"(e,t){!function(e,t,n){function r(e){var t=this,n="";t.x=0,t.y=0,t.z=0,t.w=0,t.next=function(){var e=t.x^t.x<<11;return t.x=t.y,t.y=t.z,t.z=t.w,t.w^=t.w>>>19^e^e>>>8},e===(0|e)?t.x=e:n+=e;for(var r=0;r<n.length+64;r++)t.x^=0|n.charCodeAt(r),t.next()}function a(e,t){return t.x=e.x,t.y=e.y,t.z=e.z,t.w=e.w,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xor128=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),C=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xorwow.js"(e,t){!function(e,t,n){function r(e){var t=this,n="";t.next=function(){var e=t.x^t.x>>>2;return t.x=t.y,t.y=t.z,t.z=t.w,t.w=t.v,(t.d=t.d+362437|0)+(t.v=t.v^t.v<<4^e^e<<1)|0},t.x=0,t.y=0,t.z=0,t.w=0,t.v=0,e===(0|e)?t.x=e:n+=e;for(var r=0;r<n.length+64;r++)t.x^=0|n.charCodeAt(r),r==n.length&&(t.d=t.x<<10^t.x>>>4),t.next()}function a(e,t){return t.x=e.x,t.y=e.y,t.z=e.z,t.w=e.w,t.v=e.v,t.d=e.d,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xorwow=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),T=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xorshift7.js"(e,t){!function(e,t,n){function r(e){var t=this;t.next=function(){var e,n,r=t.x,a=t.i;return e=r[a],n=(e^=e>>>7)^e<<24,n^=(e=r[a+1&7])^e>>>10,n^=(e=r[a+3&7])^e>>>3,n^=(e=r[a+4&7])^e<<7,e=r[a+7&7],n^=(e^=e<<13)^e<<9,r[a]=n,t.i=a+1&7,n},function(e,t){var n,r=[];if(t===(0|t))r[0]=t;else for(t=""+t,n=0;n<t.length;++n)r[7&n]=r[7&n]<<15^t.charCodeAt(n)+r[n+1&7]<<13;for(;r.length<8;)r.push(0);for(n=0;n<8&&0===r[n];++n);for(8==n&&(r[7]=-1),e.x=r,e.i=0,n=256;n>0;--n)e.next()}(t,e)}function a(e,t){return t.x=e.x.slice(),t.i=e.i,t}function s(e,t){null==e&&(e=+new Date);var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&(s.x&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xorshift7=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),R=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xor4096.js"(e,t){!function(e,t,n){function r(e){var t=this;t.next=function(){var e,n,r=t.w,a=t.X,s=t.i;return t.w=r=r+1640531527|0,n=a[s+34&127],e=a[s=s+1&127],n^=n<<13,e^=e<<17,n^=n>>>15,e^=e>>>12,n=a[s]=n^e,t.i=s,n+(r^r>>>16)|0},function(e,t){var n,r,a,s,i,o=[],u=128;for(t===(0|t)?(r=t,t=null):(t+="\0",r=0,u=Math.max(u,t.length)),a=0,s=-32;s<u;++s)t&&(r^=t.charCodeAt((s+32)%t.length)),0===s&&(i=r),r^=r<<10,r^=r>>>15,r^=r<<4,r^=r>>>13,s>=0&&(i=i+1640531527|0,a=0==(n=o[127&s]^=r+i)?a+1:0);for(a>=128&&(o[127&(t&&t.length||0)]=-1),a=127,s=512;s>0;--s)r=o[a+34&127],n=o[a=a+1&127],r^=r<<13,n^=n<<17,r^=r>>>15,n^=n>>>12,o[a]=r^n;e.w=i,e.X=o,e.i=a}(t,e)}function a(e,t){return t.i=e.i,t.w=e.w,t.X=e.X.slice(),t}function s(e,t){null==e&&(e=+new Date);var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&(s.X&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xor4096=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),_=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/tychei.js"(e,t){!function(e,t,n){function r(e){var t=this,n="";t.next=function(){var e=t.b,n=t.c,r=t.d,a=t.a;return e=e<<25^e>>>7^n,n=n-r|0,r=r<<24^r>>>8^a,a=a-e|0,t.b=e=e<<20^e>>>12^n,t.c=n=n-r|0,t.d=r<<16^n>>>16^a,t.a=a-e|0},t.a=0,t.b=0,t.c=-1640531527,t.d=1367130551,e===Math.floor(e)?(t.a=e/4294967296|0,t.b=0|e):n+=e;for(var r=0;r<n.length+20;r++)t.b^=0|n.charCodeAt(r),t.next()}function a(e,t){return t.a=e.a,t.b=e.b,t.c=e.c,t.d=e.d,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.tychei=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),F=x({"(disabled):crypto"(){}}),M=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/seedrandom.js"(e,t){!function(e,r){var a,s=this,i=r.pow(256,6),o=r.pow(2,52),u=2*o;function l(t,n,l){var f=[],m=p(function e(t,n){var r,a=[],s=typeof t;if(n&&"object"==s)for(r in t)try{a.push(e(t[r],n-1))}catch(e){}return a.length?a:"string"==s?t:t+"\0"}((n=1==n?{entropy:!0}:n||{}).entropy?[t,d(e)]:null==t?function(){try{var t;return a&&(t=a.randomBytes)?t=t(256):(t=new Uint8Array(256),(s.crypto||s.msCrypto).getRandomValues(t)),d(t)}catch(t){var n=s.navigator,r=n&&n.plugins;return[+new Date,s,r,s.screen,d(e)]}}():t,3),f),g=new c(f),y=function(){for(var e=g.g(6),t=i,n=0;e<o;)e=256*(e+n),t*=256,n=g.g(1);for(;e>=u;)e/=2,t/=2,n>>>=1;return(e+n)/t};return y.int32=function(){return 0|g.g(4)},y.quick=function(){return g.g(4)/4294967296},y.double=y,p(d(g.S),e),(n.pass||l||function(e,t,n,a){return a&&(a.S&&h(a,g),e.state=function(){return h(g,{})}),n?(r.random=e,t):e})(y,m,"global"in n?n.global:this==r,n.state)}function c(e){var t,n=e.length,r=this,a=0,s=r.i=r.j=0,i=r.S=[];for(n||(e=[n++]);a<256;)i[a]=a++;for(a=0;a<256;a++)i[a]=i[s=255&s+e[a%n]+(t=i[a])],i[s]=t;(r.g=function(e){for(var t,n=0,a=r.i,s=r.j,i=r.S;e--;)t=i[a=255&a+1],n=256*n+i[255&(i[a]=i[s=255&s+t])+(i[s]=t)];return r.i=a,r.j=s,n})(256)}function h(e,t){return t.i=e.i,t.j=e.j,t.S=e.S.slice(),t}function p(e,t){for(var n,r=e+"",a=0;a<r.length;)t[255&a]=255&(n^=19*t[255&a])+r.charCodeAt(a++);return d(t)}function d(e){return String.fromCharCode.apply(0,e)}if(r.seedrandom=l,p(r.random(),e),"object"==typeof t&&t.exports){t.exports=l;try{a=F()}catch(e){}}else"function"==typeof define&&n(66)&&define((function(){return l}))}([],Math)}}),D=x({"node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/index.js"(e,t){var n=E(),r=N(),a=C(),s=T(),i=R(),o=_(),u=M();u.alea=n,u.xor128=r,u.xorwow=a,u.xorshift7=s,u.xor4096=i,u.tychei=o,t.exports=u}}),O=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/alea.js"(e,t){!function(e,t,n){function r(e){var t,n=this,r=(t=4022871197,function(e){e=String(e);for(var n=0;n<e.length;n++){var r=.02519603282416938*(t+=e.charCodeAt(n));r-=t=r>>>0,t=(r*=t)>>>0,t+=4294967296*(r-=t)}return 2.3283064365386963e-10*(t>>>0)});n.next=function(){var e=2091639*n.s0+2.3283064365386963e-10*n.c;return n.s0=n.s1,n.s1=n.s2,n.s2=e-(n.c=0|e)},n.c=1,n.s0=r(" "),n.s1=r(" "),n.s2=r(" "),n.s0-=r(e),n.s0<0&&(n.s0+=1),n.s1-=r(e),n.s1<0&&(n.s1+=1),n.s2-=r(e),n.s2<0&&(n.s2+=1),r=null}function a(e,t){return t.c=e.c,t.s0=e.s0,t.s1=e.s1,t.s2=e.s2,t}function s(e,t){var n=new r(e),s=t&&t.state,i=n.next;return i.int32=function(){return 4294967296*n.next()|0},i.double=function(){return i()+11102230246251565e-32*(2097152*i()|0)},i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.alea=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),L=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xor128.js"(e,t){!function(e,t,n){function r(e){var t=this,n="";t.x=0,t.y=0,t.z=0,t.w=0,t.next=function(){var e=t.x^t.x<<11;return t.x=t.y,t.y=t.z,t.z=t.w,t.w^=t.w>>>19^e^e>>>8},e===(0|e)?t.x=e:n+=e;for(var r=0;r<n.length+64;r++)t.x^=0|n.charCodeAt(r),t.next()}function a(e,t){return t.x=e.x,t.y=e.y,t.z=e.z,t.w=e.w,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xor128=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),P=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xorwow.js"(e,t){!function(e,t,n){function r(e){var t=this,n="";t.next=function(){var e=t.x^t.x>>>2;return t.x=t.y,t.y=t.z,t.z=t.w,t.w=t.v,(t.d=t.d+362437|0)+(t.v=t.v^t.v<<4^e^e<<1)|0},t.x=0,t.y=0,t.z=0,t.w=0,t.v=0,e===(0|e)?t.x=e:n+=e;for(var r=0;r<n.length+64;r++)t.x^=0|n.charCodeAt(r),r==n.length&&(t.d=t.x<<10^t.x>>>4),t.next()}function a(e,t){return t.x=e.x,t.y=e.y,t.z=e.z,t.w=e.w,t.v=e.v,t.d=e.d,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xorwow=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),$=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xorshift7.js"(e,t){!function(e,t,n){function r(e){var t=this;t.next=function(){var e,n,r=t.x,a=t.i;return e=r[a],n=(e^=e>>>7)^e<<24,n^=(e=r[a+1&7])^e>>>10,n^=(e=r[a+3&7])^e>>>3,n^=(e=r[a+4&7])^e<<7,e=r[a+7&7],n^=(e^=e<<13)^e<<9,r[a]=n,t.i=a+1&7,n},function(e,t){var n,r=[];if(t===(0|t))r[0]=t;else for(t=""+t,n=0;n<t.length;++n)r[7&n]=r[7&n]<<15^t.charCodeAt(n)+r[n+1&7]<<13;for(;r.length<8;)r.push(0);for(n=0;n<8&&0===r[n];++n);for(8==n&&(r[7]=-1),e.x=r,e.i=0,n=256;n>0;--n)e.next()}(t,e)}function a(e,t){return t.x=e.x.slice(),t.i=e.i,t}function s(e,t){null==e&&(e=+new Date);var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&(s.x&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xorshift7=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),B=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xor4096.js"(e,t){!function(e,t,n){function r(e){var t=this;t.next=function(){var e,n,r=t.w,a=t.X,s=t.i;return t.w=r=r+1640531527|0,n=a[s+34&127],e=a[s=s+1&127],n^=n<<13,e^=e<<17,n^=n>>>15,e^=e>>>12,n=a[s]=n^e,t.i=s,n+(r^r>>>16)|0},function(e,t){var n,r,a,s,i,o=[],u=128;for(t===(0|t)?(r=t,t=null):(t+="\0",r=0,u=Math.max(u,t.length)),a=0,s=-32;s<u;++s)t&&(r^=t.charCodeAt((s+32)%t.length)),0===s&&(i=r),r^=r<<10,r^=r>>>15,r^=r<<4,r^=r>>>13,s>=0&&(i=i+1640531527|0,a=0==(n=o[127&s]^=r+i)?a+1:0);for(a>=128&&(o[127&(t&&t.length||0)]=-1),a=127,s=512;s>0;--s)r=o[a+34&127],n=o[a=a+1&127],r^=r<<13,n^=n<<17,r^=r>>>15,n^=n>>>12,o[a]=r^n;e.w=i,e.X=o,e.i=a}(t,e)}function a(e,t){return t.i=e.i,t.w=e.w,t.X=e.X.slice(),t}function s(e,t){null==e&&(e=+new Date);var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&(s.X&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.xor4096=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),z=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/tychei.js"(e,t){!function(e,t,n){function r(e){var t=this,n="";t.next=function(){var e=t.b,n=t.c,r=t.d,a=t.a;return e=e<<25^e>>>7^n,n=n-r|0,r=r<<24^r>>>8^a,a=a-e|0,t.b=e=e<<20^e>>>12^n,t.c=n=n-r|0,t.d=r<<16^n>>>16^a,t.a=a-e|0},t.a=0,t.b=0,t.c=-1640531527,t.d=1367130551,e===Math.floor(e)?(t.a=e/4294967296|0,t.b=0|e):n+=e;for(var r=0;r<n.length+20;r++)t.b^=0|n.charCodeAt(r),t.next()}function a(e,t){return t.a=e.a,t.b=e.b,t.c=e.c,t.d=e.d,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n((function(){return s})):this.tychei=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}}),U=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/seedrandom.js"(e,t){!function(e,r,a){var s,i=a.pow(256,6),o=a.pow(2,52),u=2*o;function l(t,n,l){var f=[],m=p(function e(t,n){var r,a=[],s=typeof t;if(n&&"object"==s)for(r in t)try{a.push(e(t[r],n-1))}catch(e){}return a.length?a:"string"==s?t:t+"\0"}((n=1==n?{entropy:!0}:n||{}).entropy?[t,d(r)]:null==t?function(){try{var t;return s&&(t=s.randomBytes)?t=t(256):(t=new Uint8Array(256),(e.crypto||e.msCrypto).getRandomValues(t)),d(t)}catch(t){var n=e.navigator,a=n&&n.plugins;return[+new Date,e,a,e.screen,d(r)]}}():t,3),f),g=new c(f),y=function(){for(var e=g.g(6),t=i,n=0;e<o;)e=256*(e+n),t*=256,n=g.g(1);for(;e>=u;)e/=2,t/=2,n>>>=1;return(e+n)/t};return y.int32=function(){return 0|g.g(4)},y.quick=function(){return g.g(4)/4294967296},y.double=y,p(d(g.S),r),(n.pass||l||function(e,t,n,r){return r&&(r.S&&h(r,g),e.state=function(){return h(g,{})}),n?(a.random=e,t):e})(y,m,"global"in n?n.global:this==a,n.state)}function c(e){var t,n=e.length,r=this,a=0,s=r.i=r.j=0,i=r.S=[];for(n||(e=[n++]);a<256;)i[a]=a++;for(a=0;a<256;a++)i[a]=i[s=255&s+e[a%n]+(t=i[a])],i[s]=t;(r.g=function(e){for(var t,n=0,a=r.i,s=r.j,i=r.S;e--;)t=i[a=255&a+1],n=256*n+i[255&(i[a]=i[s=255&s+t])+(i[s]=t)];return r.i=a,r.j=s,n})(256)}function h(e,t){return t.i=e.i,t.j=e.j,t.S=e.S.slice(),t}function p(e,t){for(var n,r=e+"",a=0;a<r.length;)t[255&a]=255&(n^=19*t[255&a])+r.charCodeAt(a++);return d(t)}function d(e){return String.fromCharCode.apply(0,e)}if(p(a.random(),r),"object"==typeof t&&t.exports){t.exports=l;try{s=F()}catch(e){}}else"function"==typeof define&&n(66)?define((function(){return l})):a.seedrandom=l}("undefined"!=typeof self?self:e,[],Math)}}),W=x({"node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/index.js"(e,t){var n=O(),r=L(),a=P(),s=$(),i=B(),o=z(),u=U();u.alea=n,u.xor128=r,u.xorwow=a,u.xorshift7=s,u.xor4096=i,u.tychei=o,t.exports=u}}),V=x({"(disabled):node_modules/.pnpm/string_decoder@1.1.1/node_modules/string_decoder/lib/string_decoder.js"(){}}),j=x({"(disabled):fs"(){}}),G=x({"(disabled):path"(){}}),H=x({"(disabled):worker_threads"(){}}),q=x({"(disabled):perf_hooks"(){}}),K=x({"(disabled):os"(){}}),Y=x({"node_modules/.pnpm/@tensorflow+tfjs-backend-wasm@3.13.0_@tensorflow+tfjs-core@3.13.0/node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js"(t,i){var o,u=(o=(o="undefined"!=typeof document&&document.currentScript?document.currentScript.src:void 0)||e,function(e){function t(){return M.buffer!=L&&ee(M.buffer),P}function n(){return M.buffer!=L&&ee(M.buffer),$}function i(){return M.buffer!=L&&ee(M.buffer),B}function u(){return M.buffer!=L&&ee(M.buffer),z}function l(){return M.buffer!=L&&ee(M.buffer),U}var c,h,p,d=void 0!==(e=e||{})?e:{};d.ready=new Promise((function(e,t){c=e,h=t})),void 0!==r&&r.listeners&&(p={uncaughtException:r.listeners("uncaughtException"),unhandledRejection:r.listeners("unhandledRejection")});var f,m={};for(f in d)d.hasOwnProperty(f)&&(m[f]=d[f]);var g,y,b,v,x=function(e,t){throw t};g="object"==typeof window,y="function"==typeof importScripts,b="object"==typeof r&&"object"==typeof r.versions&&"string"==typeof r.versions.node,v=!g&&!b&&!y;var w=d.ENVIRONMENT_IS_PTHREAD||!1;w&&(L=d.buffer);var k,S,I,A,E,N="";function C(e){return d.locateFile?d.locateFile(e,N):N+e}if(b){var T;N=y?G().dirname(N)+"/":a+"/",k=function(e,t){return A||(A=j()),E||(E=G()),e=E.normalize(e),A.readFileSync(e,t?null:"utf8")},I=function(e){var t=k(e,!0);return t.buffer||(t=new Uint8Array(t)),V(t.buffer),t},r.argv.length>1&&r.argv[1].replace(/\\/g,"/"),r.argv.slice(2),r.on("uncaughtException",(function(e){if(!(e instanceof kt))throw e})),r.on("unhandledRejection",ce),x=function(e){r.exit(e)},d.inspect=function(){return"[Emscripten Module object]"};try{T=H()}catch(e){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),e}s.Worker=T.Worker}else v?("undefined"!=typeof read&&(k=function(e){return read(e)}),I=function(e){var t;return"function"==typeof readbuffer?new Uint8Array(readbuffer(e)):(V("object"==typeof(t=read(e,"binary"))),t)},"undefined"!=typeof scriptArgs&&scriptArgs,"function"==typeof quit&&(x=function(e){quit(e)}),"undefined"!=typeof print&&("undefined"==typeof console&&(console={}),console.log=print,console.warn=console.error="undefined"!=typeof printErr?printErr:print)):(g||y)&&(y?N=self.location.href:"undefined"!=typeof document&&document.currentScript&&(N=document.currentScript.src),void 0!==o&&o&&(N=o),N=0!==N.indexOf("blob:")?N.substr(0,N.lastIndexOf("/")+1):"",b?(k=function(e,t){return A||(A=j()),E||(E=G()),e=E.normalize(e),A.readFileSync(e,t?null:"utf8")},I=function(e){var t=k(e,!0);return t.buffer||(t=new Uint8Array(t)),V(t.buffer),t}):(k=function(e){var t=new XMLHttpRequest;return t.open("GET",e,!1),t.send(null),t.responseText},y&&(I=function(e){var t=new XMLHttpRequest;return t.open("GET",e,!1),t.responseType="arraybuffer",t.send(null),new Uint8Array(t.response)}),S=function(e,t,n){var r=new XMLHttpRequest;r.open("GET",e,!0),r.responseType="arraybuffer",r.onload=function(){200==r.status||0==r.status&&r.response?t(r.response):n()},r.onerror=n,r.send(null)}));b&&"undefined"==typeof performance&&(s.performance=q().performance);var R,_=d.print||console.log.bind(console),F=d.printErr||console.warn.bind(console);for(f in m)m.hasOwnProperty(f)&&(d[f]=m[f]);m=null,d.quit&&(x=d.quit),d.wasmBinary&&(R=d.wasmBinary);var M,D,O=d.noExitRuntime||!0;"object"!=typeof WebAssembly&&ce("no native wasm support detected");var L,P,$,B,z,U,W=!1;function V(e,t){e||ce("Assertion failed: "+t)}function Y(e){var t=d["_"+e];return V(t,"Cannot call unknown function "+e+", make sure it is exported"),t}function X(e,n,r,a,s){var i={string:function(e){var t=0;if(null!=e&&0!==e){var n=1+(e.length<<2);J(e,t=yt(n),n)}return t},array:function(e){var n,r,a=yt(e.length);return n=e,r=a,t().set(n,r),a}},o=Y(e),u=[],l=0;if(a)for(var c=0;c<a.length;c++){var h=i[r[c]];h?(0===l&&(l=mt()),u[c]=h(a[c])):u[c]=a[c]}var p,d=o.apply(null,u);return p=d,d="string"===n?Z(p):"boolean"===n?Boolean(p):p,0!==l&&gt(l),d}function Q(e,t,n){for(var r=t+n,a="";!(t>=r);){var s=e[t++];if(!s)return a;if(128&s){var i=63&e[t++];if(192!=(224&s)){var o=63&e[t++];if((s=224==(240&s)?(15&s)<<12|i<<6|o:(7&s)<<18|i<<12|o<<6|63&e[t++])<65536)a+=String.fromCharCode(s);else{var u=s-65536;a+=String.fromCharCode(55296|u>>10,56320|1023&u)}}else a+=String.fromCharCode((31&s)<<6|i)}else a+=String.fromCharCode(s)}return a}function Z(e,t){return e?Q(n(),e,t):""}function J(e,t,r){return function(e,t,n,r){if(!(r>0))return 0;for(var a=n,s=n+r-1,i=0;i<e.length;++i){var o=e.charCodeAt(i);if(o>=55296&&o<=57343&&(o=65536+((1023&o)<<10)|1023&e.charCodeAt(++i)),o<=127){if(n>=s)break;t[n++]=o}else if(o<=2047){if(n+1>=s)break;t[n++]=192|o>>6,t[n++]=128|63&o}else if(o<=65535){if(n+2>=s)break;t[n++]=224|o>>12,t[n++]=128|o>>6&63,t[n++]=128|63&o}else{if(n+3>=s)break;t[n++]=240|o>>18,t[n++]=128|o>>12&63,t[n++]=128|o>>6&63,t[n++]=128|63&o}}return t[n]=0,n-a}(e,n(),t,r)}function ee(e){L=e,d.HEAP8=P=new Int8Array(e),d.HEAP16=new Int16Array(e),d.HEAP32=B=new Int32Array(e),d.HEAPU8=$=new Uint8Array(e),d.HEAPU16=new Uint16Array(e),d.HEAPU32=z=new Uint32Array(e),d.HEAPF32=new Float32Array(e),d.HEAPF64=U=new Float64Array(e)}var te,ne=d.INITIAL_MEMORY||16777216;if(w)M=d.wasmMemory,L=d.buffer;else if(d.wasmMemory)M=d.wasmMemory;else if(!((M=new WebAssembly.Memory({initial:ne/65536,maximum:32768,shared:!0})).buffer instanceof SharedArrayBuffer))throw F("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),b&&console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"),Error("bad memory");M&&(L=M.buffer),ne=L.byteLength,ee(L);var re=[],ae=[],se=[],ie=[];function oe(){w||ye(ae)}w||ae.push({func:function(){Qe()}});var ue=0,le=null;function ce(e){d.onAbort&&d.onAbort(e),w&&console.error("Pthread aborting at "+(new Error).stack),F(e+=""),W=!0,e="abort("+e+"). Build with -s ASSERTIONS=1 for more info.";var t=new WebAssembly.RuntimeError(e);throw h(t),t}function he(e,t){return String.prototype.startsWith?e.startsWith(t):0===e.indexOf(t)}function pe(e){return he(e,"data:application/octet-stream;base64,")}function de(e){return he(e,"file://")}d.preloadedImages={},d.preloadedAudios={};var fe="tfjs-backend-wasm-threaded-simd.wasm";function me(e){try{if(e==fe&&R)return new Uint8Array(R);if(I)return I(e);throw"both async and sync fetching of the wasm failed"}catch(e){ce(e)}}pe(fe)||(fe=C(fe));var ge={10664:function(){throw"Canceled!"},10682:function(e,t){setTimeout((function(){ot(e,t)}),0)}};function ye(e){for(;e.length>0;){var t=e.shift();if("function"!=typeof t){var n=t.func;"number"==typeof n?void 0===t.arg?te.get(n)():te.get(n)(t.arg):n(void 0===t.arg?null:t.arg)}else t(d)}}var be=71,ve=10,xe=28,we=16;function ke(e,n){if(e<=0||e>t().length||!0&e||n<0)return-28;if(0==n)return 0;n>=2147483647&&(n=1/0);var r=Atomics.load(i(),wt>>2),a=0;if(r==e&&Atomics.compareExchange(i(),wt>>2,r,0)==r&&(a=1,--n<=0))return 1;var s=Atomics.notify(i(),e>>2,n);if(s>=0)return s+a;throw"Atomics.notify returned an unexpected value "+s}function Se(e){if(w)throw"Internal Error! cleanupThread() can only ever be called from main application thread!";if(!e)throw"Internal Error! Null pthread_ptr in cleanupThread!";var t=Ae.pthreads[e];if(t){i()[e+12>>2]=0;var n=t.worker;Ae.returnWorkerToPool(n)}}d._emscripten_futex_wake=ke;var Ie,Ae={unusedWorkers:[],runningWorkers:[],initMainThreadBlock:function(){for(var e=0;e<8;++e)Ae.allocateUnusedWorker()},initRuntime:function(){for(var e=Ze(228),t=0;t<57;++t)u()[e/4+t]=0;i()[e+12>>2]=e;var n=e+152;i()[n>>2]=n;var r=Ze(512);for(t=0;t<128;++t)u()[r/4+t]=0;Atomics.store(u(),e+100>>2,r),Atomics.store(u(),e+40>>2,e),ft(e,!y,1),st(e)},initWorker:function(){},pthreads:{},threadExitHandlers:[],setThreadStatus:function(){},runExitHandlers:function(){for(;Ae.threadExitHandlers.length>0;)Ae.threadExitHandlers.pop()();w&&nt()&&rt()},runExitHandlersAndDeinitThread:function(e,t){Atomics.store(u(),e+56>>2,1),Atomics.store(u(),e+60>>2,0),Ae.runExitHandlers(),Atomics.store(u(),e+4>>2,t),Atomics.store(u(),e+0>>2,1),ke(e+0,2147483647),ft(0,0,0)},threadExit:function(e){var t=nt();t&&(Ae.runExitHandlersAndDeinitThread(t,e),w&&postMessage({cmd:"exit"}))},threadCancel:function(){Ae.runExitHandlersAndDeinitThread(nt(),-1),postMessage({cmd:"cancelDone"})},terminateAllThreads:function(){for(var e in Ae.pthreads)(r=Ae.pthreads[e])&&r.worker&&Ae.returnWorkerToPool(r.worker);Ae.pthreads={};for(var t=0;t<Ae.unusedWorkers.length;++t)(n=Ae.unusedWorkers[t]).terminate();for(Ae.unusedWorkers=[],t=0;t<Ae.runningWorkers.length;++t){var n,r=(n=Ae.runningWorkers[t]).pthread;Ae.freeThreadData(r),n.terminate()}Ae.runningWorkers=[]},freeThreadData:function(e){if(e){if(e.threadInfoStruct){var t=i()[e.threadInfoStruct+100>>2];i()[e.threadInfoStruct+100>>2]=0,Je(t),Je(e.threadInfoStruct)}e.threadInfoStruct=0,e.allocatedOwnStack&&e.stackBase&&Je(e.stackBase),e.stackBase=0,e.worker&&(e.worker.pthread=null)}},returnWorkerToPool:function(e){Ae.runWithoutMainThreadQueuedCalls((function(){delete Ae.pthreads[e.pthread.threadInfoStruct],Ae.unusedWorkers.push(e),Ae.runningWorkers.splice(Ae.runningWorkers.indexOf(e),1),Ae.freeThreadData(e.pthread),e.pthread=void 0}))},runWithoutMainThreadQueuedCalls:function(e){i()[xt>>2]=0;try{e()}finally{i()[xt>>2]=1}},receiveObjectTransfer:function(e){},loadWasmModuleToWorker:function(e,t){e.onmessage=function(n){var r=n.data,a=r.cmd;if(e.pthread&&(Ae.currentProxiedOperationCallerThread=e.pthread.threadInfoStruct),r.targetThread&&r.targetThread!=nt()){var s=Ae.pthreads[r.targetThread];return s?s.worker.postMessage(n.data,r.transferList):console.error('Internal error! Worker sent a message "'+a+'" to target pthread '+r.targetThread+", but that thread no longer exists!"),void(Ae.currentProxiedOperationCallerThread=void 0)}if("processQueuedMainThreadWork"===a)at();else if("spawnThread"===a)Ge(n.data);else if("cleanupThread"===a)Se(r.thread);else if("killThread"===a)!function(e){if(w)throw"Internal Error! killThread() can only ever be called from main application thread!";if(!e)throw"Internal Error! Null pthread_ptr in killThread!";i()[e+12>>2]=0;var t=Ae.pthreads[e];t.worker.terminate(),Ae.freeThreadData(t),Ae.runningWorkers.splice(Ae.runningWorkers.indexOf(t.worker),1),t.worker.pthread=void 0}(r.thread);else if("cancelThread"===a)!function(e){if(w)throw"Internal Error! cancelThread() can only ever be called from main application thread!";if(!e)throw"Internal Error! Null pthread_ptr in cancelThread!";Ae.pthreads[e].worker.postMessage({cmd:"cancel"})}(r.thread);else if("loaded"===a)e.loaded=!0,t&&t(e),e.runPthread&&(e.runPthread(),delete e.runPthread);else if("print"===a)_("Thread "+r.threadId+": "+r.text);else if("printErr"===a)F("Thread "+r.threadId+": "+r.text);else if("alert"===a)alert("Thread "+r.threadId+": "+r.text);else if("exit"===a)e.pthread&&Atomics.load(u(),e.pthread.threadInfoStruct+64>>2)&&Ae.returnWorkerToPool(e);else if("exitProcess"===a)try{!function(e,t){if(!t||!O||0!==e){if(!t&&w)throw postMessage({cmd:"exitProcess",returnCode:e}),new kt(e);O||(Ae.terminateAllThreads(),d.onExit&&d.onExit(e),W=!0),x(e,new kt(e))}}(r.returnCode)}catch(e){if(e instanceof kt)return;throw e}else"cancelDone"===a?Ae.returnWorkerToPool(e):"objectTransfer"===a?Ae.receiveObjectTransfer(n.data):"setimmediate"===n.data.target?e.postMessage(n.data):F("worker sent an unknown command "+a);Ae.currentProxiedOperationCallerThread=void 0},e.onerror=function(e){F("pthread sent an error! "+e.filename+":"+e.lineno+": "+e.message)},b&&(e.on("message",(function(t){e.onmessage({data:t})})),e.on("error",(function(t){e.onerror(t)})),e.on("exit",(function(e){}))),e.postMessage({cmd:"load",urlOrBlob:d.mainScriptUrlOrBlob||o,wasmMemory:M,wasmModule:D})},allocateUnusedWorker:function(){var e=C("tfjs-backend-wasm-threaded-simd.worker.js");Ae.unusedWorkers.push(new Worker(e))},getNewWorker:function(){return 0==Ae.unusedWorkers.length&&(Ae.allocateUnusedWorker(),Ae.loadWasmModuleToWorker(Ae.unusedWorkers[0])),Ae.unusedWorkers.length>0?Ae.unusedWorkers.pop():null},busySpinWait:function(e){for(var t=performance.now()+e;performance.now()<t;);}};function Ee(e,n,r){if(e<=0||e>t().length||!0&e)return-28;if(g){if(Atomics.load(i(),e>>2)!=n)return-6;var a=performance.now(),s=a+r;for(Atomics.exchange(i(),wt>>2,e);;){if((a=performance.now())>s)return Atomics.exchange(i(),wt>>2,0),-73;if(0==Atomics.exchange(i(),wt>>2,0))break;if(at(),Atomics.load(i(),e>>2)!=n)return-6;Atomics.exchange(i(),wt>>2,e)}return 0}var o=Atomics.wait(i(),e>>2,n,r);if("timed-out"===o)return-73;if("not-equal"===o)return-6;if("ok"===o)return 0;throw"Atomics.wait returned an unexpected value "+o}function Ne(e,t){for(var n=arguments.length-2,r=mt(),a=n,s=yt(8*a),i=s>>3,o=0;o<n;o++){var u=arguments[2+o];l()[i+o]=u}var c=lt(e,a,s,t);return gt(r),c}d.establishStackSpace=function(e,t){bt(e,t),gt(e)},d.getNoExitRuntime=function(){return O},d.invokeEntryPoint=function(e,t){return te.get(e)(t)},Ie=b?function(){var e=r.hrtime();return 1e3*e[0]+e[1]/1e6}:w?function(){return performance.now()-d.__performance_now_clock_drift}:"undefined"!=typeof dateNow?dateNow:function(){return performance.now()};var Ce=[],Te=[];function Re(e){try{return M.grow(e-L.byteLength+65535>>>16),ee(M.buffer),1}catch(e){}}var _e={inEventHandler:0,removeAllEventListeners:function(){for(var e=_e.eventHandlers.length-1;e>=0;--e)_e._removeHandler(e);_e.eventHandlers=[],_e.deferredCalls=[]},registerRemoveEventListeners:function(){_e.removeEventListenersRegistered||(_e.removeEventListenersRegistered=!0)},deferredCalls:[],deferCall:function(e,t,n){function r(e,t){if(e.length!=t.length)return!1;for(var n in e)if(e[n]!=t[n])return!1;return!0}for(var a in _e.deferredCalls){var s=_e.deferredCalls[a];if(s.targetFunction==e&&r(s.argsList,n))return}_e.deferredCalls.push({targetFunction:e,precedence:t,argsList:n}),_e.deferredCalls.sort((function(e,t){return e.precedence<t.precedence}))},removeDeferredCalls:function(e){for(var t=0;t<_e.deferredCalls.length;++t)_e.deferredCalls[t].targetFunction==e&&(_e.deferredCalls.splice(t,1),--t)},canPerformEventHandlerRequests:function(){return _e.inEventHandler&&_e.currentEventHandler.allowsDeferredCalls},runDeferredCalls:function(){if(_e.canPerformEventHandlerRequests())for(var e=0;e<_e.deferredCalls.length;++e){var t=_e.deferredCalls[e];_e.deferredCalls.splice(e,1),--e,t.targetFunction.apply(null,t.argsList)}},eventHandlers:[],removeAllHandlersOnTarget:function(e,t){for(var n=0;n<_e.eventHandlers.length;++n)_e.eventHandlers[n].target!=e||t&&t!=_e.eventHandlers[n].eventTypeString||_e._removeHandler(n--)},_removeHandler:function(e){var t=_e.eventHandlers[e];t.target.removeEventListener(t.eventTypeString,t.eventListenerFunc,t.useCapture),_e.eventHandlers.splice(e,1)},registerOrRemoveHandler:function(e){var t=function(t){++_e.inEventHandler,_e.currentEventHandler=e,_e.runDeferredCalls(),e.handlerFunc(t),_e.runDeferredCalls(),--_e.inEventHandler};if(e.callbackfunc)e.eventListenerFunc=t,e.target.addEventListener(e.eventTypeString,t,e.useCapture),_e.eventHandlers.push(e),_e.registerRemoveEventListeners();else for(var n=0;n<_e.eventHandlers.length;++n)_e.eventHandlers[n].target==e.target&&_e.eventHandlers[n].eventTypeString==e.eventTypeString&&_e._removeHandler(n--)},queueEventHandlerOnThread_iiii:function(e,t,n,r,a){var s=mt(),o=yt(12);i()[o>>2]=n,i()[o+4>>2]=r,i()[o+8>>2]=a,ct(0,e,637534208,t,r,o),gt(s)},getTargetThreadForEventCallback:function(e){switch(e){case 1:return 0;case 2:return Ae.currentProxiedOperationCallerThread;default:return e}},getNodeNameForTarget:function(e){return e?e==window?"#window":e==screen?"#screen":e&&e.nodeName?e.nodeName:"":""},fullscreenEnabled:function(){return document.fullscreenEnabled||document.webkitFullscreenEnabled}};function Fe(e,t,n,r){var a,s,o,u=mt(),l=yt(12),c=0;t&&(s=function(e){for(var t=0,n=0;n<e.length;++n){var r=e.charCodeAt(n);r>=55296&&r<=57343&&(r=65536+((1023&r)<<10)|1023&e.charCodeAt(++n)),r<=127?++t:t+=r<=2047?2:r<=65535?3:4}return t}(a=t)+1,o=Ze(s),J(a,o,s),c=o),i()[l>>2]=c,i()[l+4>>2]=n,i()[l+8>>2]=r,ct(0,e,657457152,0,c,l),gt(u)}var Me=[0,"undefined"!=typeof document?document:0,"undefined"!=typeof window?window:0];function De(e){var t;return e=(t=e)>2?Z(t):t,Me[e]||("undefined"!=typeof document?document.querySelector(e):void 0)}function Oe(e){return De(e)}function Le(e,t,n){var r=Oe(e);if(!r)return-4;if(r.canvasSharedPtr&&(i()[r.canvasSharedPtr>>2]=t,i()[r.canvasSharedPtr+4>>2]=n),!r.offscreenCanvas&&r.controlTransferredOffscreen)return r.canvasSharedPtr?(function(e,t,n,r){Fe(e,t=t?Z(t):"",n,r)}(i()[r.canvasSharedPtr+8>>2],e,t,n),1):-4;r.offscreenCanvas&&(r=r.offscreenCanvas);var a=!1;if(r.GLctxObject&&r.GLctxObject.GLctx){var s=r.GLctxObject.GLctx.getParameter(2978);a=0===s[0]&&0===s[1]&&s[2]===r.width&&s[3]===r.height}return r.width=t,r.height=n,a&&r.GLctxObject.GLctx.viewport(0,0,t,n),0}function Pe(e,t,n){return w?Ne(2,1,e,t,n):Le(e,t,n)}var $e,Be={counter:1,buffers:[],programs:[],framebuffers:[],renderbuffers:[],textures:[],uniforms:[],shaders:[],vaos:[],contexts:{},offscreenCanvases:{},timerQueriesEXT:[],programInfos:{},stringCache:{},unpackAlignment:4,recordError:function(e){Be.lastError||(Be.lastError=e)},getNewId:function(e){for(var t=Be.counter++,n=e.length;n<t;n++)e[n]=null;return t},getSource:function(e,t,n,r){for(var a="",s=0;s<t;++s){var o=r?i()[r+4*s>>2]:-1;a+=Z(i()[n+4*s>>2],o<0?void 0:o)}return a},createContext:function(e,t){var n=e.getContext("webgl",t);return n?Be.registerContext(n,t):0},registerContext:function(e,t){var n=Ze(8);i()[n+4>>2]=nt();var r={handle:n,attributes:t,version:t.majorVersion,GLctx:e};return e.canvas&&(e.canvas.GLctxObject=r),Be.contexts[n]=r,(void 0===t.enableExtensionsByDefault||t.enableExtensionsByDefault)&&Be.initExtensions(r),n},makeContextCurrent:function(e){return Be.currentContext=Be.contexts[e],d.ctx=$e=Be.currentContext&&Be.currentContext.GLctx,!(e&&!$e)},getContext:function(e){return Be.contexts[e]},deleteContext:function(e){Be.currentContext===Be.contexts[e]&&(Be.currentContext=null),"object"==typeof _e&&_e.removeAllHandlersOnTarget(Be.contexts[e].GLctx.canvas),Be.contexts[e]&&Be.contexts[e].GLctx.canvas&&(Be.contexts[e].GLctx.canvas.GLctxObject=void 0),Je(Be.contexts[e].handle),Be.contexts[e]=null},initExtensions:function(e){if(e||(e=Be.currentContext),!e.initExtensionsDone){e.initExtensionsDone=!0;var t,n=e.GLctx;!function(e){var t=e.getExtension("ANGLE_instanced_arrays");t&&(e.vertexAttribDivisor=function(e,n){t.vertexAttribDivisorANGLE(e,n)},e.drawArraysInstanced=function(e,n,r,a){t.drawArraysInstancedANGLE(e,n,r,a)},e.drawElementsInstanced=function(e,n,r,a,s){t.drawElementsInstancedANGLE(e,n,r,a,s)})}(n),function(e){var t=e.getExtension("OES_vertex_array_object");t&&(e.createVertexArray=function(){return t.createVertexArrayOES()},e.deleteVertexArray=function(e){t.deleteVertexArrayOES(e)},e.bindVertexArray=function(e){t.bindVertexArrayOES(e)},e.isVertexArray=function(e){return t.isVertexArrayOES(e)})}(n),function(e){var t=e.getExtension("WEBGL_draw_buffers");t&&(e.drawBuffers=function(e,n){t.drawBuffersWEBGL(e,n)})}(n),n.disjointTimerQueryExt=n.getExtension("EXT_disjoint_timer_query"),(t=n).multiDrawWebgl=t.getExtension("WEBGL_multi_draw"),(n.getSupportedExtensions()||[]).forEach((function(e){e.indexOf("lose_context")<0&&e.indexOf("debug")<0&&n.getExtension(e)}))}},populateUniformTable:function(e){for(var t=Be.programs[e],n=Be.programInfos[e]={uniforms:{},maxUniformLength:0,maxAttributeLength:-1,maxUniformBlockNameLength:-1},r=n.uniforms,a=$e.getProgramParameter(t,35718),s=0;s<a;++s){var i=$e.getActiveUniform(t,s),o=i.name;n.maxUniformLength=Math.max(n.maxUniformLength,o.length+1),"]"==o.slice(-1)&&(o=o.slice(0,o.lastIndexOf("[")));var u=$e.getUniformLocation(t,o);if(u){var l=Be.getNewId(Be.uniforms);r[o]=[i.size,l],Be.uniforms[l]=u;for(var c=1;c<i.size;++c){var h=o+"["+c+"]";u=$e.getUniformLocation(t,h),l=Be.getNewId(Be.uniforms),Be.uniforms[l]=u}}}}},ze=["default","low-power","high-performance"],Ue={mappings:{},buffers:[null,[],[]],printChar:function(e,t){var n=Ue.buffers[e];0===t||10===t?((1===e?_:F)(Q(n,0)),n.length=0):n.push(t)},varargs:void 0,get:function(){return Ue.varargs+=4,i()[Ue.varargs-4>>2]},getStr:function(e){return Z(e)},get64:function(e,t){return e}};function We(e){return w?Ne(3,1,e):0}function Ve(e,t,n,r,a){if(w)return Ne(4,1,e,t,n,r,a)}function je(e,t,r,a){if(w)return Ne(5,1,e,t,r,a);for(var s=0,o=0;o<r;o++){for(var u=i()[t+8*o>>2],l=i()[t+(8*o+4)>>2],c=0;c<l;c++)Ue.printChar(e,n()[u+c]);s+=l}return i()[a>>2]=s,0}function Ge(e){if(w)throw"Internal Error! spawnThread() can only ever be called from main application thread!";var t=Ae.getNewWorker();if(void 0!==t.pthread)throw"Internal error!";if(!e.pthread_ptr)throw"Internal error, no pthread ptr!";Ae.runningWorkers.push(t);for(var n=Ze(512),r=0;r<128;++r)i()[n+4*r>>2]=0;var a=e.stackBase+e.stackSize,s=Ae.pthreads[e.pthread_ptr]={worker:t,stackBase:e.stackBase,stackSize:e.stackSize,allocatedOwnStack:e.allocatedOwnStack,threadInfoStruct:e.pthread_ptr},o=s.threadInfoStruct>>2;Atomics.store(u(),o+16,e.detached),Atomics.store(u(),o+25,n),Atomics.store(u(),o+10,s.threadInfoStruct),Atomics.store(u(),o+20,e.stackSize),Atomics.store(u(),o+19,a),Atomics.store(u(),o+26,e.stackSize),Atomics.store(u(),o+28,a),Atomics.store(u(),o+29,e.detached);var l=tt()+40;Atomics.store(u(),o+43,l),t.pthread=s;var c={cmd:"run",start_routine:e.startRoutine,arg:e.arg,threadInfoStruct:e.pthread_ptr,stackBase:e.stackBase,stackSize:e.stackSize};t.runPthread=function(){c.time=performance.now(),t.postMessage(c,e.transferList)},t.loaded&&(t.runPthread(),delete t.runPthread)}function He(){if(w){var e=nt();if(e&&!Atomics.load(u(),e+56>>2)&&2==Atomics.load(u(),e+0>>2))throw"Canceled!"}}function qe(){b||y||function e(t){e.shown||(e.shown={}),e.shown[t]||(e.shown[t]=1,F(t))}("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")}function Ke(e){if(w)return Ne(6,1,e);switch(e){case 30:return 16384;case 85:return 131072;case 132:case 133:case 12:case 137:case 138:case 15:case 235:case 16:case 17:case 18:case 19:case 20:case 149:case 13:case 10:case 236:case 153:case 9:case 21:case 22:case 159:case 154:case 14:case 77:case 78:case 139:case 82:case 68:case 67:case 164:case 11:case 29:case 47:case 48:case 95:case 52:case 51:case 46:return 200809;case 27:case 246:case 127:case 128:case 23:case 24:case 160:case 161:case 181:case 182:case 242:case 183:case 184:case 243:case 244:case 245:case 165:case 178:case 179:case 49:case 50:case 168:case 169:case 175:case 170:case 171:case 172:case 97:case 76:case 32:case 173:case 35:case 80:case 81:case 79:return-1;case 176:case 177:case 7:case 155:case 8:case 157:case 125:case 126:case 92:case 93:case 129:case 130:case 131:case 94:case 91:return 1;case 74:case 60:case 69:case 70:case 4:return 1024;case 31:case 42:case 72:return 32;case 87:case 26:case 33:return 2147483647;case 34:case 1:return 47839;case 38:case 36:return 99;case 43:case 37:return 2048;case 0:return 2097152;case 3:return 65536;case 28:return 32768;case 44:return 32767;case 75:return 16384;case 39:return 1e3;case 89:return 700;case 71:return 256;case 40:return 255;case 2:return 100;case 180:return 64;case 25:return 20;case 5:return 16;case 6:return 6;case 73:return 4;case 84:return"object"==typeof navigator&&navigator.hardwareConcurrency||1}var t;return t=28,i()[et()>>2]=t,-1}w||Ae.initMainThreadBlock();var Ye=[null,function(e,t){if(w)return Ne(1,1,e,t)},Pe,We,Ve,je,Ke],Xe={e:function(e,t,n,r){ce("Assertion failed: "+Z(e)+", at: "+[t?Z(t):"unknown filename",n,r?Z(r):"unknown function"])},r:function(e,t){_main(e,t)},x:function(e,t){if(e==t)postMessage({cmd:"processQueuedMainThreadWork"});else if(w)postMessage({targetThread:e,cmd:"processThreadQueue"});else{var n=Ae.pthreads[e],r=n&&n.worker;if(!r)return;r.postMessage({cmd:"processThreadQueue"})}return 1},b:function(){ce()},y:function(e,t,r){var a=function(e,t){var r;for(Te.length=0,t>>=2;r=n()[e++];){var a=r<105;a&&1&t&&t++,Te.push(a?l()[t++>>1]:i()[t]),++t}return Te}(t,r);return ge[e].apply(null,a)},j:function(e,t){},d:Ee,c:ke,f:Ie,p:function(e,t,r){n().copyWithin(e,t,t+r)},A:function(){return b?K().cpus().length:navigator.hardwareConcurrency},u:function(e,t,n){Ce.length=t;for(var r=n>>3,a=0;a<t;a++)Ce[a]=l()[r+a];return(e<0?ge[-e-1]:Ye[e]).apply(null,Ce)},q:function(e){var t,r,a=n().length;if(e<=a)return!1;if(e>2147483648)return!1;for(var s=1;s<=4;s*=2){var i=a*(1+.2/s);if(i=Math.min(i,e+100663296),Re(Math.min(2147483648,((t=Math.max(e,i))%(r=65536)>0&&(t+=r-t%r),t))))return!0}return!1},v:function(e,t,n){return Oe(e)?Le(e,t,n):Pe(e,t,n)},i:function(e){},s:function(e,t){},w:function(e,t){return n=e,r=t>>2,a=i()[r+6],s={alpha:!!i()[r+0],depth:!!i()[r+1],stencil:!!i()[r+2],antialias:!!i()[r+3],premultipliedAlpha:!!i()[r+4],preserveDrawingBuffer:!!i()[r+5],powerPreference:ze[a],failIfMajorPerformanceCaveat:!!i()[r+7],majorVersion:i()[r+8],minorVersion:i()[r+9],enableExtensionsByDefault:i()[r+10],explicitSwapControl:i()[r+11],proxyContextToMainThread:i()[r+12],renderViaOffscreenBackBuffer:i()[r+13]},(o=Oe(n))?s.explicitSwapControl?0:Be.createContext(o,s):0;var n,r,a,s,o},l:We,n:Ve,g:je,o:function(){Ae.initRuntime()},a:M||d.wasmMemory,z:function(e){var t=Ae.threadExitHandlers.pop();e&&t()},k:function(e,t){Ae.threadExitHandlers.push((function(){te.get(e)(t)}))},h:function(e,t,n,r){if("undefined"==typeof SharedArrayBuffer)return F("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;if(!e)return F("pthread_create called with a null thread pointer!"),28;var a=[];if(w&&0===a.length)return ut(687865856,e,t,n,r);var s=0,o=0,l=0;t&&-1!=t?(s=i()[t>>2],s+=81920,o=i()[t+8>>2],l=0!==i()[t+12>>2]):s=2097152;var c=0==o;c?o=vt(16,s):V((o-=s)>0);for(var h=Ze(228),p=0;p<57;++p)u()[(h>>2)+p]=0;i()[e>>2]=h,i()[h+12>>2]=h;var d=h+152;i()[d>>2]=d;var f={stackBase:o,stackSize:s,allocatedOwnStack:c,detached:l,startRoutine:n,pthread_ptr:h,arg:r,transferList:a};return w?(f.cmd="spawnThread",postMessage(f,a)):Ge(f),0},m:function(e,t){return function(e,t,n){if(!e)return F("pthread_join attempted on a null thread pointer!"),be;if(w&&nt()==e)return F("PThread "+e+" is attempting to join to itself!"),we;if(!w&&it()==e)return F("Main thread "+e+" is attempting to join to itself!"),we;if(i()[e+12>>2]!==e)return F("pthread_join attempted on thread "+e+", which does not point to a valid thread, or does not exist anymore!"),be;if(Atomics.load(u(),e+64>>2))return F("Attempted to join thread "+e+", which was already detached!"),xe;for(n&&qe();;){var r=Atomics.load(u(),e+0>>2);if(1==r){var a=Atomics.load(u(),e+4>>2);return t&&(i()[t>>2]=a),Atomics.store(u(),e+64>>2,1),w?postMessage({cmd:"cleanupThread",thread:e}):Se(e),0}if(!n)return ve;He(),w||at(),Ee(e+0,r,w?100:1)}}(e,t,!0)},t:Ke};!function(){var e={a:Xe};function t(e,t){var n=e.exports;if(d.asm=n,te=d.asm.I,D=t,!w){var r=Ae.unusedWorkers.length;Ae.unusedWorkers.forEach((function(e){Ae.loadWasmModuleToWorker(e,(function(){--r||function(e){if(ue--,d.monitorRunDependencies&&d.monitorRunDependencies(ue),0==ue&&le){var t=le;le=null,t()}}()}))}))}}function n(e){t(e.instance,e.module)}function r(t){return function(){if(!R&&(g||y)){if("function"==typeof fetch&&!de(fe))return fetch(fe,{credentials:"same-origin"}).then((function(e){if(!e.ok)throw"failed to load wasm binary file at '"+fe+"'";return e.arrayBuffer()})).catch((function(){return me(fe)}));if(S)return new Promise((function(e,t){S(fe,(function(t){e(new Uint8Array(t))}),t)}))}return Promise.resolve().then((function(){return me(fe)}))}().then((function(t){return WebAssembly.instantiate(t,e)})).then(t,(function(e){F("failed to asynchronously prepare wasm: "+e),ce(e)}))}if(w||(V(!w,"addRunDependency cannot be used in a pthread worker"),ue++,d.monitorRunDependencies&&d.monitorRunDependencies(ue)),d.instantiateWasm)try{return d.instantiateWasm(e,t)}catch(e){return F("Module.instantiateWasm callback failed with error: "+e),!1}(R||"function"!=typeof WebAssembly.instantiateStreaming||pe(fe)||de(fe)||"function"!=typeof fetch?r(n):fetch(fe,{credentials:"same-origin"}).then((function(t){return WebAssembly.instantiateStreaming(t,e).then(n,(function(e){return F("wasm streaming compile failed: "+e),F("falling back to ArrayBuffer instantiation"),r(n)}))}))).catch(h)}();var Qe=d.___wasm_call_ctors=function(){return(Qe=d.___wasm_call_ctors=d.asm.B).apply(null,arguments)};d._init=function(){return(d._init=d.asm.C).apply(null,arguments)},d._init_with_threads_count=function(){return(d._init_with_threads_count=d.asm.D).apply(null,arguments)},d._get_threads_count=function(){return(d._get_threads_count=d.asm.E).apply(null,arguments)},d._register_tensor=function(){return(d._register_tensor=d.asm.F).apply(null,arguments)},d._dispose_data=function(){return(d._dispose_data=d.asm.G).apply(null,arguments)},d._dispose=function(){return(d._dispose=d.asm.H).apply(null,arguments)},d._Abs=function(){return(d._Abs=d.asm.J).apply(null,arguments)},d._Add=function(){return(d._Add=d.asm.K).apply(null,arguments)},d._AddN=function(){return(d._AddN=d.asm.L).apply(null,arguments)},d._All=function(){return(d._All=d.asm.M).apply(null,arguments)},d._Any=function(){return(d._Any=d.asm.N).apply(null,arguments)},d._ArgMax=function(){return(d._ArgMax=d.asm.O).apply(null,arguments)},d._AvgPool=function(){return(d._AvgPool=d.asm.P).apply(null,arguments)},d._BatchMatMul=function(){return(d._BatchMatMul=d.asm.Q).apply(null,arguments)},d._Ceil=function(){return(d._Ceil=d.asm.R).apply(null,arguments)},d._ClipByValue=function(){return(d._ClipByValue=d.asm.S).apply(null,arguments)},d._Conv2D=function(){return(d._Conv2D=d.asm.T).apply(null,arguments)},d._Conv2DBackpropInput=function(){return(d._Conv2DBackpropInput=d.asm.U).apply(null,arguments)},d._Cos=function(){return(d._Cos=d.asm.V).apply(null,arguments)},d._Cosh=function(){return(d._Cosh=d.asm.W).apply(null,arguments)},d._CropAndResize=function(){return(d._CropAndResize=d.asm.X).apply(null,arguments)},d._Cumsum=function(){return(d._Cumsum=d.asm.Y).apply(null,arguments)},d._DepthToSpace=function(){return(d._DepthToSpace=d.asm.Z).apply(null,arguments)},d._DepthwiseConv2dNative=function(){return(d._DepthwiseConv2dNative=d.asm._).apply(null,arguments)},d._Elu=function(){return(d._Elu=d.asm.$).apply(null,arguments)},d._Equal=function(){return(d._Equal=d.asm.aa).apply(null,arguments)},d._Exp=function(){return(d._Exp=d.asm.ba).apply(null,arguments)},d._FlipLeftRight=function(){return(d._FlipLeftRight=d.asm.ca).apply(null,arguments)},d._Floor=function(){return(d._Floor=d.asm.da).apply(null,arguments)},d._FloorDiv=function(){return(d._FloorDiv=d.asm.ea).apply(null,arguments)},d._FusedBatchNorm=function(){return(d._FusedBatchNorm=d.asm.fa).apply(null,arguments)},d._FusedConv2D=function(){return(d._FusedConv2D=d.asm.ga).apply(null,arguments)},d._FusedDepthwiseConv2D=function(){return(d._FusedDepthwiseConv2D=d.asm.ha).apply(null,arguments)},d._Gather=function(){return(d._Gather=d.asm.ia).apply(null,arguments)},d._GatherNd=function(){return(d._GatherNd=d.asm.ja).apply(null,arguments)},d._Greater=function(){return(d._Greater=d.asm.ka).apply(null,arguments)},d._GreaterEqual=function(){return(d._GreaterEqual=d.asm.la).apply(null,arguments)},d._LeakyRelu=function(){return(d._LeakyRelu=d.asm.ma).apply(null,arguments)},d._Less=function(){return(d._Less=d.asm.na).apply(null,arguments)},d._LessEqual=function(){return(d._LessEqual=d.asm.oa).apply(null,arguments)},d._Log=function(){return(d._Log=d.asm.pa).apply(null,arguments)},d._LogicalAnd=function(){return(d._LogicalAnd=d.asm.qa).apply(null,arguments)},d._Max=function(){return(d._Max=d.asm.ra).apply(null,arguments)},d._MaxPool=function(){return(d._MaxPool=d.asm.sa).apply(null,arguments)},d._Maximum=function(){return(d._Maximum=d.asm.ta).apply(null,arguments)},d._Mean=function(){return(d._Mean=d.asm.ua).apply(null,arguments)},d._Min=function(){return(d._Min=d.asm.va).apply(null,arguments)},d._Minimum=function(){return(d._Minimum=d.asm.wa).apply(null,arguments)},d._MirrorPad=function(){return(d._MirrorPad=d.asm.xa).apply(null,arguments)},d._Multiply=function(){return(d._Multiply=d.asm.ya).apply(null,arguments)},d._Neg=function(){return(d._Neg=d.asm.za).apply(null,arguments)},d._NonMaxSuppressionV3=function(){return(d._NonMaxSuppressionV3=d.asm.Aa).apply(null,arguments)},d._NonMaxSuppressionV4=function(){return(d._NonMaxSuppressionV4=d.asm.Ba).apply(null,arguments)},d._NonMaxSuppressionV5=function(){return(d._NonMaxSuppressionV5=d.asm.Ca).apply(null,arguments)},d._NotEqual=function(){return(d._NotEqual=d.asm.Da).apply(null,arguments)},d._OneHot=function(){return(d._OneHot=d.asm.Ea).apply(null,arguments)},d._PadV2=function(){return(d._PadV2=d.asm.Fa).apply(null,arguments)},d._Pow=function(){return(d._Pow=d.asm.Ga).apply(null,arguments)},d._Prelu=function(){return(d._Prelu=d.asm.Ha).apply(null,arguments)},d._Prod=function(){return(d._Prod=d.asm.Ia).apply(null,arguments)},d._RealDiv=function(){return(d._RealDiv=d.asm.Ja).apply(null,arguments)},d._Relu=function(){return(d._Relu=d.asm.Ka).apply(null,arguments)},d._Relu6=function(){return(d._Relu6=d.asm.La).apply(null,arguments)},d._ResizeBilinear=function(){return(d._ResizeBilinear=d.asm.Ma).apply(null,arguments)},d._Reverse=function(){return(d._Reverse=d.asm.Na).apply(null,arguments)},d._RotateWithOffset=function(){return(d._RotateWithOffset=d.asm.Oa).apply(null,arguments)},d._Round=function(){return(d._Round=d.asm.Pa).apply(null,arguments)},d._Rsqrt=function(){return(d._Rsqrt=d.asm.Qa).apply(null,arguments)},d._ScatterNd=function(){return(d._ScatterNd=d.asm.Ra).apply(null,arguments)},d._SelectV2=function(){return(d._SelectV2=d.asm.Sa).apply(null,arguments)},d._Sigmoid=function(){return(d._Sigmoid=d.asm.Ta).apply(null,arguments)},d._Sin=function(){return(d._Sin=d.asm.Ua).apply(null,arguments)},d._Softmax=function(){return(d._Softmax=d.asm.Va).apply(null,arguments)},d._SparseFillEmptyRows=function(){return(d._SparseFillEmptyRows=d.asm.Wa).apply(null,arguments)},d._SparseReshape=function(){return(d._SparseReshape=d.asm.Xa).apply(null,arguments)},d._SparseSegmentReduction=function(){return(d._SparseSegmentReduction=d.asm.Ya).apply(null,arguments)},d._Sqrt=function(){return(d._Sqrt=d.asm.Za).apply(null,arguments)},d._Square=function(){return(d._Square=d.asm._a).apply(null,arguments)},d._SquaredDifference=function(){return(d._SquaredDifference=d.asm.$a).apply(null,arguments)},d._Step=function(){return(d._Step=d.asm.ab).apply(null,arguments)},d._StridedSlice=function(){return(d._StridedSlice=d.asm.bb).apply(null,arguments)},d._Sub=function(){return(d._Sub=d.asm.cb).apply(null,arguments)},d._Sum=function(){return(d._Sum=d.asm.db).apply(null,arguments)},d._Tan=function(){return(d._Tan=d.asm.eb).apply(null,arguments)},d._Tanh=function(){return(d._Tanh=d.asm.fb).apply(null,arguments)},d._Tile=function(){return(d._Tile=d.asm.gb).apply(null,arguments)},d._TopK=function(){return(d._TopK=d.asm.hb).apply(null,arguments)},d._Transform=function(){return(d._Transform=d.asm.ib).apply(null,arguments)},d._Transpose=function(){return(d._Transpose=d.asm.jb).apply(null,arguments)},d.__FusedMatMul=function(){return(d.__FusedMatMul=d.asm.kb).apply(null,arguments)};var Ze=d._malloc=function(){return(Ze=d._malloc=d.asm.lb).apply(null,arguments)},Je=d._free=function(){return(Je=d._free=d.asm.mb).apply(null,arguments)},et=d.___errno_location=function(){return(et=d.___errno_location=d.asm.nb).apply(null,arguments)},tt=d._emscripten_get_global_libc=function(){return(tt=d._emscripten_get_global_libc=d.asm.ob).apply(null,arguments)},nt=d._pthread_self=function(){return(nt=d._pthread_self=d.asm.pb).apply(null,arguments)},rt=d.___pthread_tsd_run_dtors=function(){return(rt=d.___pthread_tsd_run_dtors=d.asm.qb).apply(null,arguments)},at=d._emscripten_main_thread_process_queued_calls=function(){return(at=d._emscripten_main_thread_process_queued_calls=d.asm.rb).apply(null,arguments)};d._emscripten_current_thread_process_queued_calls=function(){return(d._emscripten_current_thread_process_queued_calls=d.asm.sb).apply(null,arguments)};var st=d._emscripten_register_main_browser_thread_id=function(){return(st=d._emscripten_register_main_browser_thread_id=d.asm.tb).apply(null,arguments)},it=d._emscripten_main_browser_thread_id=function(){return(it=d._emscripten_main_browser_thread_id=d.asm.ub).apply(null,arguments)},ot=d.__emscripten_do_dispatch_to_thread=function(){return(ot=d.__emscripten_do_dispatch_to_thread=d.asm.vb).apply(null,arguments)},ut=d._emscripten_sync_run_in_main_thread_4=function(){return(ut=d._emscripten_sync_run_in_main_thread_4=d.asm.wb).apply(null,arguments)},lt=d._emscripten_run_in_main_runtime_thread_js=function(){return(lt=d._emscripten_run_in_main_runtime_thread_js=d.asm.xb).apply(null,arguments)},ct=d.__emscripten_call_on_thread=function(){return(ct=d.__emscripten_call_on_thread=d.asm.yb).apply(null,arguments)};d._emscripten_tls_init=function(){return(d._emscripten_tls_init=d.asm.zb).apply(null,arguments)};var ht,pt,dt,ft=d.__emscripten_thread_init=function(){return(ft=d.__emscripten_thread_init=d.asm.Ab).apply(null,arguments)},mt=d.stackSave=function(){return(mt=d.stackSave=d.asm.Bb).apply(null,arguments)},gt=d.stackRestore=function(){return(gt=d.stackRestore=d.asm.Cb).apply(null,arguments)},yt=d.stackAlloc=function(){return(yt=d.stackAlloc=d.asm.Db).apply(null,arguments)},bt=d._emscripten_stack_set_limits=function(){return(bt=d._emscripten_stack_set_limits=d.asm.Eb).apply(null,arguments)},vt=d._memalign=function(){return(vt=d._memalign=d.asm.Fb).apply(null,arguments)},xt=d.__emscripten_allow_main_runtime_queued_calls=10656,wt=d.__emscripten_main_thread_futex=12292;function kt(e){this.name="ExitStatus",this.message="Program terminated with exit("+e+")",this.status=e}function St(e){if(!(ue>0)){if(w)return c(d),oe(),void postMessage({cmd:"loaded"});!function(){if(!w){if(d.preRun)for("function"==typeof d.preRun&&(d.preRun=[d.preRun]);d.preRun.length;)e=d.preRun.shift(),re.unshift(e);var e;ye(re)}}(),ue>0||(d.setStatus?(d.setStatus("Running..."),setTimeout((function(){setTimeout((function(){d.setStatus("")}),1),t()}),1)):t())}function t(){ht||(ht=!0,d.calledRun=!0,W||(oe(),w||ye(se),c(d),d.onRuntimeInitialized&&d.onRuntimeInitialized(),function(){if(!w){if(d.postRun)for("function"==typeof d.postRun&&(d.postRun=[d.postRun]);d.postRun.length;)e=d.postRun.shift(),ie.unshift(e);var e;ye(ie)}}()))}}if(d.cwrap=function(e,t,n,r){var a=(n=n||[]).every((function(e){return"number"===e}));return"string"!==t&&a&&!r?Y(e):function(){return X(e,t,n,arguments)}},d.PThread=Ae,d.PThread=Ae,d.wasmMemory=M,d.ExitStatus=kt,le=function e(){ht||St(),ht||(le=e)},d.run=St,d.preInit)for("function"==typeof d.preInit&&(d.preInit=[d.preInit]);d.preInit.length>0;)d.preInit.pop()();if(w&&(O=!1,Ae.initWorker()),St(),p&&(pt={uncaughtException:r.listeners("uncaughtException").filter((function(e){return!p.uncaughtException.indexOf(e)>-1})),unhandledRejection:r.listeners("unhandledRejection").filter((function(e){return!p.unhandledRejection.indexOf(e)>-1}))}),"undefined"!=typeof WasmBackendModule)dt=WasmBackendModule;else{if(void 0===e)throw new Error("Could not find wasm module in post.js");dt=e}if(pt){var It=dt._dispose;dt._dispose=function(){It(),pt.uncaughtException.forEach((function(e){r.removeListener("uncaughtException",e)})),pt.unhandledRejection.forEach((function(e){r.removeListener("unhandledRejection",e)}))}}return e.ready});"object"==typeof t&&"object"==typeof i?i.exports=u:"function"==typeof define&&n(66)?define([],(function(){return u})):"object"==typeof t&&(t.WasmBackendModuleThreadedSimd=u)}}),X=x({"node_modules/.pnpm/@tensorflow+tfjs-backend-wasm@3.13.0_@tensorflow+tfjs-core@3.13.0/node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js"(t,s){var i,o=(i=(i="undefined"!=typeof document&&document.currentScript?document.currentScript.src:void 0)||e,function(e){var t,n,s,o=void 0!==(e=e||{})?e:{};o.ready=new Promise((function(e,r){t=e,n=r})),void 0!==r&&r.listeners&&(s={uncaughtException:r.listeners("uncaughtException"),unhandledRejection:r.listeners("unhandledRejection")});var u,l,c,h,p,d={};for(u in o)o.hasOwnProperty(u)&&(d[u]=o[u]);l="object"==typeof window,c="function"==typeof importScripts,h="object"==typeof r&&"object"==typeof r.versions&&"string"==typeof r.versions.node,p=!l&&!h&&!c;var f,m,g,y,b,v="";h?(v=c?G().dirname(v)+"/":a+"/",f=function(e,t){return y||(y=j()),b||(b=G()),e=b.normalize(e),y.readFileSync(e,t?null:"utf8")},g=function(e){var t=f(e,!0);return t.buffer||(t=new Uint8Array(t)),A(t.buffer),t},r.argv.length>1&&r.argv[1].replace(/\\/g,"/"),r.argv.slice(2),r.on("uncaughtException",(function(e){if(!(e instanceof ce))throw e})),r.on("unhandledRejection",V),o.inspect=function(){return"[Emscripten Module object]"}):p?("undefined"!=typeof read&&(f=function(e){return read(e)}),g=function(e){var t;return"function"==typeof readbuffer?new Uint8Array(readbuffer(e)):(A("object"==typeof(t=read(e,"binary"))),t)},"undefined"!=typeof scriptArgs&&scriptArgs,"undefined"!=typeof print&&("undefined"==typeof console&&(console={}),console.log=print,console.warn=console.error="undefined"!=typeof printErr?printErr:print)):(l||c)&&(c?v=self.location.href:"undefined"!=typeof document&&document.currentScript&&(v=document.currentScript.src),i&&(v=i),v=0!==v.indexOf("blob:")?v.substr(0,v.lastIndexOf("/")+1):"",f=function(e){var t=new XMLHttpRequest;return t.open("GET",e,!1),t.send(null),t.responseText},c&&(g=function(e){var t=new XMLHttpRequest;return t.open("GET",e,!1),t.responseType="arraybuffer",t.send(null),new Uint8Array(t.response)}),m=function(e,t,n){var r=new XMLHttpRequest;r.open("GET",e,!0),r.responseType="arraybuffer",r.onload=function(){200==r.status||0==r.status&&r.response?t(r.response):n()},r.onerror=n,r.send(null)});var x,w,k=o.print||console.log.bind(console),S=o.printErr||console.warn.bind(console);for(u in d)d.hasOwnProperty(u)&&(o[u]=d[u]);d=null,o.wasmBinary&&(x=o.wasmBinary),"object"!=typeof WebAssembly&&V("no native wasm support detected");var I=!1;function A(e,t){e||V("Assertion failed: "+t)}function E(e){var t=o["_"+e];return A(t,"Cannot call unknown function "+e+", make sure it is exported"),t}function N(e,t,n,r,a){var s={string:function(e){var t=0;if(null!=e&&0!==e){var n=1+(e.length<<2);!function(e,t,n){!function(e,t,n,r){if(!(r>0))return 0;for(var a=n+r-1,s=0;s<e.length;++s){var i=e.charCodeAt(s);if(i>=55296&&i<=57343){var o=e.charCodeAt(++s);i=65536+((1023&i)<<10)|1023&o}if(i<=127){if(n>=a)break;t[n++]=i}else if(i<=2047){if(n+1>=a)break;t[n++]=192|i>>6,t[n++]=128|63&i}else if(i<=65535){if(n+2>=a)break;t[n++]=224|i>>12,t[n++]=128|i>>6&63,t[n++]=128|63&i}else{if(n+3>=a)break;t[n++]=240|i>>18,t[n++]=128|i>>12&63,t[n++]=128|i>>6&63,t[n++]=128|63&i}}t[n]=0}(e,R,t,n)}(e,t=le(n),n)}return t},array:function(e){var t,n,r=le(e.length);return t=e,n=r,T.set(t,n),r}},i=E(e),o=[],u=0;if(r)for(var l=0;l<r.length;l++){var c=s[n[l]];c?(0===u&&(u=oe()),o[l]=c(r[l])):o[l]=r[l]}var h,p=i.apply(null,o);return h=p,p="string"===t?O(h):"boolean"===t?Boolean(h):h,0!==u&&ue(u),p}var C,T,R,_,F,M="undefined"!=typeof TextDecoder?new TextDecoder("utf8"):void 0;function D(e,t,n){for(var r=t+n,a=t;e[a]&&!(a>=r);)++a;if(a-t>16&&e.subarray&&M)return M.decode(e.subarray(t,a));for(var s="";t<a;){var i=e[t++];if(128&i){var o=63&e[t++];if(192!=(224&i)){var u=63&e[t++];if((i=224==(240&i)?(15&i)<<12|o<<6|u:(7&i)<<18|o<<12|u<<6|63&e[t++])<65536)s+=String.fromCharCode(i);else{var l=i-65536;s+=String.fromCharCode(55296|l>>10,56320|1023&l)}}else s+=String.fromCharCode((31&i)<<6|o)}else s+=String.fromCharCode(i)}return s}function O(e,t){return e?D(R,e,t):""}function L(e){C=e,o.HEAP8=T=new Int8Array(e),o.HEAP16=new Int16Array(e),o.HEAP32=_=new Int32Array(e),o.HEAPU8=R=new Uint8Array(e),o.HEAPU16=new Uint16Array(e),o.HEAPU32=new Uint32Array(e),o.HEAPF32=new Float32Array(e),o.HEAPF64=new Float64Array(e)}var P=[],$=[],B=[],z=[];$.push({func:function(){ne()}});var U=0,W=null;function V(e){o.onAbort&&o.onAbort(e),S(e+=""),I=!0,e="abort("+e+"). Build with -s ASSERTIONS=1 for more info.";var t=new WebAssembly.RuntimeError(e);throw n(t),t}function H(e,t){return String.prototype.startsWith?e.startsWith(t):0===e.indexOf(t)}function q(e){return H(e,"data:application/octet-stream;base64,")}function K(e){return H(e,"file://")}o.preloadedImages={},o.preloadedAudios={};var Y,X="tfjs-backend-wasm.wasm";function Q(e){try{if(e==X&&x)return new Uint8Array(x);if(g)return g(e);throw"both async and sync fetching of the wasm failed"}catch(e){V(e)}}function Z(e){for(;e.length>0;){var t=e.shift();if("function"!=typeof t){var n=t.func;"number"==typeof n?void 0===t.arg?F.get(n)():F.get(n)(t.arg):n(void 0===t.arg?null:t.arg)}else t(o)}}function J(e){try{return w.grow(e-C.byteLength+65535>>>16),L(w.buffer),1}catch(e){}}q(X)||(Y=X,X=o.locateFile?o.locateFile(Y,v):v+Y);var ee={mappings:{},buffers:[null,[],[]],printChar:function(e,t){var n=ee.buffers[e];0===t||10===t?((1===e?k:S)(D(n,0)),n.length=0):n.push(t)},varargs:void 0,get:function(){return ee.varargs+=4,_[ee.varargs-4>>2]},getStr:function(e){return O(e)},get64:function(e,t){return e}},te={a:function(){V()},d:function(e,t,n){R.copyWithin(e,t,t+n)},e:function(e){var t,n,r=R.length;if(e>2147483648)return!1;for(var a=1;a<=4;a*=2){var s=r*(1+.2/a);if(s=Math.min(s,e+100663296),J(Math.min(2147483648,((t=Math.max(e,s))%(n=65536)>0&&(t+=n-t%n),t))))return!0}return!1},f:function(e){return 0},c:function(e,t,n,r,a){},b:function(e,t,n,r){for(var a=0,s=0;s<n;s++){for(var i=_[t+8*s>>2],o=_[t+(8*s+4)>>2],u=0;u<o;u++)ee.printChar(e,R[i+u]);a+=o}return _[r>>2]=a,0},h:function(){return 6},g:function(){return 28},i:function(e){switch(e){case 30:return 16384;case 85:return 131072;case 132:case 133:case 12:case 137:case 138:case 15:case 235:case 16:case 17:case 18:case 19:case 20:case 149:case 13:case 10:case 236:case 153:case 9:case 21:case 22:case 159:case 154:case 14:case 77:case 78:case 139:case 82:case 68:case 67:case 164:case 11:case 29:case 47:case 48:case 95:case 52:case 51:case 46:return 200809;case 27:case 246:case 127:case 128:case 23:case 24:case 160:case 161:case 181:case 182:case 242:case 183:case 184:case 243:case 244:case 245:case 165:case 178:case 179:case 49:case 50:case 168:case 169:case 175:case 170:case 171:case 172:case 97:case 76:case 32:case 173:case 35:case 80:case 81:case 79:return-1;case 176:case 177:case 7:case 155:case 8:case 157:case 125:case 126:case 92:case 93:case 129:case 130:case 131:case 94:case 91:return 1;case 74:case 60:case 69:case 70:case 4:return 1024;case 31:case 42:case 72:return 32;case 87:case 26:case 33:return 2147483647;case 34:case 1:return 47839;case 38:case 36:return 99;case 43:case 37:return 2048;case 0:return 2097152;case 3:return 65536;case 28:return 32768;case 44:return 32767;case 75:return 16384;case 39:return 1e3;case 89:return 700;case 71:return 256;case 40:return 255;case 2:return 100;case 180:return 64;case 25:return 20;case 5:return 16;case 6:return 6;case 73:return 4;case 84:return"object"==typeof navigator&&navigator.hardwareConcurrency||1}var t;return t=28,_[ie()>>2]=t,-1}};!function(){var e={a:te};function t(e,t){var n=e.exports;o.asm=n,L((w=o.asm.j).buffer),F=o.asm.r,function(e){if(U--,o.monitorRunDependencies&&o.monitorRunDependencies(U),0==U&&W){var t=W;W=null,t()}}()}function r(e){t(e.instance)}function a(t){return function(){if(!x&&(l||c)){if("function"==typeof fetch&&!K(X))return fetch(X,{credentials:"same-origin"}).then((function(e){if(!e.ok)throw"failed to load wasm binary file at '"+X+"'";return e.arrayBuffer()})).catch((function(){return Q(X)}));if(m)return new Promise((function(e,t){m(X,(function(t){e(new Uint8Array(t))}),t)}))}return Promise.resolve().then((function(){return Q(X)}))}().then((function(t){return WebAssembly.instantiate(t,e)})).then(t,(function(e){S("failed to asynchronously prepare wasm: "+e),V(e)}))}if(U++,o.monitorRunDependencies&&o.monitorRunDependencies(U),o.instantiateWasm)try{return o.instantiateWasm(e,t)}catch(e){return S("Module.instantiateWasm callback failed with error: "+e),!1}(x||"function"!=typeof WebAssembly.instantiateStreaming||q(X)||K(X)||"function"!=typeof fetch?a(r):fetch(X,{credentials:"same-origin"}).then((function(t){return WebAssembly.instantiateStreaming(t,e).then(r,(function(e){return S("wasm streaming compile failed: "+e),S("falling back to ArrayBuffer instantiation"),a(r)}))}))).catch(n)}();var ne=o.___wasm_call_ctors=function(){return(ne=o.___wasm_call_ctors=o.asm.k).apply(null,arguments)};o._init=function(){return(o._init=o.asm.l).apply(null,arguments)},o._init_with_threads_count=function(){return(o._init_with_threads_count=o.asm.m).apply(null,arguments)},o._get_threads_count=function(){return(o._get_threads_count=o.asm.n).apply(null,arguments)},o._register_tensor=function(){return(o._register_tensor=o.asm.o).apply(null,arguments)},o._dispose_data=function(){return(o._dispose_data=o.asm.p).apply(null,arguments)},o._dispose=function(){return(o._dispose=o.asm.q).apply(null,arguments)},o._Abs=function(){return(o._Abs=o.asm.s).apply(null,arguments)},o._Add=function(){return(o._Add=o.asm.t).apply(null,arguments)},o._AddN=function(){return(o._AddN=o.asm.u).apply(null,arguments)},o._All=function(){return(o._All=o.asm.v).apply(null,arguments)},o._Any=function(){return(o._Any=o.asm.w).apply(null,arguments)},o._ArgMax=function(){return(o._ArgMax=o.asm.x).apply(null,arguments)},o._AvgPool=function(){return(o._AvgPool=o.asm.y).apply(null,arguments)},o._BatchMatMul=function(){return(o._BatchMatMul=o.asm.z).apply(null,arguments)},o._Ceil=function(){return(o._Ceil=o.asm.A).apply(null,arguments)},o._ClipByValue=function(){return(o._ClipByValue=o.asm.B).apply(null,arguments)},o._Conv2D=function(){return(o._Conv2D=o.asm.C).apply(null,arguments)},o._Conv2DBackpropInput=function(){return(o._Conv2DBackpropInput=o.asm.D).apply(null,arguments)},o._Cos=function(){return(o._Cos=o.asm.E).apply(null,arguments)},o._Cosh=function(){return(o._Cosh=o.asm.F).apply(null,arguments)},o._CropAndResize=function(){return(o._CropAndResize=o.asm.G).apply(null,arguments)},o._Cumsum=function(){return(o._Cumsum=o.asm.H).apply(null,arguments)},o._DepthToSpace=function(){return(o._DepthToSpace=o.asm.I).apply(null,arguments)},o._DepthwiseConv2dNative=function(){return(o._DepthwiseConv2dNative=o.asm.J).apply(null,arguments)},o._Elu=function(){return(o._Elu=o.asm.K).apply(null,arguments)},o._Equal=function(){return(o._Equal=o.asm.L).apply(null,arguments)},o._Exp=function(){return(o._Exp=o.asm.M).apply(null,arguments)},o._FlipLeftRight=function(){return(o._FlipLeftRight=o.asm.N).apply(null,arguments)},o._Floor=function(){return(o._Floor=o.asm.O).apply(null,arguments)},o._FloorDiv=function(){return(o._FloorDiv=o.asm.P).apply(null,arguments)},o._FusedBatchNorm=function(){return(o._FusedBatchNorm=o.asm.Q).apply(null,arguments)},o._FusedConv2D=function(){return(o._FusedConv2D=o.asm.R).apply(null,arguments)},o._FusedDepthwiseConv2D=function(){return(o._FusedDepthwiseConv2D=o.asm.S).apply(null,arguments)},o._Gather=function(){return(o._Gather=o.asm.T).apply(null,arguments)},o._GatherNd=function(){return(o._GatherNd=o.asm.U).apply(null,arguments)},o._Greater=function(){return(o._Greater=o.asm.V).apply(null,arguments)},o._GreaterEqual=function(){return(o._GreaterEqual=o.asm.W).apply(null,arguments)},o._LeakyRelu=function(){return(o._LeakyRelu=o.asm.X).apply(null,arguments)},o._Less=function(){return(o._Less=o.asm.Y).apply(null,arguments)},o._LessEqual=function(){return(o._LessEqual=o.asm.Z).apply(null,arguments)},o._Log=function(){return(o._Log=o.asm._).apply(null,arguments)},o._LogicalAnd=function(){return(o._LogicalAnd=o.asm.$).apply(null,arguments)},o._Max=function(){return(o._Max=o.asm.aa).apply(null,arguments)},o._MaxPool=function(){return(o._MaxPool=o.asm.ba).apply(null,arguments)},o._Maximum=function(){return(o._Maximum=o.asm.ca).apply(null,arguments)},o._Mean=function(){return(o._Mean=o.asm.da).apply(null,arguments)},o._Min=function(){return(o._Min=o.asm.ea).apply(null,arguments)},o._Minimum=function(){return(o._Minimum=o.asm.fa).apply(null,arguments)},o._MirrorPad=function(){return(o._MirrorPad=o.asm.ga).apply(null,arguments)},o._Multiply=function(){return(o._Multiply=o.asm.ha).apply(null,arguments)},o._Neg=function(){return(o._Neg=o.asm.ia).apply(null,arguments)},o._NonMaxSuppressionV3=function(){return(o._NonMaxSuppressionV3=o.asm.ja).apply(null,arguments)},o._NonMaxSuppressionV4=function(){return(o._NonMaxSuppressionV4=o.asm.ka).apply(null,arguments)},o._NonMaxSuppressionV5=function(){return(o._NonMaxSuppressionV5=o.asm.la).apply(null,arguments)},o._NotEqual=function(){return(o._NotEqual=o.asm.ma).apply(null,arguments)},o._OneHot=function(){return(o._OneHot=o.asm.na).apply(null,arguments)},o._PadV2=function(){return(o._PadV2=o.asm.oa).apply(null,arguments)},o._Pow=function(){return(o._Pow=o.asm.pa).apply(null,arguments)},o._Prelu=function(){return(o._Prelu=o.asm.qa).apply(null,arguments)},o._Prod=function(){return(o._Prod=o.asm.ra).apply(null,arguments)},o._RealDiv=function(){return(o._RealDiv=o.asm.sa).apply(null,arguments)},o._Relu=function(){return(o._Relu=o.asm.ta).apply(null,arguments)},o._Relu6=function(){return(o._Relu6=o.asm.ua).apply(null,arguments)},o._ResizeBilinear=function(){return(o._ResizeBilinear=o.asm.va).apply(null,arguments)},o._Reverse=function(){return(o._Reverse=o.asm.wa).apply(null,arguments)},o._RotateWithOffset=function(){return(o._RotateWithOffset=o.asm.xa).apply(null,arguments)},o._Round=function(){return(o._Round=o.asm.ya).apply(null,arguments)},o._Rsqrt=function(){return(o._Rsqrt=o.asm.za).apply(null,arguments)},o._ScatterNd=function(){return(o._ScatterNd=o.asm.Aa).apply(null,arguments)},o._SelectV2=function(){return(o._SelectV2=o.asm.Ba).apply(null,arguments)},o._Sigmoid=function(){return(o._Sigmoid=o.asm.Ca).apply(null,arguments)},o._Sin=function(){return(o._Sin=o.asm.Da).apply(null,arguments)},o._Softmax=function(){return(o._Softmax=o.asm.Ea).apply(null,arguments)},o._SparseFillEmptyRows=function(){return(o._SparseFillEmptyRows=o.asm.Fa).apply(null,arguments)},o._SparseReshape=function(){return(o._SparseReshape=o.asm.Ga).apply(null,arguments)},o._SparseSegmentReduction=function(){return(o._SparseSegmentReduction=o.asm.Ha).apply(null,arguments)},o._Sqrt=function(){return(o._Sqrt=o.asm.Ia).apply(null,arguments)},o._Square=function(){return(o._Square=o.asm.Ja).apply(null,arguments)},o._SquaredDifference=function(){return(o._SquaredDifference=o.asm.Ka).apply(null,arguments)},o._Step=function(){return(o._Step=o.asm.La).apply(null,arguments)},o._StridedSlice=function(){return(o._StridedSlice=o.asm.Ma).apply(null,arguments)},o._Sub=function(){return(o._Sub=o.asm.Na).apply(null,arguments)},o._Sum=function(){return(o._Sum=o.asm.Oa).apply(null,arguments)},o._Tan=function(){return(o._Tan=o.asm.Pa).apply(null,arguments)},o._Tanh=function(){return(o._Tanh=o.asm.Qa).apply(null,arguments)},o._Tile=function(){return(o._Tile=o.asm.Ra).apply(null,arguments)},o._TopK=function(){return(o._TopK=o.asm.Sa).apply(null,arguments)},o._Transform=function(){return(o._Transform=o.asm.Ta).apply(null,arguments)},o._Transpose=function(){return(o._Transpose=o.asm.Ua).apply(null,arguments)},o.__FusedMatMul=function(){return(o.__FusedMatMul=o.asm.Va).apply(null,arguments)},o._malloc=function(){return(o._malloc=o.asm.Wa).apply(null,arguments)},o._free=function(){return(o._free=o.asm.Xa).apply(null,arguments)};var re,ae,se,ie=o.___errno_location=function(){return(ie=o.___errno_location=o.asm.Ya).apply(null,arguments)},oe=o.stackSave=function(){return(oe=o.stackSave=o.asm.Za).apply(null,arguments)},ue=o.stackRestore=function(){return(ue=o.stackRestore=o.asm._a).apply(null,arguments)},le=o.stackAlloc=function(){return(le=o.stackAlloc=o.asm.$a).apply(null,arguments)};function ce(e){this.name="ExitStatus",this.message="Program terminated with exit("+e+")",this.status=e}function he(e){function n(){re||(re=!0,o.calledRun=!0,I||(Z($),Z(B),t(o),o.onRuntimeInitialized&&o.onRuntimeInitialized(),function(){if(o.postRun)for("function"==typeof o.postRun&&(o.postRun=[o.postRun]);o.postRun.length;)e=o.postRun.shift(),z.unshift(e);var e;Z(z)}()))}U>0||(function(){if(o.preRun)for("function"==typeof o.preRun&&(o.preRun=[o.preRun]);o.preRun.length;)e=o.preRun.shift(),P.unshift(e);var e;Z(P)}(),U>0||(o.setStatus?(o.setStatus("Running..."),setTimeout((function(){setTimeout((function(){o.setStatus("")}),1),n()}),1)):n()))}if(o.cwrap=function(e,t,n,r){var a=(n=n||[]).every((function(e){return"number"===e}));return"string"!==t&&a&&!r?E(e):function(){return N(e,t,n,arguments)}},W=function e(){re||he(),re||(W=e)},o.run=he,o.preInit)for("function"==typeof o.preInit&&(o.preInit=[o.preInit]);o.preInit.length>0;)o.preInit.pop()();if(he(),s&&(ae={uncaughtException:r.listeners("uncaughtException").filter((function(e){return!s.uncaughtException.indexOf(e)>-1})),unhandledRejection:r.listeners("unhandledRejection").filter((function(e){return!s.unhandledRejection.indexOf(e)>-1}))}),void 0!==e)se=e;else{if("undefined"==typeof WasmBackendModuleThreadedSimd)throw new Error("Could not find wasm module in post.js");se=WasmBackendModuleThreadedSimd}if(ae){var pe=se._dispose;se._dispose=function(){pe(),ae.uncaughtException.forEach((function(e){r.removeListener("uncaughtException",e)})),ae.unhandledRejection.forEach((function(e){r.removeListener("unhandledRejection",e)}))}}return e.ready});"object"==typeof t&&"object"==typeof s?s.exports=o:"function"==typeof define&&n(66)?define([],(function(){return o})):"object"==typeof t&&(t.WasmBackendModule=o)}}),Q=class{constructor(e,t){this.backend=e,this.dataMover=t,this.data=new WeakMap,this.dataIdsCount=0}get(e){return this.data.has(e)||this.dataMover.moveData(this.backend,e),this.data.get(e)}set(e,t){this.dataIdsCount++,this.data.set(e,t)}has(e){return this.data.has(e)}delete(e){return this.dataIdsCount--,this.data.delete(e)}numDataIds(){return this.dataIdsCount}},Z=class{refCount(e){return J("refCount")}incRef(e){return J("incRef")}timerAvailable(){return!0}time(e){return J("time")}read(e){return J("read")}readSync(e){return J("readSync")}readToGPU(e,t){return J("readToGPU")}numDataIds(){return J("numDataIds")}disposeData(e,t){return J("disposeData")}write(e,t,n){return J("write")}move(e,t,n,r,a){return J("move")}memory(){return J("memory")}floatPrecision(){return J("floatPrecision")}epsilon(){return 32===this.floatPrecision()?1e-7:1e-4}dispose(){return J("dispose")}};function J(e){throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function ee(e){let t=e.length,n=0;for(;t>0;)n=Math.random()*t|0,t--,ae(e,t,n)}function te(e,t){if(e.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);let n=e.length,r=0;for(;n>0;)r=Math.random()*n|0,n--,ae(e,n,r),ae(t,n,r)}function ne(e,t,n){return Math.max(e,Math.min(t,n))}function re(e){return e%2==0?e:e+1}function ae(e,t,n){const r=e[t];e[t]=e[n],e[n]=r}function se(e){let t=0;for(let n=0;n<e.length;n++)t+=e[n];return t}function ie(e,t){const n=Math.random();return t*n+(1-n)*e}function oe(e,t){let n=0;for(let r=0;r<e.length;r++){const a=Number(e[r])-Number(t[r]);n+=a*a}return n}function ue(e,t){if(!e)throw new Error("string"==typeof t?t:t())}function le(e,t,n=""){ue(fe(e,t),()=>n+` Shapes ${e} and ${t} must match`)}function ce(e){ue(null!=e,()=>"The input to the tensor constructor must be a non-null value.")}function he(e,t=[],n=!1){if(null==t&&(t=[]),Array.isArray(e)||Te(e)&&!n)for(let r=0;r<e.length;++r)he(e[r],t,n);else t.push(e);return t}function pe(e){if(0===e.length)return 1;let t=e[0];for(let n=1;n<e.length;n++)t*=e[n];return t}function de(e){return 0===e.length}function fe(e,t){if(e===t)return!0;if(null==e||null==t)return!1;if(e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(e[n]!==t[n])return!1;return!0}function me(e){return e%1==0}function ge(e){if(null!=Math.tanh)return Math.tanh(e);if(e===1/0)return 1;if(e===-1/0)return-1;{const t=Math.exp(2*e);return(t-1)/(t+1)}}function ye(e){const t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function be(e){const t=new Uint32Array(e);for(let n=0;n<e;++n)t[n]=n;return ee(t),t}function ve(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function xe(e,t=(e=>0),n){return new Promise((r,a)=>{let s=0;const i=()=>{if(e())return void r();s++;const o=t(s);null!=n&&s>=n?a():setTimeout(i,o)};i()})}function we(e,t){let n=1,r=-1;for(let t=0;t<e.length;++t)if(e[t]>=0)n*=e[t];else if(-1===e[t]){if(-1!==r)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${r} and dim ${t}`);r=t}else if(e[t]<0)throw Error(`Shapes can not be < 0. Found ${e[t]} at dim ${t}`);if(-1===r){if(t>0&&t!==n)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(0===n)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%n!=0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);const a=e.slice();return a[r]=t/n,a}function ke(e,t){const n=t.length;return ue((e=null==e?t.map((e,t)=>t):[].concat(e)).every(e=>e>=-n&&e<n),()=>`All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`),ue(e.every(e=>me(e)),()=>"All values in axis param must be integers but got axis "+e),e.map(e=>e<0?n+e:e)}function Se(e,t){const n=[],r=[],a=null!=t&&Array.isArray(t)&&0===t.length,s=null==t||a?null:ke(t,e).sort();let i=0;for(let t=0;t<e.length;++t){if(null!=s){if(s[i]===t&&1!==e[t])throw new Error(`Can't squeeze axis ${t} since its dim '${e[t]}' is not 1`);(null==s[i]||s[i]>t)&&1===e[t]&&(n.push(e[t]),r.push(t)),s[i]<=t&&i++}1!==e[t]&&(n.push(e[t]),r.push(t))}return{newShape:n,keptDims:r}}function Ie(e,t){let n=null;if(null==e||"float32"===e)n=new Float32Array(t);else if("int32"===e)n=new Int32Array(t);else{if("bool"!==e)throw new Error("Unknown data type "+e);n=new Uint8Array(t)}return n}function Ae(e,t){let n=null;if(null==e||"float32"===e)n=new Float32Array(t);else if("int32"===e)n=new Int32Array(t);else if("bool"===e)n=new Uint8Array(t);else{if("string"!==e)throw new Error("Unknown data type "+e);n=new Array(t)}return n}function Ee(e,t){for(let n=0;n<e.length;n++){const r=e[n];if(isNaN(r)||!isFinite(r))throw Error(`A tensor of type ${t} being uploaded contains ${r}.`)}}function Ne(e){return"bool"===e||"complex64"===e||"float32"===e||"int32"===e||"string"===e}function Ce(e,t){return"complex64"!==t&&(("float32"!==t||"complex64"===e)&&(("int32"!==t||"float32"===e||"complex64"===e)&&("bool"!==t||"bool"!==e)))}function Te(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray}function Re(e){if("float32"===e||"int32"===e)return 4;if("complex64"===e)return 8;if("bool"===e)return 1;throw new Error("Unknown dtype "+e)}function _e(e){if(null==e)return 0;let t=0;return e.forEach(e=>t+=e.length),t}function Fe(e){return"string"==typeof e||e instanceof String}function Me(e){return"boolean"==typeof e}function De(e){return"number"==typeof e}function Oe(e){return Array.isArray(e)?Oe(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray?"int32":De(e)?"float32":Fe(e)?"string":Me(e)?"bool":"float32"}function Le(e){return!!(e&&e.constructor&&e.call&&e.apply)}function Pe(e,t){for(let n=t;n<e;++n)if(e%n==0)return n;return e}function $e(e){const t=e.length;if(t<2)return[];const n=new Array(t-1);n[t-2]=e[t-1];for(let r=t-3;r>=0;--r)n[r]=n[r+1]*e[r+1];return n}function Be(e,t,n=!1){if(0===e.length)return t[0];const r=e.reduce((e,t)=>e*t)*(n?2:1);if(0===r)return[];if(r!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}${n?" for a complex tensor":""}.`);return function e(t,n,r,a=!1){const s=new Array;if(1===n.length){const e=n[0]*(a?2:1);for(let n=0;n<e;n++)s[n]=r[t+n]}else{const i=n[0],o=n.slice(1),u=o.reduce((e,t)=>e*t)*(a?2:1);for(let n=0;n<i;n++)s[n]=e(t+n*u,o,r,a)}return s}(0,e,t,n)}function ze(e,t){const n=Ue(e,t);for(let e=0;e<n.length;e++)n[e]=1;return n}function Ue(e,t){if(null==t||"float32"===t||"complex64"===t)return new Float32Array(e);if("int32"===t)return new Int32Array(e);if("bool"===t)return new Uint8Array(e);throw new Error("Unknown data type "+t)}function We(e,t){const n=e.reduce((e,t)=>e*t,1);if(null==t||"float32"===t)return Be(e,new Float32Array(n));if("int32"===t)return Be(e,new Int32Array(n));if("bool"===t)return Be(e,new Uint8Array(n));throw new Error("Unknown data type "+t)}function Ve(e){e.forEach(t=>{ue(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function je(e,t,n){if(0===t)return 0;if(1===t)return e[0];let r=e[e.length-1];for(let t=0;t<e.length-1;++t)r+=n[t]*e[t];return r}function Ge(e,t,n){if(0===t)return[];if(1===t)return[e];const r=new Array(t);for(let t=0;t<r.length-1;++t)r[t]=Math.floor(e/n[t]),e-=r[t]*n[t];return r[r.length-1]=e,r}function He(e){return e&&e.then&&"function"==typeof e.then}var qe=class{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=Ke,this.populateURLFlags()}setPlatform(e,t){null!=this.platform&&(Ye().getBool("IS_TEST")||Ye().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${t}.`)),this.platformName=e,this.platform=t}registerFlag(e,t,n){if(this.flagRegistry[e]={evaluationFn:t,setHook:n},null!=this.urlFlags[e]){const t=this.urlFlags[e];Ye().getBool("IS_TEST")||Ye().getBool("PROD")||console.warn(`Setting feature override from URL ${e}: ${t}.`),this.set(e,t)}}async getAsync(e){return e in this.flags||(this.flags[e]=await this.evaluateFlag(e)),this.flags[e]}get(e){if(e in this.flags)return this.flags[e];const t=this.evaluateFlag(e);if(He(t))throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[e]=t,this.flags[e]}getNumber(e){return this.get(e)}getBool(e){return this.get(e)}getFlags(){return this.flags}get features(){return this.flags}set(e,t){if(null==this.flagRegistry[e])throw new Error(`Cannot set flag ${e} as it has not been registered.`);this.flags[e]=t,null!=this.flagRegistry[e].setHook&&this.flagRegistry[e].setHook(t)}evaluateFlag(e){if(null==this.flagRegistry[e])throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);return this.flagRegistry[e].evaluationFn()}setFlags(e){this.flags=Object.assign({},e)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(void 0===this.global||void 0===this.global.location||void 0===this.global.location.search)return;const e=this.getQueryParams(this.global.location.search);if("tfjsflags"in e){e.tfjsflags.split(",").forEach(e=>{const[t,n]=e.split(":");this.urlFlags[t]=function(e,t){if("true"===(t=t.toLowerCase())||"false"===t)return"true"===t;if(""+ +t===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${e}.`)}(t,n)})}}};function Ke(e){const t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(e,...n)=>(function(e,t,n){e[decodeURIComponent(t)]=decodeURIComponent(n||"")}(t,n[0],n[1]),n.join("="))),t}function Ye(){return Qe}var Xe,Qe=null;function Ze(){if(null==Xe){let e;if("undefined"!=typeof window)e=window;else if(void 0!==s)e=s;else if(void 0!==r)e=r;else{if("undefined"==typeof self)throw new Error("Could not find a global object");e=self}Xe=e}return Xe}function Je(e,t){const n=function(){const e=Ze();return null==e._tfGlobals&&(e._tfGlobals=new Map),e._tfGlobals}();if(n.has(e))return n.get(e);{const r=t();return n.set(e,r),n.get(e)}}var et="Abs",tt="Acos",nt="Acosh",rt="Add",at="AddN",st="All",it="Any",ot="ArgMax",ut="ArgMin",lt="Asin",ct="Asinh",ht="Atan",pt="Atanh",dt="Atan2",ft="AvgPool",mt="AvgPoolGrad",gt="AvgPool3D",yt="AvgPool3DGrad",bt="BatchMatMul",vt="BatchToSpaceND",xt="Bincount",wt="BroadcastTo",kt="BroadcastArgs",St="Cast",It="Ceil",At="ClipByValue",Et="Complex",Nt="ComplexAbs",Ct="Concat",Tt="Conv2D",Rt="Conv2DBackpropFilter",_t="Conv2DBackpropInput",Ft="Conv3D",Mt="Conv3DBackpropFilterV2",Dt="Conv3DBackpropInputV2",Ot="Cos",Lt="Cosh",Pt="Cumsum",$t="CropAndResize",Bt="DenseBincount",zt="DepthToSpace",Ut="DepthwiseConv2dNative",Wt="DepthwiseConv2dNativeBackpropFilter",Vt="DepthwiseConv2dNativeBackpropInput",jt="Diag",Gt="Dilation2D",Ht="Dilation2DBackpropInput",qt="Dilation2DBackpropFilter",Kt="RealDiv",Yt="Einsum",Xt="Elu",Qt="EluGrad",Zt="Erf",Jt="Equal",en="Exp",tn="ExpandDims",nn="Expm1",rn="FFT",an="Fill",sn="FlipLeftRight",on="Floor",un="FloorDiv",ln="FusedBatchNorm",cn="GatherV2",hn="GatherNd",pn="Greater",dn="GreaterEqual",fn="Identity",mn="IFFT",gn="Imag",yn="IsFinite",bn="IsInf",vn="IsNan",xn="LeakyRelu",wn="Less",kn="LessEqual",Sn="LinSpace",In="Log",An="Log1p",En="LogicalAnd",Nn="LogicalNot",Cn="LogicalOr",Tn="LogSoftmax",Rn="LRN",_n="LRNGrad",Fn="Max",Mn="Maximum",Dn="MaxPool",On="MaxPoolGrad",Ln="MaxPool3D",Pn="MaxPool3DGrad",$n="MaxPoolWithArgmax",Bn="Mean",zn="Min",Un="Minimum",Wn="MirrorPad",Vn="Mod",jn="Multinomial",Gn="Multiply",Hn="Neg",qn="NotEqual",Kn="NonMaxSuppressionV3",Yn="NonMaxSuppressionV4",Xn="NonMaxSuppressionV5",Qn="OnesLike",Zn="OneHot",Jn="Pack",er="PadV2",tr="Pool",nr="Pow",rr="Prelu",ar="Prod",sr="Range",ir="Real",or="Reciprocal",ur="Relu",lr="Reshape",cr="ResizeNearestNeighbor",hr="ResizeNearestNeighborGrad",pr="ResizeBilinear",dr="ResizeBilinearGrad",fr="Relu6",mr="Reverse",gr="Round",yr="Rsqrt",br="ScatterNd",vr="Select",xr="Selu",wr="Slice",kr="Sin",Sr="Sinh",Ir="Sign",Ar="Sigmoid",Er="Softplus",Nr="Sqrt",Cr="Sum",Tr="SpaceToBatchND",Rr="SplitV",_r="Softmax",Fr="SparseFillEmptyRows",Mr="SparseReshape",Dr="SparseSegmentMean",Or="SparseSegmentSum",Lr="SparseToDense",Pr="SquaredDifference",$r="Square",Br="StridedSlice",zr="StringNGrams",Ur="StringSplit",Wr="StringToHashBucketFast",Vr="Sub",jr="Tan",Gr="Tanh",Hr="Tile",qr="TopK",Kr="Transform",Yr="Transpose",Xr="Unique",Qr="Unpack",Zr="UnsortedSegmentSum",Jr="ZerosLike",ea="Step",ta="FromPixels",na="RotateWithOffset",ra="_FusedMatMul",aa="FusedConv2D",sa="FusedDepthwiseConv2D";function ia(...e){Ye().getBool("IS_TEST")||Ye().getBool("PROD")||console.warn(...e)}function oa(...e){Ye().getBool("IS_TEST")||Ye().getBool("PROD")||console.log(...e)}var ua=Je("kernelRegistry",()=>new Map),la=Je("gradRegistry",()=>new Map);function ca(e,t){const n=ba(e,t);return ua.get(n)}function ha(e){return la.get(e)}function pa(e){const t=ua.entries(),n=[];for(;;){const{done:r,value:a}=t.next();if(r)break;const[s,i]=a,[o]=s.split("_");o===e&&n.push(i)}return n}function da(e){const{kernelName:t,backendName:n}=e,r=ba(t,n);ua.has(r)&&ia(`The kernel '${t}' for backend '${n}' is already registered`),ua.set(r,e)}function fa(e){const{kernelName:t}=e;la.has(t)&&Ye().getBool("DEBUG")&&ia(`Overriding the gradient for '${t}'`),la.set(t,e)}function ma(e,t){const n=ba(e,t);if(!ua.has(n))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);ua.delete(n)}function ga(e){if(!la.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);la.delete(e)}function ya(e,t){pa(e).forEach(e=>{da(Object.assign({},e,{backendName:t}))})}function ba(e,t){return`${t}_${e}`}var va={};w(va,{arraysEqual:()=>fe,assert:()=>ue,assertNonNegativeIntegerDimensions:()=>Ve,assertNonNull:()=>ce,assertShapesMatch:()=>le,bytesFromStringArray:()=>_e,bytesPerElement:()=>Re,checkConversionForErrors:()=>Ee,clamp:()=>ne,computeStrides:()=>$e,createScalarValue:()=>Da,createShuffledIndices:()=>be,decodeString:()=>Ba,distSquared:()=>oe,encodeString:()=>$a,fetch:()=>Pa,fingerPrint64:()=>Ma,flatten:()=>he,getArrayFromDType:()=>Ae,getTypedArrayFromDType:()=>Ie,hasEncodingLoss:()=>Ce,hexToLong:()=>ka,indexToLoc:()=>Ge,inferDtype:()=>Oe,inferFromImplicitShape:()=>we,isBoolean:()=>Me,isFunction:()=>Le,isInt:()=>me,isNumber:()=>De,isPromise:()=>He,isScalarShape:()=>de,isString:()=>Fe,isTypedArray:()=>Te,isValidDtype:()=>Ne,locToIndex:()=>je,makeOnesTypedArray:()=>ze,makeZerosNestedTypedArray:()=>We,makeZerosTypedArray:()=>Ue,nearestDivisor:()=>Pe,nearestLargerEven:()=>re,now:()=>La,parseAxisParam:()=>ke,randUniform:()=>ie,repeatedTry:()=>xe,rightPad:()=>ve,shuffle:()=>ee,shuffleCombo:()=>te,sizeFromShape:()=>pe,sizeToSquarishShape:()=>ye,squeezeShape:()=>Se,sum:()=>se,swap:()=>ae,tanh:()=>ge,toNestedArray:()=>Be,toTypedArray:()=>Oa});var xa=k(S()),wa=xa.default||xa;function ka(e){return wa.fromString(e,!0,16)}var Sa=ka("c3a5c85c97cb3127"),Ia=ka("b492b66fbe98f273"),Aa=ka("9ae16a3b2f90404f");function Ea(e){return e.xor(e.shru(47))}function Na(e,t,n){const r=e.slice(t,t+n);return wa.fromBytes(Array.from(r),!0,!0)}function Ca(e,t){return Na(e,t,8)}function Ta(e,t){return Na(e,t,4)}function Ra(e,t){return 0===t?e:e.shru(t).or(e.shl(64-t))}function _a(e,t,n=ka("9ddfea08eb382d69")){let r=e.xor(t).mul(n);r=r.xor(r.shru(47));let a=t.xor(r).mul(n);return a=a.xor(a.shru(47)),a=a.mul(n),a}function Fa(e,t,n,r){return function(e,t,n,r,a,s){a=a.add(e),s=Ra(s.add(a).add(r),21);const i=a;return a=(a=a.add(t)).add(n),s=s.add(Ra(a,44)),[a.add(r),s.add(i)]}(Ca(e,t),Ca(e,t+8),Ca(e,t+16),Ca(e,t+24),n,r)}function Ma(e,t=e.length){const n=wa.fromNumber(81,!0);if(t<=32)return t<=16?function(e,t=e.length){if(t>=8){const n=Aa.add(2*t),r=Ca(e,0).add(Aa),a=Ca(e,t-8);return _a(Ra(a,37).mul(n).add(r),Ra(r,25).add(a).mul(n),n)}if(t>=4){const n=Aa.add(2*t);return _a(Ta(e,0).shl(3).add(t),Ta(e,t-4),n)}if(t>0){const n=e[0]+(e[t>>1]<<8),r=t+(e[t-1]<<2);return Ea(Aa.mul(n).xor(Sa.mul(r))).mul(Aa)}return Aa}(e,t):function(e,t=e.length){const n=Aa.add(2*t),r=Ca(e,0).mul(Ia),a=Ca(e,8),s=Ca(e,t-8).mul(n),i=Ca(e,t-16).mul(Aa);return _a(Ra(r.add(a),43).add(Ra(s,30)).add(i),r.add(Ra(a.add(Aa),18)).add(s),n)}(e,t);if(t<=64)return function(e,t=e.length){const n=Aa.add(2*t),r=Ca(e,0).mul(Aa),a=Ca(e,8),s=Ca(e,t-8).mul(n),i=Ca(e,t-16).mul(Aa),o=Ra(r.add(a),43).add(Ra(s,30)).add(i),u=_a(o,r.add(Ra(a.add(Aa),18)).add(s),n),l=Ca(e,16).mul(n),c=Ca(e,24),h=o.add(Ca(e,t-32)).mul(n),p=u.add(Ca(e,t-24)).mul(n);return _a(Ra(l.add(c),43).add(Ra(h,30)).add(p),l.add(Ra(c.add(r),18)).add(h),n)}(e,t);let r=n,a=n.mul(Ia).add(113),s=Ea(a.mul(Aa).add(113)).mul(Aa),i=[wa.UZERO,wa.UZERO],o=[wa.UZERO,wa.UZERO];r=r.mul(Aa).add(Ca(e,0));let u=0;const l=64*(t-1>>6),c=l+(t-1&63)-63;do{r=Ra(r.add(a).add(i[0]).add(Ca(e,u+8)),37).mul(Ia),a=Ra(a.add(i[1]).add(Ca(e,u+48)),42).mul(Ia),r=r.xor(o[1]),a=a.add(i[0]).add(Ca(e,u+40)),s=Ra(s.add(o[0]),33).mul(Ia),i=Fa(e,u,i[1].mul(Ia),r.add(o[0])),o=Fa(e,u+32,s.add(o[1]),a.add(Ca(e,u+16))),[s,r]=[r,s],u+=64}while(u!==l);const h=Ia.add(s.and(255).shl(1));return u=c,o[0]=o[0].add(t-1&63),i[0]=i[0].add(o[0]),o[0]=o[0].add(i[0]),r=Ra(r.add(a).add(i[0]).add(Ca(e,u+8)),37).mul(h),a=Ra(a.add(i[1]).add(Ca(e,u+48)),42).mul(h),r=r.xor(o[1].mul(9)),a=a.add(i[0].mul(9).add(Ca(e,u+40))),s=Ra(s.add(o[0]),33).mul(h),i=Fa(e,u,i[1].mul(h),r.add(o[0])),o=Fa(e,u+32,s.add(o[1]),a.add(Ca(e,u+16))),[s,r]=[r,s],_a(_a(i[0],o[0],h).add(Ea(a).mul(Sa)).add(s),_a(i[1],o[1],h).add(r),h)}function Da(e,t){return"string"===t?$a(e):Oa([e],t)}function Oa(e,t){if("string"===t)throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=he(e)),Ye().getBool("DEBUG")&&Ee(e,t),function(e,t){return e instanceof Float32Array&&"float32"===t||e instanceof Int32Array&&"int32"===t||e instanceof Uint8Array&&"bool"===t}(e,t))return e;if(null==t||"float32"===t||"complex64"===t)return new Float32Array(e);if("int32"===t)return new Int32Array(e);if("bool"===t){const t=new Uint8Array(e.length);for(let n=0;n<t.length;++n)0!==Math.round(e[n])&&(t[n]=1);return t}throw new Error("Unknown data type "+t)}function La(){return Ye().platform.now()}function Pa(e,t){return Ye().platform.fetch(e,t)}function $a(e,t="utf-8"){return t=t||"utf-8",Ye().platform.encode(e,t)}function Ba(e,t="utf-8"){return t=t||"utf-8",Ye().platform.decode(e,t)}function za(e,t,n){if("float32"!==t)return!1;for(let t=0;t<e.length;t++){const r=e[t];if(isNaN(r)||!isFinite(r))return console.warn(`Found ${r} in the result of '${n}'`),!0}return!1}var Ua=class{logKernelProfile(e,t,n,r,a,s){const i="number"==typeof r?ve(r+"ms",9):r.error,o=ve(e,25),u=t.rank,l=t.size,c=ve(t.shape.toString(),14);let h="";for(const e in a){const n=a[e];if(null!=n){const r=n.shape||t.shape,a=r.length;h+=`${e}: ${a}D ${a>0?r:""} `}}console.log(`%c${o}\t%c${i}\t%c${u}D ${c}\t%c${l}\t%c${h}\t%c${s}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function Wa(e,t,n,r){const a=$e(t),s=function(e,t,n,r){const a=pe(t),s=r[r.length-1],i=new Array(s).fill(0),o=t.length,u="complex64"===n?Ga(e):e;if(o>1)for(let e=0;e<a/s;e++){const t=e*s;for(let e=0;e<s;e++)i[e]=Math.max(i[e],Va(u[t+e],0,n).length)}return i}(e,t,n,a),i=t.length,o=function e(t,n,r,a,s,i=!0){const o="complex64"===r?2:1,u=n[0],l=n.length;if(0===l){if("complex64"===r){return[Va(Ga(t)[0],0,r)]}return"bool"===r?[ja(t[0])]:[t[0].toString()]}if(1===l){if(u>20){const e=3*o;let n=Array.from(t.slice(0,e)),a=Array.from(t.slice((u-3)*o,u*o));return"complex64"===r&&(n=Ga(n),a=Ga(a)),["["+n.map((e,t)=>Va(e,s[t],r)).join(", ")+", ..., "+a.map((e,t)=>Va(e,s[u-3+t],r)).join(", ")+"]"]}return["["+("complex64"===r?Ga(t):Array.from(t)).map((e,t)=>Va(e,s[t],r)).join(", ")+"]"]}const c=n.slice(1),h=a.slice(1),p=a[0]*o,d=[];if(u>20){for(let n=0;n<3;n++){const a=n*p,i=a+p;d.push(...e(t.slice(a,i),c,r,h,s,!1))}d.push("...");for(let n=u-3;n<u;n++){const a=n*p,i=a+p;d.push(...e(t.slice(a,i),c,r,h,s,n===u-1))}}else for(let n=0;n<u;n++){const a=n*p,i=a+p;d.push(...e(t.slice(a,i),c,r,h,s,n===u-1))}const f=2===l?",":"";d[0]="["+d[0]+f;for(let e=1;e<d.length-1;e++)d[e]=" "+d[e]+f;let m=",\n";for(let e=2;e<l;e++)m+="\n";return d[d.length-1]=" "+d[d.length-1]+"]"+(i?"":m),d}(e,t,n,a,s),u=["Tensor"];return r&&(u.push("  dtype: "+n),u.push("  rank: "+i),u.push(`  shape: [${t}]`),u.push("  values:")),u.push(o.map(e=>"    "+e).join("\n")),u.join("\n")}function Va(e,t,n){let r;return r=Array.isArray(e)?`${parseFloat(e[0].toFixed(7))} + ${parseFloat(e[1].toFixed(7))}j`:Fe(e)?`'${e}'`:"bool"===n?ja(e):parseFloat(e.toFixed(7)).toString(),ve(r,t)}function ja(e){return 0===e?"false":"true"}function Ga(e){const t=[];for(let n=0;n<e.length;n+=2)t.push([e[n],e[n+1]]);return t}var Ha=class{constructor(e,t,n){if(this.dtype=t,this.shape=e.slice(),this.size=pe(e),null!=n){const e=n.length;ue(e===this.size,()=>`Length of values '${e}' does not match the size inferred by the shape '${this.size}'.`)}if("complex64"===t)throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");this.values=n||Ae(t,this.size),this.strides=$e(e)}set(e,...t){0===t.length&&(t=[0]),ue(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);const n=this.locToIndex(t);this.values[n]=e}get(...e){0===e.length&&(e=[0]);let t=0;for(const n of e){if(n<0||n>=this.shape[t]){const t=`Requested out of range element at ${e}.   Buffer shape=${this.shape}`;throw new Error(t)}t++}let n=e[e.length-1];for(let t=0;t<e.length-1;++t)n+=this.strides[t]*e[t];return this.values[n]}locToIndex(e){if(0===this.rank)return 0;if(1===this.rank)return e[0];let t=e[e.length-1];for(let n=0;n<e.length-1;++n)t+=this.strides[n]*e[n];return t}indexToLoc(e){if(0===this.rank)return[];if(1===this.rank)return[e];const t=new Array(this.shape.length);for(let n=0;n<t.length-1;++n)t[n]=Math.floor(e/this.strides[n]),e-=t[n]*this.strides[n];return t[t.length-1]=e,t}get rank(){return this.shape.length}toTensor(){return qa().makeTensor(this.values,this.shape,this.dtype)}},qa=null,Ka=null;var Ya=class{constructor(e,t,n,r){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=pe(e),this.strides=$e(e),this.dataId=n,this.id=r,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){const e=await this.data();return Ka.buffer(this.shape,this.dtype,e)}bufferSync(){return Ka.buffer(this.shape,this.dtype,this.dataSync())}async array(){const e=await this.data();return Be(this.shape,e,"complex64"===this.dtype)}arraySync(){return Be(this.shape,this.dataSync(),"complex64"===this.dtype)}async data(){this.throwIfDisposed();const e=qa().read(this.dataId);if("string"===this.dtype){const t=await e;try{return t.map(e=>Ba(e))}catch(e){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataToGPU(e){return this.throwIfDisposed(),qa().readToGPU(this.dataId,e)}dataSync(){this.throwIfDisposed();const e=qa().readSync(this.dataId);if("string"===this.dtype)try{return e.map(e=>Ba(e))}catch(e){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return e}async bytes(){this.throwIfDisposed();const e=await qa().read(this.dataId);return"string"===this.dtype?e:new Uint8Array(e.buffer)}dispose(){this.isDisposed||(qa().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return Ka.print(this,e)}clone(){return this.throwIfDisposed(),Ka.clone(this)}toString(e=!1){return Wa(this.dataSync(),this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),Ka.cast(this,e)}variable(e=!0,t,n){return this.throwIfDisposed(),qa().makeVariable(this,e,t,n)}};function Xa(){return Je("Tensor",()=>Ya)}Object.defineProperty(Ya,Symbol.hasInstance,{value:e=>!!e&&null!=e.data&&null!=e.dataSync&&null!=e.throwIfDisposed}),Xa();var Qa=class extends Ya{constructor(e,t,n,r){super(e.shape,e.dtype,e.dataId,r),this.trainable=t,this.name=n}assign(e){if(e.dtype!==this.dtype)throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);if(!fe(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);qa().disposeTensor(this),this.dataId=e.dataId,qa().incRef(this,null)}dispose(){qa().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(Qa,Symbol.hasInstance,{value:e=>e instanceof Ya&&null!=e.assign&&e.assign instanceof Function});var Za,Ja,es,ts,ns,rs,as,ss,is,os,us={};w(us,{assertTypesMatch:()=>ds,getTensorsInContainer:()=>ms,isTensorInList:()=>fs,makeTypesMatch:()=>ps}),(Ja=Za||(Za={})).R0="R0",Ja.R1="R1",Ja.R2="R2",Ja.R3="R3",Ja.R4="R4",Ja.R5="R5",Ja.R6="R6",(ts=es||(es={})).float32="float32",ts.int32="int32",ts.bool="int32",ts.complex64="complex64",(rs=ns||(ns={})).float32="float32",rs.int32="int32",rs.bool="bool",rs.complex64="complex64",(ss=as||(as={})).float32="float32",ss.int32="float32",ss.bool="float32",ss.complex64="complex64",(os=is||(is={})).float32="complex64",os.int32="complex64",os.bool="complex64",os.complex64="complex64";var ls={float32:as,int32:es,bool:ns,complex64:is};function cs(e,t){if("string"===e||"string"===t){if("string"===e&&"string"===t)return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return ls[e][t]}function hs(e){return cs(e,"int32")}function ps(e,t){if(e.dtype===t.dtype)return[e,t];const n=cs(e.dtype,t.dtype);return[e.cast(n),t.cast(n)]}function ds(e,t){ue(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function fs(e,t){return t.some(t=>t.id===e.id)}function ms(e){const t=[];return function e(t,n,r){if(null==t)return;if(t instanceof Ya)return void n.push(t);if(a=t,!Array.isArray(a)&&"object"!=typeof a)return;var a;const s=t;for(const t in s){const a=s[t];r.has(a)||(r.add(a),e(a,n,r))}}(e,t,new Set),t}function gs(e){return null!=e.kernelName}var ys=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(e=>e.name)))}}}dispose(){for(const e in this.registeredVariables)this.registeredVariables[e].dispose()}},bs=class{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new ys}async ready(){if(null!=this.pendingBackendInit)return this.pendingBackendInit.then(()=>{});if(null!=this.backendInstance)return;const e=this.getSortedBackends();for(let t=0;t<e.length;t++){const n=e[t];if(await this.initializeBackend(n).success)return void await this.setBackend(n)}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(null!=this.pendingBackendInit)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(null==this.backendInstance){const{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry)){if(!(e in this.registryFactory))return null;{const{asyncInit:t}=this.initializeBackend(e);if(t)return null}}return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,n=1){return e in this.registryFactory?(ia(e+" backend was already registered. Reusing existing backend factory."),!1):(this.registryFactory[e]={factory:t,priority:n},!0)}async setBackend(e){if(null==this.registryFactory[e])throw new Error(`Backend name '${e}' not found in registry`);if(this.backendName=e,null==this.registry[e]){this.backendInstance=null;const{success:t,asyncInit:n}=this.initializeBackend(e);if(!(n?await t:t))return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new class{constructor(e,t){this.backendTimer=e,this.logger=t,null==t&&(this.logger=new Ua)}profileKernel(e,t,n){let r;const a=()=>{r=n()};let s;const i=La();if(this.backendTimer.timerAvailable())s=this.backendTimer.time(a);else{a();for(const e of r)e.dataSync();s=Promise.resolve({kernelMs:La()-i})}if(Ye().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let t=0;t<r.length;t++){const n=r[t];n.data().then(t=>{za(t,n.dtype,e)})}return{kernelName:e,outputs:r,inputs:t,timeMs:s.then(e=>e.kernelMs),extraInfo:s.then(e=>null!=e.getExtraProfileInfo?e.getExtraProfileInfo():"")}}logKernelProfile(e){const{kernelName:t,outputs:n,timeMs:r,inputs:a,extraInfo:s}=e;n.forEach(e=>{Promise.all([e.data(),r,s]).then(n=>{this.logger.logKernelProfile(t,e,n[0],n[1],a,n[2])})})}}(this.backendInstance),!0}setupRegisteredKernels(){pa(this.backendName).forEach(e=>{null!=e.setupFunc&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){pa(e).forEach(t=>{null!=t.disposeFunc&&t.disposeFunc(this.registry[e])})}initializeBackend(e){const t=this.registryFactory[e];if(null==t)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{const n=t.factory();if(!n||n instanceof Z||"function"!=typeof n.then)return this.registry[e]=n,{success:!0,asyncInit:!1};{const t=++this.pendingBackendInitId,r=n.then(n=>!(t<this.pendingBackendInitId)&&(this.registry[e]=n,this.pendingBackendInit=null,!0)).catch(n=>(t<this.pendingBackendInitId||(this.pendingBackendInit=null,ia(`Initialization of backend ${e} failed`),ia(n.stack||n.message)),!1));return this.pendingBackendInit=r,{success:r,asyncInit:!0}}}catch(t){return ia(`Initialization of backend ${e} failed`),ia(t.stack||t.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(e+" backend not found in registry");this.backendName===e&&null!=this.pendingBackendInit&&this.pendingBackendInitId++,e in this.registry&&(this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e]),delete this.registryFactory[e],this.backendName===e&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(0===Object.keys(this.registryFactory).length)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((e,t)=>this.registryFactory[t].priority-this.registryFactory[e].priority)}initializeBackendsAndReturnBest(){const e=this.getSortedBackends();for(let t=0;t<e.length;t++){const n=e[t],{success:r,asyncInit:a}=this.initializeBackend(n);if(a||r)return{name:n,asyncInit:a}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){const n=this.state.tensorInfo.get(t),r=n.backend,a=this.readSync(t),s=r.refCount(t);r.disposeData(t,!0),n.backend=e,e.move(t,a,n.shape,n.dtype,s),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let n,r=null;if(null==t){if("function"!=typeof e)throw new Error("Please provide a function to tidy()");t=e}else{if("string"!=typeof e&&!(e instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if("function"!=typeof t)throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");r=e}return this.scopedRun(()=>this.startScope(r),()=>this.endScope(n),()=>(n=t(),n instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),n))}scopedRun(e,t,n){e();try{const e=n();return t(),e}catch(e){throw t(),e}}nextTensorId(){return bs.nextTensorId++}nextVariableId(){return bs.nextVariableId++}clone(e){const t=xs.runKernel(fn,{x:e}),n={x:e};return this.addTapeNode(this.state.activeScope.name,n,[t],e=>({x:()=>{const t={x:e},n={dtype:"float32"};return xs.runKernel(St,t,n)}}),[],{}),t}runKernel(e,t,n){if(!(null!=ca(e,this.backendName)))throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:e,inputs:t,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,n){const r=this.backend.numDataIds();let a=0;n.forEach(e=>{a+="complex64"===e.dtype?3:1});const s=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],i=r-t-a-s;if(i>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${i} data ids) after running '${e}'`)}runKernelFunc(e){let t,n=[];const r=this.isTapeOn(),a=this.state.numBytes,s=this.state.numTensors;let i,o;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);const u=gs(e)?e.kernelName:null!=this.state.activeScope?this.state.activeScope.name:"";if(gs(e)){const{kernelName:t,inputs:a,attrs:s}=e,u=ca(t,this.backendName);ue(null!=u,()=>`Cannot find registered kernel '${t}' for backend '${this.backendName}'`),i=()=>{const e=this.backend.numDataIds();o=u.kernelFunc({inputs:a,attrs:s,backend:this.backend});const i=Array.isArray(o)?o:[o];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(t,e,i);const l=i.map(e=>{if(null!=e.rank)return e;const{dataId:t,shape:n,dtype:r}=e;return this.makeTensorFromDataId(t,n,r)});if(r){const e=this.getTensorsForGradient(t,a,l);n=this.saveTensorsForBackwardMode(e)}return l}}else{const{forwardFunc:t}=e,a=e=>{r&&(n=e.map(e=>this.keep(this.clone(e))))};i=()=>{const e=this.backend.numDataIds();o=this.tidy(()=>t(this.backend,a));const n=Array.isArray(o)?o:[o];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(u,e,n),n}}const{inputs:l,attrs:c}=e,h=gs(e)?null:e.backwardsFunc;let p;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{this.ENV.getBool("DEBUG")||this.state.profiling?(p=this.profiler.profileKernel(u,l,()=>i()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(p),t=p.outputs):t=i()}),r&&this.addTapeNode(u,l,t,h,n,c),this.state.profiling&&this.state.activeProfile.kernels.push({name:u,bytesAdded:this.state.numBytes-a,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-s,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(l).map(e=>null!=l[e]?l[e].shape:null),outputShapes:t.map(e=>e.shape),kernelTimeMs:p.timeMs,extraInfo:p.extraInfo}),Array.isArray(o)?t:t[0]}saveTensorsForBackwardMode(e){return e.map(e=>this.keep(this.clone(e)))}getTensorsForGradient(e,t,n){const r=ha(e);if(null!=r){const e=r.inputsToSave||[],a=r.outputsToSave||[];let s;r.saveAllInputs?(ue(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),s=Object.keys(t).map(e=>t[e])):s=e.map(e=>t[e]);const i=n.filter((e,t)=>a[t]);return s.concat(i)}return[]}makeTensor(e,t,n,r){if(null==e)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",r=r||this.backend;let a=e;"string"===n&&Fe(e[0])&&(a=e.map(e=>$a(e)));const s=r.write(a,t,n),i=new Ya(t,n,s,this.nextTensorId());if(this.trackTensor(i,r),"string"===n){const e=this.state.tensorInfo.get(s),t=_e(a);this.state.numBytes+=t-e.bytes,e.bytes=t}return i}makeTensorFromDataId(e,t,n,r){const a=new Ya(t,n=n||"float32",e,this.nextTensorId());return this.trackTensor(a,r),a}makeVariable(e,t=!0,n,r){n=n||this.nextVariableId().toString(),null!=r&&r!==e.dtype&&(e=e.cast(r));const a=new Qa(e,t,n,this.nextTensorId());if(null!=this.state.registeredVariables[a.name])throw new Error(`Variable with name ${a.name} was already registered`);return this.state.registeredVariables[a.name]=a,this.incRef(a,this.backend),a}trackTensor(e,t){this.state.numTensors++,"string"===e.dtype&&this.state.numStringTensors++;let n=0;"complex64"!==e.dtype&&"string"!==e.dtype&&(n=e.size*Re(e.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(e.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:n})),e instanceof Qa||this.track(e)}incRef(e,t){this.trackTensor(e,t),this.backend.incRef(e.dataId)}removeDataId(e,t){this.state.tensorInfo.has(e)&&this.state.tensorInfo.get(e).backend===t&&(this.state.tensorInfo.delete(e),this.state.numDataBuffers--)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;const t=this.state.tensorInfo.get(e.dataId);if(this.state.numTensors--,"string"===e.dtype&&(this.state.numStringTensors--,this.state.numBytes-=t.bytes),"complex64"!==e.dtype&&"string"!==e.dtype){const t=e.size*Re(e.dtype);this.state.numBytes-=t}t.backend.disposeData(e.dataId)&&this.removeDataId(e.dataId,t.backend)}disposeVariables(){for(const e in this.state.registeredVariables){const t=this.state.registeredVariables[e];this.disposeVariable(t)}}disposeVariable(e){this.disposeTensor(e),null!=this.state.registeredVariables[e.name]&&delete this.state.registeredVariables[e.name]}memory(){const e=this.backend.memory();return e.numTensors=this.state.numTensors,e.numDataBuffers=this.state.numDataBuffers,e.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(e.unreliable=!0,null==e.reasons&&(e.reasons=[]),e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),e}async profile(e){this.state.profiling=!0;const t=this.state.numBytes,n=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await e(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(e=>e.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-n;for(const e of this.state.activeProfile.kernels)e.kernelTimeMs=await e.kernelTimeMs,e.extraInfo=await e.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&0===this.state.kernelDepth}addTapeNode(e,t,n,r,a,s){const i={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:n,saved:a},o=ha(e);null!=o&&(r=o.gradFunc),null!=r&&(i.gradient=e=>(e=e.map((e,t)=>{if(null==e){const e=n[t],r=Ue(e.size,e.dtype);return this.makeTensor(r,e.shape,e.dtype)}return e}),r(e.length>1?e:e[0],a,s))),this.state.activeTape.push(i)}keep(e){return e.kept=!0,e}startTape(){0===this.state.gradientDepth&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){const t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){const t=ms(e),n=new Set(t.map(e=>e.id));for(let e=0;e<this.state.activeScope.track.length;e++){const t=this.state.activeScope.track[e];t.kept||n.has(t.id)||t.dispose()}const r=this.state.scopeStack.pop();this.state.activeScope=0===this.state.scopeStack.length?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(e=>{e.kept||e.scopeId!==r.id||this.track(e)})}gradients(e,t,n,r=!1){if(ue(t.length>0,()=>"gradients() received an empty list of xs."),null!=n&&"float32"!==n.dtype)throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);const a=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));ue(a instanceof Ya,()=>"The result y returned by f() must be a tensor.");const s=function(e,t,n){const r={},a={};for(let e=0;e<t.length;e++)r[t[e].id]=!0;for(let n=0;n<e.length;n++){const s=e[n],i=s.inputs;for(const e in i){const n=i[e];let o=!1;for(let e=0;e<t.length;e++)if(r[n.id]){s.outputs.forEach(e=>r[e.id]=!0),o=!0,a[s.id]=!0;break}if(o)break}}const s={};s[n.id]=!0;const i={};for(let t=e.length-1;t>=0;t--){const n=e[t],r=n.inputs;for(let e=0;e<n.outputs.length;e++)if(s[n.outputs[e].id]){for(const e in r)s[r[e].id]=!0,i[n.id]=!0;break}}const o=[];for(let t=0;t<e.length;t++){const n=e[t];if(a[n.id]&&i[n.id]){const e={};for(const t in n.inputs){const a=n.inputs[t];r[a.id]&&(e[t]=a)}const t=Object.assign({},n);t.inputs=e,t.outputs=n.outputs,o.push(t)}}return o}(this.state.activeTape,t,a);if(!r&&0===s.length&&t.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{const e={};e[a.id]=null==n?function(e){const t=ze(pe(e),"float32");return xs.makeTensor(t,e,"float32")}(a.shape):n,function(e,t,n,r){for(let a=t.length-1;a>=0;a--){const s=t[a],i=[];if(s.outputs.forEach(t=>{const n=e[t.id];null!=n?i.push(n):i.push(null)}),null==s.gradient)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);const o=s.gradient(i);for(const t in s.inputs){if(!(t in o))throw new Error(`Cannot backprop through input ${t}. Available gradients found: ${Object.keys(o)}.`);const a=n(()=>o[t]());if("float32"!==a.dtype)throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${t} must have 'float32' dtype, but has '${a.dtype}'`);const i=s.inputs[t];if(!fe(a.shape,i.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${t}' has shape '${a.shape}', which does not match the shape of the input '${i.shape}'`);if(null==e[i.id])e[i.id]=a;else{const t=e[i.id];e[i.id]=r(t,a),t.dispose()}}}}(e,s,e=>this.tidy(e),ws);const r=t.map(t=>e[t.id]);return 0===this.state.gradientDepth&&(this.state.activeTape.forEach(e=>{for(const t of e.saved)t.dispose()}),this.state.activeTape=null),{value:a,grads:r}})}customGrad(e){return ue(Le(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{let n;ue(t.every(e=>e instanceof Ya),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");const r={};t.forEach((e,t)=>{r[t]=e});return this.runKernelFunc({forwardFunc:(r,a)=>(n=e(...t,a),ue(n.value instanceof Ya,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),ue(Le(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),backwardsFunc:(e,r)=>{const a=n.gradFunc(e,r),s=Array.isArray(a)?a:[a];ue(s.length===t.length,()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."),ue(s.every(e=>e instanceof Ya),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");const i={};return s.forEach((e,t)=>{i[t]=()=>e}),i},inputs:r})}}readSync(e){return this.state.tensorInfo.get(e).backend.readSync(e)}read(e){return this.state.tensorInfo.get(e).backend.read(e)}readToGPU(e,t){return this.state.tensorInfo.get(e).backend.readToGPU(e,t)}async time(e){const t=La(),n=await this.backend.time(e);return n.wallMs=La()-t,n}track(e){return null!=this.state.activeScope&&(e.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(e)),e}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new ys;for(const e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};function vs(){const e=Ze();if(null==e._tfengine){const t=new qe(e);e._tfengine=new bs(t)}var t;return t=e._tfengine.ENV,Qe=t,qa=()=>e._tfengine,e._tfengine}bs.nextTensorId=0,bs.nextVariableId=0;var xs=vs();function ws(e,t){const n={a:e,b:t};return xs.runKernel(rt,n)}var ks,Ss={};function Is(e){ks=e}function As(e){if(void 0!==ks)return ks;if(e||"undefined"!=typeof navigator&&null!=navigator){if(e||(e=navigator),"ReactNative"===e.product)return!0;const t=e.userAgent||e.vendor||("undefined"!=typeof window?window.opera:"");if(!t){const t=e;return t.userAgentData&&t.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function Es(){return"undefined"!=typeof window&&null!=window.document||"undefined"!=typeof WorkerGlobalScope}w(Ss,{isBrowser:()=>Es,isMobile:()=>As,mockIsMobile:()=>Is});var Ns=Ye();function Cs(e,t){let n=e;if(Te(e))return"string"===t?[]:[e.length];if(!Array.isArray(e))return[];const r=[];for(;Array.isArray(n)||Te(n)&&"string"!==t;)r.push(n.length),n=n[0];return Array.isArray(e)&&Ye().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&function e(t,n,r){if(r=r||[],!Array.isArray(t)&&!Te(t))return void ue(0===n.length,()=>`Element arr[${r.join("][")}] is a primitive, but should be an array/TypedArray of ${n[0]} elements`);ue(n.length>0,()=>`Element arr[${r.join("][")}] should be a primitive, but is an array of ${t.length} elements`),ue(t.length===n[0],()=>`Element arr[${r.join("][")}] should have ${n[0]} elements, but has ${t.length} elements`);const a=n.slice(1);for(let n=0;n<t.length;++n)e(t[n],a,r.concat(n))}(e,r,[]),r}function Ts(e,t,n,r){if("string_or_numeric"!==e){if(null==e)throw new Error("Expected dtype cannot be null.");if("numeric"!==e&&e!==t||"numeric"===e&&"string"===t)throw new Error(`Argument '${n}' passed to '${r}' must be ${e} tensor, but got ${t} tensor`)}}function Rs(e,t,n,r="numeric"){if(e instanceof Ya)return Ts(r,e.dtype,t,n),e;let a=Oe(e);if("string"!==a&&["bool","int32","float32"].indexOf(r)>=0&&(a=r),Ts(r,a,t,n),null==e||!Te(e)&&!Array.isArray(e)&&"number"!=typeof e&&"boolean"!=typeof e&&"string"!=typeof e){const r=null==e?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${r}'`)}const s=Cs(e,a);Te(e)||Array.isArray(e)||(e=[e]);const i="string"!==a?Oa(e,a):he(e,[],!0);return xs.makeTensor(i,s,a)}function _s(e,t,n,r="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);return e.map((e,a)=>Rs(e,`${t}[${a}]`,n,r))}Ns.registerFlag("DEBUG",()=>!1,e=>{e&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")}),Ns.registerFlag("IS_BROWSER",()=>Es()),Ns.registerFlag("IS_NODE",()=>void 0!==r&&void 0!==r.versions&&void 0!==r.versions.node),Ns.registerFlag("IS_CHROME",()=>"undefined"!=typeof navigator&&null!=navigator&&null!=navigator.userAgent&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor)),Ns.registerFlag("PROD",()=>!1),Ns.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>Ns.getBool("DEBUG")),Ns.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0),Ns.registerFlag("IS_TEST",()=>!1),Ns.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>!0),Ns.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1);var Fs="__op";function Ms(e){const t=Object.keys(e);if(1!==t.length)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let n=t[0];const r=e[n];n.endsWith("_")&&(n=n.substring(0,n.length-1)),n+=Fs;const a=(...e)=>{xs.startScope(n);try{const t=r(...e);return He(t)&&console.error("Cannot return a Promise inside of tidy."),xs.endScope(t),t}catch(e){throw xs.endScope(null),e}};return Object.defineProperty(a,"name",{value:n,configurable:!0}),a}var Ds=Ms({complex_:function(e,t){const n=Rs(e,"real","complex"),r=Rs(t,"imag","complex");le(n.shape,r.shape,`real and imag shapes, ${n.shape} and ${r.shape}, must match in call to tf.complex().`);const a={real:n,imag:r};return xs.runKernel(Et,a)}});function Os(e,t,n,r){if(null==r&&(r=Oe(e)),"complex64"===r)throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(!Te(e)&&!Array.isArray(e)&&"number"!=typeof e&&"boolean"!=typeof e&&"string"!=typeof e)throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(null!=t){Ve(t);const e=pe(t),r=pe(n);ue(e===r,()=>`Based on the provided shape, [${t}], the tensor should have ${e} values but has ${r}`);for(let e=0;e<n.length;++e){const r=n[e],a=e!==n.length-1||r!==pe(t.slice(e));ue(n[e]===t[e]||!a,()=>`Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)}}return Te(e)||Array.isArray(e)||(e=[e]),t=t||n,e="string"!==r?Oa(e,r):he(e,[],!0),xs.makeTensor(e,t,r)}function Ls(e,t,n){return Os(e,t,Cs(e,n),n)}var Ps={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8};async function $s(e,t){const n=[],r=[],a=Array.isArray(e)?e.map(e=>e.name):Object.keys(e);for(let s=0;s<a.length;++s){const i=a[s],o=Array.isArray(e)?e[s].tensor:e[i];if("float32"!==o.dtype&&"int32"!==o.dtype&&"bool"!==o.dtype&&"string"!==o.dtype&&"complex64"!==o.dtype)throw new Error(`Unsupported dtype in weight '${i}': ${o.dtype}`);const u={name:i,shape:o.shape,dtype:o.dtype};if("string"===o.dtype){const e=new Promise(async e=>{const t=await o.bytes(),n=t.reduce((e,t)=>e+t.length,0)+4*t.length,r=new Uint8Array(n);let a=0;for(let e=0;e<t.length;e++){const n=t[e],s=new Uint8Array(new Uint32Array([n.length]).buffer);r.set(s,a),a+=4,r.set(n,a),a+=n.length}e(r)});r.push(e)}else r.push(o.data());null!=t&&(u.group=t),n.push(u)}return{data:zs(await Promise.all(r)),specs:n}}function Bs(e,t){const n={};let r,a=0;for(const s of t){const t=s.name,i=s.dtype,o=s.shape,u=pe(o);let l;if("quantization"in s){const n=s.quantization;if("uint8"===n.dtype||"uint16"===n.dtype){if(!("min"in n)||!("scale"in n))throw new Error(`Weight ${s.name} with quantization ${n.dtype} doesn't have corresponding metadata min and scale.`)}else{if("float16"!==n.dtype)throw new Error(`Weight ${s.name} has unknown quantization dtype ${n.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);if("float32"!==i)throw new Error(`Weight ${s.name} is quantized with ${n.dtype} which only supports weights of type float32 not ${i}.`)}const o=Ps[n.dtype],c=e.slice(a,a+u*o),h="uint8"===n.dtype?new Uint8Array(c):new Uint16Array(c);if("float32"===i)if("uint8"===n.dtype||"uint16"===n.dtype){l=new Float32Array(h.length);for(let e=0;e<h.length;e++){const t=h[e];l[e]=t*n.scale+n.min}}else{if("float16"!==n.dtype)throw new Error(`Unsupported quantization type ${n.dtype} for weight type float32.`);void 0===r&&(r=Ks()),l=r(h)}else{if("int32"!==i)throw new Error(`Unsupported dtype in weight '${t}': ${i}`);if("uint8"!==n.dtype&&"uint16"!==n.dtype)throw new Error(`Unsupported quantization type ${n.dtype} for weight type int32.`);l=new Int32Array(h.length);for(let e=0;e<h.length;e++){const t=h[e];l[e]=Math.round(t*n.scale+n.min)}}a+=u*o}else if("string"===i){const t=pe(s.shape);l=[];for(let n=0;n<t;n++){const t=new Uint32Array(e.slice(a,a+4))[0];a+=4;const n=new Uint8Array(e.slice(a,a+t));l.push(n),a+=t}}else{const r=Ps[i],s=e.slice(a,a+u*r);if("float32"===i)l=new Float32Array(s);else if("int32"===i)l=new Int32Array(s);else if("bool"===i)l=new Uint8Array(s);else{if("complex64"!==i)throw new Error(`Unsupported dtype in weight '${t}': ${i}`);{l=new Float32Array(s);const e=new Float32Array(l.length/2),r=new Float32Array(l.length/2);for(let t=0;t<e.length;t++)e[t]=l[2*t],r[t]=l[2*t+1];const a=Ls(e,o,"float32"),i=Ls(r,o,"float32");n[t]=Ds(a,i),a.dispose(),i.dispose()}}a+=u*r}"complex64"!==i&&(n[t]=Ls(l,o,i))}return n}function zs(e){if(null===e)throw new Error("Invalid input value: "+JSON.stringify(e));let t=0;const n=[];e.forEach(e=>{if(t+=e.byteLength,n.push(e.byteLength===e.buffer.byteLength?e:new e.constructor(e)),!(e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array))throw new Error("Unsupported TypedArray subtype: "+e.constructor.name)});const r=new Uint8Array(t);let a=0;return n.forEach(e=>{r.set(new Uint8Array(e.buffer),a),a+=e.byteLength}),r.buffer}var Us=void 0!==i&&("undefined"==typeof Blob||"undefined"==typeof atob||"undefined"==typeof btoa);function Ws(e){return Us?i.byteLength(e):new Blob([e]).size}function Vs(e){if(1===e.length)return e[0];let t=0;e.forEach(e=>{t+=e.byteLength});const n=new Uint8Array(t);let r=0;return e.forEach(e=>{n.set(new Uint8Array(e),r),r+=e.byteLength}),n.buffer}function js(e){for(e=e.trim();e.endsWith("/");)e=e.slice(0,e.length-1);const t=e.split("/");return t[t.length-1]}function Gs(e,t){const n={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:t};return null!=e.signature&&(n.signature=e.signature),null!=e.userDefinedMetadata&&(n.userDefinedMetadata=e.userDefinedMetadata),null!=e.modelInitializer&&(n.modelInitializer=e.modelInitializer),null!=e.trainingConfig&&(n.trainingConfig=e.trainingConfig),n}async function Hs(e,t){const n={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy};if(null!=e.trainingConfig&&(n.trainingConfig=e.trainingConfig),null!=e.weightsManifest){const[r,a]=await t(e.weightsManifest);n.weightSpecs=r,n.weightData=a}return null!=e.signature&&(n.signature=e.signature),null!=e.userDefinedMetadata&&(n.userDefinedMetadata=e.userDefinedMetadata),null!=e.modelInitializer&&(n.modelInitializer=e.modelInitializer),n}function qs(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:null==e.modelTopology?0:Ws(JSON.stringify(e.modelTopology)),weightSpecsBytes:null==e.weightSpecs?0:Ws(JSON.stringify(e.weightSpecs)),weightDataBytes:null==e.weightData?0:e.weightData.byteLength}}function Ks(){const e=function(){const e=e=>{let t=e<<13,n=0;for(;0==(8388608&t);)n-=8388608,t<<=1;return t&=-8388609,n+=947912704,t|n},t=new Uint32Array(2048);t[0]=0;for(let n=1;n<1024;n++)t[n]=e(n);for(let e=1024;e<2048;e++)t[e]=939524096+(e-1024<<13);return t}(),t=function(){const e=new Uint32Array(64);e[0]=0,e[31]=1199570944,e[32]=2147483648,e[63]=3347054592;for(let t=1;t<31;t++)e[t]=t<<23;for(let t=33;t<63;t++)e[t]=2147483648+(t-32<<23);return e}(),n=function(){const e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}();return r=>{const a=new ArrayBuffer(4*r.length),s=new Uint32Array(a);for(let a=0;a<r.length;a++){const i=r[a],o=e[n[i>>10]+(1023&i)]+t[i>>10];s[a]=o}return new Float32Array(a)}}var Ys=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return null==Ys.instance&&(Ys.instance=new Ys),Ys.instance}static registerSaveRouter(e){Ys.getInstance().saveRouters.push(e)}static registerLoadRouter(e){Ys.getInstance().loadRouters.push(e)}static getSaveHandlers(e){return Ys.getHandlers(e,"save")}static getLoadHandlers(e,t){return Ys.getHandlers(e,"load",t)}static getHandlers(e,t,n){const r=[];return("load"===t?Ys.getInstance().loadRouters:Ys.getInstance().saveRouters).forEach(t=>{const a=t(e,n);null!==a&&r.push(a)}),r}},Xs=e=>Ys.registerSaveRouter(e),Qs=e=>Ys.registerLoadRouter(e),Zs=e=>Ys.getSaveHandlers(e),Js=(e,t)=>Ys.getLoadHandlers(e,t);function ei(){if(!Ye().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");const e="undefined"==typeof window?self:window,t=e.indexedDB||e.mozIndexedDB||e.webkitIndexedDB||e.msIndexedDB||e.shimIndexedDB;if(null==t)throw new Error("The current browser does not appear to support IndexedDB.");return t}function ti(e){const t=e.result;t.createObjectStore("models_store",{keyPath:"modelPath"}),t.createObjectStore("model_info_store",{keyPath:"modelPath"})}var ni=class{constructor(e){if(this.indexedDB=ei(),null==e||!e)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=e}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,e)}async load(){return this.databaseAction(this.modelPath)}databaseAction(e,t){return new Promise((e,n)=>{const r=this.indexedDB.open("tensorflowjs",1);r.onupgradeneeded=()=>ti(r),r.onsuccess=()=>{const a=r.result;if(null==t){const t=a.transaction("models_store","readonly"),r=t.objectStore("models_store").get(this.modelPath);r.onsuccess=()=>{if(null==r.result)return a.close(),n(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));e(r.result.modelArtifacts)},r.onerror=e=>(a.close(),n(r.error)),t.oncomplete=()=>a.close()}else{const r=qs(t),s=a.transaction("model_info_store","readwrite");let i=s.objectStore("model_info_store");const o=i.put({modelPath:this.modelPath,modelArtifactsInfo:r});let u;o.onsuccess=()=>{u=a.transaction("models_store","readwrite");const o=u.objectStore("models_store").put({modelPath:this.modelPath,modelArtifacts:t,modelArtifactsInfo:r});o.onsuccess=()=>e({modelArtifactsInfo:r}),o.onerror=e=>{i=s.objectStore("model_info_store");const t=i.delete(this.modelPath);t.onsuccess=()=>(a.close(),n(o.error)),t.onerror=e=>(a.close(),n(o.error))}},o.onerror=e=>(a.close(),n(o.error)),s.oncomplete=()=>{null==u?a.close():u.oncomplete=()=>a.close()}}},r.onerror=e=>n(r.error)})}};ni.URL_SCHEME="indexeddb://";var ri=e=>{return Ye().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(ni.URL_SCHEME)?(t=e.slice(ni.URL_SCHEME.length),new ni(t)):null;var t};Ys.registerSaveRouter(ri),Ys.registerLoadRouter(ri);var ai="tensorflowjs_models",si="info",ii="model_topology",oi="weight_specs",ui="weight_data",li="model_metadata";function ci(e){return{info:[ai,e,si].join("/"),topology:[ai,e,ii].join("/"),weightSpecs:[ai,e,oi].join("/"),weightData:[ai,e,ui].join("/"),modelMetadata:[ai,e,li].join("/")}}function hi(e){for(const t of Object.values(e))window.localStorage.removeItem(t)}function pi(e){const t=e.split("/");if(t.length<3)throw new Error("Invalid key format: "+e);return t.slice(1,t.length-1).join("/")}var di=class{constructor(e){if(!Ye().getBool("IS_BROWSER")||"undefined"==typeof window||void 0===window.localStorage)throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,null==e||!e)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=e,this.keys=ci(this.modelPath)}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{const t=JSON.stringify(e.modelTopology),n=JSON.stringify(e.weightSpecs),r=qs(e);try{this.LS.setItem(this.keys.info,JSON.stringify(r)),this.LS.setItem(this.keys.topology,t),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,function(e){if(Us)return i.from(e).toString("base64");const t=new Uint8Array(e);let n="";for(let e=0,r=t.length;e<r;e++)n+=String.fromCharCode(t[e]);return btoa(n)}(e.weightData));const a={format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,signature:null!=e.signature?e.signature:void 0,userDefinedMetadata:null!=e.userDefinedMetadata?e.userDefinedMetadata:void 0,modelInitializer:null!=e.modelInitializer?e.modelInitializer:void 0,trainingConfig:null!=e.trainingConfig?e.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(a)),{modelArtifactsInfo:r}}catch(e){throw hi(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${r.modelTopologyBytes}, weightSpecsBytes=${r.weightSpecsBytes}, weightDataBytes=${r.weightDataBytes}.`)}}}async load(){const e=JSON.parse(this.LS.getItem(this.keys.info));if(null==e)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if("JSON"!==e.modelTopologyType)throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");const t={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(null==n)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);t.modelTopology=n;const r=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(null==r)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);t.weightSpecs=r;const a=this.LS.getItem(this.keys.modelMetadata);if(null!=a){const e=JSON.parse(a);t.format=e.format,t.generatedBy=e.generatedBy,t.convertedBy=e.convertedBy,null!=e.signature&&(t.signature=e.signature),null!=e.userDefinedMetadata&&(t.userDefinedMetadata=e.userDefinedMetadata),null!=e.modelInitializer&&(t.modelInitializer=e.modelInitializer),null!=e.trainingConfig&&(t.trainingConfig=e.trainingConfig)}const s=this.LS.getItem(this.keys.weightData);if(null==s)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return t.weightData=function(e){if(Us){const t=i.from(e,"base64");return t.buffer.slice(t.byteOffset,t.byteOffset+t.byteLength)}const t=atob(e),n=new Uint8Array(t.length);for(let e=0;e<t.length;++e)n.set([t.charCodeAt(e)],e);return n.buffer}(s),t}};di.URL_SCHEME="localstorage://";var fi=e=>{return Ye().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(di.URL_SCHEME)?(t=e.slice(di.URL_SCHEME.length),new di(t)):null;var t};Ys.registerSaveRouter(fi),Ys.registerLoadRouter(fi);var mi=class{constructor(){this.managers={}}static getInstance(){return null==mi.instance&&(mi.instance=new mi),mi.instance}static registerManager(e,t){ue(null!=e,()=>"scheme must not be undefined or null."),e.endsWith("://")&&(e=e.slice(0,e.indexOf("://"))),ue(e.length>0,()=>"scheme must not be an empty string.");const n=mi.getInstance();ue(null==n.managers[e],()=>`A model store manager is already registered for scheme '${e}'.`),n.managers[e]=t}static getManager(e){const t=this.getInstance().managers[e];if(null==t)throw new Error(`Cannot find model manager for scheme '${e}'`);return t}static getSchemes(){return Object.keys(this.getInstance().managers)}};function gi(e){if(-1===e.indexOf("://"))throw new Error("The url string provided does not contain a scheme. Supported schemes are: "+mi.getSchemes().join(","));return{scheme:e.split("://")[0],path:e.split("://")[1]}}async function yi(e,t,n=!1){ue(e!==t,()=>`Old path and new path are the same: '${e}'`);const r=Ys.getLoadHandlers(e);ue(r.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),ue(r.length<2,()=>`Copying failed because more than one (${r.length}) load handlers for source URL ${e}.`);const a=r[0],s=Ys.getSaveHandlers(t);ue(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),ue(s.length<2,()=>`Copying failed because more than one (${r.length}) save handlers for destination URL ${t}.`);const i=s[0],o=gi(e).scheme,u=gi(e).path,l=o===gi(e).scheme,c=await a.load();n&&l&&await mi.getManager(o).removeModel(u);const h=await i.save(c);return n&&!l&&await mi.getManager(o).removeModel(u),h.modelArtifactsInfo}async function bi(){const e=mi.getSchemes(),t={};for(const n of e){const e=await mi.getManager(n).listModels();for(const r in e){t[n+"://"+r]=e[r]}}return t}async function vi(e){const t=gi(e);return mi.getManager(t.scheme).removeModel(t.path)}async function xi(e,t){return yi(e,t,!1)}async function wi(e,t){return yi(e,t,!0)}if(Ye().get("IS_BROWSER")){Ye().setPlatform("browser",new class{fetch(e,t){return fetch(e,t)}now(){return performance.now()}encode(e,t){if("utf-8"!==t&&"utf8"!==t)throw new Error("Browser's encoder only supports utf-8, but got "+t);return null==this.textEncoder&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(e)}decode(e,t){return new TextDecoder(t).decode(e)}});try{mi.registerManager(di.URL_SCHEME,new class{constructor(){ue(Ye().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),ue("undefined"==typeof window||void 0!==window.localStorage,()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){const e={},t=ai+"/",n="/"+si;for(let r=0;r<this.LS.length;++r){const a=this.LS.key(r);if(a.startsWith(t)&&a.endsWith(n)){e[pi(a)]=JSON.parse(this.LS.getItem(a))}}return e}async removeModel(e){var t;const n=ci(e=(t=e).startsWith(di.URL_SCHEME)?t.slice(di.URL_SCHEME.length):t);if(null==this.LS.getItem(n.info))throw new Error(`Cannot find model at path '${e}'`);const r=JSON.parse(this.LS.getItem(n.info));return hi(n),r}})}catch(e){}try{mi.registerManager(ni.URL_SCHEME,new class{constructor(){this.indexedDB=ei()}async listModels(){return new Promise((e,t)=>{const n=this.indexedDB.open("tensorflowjs",1);n.onupgradeneeded=()=>ti(n),n.onsuccess=()=>{const r=n.result,a=r.transaction("model_info_store","readonly"),s=a.objectStore("model_info_store").getAll();s.onsuccess=()=>{const t={};for(const e of s.result)t[e.modelPath]=e.modelArtifactsInfo;e(t)},s.onerror=e=>(r.close(),t(s.error)),a.oncomplete=()=>r.close()},n.onerror=e=>t(n.error)})}async removeModel(e){var t;return e=(t=e).startsWith(ni.URL_SCHEME)?t.slice(ni.URL_SCHEME.length):t,new Promise((t,n)=>{const r=this.indexedDB.open("tensorflowjs",1);r.onupgradeneeded=()=>ti(r),r.onsuccess=()=>{const a=r.result,s=a.transaction("model_info_store","readwrite"),i=s.objectStore("model_info_store"),o=i.get(e);let u;o.onsuccess=()=>{if(null==o.result)return a.close(),n(new Error(`Cannot find model with path '${e}' in IndexedDB.`));{const r=i.delete(e),s=()=>{u=a.transaction("models_store","readwrite");const r=u.objectStore("models_store").delete(e);r.onsuccess=()=>t(o.result.modelArtifactsInfo),r.onerror=e=>n(o.error)};r.onsuccess=s,r.onerror=e=>(s(),a.close(),n(o.error))}},o.onerror=e=>(a.close(),n(o.error)),s.oncomplete=()=>{null==u?a.close():u.oncomplete=()=>a.close()}},r.onerror=e=>n(r.error)})}})}catch(e){}}var ki,Si=()=>I();function Ii(e,t="float32",n){return t=t||"float32",Ve(e),new Ha(e,t,n)}Ye().get("IS_NODE")&&Ye().setPlatform("node",new class{constructor(){this.util=A(),this.textEncoder=new this.util.TextEncoder}fetch(e,t){return null!=Ye().global.fetch?Ye().global.fetch(e,t):(null==ki&&(ki=Si()),ki(e,t))}now(){const e=r.hrtime();return 1e3*e[0]+e[1]/1e6}encode(e,t){if("utf-8"!==t&&"utf8"!==t)throw new Error("Node built-in encoder only supports utf-8, but got "+t);return this.textEncoder.encode(e)}decode(e,t){return 0===e.length?"":new this.util.TextDecoder(t).decode(e)}});var Ai=Ms({cast_:function(e,t){const n=Rs(e,"x","cast");if(!Ne(t))throw new Error("Failed to cast to unknown dtype "+t);if("string"===t&&"string"!==n.dtype||"string"!==t&&"string"===n.dtype)throw new Error("Only strings can be casted to strings");const r={x:n},a={dtype:t};return xs.runKernel(St,r,a)}});var Ei=Ms({clone_:function(e){const t={x:Rs(e,"x","clone","string_or_numeric")};return xs.runKernel(fn,t)}});function Ni(e,t=!1){console.log(e.toString(t))}vs(),Ka={buffer:Ii,cast:Ai,clone:Ei,print:Ni};var Ci={};w(Ci,{browserFiles:()=>_i,browserHTTPRequest:()=>zi,concatenateArrayBuffers:()=>Vs,copyModel:()=>xi,decodeWeights:()=>Bs,encodeWeights:()=>$s,fromMemory:()=>Wi,getLoadHandlers:()=>Js,getModelArtifactsForJSON:()=>Hs,getModelArtifactsInfoForJSON:()=>qs,getSaveHandlers:()=>Zs,http:()=>Bi,isHTTPScheme:()=>Pi,listModels:()=>bi,loadWeights:()=>Di,moveModel:()=>wi,registerLoadRouter:()=>Qs,registerSaveRouter:()=>Xs,removeModel:()=>vi,weightsLoaderFactory:()=>Oi,withSaveHandler:()=>Vi});function Ti(e){return new Promise(e=>setTimeout(e)).then(e)}var Ri=class{constructor(e){if(!Ye().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");e.startsWith(Ri.URL_SCHEME)&&(e=e.slice(Ri.URL_SCHEME.length)),null!=e&&0!==e.length||(e="model"),this.modelJsonFileName=e+".json",this.weightDataFileName=e+".weights.bin"}async save(e){if("undefined"==typeof document)throw new Error("Browser downloads are not supported in this environment since `document` is not present");const t=window.URL.createObjectURL(new Blob([e.weightData],{type:"application/octet-stream"}));if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{const n=Gs(e,[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}]),r=window.URL.createObjectURL(new Blob([JSON.stringify(n)],{type:"application/json"})),a=null==this.modelJsonAnchor?document.createElement("a"):this.modelJsonAnchor;if(a.download=this.modelJsonFileName,a.href=r,await Ti(()=>a.dispatchEvent(new MouseEvent("click"))),null!=e.weightData){const e=null==this.weightDataAnchor?document.createElement("a"):this.weightDataAnchor;e.download=this.weightDataFileName,e.href=t,await Ti(()=>e.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:qs(e)}}}};Ri.URL_SCHEME="downloads://";function _i(e){return new class{constructor(e){if(null==e||e.length<1)throw new Error("When calling browserFiles, at least 1 file is required, but received "+e);this.jsonFile=e[0],this.weightsFiles=e.slice(1)}async load(){return new Promise((e,t)=>{const n=new FileReader;n.onload=n=>{const r=JSON.parse(n.target.result),a=r.modelTopology;if(null==a)return void t(new Error("modelTopology field is missing from file "+this.jsonFile.name));if(null==r.weightsManifest)return void t(new Error("weightManifest field is missing from file "+this.jsonFile.name));if(0===this.weightsFiles.length)return void e({modelTopology:a});const s=Hs(r,e=>this.loadWeights(e));e(s)},n.onerror=e=>t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(e){const t=[],n=[];for(const r of e)t.push(...r.weights),n.push(...r.paths);const r=this.checkManifestAndWeightFiles(e),a=n.map(e=>this.loadWeightsFile(e,r[e]));return Promise.all(a).then(e=>[t,Vs(e)])}loadWeightsFile(e,t){return new Promise((n,r)=>{const a=new FileReader;a.onload=e=>{const t=e.target.result;n(t)},a.onerror=t=>r(`Failed to weights data from file of path '${e}'.`),a.readAsArrayBuffer(t)})}checkManifestAndWeightFiles(e){const t=[],n=this.weightsFiles.map(e=>js(e.name)),r={};for(const a of e)a.paths.forEach(e=>{const a=js(e);if(-1!==t.indexOf(a))throw new Error(`Duplicate file basename found in weights manifest: '${a}'`);if(t.push(a),-1===n.indexOf(a))throw new Error(`Weight file with basename '${a}' is not provided.`);r[e]=this.weightsFiles[n.indexOf(a)]});if(t.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return r}}(e)}function Fi(e,t,n,r){var a,s,i;ue(null!=(a=e)&&Array.isArray(a)&&a.length>0,()=>"promises must be a none empty array"),i=r=null==r?1:r,ue((s=n=null==n?0:n)>=0&&s<=1,()=>"Progress fraction must be in range [0, 1], but got startFraction "+s),ue(i>=0&&i<=1,()=>"Progress fraction must be in range [0, 1], but got endFraction "+i),ue(i>=s,()=>`startFraction must be no more than endFraction, but got startFraction ${s} and endFraction ${i}`);let o=0;return Promise.all(e.map(a=>(a.then(a=>{const s=n+ ++o/e.length*(r-n);return t(s),a}),a)))}async function Mi(e,t){null==t&&(t={});const n=null==t.fetchFunc?Ye().platform.fetch:t.fetchFunc,r=e.map(e=>n(e,t.requestInit,{isBinary:!0})),a=(null==t.onProgress?await Promise.all(r):await Fi(r,t.onProgress,0,.5)).map(e=>e.arrayBuffer());return null==t.onProgress?await Promise.all(a):await Fi(a,t.onProgress,.5,1)}async function Di(e,t="",n,r){return Oi(e=>Mi(e,{requestInit:r}))(e,t,n)}function Oi(e){return async(t,n="",r)=>{const a=t.map(()=>!1),s={},i=null!=r?r.map(()=>!1):[],o=[];if(t.forEach((e,t)=>{let n=0;e.weights.forEach(e=>{const u="quantization"in e?e.quantization.dtype:e.dtype,l=Ps[u]*pe(e.shape),c=()=>{a[t]=!0,null==s[t]&&(s[t]=[]),s[t].push({manifestEntry:e,groupOffset:n,sizeBytes:l})};null!=r?r.forEach((t,n)=>{t===e.name&&(c(),i[n]=!0)}):c(),o.push(e.name),n+=l})}),!i.every(e=>e)){const e=r.filter((e,t)=>!i[t]);throw new Error(`Could not find weights in manifest with names: ${e.join(", ")}. \nManifest JSON has weights with names: ${o.join(", ")}.`)}const u=a.reduce((e,t,n)=>(t&&e.push(n),e),[]),l=[];u.forEach(e=>{t[e].paths.forEach(e=>{const t=n+(n.endsWith("/")?"":"/")+e;l.push(t)})});const c=await e(l),h={};let p=0;return u.forEach(e=>{const n=t[e].paths.length;let r=0;for(let e=0;e<n;e++)r+=c[p+e].byteLength;const a=new ArrayBuffer(r),i=new Uint8Array(a);let o=0;for(let e=0;e<n;e++){const t=new Uint8Array(c[p+e]);i.set(t,o),o+=t.byteLength}s[e].forEach(e=>{const t=Bs(a.slice(e.groupOffset,e.groupOffset+e.sizeBytes),[e.manifestEntry]);for(const e in t)h[e]=t[e]}),p+=n}),h}}Ys.registerSaveRouter(e=>Ye().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(Ri.URL_SCHEME)?function(e="model"){return new Ri(e)}(e.slice(Ri.URL_SCHEME.length)):null);var Li=class{constructor(e,t){if(this.DEFAULT_METHOD="POST",null==t&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.onProgress=t.onProgress,this.weightUrlConverter=t.weightUrlConverter,null!=t.fetchFunc?(ue("function"==typeof t.fetchFunc,()=>"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"),this.fetch=t.fetchFunc):this.fetch=Ye().platform.fetch,ue(null!=e&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&ue(2===e.length,()=>`URL paths for http must have a length of 2, (actual length is ${e.length}).`),this.path=e,null!=t.requestInit&&null!=t.requestInit.body)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=t.requestInit||{}}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");const t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData;const n=Gs(e,[{paths:["./model.weights.bin"],weights:e.weightSpecs}]);t.body.append("model.json",new Blob([JSON.stringify(n)],{type:"application/json"}),"model.json"),null!=e.weightData&&t.body.append("model.weights.bin",new Blob([e.weightData],{type:"application/octet-stream"}),"model.weights.bin");const r=await this.fetch(this.path,t);if(r.ok)return{modelArtifactsInfo:qs(e),responses:[r]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${r.status}.`)}async load(){const e=await this.fetch(this.path,this.requestInit);if(!e.ok)throw new Error(`Request to ${this.path} failed with status code ${e.status}. Please verify this URL points to the model JSON of the model to load.`);let t;try{t=await e.json()}catch(e){let t=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?t+=" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.":t+=" Please make sure the server is serving valid JSON for this request.",new Error(t)}const n=t.modelTopology,r=t.weightsManifest;if(null==n&&null==r)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return Hs(t,e=>this.loadWeights(e))}async loadWeights(e){const t=Array.isArray(this.path)?this.path[1]:this.path,[n,r]=function(e){const t=e.lastIndexOf("/"),n=e.lastIndexOf("?"),r=e.substring(0,t),a=n>t?e.substring(n):"";return[r+"/",a]}(t),a=this.weightPathPrefix||n,s=[];for(const t of e)s.push(...t.weights);const i=[],o=[];for(const t of e)for(const e of t.paths)null!=this.weightUrlConverter?o.push(this.weightUrlConverter(e)):i.push(a+e+r);this.weightUrlConverter&&i.push(...await Promise.all(o));return[s,Vs(await Mi(i,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress}))]}};function Pi(e){return null!=e.match(Li.URL_SCHEME_REGEX)}Li.URL_SCHEME_REGEX=/^https?:\/\//;var $i=(e,t)=>{if("undefined"==typeof fetch&&(null==t||null==t.fetchFunc))return null;{let n=!0;if(n=Array.isArray(e)?e.every(e=>Pi(e)):Pi(e),n)return Bi(e,t)}return null};function Bi(e,t){return new Li(e,t)}function zi(e,t){return Bi(e,t)}Ys.registerSaveRouter($i),Ys.registerLoadRouter($i);var Ui=class{constructor(e){this.modelArtifacts=e}async load(){return this.modelArtifacts}};function Wi(e,t,n,r){if(1===arguments.length){return null!=e.modelTopology||null!=e.weightSpecs?new Ui(e):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new Ui({modelTopology:e}))}return console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new Ui({modelTopology:e,weightSpecs:t,weightData:n,trainingConfig:r})}function Vi(e){return new class{constructor(e){this.saveHandler=e}async save(e){return this.saveHandler(e)}}(e)}var ji={};w(ji,{confusionMatrix:()=>Ki});var Gi=Ms({matMul_:function(e,t,n=!1,r=!1){let a=Rs(e,"a","matMul"),s=Rs(t,"b","matMul");[a,s]=ps(a,s);const i={a:a,b:s},o={transposeA:n,transposeB:r};return xs.runKernel(bt,i,o)}});var Hi=Ms({oneHot_:function(e,t,n=1,r=0){if(t<2)throw new Error("Error in oneHot: depth must be >=2, but it is "+t);const a={indices:Rs(e,"indices","oneHot","int32")},s={depth:t,onValue:n,offValue:r};return xs.runKernel(Zn,a,s)}});var qi=Ms({transpose_:function(e,t){const n=Rs(e,"x","transpose");if(null==t&&(t=n.shape.map((e,t)=>t).reverse()),ue(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of perm ${t}.`),t.forEach(e=>{ue(e>=0&&e<n.rank,()=>`All entries in 'perm' must be between 0 and ${n.rank-1} but got ${t}`)}),n.rank<=1)return n.clone();const r={x:n},a={perm:t};return xs.runKernel(Yr,r,a)}});var Ki=Ms({confusionMatrix_:function(e,t,n){const r=Rs(e,"labels","confusionMatrix"),a=Rs(t,"predictions","confusionMatrix");ue(null==n||n>0&&Number.isInteger(n),()=>"If provided, numClasses must be a positive integer, but got "+n),ue(1===r.rank,()=>"Expected the rank of labels to be 1, but got "+r.rank),ue(1===a.rank,()=>"Expected the rank of predictions to be 1, but got "+a.rank),ue(r.shape[0]===a.shape[0],()=>`Mismatch in the number of examples: ${r.shape[0]} vs. ${a.shape[0]}. Labels and predictions should have the same number of elements.`),ue(n>0&&Number.isInteger(n),()=>"numClasses is required to be a positive integer, but got "+n);const s=Hi(Ai(r,"int32"),n),i=Hi(Ai(a,"int32"),n),o=qi(s),u=Gi(o,i);return Ai(u,"int32")}}),Yi={};function Xi(e,t){const n=e.length,r=[];for(let a=0;a<n;a++){const s=n-1-a,i=e[s]||1;(t[t.length-1-a]||1)>1&&1===i&&r.unshift(s)}return r}function Qi(e,t){const n=[];for(let r=0;r<t.length;r++){const a=e[e.length-r-1],s=t.length-r-1,i=t[s];(null==a||1===a&&i>1)&&n.unshift(s)}return n}function Zi(e,t){const n=[],r=Math.max(e.length,t.length);for(let a=0;a<r;a++){let r=e[e.length-a-1];null==r&&(r=1);let s=t[t.length-a-1];if(null==s&&(s=1),1===r)n.unshift(s);else if(1===s)n.unshift(r);else{if(r!==s){throw Error(`Operands could not be broadcast together with shapes ${e} and ${t}.`)}n.unshift(r)}}return n}w(Yi,{assertAndGetBroadcastShape:()=>Zi,getBroadcastDims:()=>Xi,getReductionAxes:()=>Qi});var Ji,eo={};function to(e,t,n){if(ce(e),null!=t&&3!==t.length)throw new Error("tensor3d() requires shape to have three numbers");const r=Cs(e,n);if(3!==r.length&&1!==r.length)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return Os(e,t,r,n)}function no(e,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(null==e)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let n=!1,r=!1,a=!1,s=!1,i=!1,o=!1;if(e.data instanceof Uint8Array)n=!0;else if("undefined"!=typeof ImageData&&e instanceof ImageData)r=!0;else if("undefined"!=typeof HTMLVideoElement&&e instanceof HTMLVideoElement)a=!0;else if("undefined"!=typeof HTMLImageElement&&e instanceof HTMLImageElement)s=!0;else if(null!=e.getContext)i=!0;else{if(!("undefined"!=typeof ImageBitmap&&e instanceof ImageBitmap))throw new Error("pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was "+e.constructor.name);o=!0}if(a){const t=2;if(a&&e.readyState<t)throw new Error("The video element has not loaded data yet. Please wait for `loadeddata` event on the <video> element.")}if(null!=ca(ta,xs.backendName)){const n={pixels:e},r={numChannels:t};return xs.runKernel(ta,n,r)}const[u,l]=a?[e.videoWidth,e.videoHeight]:[e.width,e.height];let c,h;if(i)c=e.getContext("2d").getImageData(0,0,u,l).data;else if(r||n)c=e.data;else if(s||a||o){if(null==Ji)if("undefined"==typeof document){if("undefined"==typeof OffscreenCanvas||"undefined"==typeof OffscreenCanvasRenderingContext2D)throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");Ji=new OffscreenCanvas(1,1).getContext("2d")}else Ji=document.createElement("canvas").getContext("2d");Ji.canvas.width=u,Ji.canvas.height=l,Ji.drawImage(e,0,0,u,l),c=Ji.getImageData(0,0,u,l).data}if(4===t)h=new Int32Array(c);else{const e=u*l;h=new Int32Array(e*t);for(let n=0;n<e;n++)for(let e=0;e<t;++e)h[n*t+e]=c[4*n+e]}return to(h,[l,u,t],"int32")}function ro(e){return"undefined"!=typeof window&&"undefined"!=typeof ImageBitmap&&window.hasOwnProperty("createImageBitmap")&&!(e instanceof ImageBitmap)&&function(e){return null!=e&&0!==e.width&&0!==e.height}(e)&&!function(e){return null!=e&&e.data instanceof Uint8Array}(e)}async function ao(e,t=3){let n=null;if(Ye().getBool("WRAP_TO_IMAGEBITMAP")&&ro(e)){let t;try{t=await createImageBitmap(e,{premultiplyAlpha:"none"})}catch(e){t=null}n=null!=t&&t.width===e.width&&t.height===e.height?t:e}else n=e;return no(n,t)}async function so(e,t){let n=Rs(e,"img","toPixels");if(!(e instanceof Ya)){const e=n;n=Ai(e,"int32"),e.dispose()}if(2!==n.rank&&3!==n.rank)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${n.rank}.`);const[r,a]=n.shape.slice(0,2),s=2===n.rank?1:n.shape[2];if(s>4||2===s)throw new Error("toPixels only supports depth of size 1, 3 or 4 but got "+s);if("float32"!==n.dtype&&"int32"!==n.dtype)throw new Error(`Unsupported type for toPixels: ${n.dtype}. Please use float32 or int32 tensors.`);const i=await n.data(),o="float32"===n.dtype?255:1,u=new Uint8ClampedArray(a*r*4);for(let e=0;e<r*a;++e){const t=[0,0,0,255];for(let r=0;r<s;r++){const a=i[e*s+r];if("float32"===n.dtype){if(a<0||a>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${a}.`)}else if("int32"===n.dtype&&(a<0||a>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${a}.`);1===s?(t[0]=a*o,t[1]=a*o,t[2]=a*o):t[r]=a*o}const r=4*e;u[r+0]=Math.round(t[0]),u[r+1]=Math.round(t[1]),u[r+2]=Math.round(t[2]),u[r+3]=Math.round(t[3])}if(null!=t){t.width=a,t.height=r;const e=t.getContext("2d"),n=new ImageData(u,a,r);e.putImageData(n,0,0)}return n!==e&&n.dispose(),u}w(eo,{fromPixels:()=>io,fromPixelsAsync:()=>ao,toPixels:()=>so});var io=Ms({fromPixels_:no}),oo={};function uo(e,t){const n=e.shape.length,r=t.shape.length;if(n<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${n}.`);if(r<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${r}.`);if("int32"!==t.dtype)throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[r-1]>n)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[r-1]} vs. ${n}`);if(0===pe(e.shape))throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);const a=t.shape,s=a[a.length-1];let i=1;for(let e=0;e<a.length-1;++e)i*=a[e];const o=e.shape,u=a.slice();u.pop();let l=1;for(let e=s;e<n;++e)l*=o[e],u.push(o[e]);const c=[...$e(e.shape).map(e=>e/l),1].slice(0,s);return[u,i,l,c]}w(oo,{prepareAndValidate:()=>uo});var lo={};function co(e,t,n){const r=t.rank>1?t.shape[t.rank-1]:1,a=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${r}, and batchDim: ${a}.`;if(n.rank<a)throw new Error(s+` update.rank < ${a}. `);if(e.length<r+(n.rank-a))throw new Error(s+" Output shape length < "+(r+(n.rank-a)));if(n.rank!==a+e.length-r)throw new Error(s+" update.rank != "+(a+e.length-r));for(let e=0;e<a;++e)if(n.shape[e]!==t.shape[e])throw new Error(s+` updates.shape[${e}] (${n.shape[e]}) != indices.shape[${e}] (${t.shape[e]}).`);for(let t=0;t<n.rank-a;++t)if(n.shape[t+a]!==e[t+r])throw new Error(s+` updates.shape[${t+a}] (${n.shape[t+a]}) != shape[${t+a}] (${e[t+a]})`)}function ho(e,t,n){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(e.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${e.rank}.`);if("int32"!==t.dtype)throw new Error("The dtype of 'indices' should be int32, but got dtype: "+t.dtype);if(n.length<1)throw new Error("Output rank must be greater or equal to 1, but got shape: "+n);if(0===n.length){if(0===t.size)throw new Error("Indices specified for empty output. indices shape: "+t.shape);if(0===e.size)throw new Error("Updates specified for empty output. updates shape: "+e.shape)}co(n,t,e)}function po(e,t,n){const r=t.shape.length,a=r>1?t.shape[r-1]:1,s=n.length;let i=1;for(let e=a;e<s;++e)i*=n[e];const o=a<1?1:a;return{sliceRank:a,numUpdates:pe(t.shape)/o,sliceSize:i,strides:[...$e(n.slice(0,a)),1],outputSize:pe(n)}}w(lo,{calculateShapes:()=>po,validateInput:()=>ho,validateUpdateShape:()=>co});var fo={};w(fo,{assertParamsValid:()=>mo,computeFlatOffset:()=>Co,computeOutShape:()=>yo,getNormalizedAxes:()=>wo,isSliceContinous:()=>No,maskToAxes:()=>go,parseSliceParams:()=>To,sliceInfo:()=>Ro,startForAxis:()=>Ao,startIndicesWithElidedDims:()=>ko,stopForAxis:()=>Eo,stopIndicesWithElidedDims:()=>So,stridesForAxis:()=>Io,stridesWithElidedDims:()=>bo});function mo(e,t,n){const r=e.shape.length;ue(r===t.length,()=>`Error in slice${r}D: Length of begin ${t} must match the rank of the array (${r}).`),ue(r===n.length,()=>`Error in slice${r}D: Length of size ${n} must match the rank of the array (${r}).`);for(let a=0;a<r;++a)ue(t[a]+n[a]<=e.shape[a],()=>`Error in slice${r}D: begin[${a}] + size[${a}] (${t[a]+n[a]}) would overflow input.shape[${a}] (${e.shape[a]})`)}function go(e){const t=[];let n=0;for(;e>0;)1&e&&t.push(n),e/=2,n++;return t}function yo(e,t,n){const r=[];for(let a=0;a<e.length;a++)r[a]=Math.ceil((t[a]-e[a])/n[a]);return r}function bo(e,t,n,r){const a=[...e];for(let e=a.length;e<r.length;e++)a.push(1);for(let e=0;e<n;e++)0===e?a[t]=1:(a.splice(t,0,1),a.pop());return a}function vo(e,t,n){return n<=e?n:n-(t-1)}function xo(e,t){const n=[];for(let r=0;r<e;r++)n.push(t+r);return n}function wo(e,t,n,r,a,s,i,o,u){const l=e.length;let c=new Array(l),h=new Array(l),p=new Array(l);if(t.length&&n>0){const u=t[0],l=n+1;c=ko(i,u,l,r,e),h=So(o,u,l,a,e),p=bo(s,u,l,e)}else for(let t=0;t<l;t++)c[t]=Ao(i,r,s,e,t,u),h[t]=Eo(o,a,s,e,t,u),p[t]=Io(s,t,u);return{begin:c,end:h,strides:p}}function ko(e,t,n,r,a){const s=[...a],i=xo(n,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=0;else{const i=vo(t,n,a);let o=r[i];e&1<<i&&(o=0),s[a]=o}return s}function So(e,t,n,r,a){const s=[...a],i=xo(n,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=Number.MAX_SAFE_INTEGER;else{const i=vo(t,n,a);let o=r[i];e&1<<i&&(o=Number.MAX_SAFE_INTEGER),s[a]=o}for(let e=0;e<s.length;e++){const t=a[e];s[e]<0&&(s[e]+=t),s[e]=ne(0,s[e],a[e])}return s}function Io(e,t,n){let r=e[t];return(n&1<<t||null==r)&&(r=1),r}function Ao(e,t,n,r,a,s){let i=t[a];const o=n[a]||1;(e&1<<a||s&1<<a||null==i)&&(i=o>0?Number.MIN_SAFE_INTEGER:Number.MAX_SAFE_INTEGER);const u=r[a];return i<0&&(i+=u),i=ne(0,i,u-1),i}function Eo(e,t,n,r,a,s){let i=t[a];const o=n[a]||1;(e&1<<a||s&1<<a||null==i)&&(i=o>0?Number.MAX_SAFE_INTEGER:Number.MIN_SAFE_INTEGER);const u=r[a];return i<0&&(i+=u),i=o>0?ne(0,i,u):ne(-1,i,u-1),i}function No(e,t,n){let r=n.length;for(let e=0;e<n.length;e++)if(n[e]>1){r=e;break}for(let a=r+1;a<n.length;a++)if(t[a]>0||n[a]!==e[a])return!1;return!0}function Co(e,t){let n=e.length>0?e[e.length-1]:1;for(let r=0;r<e.length-1;r++)n+=e[r]*t[r];return n}function To(e,t,n){let r;const a=e.shape.length;let s;return r="number"==typeof t?[t,...new Array(a-1).fill(0)]:t.length<a?t.concat(new Array(a-t.length).fill(0)):t.slice(),r.forEach(e=>{ue(-1!==e,()=>"slice() does not support negative begin indexing.")}),s=null==n?new Array(a).fill(-1):"number"==typeof n?[n,...new Array(a-1).fill(-1)]:n.length<a?n.concat(new Array(a-n.length).fill(-1)):n,s=s.map((t,n)=>t>=0?t:(ue(-1===t,()=>`Negative size values should be exactly -1 but got ${t} for the slice() size at index ${n}.`),e.shape[n]-r[n])),[r,s]}function Ro(e,t,n,r,a,s,i,o,u){let l;if(null==r?(l=new Array(t.length),l.fill(1)):l=r,null!=i&&0!=(i&i-1))throw new Error("Multiple ellipses in slice is not allowed.");let c=!1;const h={dims:l.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:n.slice(),strides:l.slice(),beginMask:a,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:u};for(let e=0;e<h.dims;e++)c&&0!=(1<<e&o)&&h.numAddAxisAfterEllipsis++,1<<e&i&&(c=!0);c||(h.ellipsisMask|=1<<h.dims,h.dims++);const p={dims:e.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};!function(e,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let n=0;t.beginValid=null!=e.begin,t.endValid=null!=e.end,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let r=0;r<e.dims;r++)if(1<<r&e.ellipsisMask){const a=Math.min(t.dims-(e.dims-r)+1+e.numAddAxisAfterEllipsis,t.dims);for(;n<a;n++)t.begin[n]=0,t.end[n]=0,t.strides[n]=1,t.beginMask|=1<<n,t.endMask|=1<<n,t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[n]=r}else if(1<<r&e.newAxisMask)t.finalShapeGatherIndices.push(-2),t.finalShapeGatherIndicesSparse.push(-1);else{if(n===t.begin.length)throw Error(`Index out of range using input dim ${n}; input has only ${t.dims} dims, ${t.begin.length}.`);null!=e.begin&&(t.begin[n]=e.begin[r]),null!=e.end&&(t.end[n]=e.end[r]),t.strides[n]=e.strides[r],e.beginMask&1<<r&&(t.beginMask|=1<<n),e.endMask&1<<r&&(t.endMask|=1<<n),e.shrinkAxisMask&1<<r?(t.finalShapeGatherIndices.push(-1),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<n):(t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(r)),t.inputShapeGatherIndicesSparse[n]=r,n++}}(h,p);let d=!0,f=!0,m=!0;const g=[],y=[];for(let t=0;t<e.length;++t){if(0===p.strides[t])throw Error(`strides[${t}] must be non-zero`);const n=!!(p.shrinkAxisMask&1<<t),r=e[t];if(-1===r){g.push(n?1:-1);continue}const a=[p.beginMask&1<<t,p.endMask&1<<t],s=[p.strides[t]>0?0:-1,p.strides[t]>0?r:r-1];if(n&&p.strides[t]<=0)throw Error("only stride 1 allowed on non-range indexing.");m=m&&1===p.strides[t];const i=!!(p.beginMask&1<<t&&p.endMask&1<<t);if(p.beginValid&&p.endValid){if(n){const e=p.begin[t]<0?r+p.begin[t]:p.begin[t];if(p.begin[t]=e,p.end[t]=p.begin[t]+1,e<0||e>=r)throw Error(`slice index ${p.begin[t]} of dimension ${t} out of bounds.`)}else p.begin[t]=_o(p.begin[t],0,p.strides[t],r,a,s),p.end[t]=_o(p.end[t],1,p.strides[t],r,a,s);const e=1===p.strides[t]&&0===p.begin[t]&&p.end[t]===r;d=d&&e,f=f&&(0===t&&1===p.strides[t]||e)}else d=d&&1===p.strides[t]&&i,f=f&&(0===t&&1===p.strides[t]||i);let o,u=!1;if(p.beginValid&&p.endValid?(o=p.end[t]-p.begin[t],u=!0):n?(o=1,u=!0):i&&r>=0&&(o=p.strides[t]<0?-r:r,u=!0),u){let e;e=0===o||o<0!=p.strides[t]<0?0:Math.trunc(o/p.strides[t])+(o%p.strides[t]!=0?1:0),g.push(e)}else g.push(-1)}for(let e=0;e<p.finalShapeGatherIndices.length;++e){const t=p.finalShapeGatherIndices[e];t>=0?y.push(g[t]):-2===t&&y.push(1)}return{finalShapeSparse:y.filter((e,t)=>-2!==p.finalShapeGatherIndices[t]),finalShape:y,isIdentity:d,sliceDim0:f,isSimpleSlice:m,begin:p.begin,end:p.end,strides:p.strides}}function _o(e,t,n,r,a,s){if(a[t])return n>0?s[t]:s[t+1&1];{const t=e<0?r+e:e;return t<s[0]?s[0]:t>s[1]?s[1]:t}}var Fo={};w(Fo,{Serializable:()=>Mo,SerializationMap:()=>Do,registerClass:()=>Oo});var Mo=class{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}},Do=class{constructor(){this.classNameMap={}}static getMap(){return null==Do.instance&&(Do.instance=new Do),Do.instance}static register(e){Do.getMap().classNameMap[e.className]=[e,e.fromConfig]}};function Oo(e){ue(null!=e.className,()=>"Class being registered does not have the static className property defined."),ue("string"==typeof e.className,()=>"className is required to be a string, but got type "+typeof e.className),ue(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),Do.register(e)}var Lo={};w(Lo,{TEST_EPSILON_FLOAT16:()=>Po,encodeStrings:()=>qo,expectArrayBuffersEqual:()=>Ho,expectArraysClose:()=>$o,expectArraysEqual:()=>Wo,expectNumbersClose:()=>Vo,expectPromiseToFail:()=>Uo,expectValuesInRange:()=>Go,testEpsilon:()=>Bo});var Po=.1;function $o(e,t,n){return null==n&&(n=Bo()),zo(e,t,(e,t)=>jo(e,t,n))}function Bo(){return 32===xs.backend.floatPrecision()?.001:Po}function zo(e,t,n){let r=!0;if((Te(e)||Te(t))&&(r=!1),Te(e)&&Te(t)&&(r=!0),r){const n=e.constructor.name,r=t.constructor.name;if(n!==r)throw new Error(`Arrays are of different type. Actual: ${n}. Expected: ${r}`)}if(Array.isArray(e)&&Array.isArray(t)){const n=Cs(e),r=Cs(t);if(!fe(n,r))throw new Error(`Arrays have different shapes. Actual: [${n}]. Expected: [${r}]`)}const a=Te(e)?e:he(e),s=Te(t)?t:he(t);if(a.length!==s.length)throw new Error(`Arrays have different lengths actual: ${a.length} vs expected: ${s.length}.\nActual:   ${a}.\nExpected: ${s}.`);for(let e=0;e<s.length;++e){const t=a[e],r=s[e];if(!n(t,r))throw new Error(`Arrays differ: actual[${e}] = ${t}, expected[${e}] = ${r}.\nActual:   ${a}.\nExpected: ${s}.`)}}function Uo(e,t){e().then(()=>t.fail(),()=>t())}function Wo(e,t){const n="string"==typeof t||"number"==typeof t||"boolean"==typeof t?[t]:t;return Fe(e)||Fe(e[0])||Fe(t)||Fe(t[0])?zo(e,n,(e,t)=>e==t):zo(e,t,(e,t)=>jo(e,t,0))}function Vo(e,t,n){if(null==n&&(n=Bo()),!jo(e,t,n))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`)}function jo(e,t,n){return!isFinite(e)&&!isFinite(t)||!(isNaN(e)||isNaN(t)||Math.abs(e-t)>n)}function Go(e,t,n){for(let r=0;r<e.length;r++)if(e[r]<t||e[r]>n)throw new Error(`Value out of range:${e[r]} low: ${t}, high: ${n}`)}function Ho(e,t){const n=new Float32Array(e),r=new Float32Array(t);if(n.length!==r.length)throw new Error(`Expected ArrayBuffer to be of length ${r.length}, but it was ${n.length}`);for(let e=0;e<r.length;e++)if(n[e]!==r[e])throw new Error(`Expected ArrayBuffer value at ${e} to be ${r[e]} but got ${n[e]} instead`)}function qo(e){for(let t=0;t<e.length;t++){const n=e[t];Array.isArray(n)?qo(n):e[t]=$a(n)}return e}var Ko="3.13.0";function Yo(){Ye().set("PROD",!0)}function Xo(){Ye().set("DEBUG",!0)}function Qo(){Ye().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function Zo(e){Ye().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}function Jo(){xs.disposeVariables()}function eu(){return xs}function tu(){return xs.memory()}function nu(e){return xs.profile(e)}function ru(e,t){return xs.tidy(e,t)}function au(e){ms(e).forEach(e=>e.dispose())}function su(e){return xs.keep(e)}function iu(e){return xs.time(e)}function ou(e){return xs.setBackend(e)}function uu(){return xs.ready()}function lu(){return xs.backendName}function cu(e){xs.removeBackend(e)}function hu(e){return xs.findBackend(e)}function pu(e){return xs.findBackendFactory(e)}function du(e,t,n=1){return xs.registerBackend(e,t,n)}function fu(){return xs.backend}function mu(e,t){Ye().setPlatform(e,t)}var gu=Ms({add_:function(e,t){let n=Rs(e,"a","add"),r=Rs(t,"b","add");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(rt,a)}});var yu=Ms({floorDiv_:function(e,t){let n=Rs(e,"a","floorDiv"),r=Rs(t,"b","floorDiv");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(un,a)}});var bu=Ms({div_:function(e,t){let n=Rs(e,"a","div"),r=Rs(t,"b","div");if([n,r]=ps(n,r),"int32"===n.dtype&&"int32"===r.dtype)return yu(n,r);const a={a:n,b:r};return xs.runKernel(Kt,a,{})}});var vu=Ms({mul_:function(e,t){let n=Rs(e,"a","mul"),r=Rs(t,"b","mul");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(Gn,a)}});var xu=Ms({abs_:function(e){const t=Rs(e,"x","abs");if("complex64"===t.dtype){const e={x:t};return xs.runKernel(Nt,e)}{const e={x:t};return xs.runKernel(et,e)}}});var wu=Ms({acos_:function(e){const t={x:Rs(e,"x","acos")};return xs.runKernel(tt,t)}});var ku=Ms({acosh_:function(e){const t={x:Rs(e,"x","acosh")};return xs.runKernel(nt,t)}});var Su=Ms({addN_:function(e){ue(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),ue(e.length>=1,()=>"Must pass at least one tensor to tf.addN(), but got "+e.length);const t=e.map((e,t)=>Rs(e,"tensors"+t,"addN")),n=t[0];t.forEach(e=>{if(e.dtype!==n.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(e=>{if(!fe(e.shape,n.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});const r=t;return xs.runKernel(at,r)}});var Iu=Ms({all_:function(e,t=null,n=!1){const r={x:Rs(e,"x","all","bool")},a={axis:t,keepDims:n};return xs.runKernel(st,r,a)}});var Au=Ms({any_:function(e,t=null,n=!1){const r={x:Rs(e,"x","any","bool")},a={axis:t,keepDims:n};return xs.runKernel(it,r,a)}});var Eu=Ms({argMax_:function(e,t=0){const n={x:Rs(e,"x","argMax")},r={axis:t};return xs.runKernel(ot,n,r)}});var Nu=Ms({argMin_:function(e,t=0){const n={x:Rs(e,"x","argMin")},r={axis:t};return xs.runKernel(ut,n,r)}});var Cu=Ms({asin_:function(e){const t={x:Rs(e,"x","asin")};return xs.runKernel(lt,t)}});var Tu=Ms({asinh_:function(e){const t={x:Rs(e,"x","asinh")};return xs.runKernel(ct,t)}});var Ru=Ms({atan_:function(e){const t={x:Rs(e,"x","atan")};return xs.runKernel(ht,t)}});var _u=Ms({atan2_:function(e,t){let n=Rs(e,"a","atan2"),r=Rs(t,"b","atan2");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(dt,a)}});var Fu=Ms({atanh_:function(e){const t={x:Rs(e,"x","atanh")};return xs.runKernel(pt,t)}});function Mu(e,t,n,r,a="NHWC",s){return Lu(e,[...t,e[3]],n,s,r,null,null,Gu(a))}function Du(e,t,n,r,a,s,i="channelsLast"){const[o,u]=Bu(t);let l;if("channelsLast"===i)l=[o,u,e[3],e[3]];else{if("channelsFirst"!==i)throw new Error("Unknown dataFormat "+i);l=[o,u,e[1],e[1]]}return Lu(e,l,n,r,a,s,!1,i)}function Ou(e,t,n,r,a,s,i="NDHWC"){const[o,u,l]=zu(t);let c,h;if("NDHWC"===i)h="channelsLast",c=[o,u,l,e[4],e[4]];else{if("NCDHW"!==i)throw new Error("Unknown dataFormat "+i);h="channelsFirst",c=[o,u,l,e[1],e[1]]}return Pu(e,c,n,r,a,!1,h,s)}function Lu(e,t,n,r,a,s,i=!1,o="channelsLast"){let[u,l,c,h]=[-1,-1,-1,-1];if("channelsLast"===o)[u,l,c,h]=e;else{if("channelsFirst"!==o)throw new Error("Unknown dataFormat "+o);[u,h,l,c]=e}const[p,d,,f]=t,[m,g]=Bu(n),[y,b]=Bu(r),v=Uu(p,y),x=Uu(d,b),{padInfo:w,outHeight:k,outWidth:S}=function(e,t,n,r,a,s,i,o,u){let l,c,h;if("number"==typeof e){l={top:e,bottom:e,left:e,right:e,type:0===e?"VALID":"NUMBER"};const a=function(e,t,n,r,a){null==r&&(r=$u(e,t,n));const s=e[0],i=e[1],o=Wu((s-t+2*r)/n+1,a),u=Wu((i-t+2*r)/n+1,a);return[o,u]}([t,n],s,r,e,o);c=a[0],h=a[1]}else if("same"===e){c=Math.ceil(t/r),h=Math.ceil(n/a);const e=Math.max(0,(c-1)*r+s-t),o=Math.max(0,(h-1)*a+i-n),u=Math.floor(e/2),p=e-u,d=Math.floor(o/2);l={top:u,bottom:p,left:d,right:o-d,type:"SAME"}}else if("valid"===e)l={top:0,bottom:0,left:0,right:0,type:"VALID"},c=Math.ceil((t-s+1)/r),h=Math.ceil((n-i+1)/a);else{if("object"!=typeof e)throw Error("Unknown padding parameter: "+e);{const p="channelsLast"===u?e[1][0]:e[2][0],d="channelsLast"===u?e[1][1]:e[2][1],f="channelsLast"===u?e[2][0]:e[3][0],m="channelsLast"===u?e[2][1]:e[3][1];l={top:p,bottom:d,left:f,right:m,type:0===p&&0===d&&0===f&&0===m?"VALID":"EXPLICIT"},c=Wu((t-s+p+d)/r+1,o),h=Wu((n-i+f+m)/a+1,o)}}return{padInfo:l,outHeight:c,outWidth:h}}(a,l,c,m,g,v,x,s,o),I=i?f*h:f;let A;return"channelsFirst"===o?A=[u,I,k,S]:"channelsLast"===o&&(A=[u,k,S,I]),{batchSize:u,dataFormat:o,inHeight:l,inWidth:c,inChannels:h,outHeight:k,outWidth:S,outChannels:I,padInfo:w,strideHeight:m,strideWidth:g,filterHeight:p,filterWidth:d,effectiveFilterHeight:v,effectiveFilterWidth:x,dilationHeight:y,dilationWidth:b,inShape:e,outShape:A,filterShape:t}}function Pu(e,t,n,r,a,s=!1,i="channelsLast",o){let[u,l,c,h,p]=[-1,-1,-1,-1,-1];if("channelsLast"===i)[u,l,c,h,p]=e;else{if("channelsFirst"!==i)throw new Error("Unknown dataFormat "+i);[u,p,l,c,h]=e}const[d,f,m,,g]=t,[y,b,v]=zu(n),[x,w,k]=zu(r),S=Uu(d,x),I=Uu(f,w),A=Uu(m,k),{padInfo:E,outDepth:N,outHeight:C,outWidth:T}=function(e,t,n,r,a,s,i,o,u,l,c){let h,p,d,f;if("number"==typeof e){h={top:e,bottom:e,left:e,right:e,front:e,back:e,type:0===e?"VALID":"NUMBER"};const s=function(e,t,n,r,a,s){null==a&&(a=$u(e,t,r));const i=e[0],o=e[1],u=e[2],l=Wu((i-t+2*a)/r+1,s),c=Wu((o-t+2*a)/r+1,s),h=Wu((u-t+2*a)/r+1,s);return[l,c,h,n]}([t,n,r,1],o,1,a,e,c);p=s[0],d=s[1],f=s[2]}else if("same"===e){p=Math.ceil(t/a),d=Math.ceil(n/s),f=Math.ceil(r/i);const e=(p-1)*a+o-t,c=(d-1)*s+u-n,m=(f-1)*i+l-r,g=Math.floor(e/2),y=e-g,b=Math.floor(c/2),v=c-b,x=Math.floor(m/2);h={top:b,bottom:v,left:x,right:m-x,front:g,back:y,type:"SAME"}}else{if("valid"!==e)throw Error("Unknown padding parameter: "+e);h={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},p=Math.ceil((t-o+1)/a),d=Math.ceil((n-u+1)/s),f=Math.ceil((r-l+1)/i)}return{padInfo:h,outDepth:p,outHeight:d,outWidth:f}}(a,l,c,h,y,b,v,S,I,A,o),R=s?g*p:g;let _;return"channelsFirst"===i?_=[u,R,N,C,T]:"channelsLast"===i&&(_=[u,N,C,T,R]),{batchSize:u,dataFormat:i,inDepth:l,inHeight:c,inWidth:h,inChannels:p,outDepth:N,outHeight:C,outWidth:T,outChannels:R,padInfo:E,strideDepth:y,strideHeight:b,strideWidth:v,filterDepth:d,filterHeight:f,filterWidth:m,effectiveFilterDepth:S,effectiveFilterHeight:I,effectiveFilterWidth:A,dilationDepth:x,dilationHeight:w,dilationWidth:k,inShape:e,outShape:_,filterShape:t}}function $u(e,t,n,r=1){const a=Uu(t,r);return Math.floor((e[0]*(n-1)-n+a)/2)}function Bu(e){return"number"==typeof e?[e,e,e]:2===e.length?[e[0],e[1],1]:e}function zu(e){return"number"==typeof e?[e,e,e]:e}function Uu(e,t){return t<=1?e:e+(e-1)*(t-1)}function Wu(e,t){if(!t)return Math.trunc(e);switch(t){case"round":return Math.round(e);case"ceil":return Math.ceil(e);case"floor":return Math.floor(e);default:throw new Error("Unknown roundingMode "+t)}}function Vu(e){const[t,n,r]=Bu(e);return 1===t&&1===n&&1===r}function ju(e,t){return Vu(e)||Vu(t)}function Gu(e){if("NHWC"===e)return"channelsLast";if("NCHW"===e)return"channelsFirst";throw new Error("Unknown dataFormat "+e)}function Hu(e,t,n){if(null!=n){if("string"==typeof t)throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);if("number"==typeof t)ue(me(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);else{if("object"!=typeof t)throw Error(`Error in ${e}: Unknown padding parameter: ${t}`);t.forEach(t=>{t.forEach(t=>{ue(me(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`)})})}}}var qu=Ms({reshape_:function(e,t){const n={x:Rs(e,"x","reshape","string_or_numeric")},r={shape:t};return xs.runKernel(lr,n,r)}});var Ku=Ms({avgPool_:function(e,t,n,r,a){const s=Rs(e,"x","avgPool","float32");ue(ju(n,1),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '1'`);let i=s,o=!1;3===s.rank&&(o=!0,i=qu(s,[1,s.shape[0],s.shape[1],s.shape[2]])),ue(4===i.rank,()=>`Error in avgPool: x must be rank 4 but got rank ${i.rank}.`),Hu("avgPool",r,a);const u={x:i},l={filterSize:t,strides:n,pad:r,dimRoundingMode:a};let c=xs.runKernel(ft,u,l);return c=Ai(c,s.dtype),o?qu(c,[c.shape[1],c.shape[2],c.shape[3]]):c}});var Yu=Ms({avgPool3d_:function(e,t,n,r,a,s="NDHWC"){const i=Rs(e,"x","avgPool3d","float32");let o=i,u=!1;4===i.rank&&(u=!0,o=qu(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),ue(5===o.rank,()=>`Error in avgPool3d: x must be rank 5 but got rank ${o.rank}.`),ue("NDHWC"===s,()=>"Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of "+s),Hu("avgPool3d",r,a);const l={x:o},c={filterSize:t,strides:n,pad:r,dimRoundingMode:a,dataFormat:s};let h=xs.runKernel(gt,l,c);return h=Ai(h,o.dtype),u?qu(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}});var Xu=Ms({concat_:function(e,t=0){ue(e.length>=1,()=>"Pass at least one tensor to concat");const n=_s(e,"tensors","concat","string_or_numeric");if("complex64"===n[0].dtype&&n.forEach(e=>{if("complex64"!==e.dtype)throw new Error(`Cannot concatenate complex64 tensors with a tensor\n          with dtype ${e.dtype}. `)}),1===n.length)return Ei(n[0]);const r=n,a={axis:t};return xs.runKernel(Ct,r,a)}});var Qu=Ms({sigmoid_:function(e){const t={x:Rs(e,"x","sigmoid","float32")};return xs.runKernel(Ar,t)}});var Zu=Ms({slice_:function(e,t,n){const r=Rs(e,"x","slice","string_or_numeric");if(0===r.rank)throw new Error("Slicing scalar is not possible");const a={x:r},s={begin:t,size:n};return xs.runKernel(wr,a,s)}});var Ju=Ms({tanh_:function(e){const t={x:Rs(e,"x","tanh","float32")};return xs.runKernel(Gr,t)}});var el=Ms({basicLSTMCell_:function(e,t,n,r,a,s){const i=Rs(e,"forgetBias","basicLSTMCell"),o=Rs(t,"lstmKernel","basicLSTMCell"),u=Rs(n,"lstmBias","basicLSTMCell"),l=Rs(r,"data","basicLSTMCell"),c=Rs(a,"c","basicLSTMCell"),h=Rs(s,"h","basicLSTMCell"),p=Xu([l,h],1),d=Gi(p,o),f=gu(d,u),m=f.shape[0],g=f.shape[1]/4,y=[m,g],b=Zu(f,[0,0],y),v=Zu(f,[0,g],y),x=Zu(f,[0,2*g],y),w=Zu(f,[0,3*g],y),k=gu(vu(Qu(b),Ju(v)),vu(c,Qu(gu(i,x))));return[k,vu(Ju(k),Qu(w))]}});var tl=Ms({batchToSpaceND_:function(e,t,n){const r=Rs(e,"x","batchToSpaceND"),a=t.reduce((e,t)=>e*t);ue(r.rank>=1+t.length,()=>`input rank is ${r.rank} but should be > than blockShape.length ${t.length}`),ue(n.length===t.length,()=>`crops.length is ${n.length} but should be equal to blockShape.length  ${t.length}`),ue(r.shape[0]%a==0,()=>`input tensor batch is ${r.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${a}`);const s={x:r},i={blockShape:t,crops:n};return xs.runKernel(vt,s,i)}});var nl=Ms({batchNorm_:function(e,t,n,r,a,s){null==s&&(s=.001);const i=Rs(e,"x","batchNorm"),o=Rs(t,"mean","batchNorm"),u=Rs(n,"variance","batchNorm");let l,c;null!=a&&(l=Rs(a,"scale","batchNorm")),null!=r&&(c=Rs(r,"offset","batchNorm")),ue(o.rank===u.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),ue(null==c||o.rank===c.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),ue(null==l||o.rank===l.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");const h={x:function(e){let t;return t=0===e.rank||1===e.rank?qu(e,[1,1,1,e.size]):2===e.rank?qu(e,[1,1,e.shape[0],e.shape[1]]):3===e.rank?qu(e,[1,e.shape[0],e.shape[1],e.shape[2]]):e,t}(i),scale:l,offset:c,mean:o,variance:u},p={varianceEpsilon:s},d=xs.runKernel(ln,h,p);return qu(d,i.shape)}});var rl=Ms({batchNorm2d_:function(e,t,n,r,a,s){const i=Rs(e,"x","batchNorm"),o=Rs(t,"mean","batchNorm"),u=Rs(n,"variance","batchNorm");let l,c;return null!=a&&(l=Rs(a,"scale","batchNorm")),null!=r&&(c=Rs(r,"offset","batchNorm")),ue(2===i.rank,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`),ue(2===o.rank||1===o.rank,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${o.rank}.`),ue(2===u.rank||1===u.rank,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${u.rank}.`),null!=l&&ue(2===l.rank||1===l.rank,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${l.rank}.`),null!=c&&ue(2===c.rank||1===c.rank,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${c.rank}.`),nl(i,o,u,c,l,s)}});var al=Ms({batchNorm3d_:function(e,t,n,r,a,s){const i=Rs(e,"x","batchNorm"),o=Rs(t,"mean","batchNorm"),u=Rs(n,"variance","batchNorm");let l,c;return null!=a&&(l=Rs(a,"scale","batchNorm")),null!=r&&(c=Rs(r,"offset","batchNorm")),ue(3===i.rank,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`),ue(3===o.rank||1===o.rank,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${o.rank}.`),ue(3===u.rank||1===u.rank,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${u.rank}.`),null!=l&&ue(3===l.rank||1===l.rank,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${l.rank}.`),null!=c&&ue(3===c.rank||1===c.rank,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${c.rank}.`),nl(i,o,u,c,l,s)}});var sl=Ms({batchNorm4d_:function(e,t,n,r,a,s){const i=Rs(e,"x","batchNorm"),o=Rs(t,"mean","batchNorm"),u=Rs(n,"variance","batchNorm");let l,c;return null!=a&&(l=Rs(a,"scale","batchNorm")),null!=r&&(c=Rs(r,"offset","batchNorm")),ue(4===i.rank,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`),ue(4===o.rank||1===o.rank,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${o.rank}.`),ue(4===u.rank||1===u.rank,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${u.rank}.`),null!=l&&ue(4===l.rank||1===l.rank,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${l.rank}.`),null!=c&&ue(4===c.rank||1===c.rank,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${c.rank}.`),nl(i,o,u,c,l,s)}});var il=Ms({bincount_:function(e,t,n){const r=Rs(e,"x","bincount"),a=Rs(t,"weights","bincount");ue("int32"===r.dtype,()=>"Error in bincount: input dtype must be int32, but got "+r.dtype),ue(n>=0,()=>`size must be non-negative, but got ${n}.`),ue(a.size===r.size||0===a.size,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${r.shape}, weights shape: ${a.shape}.`);const s={x:r,weights:a},i={size:n};return xs.runKernel(xt,s,i)}});var ol=Ms({broadcastArgs_:function(e,t){const n=Rs(e,"s0","broadcastArgs","int32"),r=Rs(t,"s1","broadcastArgs","int32");if(1!==n.rank)throw new Error("broadcastArgs(): first input must be a vector (rank=1). Has rank "+n.rank);if(1!==r.rank)throw new Error("broadcastArgs(): second input must be a vector (rank=1). Has rank "+r.rank);const a={s0:n,s1:r};return xs.runKernel(kt,a)}});var ul=Ms({broadcastTo_:function(e,t){let n=Rs(e,"broadcastTo","x");const r=n.shape;if(t.some(e=>!(e>0)||e%1!=0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<n.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);if(t.length>n.rank){const e=n.shape.slice();for(;e.length<t.length;)e.unshift(1);n=qu(n,e)}const a=n.shape,s=Array.from(t);for(let e=t.length-1;e>=0;e--)if(a[e]===t[e])s[e]=1;else if(1!==n.shape[e])throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${t}].`);if(0===s.map((e,t)=>e>1?t:-1).filter(e=>e>=0).length)return Ei(n);const i={x:n},o={reps:s};return xs.runKernel(Hr,i,o)}});var ll=Ms({ceil_:function(e){const t={x:Rs(e,"x","ceil","float32")};return xs.runKernel(It,t)}});var cl=Ms({clipByValue_:function(e,t,n){const r=Rs(e,"x","clipByValue");ue(t<=n,()=>`Error in clip: min (${t}) must be less than or equal to max (${n}).`);const a={x:r},s={clipValueMin:t,clipValueMax:n};return xs.runKernel(At,a,s)}});var hl=Ms({concat1d_:function(e){return Xu(e,0)}});var pl=Ms({concat2d_:function(e,t){return Xu(e,t)}});var dl=Ms({concat3d_:function(e,t){return Xu(e,t)}});var fl=Ms({concat4d_:function(e,t){return Xu(e,t)}});var ml=Ms({conv2d_:function(e,t,n,r,a="NHWC",s=[1,1],i){const o=Rs(e,"x","conv2d","float32"),u=Rs(t,"filter","conv2d","float32");let l=o,c=!1;3===o.rank&&(c=!0,l=qu(o,[1,o.shape[0],o.shape[1],o.shape[2]])),ue(4===l.rank,()=>`Error in conv2d: input must be rank 4, but got rank ${l.rank}.`),ue(4===u.rank,()=>`Error in conv2d: filter must be rank 4, but got rank ${u.rank}.`),Hu("conv2d",r,i);const h="NHWC"===a?l.shape[3]:l.shape[1];ue(h===u.shape[2],()=>`Error in conv2d: depth of input (${h}) must match input depth for filter ${u.shape[2]}.`),ue(ju(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);const p={x:l,filter:u},d={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i},f=xs.runKernel(Tt,p,d);return c?qu(f,[f.shape[1],f.shape[2],f.shape[3]]):f}});var gl=Ms({conv1d_:function(e,t,n,r,a="NWC",s=1,i){const o=Rs(e,"x","conv1d"),u=Rs(t,"filter","conv1d");let l=o,c=!1;2===o.rank&&(c=!0,l=qu(o,[1,o.shape[0],o.shape[1]])),ue(3===l.rank,()=>`Error in conv1d: input must be rank 3, but got rank ${l.rank}.`),ue(3===u.rank,()=>`Error in conv1d: filter must be rank 3, but got rank ${u.rank}.`),Hu("conv1d",r,i),ue(l.shape[2]===u.shape[1],()=>`Error in conv1d: depth of input (${l.shape[2]}) must match input depth for filter ${u.shape[1]}.`),ue(ju(n,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${s}'`),ue("NWC"===a,()=>`Error in conv1d: got dataFormat of ${a} but only NWC is currently supported.`);const h=qu(u,[1,u.shape[0],u.shape[1],u.shape[2]]),p=qu(l,[l.shape[0],1,l.shape[1],l.shape[2]]),d=ml(p,h,[1,n],r,"NHWC",[1,s],i);return qu(d,c?[d.shape[2],d.shape[3]]:[d.shape[0],d.shape[2],d.shape[3]])}});var yl=Ms({conv2DBackpropInput_:function(e,t,n,r,a,s="NHWC",i){ue(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let o=e,u=t,l=!1;3===t.rank&&(l=!0,u=qu(t,[1,t.shape[0],t.shape[1],t.shape[2]]),o=[1,e[0],e[1],e[2]]),ue(4===o.length,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${o.length}.`),ue(4===u.rank,()=>"Error in conv2dDerInput: dy must be rank 4, but got rank "+u.rank),ue(4===n.rank,()=>"Error in conv2dDerInput: filter must be rank 4, but got rank "+n.rank);const c="NHWC"===s?o[3]:o[1],h="NHWC"===s?u.shape[3]:u.shape[1];ue(c===n.shape[2],()=>`Error in conv2dDerInput: depth of input (${c}) must match input depth for filter ${n.shape[2]}.`),ue(h===n.shape[3],()=>`Error in conv2dDerInput: depth of output (${h}) must match output depth for filter ${n.shape[3]}.`),Hu("conv2dDerInput",a,i);const p={dy:u,filter:n},d={strides:r,pad:a,dataFormat:s,dimRoundingMode:i,inputShape:o},f=xs.runKernel(_t,p,d);return l?qu(f,[f.shape[1],f.shape[2],f.shape[3]]):f}});var bl=Ms({conv2dTranspose_:function(e,t,n,r,a,s){const i=Rs(e,"x","conv2dTranspose"),o=Rs(t,"filter","conv2dTranspose");return yl(n,i,o,r,a,"NHWC",s)}});var vl=Ms({conv3d_:function(e,t,n,r,a="NDHWC",s=[1,1,1]){const i=Rs(e,"x","conv3d"),o=Rs(t,"filter","conv3d");let u=i,l=!1;4===i.rank&&(l=!0,u=qu(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),ue(5===u.rank,()=>`Error in conv3d: input must be rank 5, but got rank ${u.rank}.`),ue(5===o.rank,()=>`Error in conv3d: filter must be rank 5, but got rank ${o.rank}.`),ue(u.shape[4]===o.shape[3],()=>`Error in conv3d: depth of input (${u.shape[4]}) must match input depth for filter ${o.shape[3]}.`),ue(ju(n,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),ue("NDHWC"===a,()=>`Error in conv3d: got dataFormat of ${a} but only NDHWC is currently supported.`);const c={x:u,filter:o},h={strides:n,pad:r,dataFormat:a,dilations:s},p=xs.runKernel(Ft,c,h);return l?qu(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}});var xl=Ms({conv3DBackpropInput_:function(e,t,n,r,a){ue(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let s=e,i=t,o=!1;4===t.rank&&(o=!0,i=qu(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,e[0],e[1],e[2],e[3]]);const u=s[4],l=i.shape[4];ue(5===s.length,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),ue(5===i.rank,()=>"Error in conv3dDerInput: dy must be rank 5, but got rank "+i.rank),ue(5===n.rank,()=>"Error in conv3dDerInput: filter must be rank 5, but got rank "+n.rank),ue(u===n.shape[3],()=>`Error in conv3dDerInput: depth of input (${u}) must match input depth for filter ${n.shape[3]}.`),ue(l===n.shape[4],()=>`Error in conv3dDerInput: depth of output (${l}) must match output depth for filter ${n.shape[4]}.`);const c={dy:i,filter:n},h={pad:a,strides:r,inputShape:s},p=xs.runKernel(Dt,c,h);return o?qu(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}});var wl=Ms({conv3dTranspose_:function(e,t,n,r,a){const s=Rs(e,"x","conv3dTranspose"),i=Rs(t,"filter","conv3dTranspose");return xl(n,s,i,r,a)}});var kl=Ms({cos_:function(e){const t={x:Rs(e,"x","cos","float32")};return xs.runKernel(Ot,t)}});var Sl=Ms({cosh_:function(e){const t={x:Rs(e,"x","cosh","float32")};return xs.runKernel(Lt,t)}});var Il=Ms({cumsum_:function(e,t=0,n=!1,r=!1){const a={x:Rs(e,"x","cumsum")},s={axis:t,exclusive:n,reverse:r};return xs.runKernel(Pt,a,s)}});var Al=Ms({denseBincount_:function(e,t,n,r=!1){const a=Rs(e,"x","denseBincount"),s=Rs(t,"weights","denseBincount");ue("int32"===a.dtype,()=>"Error in denseBincount: input dtype must be int32, but got "+a.dtype),ue(a.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${a.rank}.`),ue(n>=0,()=>`size must be non-negative, but got ${n}.`),ue(s.size===a.size||0===s.size,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${a.shape}, weights shape: ${s.shape}.`);const i={x:a,weights:s},o={size:n,binaryOutput:r};return xs.runKernel(Bt,i,o)}});var El=Ms({depthToSpace_:function(e,t,n="NHWC"){const r=Rs(e,"x","depthToSpace","float32"),a="NHWC"===n?r.shape[1]:r.shape[2],s="NHWC"===n?r.shape[2]:r.shape[3],i="NHWC"===n?r.shape[3]:r.shape[1];ue(t>1,()=>"blockSize should be > 1 for depthToSpace, but was: "+t),ue(a*t>=0,()=>`Negative dimension size caused by overflow when multiplying\n    ${a} and ${t}  for depthToSpace with input shape\n    ${r.shape}`),ue(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying\n    ${s} and ${t} for depthToSpace with input shape\n        ${r.shape}`),ue(i%(t*t)==0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${r.shape}`);const o={x:r},u={blockSize:t,dataFormat:n};return xs.runKernel(zt,o,u)}});var Nl=Ms({depthwiseConv2d_:function(e,t,n,r,a="NHWC",s=[1,1],i){const o=Rs(e,"x","depthwiseConv2d","float32"),u=Rs(t,"filter","depthwiseConv2d","float32");let l=o,c=!1;3===o.rank&&(c=!0,l=qu(o,[1,o.shape[0],o.shape[1],o.shape[2]])),ue(4===l.rank,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${l.rank}.`),ue(4===u.rank,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${u.rank}.`),ue(l.shape[3]===u.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${l.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),Hu("depthwiseConv2d",r,i);const h={x:l,filter:u},p={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i},d=xs.runKernel(Ut,h,p);return c?qu(d,[d.shape[1],d.shape[2],d.shape[3]]):d}});var Cl=Ms({diag_:function(e){const t={x:Rs(e,"x","diag")};return xs.runKernel(jt,t)}});var Tl=Ms({dilation2d_:function(e,t,n,r,a=[1,1],s="NHWC"){const i=Rs(e,"x","dilation2d"),o=Rs(t,"filter","dilation2d");ue(3===i.rank||4===i.rank,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),ue(3===o.rank,()=>`Error in dilation2d: filter must be rank 3, but got rank ${o.rank}.`),ue("NHWC"===s,()=>"Error in dilation2d: Only NHWC is currently supported, but got dataFormat of "+s);let u=i,l=!1;3===i.rank&&(u=qu(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=!0);const c={x:u,filter:o},h={strides:n,pad:r,dilations:a},p=xs.runKernel(Gt,c,h);return l?qu(p,[p.shape[1],p.shape[2],p.shape[3]]):p}});var Rl=Ms({equal_:function(e,t){let n=Rs(e,"a","equal","string_or_numeric"),r=Rs(t,"b","equal","string_or_numeric");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(Jt,a)}});var _l=Ms({where_:function(e,t,n){const r=Rs(t,"a","where"),a=Rs(n,"b","where"),s=Rs(e,"condition","where","bool"),i=Zi(Zi(s.shape,r.shape),a.shape),o={condition:ul(s,i),t:ul(r,i),e:ul(a,i)};return xs.runKernel(vr,o)}});var Fl=Ms({zerosLike_:function(e){const t={x:Rs(e,"x","zerosLike")};return xs.runKernel(Jr,t)}});var Ml=Ms({divNoNan_:function(e,t){let n=Rs(e,"a","div"),r=Rs(t,"b","div");[n,r]=ps(n,r);const a=bu(n,r),s=Fl(a),i=Rl(r,s);return _l(i,s,a)}});var Dl=Ms({dot_:function(e,t){const n=Rs(e,"t1","dot"),r=Rs(t,"t2","dot");ue(!(1!==n.rank&&2!==n.rank||1!==r.rank&&2!==r.rank),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${r.rank}.`);const a=1===n.rank?n.size:n.shape[1],s=1===r.rank?r.size:r.shape[0];if(ue(a===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${a} and ${s}.`),1===n.rank&&1===r.rank){const e=qu(n,[1,-1]),t=qu(r,[-1,1]),a=Gi(e,t);return qu(a,[])}if(1===n.rank&&2===r.rank){const e=qu(n,[1,-1]),t=qu(r,[r.shape[0],r.shape[1]]),a=Gi(e,t);return qu(a,[a.size])}if(2===n.rank&&1===r.rank){const e=qu(r,[-1,1]),t=Gi(n,e);return qu(t,[t.size])}{const e=qu(r,[r.shape[0],r.shape[1]]);return Gi(n,e)}}});var Ol=Ms({einsum_:function(e,...t){const n=t.map((e,t)=>Rs(e,"tensors"+t,"einsum")),r={equation:e};return xs.runKernel(Yt,n,r)}});var Ll=Ms({elu_:function(e){const t={x:Rs(e,"x","elu","float32")};return xs.runKernel(Xt,t)}});var Pl=Ms({erf_:function(e){let t=Rs(e,"x","erf");ue("int32"===t.dtype||"float32"===t.dtype,()=>"Input dtype must be `int32` or `float32`."),"int32"===t.dtype&&(t=Ai(t,"float32"));const n={x:t};return xs.runKernel(Zt,n)}});var $l=Ms({exp_:function(e){const t={x:Rs(e,"x","exp")};return xs.runKernel(en,t)}});var Bl=Ms({expandDims_:function(e,t=0){const n=Rs(e,"x","expandDims","string_or_numeric");ue(t<=n.rank,()=>"Axis must be <= rank of the tensor");const r={input:n},a={dim:t};return xs.runKernel(tn,r,a)}});var zl=Ms({expm1_:function(e){const t={x:Rs(e,"x","expm1")};return xs.runKernel(nn,t)}});var Ul=Ms({tile_:function(e,t){const n=Rs(e,"x","tile","string_or_numeric");ue(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of reps ${t}.`);const r={x:n},a={reps:t};return xs.runKernel(Hr,r,a)}});var Wl=Ms({eye_:function(e,t,n,r="float32"){null==t&&(t=e);const a=Ii([e,t],r),s=e<=t?e:t;for(let e=0;e<s;++e)a.set(1,e,e);const i=qu(a.toTensor(),[e,t]);if(null==n)return i;if(1===n.length)return Ul(Bl(i,0),[n[0],1,1]);if(2===n.length)return Ul(Bl(Bl(i,0),0),[n[0],n[1],1,1]);if(3===n.length)return Ul(Bl(Bl(Bl(i,0),0),0),[n[0],n[1],n[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${n.length}D.`)}});function Vl(e,t,n){const r={shape:e,value:t,dtype:n};return xs.runKernel(an,{},r)}var jl=Ms({floor_:function(e){const t={x:Rs(e,"x","floor","float32")};return xs.runKernel(on,t)}});var Gl=Ms({gather_:function(e,t,n=0,r=0){const a={x:Rs(e,"x","gather"),indices:Rs(t,"indices","gather","int32")},s={axis:n,batchDims:r};return xs.runKernel(cn,a,s)}});var Hl=Ms({greater_:function(e,t){let n=Rs(e,"a","greater","string_or_numeric"),r=Rs(t,"b","greater","string_or_numeric");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(pn,a)}});var ql=Ms({greaterEqual_:function(e,t){let n=Rs(e,"a","greaterEqual","string_or_numeric"),r=Rs(t,"b","greaterEqual","string_or_numeric");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(dn,a)}});var Kl=Ms({imag_:function(e){const t={input:Rs(e,"input","imag")};return xs.runKernel(gn,t)}});var Yl=Ms({isFinite_:function(e){const t={x:Rs(e,"x","isFinite")};return xs.runKernel(yn,t)}});var Xl=Ms({isInf_:function(e){const t={x:Rs(e,"x","isInf")};return xs.runKernel(bn,t)}});var Ql=Ms({isNaN_:function(e){const t={x:Rs(e,"x","isNaN")};return xs.runKernel(vn,t)}});var Zl=Ms({leakyRelu_:function(e,t=.2){const n={x:Rs(e,"x","leakyRelu")},r={alpha:t};return xs.runKernel(xn,n,r)}});var Jl=Ms({less_:function(e,t){let n=Rs(e,"a","less","string_or_numeric"),r=Rs(t,"b","less","string_or_numeric");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(wn,a)}});var ec=Ms({lessEqual_:function(e,t){let n=Rs(e,"a","lessEqual","string_or_numeric"),r=Rs(t,"b","lessEqual","string_or_numeric");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(kn,a)}});function tc(e,t,n){if(n<=0)throw new Error("The number of values should be positive.");const r={start:e,stop:t,num:n};return xs.runKernel(Sn,{},r)}var nc=Ms({localResponseNormalization_:function(e,t=5,n=1,r=1,a=.5){const s=Rs(e,"x","localResponseNormalization");ue(4===s.rank||3===s.rank,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got\n               rank ${s.rank}.`),ue(me(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,o=!1;3===s.rank&&(o=!0,i=qu(s,[1,s.shape[0],s.shape[1],s.shape[2]]));const u={x:i},l={depthRadius:t,bias:n,alpha:r,beta:a},c=xs.runKernel(Rn,u,l);return o?qu(c,[c.shape[1],c.shape[2],c.shape[3]]):c}});var rc=Ms({log_:function(e){const t={x:Rs(e,"x","log","float32")};return xs.runKernel(In,t)}});var ac=Ms({log1p_:function(e){const t={x:Rs(e,"x","log1p")};return xs.runKernel(An,t)}});function sc(e){return ue(Le(e),()=>"The f passed in grad(f) must be a function"),(t,n)=>{const r=Rs(t,"x","tf.grad","string_or_numeric"),a=null!=n?Rs(n,"dy","tf.grad"):null;return xs.tidy(()=>{const{value:t,grads:n}=xs.gradients(()=>e(r),[r],a);return null!=a&&le(t.shape,a.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),hc(n),n[0]})}}function ic(e){return ue(Le(e),()=>"The f passed in grads(f) must be a function"),(t,n)=>{ue(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");const r=_s(t,"args","tf.grads","string_or_numeric"),a=null!=n?Rs(n,"dy","tf.grads"):null;return xs.tidy(()=>{const{value:t,grads:n}=xs.gradients(()=>e(...r),r,a);return null!=a&&le(t.shape,a.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),hc(n),n})}}function oc(e){return ue(Le(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,n)=>{ue(t instanceof Ya,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),ue(null==n||n instanceof Ya,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");const{grads:r,value:a}=xs.gradients(()=>e(t),[t],n);return hc(r),{grad:r[0],value:a}}}function uc(e){return ue(Le(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,n)=>{ue(Array.isArray(t)&&t.every(e=>e instanceof Ya),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),ue(null==n||n instanceof Ya,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");const r=xs.gradients(()=>e(...t),t,n);return null!=n&&le(r.value.shape,n.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),hc(r.grads),r}}function lc(e,t){ue(Le(e),()=>"The f passed in variableGrads(f) must be a function"),ue(null==t||Array.isArray(t)&&t.every(e=>e instanceof Qa),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");const n=null!=t;if(!n){t=[];for(const e in xs.registeredVariables)t.push(xs.registeredVariables[e])}const r=n?t.filter(e=>!e.trainable):null,a=t.length;ue((t=t.filter(e=>e.trainable)).length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${a} variables is trainable.`);const{value:s,grads:i}=xs.gradients(e,t,null,!0);ue(i.some(e=>null!=e),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),ue(0===s.rank,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${s.rank} tensor`);const o={};return t.forEach((e,t)=>{null!=i[t]&&(o[e.name]=i[t])}),null!=r&&r.forEach(e=>o[e.name]=null),{value:s,grads:o}}function cc(e){return xs.customGrad(e)}function hc(e){if(e.filter(e=>null==e).length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that\n    the f you passed encloses all operations that lead from x to y.")}var pc=Ms({neg_:function(e){const t={x:Rs(e,"x","neg")};return xs.runKernel(Hn,t)}});var dc=Ms({softplus_:function(e){const t={x:Rs(e,"x","softplus")};return xs.runKernel(Er,t)}});var fc=Ms({logSigmoid_:function(e){const t=Rs(e,"x","logSigmoid");return cc(e=>({value:pc(dc(pc(e))),gradFunc:t=>vu(t,Qu(pc(e)))}))(t)}});var mc=Ms({max_:function(e,t=null,n=!1){const r={x:Rs(e,"x","max")},a={reductionIndices:t,keepDims:n};return xs.runKernel(Fn,r,a)}});var gc=Ms({sub_:function(e,t){let n=Rs(e,"a","sub"),r=Rs(t,"b","sub");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(Vr,a)}});var yc=Ms({sum_:function(e,t=null,n=!1){let r=Rs(e,"x","sum");"bool"===r.dtype&&(r=Ai(r,"int32"));const a={x:r},s={axis:t,keepDims:n};return xs.runKernel(Cr,a,s)}});var bc=Ms({logSoftmax_:function(e,t=-1){const n=Rs(e,"logits","logSoftmax");if(-1===t&&(t=n.rank-1),t!==n.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and axis was ${t}`);return cc((e,n)=>{const r=mc(e,t,!0),a=gc(e,r),s=gc(Ai(a,"float32"),rc(yc($l(a),t,!0)));n([s]);return{value:s,gradFunc:(e,n)=>{const[r]=n,a=$l(r);return gc(e,vu(yc(e,t,!0),a))}}})(n)}});function vc(e,t){for(let n=0;n<e.length;++n)if(e[e.length-n-1]!==t-1-n)return!1;return!0}function xc(e,t,n){const r=e.length+t.length,a=[];let s=0,i=0;for(let o=0;o<r;o++)-1===n.indexOf(o)?a.push(e[s++]):a.push(t[i++]);return a}function wc(e,t){const n=[],r=e.length;for(let a=0;a<r;a++)-1===t.indexOf(a)&&n.push(e[a]);return[n,t.map(t=>e[t])]}function kc(e,t){return xc(e,t.map(e=>1),t)}function Sc(e,t,n){ue(vc(t,n),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)}function Ic(e,t){if(vc(e,t))return null;const n=[];for(let r=0;r<t;++r)-1===e.indexOf(r)&&n.push(r);return e.forEach(e=>n.push(e)),n}function Ac(e){return e.map((e,t)=>[t,e]).sort((e,t)=>e[1]-t[1]).map(e=>e[0])}function Ec(e,t){const n=[];for(let r=t-e;r<t;++r)n.push(r);return n}var Nc=Ms({logSumExp_:function(e,t=null,n=!1){const r=Rs(e,"x","logSumExp"),a=ke(t,r.shape),s=mc(r,a,!0),i=gc(r,s),o=$l(i),u=yc(o,a),l=rc(u),c=gu(qu(s,l.shape),l);if(n){const e=kc(c.shape,a);return qu(c,e)}return c}});var Cc=Ms({logicalAnd_:function(e,t){const n=Rs(e,"a","logicalAnd","bool"),r=Rs(t,"b","logicalAnd","bool");Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(En,a)}});var Tc=Ms({logicalNot_:function(e){const t={x:Rs(e,"x","logicalNot","bool")};return xs.runKernel(Nn,t)}});var Rc=Ms({logicalOr_:function(e,t){const n=Rs(e,"a","logicalOr","bool"),r=Rs(t,"b","logicalOr","bool");Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(Cn,a)}});var _c=Ms({logicalXor_:function(e,t){const n=Rs(e,"a","logicalXor","bool"),r=Rs(t,"b","logicalXor","bool");return Zi(n.shape,r.shape),Cc(Rc(e,t),Tc(Cc(e,t)))}});var Fc=Ms({maxPool_:function(e,t,n,r,a){const s=Rs(e,"x","maxPool");let i=s,o=!1;3===s.rank&&(o=!0,i=qu(s,[1,s.shape[0],s.shape[1],s.shape[2]])),ue(4===i.rank,()=>`Error in maxPool: input must be rank 4 but got rank ${i.rank}.`),ue(ju(n,1),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '1'`),Hu("maxPool",r,a);const u={x:i},l={filterSize:t,strides:n,pad:r,dimRoundingMode:a},c=xs.runKernel(Dn,u,l);return o?qu(c,[c.shape[1],c.shape[2],c.shape[3]]):c}});var Mc=Ms({maxPool3d_:function(e,t=[1,1,1],n,r,a,s="NDHWC"){const i=Rs(e,"x","maxPool3d");let o=i,u=!1;4===i.rank&&(u=!0,o=qu(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),ue(5===o.rank,()=>`Error in maxPool3d: x must be rank 5 but got rank ${o.rank}.`),ue("NDHWC"===s,()=>"Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of "+s),Hu("maxPool3d",r,a);const l={x:o},c={filterSize:t,strides:n,pad:r,dimRoundingMode:a,dataFormat:s},h=xs.runKernel(Ln,l,c);return u?qu(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}});var Dc=Ms({maxPoolWithArgmax_:function(e,t,n,r,a=!1){const s={x:Rs(e,"x","maxPoolWithArgmax")},i={filterSize:t,strides:n,pad:r,includeBatchInIndex:a},o=xs.runKernel($n,s,i);return{result:o[0],indexes:o[1]}}});var Oc=Ms({maximum_:function(e,t){let n=Rs(e,"a","maximum"),r=Rs(t,"b","maximum");[n,r]=ps(n,r),"bool"===n.dtype&&(n=Ai(n,"int32"),r=Ai(r,"int32")),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(Mn,a)}});var Lc=Ms({mean_:function(e,t=null,n=!1){const r={x:Rs(e,"x","mean")},a={axis:t,keepDims:n};return xs.runKernel(Bn,r,a)}});function Pc(e,t="float32"){if("complex64"===t){const t=Pc(e,"float32"),n=Pc(e,"float32");return Ds(t,n)}const n=Ue(pe(e),t);return xs.makeTensor(n,e,t)}function $c(e,t="float32"){if("complex64"===t){const t=$c(e,"float32"),n=Pc(e,"float32");return Ds(t,n)}const n=ze(pe(e),t);return xs.makeTensor(n,e,t)}function Bc(e,t,{indexing:n="xy"}={}){if("xy"!==n&&"ij"!==n)throw new TypeError(n+" is not a valid third argument to meshgrid");if(void 0===e)return[];let r=Rs(e,"x","meshgrid",e instanceof Ya?e.dtype:"float32");if(void 0===t)return[r];let a=Rs(t,"y","meshgrid",t instanceof Ya?t.dtype:"float32");const s=pe(r.shape),i=pe(a.shape);return"xy"===n?(r=qu(r,[1,-1]),a=qu(a,[-1,1]),[Gi($c([i,1],r.dtype),r),Gi(a,$c([1,s],a.dtype))]):(r=qu(r,[-1,1]),a=qu(a,[1,-1]),[Gi(r,$c([1,i],r.dtype)),Gi($c([s,1],a.dtype),a)])}var zc=Ms({min_:function(e,t=null,n=!1){const r={x:Rs(e,"x","min")},a={axis:t,keepDims:n};return xs.runKernel(zn,r,a)}});var Uc=Ms({minimum_:function(e,t){let n=Rs(e,"a","minimum"),r=Rs(t,"b","minimum");[n,r]=ps(n,r),"bool"===n.dtype&&(n=Ai(n,"int32"),r=Ai(r,"int32")),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(Un,a)}});var Wc=Ms({mirrorPad_:function(e,t,n){ue("reflect"===n||"symmetric"===n,()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${n}.`);const r=Rs(e,"x","mirrorPad");if(0===r.rank)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");ue(t.length===r.rank,()=>`Padding doesn't match input. Must be ${r.rank}. Got ${t.length}.`);const a="reflect"===n?1:0;for(let e=0;e<r.rank;e++)ue(2===t[e].length,()=>"Invalid number of paddings. Must be length of 2 each."),ue(t[e][0]>=0&&t[e][0]<=r.shape[e]-a&&t[e][1]>=0&&t[e][1]<=r.shape[e]-a,()=>`Padding in dimension ${e} cannot be greater than or equal to ${r.shape[e]-a} or less than 0 for input of shape ${r.shape}`);const s={paddings:t,mode:n},i={x:r};return xs.runKernel(Wn,i,s)}});var Vc=Ms({mod_:function(e,t){let n=Rs(e,"a","mod"),r=Rs(t,"b","mod");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(Vn,a)}});var jc=Ms({square_:function(e){const t=Rs(e,"x","square");return xs.runKernel("Square",{x:t},{})}});var Gc=Ms({moments_:function(e,t=null,n=!1){const r=ke(t,(e=Rs(e,"x","moments")).shape),a=Lc(e,r,n);let s=a.shape;n||(s=kc(a.shape,r));const i=jc(gc(Ai(e,"float32"),qu(a,s)));return{mean:a,variance:Lc(i,r,n)}}});var Hc=Ms({multiRNNCell_:function(e,t,n,r){const a=Rs(t,"data","multiRNNCell"),s=_s(n,"c","multiRNNCell"),i=_s(r,"h","multiRNNCell");let o=a;const u=[];for(let t=0;t<e.length;t++){const n=e[t](o,s[t],i[t]);u.push(n[0]),u.push(n[1]),o=n[1]}const l=[],c=[];for(let e=0;e<u.length;e+=2)l.push(u[e]),c.push(u[e+1]);return[l,c]}});var qc=Ms({multinomial_:function(e,t,n,r=!1){const a=Rs(e,"logits","multinomial"),s=a.size,i=a.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error("Rank of probabilities must be 1 or 2, but is "+i);n=n||Math.random();const o={logits:1===i?qu(a,[1,-1]):a},u={numSamples:t,seed:n,normalized:r},l=xs.runKernel(jn,o,u);return 1===i?qu(l,[l.size]):l}});var Kc=Ms({notEqual_:function(e,t){let n=Rs(e,"a","notEqual","string_or_numeric"),r=Rs(t,"b","notEqual","string_or_numeric");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(qn,a)}});var Yc=Ms({onesLike_:function(e){const t={x:Rs(e,"x","onesLike")};return xs.runKernel(Qn,t)}});var Xc=Ms({outerProduct_:function(e,t){const n=Rs(e,"v1","outerProduct"),r=Rs(t,"v2","outerProduct");ue(1===n.rank&&1===r.rank,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${r.rank}.`);const a=qu(n,[-1,1]),s=qu(r,[1,-1]);return Gi(a,s)}});var Qc=Ms({pad_:function(e,t,n=0){const r=Rs(e,"x","pad");if(0===r.rank)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");const a={paddings:t,constantValue:n},s={x:r};return xs.runKernel(er,s,a)}});var Zc=Ms({pad1d_:function(e,t,n=0){return ue(2===t.length,()=>"Invalid number of paddings. Must be length of 2."),Qc(e,[t],n)}});var Jc=Ms({pad2d_:function(e,t,n=0){return ue(2===t.length&&2===t[0].length&&2===t[1].length,()=>"Invalid number of paddings. Must be length of 2 each."),Qc(e,t,n)}});var eh=Ms({pad3d_:function(e,t,n=0){return ue(3===t.length&&2===t[0].length&&2===t[1].length&&2===t[2].length,()=>"Invalid number of paddings. Must be length of 2 each."),Qc(e,t,n)}});var th=Ms({pad4d_:function(e,t,n=0){return ue(4===t.length&&2===t[0].length&&2===t[1].length&&2===t[2].length&&2===t[3].length,()=>"Invalid number of paddings. Must be length of 2 each."),Qc(e,t,n)}});var nh=Ms({spaceToBatchND_:function(e,t,n){const r=Rs(e,"x","spaceToBatchND");ue(r.rank>=1+t.length,()=>`input rank ${r.rank} should be > than [blockShape] ${t.length}`),ue(n.length===t.length,()=>`paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`),ue(r.shape.reduce((e,r,a)=>a>0&&a<=t.length?e&&(r+n[a-1][0]+n[a-1][1])%t[a-1]==0:e,!0),()=>`input spatial dimensions ${r.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);const a={x:r},s={blockShape:t,paddings:n};return xs.runKernel(Tr,a,s)}});var rh=Ms({pool_:function(e,t,n,r,a,s,i){null==a&&(a=[1,1]),null==s&&(s=1),0===r&&(r="valid");const o=Rs(e,"x","maxPool");let u=o,l=!1;3===o.rank&&(l=!0,u=qu(o,[1,o.shape[0],o.shape[1],o.shape[2]])),ue(ju(s,a),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${a}'`);const c=Du(u.shape,t,s,a,r),h=[c.dilationHeight,c.dilationWidth];let p;p="same"===r?function(e,t){const n=e.map((e,n)=>e+(e-1)*(t[n]-1)).map(e=>e-1),r=n.map(e=>Math.floor(e/2)),a=n.map((e,t)=>e-r[t]);return n.map((e,t)=>[r[t],a[t]])}([c.filterHeight,c.filterWidth],h):[[0,0],[0,0]];const d=1===h[0]&&1===h[1],[f,m]=function(e,t,n){const r=n.map(e=>e[0]),a=n.map(e=>e[1]),s=e.concat(r,a),i=t.map((e,t)=>(e-s[t]%e)%e),o=a.map((e,t)=>e+i[t]),u=t.map((e,t)=>[r[t],o[t]]),l=t.map((e,t)=>[0,i[t]]);return[u,l]}([c.inHeight,c.inWidth],h,p),g=d?r:"valid",y=d?u:nh(u,h,f),b=("avg"===n?()=>Ku(y,t,s,g,i):()=>Fc(y,t,s,g,i))(),v=d?b:tl(b,h,m);return l?qu(v,[v.shape[1],v.shape[2],v.shape[3]]):v}});var ah=Ms({pow_:function(e,t){let n=Rs(e,"base","pow"),r=Rs(t,"exp","pow");[n,r]=ps(n,r);const a={a:n,b:r};return xs.runKernel(nr,a)}});var sh=Ms({prelu_:function(e,t){const n={x:Rs(e,"x","prelu"),alpha:Rs(t,"alpha","prelu")};return xs.runKernel(rr,n)}});var ih=Ms({prod_:function(e,t=null,n=!1){let r=Rs(e,"x","prod");"bool"===r.dtype&&(r=Ai(r,"int32"));const a={x:r},s={axis:t,keepDims:n};return xs.runKernel(ar,a,s)}});var oh=Ms({rand_:function(e,t,n){const r=pe(e);let a=null;if(null==n||"float32"===n)a=new Float32Array(r);else if("int32"===n)a=new Int32Array(r);else{if("bool"!==n)throw new Error("Unknown data type "+n);a=new Uint8Array(r)}for(let e=0;e<r;e++)a[e]=t();return xs.makeTensor(a,e,n)}}),uh=k(D()),lh=class{constructor(e,t,n,r,a){this.mean=e,this.stdDev=t,this.dtype=n,this.nextVal=NaN,this.truncated=r,this.truncated&&(this.upper=this.mean+2*this.stdDev,this.lower=this.mean-2*this.stdDev);const s=a||Math.random();this.random=uh.alea(s.toString())}nextValue(){if(!isNaN(this.nextVal)){const e=this.nextVal;return this.nextVal=NaN,e}let e,t,n=!1;for(;!n;){let r,a,s;do{r=2*this.random()-1,a=2*this.random()-1,s=r*r+a*a}while(s>=1||0===s);const i=Math.sqrt(-2*Math.log(s)/s);e=this.mean+this.stdDev*r*i,t=this.mean+this.stdDev*a*i,this.truncated&&!this.isValidTruncated(e)||(n=!0)}return this.truncated&&!this.isValidTruncated(t)||(this.nextVal=this.convertValue(t)),this.convertValue(e)}convertValue(e){return null==this.dtype||"float32"===this.dtype?e:Math.round(e)}isValidTruncated(e){return e<=this.upper&&e>=this.lower}};var ch=Ms({randomGamma_:function(e,t,n=1,r="float32",a){if(null==n&&(n=1),null==r&&(r="float32"),"float32"!==r&&"int32"!==r)throw new Error("Unsupported data type "+r);const s=new class{constructor(e,t,n,r){this.alpha=e,this.beta=1/t,this.dtype=n;const a=r||Math.random();this.randu=uh.alea(a.toString()),this.randn=new lh(0,1,n,!1,this.randu()),this.d=e<1?e+2/3:e-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let e,t,n,r,a,s;for(;;){do{r=this.randn.nextValue(),s=1+this.c*r}while(s<=0);if(s*=s*s,e=r*r,t=1-.331*e*e,n=.5*e+this.d*(1-s+Math.log(s)),a=this.randu(),a<t||Math.log(a)<n)break}return s=1/this.beta*this.d*s,this.alpha<1&&(s*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(s)}convertValue(e){return"float32"===this.dtype?e:Math.round(e)}}(t,n,r,a),i=Ii(e,r);for(let e=0;e<i.values.length;e++)i.values[e]=s.nextValue();return i.toTensor()}});var hh=Ms({randomNormal_:function(e,t=0,n=1,r,a){if(null!=r&&"bool"===r)throw new Error("Unsupported data type "+r);const s=new lh(t,n,r,!1,a),i=Ii(e,r);for(let e=0;e<i.values.length;e++)i.values[e]=s.nextValue();return i.toTensor()}});var ph=Ms({randomUniform_:function(e,t=0,n=1,r="float32",a){const s=Ii(e,r),i=new class{constructor(e=0,t=1,n,r){if(this.canReturnFloat=()=>null==this.dtype||"float32"===this.dtype,this.min=e,this.range=t-e,this.dtype=n,null==r&&(r=Math.random()),"number"==typeof r&&(r=r.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=uh.alea(r)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}}(t,n,null,a);for(let e=0;e<s.values.length;e++)s.values[e]=i.nextValue();return s.toTensor()}});function dh(e,t,n=1,r="float32"){if(0===n)throw new Error("Cannot have a step of zero");const a={start:e,stop:t,step:n,dtype:r};return xs.runKernel(sr,{},a)}var fh=Ms({real_:function(e){const t={input:Rs(e,"input","real")};return xs.runKernel(ir,t)}});var mh=Ms({reciprocal_:function(e){const t={x:Rs(e,"x","reciprocal")};return xs.runKernel(or,t)}});var gh=Ms({relu_:function(e){const t={x:Rs(e,"x","relu")};return xs.runKernel(ur,t)}});var yh=Ms({relu6_:function(e){const t={x:Rs(e,"x","relu6")};return xs.runKernel(fr,t)}});var bh=Ms({reverse_:function(e,t){const n={x:Rs(e,"x","reverse")},r={dims:t};return xs.runKernel(mr,n,r)}});var vh=Ms({reverse1d_:function(e){const t=Rs(e,"x","reverse");return ue(1===t.rank,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),bh(t,0)}});var xh=Ms({reverse2d_:function(e,t){const n=Rs(e,"x","reverse");return ue(2===n.rank,()=>`Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`),bh(n,t)}});var wh=Ms({reverse3d_:function(e,t){const n=Rs(e,"x","reverse");return ue(3===n.rank,()=>`Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`),bh(n,t)}});var kh=Ms({reverse4d_:function(e,t){const n=Rs(e,"x","reverse");return ue(4===n.rank,()=>`Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`),bh(n,t)}});var Sh=Ms({round_:function(e){const t={x:Rs(e,"x","round")};return xs.runKernel(gr,t)}});var Ih=Ms({rsqrt_:function(e){const t={x:Rs(e,"x","rsqrt","float32")};return xs.runKernel(yr,t)}});function Ah(e,t){if((Te(e)&&"string"!==t||Array.isArray(e))&&"complex64"!==t)throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if("string"===t&&Te(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return Os(e,[],[],t)}var Eh=Ms({selu_:function(e){const t={x:Rs(e,"x","selu")};return xs.runKernel(xr,t)}});var Nh=Ms({separableConv2d_:function(e,t,n,r,a,s=[1,1],i="NHWC"){const o=Rs(e,"x","separableConv2d"),u=Rs(t,"depthwiseFilter","separableConv2d"),l=Rs(n,"pointwiseFilter","separableConv2d");let c=o,h=!1;if(3===o.rank&&(h=!0,c=qu(o,[1,o.shape[0],o.shape[1],o.shape[2]])),"NCHW"===i)throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");ue(4===c.rank,()=>`Error in separableConv2d: input must be rank 4, but got rank ${c.rank}.`),ue(4===u.rank,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${u.rank}.`),ue(4===l.rank,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${u.rank}.`),ue(1===l.shape[0],()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${l.shape[0]}.`),ue(1===l.shape[1],()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${l.shape[1]}.`);const p=u.shape[2],d=u.shape[3];ue(l.shape[2]===p*d,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${p*d}, but got ${l.shape[2]}.`);const f=Nl(c,u,r,a,i,s),m=ml(f,l,1,"valid",i);return h?qu(m,[m.shape[1],m.shape[2],m.shape[3]]):m}});var Ch=async function(e,t){const n=Rs(e,"x","setdiff1d"),r=Rs(t,"y","setdiff1d");ue(n.dtype===r.dtype,()=>`x and y should have the same dtype, but got x (${n.dtype}) and y (${r.dtype}).`),ue(1===n.rank,()=>`x should be 1D tensor, but got x (${n.shape}).`),ue(1===r.rank,()=>`y should be 1D tensor, but got y (${r.shape}).`);const a=await n.data(),s=await r.data(),i=new Set(s);let o=0;for(let e=0;e<a.length;e++)i.has(a[e])||o++;const u=new Ha([o],n.dtype),l=new Ha([o],"int32");for(let e=0,t=0;e<a.length;e++)i.has(a[e])||(u.values[t]=a[e],l.values[t]=e,t++);return[u.toTensor(),l.toTensor()]};var Th=Ms({sign_:function(e){const t={x:Rs(e,"x","sign")};return xs.runKernel(Ir,t)}});var Rh=Ms({sin_:function(e){const t={x:Rs(e,"x","sin","float32")};return xs.runKernel(kr,t)}});var _h=Ms({sinh_:function(e){const t={x:Rs(e,"x","sinh")};return xs.runKernel(Sr,t)}});var Fh=Ms({slice1d_:function(e,t,n){const r=Rs(e,"x","slice1d");return ue(1===r.rank,()=>`slice1d expects a rank-1 tensor, but got a rank-${r.rank} tensor`),Zu(r,[t],[n])}});var Mh=Ms({slice2d_:function(e,t,n){const r=Rs(e,"x","slice2d");return ue(2===r.rank,()=>`slice2d expects a rank-2 tensor, but got a rank-${r.rank} tensor`),Zu(r,t,n)}});var Dh=Ms({slice3d_:function(e,t,n){const r=Rs(e,"x","slice3d");return ue(3===r.rank,()=>`slice3d expects a rank-3 tensor, but got a rank-${r.rank} tensor`),Zu(r,t,n)}});var Oh=Ms({slice4d_:function(e,t,n){const r=Rs(e,"x","slice4d");return ue(4===r.rank,()=>`slice4d expects a rank-4 tensor, but got a rank-${r.rank} tensor`),Zu(r,t,n)}});var Lh=Ms({softmax_:function(e,t=-1){const n=Rs(e,"logits","softmax","float32");if(-1===t&&(t=n.rank-1),t!==n.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and dim was ${t}`);const r={logits:n},a={dim:t};return xs.runKernel(_r,r,a)}});var Ph=Ms({fft_:function(e){ue("complex64"===e.dtype,()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);const t={input:e};return xs.runKernel(rn,t)}});var $h=Ms({ifft_:function(e){ue("complex64"===e.dtype,()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);const t={input:e};return xs.runKernel(mn,t)}});var Bh=Ms({irfft_:function(e){const t=e.shape[e.shape.length-1],n=e.size/t;let r;if(t<=2){const a=qu(e,[n,t]);r=$h(a)}else{const a=[n,2*(t-1)],s=qu(fh(e),[n,t]),i=qu(Kl(e),[n,t]),o=bh(Zu(s,[0,1],[n,t-2]),1),u=vu(bh(Zu(i,[0,1],[n,t-2]),1),Ah(-1)),l=Xu([s,o],1),c=Xu([i,u],1),h=qu(Ds(l,c),[a[0],a[1]]);r=$h(h)}if(r=fh(r),3===e.rank&&0!==e.shape[0]){const t=r,n=e.shape[0];r=qu(r,[n,r.shape[0]/n,r.shape[1]]),t.dispose()}return r}});var zh=Ms({split_:function(e,t,n=0){const r={x:Rs(e,"x","split")},a={numOrSizeSplits:t,axis:n};return xs.runKernel(Rr,r,a)}});var Uh=Ms({rfft_:function(e,t){ue("float32"===e.dtype,()=>"The dtype for rfft() must be real value but got "+e.dtype);let n=e.shape[e.shape.length-1];const r=e.size/n;let a;if(null!=t&&t<n){const r=e.shape.map(e=>0),s=e.shape.map(e=>e);s[e.shape.length-1]=t,a=Zu(e,r,s),n=t}else if(null!=t&&t>n){const r=e.shape.map(e=>e);r[e.shape.length-1]=t-n,a=Xu([e,Pc(r)],e.shape.length-1),n=t}else a=e;const s=Fl(a),i=qu(Ds(a,s),[r,n]),o=Ph(i),u=Math.floor(n/2)+1,l=fh(o),c=Kl(o),h=zh(l,[u,n-u],l.shape.length-1),p=zh(c,[u,n-u],c.shape.length-1),d=a.shape.slice();return d[a.shape.length-1]=u,qu(Ds(h[0],p[0]),d)}});var Wh=Ms({sqrt_:function(e){const t={x:Rs(e,"x","sqrt","float32")};return xs.runKernel(Nr,t)}});var Vh=Ms({squaredDifference_:function(e,t){let n=Rs(e,"a","squaredDifference"),r=Rs(t,"b","squaredDifference");[n,r]=ps(n,r),Zi(n.shape,r.shape);const a={a:n,b:r};return xs.runKernel(Pr,a,{})}});var jh=Ms({squeeze_:function(e,t){const n=Rs(e,"x","squeeze");return qu(n,Se(n.shape,t).newShape)}});var Gh=Ms({stack_:function(e,t=0){const n=_s(e,"tensors","stack","string_or_numeric");ue(n.length>=1,()=>"Pass at least one tensor to tf.stack"),n.length>0&&ue(t<=n[0].rank,()=>"Axis must be <= rank of the tensor");const r=n,a={axis:t};return xs.runKernel(Jn,r,a)}});var Hh=Ms({step_:function(e,t=0){const n={x:Rs(e,"x","step")},r={alpha:t};return xs.runKernel(ea,n,r)}});var qh=Ms({stridedSlice_:function(e,t,n,r,a=0,s=0,i=0,o=0,u=0){const l={x:Rs(e,"x","stridedSlice","string_or_numeric")},c={begin:t,end:n,strides:r,beginMask:a,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:u};return xs.runKernel(Br,l,c)}});var Kh=Ms({tan_:function(e){const t={x:Rs(e,"x","tan","float32")};return xs.runKernel(jr,t)}});function Yh(e,t){ce(e);const n=Cs(e,t);if(1!==n.length)throw new Error("tensor1d() requires values to be a flat/TypedArray");return Os(e,null,n,t)}function Xh(e,t,n){if(ce(e),null!=t&&2!==t.length)throw new Error("tensor2d() requires shape to have two numbers");const r=Cs(e,n);if(2!==r.length&&1!==r.length)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return Os(e,t,r,n)}function Qh(e,t,n){if(ce(e),null!=t&&4!==t.length)throw new Error("tensor4d() requires shape to have four numbers");const r=Cs(e,n);if(4!==r.length&&1!==r.length)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return Os(e,t,r,n)}function Zh(e,t,n){if(ce(e),null!=t&&5!==t.length)throw new Error("tensor5d() requires shape to have five numbers");const r=Cs(e,n);if(5!==r.length&&1!==r.length)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return Os(e,t,r,n)}function Jh(e,t,n){if(ce(e),null!=t&&6!==t.length)throw new Error("tensor6d() requires shape to have six numbers");const r=Cs(e,n);if(6!==r.length&&1!==r.length)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return Os(e,t=t||r,r,n)}var ep=Ms({topk_:function(e,t=1,n=!0){const r=Rs(e,"x","topk");if(0===r.rank)throw new Error("topk() expects the input to be of rank 1 or higher");const a=r.shape[r.shape.length-1];if(t<0)throw new Error("'k' passed to topk() must be >= 0 but got "+t);if(t>a)throw new Error(`'k' passed to topk() must be <= the last dimension (${a}) but got ${t}`);const s={x:r},i={k:t,sorted:n},[o,u]=xs.runKernel(qr,s,i);return{values:o,indices:u}}});var tp=Ms({truncatedNormal_:function(e,t=0,n=1,r,a){if(null!=r&&"bool"===r)throw new Error("Unsupported data type $ { dtype }");const s=new lh(t,n,r,!0,a),i=Ii(e,r);for(let e=0;e<i.values.length;e++)i.values[e]=s.nextValue();return i.toTensor()}});var np=Ms({unique_:function(e,t=0){const n=Rs(e,"x","unique","string_or_numeric");ue(n.rank>0,()=>"The input tensor must be at least 1D");const r={x:n},a={axis:t},[s,i]=xs.runKernel(Xr,r,a);return{values:s,indices:i}}});var rp=Ms({unsortedSegmentSum_:function(e,t,n){const r=Rs(e,"x","unsortedSegmentSum"),a=Rs(t,"segmentIds","unsortedSegmentSum","int32");ue(me(n),()=>"numSegments must be of dtype int");const s={x:r,segmentIds:a},i={numSegments:n};return xs.runKernel(Zr,s,i)}});var ap=Ms({unstack_:function(e,t=0){const n=Rs(e,"x","unstack","string_or_numeric");ue(t>=-n.shape.length&&t<n.shape.length,()=>`Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`);const r={value:n},a={axis:t};return xs.runKernel(Qr,r,a)}});function sp(e,t=!0,n,r){return xs.makeVariable(e,t,n,r)}function ip(e,t){const n=[];for(let e=0;e<t.length;e++)t[e]&&n.push(e);const r=Ii(e,"int32"),a=Ii([n.length,e.length],"int32");for(let t=0;t<n.length;t++){const s=r.indexToLoc(n[t]),i=t*e.length;a.values.set(s,i)}return a.toTensor()}var op=async function(e){const t=Rs(e,"condition","whereAsync","bool"),n=await t.data(),r=ip(t.shape,n);return e!==t&&t.dispose(),r};var up=async function(e,t,n){const r=Rs(e,"tensor","boolMask"),a=Rs(t,"mask","boolMask","bool"),s=null==n?0:n,i=a.rank,o=r.shape;ue(i>0,()=>"mask cannot be scalar"),le(o.slice(s,s+i),a.shape,"mask's shape must match the first K dimensions of tensor's shape,");let u=1;for(let e=s;e<s+i;e++)u*=o[e];const l=o.slice(0,s).concat([u],o.slice(s+i)),c=qu(r,l),h=qu(a,[-1]),p=await op(h),d=jh(p,[1]),f=Gl(c,d,s);return e!==r&&r.dispose(),t!==a&&a.dispose(),d.dispose(),c.dispose(),h.dispose(),p.dispose(),f};var lp=Ms({norm_:function(e,t="euclidean",n=null,r=!1){const a=function e(t,n,r=null){if(0===t.rank)return xu(t);if(1!==t.rank&&null===r)return e(qu(t,[-1]),n,r);if(1===t.rank||"number"==typeof r||Array.isArray(r)&&1===r.length){if(1===n)return yc(xu(t),r);if(n===1/0)return mc(xu(t),r);if(n===-1/0)return zc(xu(t),r);if("euclidean"===n||2===n)return Wh(yc(ah(xu(t),Ah(2,"int32")),r));throw new Error("Error in norm: invalid ord value: "+n)}if(Array.isArray(r)&&2===r.length){if(1===n)return mc(yc(xu(t),r[0]),r[1]-1);if(n===1/0)return mc(yc(xu(t),r[1]),r[0]);if(n===-1/0)return zc(yc(xu(t),r[1]),r[0]);if("fro"===n||"euclidean"===n)return Wh(yc(jc(t),r));throw new Error("Error in norm: invalid ord value: "+n)}throw new Error("Error in norm: invalid axis: "+r)}(e=Rs(e,"x","norm"),t,n);let s=a.shape;if(r){const t=ke(n,e.shape);s=kc(a.shape,t)}return qu(a,s)}});var cp=Ms({movingAverage_:function(e,t,n,r,a=!0){const s=Rs(e,"v","movingAverage"),i=Rs(t,"x","movingAverage"),o=Rs(n,"decay","movingAverage");ds(s,i),ue(fe(s.shape,i.shape),()=>"Shape mismatch in v and x");const u=Ah(1),l=gc(u,o);let c=vu(gc(i,s),l);if(a){ue(null!=r,()=>"When using zeroDebias: true, step is required.");const e=Rs(r,"step","movingAverage");c=bu(c,gc(u,ah(o,e)))}return gu(s,c)}});var hp=Ms({scatterND_:function(e,t,n){const r=Rs(e,"indices","scatterND","int32"),a=Rs(t,"updates","scatterND");ho(a,r,n);const s={indices:r,updates:a},i={shape:n};return xs.runKernel(br,s,i)}});var pp=Ms({sparseToDense_:function(e,t,n,r=0){const a=Rs(e,"sparseIndices","sparseToDense","int32"),s=Rs(t,"sparseValues","sparseToDense"),i=Rs(r,"defaultValue","sparseToDense",s.dtype);!function(e,t,n,r){if("int32"!==e.dtype)throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);if(e.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);const a=e.rank>0?e.shape[0]:1,s=e.rank>1?e.shape[1]:1;if(n.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${s}.`);const i=t.size;if(0!==t.rank&&(1!==t.rank||i!==a))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${a}]`);if(t.dtype!==r.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}(a,s,n,i);const o={sparseIndices:a,sparseValues:s,defaultValue:i},u={outputShape:n};return xs.runKernel(Lr,o,u)}});var dp=Ms({gatherND_:function(e,t){const n=Rs(t,"indices","gatherND","int32"),r={params:Rs(e,"x","gatherND","string_or_numeric"),indices:n};return xs.runKernel(hn,r)}});var fp=Ms({dropout_:function(e,t,n,r){const a=Rs(e,"x","dropout");if(ue("float32"===a.dtype,()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${a.dtype} tensor instead.`),ue(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),0===t)return e instanceof Ya?a.clone():a;const s=function(e,t){if(null==t)return e.shape.slice();if(fe(e.shape,t))return t;if(e.shape.length===t.length){const n=[];for(let r=0;r<e.shape.length;r++)null==t[r]&&null!=e.shape[r]?n.push(e.shape[r]):n.push(t[r]);return n}return t}(a,n),i=1-t,o=bu(jl(gu(ph(s,0,1,"float32",r),i)),i);return vu(a,o)}});function mp(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function gp(e,t,n){const r=1-e%2,a=new Float32Array(e);for(let s=0;s<e;++s){const i=2*Math.PI*s/(e+r-1);a[s]=t-n*Math.cos(i)}return Yh(a,"float32")}var yp=async function(e,t,n=1){const r=Rs(e,"predictions","inTopK"),a=Rs(t,"targets","inTopK");ue(r.rank>1,()=>"inTopK() expects the predictions to be of rank 2 or higher, but got "+r.rank),ue(r.rank-1===a.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${r.rank} and targets rank ${a.rank}`),le(r.shape.slice(0,r.shape.length-1),a.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");const s=r.shape[r.shape.length-1];ue(n>0&&n<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${n}`);const i=await r.data(),o=await a.data(),[u,l]=[i.length/s,s],c=Ie("bool",u);for(let e=0;e<u;e++){const t=e*l,r=i.subarray(t,t+l),a=[];for(let e=0;e<r.length;e++)a.push({value:r[e],index:e});a.sort((e,t)=>t.value-e.value),c[e]=0;for(let t=0;t<n;t++)if(a[t].index===o[e]){c[e]=1;break}}return e!==r&&r.dispose(),t!==a&&a.dispose(),Ls(c,a.shape,"bool")},bp={};w(bp,{conv2d:()=>Ip,depthwiseConv2d:()=>Np,matMul:()=>Cp});var vp=Ms({conv2DBackpropFilter_:function(e,t,n,r,a,s="NHWC",i){let o=e;3===e.rank&&(o=qu(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let u=t;3===u.rank&&(u=qu(t,[1,t.shape[0],t.shape[1],t.shape[2]])),ue(4===o.rank,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${o.shape}.`),ue(4===u.rank,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${u.shape}.`),ue(4===n.length,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);const l="NHWC"===s?o.shape[3]:o.shape[1],c="NHWC"===s?u.shape[3]:u.shape[1];ue(l===n[2],()=>`Error in conv2dDerFilter: depth of input ${l}) must match input depth in filter (${n[2]}.`),ue(c===n[3],()=>`Error in conv2dDerFilter: depth of dy (${c}) must match output depth for filter (${n[3]}).`),Hu("conv2dDerFilter",a,i);const h={x:o,dy:u},p={strides:r,pad:a,dataFormat:s,dimRoundingMode:i,filterShape:n};return xs.runKernel(Rt,h,p)}});function xp(e,t,n){if(null==n||"linear"===n)return e;if("relu"===n)return vu(e,Hh(t));throw new Error(`Cannot compute gradient for fused activation ${n}.`)}function wp(e,t){let n=t;const r=Qi(e.shape,t.shape);return r.length>0&&(n=yc(n,r)),qu(n,e.shape)}function kp(e,t,n,r){if("linear"===t)return e;if("relu"===t)return gh(e);if("elu"===t)return Ll(e);if("relu6"===t)return yh(e);if("prelu"===t)return sh(e,n);if("leakyrelu"===t)return Zl(e,r);if("sigmoid"===t)return Qu(e);throw new Error(`Unknown fused activation ${t}.`)}var Sp=(e,t)=>!(e>0)||"linear"===t;var Ip=Ms({fusedConv2d_:function({x:e,filter:t,strides:n,pad:r,dataFormat:a="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:u="linear",preluActivationWeights:l,leakyreluAlpha:c}){if(u=u||"linear",!1===Sp(xs.state.gradientDepth,u)){let h=ml(e,t,n,r,a,s,i);return null!=o&&(h=gu(h,o)),kp(h,u,l,c)}const h=Rs(e,"x","conv2d","float32"),p=Rs(t,"filter","conv2d","float32");let d=h,f=!1;3===h.rank&&(f=!0,d=qu(h,[1,h.shape[0],h.shape[1],h.shape[2]])),ue(4===d.rank,()=>`Error in fused conv2d: input must be rank 4, but got rank ${d.rank}.`),ue(4===p.rank,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${p.rank}.`),Hu("fused conv2d",r,i),ue(d.shape[3]===p.shape[2],()=>`Error in conv2d: depth of input (${d.shape[3]}) must match input depth for filter ${p.shape[2]}.`),ue(ju(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),ue("NHWC"===a,()=>`Error in conv2d: got dataFormat of ${a} but only NHWC is currently supported.`);const m=Lu(d.shape,p.shape,n,s,r,i);let g,y;null!=o&&(g=Rs(o,"bias","fused conv2d"),[g]=ps(g,h),Zi(m.outShape,g.shape)),null!=l&&(y=Rs(l,"prelu weights","fused conv2d"));const b=(e,t)=>{const[a,i,o,l]=t,c=xp(e,o,u);ue(Vu(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);const h=[yl(i.shape,c,a,n,r),vp(i,c,a.shape,n,r)];if(null!=l){const e=wp(l,c);h.push(e)}return h},v={x:d,filter:p,bias:g,preluActivationWeights:y},x={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i,activation:u,leakyreluAlpha:c};if(null==o){return cc((e,t,n)=>{let r=xs.runKernel(aa,v,x);return n([t,e,r]),f&&(r=qu(r,[r.shape[1],r.shape[2],r.shape[3]])),{value:r,gradFunc:b}})(d,p)}return cc((e,t,n,r)=>{let a=xs.runKernel(aa,v,x);return r([t,e,a,n]),f&&(a=qu(a,[a.shape[1],a.shape[2],a.shape[3]])),{value:a,gradFunc:b}})(d,p,g)}});var Ap=Ms({depthwiseConv2dNativeBackpropFilter_:function(e,t,n,r,a,s=[1,1],i){let o=e;3===e.rank&&(o=qu(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let u=t;3===u.rank&&(u=qu(t,[1,t.shape[0],t.shape[1],t.shape[2]]));const l={x:o,dy:u},c={strides:r,pad:a,dimRoundingMode:i,dilations:s,filterShape:n};return xs.runKernel(Wt,l,c)}});var Ep=Ms({depthwiseConv2dNativeBackpropInput_:function(e,t,n,r,a,s=[1,1],i){let o=t,u=!1;3===t.rank&&(u=!0,o=qu(t,[1,t.shape[0],t.shape[1],t.shape[2]]));const l={dy:o,filter:n},c={strides:r,pad:a,dimRoundingMode:i,dilations:s,inputShape:e},h=xs.runKernel(Vt,l,c);return u?qu(h,[h.shape[1],h.shape[2],h.shape[3]]):h}});var Np=Ms({fusedDepthwiseConv2d_:function({x:e,filter:t,strides:n,pad:r,dataFormat:a="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:u="linear",preluActivationWeights:l,leakyreluAlpha:c}){if(!1===Sp(xs.state.gradientDepth,u)){let h=Nl(e,t,n,r,a,s,i);return null!=o&&(h=gu(h,o)),kp(h,u,l,c)}const h=Rs(e,"x","depthwiseConv2d","float32"),p=Rs(t,"filter","depthwiseConv2d","float32");let d=h,f=!1;3===h.rank&&(f=!0,d=qu(h,[1,h.shape[0],h.shape[1],h.shape[2]])),ue(4===d.rank,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${d.rank}.`),ue(4===p.rank,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${p.rank}.`),ue(d.shape[3]===p.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${d.shape[3]}) must match the inChannels dimension in filter ${p.shape[2]}.`),null==s&&(s=[1,1]),ue(ju(n,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),Hu("fused depthwiseConv2d",r,i);const m=Lu(d.shape,p.shape,n,s,r,i,!0);let g,y;null!=o&&(g=Rs(o,"bias","fused conv2d"),[g]=ps(g,h),Zi(m.outShape,g.shape)),null!=l&&(y=Rs(l,"prelu weights","fused depthwiseConv2d"));const b=(e,t)=>{ue(Vu(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);const[a,o,l,c]=t,h=xp(e,l,u),p=Ep(o.shape,h,a,n,r,s,i),d=Ap(o,h,a.shape,n,r,s,i);if(null!=c){return[p,d,wp(g,h)]}return[p,d]},v={x:d,filter:p,bias:g,preluActivationWeights:y},x={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i,activation:u,leakyreluAlpha:c};if(null==o){return cc((e,t,n)=>{let r=xs.runKernel(sa,v,x);return n([t,e,r]),f&&(r=qu(r,[r.shape[1],r.shape[2],r.shape[3]])),{value:r,gradFunc:b}})(d,p)}return cc((e,t,n,r)=>{let a=xs.runKernel(sa,v,x);return r([t,e,a,n]),f&&(a=qu(a,[a.shape[1],a.shape[2],a.shape[3]])),{value:a,gradFunc:b}})(d,p,g)}});var Cp=Ms({fusedMatMul_:function({a:e,b:t,transposeA:n=!1,transposeB:r=!1,bias:a,activation:s="linear",preluActivationWeights:i,leakyreluAlpha:o}){if(!1===Sp(xs.state.gradientDepth,s)){let u=Gi(e,t,n,r);return null!=a&&(u=gu(u,a)),kp(u,s,i,o)}let u=Rs(e,"a","fused matMul"),l=Rs(t,"b","fused matMul");[u,l]=ps(u,l);const c=n?u.shape[u.rank-2]:u.shape[u.rank-1],h=r?l.shape[l.rank-1]:l.shape[l.rank-2],p=n?u.shape[u.rank-1]:u.shape[u.rank-2],d=r?l.shape[l.rank-2]:l.shape[l.rank-1],f=u.shape.slice(0,-2),m=l.shape.slice(0,-2),g=pe(f),y=pe(m);ue(c===h,()=>`Error in fused matMul: inner shapes (${c}) and (${h}) of Tensors with shapes ${u.shape} and ${l.shape} and transposeA=${n} and transposeB=${r} must match.`);const b=Zi(u.shape.slice(0,-2),l.shape.slice(0,-2)).concat([p,d]),v=qu(u,n?[g,c,p]:[g,p,c]),x=qu(l,r?[y,d,h]:[y,h,d]);let w,k;null!=a&&(w=Rs(a,"bias","fused matMul"),[w]=ps(w,u),Zi(b,w.shape)),null!=i&&(k=Rs(i,"prelu weights","fused matMul"));const S=(e,t)=>{const[i,o,u,l]=t,c=xp(qu(e,u.shape),u,s);let h,p;if(n||r?!n&&r?(h=Gi(c,o,!1,!1),p=Gi(c,i,!0,!1)):n&&!r?(h=Gi(o,c,!1,!0),p=Gi(i,c,!1,!1)):(h=Gi(o,c,!0,!0),p=Gi(c,i,!0,!0)):(h=Gi(c,o,!1,!0),p=Gi(i,c,!0,!1)),null!=a){return[h,p,wp(l,c)]}return[h,p]},I={a:v,b:x,bias:w,preluActivationWeights:k},A={transposeA:n,transposeB:r,activation:s,leakyreluAlpha:o};if(null==a){return cc((e,t,n)=>{const r=xs.runKernel(ra,I,A);return n([e,t,r]),{value:qu(r,b),gradFunc:S}})(v,x)}return cc((e,t,n,r)=>{const a=xs.runKernel(ra,I,A);return r([e,t,a,n]),{value:qu(a,b),gradFunc:S}})(v,x,w)}});var Tp=Ms({hammingWindow_:function(e){return gp(e,.54,.46)}});var Rp=Ms({hannWindow_:function(e){return gp(e,.5,.5)}});var _p=Ms({frame_:function(e,t,n,r=!1,a=0){let s=0;const i=[];for(;s+t<=e.size;)i.push(Zu(e,s,t)),s+=n;if(r)for(;s<e.size;){const r=s+t-e.size,o=Xu([Zu(e,s,t-r),Vl([r],a)]);i.push(o),s+=n}return 0===i.length?Xh([],[0,t]):qu(Xu(i),[i.length,t])}});var Fp=Ms({stft_:function(e,t,n,r,a=Rp){null==r&&(r=mp(t));const s=_p(e,t,n),i=vu(s,a(t));return Uh(i,r)}});var Mp=Ms({cropAndResize_:function(e,t,n,r,a="bilinear",s=0){const i=Rs(e,"image","cropAndResize"),o=Rs(t,"boxes","cropAndResize","float32"),u=Rs(n,"boxInd","cropAndResize","int32"),l=o.shape[0];ue(4===i.rank,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),ue(2===o.rank&&4===o.shape[1],()=>`Error in cropAndResize: boxes must be have size [${l},4] but had shape ${o.shape}.`),ue(1===u.rank&&u.shape[0]===l,()=>`Error in cropAndResize: boxInd must be have size [${l}] but had shape ${o.shape}.`),ue(2===r.length,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${r.length}.`),ue(r[0]>=1&&r[1]>=1,()=>"cropSize must be atleast [1,1], but was "+r),ue("bilinear"===a||"nearest"===a,()=>"method must be bilinear or nearest, but was "+a);const c={image:i,boxes:o,boxInd:u},h={method:a,extrapolationValue:s,cropSize:r};return xs.runKernel($t,c,h)}});var Dp=Ms({flipLeftRight_:function(e){const t=Rs(e,"image","flipLeftRight","float32");ue(4===t.rank,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);const n={image:t};return xs.runKernel(sn,n,{})}});var Op=Ms({grayscaleToRGB_:function(e){const t=Rs(e,"image","grayscaleToRGB"),n=t.rank-1,r=t.shape[n];ue(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),ue(1===r,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${r}.`);const a=new Array(t.rank);return a.fill(1,0,n),a[n]=3,Ul(t,a)}});var Lp=Ms({rotateWithOffset_:function(e,t,n=0,r=.5){const a=Rs(e,"image","rotateWithOffset","float32");ue(4===a.rank,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${a.rank}.`);const s={image:a},i={radians:t,fillValue:n,center:r};return xs.runKernel(na,s,i)}});function Pp(e,t,n,r,a,s){null==r&&(r=.5),null==a&&(a=Number.NEGATIVE_INFINITY),null==s&&(s=0);const i=e.shape[0];return n=Math.min(n,i),ue(0<=r&&r<=1,()=>`iouThreshold must be in [0, 1], but was '${r}'`),ue(2===e.rank,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),ue(4===e.shape[1],()=>"boxes must have 4 columns, but 2nd dimension was "+e.shape[1]),ue(1===t.rank,()=>"scores must be a 1D tensor"),ue(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),ue(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:n,iouThreshold:r,scoreThreshold:a,softNmsSigma:s}}var $p=Ms({nonMaxSuppression_:function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY){const s=Rs(e,"boxes","nonMaxSuppression","float32"),i=Rs(t,"scores","nonMaxSuppression","float32"),o=Pp(s,i,n,r,a),u={maxOutputSize:n=o.maxOutputSize,iouThreshold:r=o.iouThreshold,scoreThreshold:a=o.scoreThreshold};return xs.runKernel(Kn,{boxes:s,scores:i},u)}});function Bp(e,t,n){const r=function(e,t,n){return function(e,t,n){let r=0,a=e.length,s=0,i=!1;for(;r<a;){s=r+(a-r>>>1);const o=n(t,e[s]);o>0?r=s+1:(a=s,i=!o)}return i?r:-r-1}(e,t,n||zp)}(e,t,n),a=r<0?-(r+1):r;e.splice(a,0,t)}function zp(e,t){return e>t?1:e<t?-1:0}function Up(e,t,n,r,a){return jp(e,t,n,r,a,0)}function Wp(e,t,n,r,a,s){return jp(e,t,n,r,a,0,!1,s,!0)}function Vp(e,t,n,r,a,s){return jp(e,t,n,r,a,s,!0)}function jp(e,t,n,r,a,s,i=!1,o=!1,u=!1){const l=[];for(let e=0;e<t.length;e++)t[e]>a&&l.push({score:t[e],boxIndex:e,suppressBeginIndex:0});l.sort(qp);const c=s>0?-.5/s:0,h=[],p=[];for(;h.length<n&&l.length>0;){const t=l.pop(),{score:n,boxIndex:s,suppressBeginIndex:i}=t;if(n<a)break;let o=!1;for(let n=h.length-1;n>=i;--n){const i=Gp(e,s,h[n]);if(i>=r){o=!0;break}if(t.score=t.score*Hp(r,c,i),t.score<=a)break}t.suppressBeginIndex=h.length,o||(t.score===n?(h.push(s),p.push(t.score)):t.score>a&&Bp(l,t,qp))}const d=h.length,f=n-d;o&&f>0&&(h.push(...new Array(f).fill(0)),p.push(...new Array(f).fill(0)));const m={selectedIndices:h};return i&&(m.selectedScores=p),u&&(m.validOutputs=d),m}function Gp(e,t,n){const r=e.subarray(4*t,4*t+4),a=e.subarray(4*n,4*n+4),s=Math.min(r[0],r[2]),i=Math.min(r[1],r[3]),o=Math.max(r[0],r[2]),u=Math.max(r[1],r[3]),l=Math.min(a[0],a[2]),c=Math.min(a[1],a[3]),h=Math.max(a[0],a[2]),p=Math.max(a[1],a[3]),d=(o-s)*(u-i),f=(h-l)*(p-c);if(d<=0||f<=0)return 0;const m=Math.max(s,l),g=Math.max(i,c),y=Math.min(o,h),b=Math.min(u,p),v=Math.max(y-m,0)*Math.max(b-g,0);return v/(d+f-v)}function Hp(e,t,n){const r=Math.exp(t*n*n);return n<=e?r:0}function qp(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}var Kp=async function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY){const s=Rs(e,"boxes","nonMaxSuppressionAsync"),i=Rs(t,"scores","nonMaxSuppressionAsync"),o=Pp(s,i,n,r,a);n=o.maxOutputSize,r=o.iouThreshold,a=o.scoreThreshold;const u=await Promise.all([s.data(),i.data()]),l=u[0],c=u[1],{selectedIndices:h}=Up(l,c,n,r,a);return s!==e&&s.dispose(),i!==t&&i.dispose(),Yh(h,"int32")};var Yp=Ms({nonMaxSuppressionWithScore_:function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=0){const i=Rs(e,"boxes","nonMaxSuppression"),o=Rs(t,"scores","nonMaxSuppression"),u=Pp(i,o,n,r,a,s),l={boxes:i,scores:o},c={maxOutputSize:n=u.maxOutputSize,iouThreshold:r=u.iouThreshold,scoreThreshold:a=u.scoreThreshold,softNmsSigma:s=u.softNmsSigma},h=xs.runKernel(Xn,l,c);return{selectedIndices:h[0],selectedScores:h[1]}}});var Xp=async function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=0){const i=Rs(e,"boxes","nonMaxSuppressionAsync"),o=Rs(t,"scores","nonMaxSuppressionAsync"),u=Pp(i,o,n,r,a,s);n=u.maxOutputSize,r=u.iouThreshold,a=u.scoreThreshold,s=u.softNmsSigma;const l=await Promise.all([i.data(),o.data()]),c=l[0],h=l[1],{selectedIndices:p,selectedScores:d}=Vp(c,h,n,r,a,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Yh(p,"int32"),selectedScores:Yh(d)}};var Qp=Ms({nonMaxSuppressionPadded_:function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=!1){const i=Rs(e,"boxes","nonMaxSuppression"),o=Rs(t,"scores","nonMaxSuppression"),u=Pp(i,o,n,r,a,null),l={boxes:i,scores:o},c={maxOutputSize:u.maxOutputSize,iouThreshold:u.iouThreshold,scoreThreshold:u.scoreThreshold,padToMaxOutputSize:s},h=xs.runKernel(Yn,l,c);return{selectedIndices:h[0],validOutputs:h[1]}}});var Zp=async function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=!1){const i=Rs(e,"boxes","nonMaxSuppressionAsync"),o=Rs(t,"scores","nonMaxSuppressionAsync"),u=Pp(i,o,n,r,a,null),l=u.maxOutputSize,c=u.iouThreshold,h=u.scoreThreshold,[p,d]=await Promise.all([i.data(),o.data()]),{selectedIndices:f,validOutputs:m}=Wp(p,d,l,c,h,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Yh(f,"int32"),validOutputs:Ah(m,"int32")}};var Jp=Ms({resizeBilinear_:function(e,t,n=!1,r=!1){const a=Rs(e,"images","resizeBilinear");ue(3===a.rank||4===a.rank,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${a.rank}.`),ue(2===t.length,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),ue(!1===r||!1===n,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let s=a,i=!1;3===a.rank&&(i=!0,s=qu(a,[1,a.shape[0],a.shape[1],a.shape[2]]));const o={images:s},u={alignCorners:n,halfPixelCenters:r,size:t},l=xs.runKernel(pr,o,u);return i?qu(l,[l.shape[1],l.shape[2],l.shape[3]]):l}});var ed=Ms({resizeNearestNeighbor_:function(e,t,n=!1,r=!1){const a=Rs(e,"images","resizeNearestNeighbor");ue(3===a.rank||4===a.rank,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${a.rank}.`),ue(2===t.length,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),ue("float32"===a.dtype||"int32"===a.dtype,()=>"`images` must have `int32` or `float32` as dtype"),ue(!1===r||!1===n,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let s=a,i=!1;3===a.rank&&(i=!0,s=qu(a,[1,a.shape[0],a.shape[1],a.shape[2]]));const o={images:s},u={alignCorners:n,halfPixelCenters:r,size:t},l=xs.runKernel(cr,o,u);return i?qu(l,[l.shape[1],l.shape[2],l.shape[3]]):l}});var td=Ms({threshold_:function(e,t="binary",n=!1,r=.5){const a=Rs(e,"image","threshold"),s=a.shape[0]*a.shape[1];let i,o,u,l,c=vu(Yh([r]),255);if(ue(3===a.rank,()=>`Error in threshold: image must be rank 3,but got rank ${a.rank}.`),ue(3===a.shape[2]||1===a.shape[2],()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${a.shape[2]}.`),ue("int32"===a.dtype||"float32"===a.dtype,()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${a.dtype}.`),ue("otsu"===t||"binary"===t,()=>"Method must be binary or otsu, but was "+t),3===a.shape[2]){[i,o,u]=zh(a,[1,1,1],-1);const e=vu(i,.2989),t=vu(o,.587),n=vu(u,.114);l=gu(gu(e,t),n)}else l=e;if("otsu"===t){c=function(e,t){let n,r,a,s,i,o,u=Yh([-1]),l=Yh([0]),c=Yh([0]);for(let h=0;h<e.size-1;h++){n=Zu(e,0,h+1),r=Zu(e,h+1),i=bu(yc(n),t),o=bu(yc(r),t);const p=yc(vu(n,dh(0,n.size)));a=bu(p,yc(n));const d=Vl(r.shape,n.size),f=gu(dh(0,r.size),d),m=vu(r,f);s=bu(yc(m),yc(r));const g=gc(a,s),y=gc(a,s),b=vu(i,o);c=vu(vu(b,g),y);const v=Hl(c,l);l=_l(v,c,l),u=_l(v,Yh([h]),u)}return u}(il(Ai(Sh(l),"int32"),Ls([]),256),s)}const h=n?ec(l,c):Hl(l,c);return Ai(vu(h,255),"int32")}});var nd=Ms({transform_:function(e,t,n="nearest",r="constant",a=0,s){const i=Rs(e,"image","transform","float32"),o=Rs(t,"transforms","transform","float32");ue(4===i.rank,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),ue(2===o.rank&&(o.shape[0]===i.shape[0]||1===o.shape[0])&&8===o.shape[1],()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),ue(null==s||2===s.length,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);const u={image:i,transforms:o},l={interpolation:n,fillMode:r,fillValue:a,outputShape:s};return xs.runKernel(Kr,u,l)}});var rd=Ms({bandPart_:function(e,t,n){ue(t%1==0,()=>`bandPart(): numLower must be an integer, got ${t}.`),ue(n%1==0,()=>`bandPart(): numUpper must be an integer, got ${n}.`);const r=Rs(e,"a","bandPart");ue(r.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${r.rank}.`);const a=r.shape,[s,i]=r.shape.slice(-2);if(!(t<=s))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);if(!(n<=i))throw new Error(`bandPart(): numUpper (${n}) must not be greater than the number of columns (${i}).`);t<0&&(t=s),n<0&&(n=i);const o=qu(dh(0,s,1,"int32"),[-1,1]),u=dh(0,i,1,"int32"),l=gc(o,u),c=Cc(ec(l,Ah(+t,"int32")),ql(l,Ah(-n,"int32"))),h=Pc([s,i],r.dtype);return qu(Gh(ap(qu(r,[-1,s,i])).map(e=>_l(c,e,h))),a)}});var ad=Ms({gramSchmidt_:function(e){let t;if(Array.isArray(e)){t=!1,ue(null!=e&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");const n=e[0].shape[0];for(let t=1;t<e.length;++t)ue(e[t].shape[0]===n,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[t].shape[0]} vs. ${n})`)}else t=!0,e=zh(e,e.shape[0],0).map(e=>jh(e,[0]));ue(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);const n=[],r=e;for(let t=0;t<e.length;++t)n.push(xs.tidy(()=>{let e=r[t];if(t>0)for(let r=0;r<t;++r){const t=vu(yc(vu(n[r],e)),n[r]);e=gc(e,t)}return bu(e,lp(e,"euclidean"))}));return t?Gh(n,0):n}});function sd(e,t=!1){return xs.tidy(()=>{ue(2===e.shape.length,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);const n=e.shape[0],r=e.shape[1];let a=Wl(n),s=Ei(e);const i=Xh([[1]],[1,1]);let o=Ei(i);const u=n>=r?r:n;for(let e=0;e<u;++e){const t=s,u=o,l=a;[o,s,a]=xs.tidy(()=>{const t=Zu(s,[e,e],[n-e,1]),u=lp(t),l=Zu(s,[e,e],[1,1]),c=_l(Hl(l,0),Xh([[-1]]),Xh([[1]])),h=gc(l,vu(c,u)),p=bu(t,h);o=1===p.shape[0]?Ei(i):Xu([i,Zu(p,[1,0],[p.shape[0]-1,p.shape[1]])],0);const d=pc(bu(Gi(c,h),u)),f=Zu(s,[e,0],[n-e,r]),m=vu(d,o),g=qi(o);if(0===e)s=gc(f,Gi(m,Gi(g,f)));else{const t=gc(f,Gi(m,Gi(g,f)));s=Xu([Zu(s,[0,0],[e,r]),t],0)}const y=qi(m),b=Zu(a,[0,e],[n,a.shape[1]-e]);if(0===e)a=gc(b,Gi(Gi(b,o),y));else{const t=gc(b,Gi(Gi(b,o),y));a=Xu([Zu(a,[0,0],[n,e]),t],1)}return[o,s,a]}),au([t,u,l])}return!t&&n>r&&(a=Zu(a,[0,0],[n,r]),s=Zu(s,[0,0],[r,r])),[a,s]})}var id,od,ud=Ms({qr_:function(e,t=!1){if(ue(e.rank>=2,()=>"qr() requires input tensor to have a rank >= 2, but got rank "+e.rank),2===e.rank)return sd(e,t);{const n=e.shape.slice(0,e.shape.length-2).reduce((e,t)=>e*t),r=ap(qu(e,[n,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),a=[],s=[];r.forEach(e=>{const[n,r]=sd(e,t);a.push(n),s.push(r)});return[qu(Gh(a,0),e.shape),qu(Gh(s,0),e.shape)]}}});(od=id||(id={}))[od.NONE=0]="NONE",od[od.MEAN=1]="MEAN",od[od.SUM=2]="SUM",od[od.SUM_BY_NONZERO_WEIGHTS=3]="SUM_BY_NONZERO_WEIGHTS";var ld=Ms({computeWeightedLoss_:function(e,t,n=id.SUM_BY_NONZERO_WEIGHTS){const r=Rs(e,"losses","computeWeightedLoss");let a=null;null!=t&&(a=Rs(t,"weights","computeWeightedLoss"));const s=null==a?r:vu(r,a);if(n===id.NONE)return s;if(n===id.SUM)return yc(s);if(n===id.MEAN){if(null==a)return Lc(s);{const e=r.size/a.size,t=bu(yc(s),yc(a));return e>1?bu(t,Ah(e)):t}}if(n===id.SUM_BY_NONZERO_WEIGHTS){if(null==a)return bu(yc(s),Ah(r.size));{const e=vu(a,$c(r.shape)),t=Ai(yc(Kc(e,Ah(0))),"float32");return bu(yc(s),t)}}throw Error("Unknown reduction: "+n)}});var cd=Ms({absoluteDifference_:function(e,t,n,r=id.SUM_BY_NONZERO_WEIGHTS){const a=Rs(e,"labels","absoluteDifference"),s=Rs(t,"predictions","absoluteDifference");let i=null;null!=n&&(i=Rs(n,"weights","absoluteDifference")),le(a.shape,s.shape,"Error in absoluteDifference: ");const o=xu(gc(a,s));return ld(o,i,r)}});var hd=Ms({cosineDistance_:function(e,t,n,r,a=id.SUM_BY_NONZERO_WEIGHTS){const s=Rs(e,"labels","cosineDistance"),i=Rs(t,"predictions","cosineDistance");let o=null;null!=r&&(o=Rs(r,"weights","cosineDistance")),le(s.shape,i.shape,"Error in cosineDistance: ");const u=Ah(1),l=gc(u,yc(vu(s,i),n,!0));return ld(l,o,a)}});var pd=Ms({hingeLoss_:function(e,t,n,r=id.SUM_BY_NONZERO_WEIGHTS){let a=Rs(e,"labels","hingeLoss");const s=Rs(t,"predictions","hingeLoss");let i=null;null!=n&&(i=Rs(n,"weights","hingeLoss")),le(a.shape,s.shape,"Error in hingeLoss: ");const o=Ah(1);a=gc(vu(Ah(2),a),o);const u=gh(gc(o,vu(a,s)));return ld(u,i,r)}});var dd=Ms({huberLoss_:function(e,t,n,r=1,a=id.SUM_BY_NONZERO_WEIGHTS){const s=Rs(e,"labels","huberLoss"),i=Rs(t,"predictions","huberLoss");let o=null;null!=n&&(o=Rs(n,"weights","huberLoss")),le(s.shape,i.shape,"Error in huberLoss: ");const u=Ah(r),l=xu(gc(i,s)),c=Uc(l,u),h=gc(l,c),p=gu(vu(Ah(.5),jc(c)),vu(u,h));return ld(p,o,a)}});var fd=Ms({logLoss_:function(e,t,n,r=1e-7,a=id.SUM_BY_NONZERO_WEIGHTS){const s=Rs(e,"labels","logLoss"),i=Rs(t,"predictions","logLoss");let o=null;null!=n&&(o=Rs(n,"weights","logLoss")),le(s.shape,i.shape,"Error in logLoss: ");const u=Ah(1),l=Ah(r),c=pc(vu(s,rc(gu(i,l)))),h=vu(gc(u,s),rc(gu(gc(u,i),l))),p=gc(c,h);return ld(p,o,a)}});var md=Ms({meanSquaredError_:function(e,t,n,r=id.SUM_BY_NONZERO_WEIGHTS){const a=Rs(e,"labels","meanSquaredError"),s=Rs(t,"predictions","meanSquaredError");let i=null;null!=n&&(i=Rs(n,"weights","meanSquaredError")),le(a.shape,s.shape,"Error in meanSquaredError: ");const o=Vh(a,s);return ld(o,i,r)}});var gd=Ms({sigmoidCrossEntropy_:function(e,t,n,r=0,a=id.SUM_BY_NONZERO_WEIGHTS){let s=Rs(e,"multiClassLabels","sigmoidCrossEntropy");const i=Rs(t,"logits","sigmoidCrossEntropy");let o=null;if(null!=n&&(o=Rs(n,"weights","sigmoidCrossEntropy")),le(s.shape,i.shape,"Error in sigmoidCrossEntropy: "),r>0){const e=Ah(r),t=Ah(1),n=Ah(.5);s=gu(vu(s,gc(t,e)),vu(n,e))}const u=function(e,t){const n=Rs(e,"labels","sigmoidCrossEntropyWithLogits"),r=Rs(t,"logits","sigmoidCrossEntropyWithLogits");le(n.shape,r.shape,"Error in sigmoidCrossEntropyWithLogits: ");const a=gh(r),s=vu(r,n),i=ac($l(pc(xu(r))));return gu(gc(a,s),i)}(s,i);return ld(u,o,a)}});var yd=Ms({softmaxCrossEntropy_:function(e,t,n,r=0,a=id.SUM_BY_NONZERO_WEIGHTS){let s=Rs(e,"onehotLabels","softmaxCrossEntropy");const i=Rs(t,"logits","softmaxCrossEntropy");let o=null;if(null!=n&&(o=Rs(n,"weights","softmaxCrossEntropy")),le(s.shape,i.shape,"Error in softmaxCrossEntropy: "),r>0){const e=Ah(r),t=Ah(1),n=Ah(s.shape[1]);s=gu(vu(s,gc(t,e)),bu(e,n))}const u=function(e,t,n=-1){if(-1===n&&(n=t.rank-1),n!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${n}`);return cc((e,t,r)=>{const a=Nc(t,[n],!0),s=gc(Ai(t,"float32"),a);r([e,s]);const i=pc(vu(s,e));return{value:yc(i,[n]),gradFunc:(e,t)=>{const[r,a]=t,s=kc(e.shape,[n]);return[vu(qu(e,s),gc(Ai(r,"float32"),$l(a))),vu(qu(e,s),gc($l(a),Ai(r,"float32")))]}}})(e,t)}(s,i);return ld(u,o,a)}});var bd=Ms({sparseFillEmptyRows_:function(e,t,n,r){const a=Rs(e,"indices","sparseFillEmptyRows","int32"),s=Rs(t,"values","sparseFillEmptyRows"),i=Rs(n,"denseShape","sparseFillEmptyRows","int32"),o=Rs(r,"defaultValue","sparseFillEmptyRows",s.dtype);if(2!==a.rank)throw new Error("Indices should be Tensor2D but received shape\n        "+a.shape);if(1!==s.rank)throw new Error("Values should be Tensor1D but received shape "+s.shape);if(1!==i.rank)throw new Error("Dense shape should be Tensor1D but received shape "+i.shape);if(0!==o.rank)throw new Error("Default value should be a scalar but received shape "+o.shape);const u={indices:a,values:s,denseShape:i,defaultValue:o},l=xs.runKernel(Fr,u);return{outputIndices:l[0],outputValues:l[1],emptyRowIndicator:l[2],reverseIndexMap:l[3]}}});var vd=Ms({sparseReshape_:function(e,t,n){const r=Rs(e,"inputIndices","sparseReshape","int32"),a=Rs(t,"inputShape","sparseReshape","int32"),s=Rs(n,"newShape","sparseReshape","int32");if(2!==r.rank)throw new Error("Input indices should be Tensor2D but received shape\n        "+r.shape);if(1!==a.rank)throw new Error("Input shape should be Tensor1D but received shape "+a.shape);if(1!==s.rank)throw new Error("New shape should be Tensor1D but received shape "+s.shape);const i={inputIndices:r,inputShape:a,newShape:s},o=xs.runKernel(Mr,i);return{outputIndices:o[0],outputShape:o[1]}}});var xd=Ms({sparseSegmentMean_:function(e,t,n){const r=Rs(e,"data","sparseSegmentMean"),a=Rs(t,"indices","sparseSegmentMean","int32"),s=Rs(n,"segmentIds","sparseSegmentMean","int32");if(r.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.rank)throw new Error("Indices should be Tensor1D but received shape\n          "+a.shape);if(1!==s.rank)throw new Error("Segment ids should be Tensor1D but received shape\n          "+s.shape);const i={data:r,indices:a,segmentIds:s};return xs.runKernel(Dr,i)}});var wd=Ms({sparseSegmentSum_:function(e,t,n){const r=Rs(e,"data","sparseSegmentSum"),a=Rs(t,"indices","sparseSegmentSum","int32"),s=Rs(n,"segmentIds","sparseSegmentSum","int32");if(r.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.rank)throw new Error("Indices should be Tensor1D but received shape\n         "+a.shape);if(1!==s.rank)throw new Error("Segment ids should be Tensor1D but received shape\n         "+s.shape);const i={data:r,indices:a,segmentIds:s};return xs.runKernel(Or,i)}});var kd=Ms({stringNGrams_:function(e,t,n,r,a,s,i,o){const u=Rs(e,"data","stringNGrams","string");if("string"!==u.dtype)throw new Error("Data must be of datatype string");if(1!==u.shape.length)throw new Error("Data must be a vector, saw: "+u.shape);const l=Rs(t,"dataSplits","stringNGrams");if("int32"!==l.dtype)throw new Error("Data splits must be of datatype int32");const c={separator:n,nGramWidths:r,leftPad:a,rightPad:s,padWidth:i,preserveShortSequences:o},h={data:u,dataSplits:l},p=xs.runKernel(zr,h,c);return{nGrams:p[0],nGramsSplits:p[1]}}});var Sd=Ms({stringSplit_:function(e,t,n=!0){const r=Rs(e,"input","stringSplit","string"),a=Rs(t,"delimiter","stringSplit","string");if(1!==r.rank)throw new Error("Input should be Tensor1D but received shape "+r.shape);if(0!==a.rank)throw new Error("Delimiter should be a scalar but received shape "+a.shape);const s={skipEmpty:n},i={input:r,delimiter:a},o=xs.runKernel(Ur,i,s);return{indices:o[0],values:o[1],shape:o[2]}}});var Id=Ms({stringToHashBucketFast_:function(e,t){const n=Rs(e,"input","stringToHashBucketFast","string"),r={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");const a={input:n};return xs.runKernel(Wr,a,r)}}),Ad={fft:Ph,ifft:$h,rfft:Uh,irfft:Bh},Ed={hammingWindow:Tp,hannWindow:Rp,frame:_p,stft:Fp},Nd={flipLeftRight:Dp,grayscaleToRGB:Op,resizeNearestNeighbor:ed,resizeBilinear:Jp,rotateWithOffset:Lp,cropAndResize:Mp,nonMaxSuppression:$p,nonMaxSuppressionAsync:Kp,nonMaxSuppressionWithScore:Yp,nonMaxSuppressionWithScoreAsync:Xp,nonMaxSuppressionPadded:Qp,nonMaxSuppressionPaddedAsync:Zp,threshold:td,transform:nd},Cd={bandPart:rd,gramSchmidt:ad,qr:ud},Td={absoluteDifference:cd,computeWeightedLoss:ld,cosineDistance:hd,hingeLoss:pd,huberLoss:dd,logLoss:fd,meanSquaredError:md,sigmoidCrossEntropy:gd,softmaxCrossEntropy:yd},Rd={sparseFillEmptyRows:bd,sparseReshape:vd,sparseSegmentMean:xd,sparseSegmentSum:wd},_d={stringNGrams:kd,stringSplit:Sd,stringToHashBucketFast:Id},Fd=class extends Mo{minimize(e,t=!1,n){const{value:r,grads:a}=this.computeGradients(e,n);if(null!=n){const e=n.map(e=>({name:e.name,tensor:a[e.name]}));this.applyGradients(e)}else this.applyGradients(a);return au(a),t?r:(r.dispose(),null)}get iterations(){return null==this.iterations_&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return lc(e,t)}dispose(){null!=this.iterations_&&au(this.iterations_)}async saveIterations(){return null==this.iterations_&&(this.iterations_=0),{name:"iter",tensor:Ah(this.iterations_,"int32")}}async getWeights(){throw new Error("getWeights() is not implemented for this optimizer yet.")}async setWeights(e){throw new Error("setWeights() is not implemented for this optimizer class "+this.getClassName())}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}};Object.defineProperty(Fd,Symbol.hasInstance,{value:e=>null!=e.minimize&&null!=e.computeGradients&&null!=e.applyGradients});var Md=class extends Fd{constructor(e,t,n=null){super(),this.learningRate=e,this.rho=t,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],null==n&&(this.epsilon=xs.backend.epsilon())}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{const r=xs.registeredVariables[t];null==this.accumulatedGrads[n]&&(this.accumulatedGrads[n]={originalName:t+"/accum_grad",variable:ru(()=>Fl(r).variable(!1))}),null==this.accumulatedUpdates[n]&&(this.accumulatedUpdates[n]={originalName:t+"/accum_var",variable:ru(()=>Fl(r).variable(!1))});const a=Array.isArray(e)?e[n].tensor:e[t];if(null==a)return;const s=this.accumulatedGrads[n].variable,i=this.accumulatedUpdates[n].variable;ru(()=>{const e=gu(vu(s,this.rho),vu(jc(a),1-this.rho)),t=vu(bu(Wh(gu(i,this.epsilon)),Wh(gu(s,this.epsilon))),a),n=gu(vu(i,this.rho),vu(jc(t),1-this.rho));s.assign(e),i.assign(n);const o=gu(vu(t,-this.learningRate),r);r.assign(o)})}),this.incrementIterations()}dispose(){null!=this.accumulatedUpdates&&(au(this.accumulatedGrads.map(e=>e.variable)),au(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){const e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){const t=(e=await this.extractIterations(e)).length/2;this.accumulatedGrads=e.slice(0,t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)})),this.accumulatedUpdates=e.slice(t,2*t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}};Md.className="Adadelta",Oo(Md);var Dd=class extends Fd{constructor(e,t=.1){super(),this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{const r=xs.registeredVariables[t];if(null==this.accumulatedGrads[n]){const e=!1;this.accumulatedGrads[n]={originalName:t+"/accumulator",variable:ru(()=>Vl(r.shape,this.initialAccumulatorValue).variable(e))}}const a=Array.isArray(e)?e[n].tensor:e[t];if(null==a)return;const s=this.accumulatedGrads[n].variable;ru(()=>{const e=gu(s,jc(a));s.assign(e);const t=gu(vu(bu(a,Wh(gu(e,xs.backend.epsilon()))),-this.learningRate),r);r.assign(t)})}),this.incrementIterations()}dispose(){null!=this.accumulatedGrads&&au(this.accumulatedGrads.map(e=>e.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);this.accumulatedGrads=e.map(e=>({originalName:e.name,variable:e.tensor.variable(!1)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}};Dd.className="Adagrad",Oo(Dd);var Od=class extends Fd{constructor(e,t,n,r=null){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],ru(()=>{this.accBeta1=Ah(t).variable(),this.accBeta2=Ah(n).variable()}),null==r&&(this.epsilon=xs.backend.epsilon())}applyGradients(e){const t=Array.isArray(e)?e.map(e=>e.name):Object.keys(e);ru(()=>{const n=gc(1,this.accBeta1),r=gc(1,this.accBeta2);t.forEach((t,a)=>{const s=xs.registeredVariables[t];null==this.accumulatedFirstMoment[a]&&(this.accumulatedFirstMoment[a]={originalName:t+"/m",variable:ru(()=>Fl(s).variable(!1))}),null==this.accumulatedSecondMoment[a]&&(this.accumulatedSecondMoment[a]={originalName:t+"/v",variable:ru(()=>Fl(s).variable(!1))});const i=Array.isArray(e)?e[a].tensor:e[t];if(null==i)return;const o=this.accumulatedFirstMoment[a].variable,u=this.accumulatedSecondMoment[a].variable,l=gu(vu(o,this.beta1),vu(i,1-this.beta1)),c=gu(vu(u,this.beta2),vu(jc(i),1-this.beta2)),h=bu(l,n),p=bu(c,r);o.assign(l),u.assign(c);const d=gu(vu(bu(h,gu(Wh(p),this.epsilon)),-this.learningRate),s);s.assign(d)}),this.accBeta1.assign(vu(this.accBeta1,this.beta1)),this.accBeta2.assign(vu(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),null!=this.accumulatedFirstMoment&&au(this.accumulatedFirstMoment.map(e=>e.variable)),null!=this.accumulatedSecondMoment&&au(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){const e=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(e.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e),ru(()=>{this.accBeta1.assign(ah(this.beta1,this.iterations_+1)),this.accBeta2.assign(ah(this.beta2,this.iterations_+1))});const t=e.length/2;this.accumulatedFirstMoment=e.slice(0,t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)})),this.accumulatedSecondMoment=e.slice(t,2*t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon)}};Od.className="Adam",Oo(Od);var Ld=class extends Fd{constructor(e,t,n,r=null,a=0){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.decay=a,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],ru(()=>{this.iteration=Ah(0).variable(),this.accBeta1=Ah(t).variable()}),null==r&&(this.epsilon=xs.backend.epsilon())}applyGradients(e){const t=Array.isArray(e)?e.map(e=>e.name):Object.keys(e);ru(()=>{const n=gc(1,this.accBeta1),r=bu(-this.learningRate,gu(vu(this.iteration,this.decay),1));t.forEach((t,a)=>{const s=xs.registeredVariables[t];null==this.accumulatedFirstMoment[a]&&(this.accumulatedFirstMoment[a]={originalName:t+"/m",variable:Fl(s).variable(!1)}),null==this.accumulatedWeightedInfNorm[a]&&(this.accumulatedWeightedInfNorm[a]={originalName:t+"/v",variable:Fl(s).variable(!1)});const i=Array.isArray(e)?e[a].tensor:e[t];if(null==i)return;const o=this.accumulatedFirstMoment[a].variable,u=this.accumulatedWeightedInfNorm[a].variable,l=gu(vu(o,this.beta1),vu(i,1-this.beta1)),c=vu(u,this.beta2),h=xu(i),p=Oc(c,h);o.assign(l),u.assign(p);const d=gu(vu(bu(r,n),bu(l,gu(p,this.epsilon))),s);s.assign(d)}),this.iteration.assign(gu(this.iteration,1)),this.accBeta1.assign(vu(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),null!=this.accumulatedFirstMoment&&au(this.accumulatedFirstMoment.map(e=>e.variable)),null!=this.accumulatedWeightedInfNorm&&au(this.accumulatedWeightedInfNorm.map(e=>e.variable))}async getWeights(){throw new Error("getWeights() is not implemented for Adamax yet.")}async setWeights(e){throw new Error("setWeights() is not implemented for Adamax yet.")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon,t.decay)}};Ld.className="Adamax",Oo(Ld);var Pd=class extends Fd{constructor(e){super(),this.learningRate=e,this.setLearningRate(e)}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{const r=Array.isArray(e)?e[n].tensor:e[t];if(null==r)return;const a=xs.registeredVariables[t];ru(()=>{const e=gu(vu(this.c,r),a);a.assign(e)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,null!=this.c&&this.c.dispose(),this.c=su(Ah(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(0!==(e=await this.extractIterations(e)).length)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}};Pd.className="SGD",Oo(Pd);var $d=class extends Pd{constructor(e,t,n=!1){super(e),this.learningRate=e,this.momentum=t,this.useNesterov=n,this.accumulations=[],this.m=Ah(this.momentum)}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{const r=xs.registeredVariables[t];if(null==this.accumulations[n]){const e=!1;this.accumulations[n]={originalName:t+"/momentum",variable:ru(()=>Fl(r).variable(e))}}const a=this.accumulations[n].variable,s=Array.isArray(e)?e[n].tensor:e[t];null!=s&&ru(()=>{let e;const t=gu(vu(this.m,a),s);e=this.useNesterov?gu(vu(this.c,gu(s,vu(t,this.m))),r):gu(vu(this.c,t),r),a.assign(t),r.assign(e)})}),this.incrementIterations()}dispose(){this.m.dispose(),null!=this.accumulations&&au(this.accumulations.map(e=>e.variable))}setMomentum(e){this.momentum=e}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);this.accumulations=e.map(e=>({originalName:e.name,variable:e.tensor.variable(!1)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(e,t){return new e(t.learningRate,t.momentum,t.useNesterov)}};$d.className="Momentum",Oo($d);var Bd=class extends Fd{constructor(e,t=.9,n=0,r=null,a=!1){if(super(),this.learningRate=e,this.decay=t,this.momentum=n,this.epsilon=r,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=a,null==r&&(this.epsilon=xs.backend.epsilon()),null==e)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{const r=xs.registeredVariables[t];null==this.accumulatedMeanSquares[n]&&(this.accumulatedMeanSquares[n]={originalName:t+"/rms",variable:ru(()=>Fl(r).variable(!1))}),null==this.accumulatedMoments[n]&&(this.accumulatedMoments[n]={originalName:t+"/momentum",variable:ru(()=>Fl(r).variable(!1))}),null==this.accumulatedMeanGrads[n]&&this.centered&&(this.accumulatedMeanGrads[n]={originalName:t+"/mg",variable:ru(()=>Fl(r).variable(!1))});const a=Array.isArray(e)?e[n].tensor:e[t];if(null==a)return;const s=this.accumulatedMeanSquares[n].variable,i=this.accumulatedMoments[n].variable;ru(()=>{const e=gu(vu(s,this.decay),vu(jc(a),1-this.decay));if(this.centered){const t=this.accumulatedMeanGrads[n].variable,o=gu(vu(t,this.decay),vu(a,1-this.decay)),u=bu(vu(a,this.learningRate),Wh(gc(e,gu(jc(o),this.epsilon)))),l=gu(vu(i,this.momentum),u);s.assign(e),t.assign(o),i.assign(l);const c=gc(r,l);r.assign(c)}else{const e=gu(vu(s,this.decay),vu(jc(a),1-this.decay)),t=gu(vu(i,this.momentum),bu(vu(a,this.learningRate),Wh(gu(e,this.epsilon))));s.assign(e),i.assign(t);const n=gc(r,t);r.assign(n)}})}),this.incrementIterations()}dispose(){null!=this.accumulatedMeanSquares&&au(this.accumulatedMeanSquares.map(e=>e.variable)),null!=this.accumulatedMeanGrads&&this.centered&&au(this.accumulatedMeanGrads.map(e=>e.variable)),null!=this.accumulatedMoments&&au(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){const e=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&e.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(e.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);const t=this.centered?e.length/3:e.length/2;this.accumulatedMeanSquares=e.slice(0,t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)})),this.accumulatedMoments=e.slice(t,2*t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)})),this.centered&&(this.accumulatedMeanGrads=e.slice(2*t,3*t).map(e=>({originalName:e.name,variable:e.tensor.variable(!1)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(e,t){return new e(t.learningRate,t.decay,t.momentum,t.epsilon,t.centered)}};Bd.className="RMSProp",Oo(Bd);var zd=class{static sgd(e){return new Pd(e)}static momentum(e,t,n=!1){return new $d(e,t,n)}static rmsprop(e,t=.9,n=0,r=null,a=!1){return new Bd(e,t,n,r,a)}static adam(e=.001,t=.9,n=.999,r=null){return new Od(e,t,n,r)}static adadelta(e=.001,t=.95,n=null){return new Md(e,t,n)}static adamax(e=.002,t=.9,n=.999,r=null,a=0){return new Ld(e,t,n,r,a)}static adagrad(e,t=.1){return new Dd(e,t)}},Ud={sgd:zd.sgd,momentum:zd.momentum,adadelta:zd.adadelta,adagrad:zd.adagrad,rmsprop:zd.rmsprop,adamax:zd.adamax,adam:zd.adam},Wd="undefined"!=typeof requestAnimationFrame?requestAnimationFrame:void 0!==o?o:e=>e();function Vd(){return new Promise(e=>Wd(()=>e()))}var jd={};function Gd(e,t){const n=e[0].length;e.forEach((e,t)=>{ue(e.length===n,()=>`Error in concat${n}D: rank of tensors[${t}] must be the same as the rank of the rest (${n})`)}),ue(t>=0&&t<n,()=>`Error in concat${n}D: axis must be between 0 and ${n-1}.`);const r=e[0];e.forEach((e,a)=>{for(let s=0;s<n;s++)ue(s===t||e[s]===r[s],()=>`Error in concat${n}D: Shape of tensors[${a}] (${e}) does not match the shape of the rest (${r}) along the non-concatenated axis ${a}.`)})}function Hd(e,t){const n=e[0].slice();for(let r=1;r<e.length;r++)n[t]+=e[r][t];return n}w(jd,{ERF_A1:()=>af,ERF_A2:()=>sf,ERF_A3:()=>of,ERF_A4:()=>uf,ERF_A5:()=>lf,ERF_P:()=>rf,PARALLELIZE_THRESHOLD:()=>qd,SELU_SCALE:()=>nf,SELU_SCALEALPHA:()=>tf,applyActivation:()=>kp,assertAndGetBroadcastShape:()=>Zi,assertAxesAreInnerMostDims:()=>Sc,assertParamsConsistent:()=>Gd,assignToTypedArray:()=>mf,axesAreInnerMostDims:()=>vc,calculateShapes:()=>po,checkEinsumDimSizes:()=>wf,checkPadOnDimRoundingMode:()=>Hu,combineLocations:()=>xc,complexWithEvenIndex:()=>pf,complexWithOddIndex:()=>df,computeConv2DInfo:()=>Lu,computeConv3DInfo:()=>Pu,computeDefaultPad:()=>$u,computeDilation2DInfo:()=>Mu,computeOptimalWindowSize:()=>Kd,computeOutAndReduceShapes:()=>wc,computeOutShape:()=>Hd,computePool2DInfo:()=>Du,computePool3DInfo:()=>Ou,convertConv2DDataFormat:()=>Gu,decodeEinsumEquation:()=>vf,eitherStridesOrDilationsAreOne:()=>ju,expandShapeToKeepDim:()=>kc,exponent:()=>yf,exponents:()=>gf,fromStringArrayToUint8:()=>Vf,fromUint8ToStringArray:()=>Wf,getAxesPermutation:()=>Ic,getBroadcastDims:()=>Xi,getComplexWithIndex:()=>ff,getEinsumComputePath:()=>kf,getEinsumPermutation:()=>xf,getFusedBiasGradient:()=>wp,getFusedDyActivation:()=>xp,getImageCenter:()=>Yd,getInnerMostAxes:()=>Ec,getPermuted:()=>Qd,getReductionAxes:()=>Qi,getReshaped:()=>Xd,getReshapedPermuted:()=>Zd,getSliceBeginCoords:()=>Jd,getSliceSize:()=>ef,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>Ef,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>Nf,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>Cf,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>_f,getSparseReshapeInputOutputMismatchErrorMessage:()=>Mf,getSparseReshapeInputOutputMultipleErrorMessage:()=>Ff,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>Tf,getSparseReshapeNegativeOutputDimErrorMessage:()=>Rf,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>Pf,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>Df,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>Of,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>Lf,getUndoAxesPermutation:()=>Ac,isIdentityPermutation:()=>Sf,log:()=>oa,mergeRealAndImagArrays:()=>cf,prepareAndValidate:()=>uo,prepareSplitSize:()=>Af,segment_util:()=>$f,shouldFuse:()=>Sp,slice_util:()=>fo,splitRealAndImagArrays:()=>hf,tupleValuesAreOne:()=>Vu,upcastType:()=>cs,validateInput:()=>ho,validateUpdateShape:()=>co,warn:()=>ia});var qd=30;function Kd(e){return e<=qd?e:Pe(e,Math.floor(Math.sqrt(e)))}function Yd(e,t,n){return[n*("number"==typeof e?e:e[0]),t*("number"==typeof e?e:e[1])]}function Xd(e,t,n,r=!0){let a=[];if(r)a=a.concat(t.slice(0)),a.push(e[0]/n),a=a.concat(e.slice(1));else{a=a.concat(e[0]);const n=t.length;for(let r=0;r<n;++r)a=a.concat([e[r+1]/t[r],t[r]]);a=a.concat(e.slice(n+1))}return a}function Qd(e,t,n=!0){const r=[];if(n){r.push(t);for(let n=t+1;n<e;++n)n<=2*t?(r.push(n),r.push(n-(t+1))):r.push(n)}else{const n=[],a=[];for(let r=1;r<e;++r)r>=2*t+1||r%2==1?a.push(r):n.push(r);r.push(...n),r.push(0),r.push(...a)}return r}function Zd(e,t,n,r=!0){const a=[];r?a.push(e[0]/n):a.push(e[0]*n);for(let n=1;n<e.length;++n)n<=t.length?r?a.push(t[n-1]*e[n]):a.push(e[n]/t[n-1]):a.push(e[n]);return a}function Jd(e,t){const n=[0];for(let r=0;r<t;++r)n.push(e[r][0]);return n}function ef(e,t,n){const r=e.slice(0,1);for(let a=0;a<n;++a)r.push(e[a+1]-t[a][0]-t[a][1]);return r}var tf=1.7580993408473768,nf=1.0507009873554805,rf=.3275911,af=.254829592,sf=-.284496736,of=1.421413741,uf=-1.453152027,lf=1.061405429;function cf(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);const n=new Float32Array(2*e.length);for(let r=0;r<n.length;r+=2)n[r]=e[r/2],n[r+1]=t[r/2];return n}function hf(e){const t=new Float32Array(e.length/2),n=new Float32Array(e.length/2);for(let r=0;r<e.length;r+=2)t[r/2]=e[r],n[r/2]=e[r+1];return{real:t,imag:n}}function pf(e){const t=Math.ceil(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let t=0;t<e.length;t+=4)n[Math.floor(t/4)]=e[t],r[Math.floor(t/4)]=e[t+1];return{real:n,imag:r}}function df(e){const t=Math.floor(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let t=2;t<e.length;t+=4)n[Math.floor(t/4)]=e[t],r[Math.floor(t/4)]=e[t+1];return{real:n,imag:r}}function ff(e,t){return{real:e[2*t],imag:e[2*t+1]}}function mf(e,t,n,r){e[2*r]=t,e[2*r+1]=n}function gf(e,t){const n=new Float32Array(e/2),r=new Float32Array(e/2);for(let a=0;a<Math.ceil(e/2);a++){const s=(t?2:-2)*Math.PI*(a/e);n[a]=Math.cos(s),r[a]=Math.sin(s)}return{real:n,imag:r}}function yf(e,t,n){const r=(n?2:-2)*Math.PI*(e/t);return{real:Math.cos(r),imag:Math.sin(r)}}var bf=/->/g;function vf(e,t){const n=((e=e.replace(/\s/g,"")).length-e.replace(bf,"").length)/"->".length;if(n<1)throw new Error("Equations without an arrow are not supported.");if(n>1)throw new Error('Equation must contain exactly one arrow ("->").');const[r,a]=e.split("->");ue(-1===r.indexOf("..."),()=>'The ellipsis notation ("...") is not supported yet.');const s=r.split(","),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");const o=[];for(let e=0;e<a.length;++e){const t=a[e];if(!s.some(e=>-1!==e.indexOf(t)))throw new Error(`Output subscripts contain the label ${t} not present in the input subscripts.`);-1===o.indexOf(t)&&o.push(t)}for(let e=0;e<r.length;++e){const t=r[e];-1===o.indexOf(t)&&","!==t&&o.push(t)}const u=new Array(s.length);for(let e=0;e<i;++e){if(new Set(s[e].split("")).size!==s[e].length)throw new Error(`Found duplicate axes in input component ${s[e]}. Support for duplicate axes in input is not implemented yet.`);u[e]=[];for(let t=0;t<s[e].length;++t)u[e].push(o.indexOf(s[e][t]))}const l=o.length,c=[];for(let e=a.length;e<l;++e)c.push(e);return{allDims:o,summedDims:c,idDims:u}}function xf(e,t){let n=new Array(e);n.fill(-1);for(let e=0;e<t.length;++e)n[t[e]]=e;const r=[];for(let t=0;t<e;++t)-1===n[t]&&r.push(t);return n=n.filter(e=>-1!==e),{permutationIndices:n,expandDims:r}}function wf(e,t,n){const r=new Array(e);for(let e=0;e<n.length;++e){const a=n[e].shape;for(let n=0;n<t[e].length;++n)void 0===r[t[e][n]]?r[t[e][n]]=a[n]:ue(r[t[e][n]]===a[n],()=>`Expected dimension ${r[t[e][n]]} at axis ${n} of input shaped ${JSON.stringify(a)}, but got dimension ${a[n]}`)}}function kf(e,t){const n=e,r=[];let a=0;0===e.length&&n.push(-1),a=e.length+1;for(let e=0;e<a;++e)r.push([]);const s=[];for(let e=0;e<n.length;++e){const a=If(t,n[e]);for(const t of a)-1===s.indexOf(t)&&(r[e].push(t),s.push(t))}return{path:n,steps:r}}function Sf(e){return e.every((e,t)=>e===t)}function If(e,t){const n=[];for(let r=0;r<e.length;++r)0!==e[r].length&&-1===e[r].indexOf(t)&&-1!==t||n.push(r);return n}function Af(e,t,n=0){let r=[];if("number"==typeof t)ue(e.shape[n]%t==0,()=>"Number of splits must evenly divide the axis."),r=new Array(t).fill(e.shape[n]/t);else{ue(t.reduce((e,t)=>(-1===t&&(e+=1),e),0)<=1,()=>"There should be only one negative value in split array.");const a=t.indexOf(-1);if(-1!==a){const r=t.reduce((e,t)=>t>0?e+t:e);t[a]=e.shape[n]-r}ue(e.shape[n]===t.reduce((e,t)=>e+t),()=>"The sum of sizes must match the size of the axis dimension."),r=t}return r}function Ef(e){return"Received SparseTensor with denseShape[0] = 0 but\n  indices.shape[0] = "+e}function Nf(e,t){return`indices(${e}, 0) is invalid: ${t} < 0`}function Cf(e,t,n){return`indices(${e}, 0) is invalid: ${t} >= ${n}`}function Tf(e,t){return`only one output dimension may be -1, not both ${e} and ${t}`}function Rf(e,t){return`size ${e} must be non-negative, not ${t}`}function _f(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function Ff(e,t){return`Input to reshape is a SparseTensor with ${pe(e)}\n  dense values, but the requested shape requires a multiple of ${pe(t)}. inputShape=${e} outputShape= ${t}`}function Mf(e,t){return`Input to reshape is a tensor with ${pe(e)} dense values, but the requested shape has ${pe(t)}. inputShape=${e} outputShape=${t}`}function Df(){return"segment ids must be >= 0"}function Of(){return"segment ids are not increasing"}function Lf(e,t){return`Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function Pf(e,t,n){return`Bad: indices[${e}] == ${t} out of range [0, ${n})`}var $f={};function Bf(e,t){let n,r=!1;for(e<=qd?(n=e,r=!0):n=Pe(e,Math.floor(Math.sqrt(e)));!r;)n>t||n===e?r=!0:n=Pe(e,n+1);return n}function zf(e,t,n){const r=[],a=e.length;for(let s=0;s<a;s++)s!==t?r.push(e[s]):r.push(n);return r}function Uf(e,t,n,r){const a=t.shape.length,s=e.shape.length;if(0!==r&&(r<-a||r>a))throw new Error(`Expect batchDims in the range of [-${a}, ${a}], but got ${r}`);if(r<0&&(r+=a),r>s)throw new Error(`batchDims (${r}) must be less than rank(x) (\n    ${s}).`);if(n<r)throw new Error(`batchDims (${r}) must be less than or equal to axis (${n}).`);for(let n=0;n<r;++n)if(e.shape[n]!==t.shape[n])throw new Error(`x.shape[${n}]: ${e.shape[n]} should be equal to indices.shape[${n}]: ${t.shape[n]}.`);const i=e.shape[n],o=[];let u=1,l=1,c=1;for(let t=0;t<r;++t)o.push(e.shape[t]),u*=e.shape[t];for(let t=r;t<n;t++)o.push(e.shape[t]),l*=e.shape[t];for(let e=r;e<a;e++)o.push(t.shape[e]);for(let t=n+1;t<s;t++)o.push(e.shape[t]),c*=e.shape[t];return{batchSize:u,sliceSize:c,outerSize:l,dimSize:i,outputShape:o}}function Wf(e){try{return e.map(e=>Ba(e))}catch(e){throw new Error("Failed to decode encoded string bytes into utf-8, error: "+e)}}function Vf(e){return e.map(e=>$a(e))}w($f,{collectGatherOpShapeInfo:()=>Uf,computeOutShape:()=>zf,segOpComputeOptimalWindowSize:()=>Bf});var jf={};w(jf,{nonMaxSuppressionV3Impl:()=>Up,nonMaxSuppressionV4Impl:()=>Wp,nonMaxSuppressionV5Impl:()=>Vp,whereImpl:()=>ip});var Gf={kernelName:et,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,Hh(Ai(n,"float32"),-1))}}},Hf={kernelName:tt,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const t=jc(Ai(n,"float32")),r=Wh(gc(Ah(1),t));return pc(bu(e,r))}}}},qf={kernelName:nt,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const t=Wh(gc(jc(Ai(n,"float32")),1));return bu(e,t)}}}},Kf={kernelName:rt,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{let t=e;const r=Qi(n.shape,a);return r.length>0&&(t=yc(t,r)),qu(t,n.shape)},b:()=>{let t=e;const n=Qi(r.shape,a);return n.length>0&&(t=yc(t,n)),qu(t,r.shape)}}}},Yf={kernelName:at,saveAllInputs:!0,gradFunc:(e,t)=>{const n={};return t.forEach((t,r)=>{n[r]=()=>e.clone()}),n}},Xf={kernelName:ot,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>Fl(n)}}},Qf={kernelName:ut,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>Fl(n)}}},Zf={kernelName:lt,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,Wh(gc(Ah(1),jc(Ai(n,"float32")))))}}},Jf={kernelName:ct,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const t=Wh(gu(Ah(1),jc(Ai(n,"float32"))));return bu(e,t)}}}},em={kernelName:dt,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{const t=gu(jc(n),jc(r));let s=vu(e,bu(r,t));const i=Qi(n.shape,a);return i.length>0&&(s=yc(s,i)),qu(s,n.shape)},b:()=>{const t=gu(jc(n),jc(r));let s=pc(vu(e,bu(n,t)));const i=Qi(r.shape,a);return i.length>0&&(s=yc(s,i)),qu(s,r.shape)}}}},tm={kernelName:ht,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,gu(jc(Ai(n,"float32")),1))}}},nm={kernelName:pt,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,gc(Ah(1),jc(Ai(n,"float32"))))}}};var rm=Ms({avgPool3dGrad_:function(e,t,n,r,a,s){const i=Rs(e,"dy","avgPool3dGrad"),o=Rs(t,"input","avgPool3dGrad");let u=i,l=o,c=!1;4===o.rank&&(c=!0,u=qu(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),l=qu(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),ue(5===u.rank,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${u.rank}.`),ue(5===l.rank,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${l.rank}.`),Hu("avgPool3dGrad",a,s);const h={dy:u,input:l},p={filterSize:n,strides:r,pad:a,dimRoundingMode:s},d=xs.runKernel(yt,h,p);return c?qu(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}}),am={kernelName:gt,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{filterSize:a,strides:s,pad:i,dimRoundingMode:o}=n;return{x:()=>rm(e,r,a,s,i,o)}}};var sm=Ms({avgPoolGrad_:function(e,t,n,r,a){const s=Rs(e,"dy","avgPoolGrad"),i=Rs(t,"input","avgPoolGrad");ue(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let o=i,u=s,l=!1;3===i.rank&&(l=!0,o=qu(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=qu(s,[1,s.shape[0],s.shape[1],s.shape[2]])),ue(4===u.rank,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${u.rank}.`),ue(4===o.rank,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${o.rank}.`);const c={dy:u,input:o},h={filterSize:n,strides:r,pad:a},p=xs.runKernel(mt,c,h);return l?qu(p,[p.shape[1],p.shape[2],p.shape[3]]):p}}),im={kernelName:ft,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{filterSize:a,strides:s,pad:i}=n;return{x:()=>sm(e,r,a,s,i)}}},om={kernelName:bt,inputsToSave:["a","b"],gradFunc:(e,t,n)=>{const[r,a]=t,{transposeA:s,transposeB:i}=n;return s||i?!s&&i?{a:()=>Gi(e,a,!1,!1),b:()=>Gi(e,r,!0,!1)}:s&&!i?{a:()=>Gi(a,e,!1,!0),b:()=>Gi(r,e,!1,!1)}:{a:()=>Gi(a,e,!0,!0),b:()=>Gi(e,r,!0,!0)}:{a:()=>Gi(e,a,!1,!0),b:()=>Gi(r,e,!0,!1)}}},um={kernelName:vt,gradFunc:(e,t,n)=>{const{blockShape:r,crops:a}=n;return{x:()=>nh(e,r,a)}}},lm={kernelName:wt,gradFunc:(e,t,n)=>{const r=n,a=r.inputShape,s=r.shape,i=Array.from(s);for(let e=a.length-1;e>=0;e--)if(a[e]===s[e])i[e]=1;else if(1!==a[e])throw new Error(`broadcastTo(): [${a}] cannot be broadcast to [${s}].`);const o=[];for(let e=0;e<i.length;e++)i[e]>1&&o.push(e);return{x:()=>yc(e,o,!0)}}},cm={kernelName:St,gradFunc:e=>({x:()=>e.clone()})},hm={kernelName:It,gradFunc:e=>({x:()=>Fl(e)})},pm={kernelName:At,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{clipValueMin:a,clipValueMax:s}=n;return{x:()=>_l(Cc(ql(r,a),ec(r,s)),e,Fl(e))}}},dm={kernelName:Nt,inputsToSave:["x"],gradFunc:Gf.gradFunc},fm={kernelName:Ct,saveAllInputs:!0,gradFunc:(e,t,n)=>{const r=t.map(e=>e.shape),{axis:a}=n,s=ke(a,t[0].shape)[0],i=r.map(e=>e[s]);return zh(e,i,s).map(e=>()=>e)}},mm={kernelName:Tt,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const[r,a]=t,{dilations:s,strides:i,pad:o,dataFormat:u}=n;return ue(Vu(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>yl(r.shape,e,a,i,o,u),filter:()=>vp(r,e,a.shape,i,o,u)}}},gm={kernelName:_t,inputsToSave:["dy","filter"],gradFunc:(e,t,n)=>{const[r,a]=t,{strides:s,pad:i,dataFormat:o,dimRoundingMode:u}=n;return{dy:()=>ml(e,a,s,i,o,1,u),filter:()=>vp(e,r,a.shape,s,i,o,u)}}};var ym=Ms({conv3DBackpropFilter_:function(e,t,n,r,a){let s=e;4===e.rank&&(s=qu(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let i=t;4===i.rank&&(i=qu(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),ue(5===s.rank,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),ue(5===i.rank,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),ue(5===n.length,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`),ue(s.shape[4]===n[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${n[3]}.`),ue(i.shape[4]===n[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${n[4]}).`);const o={x:s,dy:i},u={strides:r,pad:a,filterShape:n};return xs.runKernel(Mt,o,u)}}),bm={kernelName:Ft,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const{dilations:r,strides:a,pad:s}=n;ue(Vu(r),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${r}'`);const[i,o]=t;return{x:()=>xl(i.shape,e,o,a,s),filter:()=>ym(i,e,o.shape,a,s)}}},vm={kernelName:Ot,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(pc(Rh(Ai(n,"float32"))),e)}}},xm={kernelName:Lt,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(_h(Ai(n,"float32")),e)}}},wm={kernelName:Pt,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{axis:a,exclusive:s,reverse:i}=n;return{x:()=>{const t=Ic([a],r.rank);let n=Il(e,a,s,!i);return null!=t&&(n=qi(n,t)),n}}}},km={kernelName:Ut,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const{dilations:r,strides:a,pad:s,dimRoundingMode:i}=n,o=null==r?[1,1]:r;ue(Vu(o),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);const[u,l]=t;return ue(4===u.rank,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${u.rank}.`),ue(4===l.rank,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${l.rank}.`),ue(u.shape[3]===l.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${u.shape[3]}) must match the inChannels dimension in filter ${l.shape[2]}.`),ue(ju(a,o),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${a} and dilations '${o}'.`),Hu("depthwiseConv2d",s,i),{x:()=>Ep(u.shape,e,l,a,s,o,i),filter:()=>Ap(u,e,l.shape,a,s,o,i)}}},Sm={kernelName:Gt,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const[r,a]=t,s={x:r,filter:a,dy:e},i={x:r,filter:a,dy:e};return{x:()=>xs.runKernel(Ht,s,n),filter:()=>xs.runKernel(qt,i,n)}}},Im={kernelName:Xt,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t,r={dy:e,y:n};return{x:()=>xs.runKernel(Qt,r)}}},Am={kernelName:Zt,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t,r=vu($l(pc(jc(n))),2/Math.sqrt(Math.PI));return{x:()=>vu(e,r)}}},Em={kernelName:en,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,n)}}},Nm={kernelName:tn,inputsToSave:["input"],gradFunc:(e,t)=>{const[n]=t;return{input:()=>qu(e,n.shape)}}},Cm={kernelName:nn,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,$l(n))}}},Tm={kernelName:on,gradFunc:e=>({x:()=>Fl(e)})},Rm={kernelName:un,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{const t=bu(e,Ai(r,"float32")),s=Qi(n.shape,a);return s.length>0?qu(yc(t,s),n.shape):t},b:()=>{let t=vu(e,Ai(n,"float32"));const s=Qi(r.shape,a);s.length>0&&(t=qu(yc(t,s),r.shape));const i=jc(r);return pc(bu(t,Ai(i,"float32")))}}}},_m={kernelName:ln,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,n)=>{const{varianceEpsilon:r}=n,[a,s,i,o]=t,u=null==o?Ah(1):o,l=Qi(s.shape,a.shape),c=[];if(1===s.rank){for(let e=0;e<a.shape.length-1;++e)c.push(a.shape[e]);c.push(1)}const h=gc(a,s),p=vu(e,u),d=Ih(gu(i,Ah(r))),f=vu(vu(vu(d,d),d),Ah(-.5));return{x:()=>1===s.rank?qu(vu(vu(e,Ul(qu(d,[1,1,1,s.shape[0]]),c)),u),a.shape):qu(vu(vu(e,d),u),a.shape),mean:()=>{let e=vu(vu(d,Ah(-1)),p);return 1===s.rank&&(e=yc(e,l)),qu(e,s.shape)},variance:()=>{let e=vu(vu(f,h),p);return 1===s.rank&&(e=yc(e,l)),qu(e,s.shape)},scale:()=>{const t=vu(h,d);let n=vu(e,t);return 1===s.rank&&(n=yc(n,l)),qu(n,s.shape)},offset:()=>{let t=e;return 1===s.rank&&(t=yc(t,l)),qu(t,s.shape)}}}},Fm={kernelName:cn,inputsToSave:["x","indices"],gradFunc:(e,t,n)=>{const[r,a]=t,{axis:s}=n,i=ke(s,r.shape)[0];return{x:()=>{const t=r.shape,n=a.size,o=t.slice(0,i),u=o.length,l=t.slice(s,t.length).slice(1),c=l.length,h=Mm(0,u),p=Mm(u+1,u+1+c),d=Dm([o,[n],l]),f=qu(e,d),m=qu(a,[n]),g=Dm([[u],h,p]),y=qi(f,g);let b=rp(y,m,r.shape[i]);const v=Ac(g);return b=qi(b,v),b},indices:()=>a}}};function Mm(e,t){const n=[];for(let r=e;r<t;++r)n.push(r);return n}function Dm(e){const t=[];for(let n=0;n<e.length;++n)for(let r=0;r<e[n].length;++r)t.push(e[n][r]);return t}var Om={kernelName:dn,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t;return{a:()=>Fl(n),b:()=>Fl(r)}}},Lm={kernelName:fn,gradFunc:e=>({x:()=>Ai(e,"float32")})},Pm={kernelName:yn,gradFunc:e=>({x:()=>Fl(e)})},$m={kernelName:bn,gradFunc:e=>({x:()=>Fl(e)})},Bm={kernelName:vn,gradFunc:e=>({x:()=>Fl(e)})},zm={kernelName:xn,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{alpha:a}=n,s=Hl(r,0);return{x:()=>_l(s,e,vu(e,a))}}},Um={kernelName:An,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,gu(n,1))}}},Wm={kernelName:In,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,Ai(n,"float32"))}}},Vm={kernelName:Tn,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r]=t,{axis:a}=n;return{logits:()=>{const t=$l(r);return gc(e,vu(yc(e,a,!0),t))}}}};var jm=Ms({localResponseNormalizationBackprop_:function(e,t,n,r=5,a=1,s=1,i=.5){const o={x:e,y:t,dy:n},u={depthRadius:r,bias:a,alpha:s,beta:i};return xs.runKernel(_n,o,u)}}),Gm={kernelName:Rn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r,a]=t,{depthRadius:s,bias:i,alpha:o,beta:u}=n;return{x:()=>jm(r,a,e,s,i,o,u)}}};function Hm(e,t,n,r){return t.rank<n.rank&&(t=qu(t,kc(t.shape,r))),e.rank<n.rank&&(e=qu(e,kc(e.shape,r))),{x:()=>vu(e,Ai(Rl(n,t),e.dtype))}}var qm={kernelName:Fn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const r=n,{reductionIndices:a}=r,s=t[0],i=Hm(e,t[1],s,ke(a,s.shape));return{x:()=>i.x()}}},Km={kernelName:Mn,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t;return{a:()=>vu(e,Ai(ql(n,r),"float32")),b:()=>vu(e,Ai(Jl(n,r),"float32"))}}};var Ym=Ms({maxPool3dGrad_:function(e,t,n,r,a,s,i){const o=Rs(e,"dy","maxPool3dGrad"),u=Rs(t,"input","maxPool3dGrad"),l=Rs(n,"output","maxPool3dGrad");let c=o,h=u,p=l,d=!1;4===u.rank&&(d=!0,c=qu(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]]),h=qu(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]]),p=qu(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]])),ue(5===c.rank,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${c.rank}.`),ue(5===h.rank,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${h.rank}.`),ue(5===p.rank,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${p.rank}.`),Hu("maxPool3dGrad",s,i);const f={dy:c,input:h,output:p},m={filterSize:r,strides:a,pad:s,dimRoundingMode:i},g=xs.runKernel(Pn,f,m);return d?qu(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}}),Xm={kernelName:Ln,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r,a]=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:u}=n;return{x:()=>Ym(e,r,a,s,i,o,u)}}};var Qm=Ms({maxPoolGrad_:function(e,t,n,r,a,s,i){const o=Rs(e,"dy","maxPoolGrad"),u=Rs(t,"input","maxPoolGrad"),l=Rs(n,"output","maxPoolGrad");ue(u.rank===o.rank,()=>`Rank of input (${u.rank}) does not match rank of dy (${o.rank})`),ue(4===o.rank,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${o.rank}.`),ue(4===u.rank,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${u.rank}.`),Hu("maxPoolGrad",s,i);const c={dy:o,input:u,output:l},h={filterSize:r,strides:a,pad:s,dimRoundingMode:i};return xs.runKernel(On,c,h)}}),Zm={kernelName:er,inputsToSave:["x"],gradFunc:(e,t,n)=>{const r=t[0],{paddings:a}=n,s=a.map(e=>e[0]);return{x:()=>Zu(e,s,r.shape)}}},Jm={kernelName:Tr,gradFunc:(e,t,n)=>{const{blockShape:r,paddings:a}=n;return{x:()=>tl(e,r,a)}}},eg={kernelName:Rr,gradFunc:(e,t,n)=>{const{axis:r}=n;return{x:()=>Xu(e,r)}}};var tg=[Gf,Hf,qf,Kf,Yf,Xf,Qf,Zf,Jf,em,tm,nm,am,im,om,um,lm,cm,hm,pm,dm,fm,gm,mm,bm,vm,xm,wm,km,Sm,{kernelName:Kt,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{const t=bu(e,Ai(r,"float32")),s=Qi(n.shape,a);return s.length>0?qu(yc(t,s),n.shape):t},b:()=>{let t=vu(e,Ai(n,"float32"));const s=Qi(r.shape,a);s.length>0&&(t=qu(yc(t,s),r.shape));const i=jc(r);return pc(bu(t,Ai(i,"float32")))}}}},Im,Am,Em,Nm,Cm,Rm,Tm,_m,Fm,Om,Lm,Pm,$m,Bm,zm,Um,Wm,Vm,Gm,qm,qm,Km,Xm,{kernelName:Dn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r,a]=t,{filterSize:s,strides:i,pad:o}=n;return{x:()=>Qm(e,r,a,s,i,o)}}},{kernelName:Bn,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{axis:a}=n,s=ke(a,r.shape),i=pe(wc(r.shape,s)[1]);return{x:()=>{const t=r.shape.slice();s.forEach(e=>{t[e]=1});const n=qu(e,t);return bu(vu(n,$c(r.shape,"float32")),i)}}}},{kernelName:zn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const r=n,{axis:a}=r,[s,i]=t,o=Hm(e,i,s,ke(a,s.shape));return{x:()=>o.x()}}},{kernelName:Un,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t;return{a:()=>vu(e,Ai(ec(n,r),"float32")),b:()=>vu(e,Ai(Hl(n,r),"float32"))}}},{kernelName:Wn,inputsToSave:["x"],gradFunc:(e,t,n)=>{const r=t[0],{paddings:a}=n,s=a.map(e=>e[0]);return{x:()=>Zu(e,s,r.shape)}}},{kernelName:Vn,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{const t=Qi(n.shape,a);return t.length>0?qu(yc(e,t),n.shape):e},b:()=>{const t=vu(e,pc(jl(bu(n,r)))),s=Qi(r.shape,a);return s.length>0?qu(yc(t,s),r.shape):t}}}},{kernelName:Gn,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{const t=vu(e,Ai(r,"float32")),s=Qi(n.shape,a);return s.length>0?qu(yc(t,s),n.shape):t},b:()=>{const t=vu(e,Ai(n,"float32")),s=Qi(r.shape,a);return s.length>0?qu(yc(t,s),r.shape):t}}}},{kernelName:Hn,gradFunc:e=>({x:()=>pc(e)})},{kernelName:Zn,inputsToSave:["indices"],gradFunc:(e,t)=>{const n=t[0];return{indices:()=>Pc(n.shape,"float32")}}},{kernelName:Qn,gradFunc:e=>({x:()=>Fl(e)})},{kernelName:Jn,saveAllInputs:!0,gradFunc:(e,t,n)=>{const{axis:r}=n;return ap(e,r).map(e=>()=>e)}},Zm,Zm,{kernelName:nr,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{const[n,r,a]=t,s=n,i=r,o=Zi(s.shape,i.shape);return{a:()=>{const t=Ai(i,"float32");let n=vu(e,vu(t,ah(s,gc(t,Ah(1)))));const r=Qi(s.shape,o);return r.length>0&&(n=yc(n,r)),qu(n,s.shape)},b:()=>{const t=Hl(s,0),n=_l(t,rc(s),Fl(s));let r=vu(e,vu(a,n));const u=Qi(i.shape,o);return u.length>0&&(r=yc(r,u)),qu(r,i.shape)}}}},{kernelName:rr,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{const[n,r]=t,a=Hl(n,0);return{x:()=>_l(a,e,vu(e,r)),alpha:()=>{let t=_l(a,Fl(e),vu(e,n));const s=Qi(r.shape,e.shape);return s.length>0&&(t=yc(t,s)),qu(t,r.shape)}}}},{kernelName:or,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,pc(jc(n)))}}},{kernelName:fr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t,r=vu(ec(n,6),Hh(n));return{x:()=>vu(e,Ai(r,"float32"))}}},{kernelName:ur,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,Ai(Hh(n),"float32"))}}},{kernelName:lr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>qu(e,n.shape)}}},{kernelName:pr,inputsToSave:["images"],gradFunc:(e,t,n)=>{const[r]=t,a={dy:e,images:r};return{images:()=>xs.runKernel(dr,a,n)}}},{kernelName:cr,inputsToSave:["images"],gradFunc:(e,t,n)=>{const[r]=t,a={dy:e,images:r};return{images:()=>xs.runKernel(hr,a,n)}}},{kernelName:mr,gradFunc:(e,t,n)=>{const{dims:r}=n,a=ke(r,e.shape);return{x:()=>bh(e,a)}}},{kernelName:gr,gradFunc:e=>({x:()=>Fl(e)})},{kernelName:yr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>pc(bu(e,vu(ah(n,1.5),2)))}}},{kernelName:vr,inputsToSave:["condition"],gradFunc:(e,t)=>{const[n]=t;return{condition:()=>Ai(Fl(n),"float32"),t:()=>vu(e,Ai(n,e.dtype)),e:()=>vu(e,Ai(Tc(n),e.dtype))}}},{kernelName:xr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const t=Hl(n,Ah(0)),r=Ah(tf),a=Ah(nf),s=vu(e,a),i=vu(vu(e,r),$l(Ai(n,"float32")));return _l(t,s,i)}}}},{kernelName:Ar,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,vu(n,gc(Ah(1),n)))}}},{kernelName:Ir,gradFunc:e=>({x:()=>Fl(e)})},{kernelName:kr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(kl(Ai(n,"float32")),e)}}},{kernelName:Sr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(Sl(Ai(n,"float32")),e)}}},{kernelName:wr,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{begin:a,size:s}=n,i=r.shape,[o,u]=To(r,a,s),l=[];for(let t=0;t<e.rank;t++)l.push([o[t],i[t]-o[t]-u[t]]);return{x:()=>Qc(e,l)}}},{kernelName:_r,outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r]=t,{dim:a}=n,s=vu(e,r);return{logits:()=>gc(s,vu(yc(s,[a],!0),r))}}},{kernelName:Er,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,Qu(n))}}},Jm,Jm,eg,eg,{kernelName:Nr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,vu(Wh(Ai(n,"float32")),2))}}},{kernelName:Pr,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Ah(2);return{a:()=>vu(e,vu(a,gc(n,r))),b:()=>vu(e,vu(a,gc(r,n)))}}},{kernelName:$r,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(e,vu(Ai(n,"float32"),2))}}},{kernelName:ea,gradFunc:e=>({x:()=>Fl(e)})},{kernelName:Vr,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,a=Zi(n.shape,r.shape);return{a:()=>{let t=e;const r=Qi(n.shape,a);return r.length>0&&(t=yc(t,r)),qu(t,n.shape)},b:()=>{let t=e;const n=Qi(r.shape,a);return n.length>0&&(t=yc(t,n)),qu(pc(t),r.shape)}}}},{kernelName:Cr,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,a=r.shape.slice(),{axis:s}=n;ke(s,r.shape).forEach(e=>{a[e]=1});const i=qu(e,a),o=vu(i,$c(r.shape,"float32"));return{x:()=>o}}},{kernelName:jr,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>bu(e,jc(kl(n)))}}},{kernelName:Gr,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t;return{x:()=>vu(gc(Ah(1),jc(n)),e)}}},{kernelName:Hr,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{reps:a}=n;return{x:()=>{let t=Fl(r);if(1===r.rank)for(let n=0;n<a[0];++n)t=gu(t,Zu(e,[n*r.shape[0]],[r.shape[0]]));else if(2===r.rank)for(let n=0;n<a[0];++n)for(let s=0;s<a[1];++s)t=gu(t,Zu(e,[n*r.shape[0],s*r.shape[1]],[r.shape[0],r.shape[1]]));else if(3===r.rank)for(let n=0;n<a[0];++n)for(let s=0;s<a[1];++s)for(let i=0;i<a[2];++i)t=gu(t,Zu(e,[n*r.shape[0],s*r.shape[1],i*r.shape[2]],[r.shape[0],r.shape[1],r.shape[2]]));else{if(4!==r.rank)throw new Error(`Gradient for tile operation is not implemented for rank-${r.rank} tensors yet.`);for(let n=0;n<a[0];++n)for(let s=0;s<a[1];++s)for(let i=0;i<a[2];++i)for(let o=0;o<a[3];++o)t=gu(t,Zu(e,[n*r.shape[0],s*r.shape[1],i*r.shape[2],o*r.shape[3]],[r.shape[0],r.shape[1],r.shape[2],r.shape[3]]))}return t}}}},{kernelName:Yr,gradFunc:(e,t,n)=>{const r=n,{perm:a}=r,s=Ac(a);return{x:()=>qi(e,s)}}},{kernelName:Qr,gradFunc:(e,t,n)=>{const r=n,{axis:a}=r;return{value:()=>Gh(e,a)}}},{kernelName:Zr,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>function(e,t){const n=Oc(t,Fl(t)),r=Gl(e,n);let a=ql(t,Ah(0,"int32"));const s=r.rank-a.rank;for(let e=0;e<s;++e)a=Bl(a,e+1);a=Cc(a,$c(r.shape,"bool"));const i=Fl(r);return _l(a,r,i)}(e,n)}}},{kernelName:Jr,gradFunc:e=>({x:()=>Fl(e)})}];for(const e of tg)fa(e);Xa().prototype.abs=function(){return this.throwIfDisposed(),xu(this)},Xa().prototype.acos=function(){return this.throwIfDisposed(),wu(this)},Xa().prototype.acosh=function(){return this.throwIfDisposed(),ku(this)},Xa().prototype.add=function(e){return this.throwIfDisposed(),gu(this,e)},Xa().prototype.all=function(e,t){return this.throwIfDisposed(),Iu(this,e,t)},Xa().prototype.any=function(e,t){return this.throwIfDisposed(),Au(this,e,t)},Xa().prototype.argMax=function(e){return this.throwIfDisposed(),Eu(this,e)},Xa().prototype.argMin=function(e){return this.throwIfDisposed(),Nu(this,e)},Xa().prototype.asScalar=function(){return this.throwIfDisposed(),ue(1===this.size,()=>"The array must have only 1 element."),qu(this,[])},Xa().prototype.asType=function(e){return this.throwIfDisposed(),Ai(this,e)},Xa().prototype.as1D=function(){return this.throwIfDisposed(),qu(this,[this.size])},Xa().prototype.as2D=function(e,t){return this.throwIfDisposed(),qu(this,[e,t])},Xa().prototype.as3D=function(e,t,n){return this.throwIfDisposed(),qu(this,[e,t,n])},Xa().prototype.as4D=function(e,t,n,r){return this.throwIfDisposed(),qu(this,[e,t,n,r])},Xa().prototype.as5D=function(e,t,n,r,a){return this.throwIfDisposed(),qu(this,[e,t,n,r,a])},Xa().prototype.asin=function(){return this.throwIfDisposed(),Cu(this)},Xa().prototype.asinh=function(){return this.throwIfDisposed(),Tu(this)},Xa().prototype.atan=function(){return this.throwIfDisposed(),Ru(this)},Xa().prototype.atan2=function(e){return this.throwIfDisposed(),_u(this,e)},Xa().prototype.atanh=function(){return this.throwIfDisposed(),Fu(this)},Xa().prototype.avgPool=function(e,t,n,r){return this.throwIfDisposed(),Ku(this,e,t,n,r)},Xa().prototype.batchToSpaceND=function(e,t){return this.throwIfDisposed(),tl(this,e,t)},Xa().prototype.batchNorm=function(e,t,n,r,a){return this.throwIfDisposed(),nl(this,e,t,n,r,a)},Xa().prototype.broadcastTo=function(e){return this.throwIfDisposed(),ul(this,e)},Xa().prototype.cast=function(e){return this.throwIfDisposed(),Ai(this,e)},Xa().prototype.ceil=function(){return this.throwIfDisposed(),ll(this)},Xa().prototype.clipByValue=function(e,t){return this.throwIfDisposed(),cl(this,e,t)},Xa().prototype.concat=function(e,t){return this.throwIfDisposed(),e instanceof Ya&&(e=[e]),Xu([this,...e],t)},Xa().prototype.conv1d=function(e,t,n,r,a,s){return this.throwIfDisposed(),gl(this,e,t,n,r,a,s)},Xa().prototype.conv2dTranspose=function(e,t,n,r,a){return this.throwIfDisposed(),bl(this,e,t,n,r,a)},Xa().prototype.conv2d=function(e,t,n,r,a,s){return this.throwIfDisposed(),ml(this,e,t,n,r,a,s)},Xa().prototype.cos=function(){return this.throwIfDisposed(),kl(this)},Xa().prototype.cosh=function(){return this.throwIfDisposed(),Sl(this)},Xa().prototype.cumsum=function(e,t,n){return this.throwIfDisposed(),Il(this,e,t,n)},Xa().prototype.depthToSpace=function(e,t){return this.throwIfDisposed(),El(this,e,t)},Xa().prototype.depthwiseConv2d=function(e,t,n,r,a,s){return this.throwIfDisposed(),Nl(this,e,t,n,r,a,s)},Xa().prototype.dilation2d=function(e,t,n,r,a){return this.throwIfDisposed(),Tl(this,e,t,n,r,a)},Xa().prototype.divNoNan=function(e){return this.throwIfDisposed(),Ml(this,e)},Xa().prototype.div=function(e){return this.throwIfDisposed(),bu(this,e)},Xa().prototype.dot=function(e){return this.throwIfDisposed(),Dl(this,e)},Xa().prototype.elu=function(){return this.throwIfDisposed(),Ll(this)},Xa().prototype.equal=function(e){return this.throwIfDisposed(),Rl(this,e)},Xa().prototype.erf=function(){return this.throwIfDisposed(),Pl(this)},Xa().prototype.exp=function(){return this.throwIfDisposed(),$l(this)},Xa().prototype.expandDims=function(e){return this.throwIfDisposed(),Bl(this,e)},Xa().prototype.expm1=function(){return this.throwIfDisposed(),zl(this)},Xa().prototype.fft=function(){return this.throwIfDisposed(),Ph(this)},Xa().prototype.flatten=function(){return this.throwIfDisposed(),qu(this,[this.size])},Xa().prototype.floor=function(){return this.throwIfDisposed(),jl(this)},Xa().prototype.floorDiv=function(e){return this.throwIfDisposed(),yu(this,e)},Xa().prototype.gather=function(e,t){return this.throwIfDisposed(),Gl(this,e,t)},Xa().prototype.greaterEqual=function(e){return this.throwIfDisposed(),ql(this,e)},Xa().prototype.greater=function(e){return this.throwIfDisposed(),Hl(this,e)},Xa().prototype.ifft=function(){return this.throwIfDisposed(),$h(this)},Xa().prototype.irfft=function(){return this.throwIfDisposed(),Bh(this)},Xa().prototype.isFinite=function(){return this.throwIfDisposed(),Yl(this)},Xa().prototype.isInf=function(){return this.throwIfDisposed(),Xl(this)},Xa().prototype.isNaN=function(){return this.throwIfDisposed(),Ql(this)},Xa().prototype.leakyRelu=function(e){return this.throwIfDisposed(),Zl(this,e)},Xa().prototype.lessEqual=function(e){return this.throwIfDisposed(),ec(this,e)},Xa().prototype.less=function(e){return this.throwIfDisposed(),Jl(this,e)},Xa().prototype.localResponseNormalization=function(e,t,n,r){return this.throwIfDisposed(),nc(this,e,t,n,r)},Xa().prototype.logSigmoid=function(){return this.throwIfDisposed(),fc(this)},Xa().prototype.logSoftmax=function(e){return this.throwIfDisposed(),bc(this,e)},Xa().prototype.logSumExp=function(e,t){return this.throwIfDisposed(),Nc(this,e,t)},Xa().prototype.log=function(){return this.throwIfDisposed(),rc(this)},Xa().prototype.log1p=function(){return this.throwIfDisposed(),ac(this)},Xa().prototype.logicalAnd=function(e){return this.throwIfDisposed(),Cc(this,e)},Xa().prototype.logicalNot=function(){return this.throwIfDisposed(),Tc(this)},Xa().prototype.logicalOr=function(e){return this.throwIfDisposed(),Rc(this,e)},Xa().prototype.logicalXor=function(e){return this.throwIfDisposed(),_c(this,e)},Xa().prototype.matMul=function(e,t,n){return this.throwIfDisposed(),Gi(this,e,t,n)},Xa().prototype.maxPool=function(e,t,n,r){return this.throwIfDisposed(),Fc(this,e,t,n,r)},Xa().prototype.max=function(e,t){return this.throwIfDisposed(),mc(this,e,t)},Xa().prototype.maximum=function(e){return this.throwIfDisposed(),Oc(this,e)},Xa().prototype.mean=function(e,t){return this.throwIfDisposed(),Lc(this,e,t)},Xa().prototype.min=function(e,t){return this.throwIfDisposed(),zc(this,e,t)},Xa().prototype.minimum=function(e){return this.throwIfDisposed(),Uc(this,e)},Xa().prototype.mirrorPad=function(e,t){return this.throwIfDisposed(),Wc(this,e,t)},Xa().prototype.mod=function(e){return this.throwIfDisposed(),Vc(this,e)},Xa().prototype.mul=function(e){return this.throwIfDisposed(),vu(this,e)},Xa().prototype.neg=function(){return this.throwIfDisposed(),pc(this)},Xa().prototype.norm=function(e,t,n){return this.throwIfDisposed(),lp(this,e,t,n)},Xa().prototype.notEqual=function(e){return this.throwIfDisposed(),Kc(this,e)},Xa().prototype.oneHot=function(e,t=1,n=0){return this.throwIfDisposed(),Hi(this,e,t,n)},Xa().prototype.onesLike=function(){return this.throwIfDisposed(),Yc(this)},Xa().prototype.pad=function(e,t){return this.throwIfDisposed(),Qc(this,e,t)},Xa().prototype.pool=function(e,t,n,r,a,s){return this.throwIfDisposed(),rh(this,e,t,n,r,a,s)},Xa().prototype.pow=function(e){return this.throwIfDisposed(),ah(this,e)},Xa().prototype.prelu=function(e){return this.throwIfDisposed(),sh(this,e)},Xa().prototype.prod=function(e,t){return this.throwIfDisposed(),ih(this,e,t)},Xa().prototype.reciprocal=function(){return this.throwIfDisposed(),mh(this)},Xa().prototype.relu=function(){return this.throwIfDisposed(),gh(this)},Xa().prototype.relu6=function(){return this.throwIfDisposed(),yh(this)},Xa().prototype.reshapeAs=function(e){return this.throwIfDisposed(),qu(this,e.shape)},Xa().prototype.reshape=function(e){return this.throwIfDisposed(),qu(this,e)},Xa().prototype.resizeBilinear=function(e,t,n){return this.throwIfDisposed(),Jp(this,e,t,n)},Xa().prototype.resizeNearestNeighbor=function(e,t,n){return this.throwIfDisposed(),ed(this,e,t,n)},Xa().prototype.reverse=function(e){return this.throwIfDisposed(),bh(this,e)},Xa().prototype.rfft=function(){return this.throwIfDisposed(),Uh(this)},Xa().prototype.round=function(){return this.throwIfDisposed(),Sh(this)},Xa().prototype.rsqrt=function(){return this.throwIfDisposed(),Ih(this)},Xa().prototype.selu=function(){return this.throwIfDisposed(),Eh(this)},Xa().prototype.separableConv2d=function(e,t,n,r,a,s){return this.throwIfDisposed(),Nh(this,e,t,n,r,a,s)},Xa().prototype.sigmoid=function(){return this.throwIfDisposed(),Qu(this)},Xa().prototype.sign=function(){return this.throwIfDisposed(),Th(this)},Xa().prototype.sin=function(){return this.throwIfDisposed(),Rh(this)},Xa().prototype.sinh=function(){return this.throwIfDisposed(),_h(this)},Xa().prototype.slice=function(e,t){return this.throwIfDisposed(),Zu(this,e,t)},Xa().prototype.softmax=function(e){return this.throwIfDisposed(),Lh(this,e)},Xa().prototype.softplus=function(){return this.throwIfDisposed(),dc(this)},Xa().prototype.spaceToBatchND=function(e,t){return this.throwIfDisposed(),nh(this,e,t)},Xa().prototype.split=function(e,t){return this.throwIfDisposed(),zh(this,e,t)},Xa().prototype.sqrt=function(){return this.throwIfDisposed(),Wh(this)},Xa().prototype.square=function(){return this.throwIfDisposed(),jc(this)},Xa().prototype.squaredDifference=function(e){return this.throwIfDisposed(),Vh(this,e)},Xa().prototype.squeeze=function(e){return this.throwIfDisposed(),jh(this,e)},Xa().prototype.stack=function(e,t){this.throwIfDisposed();const n=e instanceof Ya?[this,e]:[this,...e];return Gh(n,t)},Xa().prototype.step=function(e){return this.throwIfDisposed(),Hh(this,e)},Xa().prototype.stridedSlice=function(e,t,n,r,a,s,i,o){return this.throwIfDisposed(),qh(this,e,t,n,r,a,s,i,o)},Xa().prototype.sub=function(e){return this.throwIfDisposed(),gc(this,e)},Xa().prototype.sum=function(e,t){return this.throwIfDisposed(),yc(this,e,t)},Xa().prototype.tan=function(){return this.throwIfDisposed(),Kh(this)},Xa().prototype.tanh=function(){return this.throwIfDisposed(),Ju(this)},Xa().prototype.tile=function(e){return this.throwIfDisposed(),Ul(this,e)},Xa().prototype.toBool=function(){return this.throwIfDisposed(),Ai(this,"bool")},Xa().prototype.toFloat=function(){return this.throwIfDisposed(),Ai(this,"float32")},Xa().prototype.toInt=function(){return this.throwIfDisposed(),Ai(this,"int32")},Xa().prototype.topk=function(e,t){return this.throwIfDisposed(),ep(this,e,t)},Xa().prototype.transpose=function(e){return this.throwIfDisposed(),qi(this,e)},Xa().prototype.unique=function(e){return this.throwIfDisposed(),np(this,e)},Xa().prototype.unsortedSegmentSum=function(e,t){return this.throwIfDisposed(),rp(this,e,t)},Xa().prototype.unstack=function(e){return this.throwIfDisposed(),ap(this,e)},Xa().prototype.where=function(e,t){return this.throwIfDisposed(),_l(e,this,t)},Xa().prototype.zerosLike=function(){return this.throwIfDisposed(),Fl(this)};var ng,rg={};function ag(){return null==ng&&(ng=fu().epsilon()),ng}w(rg,{maxNorm:()=>Pg,minMaxNorm:()=>zg,nonNeg:()=>Bg,unitNorm:()=>$g});var sg=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,sg.prototype)}},ig=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,ig.prototype)}},og=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,og.prototype)}},ug=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,ug.prototype)}},lg=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,lg.prototype)}};function cg(e,t){if(Array.isArray(e)){let n=[];for(let r=0;r<t;r++)n=n.concat(e);return n}{const n=new Array(t);return n.fill(e),n}}function hg(e,t){if(!e)throw new lg(t)}function pg(e,t){let n=0;for(const r of e)r===t&&n++;return n}function dg(e){return 1===e.length?e[0]:e}function fg(e){return Array.isArray(e)?e:[e]}function mg(e){const t=e.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return"_"!==t[0]?t:"private"+t}function gg(e){return e.length<=1||-1===e.indexOf("_")?e:e.replace(/[_]+(\w|$)/g,(e,t)=>t.toUpperCase())}var yg={};function bg(e){if(null==e)return null;const t={};return t.className=e.getClassName(),t.config=e.getConfig(),t}function vg(e,t={},n={},r="object",a=!1){if("string"==typeof e){const a=e;let s;if(a in n)s=n[a];else if(a in yg)s=yg[a];else if(s=t[a],null==s)throw new og(`Unknown ${r}: ${e}. This may be due to one of the following reasons:\n1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return s}{const s=e;if(null==s.className||null==s.config)throw new og(`${r}: Improper config format: ${JSON.stringify(s)}.\n'className' and 'config' must set.`);const i=s.className;let o,u;if(i in n?[o,u]=n[i]:i in yg?[o,u]=yg.className:i in t&&([o,u]=t[i]),null==o)throw new og(`Unknown ${r}: ${i}. This may be due to one of the following reasons:\n1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(null!=u){const e={};for(const t of Object.keys(yg))e[t]=yg[t];for(const t of Object.keys(n))e[t]=n[t];s.config.customObjects=e;const t=Object.assign({},yg);for(const e of Object.keys(n))yg[e]=n[e];!function e(t){if(null!=t&&"object"==typeof t)if(Array.isArray(t))t.forEach(t=>e(t));else{const n=Object.keys(t);for(const r of n){const n=t[r];null!=n&&"object"==typeof n&&(Array.isArray(n)||"ndarray"!==n.type||"number"!=typeof n.value?e(n):t[r]=n.value)}}}(s.config);const r=u(o,s.config,n,a);return yg=Object.assign({},t),r}{const e=Object.assign({},yg);for(const e of Object.keys(n))yg[e]=n[e];const t=new o(s.config);return yg=Object.assign({},e),t}}}function xg(e,t){return-1*function(e,t){return e<t?-1:e>t?1:0}(e,t)}function wg(e){if(null==e)return e;const t=[];for(const n of e)-1===t.indexOf(n)&&t.push(n);return t}function kg(e){if(null==e)throw new og("Invalid value in obj: "+JSON.stringify(e));for(const t in e)if(e.hasOwnProperty(t))return!1;return!0}function Sg(e,t,n){if(null!=n&&e.indexOf(n)<0)throw new og(`${n} is not a valid ${t}.  Valid values are ${e} or null/undefined.`)}function Ig(e,t,n=0,r=1/0){return hg(n>=0),hg(r>=n),Array.isArray(e)&&e.length>=n&&e.length<=r&&e.every(e=>typeof e===t)}function Ag(e,t){Array.isArray(e)?(va.assert(e.length>0,()=>t+" is unexpectedly an empty array."),e.forEach((e,n)=>Ag(e,`element ${n+1} of ${t}`))):va.assert(Number.isInteger(e)&&e>0,()=>`Expected ${t} to be a positive integer, but got ${function e(t){return null===t?"null":Array.isArray(t)?"["+t.map(t=>e(t)).join(",")+"]":"string"==typeof t?`"${t}"`:""+t}(e)}.`)}function Eg(e){return"relu"===e?"relu":"linear"===e?"linear":"elu"===e?"elu":null}function Ng(e,t){return ru(()=>Wh(yc(vu(e,e),t,!0)))}var Cg=class extends Fo.Serializable{getConfig(){return{}}},Tg=class extends Cg{constructor(e){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=null!=e.maxValue?e.maxValue:this.defaultMaxValue,this.axis=null!=e.axis?e.axis:this.defaultAxis}apply(e){return ru(()=>{const t=Ng(e,this.axis),n=cl(t,0,this.maxValue);return vu(e,bu(n,gu(ag(),t)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};Tg.className="MaxNorm",Fo.registerClass(Tg);var Rg=class extends Cg{constructor(e){super(),this.defaultAxis=0,this.axis=null!=e.axis?e.axis:this.defaultAxis}apply(e){return ru(()=>bu(e,gu(ag(),Ng(e,this.axis))))}getConfig(){return{axis:this.axis}}};Rg.className="UnitNorm",Fo.registerClass(Rg);var _g=class extends Cg{apply(e){return gh(e)}};_g.className="NonNeg",Fo.registerClass(_g);var Fg=class extends Cg{constructor(e){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=null!=e.minValue?e.minValue:this.defaultMinValue,this.maxValue=null!=e.maxValue?e.maxValue:this.defaultMaxValue,this.rate=null!=e.rate?e.rate:this.defaultRate,this.axis=null!=e.axis?e.axis:this.defaultAxis}apply(e){return ru(()=>{const t=Ng(e,this.axis),n=gu(vu(this.rate,cl(t,this.minValue,this.maxValue)),vu(1-this.rate,t));return vu(e,bu(n,gu(ag(),t)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};Fg.className="MinMaxNorm",Fo.registerClass(Fg);var Mg={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function Dg(e){return bg(e)}function Og(e,t={}){return vg(e,Fo.SerializationMap.getMap().classNameMap,t,"constraint")}function Lg(e){if(null==e)return null;if("string"==typeof e){return Og({className:e in Mg?Mg[e]:e,config:{}})}return e instanceof Cg?e:Og(e)}function Pg(e){return new Tg(e)}function $g(e){return new Rg(e)}function Bg(){return new _g}function zg(e){return new Fg(e)}var Ug={};w(Ug,{constant:()=>qy,glorotNormal:()=>eb,glorotUniform:()=>Jy,heNormal:()=>tb,heUniform:()=>nb,identity:()=>Qy,leCunNormal:()=>rb,leCunUniform:()=>ab,ones:()=>Hy,orthogonal:()=>sb,randomNormal:()=>Yy,randomUniform:()=>Ky,truncatedNormal:()=>Xy,varianceScaling:()=>Zy,zeros:()=>Gy});var Wg=["channelsFirst","channelsLast"],Vg=["nearest","bilinear"],jg=["valid","same","causal"],Gg=["max","avg"],Hg=["sum","mul","concat","ave"],qg=new Map;function Kg(e){Sg(Wg,"DataFormat",e)}function Yg(e){Sg(jg,"PaddingMode",e)}function Xg(e){Sg(Gg,"PoolMode",e)}var Qg=[];function Zg(e,t){Qg.push(e);try{const e=t();return Qg.pop(),e}catch(e){throw Qg.pop(),e}}function Jg(e){if(!ny(e))throw new Error("Not a valid tensor name: '"+e+"'");return(0===Qg.length?"":Qg.join("/")+"/")+e}function ey(e){if(!ny(e))throw new Error("Not a valid tensor name: '"+e+"'");qg.has(e)||qg.set(e,0);const t=qg.get(e);if(qg.set(e,qg.get(e)+1),t>0){const n=`${e}_${t}`;return qg.set(n,1),n}return e}var ty=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function ny(e){return!!e.match(ty)}function ry(e,t,n){null==t&&(t=0),null==n&&(n=e.length);let r=1;for(let a=t;a<n;++a)r*=e[a];return r}function ay(e){if(0===e.length)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let n=0;n<e.length;n++){const r=e[n];r<t&&(t=r)}return t}function sy(e){if(0===e.length)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let n=0;n<e.length;n++){const r=e[n];r>t&&(t=r)}return t}function iy(e,t){if(t<e)throw new og(`end (${t}) < begin (${e}) is forbidden.`);const n=[];for(let r=e;r<t;++r)n.push(r);return n}function oy(e,t){return Ai(e,t)}function uy(e,t=-1){const n=e.shape.slice();return t<0&&(t=n.length+t+1),n.splice(t,0,1),qu(e,n)}function ly(e,t,n){return ru(()=>{switch(e.rank){case 1:return Fh(e,t,n);case 2:return Mh(e,[t,0],[n,e.shape[1]]);case 3:return Dh(e,[t,0,0],[n,e.shape[1],e.shape[2]]);case 4:return Oh(e,[t,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3]]);case 5:return Zu(e,[t,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4]]);case 6:return Zu(e,[t,0,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4],e.shape[5]]);default:throw new og("sliceAlongFirstAxis() received an unsupported tensor rank: "+e.rank)}})}function cy(e,t,n){return ru(()=>{switch(e.rank){case 1:return Fh(e,t,n);case 2:return Mh(e,[0,t],[e.shape[0],n]);case 3:return Dh(e,[0,0,t],[e.shape[0],e.shape[1],n]);case 4:return Oh(e,[0,0,0,t],[e.shape[0],e.shape[1],e.shape[2],n]);default:throw new og("sliceAlongLastAxis() received an unsupported tensor rank: "+e.rank)}})}function hy(e,t,n,r){return ru(()=>{switch(e.rank){case 1:return Fh(e,t,n);case 2:switch(r){case 1:return ly(e,t,n);case 2:return cy(e,t,n);default:throw new og("The axis is not within the rank of the tensor "+r)}case 3:switch(r){case 1:return ly(e,t,n);case 2:return Dh(e,[0,t,0],[e.shape[0],n,e.shape[2]]);case 3:return cy(e,t,n);default:throw new og("The axis is not within the rank of the tensor "+r)}case 4:switch(r){case 1:return ly(e,t,n);case 2:return Oh(e,[0,t,0,0],[e.shape[0],n,e.shape[2],e.shape[3]]);case 3:return Oh(e,[0,0,t,0],[e.shape[0],e.shape[1],n,e.shape[3]]);case 4:return cy(e,t,n);default:throw new og("The axis is not within the rank of the tensor "+r)}default:throw new og("sliceAlongLastAxis() received an unsupported tensor rank: "+e.rank)}})}function py(e,t=-1){let n;return t<0&&(n=e[0].rank,t=0!==n?n:0),t===e[0].rank&&(t=-1),Xu(e,t)}function dy(e,t){switch(e.rank){case 1:return hl([e,t]);case 2:return pl([e,t],0);case 3:return dl([e,t],0);case 4:return fl([e,t],0);default:throw new og("concatAlongFirstAxis() received an unsupported tensor rank: "+e.rank)}}function fy(e,t){if(Array.isArray(t)||(t=[t]),e.rank!==t.length)throw new og(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);return Ul(e,t)}function my(e,t=0,n=1,r,a){return hh(e,t,n,r,a)}function gy(e,t,n,r){if(e.rank<2||t.rank<2)throw new ug(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);if(t.rank>=3){if(e.shape.slice(-1)[0]!==t.shape.slice(-2)[0])throw new ug(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and  y shape = ${t.shape}`)}if(2===e.rank&&2===t.rank){const a=!1,s=!1;return bp.matMul({a:e,b:t,transposeA:a,transposeB:s,bias:r?vy(e.rank,r,"channelsLast"):null,activation:n})}{const a=e.shape.slice(),s=a.pop();e=qu(e,[-1,s]);const i=t.shape.slice(),o=i.pop(),u=i.pop(),l=[...i,o],c=Array.from({length:t.rank},(e,n)=>0===n?t.rank-2:n<=t.rank-2?n-1:n);t=qu(qi(t,c),[u,-1]);const h=[...a,...l],p=!1,d=!1;return qu(bp.matMul({a:e,b:t,transposeA:p,transposeB:d,bias:r?vy(e.rank,r,"channelsLast"):null,activation:n}),h)}}function yy(e,t,n){return ru(()=>(t=Array.isArray(t)?Yh(t,"int32"):Ai(t,"int32"),Gl(e,t,n)))}function by(e){return vu(e,e)}function vy(e,t,n){const r=t.shape;if(1!==t.rank&&t.rank!==e)throw new og(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);if(5===e){if("channelsFirst"===n)return 1===r.length?qu(t,[1,r[0],1,1,1]):qu(t,[1,r[3],r[0],r[1],r[2]]);if("channelsLast"===n)return 1===r.length?qu(t,[1,1,1,1,r[0]]):qu(t,[1].concat(r))}else if(4===e){if("channelsFirst"===n)return 1===r.length?qu(t,[1,r[0],1,1]):qu(t,[1,r[2],r[0],r[1]]);if("channelsLast"===n)return 1===r.length?qu(t,[1,1,1,r[0]]):qu(t,[1].concat(r))}else if(3===e){if("channelsFirst"===n)return 1===r.length?qu(t,[1,r[0],1]):qu(t,[1,r[1],r[0]]);if("channelsLast"===n)return 1===r.length?qu(t,[1,1,r[0]]):qu(t,[1].concat(r))}else if(e<3)return t;throw new og("Unsupported input rank by biasAdd: "+t.rank)}function xy(e,t,n){return ru(()=>(null==n&&(n="channelsLast"),Kg(n),gu(e,vy(e.rank,t,n))))}function wy(e,t,n,r){return ru(()=>fp(e,t,n,r))}function ky(e,t,n=!1){return n?e():t()}var Sy=["fanIn","fanOut","fanAvg"],Iy=["normal","uniform","truncatedNormal"];var Ay=class extends Fo.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},Ey=class extends Ay{apply(e,t){return Pc(e,t)}};Ey.className="Zeros",Fo.registerClass(Ey);var Ny=class extends Ay{apply(e,t){return $c(e,t)}};Ny.className="Ones",Fo.registerClass(Ny);var Cy=class extends Ay{constructor(e){if(super(),"object"!=typeof e)throw new og("Expected argument of type ConstantConfig but got "+e);if(void 0===e.value)throw new og("config must have value set but got "+e);this.value=e.value}apply(e,t){return ru(()=>vu(Ah(this.value),$c(e,t)))}getConfig(){return{value:this.value}}};Cy.className="Constant",Fo.registerClass(Cy);var Ty=class extends Ay{constructor(e){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=e.minval||this.DEFAULT_MINVAL,this.maxval=e.maxval||this.DEFAULT_MAXVAL,this.seed=e.seed}apply(e,t){return ph(e,this.minval,this.maxval,t)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};Ty.className="RandomUniform",Fo.registerClass(Ty);var Ry=class extends Ay{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if("float32"!==(t=t||"float32")&&"int32"!==t)throw new ug(`randomNormal does not support dType ${t}.`);return my(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};Ry.className="RandomNormal",Fo.registerClass(Ry);var _y=class extends Ay{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if("float32"!==(t=t||"float32")&&"int32"!==t)throw new ug(`truncatedNormal does not support dType ${t}.`);return tp(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};_y.className="TruncatedNormal",Fo.registerClass(_y);var Fy=class extends Ay{constructor(e){super(),this.gain=null!=e.gain?e.gain:1}apply(e,t){return ru(()=>{if(2!==e.length||e[0]!==e[1])throw new og("Identity matrix initializer can only be used for 2D square matrices.");return vu(this.gain,Wl(e[0]))})}getConfig(){return{gain:this.gain}}};Fy.className="Identity",Fo.registerClass(Fy);var My=class extends Ay{constructor(e){if(super(),e.scale<0)throw new og("scale must be a positive float. Got: "+e.scale);var t;this.scale=null==e.scale?1:e.scale,this.mode=null==e.mode?"fanIn":e.mode,t=this.mode,Sg(Sy,"FanMode",t),this.distribution=null==e.distribution?"normal":e.distribution,function(e){Sg(Iy,"Distribution",e)}(this.distribution),this.seed=e.seed}apply(e,t){const n=function(e,t="channelsLast"){let n,r;if(Kg(t),2===e.length)n=e[0],r=e[1];else if(-1!==[3,4,5].indexOf(e.length)){if("channelsFirst"===t){const t=ry(e,2);n=e[1]*t,r=e[0]*t}else if("channelsLast"===t){const t=ry(e,0,e.length-2);n=e[e.length-2]*t,r=e[e.length-1]*t}}else{const t=ry(e);n=Math.sqrt(t),r=Math.sqrt(t)}return[n,r]}(e),r=n[0],a=n[1];let s=this.scale;if("fanIn"===this.mode?s/=Math.max(1,r):"fanOut"===this.mode?s/=Math.max(1,a):s/=Math.max(1,(r+a)/2),"normal"===this.distribution){const n=Math.sqrt(s);if("float32"!==(t=t||"float32")&&"int32"!==t)throw new ug(`${this.getClassName()} does not support dType ${t}.`);return tp(e,0,n,t,this.seed)}{const n=Math.sqrt(3*s);return ph(e,-n,n,t)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};My.className="VarianceScaling",Fo.registerClass(My);var Dy=class extends My{constructor(e){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:null==e?null:e.seed})}getClassName(){return My.className}};Dy.className="GlorotUniform",Fo.registerClass(Dy);var Oy=class extends My{constructor(e){super({scale:1,mode:"fanAvg",distribution:"normal",seed:null==e?null:e.seed})}getClassName(){return My.className}};Oy.className="GlorotNormal",Fo.registerClass(Oy);var Ly=class extends My{constructor(e){super({scale:2,mode:"fanIn",distribution:"normal",seed:null==e?null:e.seed})}getClassName(){return My.className}};Ly.className="HeNormal",Fo.registerClass(Ly);var Py=class extends My{constructor(e){super({scale:2,mode:"fanIn",distribution:"uniform",seed:null==e?null:e.seed})}getClassName(){return My.className}};Py.className="HeUniform",Fo.registerClass(Py);var $y=class extends My{constructor(e){super({scale:1,mode:"fanIn",distribution:"normal",seed:null==e?null:e.seed})}getClassName(){return My.className}};$y.className="LeCunNormal",Fo.registerClass($y);var By=class extends My{constructor(e){super({scale:1,mode:"fanIn",distribution:"uniform",seed:null==e?null:e.seed})}getClassName(){return My.className}};By.className="LeCunNormal",Fo.registerClass(By);var zy=class extends Ay{constructor(e){if(super(),this.DEFAULT_GAIN=1,this.gain=null==e.gain?this.DEFAULT_GAIN:e.gain,this.seed=e.seed,null!=this.seed)throw new ug("Random seed is not implemented for Orthogonal Initializer yet.")}apply(e,t){return ru(()=>{if(e.length<2)throw new ug("Shape must be at least 2D.");e[0]*e[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${e[0]*e[1]}) elements: Slowness may result.`);const t=my(e[0]>e[1]?[e[1],e[0]]:e,0,1,"float32");let n=Cd.gramSchmidt(t);return e[0]>e[1]&&(n=qi(n)),vu(this.gain,n)})}getConfig(){return{gain:this.gain,seed:this.seed}}};zy.className="Orthogonal",Fo.registerClass(zy);var Uy={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function Wy(e,t={}){return vg(e,Fo.SerializationMap.getMap().classNameMap,t,"initializer")}function Vy(e){return bg(e)}function jy(e){if("string"==typeof e){const t=e in Uy?Uy[e]:e;if("GlorotNormal"===t)return new Oy;if("GlorotUniform"===t)return new Dy;if("HeNormal"===t)return new Ly;if("HeUniform"===t)return new Py;if("LeCunNormal"===t)return new $y;if("LeCunUniform"===t)return new By;{const e={};return e.className=t,e.config={},Wy(e)}}return e instanceof Ay?e:Wy(e)}function Gy(){return new Ey}function Hy(){return new Ny}function qy(e){return new Cy(e)}function Ky(e){return new Ty(e)}function Yy(e){return new Ry(e)}function Xy(e){return new _y(e)}function Qy(e){return new Fy(e)}function Zy(e){return new My(e)}function Jy(e){return new Dy(e)}function eb(e){return new Oy(e)}function tb(e){return new Ly(e)}function nb(e){return new Py(e)}function rb(e){return new $y(e)}function ab(e){return new By(e)}function sb(e){return new zy(e)}var ib={};w(ib,{Layer:()=>Ib,RNN:()=>Yx,RNNCell:()=>Xx,activation:()=>yk,add:()=>Ek,alphaDropout:()=>gS,average:()=>Nk,averagePooling1d:()=>Lk,averagePooling2d:()=>Bk,averagePooling3d:()=>Wk,avgPool1d:()=>Pk,avgPool2d:()=>zk,avgPool3d:()=>Vk,avgPooling1d:()=>$k,avgPooling2d:()=>Uk,avgPooling3d:()=>jk,batchNormalization:()=>Mk,bidirectional:()=>uS,concatenate:()=>Ck,conv1d:()=>uk,conv2d:()=>lk,conv2dTranspose:()=>ck,conv3d:()=>hk,conv3dTranspose:()=>pk,convLstm2d:()=>aS,convLstm2dCell:()=>sS,cropping2D:()=>fk,dense:()=>bk,depthwiseConv2d:()=>gk,dot:()=>Fk,dropout:()=>vk,elu:()=>nk,embedding:()=>Ak,flatten:()=>wk,gaussianDropout:()=>mS,gaussianNoise:()=>fS,globalAveragePooling1d:()=>Gk,globalAveragePooling2d:()=>Hk,globalMaxPool1d:()=>cS,globalMaxPool2d:()=>hS,globalMaxPooling1d:()=>qk,globalMaxPooling2d:()=>Kk,gru:()=>Zk,gruCell:()=>Jk,input:()=>Xv,inputLayer:()=>tk,layerNormalization:()=>Dk,leakyReLU:()=>ak,lstm:()=>eS,lstmCell:()=>tS,masking:()=>yS,maxPool1d:()=>pS,maxPool2d:()=>dS,maxPooling1d:()=>Yk,maxPooling2d:()=>Xk,maxPooling3d:()=>Qk,maximum:()=>Tk,minimum:()=>Rk,multiply:()=>_k,permute:()=>Ik,prelu:()=>sk,reLU:()=>rk,repeatVector:()=>kk,reshape:()=>Sk,rnn:()=>iS,separableConv2d:()=>dk,simpleRNN:()=>nS,simpleRNNCell:()=>rS,softmax:()=>ik,spatialDropout1d:()=>xk,stackedRNNCells:()=>oS,thresholdedReLU:()=>ok,timeDistributed:()=>lS,upSampling2d:()=>mk,zeroPadding2d:()=>Ok});var ob=0;function ub(){return ob++}var lb={};function cb(e=""){return e in lb||(lb[e]=0),lb[e]+=1,e+lb[e].toString()}function hb(e){return Array.isArray(e)&&Array.isArray(e[0])}function pb(e){return 0===e.length?[]:Array.isArray(e[0])?e:[e]}function db(e){let t;if(Array.isArray(e)){if(1!==e.length)throw new og("Expected Tensor length to be 1; got "+e.length);t=e[0]}else t=e;return t}function fb(e){if(Array.isArray(e)&&Array.isArray(e[0])){if(1===e.length)return(e=e)[0];throw new og("Expected exactly 1 Shape; got "+e.length)}return e}function mb(e){let t=0;for(const n of e)0===n.shape.length?t+=1:t+=n.shape.reduce((e,t)=>e*t);return t}var gb=class{constructor(e,t="float32",n="Variable",r=!0,a=null){this.dtype=null==t?"float32":t,this.shape=e.shape,this.id=ub(),n=null==n?"Variable":n,this.originalName=Jg(n),this.name=ey(this.originalName),this.trainable_=r,this.constraint=a,this.val=sp(e,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(e){return this.assertNotDisposed(),function(e,t){if(e.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(e.shape)+" vs. "+JSON.stringify(t.shape))}(this.val,e),this.val.id!==e.id&&(this.val.assign(e),null!=this.constraint&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(e){this.trainable_=e,this.val.trainable=e}};function yb(e){return e.map(e=>e.read())}function bb(e){e.forEach(e=>{e[0].write(e[1])})}var vb=class{constructor(e){this.dtype=e.dtype,this.shape=e.shape,null!=e.shape?this.ndim=e.shape.length:this.ndim=e.ndim,this.maxNDim=e.maxNDim,this.minNDim=e.minNDim,this.axes=e.axes||{}}},xb=class{constructor(e,t,n,r,a,s,i){this.dtype=e,this.shape=t,this.sourceLayer=n,this.inputs=r,this.callArgs=a,this.outputTensorIndex=i,this.id=ub(),null!=s&&(this.originalName=Jg(s),this.name=ey(this.originalName)),this.rank=t.length}},wb=0,kb=class{constructor(e,t){this.callArgs=t,this.id=wb++,this.outboundLayer=e.outboundLayer,this.inboundLayers=e.inboundLayers,this.nodeIndices=e.nodeIndices,this.tensorIndices=e.tensorIndices,this.inputTensors=e.inputTensors,this.outputTensors=e.outputTensors,this.inputMasks=e.inputMasks,this.outputMasks=e.outputMasks,this.inputShapes=e.inputShapes,this.outputShapes=e.outputShapes;for(const t of e.inboundLayers)null!=t&&t.outboundNodes.push(this);e.outboundLayer.inboundNodes.push(this)}getConfig(){const e=[];for(const t of this.inboundLayers)null!=t?e.push(t.name):e.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:e,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},Sb=0,Ib=class extends Fo.Serializable{constructor(e={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=Sb++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let t=e.name;if(!t){const e=this.getClassName();t=mg(e)+"_"+cb(e)}if(this.name=t,this.trainable_=null==e.trainable||e.trainable,null!=e.inputShape||null!=e.batchInputShape){let t;if(null!=e.batchInputShape)t=e.batchInputShape;else if(null!=e.inputShape){let n=null;null!=e.batchSize&&(n=e.batchSize),t=[n].concat(e.inputShape)}this.batchInputShape=t;let n=e.dtype;null==n&&(n=e.inputDType),null==n&&(n="float32"),this.dtype=n}null!=e.weights?this.initialWeights=e.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(e,t){return e.name+"_ib-"+t.toString()}getNodeAtIndex(e,t){if(0===this.inboundNodes.length)throw new ig(`The layer has never been called and thus has no defined ${t}.`);if(this.inboundNodes.length<=e)throw new og(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[e]}getInputAt(e){return dg(this.getNodeAtIndex(e,"input").inputTensors)}getOutputAt(e){return dg(this.getNodeAtIndex(e,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new sg(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(0===this.inboundNodes.length)throw new sg(`Layer ${this.name} is not connected, no input to return.`);return dg(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(0===this.inboundNodes.length)throw new sg(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new sg(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return dg(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(e=>e())}get updates(){return this._updates}get built(){return this._built}set built(e){this._built=e}get trainable(){return this.trainable_}set trainable(e){this._trainableWeights.forEach(t=>t.trainable=e),this.trainable_=e}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(e=>e.trainable):[]}set trainableWeights(e){this._trainableWeights=e}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(e=>!e.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(e){this._nonTrainableWeights=e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(e){if(e=fg(e),null==this.inputSpec||0===this.inputSpec.length)return;const t=fg(this.inputSpec);if(e.length!==t.length)throw new og(`Layer ${this.name} expects ${t.length} inputs, but it received ${e.length} input tensors. Input received: ${e}`);for(let n=0;n<e.length;n++){const r=e[n],a=t[n];if(null==a)continue;const s=r.rank;if(null!=a.ndim&&s!==a.ndim)throw new og(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${a.ndim}, found ndim=${s}`);if(null!=a.maxNDim&&s>a.maxNDim)throw new og(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${a.maxNDim}, found ndim=${s}`);if(null!=a.minNDim&&s<a.minNDim)throw new og(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${a.minNDim}, found ndim=${s}.`);if(null!=a.dtype&&r.dtype!==a.dtype)throw new og(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${a.dtype}, found dtype=${r.dtype}.`);if(a.axes){const e=r.shape;for(const t in a.axes){const r=Number(t),s=a.axes[t],i=r>=0?e[r]:e[e.length+r];if(null!=s&&-1===[s,null].indexOf(i))throw new og(`Input ${n} is incompatible with layer ${this.name}: expected axis ${r} of input shape to have value ${s} but got shape ${e}.`)}}if(null!=a.shape)for(let e=0;e<a.shape.length;++e){const t=a.shape[e],s=r.shape[e];if(null!=t&&null!=s&&t!==s)throw new og(`Input ${n} is incompatible with layer ${this.name}: expected shape=${a.shape}, found shape=${r.shape}.`)}}}call(e,t){return e}invokeCallHook(e,t){null!=this._callHook&&this._callHook(e,t)}setCallHook(e){this._callHook=e}clearCallHook(){this._callHook=null}apply(e,t){t=t||{},this.assertNotDisposed();const n=fg(e);let r=!0;for(const e of n)if(!(e instanceof xb)){r=!1;break}let a=!0;for(const e of n)if(e instanceof xb){a=!1;break}if(r===a)throw new og("Arguments to apply() must be all SymbolicTensors or all Tensors");return Zg(this.name,()=>{if(!this.built){this.assertInputCompatibility(e);const t=[];for(const n of fg(e))t.push(n.shape);this.build(dg(t)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),null===this._refCount&&a&&(this._refCount=1)}if(this.assertInputCompatibility(e),a){let r=this.call(e,t);const a=fg(r),s=[];for(let e of a)-1!==n.indexOf(e)&&(e=e.clone()),s.push(e);if(r=dg(s),null!=this.activityRegularizer)throw new ug("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return r}{const n=function(e){e=fg(e);const t=[];for(const n of e)t.push(n.shape);return dg(t)}(e),r=this.computeOutputShape(n);let a;const s="float32";if(this.warnOnIncompatibleInputShape(Array.isArray(e)?n[0]:n),a=null!=r&&r.length>0&&Array.isArray(r[0])?r.map((n,r)=>new xb(s,n,this,fg(e),t,this.name,r)):new xb(s,r,this,fg(e),t,this.name),this.addInboundNode(e,a,null,null,n,r,t),this._refCount++,null!=this.activityRegularizer)throw new ug("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return a}})}warnOnIncompatibleInputShape(e){if(null!=this.batchInputShape)if(e.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let t=!1;this.batchInputShape.forEach((n,r)=>{null!=n&&null!=e[r]&&e[r]!==n&&(t=!0)}),t&&console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(null==this.inboundNodes||0===this.inboundNodes.length)throw new sg(`The layer ${this.name} has never been called and thus has no defined output shape.`);const e=[];for(const t of this.inboundNodes){const n=JSON.stringify(t.outputShapes);-1===e.indexOf(n)&&e.push(n)}if(1===e.length){const e=this.inboundNodes[0].outputShapes;return Array.isArray(e)&&Array.isArray(e[0])&&1===e.length?e[0]:e}throw new sg(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new ig(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return mb(this.weights)}build(e){this.built=!0}getWeights(e=!1){return yb(e?this.trainableWeights:this.weights)}setWeights(e){ru(()=>{const t=this.weights;if(t.length!==e.length)throw new og(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);if(0===t.length)return;const n=[],r=yb(t);for(let a=0;a<r.length;++a){const s=r[a],i=t[a],o=e[a];if(!va.arraysEqual(s.shape,o.shape))throw new og(`Layer weight shape ${s.shape} not compatible with provided weight shape ${o.shape}`);n.push([i,o])}bb(n)})}addWeight(e,t,n,r,a,s,i,o){if(-1!==this._addedWeightNames.indexOf(e))throw new og(`Duplicate weight name ${e} for layer ${this.name}`);this._addedWeightNames.push(e),null==n&&(n="float32"),this.fastWeightInitDuringBuild&&(r=null!=o?o():jy("zeros"));const u=r.apply(t,n),l=new gb(u,n,e,s,i);return u.dispose(),null!=a&&this.addLoss(()=>a.apply(l.read())),null==s&&(s=!0),s?this._trainableWeights.push(l):this._nonTrainableWeights.push(l),l}setFastWeightInitDuringBuild(e){this.fastWeightInitDuringBuild=e}addLoss(e){null==e||Array.isArray(e)&&0===e.length||(e=fg(e),void 0!==this._losses&&null!==this._losses&&this.losses.push(...e))}computeOutputShape(e){return e}computeMask(e,t){if(!this.supportsMasking){if(null!=t){if(!Array.isArray(t))throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);t.forEach(e=>{if(null!=e)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)})}return null}return t}addInboundNode(e,t,n,r,a,s,i=null){const o=fg(e);t=fg(t),n=fg(n),r=fg(r),a=pb(a),s=pb(s);const u=[],l=[],c=[];for(const e of o)u.push(e.sourceLayer),l.push(e.nodeIndex),c.push(e.tensorIndex);new kb({outboundLayer:this,inboundLayers:u,nodeIndices:l,tensorIndices:c,inputTensors:o,outputTensors:t,inputMasks:n,outputMasks:r,inputShapes:a,outputShapes:s},i);for(let e=0;e<t.length;e++)t[e].sourceLayer=this,t[e].nodeIndex=this.inboundNodes.length-1,t[e].tensorIndex=e}getConfig(){const e={name:this.name,trainable:this.trainable};return null!=this.batchInputShape&&(e.batchInputShape=this.batchInputShape),null!=this.dtype&&(e.dtype=this.dtype),e}disposeWeights(){return this.weights.forEach(e=>e.dispose()),this.weights.length}assertNotDisposed(){if(0===this._refCount)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(null===this._refCount)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let e=0;return 0==--this._refCount&&(e=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:e}}};var Ab,Eb,Nb=class extends Ib{constructor(e){if(super({dtype:e.dtype,name:null!=e.name?e.name:cb("input").toString()}),null==e.batchSize&&(e.batchSize=null),null==e.sparse&&(e.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=e.sparse,null!=e.inputShape&&null!=e.batchInputShape)throw new og("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let t=e.batchInputShape;if(null==t){if(null==e.inputShape)throw new og("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");t=[e.batchSize].concat(e.inputShape)}else if(null!=e.batchSize)throw new og("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");const n=e.dtype||"float32";this.batchInputShape=t,this.dtype=n,this.inputSpec=[{shape:t}];const r=new xb(this.dtype,this.batchInputShape,this,[],{},this.name);r.nodeIndex=0,r.tensorIndex=0,new kb({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[r],outputTensors:[r],inputMasks:[null],outputMasks:[null],inputShapes:[t],outputShapes:[t]})}apply(e,t){throw new og("Cannot pass any input to an InputLayer's apply() method. InputLayer name: "+this.name)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};function Cb(e){if(null==e.batchShape&&null==e.shape)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(null!=e.batchShape&&null!=e.shape)throw new og("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=e.batchShape;null!=e.shape&&null==t&&(t=[null].concat(e.shape));let n=e.dtype;null==n&&(n="float32");return new Nb({batchInputShape:t,name:e.name,dtype:n,sparse:e.sparse}).inboundNodes[0].outputTensors[0]}async function Tb(e){if(null==e)return;const t=[],n=[],r=[];for(const a in e){const s=e[a];if("number"!=typeof s){const e=s;t.push(e.data()),n.push(a),r.push(e)}}if(t.length>0){const a=await Promise.all(t);for(let t=0;t<a.length;++t)e[n[t]]=a[t][0];au(r)}}function Rb(e){if(null!=e)for(const t in e){const n=e[t];"number"!=typeof n&&n.dispose()}}Nb.className="InputLayer",Fo.registerClass(Nb),(Eb=Ab||(Ab={}))[Eb.SILENT=0]="SILENT",Eb[Eb.VERBOSE=1]="VERBOSE";var _b=class{constructor(){this.validationData=null}setParams(e){this.params=e}async onEpochBegin(e,t){}async onEpochEnd(e,t){}async onBatchBegin(e,t){}async onBatchEnd(e,t){}async onTrainBegin(e){}async onTrainEnd(e){}setModel(e){}},Fb=class{constructor(e,t=10){null==e&&(e=[]),this.callbacks=e,this.queueLength=t}append(e){this.callbacks.push(e)}setParams(e){for(const t of this.callbacks)t.setParams(e)}setModel(e){for(const t of this.callbacks)t.setModel(e)}async onEpochBegin(e,t){null==t&&(t={});for(const n of this.callbacks)await n.onEpochBegin(e,t)}async onEpochEnd(e,t){null==t&&(t={});for(const n of this.callbacks)await n.onEpochEnd(e,t)}async onBatchBegin(e,t){null==t&&(t={});for(const n of this.callbacks)await n.onBatchBegin(e,t)}async onBatchEnd(e,t){null==t&&(t={});for(const n of this.callbacks)await n.onBatchEnd(e,t)}async onTrainBegin(e){null==e&&(e={});for(const t of this.callbacks)await t.onTrainBegin(e)}async onTrainEnd(e){null==e&&(e={});for(const t of this.callbacks)await t.onTrainEnd(e)}},Mb=class extends _b{constructor(){super()}async onEpochBegin(e){this.seen=0,this.totals={}}async onBatchEnd(e,t){null==t&&(t={});const n=null==t.size?0:t.size;this.seen+=n;for(const e in t){const r=t[e];if("number"==typeof r)this.totals.hasOwnProperty(e)||(this.totals[e]=0),this.totals[e]=this.totals[e]+r*n;else{let t;e in this.totals?t=this.totals[e]:this.totals[e]=0;const a=ru(()=>gu(this.totals[e],vu(r,n)));this.totals[e]=a,null!=t&&t.dispose()}}}async onEpochEnd(e,t){if(null!=t)for(const e of this.params.metrics)null!=this.totals[e]&&("number"==typeof this.totals[e]?t[e]=this.totals[e]/this.seen:ru(()=>{const n=vu(bu(1,this.seen),this.totals[e]);t[e]=n,this.totals[e].dispose(),su(t[e])}))}},Db=class extends _b{async onTrainBegin(e){this.epoch=[],this.history={}}async onEpochEnd(e,t){null==t&&(t={}),this.epoch.push(e);for(const e in t)null==this.history[e]&&(this.history[e]=[]),this.history[e].push(t[e])}async syncData(){const e=[],t=[],n=[];for(const r in this.history){const a=this.history[r];for(let s=0;s<a.length;++s)if("number"!=typeof a[s]){const i=a[s];e.push(i.data()),t.push(r),n.push(s)}}const r=await Promise.all(e);for(let e=0;e<r.length;++e){this.history[t[e]][n[e]].dispose(),this.history[t[e]][n[e]]=r[e][0]}}},Ob=class extends _b{constructor(e,t){if(super(),this.currentEpoch=0,this.nowFunc=e.nowFunc,this.nextFrameFunc=e.nextFrameFunc||Vd,this.yieldEvery=t||"auto","auto"===this.yieldEvery&&(this.yieldEvery=125),"never"===this.yieldEvery&&null!=e.onYield)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");va.isNumber(this.yieldEvery)&&(this.maybeWait=function(e,t,n){let r,a=null!=n?n():va.now();return(...s)=>{const i=null!=n?n():va.now();return i-a<t||(a=i,r=e(...s)),r}}(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=e.onTrainBegin,this.trainEnd=e.onTrainEnd,this.epochBegin=e.onEpochBegin,this.epochEnd=e.onEpochEnd,this.batchBegin=e.onBatchBegin,this.batchEnd=e.onBatchEnd,this.yield=e.onYield}async maybeWait(e,t,n){const r=[];null!=this.yield&&(await Tb(n),r.push(this.yield(e,t,n))),r.push(this.nextFrameFunc()),await Promise.all(r)}async onEpochBegin(e,t){this.currentEpoch=e,null!=this.epochBegin&&(await Tb(t),await this.epochBegin(e,t))}async onEpochEnd(e,t){const n=[];null!=this.epochEnd&&(await Tb(t),n.push(this.epochEnd(e,t))),"epoch"===this.yieldEvery&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(e,t){null!=this.batchBegin&&(await Tb(t),await this.batchBegin(e,t))}async onBatchEnd(e,t){const n=[];null!=this.batchEnd&&(await Tb(t),n.push(this.batchEnd(e,t))),"batch"===this.yieldEvery?n.push(this.nextFrameFunc()):va.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,e,t)),await Promise.all(n)}async onTrainBegin(e){null!=this.trainBegin&&(await Tb(e),await this.trainBegin(e))}async onTrainEnd(e){null!=this.trainEnd&&(await Tb(e),await this.trainEnd(e))}};function Lb(e,t){if(null==e&&(e={}),e instanceof _b)return[e];if(Array.isArray(e)&&e[0]instanceof _b)return e;return fg(e).map(e=>new Ob(e,t))}var Pb=class{constructor(){}static registerCallbackConstructor(e,t){va.assert(e>=0&&Number.isInteger(e),()=>"Verbosity level is expected to be an integer >= 0, but got "+e),Pb.checkForDuplicate(t),null==Pb.constructors[e]&&(Pb.constructors[e]=[]),Pb.constructors[e].push(t)}static checkForDuplicate(e){for(const t in Pb.constructors){Pb.constructors[+t].forEach(t=>{if(t===e)throw new og("Duplicate callback constructor.")})}}static clear(){Pb.constructors={}}static createCallbacks(e){const t=[];for(const n in Pb.constructors){const r=+n;e>=r&&t.push(...Pb.constructors[r])}return t.map(e=>new e)}};function $b(e,t,n,r,a,s,i,o,u){const l=new Db,c=[new Mb,...Pb.createCallbacks(t)];null!=e&&c.push(...e),c.push(l);const h=new Fb(c);return h.setParams({epochs:n,initialEpoch:r,samples:a,steps:s,batchSize:i,verbose:t,doValidation:o,metrics:u}),{callbackList:h,history:l}}function Bb(e,t={},n=!1){return vg(e,Fo.SerializationMap.getMap().classNameMap,t,"layer",n)}function zb(e,t){return ru(()=>{"float32"!==e.dtype&&(e=Ai(e,"float32"));const n=yc(by(e),t,!0),r=Vl(n.shape,ag()),a=Wh(Oc(n,r));return bu(e,a)})}function Ub(e,t){return ru(()=>Lc(by(gc(t,e)),-1))}function Wb(e,t){return ru(()=>Lc(xu(gc(t,e)),-1))}function Vb(e,t){return ru(()=>{const n=gc(e,t),r=cl(xu(e),ag(),Number.MAX_VALUE),a=xu(bu(n,r));return vu(100,Lc(a,-1))})}function jb(e,t,n=!1){return ru(()=>{if(n)t=Lh(t);else{const e=yc(t,t.shape.length-1,!0);t=bu(t,e)}return t=cl(t,ag(),1-ag()),pc(yc(vu(Ai(e,"float32"),rc(t)),t.shape.length-1))})}function Gb(e,t,n=!1){return ru(()=>{const r=Ai(jl(function(e){const t=[ry(e.shape)];return qu(e,t)}(e)),"int32"),a=(t=cl(t,ag(),1-ag())).shape;return jb(qu(Hi(r,a[a.length-1]),a),t,n)})}function Hb(e,t){return ru(()=>{let n;return n=cl(t,ag(),1-ag()),n=rc(bu(n,gc(1,n))),Lc(function(e,t){if(!va.arraysEqual(e.shape,t.shape))throw new og(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);return ru(()=>{const n=gh(t),r=pc(xu(t));return gu(gc(n,vu(t,e)),ac($l(r)))})}(e,n),-1)})}function qb(e,t){return ru(()=>{const n=zb(e,-1),r=zb(t,-1),a=vu(n,r);return pc(yc(a,-1))})}Pb.constructors={};var Kb={meanSquaredError:Ub,meanAbsoluteError:Wb,meanAbsolutePercentageError:Vb,meanSquaredLogarithmicError:function(e,t){return ru(()=>{const n=cl(t,ag(),Number.MAX_VALUE),r=rc(gu(1,n)),a=cl(e,ag(),Number.MAX_VALUE),s=rc(gu(1,a));return Lc(by(gc(r,s)),-1)})},squaredHinge:function(e,t){return ru(()=>{const n=Oc(0,gc(1,vu(e,t)));return Lc(by(n),-1)})},hinge:function(e,t){return ru(()=>{const n=Oc(0,gc(1,vu(e,t)));return Lc(n,-1)})},categoricalHinge:function(e,t){return ru(()=>{const n=yc(vu(e,t),-1),r=mc(vu(gc(1,e),t),-1);return Oc(0,gu(1,gc(r,n)))})},logcosh:function(e,t){return ru(()=>{const n=Math.log(2),r=gc(t,e),a=gc(gu(r,dc(vu(-2,r))),n);return Lc(a,-1)})},categoricalCrossentropy:jb,sparseCategoricalCrossentropy:Gb,binaryCrossentropy:Hb,kullbackLeiblerDivergence:function(e,t){return ru(()=>{const n=cl(e,ag(),1),r=cl(t,ag(),1);return yc(vu(e,rc(bu(n,r))),-1)})},poisson:function(e,t){return ru(()=>{const n=rc(gu(ag(),t));return Lc(gc(t,vu(e,n)),-1)})},cosineProximity:qb};function Yb(e){if("string"==typeof e){if(e in Kb)return Kb[e];let t="Unknown loss "+e;throw e.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new og(t)}return e}function Xb(e,t){return ru(()=>{const n=vu(.5,Yc(t)),r=oy(Hl(t,n),e.dtype);return Lc(Rl(e,r),-1)})}function Qb(e,t){return ru(()=>oy(Rl(Eu(e,-1),Eu(t,-1)),"float32"))}function Zb(e,t){return ru(()=>Ai(yc(Cc(Rl(e,1),Rl(t,1))),"float32"))}function Jb(e,t){return ru(()=>{const n=Zb(e,t),r=function(e,t){return ru(()=>Ai(yc(Cc(Rl(e,0),Rl(t,1))),"float32"))}(e,t),a=gu(n,r);return Ai(_l(Hl(a,0),bu(n,a),0),"float32")})}function ev(e,t){return ru(()=>{const n=Zb(e,t),r=function(e,t){return ru(()=>Ai(yc(Cc(Rl(e,1),Rl(t,0))),"float32"))}(e,t),a=gu(n,r);return Ai(_l(Hl(a,0),bu(n,a),0),"float32")})}function tv(e,t){return Hb(e,t)}function nv(e,t){return e.rank===t.rank&&(e=jh(e,[e.rank-1])),(t=Eu(t,-1)).dtype!==e.dtype&&(t=Ai(t,e.dtype)),Ai(Rl(e,t),"float32")}var rv=jb,av=Gb,sv={binaryAccuracy:Xb,categoricalAccuracy:Qb,precision:Jb,categoricalCrossentropy:rv,sparseCategoricalCrossentropy:av,mse:Ub,MSE:Ub,mae:Wb,MAE:Wb,mape:Vb,MAPE:Vb,cosine:qb};function iv(e){if("string"==typeof e&&e in sv)return sv[e];if("string"!=typeof e&&null!=e)return e;throw new og("Unknown metric "+e)}function ov(e){if(hg(null!==e,"Unknown LossOrMetricFn "+e),"string"==typeof e)return e;{let t;for(const n of Object.keys(Kb))if(Kb[n]===e){t=n;break}if(void 0!==t)return t;for(const n of Object.keys(sv))if(sv[n]===e){t=n;break}return void 0!==t?t:e.name}}function uv(e,t,n=!1){if(null==e||"object"!=typeof e||Object.getPrototypeOf(e)!==Object.prototype||!function e(t){if(null===t)return!0;if("object"==typeof t){if(Object.getPrototypeOf(t)===Object.prototype){const n=Object.keys(t);for(const r of n){if("string"!=typeof r)return!1;if(!e(t[r]))return!1}return!0}if(Array.isArray(t)){for(const n of t)if(!e(n))return!1;return!0}return!1}{const e=typeof t;return"string"===e||"number"===e||"boolean"===e}}(e))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(n){const n=JSON.stringify(e);n.length>1048576&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${n.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= 1048576.`)}}function lv(e,t,n,r=console.log){const a=function(e){let t=!0;const n=[],r=[];for(const t in e.nodesByDepth)n.push(e.nodesByDepth[t]);for(const e of n){if(e.length>1||1===e.length&&e[0].inboundLayers.length>1){t=!1;break}r.push(...e)}if(t)for(const n of e.layers){let e=!1;for(const a of n.inboundNodes)if(-1!==r.indexOf(a)){if(e){t=!1;break}e=!0}if(!t)break}return t}(e),s=["Layer (type)","Output shape","Param #"];let i;if(a?(t=t||65,n=n||[.45,.85,1]):(t=t||98,n=n||[.33,.55,.67,1]),n[n.length-1]<=1&&(n=n.map(e=>Math.floor(t*e))),!a){s.push("Receives inputs"),i=[];for(const t in e.nodesByDepth)i.push(...e.nodesByDepth[t])}r("_".repeat(t)),cv(s,n,r),r("=".repeat(t));const o=e.layers;for(let e=0;e<o.length;++e)a?hv(o[e],n,r):pv(o[e],n,i,r),r((e===o.length-1?"=":"_").repeat(t));e.checkTrainableWeightsConsistency();const u=function(e){let t;t=null!=e.collectedTrainableWeights?mb(e.collectedTrainableWeights):mb(e.trainableWeights);return t}(e),l=mb(e.nonTrainableWeights);r("Total params: "+(u+l)),r("Trainable params: "+u),r("Non-trainable params: "+l),r("_".repeat(t))}function cv(e,t,n=console.log){let r="";for(let n=0;n<e.length;++n)n>0&&(r=r.slice(0,r.length-1)+" "),r+=e[n],r=r.slice(0,t[n]),r+=" ".repeat(t[n]-r.length);n(r)}function hv(e,t,n){let r;try{r=JSON.stringify(e.outputShape)}catch(e){r="multiple"}cv([`${e.name} (${e.getClassName()})`,r,e.countParams().toString()],t,n)}function pv(e,t,n,r){let a;try{a=JSON.stringify(e.outputShape)}catch(e){a="multiple"}const s=[];for(const t of e.inboundNodes)if(!(null!=n&&n.length>0&&-1===n.indexOf(t)))for(let e=0;e<t.inboundLayers.length;++e){const n=t.inboundLayers[e].name,r=t.nodeIndices[e],a=t.tensorIndices[e];s.push(`${n}[${r}][${a}]`)}const i=e.name,o=e.getClassName(),u=0===s.length?"":s[0];cv([`${i} (${o})`,a,e.countParams().toString(),u],t,r);for(let e=1;e<s.length;++e)cv(["","","",s[e]],t,r)}function dv(e,t,n){return("inboundNodes"===e||"outputLayers"===e||"inputLayers"===e)&&0===t&&"string"==typeof n}function fv(e,t){if(null===e)return null;if("string"==typeof e)return gg(e);if("number"==typeof e||"boolean"==typeof e)return e;if(e instanceof Array){const n=[],r=e.length;for(let a=0;a<r;++a){const r=e[a];dv(t,a,r)?n.push(r):n.push(fv(r,t))}return n}{const t={};for(const n of Object.keys(e)){const r=e[n];if("name"===n&&"string"==typeof r)t[n]=r;else{const e=gg(n);t[e]=fv(r,e)}}return t}}var mv="3.13.0";var gv=class{constructor(e){if(this.id2Value={},this.id2Mask={},this.name2Id={},e instanceof gv)for(const t in e.id2Value)this.id2Value[t]=e.id2Value[t],t in e.id2Mask&&(this.id2Mask[t]=e.id2Mask[t]);else{if(null==e)return;for(const t of e)this.add(t.key,t.value)}}add(e,t,n){if(null!=this.id2Value[e.id])throw new og(`Duplicate key: name=${e.name}, id=${e.id}`);return this.id2Value[e.id]=function(e,t){if(null==e.dtype||e.dtype===t.dtype)return t;try{return Ai(t,e.dtype)}catch(n){throw new og(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)}}(e,t),this.name2Id[e.name]=e.id,null!=n&&(this.id2Mask[e.id]=n),this}addFeed(e){this.add(e.key,e.value)}hasKey(e){return null!=this.id2Value[e.id]}names(){return Object.keys(this.name2Id)}getValue(e){if(e instanceof xb){if(null==this.id2Value[e.id])throw new og("Nonexistent key: "+e.name);return this.id2Value[e.id]}{const t=this.name2Id[e];if(null==t)throw new og("Feed dict has no SymbolicTensor name: "+e);return this.id2Value[t]}}getMask(e){if(e instanceof xb){if(null==this.id2Value[e.id])throw new og("Nonexistent key: "+e.name);return this.id2Mask[e.id]}{const t=this.name2Id[e];if(null==t)throw new og("Feed dict has no SymbolicTensor name: "+e);return this.id2Mask[t]}}disposeMasks(){null!=this.id2Mask&&au(this.id2Mask)}},yv={},bv={};function vv(e,t,n,r){const a=null!=n&&n.training,s=Array.isArray(e),i=s?e:[e],o=i.map(e=>e.name),u=[],l=t.names();for(const e of o)-1!==l.indexOf(e)?u.push(t.getValue(e)):u.push(null);null!=r&&(r.maxNumTensors=-1/0,r.minNumTensors=1/0);const c=o.join(",")+"|"+t.names().join(",");let h,p;if(null==yv[c]){const e=function(e,t){va.assert(null!=e&&e.length>0,()=>"Expected at least one fetch, got none");let n=[],r={};if(1===e.length){const a=wv(e[0],t);n=a.sorted,r=a.recipientMap}else{const a=new Set;for(const s of e){const{sorted:e,recipientMap:i}=wv(s,t);for(const t of e)a.has(t.name)||(n.push(t),a.add(t.name));for(const e in i)null==r[e]&&(r[e]=new Set),i[e].forEach(t=>r[e].add(t))}}return{sorted:n,recipientCounts:xv(r)}}(i,t);h=e.sorted,p=e.recipientCounts,yv[c]=h,bv[c]=p}h=yv[c],p={},a||Object.assign(p,bv[c]);const d=new gv(t);for(let e=0;e<h.length;++e){if(null!=r){const e=tu().numTensors;e>r.maxNumTensors&&(r.maxNumTensors=e),e<r.minNumTensors&&(r.minNumTensors=e)}const s=h[e],i=s.sourceLayer;if(i instanceof Nb)continue;const l=[],c=[],f=[];let m=!1;for(const e of s.inputs){const n=d.getValue(e),r=d.getMask(e);l.push(n),c.push(r),null!=r&&(m=!0),a||(p[e.name]--,0!==p[e.name]||t.hasKey(e)||-1!==o.indexOf(e.name)||n.isDisposed||!0===e.sourceLayer.stateful||f.push(n))}m&&((n=n||{}).mask=c[0]);const g=fg(i.apply(l,n));let y=null;i.supportsMasking&&(y=i.computeMask(l,c));const b=kv(s),v=Array.isArray(b)?b:[b];for(let e=0;e<v.length;++e){d.hasKey(v[e])||d.add(v[e],g[e],Array.isArray(y)?y[0]:y);const t=o.indexOf(v[e].name);-1!==t&&(u[t]=g[e])}a||au(f)}return d.disposeMasks(),s?u:u[0]}function xv(e){const t={};for(const n in e)t[n]=e[n].size;return t}function wv(e,t){const n=new Set,r=[],a={};for(const e of t.names())n.add(e);const s=[],i=[];for(s.push(e);s.length>0;){const e=s[s.length-1];if(n.has(e.name)){s.pop();continue}const t=i[i.length-1]===s.length-1;if(0===e.inputs.length||t)s.pop(),r.push(e),n.add(e.name),t&&i.pop();else{i.push(s.length-1);for(const t of e.inputs)null==a[t.name]&&(a[t.name]=new Set),a[t.name].add(e.name),n.has(t.name)||s.push(t)}}return{sorted:r,recipientMap:a}}function kv(e){let t;if(1===e.sourceLayer.inboundNodes.length)t=e.sourceLayer.output;else{let n=null;for(let t=0;t<e.sourceLayer.inboundNodes.length;++t)for(const r of e.sourceLayer.inboundNodes[t].outputTensors)if(r.id===e.id){n=t;break}t=e.sourceLayer.getOutputAt(n)}return t}var Sv=class extends Ib{constructor(e){if(super({}),this.containerNodes=new Set,this.name=e.name,null==this.name){const e=this.getClassName().toLowerCase();this.name=cb(e)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(e.inputs)?this.inputs=e.inputs.slice():this.inputs=[e.inputs],Array.isArray(e.outputs)?this.outputs=e.outputs.slice():this.outputs=[e.outputs],wg(this.inputs).length!==this.inputs.length)throw new og("The list of inputs passed to the model is redundant. All inputs should only appear once. Found: "+this.inputs.map(e=>e.name));wg(this.outputs).length!==this.outputs.length&&console.warn("The list of outputs passed to the model is redundant. All outputs should only appear once. Found: "+this.outputs.map(e=>e.name)),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(const e of this.outputs){const t=e.sourceLayer,n=e.nodeIndex,r=e.tensorIndex;this.outputLayers.push(t),this.outputLayersNodeIndices.push(n),this.outputLayersTensorIndices.push(r)}for(const e of this.inputs){const t=e.sourceLayer,n=e.nodeIndex,r=e.tensorIndex;hg(0===n,"input layer has >1 nodes"),hg(0===r,"input layer has >1 tensors"),this.inputLayers.push(t),this.inputLayersNodeIndices.push(n),this.inputLayersTensorIndices.push(r)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let t=0;t<this.inputLayers.length;t++){const n=this.inputLayers[t];if(!(n instanceof Nb))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${e.inputs}. Input ${t} (0-based) originates from layer type ${n.getClassName()}.`);this.inputNames.push(n.name),this.feedInputShapes.push(n.batchInputShape),this.feedInputNames.push(n.name)}for(const e of this.outputLayers)this.outputNames.push(e.name);this.internalInputShapes=this.inputs.map(e=>e.shape),this.internalOutputShapes=this.outputs.map(e=>e.shape);const t={},n={},r={},a={},s={},i=[],o=(e,t,n,r,a,u)=>{null!=r&&null!=a&&null!=u||(r=e.sourceLayer,a=e.nodeIndex,u=e.tensorIndex);const l=r.inboundNodes[a];if(-1!==n.indexOf(l))throw new ig(`The tensor ${e.name} at layer "${r.name}" is part of a cycle.`);if(-1!==t.indexOf(l))return;this.containerNodes.add(Sv.nodeKey(r,a)),r.id in s||(s[r.id]=Object.keys(s).length),-1===n.indexOf(l)&&n.push(l);const c=l.inboundLayers.length;for(let e=0;e<c;e++){const r=l.inputTensors[e],a=l.inboundLayers[e],s=l.nodeIndices[e],i=l.tensorIndices[e];o(r,t,n,a,s,i)}for(t.push(l);n.indexOf(l)>=0;)n.splice(n.indexOf(l),1);i.push(l)},u=[],l=[];for(const e of this.outputs)o(e,u,l);const c=i.slice().reverse();for(const e of c){n[e.id]=e,e.id in t||(t[e.id]=0);let s=t[e.id];const i=null==r[e.outboundLayer.id]?0:r[e.outboundLayer.id];s=Math.max(s,i),r[e.outboundLayer.id]=s,a[e.outboundLayer.id]=e.outboundLayer,t[e.id]=s;for(let r=0;r<e.inboundLayers.length;r++){const a=e.inboundLayers[r],i=e.nodeIndices[r],o=a.inboundNodes[i],u=null==t[o.id]?0:t[o.id];t[o.id]=Math.max(s+1,u),n[o.id]=o}}const h={};for(const e in t){const r=t[e];r in h||(h[r]=[]),h[r].push(n[e])}const p={};for(const e in r){const t=r[e];t in p||(p[t]=[]),p[t].push(a[e])}let d=Object.keys(p).map(e=>parseInt(e,10)).sort(xg);this.layers=[];for(const e of d){const t=p[e];t.sort((e,t)=>{const n=s[e.id],r=s[t.id];return n<r?-1:n>r?1:0});for(const e of t)e instanceof Sv&&this.internalContainerRefs.push(e),this.layers.push(e)}this.layersByDepth=p,d=Object.keys(h).map(e=>parseInt(e,10)).sort(xg);const f=this.inputs.slice(),m=[];for(const e of d)for(const t of h[e]){const e=t.outboundLayer;if(null!=e){for(const n of t.inputTensors)if(-1===f.indexOf(n))throw new ig(`Graph disconnected: cannot obtain value for tensor ${n} at layer "${e.name}". The following previous layers were accessed without issue: ${m}`);for(const e of t.outputTensors)f.push(e);m.push(e.name)}}this.nodesByDepth=h;const g=this.layers.map(e=>e.name);for(const e of g){const t=g.filter(t=>t===e).length;if(1!==t)throw new ig(`The name "${e}" is used ${t} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(g))}this.outboundNodes=[],this.inboundNodes=[],new kb({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(e=>null),outputMasks:this.outputs.map(e=>null),inputShapes:this.inputs.map(e=>e.shape),outputShapes:this.outputs.map(e=>e.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(0===this._refCount)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();const e={refCountAfterDispose:null,numDisposedVariables:0};if(0==--this._refCount){for(const t of this.layers)e.numDisposedVariables+=t.dispose().numDisposedVariables;for(const t of this.internalContainerRefs)e.numDisposedVariables+=t.dispose().numDisposedVariables}return e.refCountAfterDispose=this._refCount,e}get trainable(){return this.trainable_}set trainable(e){this.layers.forEach(t=>{t._trainableWeights.forEach(t=>t.trainable=e)}),this.trainable_=e}get trainableWeights(){if(this._trainableWeights.length>0)throw new og("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let e=[];for(const t of this.layers)e=e.concat(t.trainableWeights);return e}get nonTrainableWeights(){const e=[];for(const t of this.layers)e.push(...t.nonTrainableWeights);if(!this.trainable){const t=[];for(const e of this.layers)t.push(...e.trainableWeights);return t.concat(e)}return e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(e,t=!0){const n={};let r=0;for(const e of this.layers)for(const t of e.weights){if(null!=n[t.originalName])throw new og("Duplicate weight name: "+t.originalName);n[t.originalName]=t,r++}const a=[];for(const r in e){let s=r;if(null==n[r]){const e=r.split("/");s=e.slice(0,-2).concat([e[e.length-1]]).join("/")}if(null!=n[s])a.push([n[s],e[r]]);else if(t)throw new og("Provided weight data has no target variable: "+r);delete n[s]}if(t){const e=[];for(const t in n)e.push(t);if(e.length>0)throw new og(`${e.length} of ${r} weights are not set: ${e}`)}bb(a)}updatedConfig(){const e=this.getConfig(),t={};return t.className=this.getClassName(),t.config=e,t.kerasVersion="tfjs-layers "+mv,t.backend="TensorFlow.js",t}toJSON(e,t=!0){const n=function e(t,n){if(null==t)return null;if("string"==typeof t)return mg(t);if("number"==typeof t||"boolean"==typeof t)return t;if(t instanceof Array){const r=[],a=t.length;for(let s=0;s<a;++s){const a=t[s];dv(n,s,a)?r.push(a):r.push(e(a,n))}return r}{const n={};for(const r of Object.keys(t)){const a=t[r],s=mg(r);n[s]="name"!==r&&"className"!==r||"string"!=typeof a?e(a,r):a}return n}}(this.updatedConfig());return t?JSON.stringify(n):n}call(e,t){return ru(()=>{e=fg(e);const n=new gv;for(let t=0;t<this.inputs.length;++t)n.add(this.inputs[t],e[t]);return vv(this.outputs,n,t)})}computeMask(e,t){return ru(()=>{let n;return e=fg(e),n=null==t?cg(null,e.length):fg(t),this.runInternalGraph(e,n)[1]})}computeOutputShape(e){const t=pb(e);if(t.length!==this.inputLayers.length)throw new og(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);const n={};for(let e=0;e<t.length;e++){const r=this.inputLayers[e],a=t[e];n[r.name+"_0_0"]=a}const r=Object.keys(this.nodesByDepth).map(e=>parseInt(e,10)).sort(xg);if(r.length>1)for(const e of r){const t=this.nodesByDepth[e];for(const e of t){const t=e.outboundLayer;if(-1!==this.inputLayers.map(e=>e.id).indexOf(t.id))continue;const r=[];for(let t=0;t<e.inboundLayers.length;t++){const a=e.inboundLayers[t],s=e.nodeIndices[t],i=e.tensorIndices[t],o=n[`${a.name}_${s}_${i}`];r.push(o)}const a=pb(t.computeOutputShape(dg(r))),s=t.inboundNodes.indexOf(e);for(let e=0;e<a.length;e++){n[`${t.name}_${s}_${e}`]=a[e]}}}const a=[],s=[];for(let e=0;e<this.outputLayers.length;e++){const t=this.outputLayers[e],n=this.outputLayersNodeIndices[e],r=this.outputLayersTensorIndices[e],a=`${t.name}_${n}_${r}`;s.push(a)}for(let e=0;e<s.length;e++){const t=s[e];hg(t in n),a.push(n[t])}return dg(a)}runInternalGraph(e,t){null==t&&(t=cg(null,e.length));const n={};for(let r=0;r<this.inputs.length;++r){const a=this.inputs[r],s=e[r],i=t[r];n[a.id]=[s,i]}const r=Object.keys(this.nodesByDepth).map(e=>parseInt(e,10)).sort(xg);for(const e of r){const t=this.nodesByDepth[e];for(const e of t){const t=e.outboundLayer,r=e.inputTensors,a=e.outputTensors,s=new Array;for(const e of r)e.id in n&&s.push(n[e.id]);if(s.length===r.length){let r,i,o,u,l={};if(null!=e.callArgs&&(l=e.callArgs),1===s.length){const[e,n]=s[0];null==l.mask&&(l.mask=n),o=fg(t.call(e,l)),u=fg(t.computeMask(e,n)),r=[e],i=[n]}else r=s.map(e=>e[0]),i=s.map(e=>e[1]),null==l.mask&&(l.mask=i),o=fg(t.call(r,l)),u=fg(t.computeMask(r,i));if(t.activityRegularizer)throw new ug("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let e=0;e<a.length;++e){const t=a[e],r=o[e],s=u[e];n[t.id]=[r,s]}}}}const a=[],s=[],i=[];for(const e of this.outputs){hg(e.id in n,`Could not compute output ${e.name} : ${e.id}`);const[t,r]=n[e.id];i.push(t.shape),a.push(t),s.push(r)}return[a,s,i]}buildNodeConversionMap(e){const t={};let n;for(const e of this.layers){n=e instanceof Sv?1:0;for(let r=0;r<e.inboundNodes.length;r++){const a=Sv.nodeKey(e,r);this.containerNodes.has(a)&&(t[a]=n,n+=1)}}return t}getLayer(e,t){if(null!=t){if(this.layers.length<=t)throw new og(`Was asked to retrieve layer at index ${t}, but model only has ${this.layers.length} layer(s).`);return this.layers[t]}if(null==e)throw new og("Provide either a layer name or layer index");for(const t of this.layers)if(t.name===e)return t;throw new og("No such layer: "+e)}calculateLosses(){return ru(()=>{const e=[];for(const t of this.layers)for(let n=0;n<t.inboundNodes.length;++n){const r=Sv.nodeKey(t,n);this.containerNodes.has(r)&&e.push(...t.calculateLosses())}return e})}getConfig(){const e={name:this.name},t=this.buildNodeConversionMap(this.layers),n=[];for(const e of this.layers){const r=e.getClassName(),a=e.getConfig(),s=[];for(let n=0;n<e.inboundNodes.length;n++){const r=e.inboundNodes[n],a=Sv.nodeKey(e,n);let i={};if(this.containerNodes.has(a)){if(r.callArgs)try{JSON.stringify(r.callArgs),i=r.callArgs}catch(t){console.warn(`Layer ${e.name} was passed non-serializable keyword arguments: ${r.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),i={}}if(r.inboundLayers.length>0){const e=[];for(let n=0;n<r.inboundLayers.length;n++){const a=r.inboundLayers[n],s=r.nodeIndices[n],o=r.tensorIndices[n];let u=t[Sv.nodeKey(a,s)];null==u&&(u=0),e.push([a.name,u,o,i])}s.push(e)}}}const i={};i.name=e.name,i.className=r,i.config=a,i.inboundNodes=s,n.push(i)}e.layers=n;const r=[];for(let e=0;e<this.inputLayers.length;e++){const n=this.inputLayers[e],a=this.inputLayersNodeIndices[e],s=Sv.nodeKey(n,a);if(!this.containerNodes.has(s))continue;let i=t[s];null==i&&(i=0);const o=this.inputLayersTensorIndices[e];r.push([n.name,i,o])}e.inputLayers=r;const a=[];for(let e=0;e<this.outputLayers.length;e++){const n=this.outputLayers[e],r=this.outputLayersNodeIndices[e],s=Sv.nodeKey(n,r);if(!this.containerNodes.has(s))continue;let i=t[s];null==i&&(i=0);const o=this.outputLayersTensorIndices[e];a.push([n.name,i,o])}return e.outputLayers=a,e}static fromConfig(e,t,n={},r=!1){const a={},s={};function i(e,t){e.name in s?s[e.name].push(t):s[e.name]=[t]}function o(e,t){const n=[];let r;for(const s of t){const o=s[0],u=s[1],l=s[2];if(r=null==s[3]?{}:s[3],!(o in a))return void i(e,t);const c=a[o];if(c.inboundNodes.length<=u)return void i(e,t);const h=c.inboundNodes[u];n.push(h.outputTensors[l])}n.length>0&&e.apply(dg(n),r)}function u(e){const n=e.name,s=Bb(e,null!=t.customObjects?t.customObjects:{});s.setFastWeightInitDuringBuild(r),a[n]=s;e.inboundNodes.forEach(e=>{if(!(e instanceof Array))throw new og("Corrupted configuration, expected array for nodeData: "+e);i(s,e)})}const l=t.name,c=t.layers;for(const e of c)u(e);for(;!kg(s);)for(const e of c){const t=a[e.name];if(t.name in s){const e=s[t.name];delete s[t.name];for(const n of e)o(t,n)}}const h=[],p=[],d=t.inputLayers;for(const e of d){const t=e[0],n=e[1],r=e[2];hg(t in a);const s=a[t].inboundNodes[n].outputTensors;h.push(s[r])}const f=t.outputLayers;for(const e of f){const t=e[0],n=e[1],r=e[2];hg(t in a);const s=a[t].inboundNodes[n].outputTensors;p.push(s[r])}return new e({inputs:h,outputs:p,name:l})}get stateful(){if(this._stateful)throw new og("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(const e of this.layers)if(e.stateful)return!0;return!1}resetStates(){ru(()=>{this.layers.forEach(e=>{e.stateful&&e.resetStates()})})}};function Iv(e,t){return function(e,t,n){const r=t.length;if(null==e||Array.isArray(e)&&0===e.length)return t.map(e=>null);if(1===r)return Array.isArray(e)&&1===e.length?e:"object"==typeof e&&t[0]in e?[e[t[0]]]:[e];if(Array.isArray(e)){if(e.length!==r)throw new Error(`Provided ${n} is an array of ${e.length} element(s), but the model has ${r} outputs. Make sure a set of weights is provided for each model output.`);return e}if("object"==typeof e&&Object.keys(e).length>0&&"object"==typeof e[Object.keys(e)[0]]){const n=[];return t.forEach(t=>{t in e?n.push(e[t]):n.push(null)}),n}throw new Error(`The model has multiple (${r}) outputs, so ${n} must be either an array with ${r} elements or an object with ${t} keys. Provided ${n} not understood: ${JSON.stringify(e)}`)}(e,t,"classWeight")}async function Av(e,t,n,r){if(null!=t||null!=r)throw new Error("Support sampleWeight is not implemented yet");if(null!=n){const t=ru(()=>{if(1===e.shape.length)return Ei(e);if(2===e.shape.length){if(e.shape[1]>1){return Eu(e,1)}if(1===e.shape[1])return qu(e,[e.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),r=Array.from(await t.data());au(t);const a=[];return r.forEach(e=>{if(null==n[e])throw new Error(`classWeight must contain all classes in the training data. The class ${e} exists in the data but not in classWeight`);a.push(n[e])}),Yh(a,"float32")}return null}function Ev(e,t){return vu(e,t)}function Nv(e,t){let n,r;const a=t;n=a.xs,r=a.ys,va.assert(null!=n&&null!=r,()=>"A Dataset iterator for fitDataset() is expected to generate objects of the form `{xs: xVal, ys: yVal}`, where the two values may be `tf.Tensor`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates "+t);const s=Cv("input",e.inputNames,n),i=Cv("output",e.outputNames,r),o=s[0].shape[0];va.assert(s.length===e.inputs.length,()=>`LayersModel has ${e.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(e.inputNames)})`),va.assert(i.length===e.outputs.length,()=>`LayersModel has ${e.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(e.outputNames)})`);for(let t=0;t<s.length;t++)va.assert(s[t].shape[0]===o,()=>`Batch size mismatch: input ${e.inputNames[t]} has ${s[t].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);for(let t=0;t<i.length;t++)va.assert(i[t].shape[0]===o,()=>`Batch size mismatch: output ${e.outputNames[t]} has ${i[t].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);return{xs:s,ys:i}}function Cv(e,t,n){if(n instanceof Ya)return[n];if(Array.isArray(n))return va.assert(n.length===t.length,()=>`Received an array of ${n.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`),n;{const r=[];for(const a of t){if(null==n[a])throw new og(`The feature data generated by the dataset lacks the required ${e} key '${a}'.`);r.push(n[a])}return r}}async function Tv(e,t,n){const r=null!=n.batchesPerEpoch;if(va.assert(null!=e.optimizer,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),va.assert(null!=n,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),va.assert(null!=n.epochs&&n.epochs>0&&Number.isInteger(n.epochs),()=>"For fitDataset(), config.epochs is expected to be a positive integer, but got "+n.epochs),va.assert(!r||n.batchesPerEpoch>0&&Number.isInteger(n.batchesPerEpoch),()=>"For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got "+n.batchesPerEpoch),va.assert(null==n.validationSplit,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;try{const a=null!=n.validationData;let s,i;if(a)if(Rv(n.validationData))va.assert(null==n.validationBatches||n.validationBatches>0&&Number.isInteger(n.validationBatches),()=>"For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got "+n.validationBatches);else{const e=function(e){if(3===e.length)throw new ug("Validation with sample weights is not implemented yet.");return{xs:e[0],ys:e[1]}}(n.validationData);s=e.xs,i=e.ys}const o=e.makeTrainFunction(),u=e.getDedupedMetricsNames();let l;l=a?u.slice().concat(u.map(e=>"val_"+e)):u.slice();const c=Lb(n.callbacks,n.yieldEvery),h=null==n.verbose?1:n.verbose,{callbackList:p,history:d}=$b(c,h,n.epochs,null,null,function(e,t){let n=null;null!=t.batchesPerEpoch?n=t.batchesPerEpoch:Number.isFinite(e.size)&&(n=e.size);return n}(t,n),null,a,l);p.setModel(e),e.history=d,await p.onTrainBegin(),e.stopTraining_=!1;let f=null==n.initialEpoch?0:n.initialEpoch,m=await t.iterator();for(;f<n.epochs;){const l={};await p.onEpochBegin(f);let c=0,h=0;for(r||(m=await t.iterator());!r||c<n.batchesPerEpoch;){const t=await m.next();if(r&&t.done){console.warn(`You provided \`batchesPerEpoch\` as ${n.batchesPerEpoch}, but your dataset iterator ran out of data after ${c} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${n.batchesPerEpoch*n.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(null!=t.value){const{xs:r,ys:a}=Nv(e,t.value),s={};s.batch=h,s.size=r[0].shape[0],await p.onBatchBegin(h,s);const i=[];if(null!=n.classWeight){const t=Iv(n.classWeight,e.outputNames);for(let e=0;e<t.length;++e)i.push(await Av(a[e],null,t[e]))}const l=r.concat(a).concat(i),d=o(l);au(l);for(let e=0;e<u.length;++e){const t=u[e],n=d[e];s[t]=n,su(n)}await p.onBatchEnd(h,s),Rb(s),h++,c++}if(r?c>=n.batchesPerEpoch:t.done){if(a){let t;t=Rv(n.validationData)?fg(await e.evaluateDataset(n.validationData,{batches:n.validationBatches})):fg(e.evaluate(s,i,{batchSize:null==n.validationBatchSize?32:n.validationBatchSize,verbose:0}));for(let n=0;n<e.metricsNames.length;++n)l["val_"+e.metricsNames[n]]=t[n]}break}if(e.stopTraining_)break}if(await p.onEpochEnd(f,l),f++,e.stopTraining_)break}return await p.onTrainEnd(),await e.history.syncData(),e.history}finally{e.isTraining=!1}}function Rv(e){return"function"==typeof e.iterator}function _v(e){va.assert(e>0&&Number.isInteger(e),()=>"batchSize is required to be a positive integer, but got "+e)}function Fv(e,t,n){return null==e?[null]:Array.isArray(e)?e.map(e=>ly(e,t,n-t)):ly(e,t,n-t)}function Mv(e,t){return ru(()=>null==e?null:Array.isArray(e)?e.map(e=>Mv(e,t)):yy(e,"int32"===t.dtype?t:Ai(t,"int32")))}function Dv(e,t){const n=[];let r=0,a=null;for(;r<e;)a=r+t,a>=e&&(a=e),n.push([r,a]),r=a;return n}async function Ov(e,t,n,r={}){if(e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");let a,s,i,o,u,l,c,h,p;e.isTraining=!0;try{const d=null==r.batchSize?32:r.batchSize;_v(d);const f=!1,m=await e.standardizeUserData(t,n,r.sampleWeight,r.classWeight,f,d);a=m[0],s=m[1],p=m[2];let g,y=!1;if(null!=r.validationData&&r.validationData.length>0){if(y=!0,2!==r.validationData.length)throw 3===r.validationData.length?new ug("validationData including sample weights is not supported yet."):new og(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${r.validationData} is invalid.`);u=r.validationData[0],l=r.validationData[1];const t=!0,n=await e.standardizeUserData(u,l,null,null,t,d);c=n[0],h=n[1],g=c.concat(h)}else if(null!=r.validationSplit&&r.validationSplit>0&&r.validationSplit<1){y=!0;const e=Math.floor(a[0].shape[0]*(1-r.validationSplit)),t=a[0].shape[0];c=Fv(a,e,t),i=a,a=Fv(a,0,e),h=Fv(s,e,t),o=s,s=Fv(s,0,e),g=c.concat(h)}else null!=r.validationSteps&&(y=!0);const b=a.concat(s).concat(p);e.checkTrainableWeightsConsistency();const v=e.makeTrainFunction(),x=e.getDedupedMetricsNames();let w,k;y?(e.makeTestFunction(),w=e.testFunction,k=x.slice().concat(x.map(e=>"val_"+e))):(w=null,g=[],k=x.slice());const S=Lb(r.callbacks,r.yieldEvery);return await async function(e,t,n,r,a,s,i,o,u,l,c,h,p,d,f){null==a&&(a=32),null==s&&(s=1),null==c&&(c=!0),null==p&&(p=0);let m=!1;if(null!=u&&null!=l&&(m=!0),null!=f&&(m=!0,null==d))throw new og("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");const g=e.checkNumSamples(n,a,d,"steps_per_epoch");let y;null!=g&&(y=iy(0,g)),null==i&&(i=1);const{callbackList:b,history:v}=$b(o,i,s,p,g,d,a,m,h);b.setModel(e),e.history=v,await b.onTrainBegin(),e.stopTraining_=!1;for(let i=p;i<s;++i){await b.onEpochBegin(i);const s={};if(null!=d)throw new ug("stepsPerEpoch mode is not implemented yet.");{if("batch"===c)throw new ug("batch shuffling is not implemneted yet");c&&va.shuffle(y);const i=Yh(y),o=Dv(g,a);for(let c=0;c<o.length;++c){const h={};if(await b.onBatchBegin(c,h),ru(()=>{const p=o[c][0],d=o[c][1],f=ly(i,p,d-p);h.batch=c,h.size=d-p;const g=Mv(n,f),y=t(g);for(let e=0;e<r.length;++e){const t=r[e],n=y[e];h[t]=n,su(n)}if(c===o.length-1&&m){const t=e.testLoop(u,l,a);for(let e=0;e<r.length;++e){const n=r[e],a=t[e];su(a),s["val_"+n]=a}}}),await b.onBatchEnd(c,h),Rb(h),e.stopTraining_)break}i.dispose()}if(await b.onEpochEnd(i,s),e.stopTraining_)break}return await b.onTrainEnd(),await e.history.syncData(),e.history}(e,v,b,x,d,r.epochs,r.verbose,S,w,g,r.shuffle,k,r.initialEpoch,null,null)}finally{e.isTraining=!1,Pv(a,t),Pv(s,n),Pv(i,t),Pv(o,n),Pv(c,u),Pv(h,l),null!=p&&au(p)}}function Lv(e){const t=[];e instanceof Ya&&(e=[e]);for(let n=0;n<e.length;++n){const r=e[n];if(1===r.rank)t.push(uy(r,1));else{if(0===r.rank)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(r)}}return t}function Pv(e,t){if(null==e)return;const n=[];if(t instanceof Ya)n.push(t.id);else if(Array.isArray(t))t.forEach(e=>n.push(e.id));else if(null!=t)for(const e in t){const r=t[e];n.push(r.id)}const r=[];if(e instanceof Ya)-1===n.indexOf(e.id)&&r.push(e);else if(Array.isArray(e))e.forEach(e=>{-1===n.indexOf(e.id)&&r.push(e)});else if(null!=e)for(const t in e){const a=e[t];-1===n.indexOf(a.id)&&r.push(a)}r.forEach(e=>{e.isDisposed||e.dispose()})}function $v(e){return Array.isArray(e)}function Bv(e){return!function(e){return e instanceof Ya}(e)&&!$v(e)}function zv(e,t,n,r=!0,a=""){if(null==t||0===t.length){if(null!=e){let t=!1;if($v(e)&&e.length>0)t=!0;else if(Bv(e)){for(const n in e)if(e.hasOwnProperty(n)){t=!0;break}}else t=!0;if(t)throw new og(`Error when checking model ${a} expected no data, but got ${e}`)}return[]}if(null==e)return t.map(e=>null);let s;if(Bv(e)){e=e,s=[];for(const n of t){if(null==e[n])throw new og(`No data provided for "${n}". Need data for each key in: ${t}`);s.push(e[n])}}else if($v(e)){if((e=e).length!==t.length)throw new og(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);s=e}else{if(e=e,t.length>1)throw new og(`The model ${a} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);s=[e]}if(s=Lv(s),null!=n)for(let e=0;e<t.length;++e){if(null==n[e])continue;const i=s[e];if(i.shape.length!==n[e].length)throw new og(`Error when checking ${a}: expected ${t[e]} to have ${n[e].length} dimension(s). but got array with shape ${i.shape}`);for(let t=0;t<n[e].length;++t){if(0===t&&!r)continue;const s=i.shape[t],o=n[e][t];if(null!=o&&o>=0&&s!==o)throw new og(`${a} expected a batch of elements where each example has shape [${n[e].slice(1,n[e].length)}] (i.e.,tensor shape [*,${n[e].slice(1,n[e].length)}]) but the ${a} received an input with ${i.shape[0]} examples, each with shape [${i.shape.slice(1,i.shape.length)}] (tensor shape [${i.shape}])`)}}return s}function Uv(e,t,n,r=!0,a=""){let s;if(Array.isArray(e)){if(e.length!==t.length)throw new og(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);s=e}else{if(t.length>1)throw new og(`The model expects ${t.length} ${a} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);s=[e]}if(null!=n)for(let e=0;e<t.length;++e){if(null==n[e])continue;const i=s[e];if(i.shape.length!==n[e].length)throw new og(`Error when checking ${a}: expected ${t[e]} to have ${n[e].length} dimension(s), but got array with shape ${JSON.stringify(i.shape)}`);for(let s=0;s<n[e].length;++s){if(0===s&&!r)continue;const o=i.shape[s],u=n[e][s];if(null!=u&&u!==o)throw new og(`Error when checking ${a}: expected ${t[e]} to have shape ${JSON.stringify(n[e])} but got array with shape ${JSON.stringify(i.shape)}.`)}}}var Wv=class extends Sv{constructor(e){super(e),this.isTraining=!1}summary(e,t,n=console.log){if(!this.built)throw new og("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");lv(this,e,t,n)}compile(e){if(null==e.loss&&(e.loss=[]),this.loss=e.loss,"string"==typeof e.optimizer)this.optimizer_=function(e){const t={Adagrad:()=>Ud.adagrad(.01),Adadelta:()=>Ud.adadelta(1,.95,ag()),Adam:()=>Ud.adam(.001,.9,.999,ag()),Adamax:()=>Ud.adamax(.002,.9,.999,ag(),0),RMSProp:()=>Ud.rmsprop(.001,.9,0,ag()),SGD:()=>Ud.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,e in t)return t[e]();throw new og("Unknown Optimizer "+e)}(e.optimizer),this.isOptimizerOwned=!0;else{if(!(e.optimizer instanceof Fd))throw new og("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=e.optimizer,this.isOptimizerOwned=!1}let t=[];if(Array.isArray(e.loss)||"string"==typeof e.loss||"function"==typeof e.loss)if(Array.isArray(e.loss)){if(e.loss.length!==this.outputs.length)throw new og(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);const n=e.loss;t=n.map(e=>Yb(e))}else{const n=Yb(e.loss);this.outputs.forEach(e=>{t.push(n)})}else{e.loss=e.loss;for(const t in e.loss)if(-1===this.outputNames.indexOf(t))throw new og(`Unknown entry in loss dictionary: "${t}". Only expected the following keys: ${this.outputNames}`);for(const n of this.outputNames)null==e.loss[n]&&console.warn(`Output "${n}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${n} during training`),t.push(Yb(e.loss[n]))}this.lossFunctions=t,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let e=0;e<this.outputs.length;++e){const t=this.internalOutputShapes[e],n=this.outputNames[e];this.feedOutputNames.push(n),this.feedOutputShapes.push(t),this.feedLossFns.push(this.lossFunctions[e])}const n=[];this.metrics=e.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Zg("loss",()=>{for(let e=0;e<this.outputs.length;++e){if(-1!==n.indexOf(e))continue;const t=this.lossFunctions[e];this.outputs.length>1&&(this.metricsTensors.push([t,e]),this.metricsNames.push(this.outputNames[e]+"_loss"))}});const r=function(e,t){if(null==e||Array.isArray(e)&&0===e.length)return t.map(e=>[]);let n;if("string"==typeof e||"function"==typeof e)n=[e];else{if(!Array.isArray(e)&&"object"!=typeof e)throw new TypeError("Type of metrics argument not understood. Expected an string,function, Array, or Object, found: "+e);n=e}if(Array.isArray(n))return t.map(e=>n);{const e=[];for(const r of t){let t=n.hasOwnProperty(r)?n[r]:[];Array.isArray(t)||(t=[t]),e.push(t)}return e}}(e.metrics,this.outputNames),a=(e,t,n)=>{this.outputNames.length>1&&(t=this.outputNames[e]+"_"+t),this.metricsNames.push(t),this.metricsTensors.push([n,e])};Zg("metric",()=>{for(let e=0;e<this.outputs.length;++e){if(-1!==n.indexOf(e))continue;(t=>{let n,r,s;for(const i of t){if("string"==typeof i&&-1!==["accuracy","acc","crossentropy","ce"].indexOf(i)){const t=this.internalOutputShapes[e];let a;1===t[t.length-1]||this.lossFunctions[e]===Hb?-1!==["accuracy","acc"].indexOf(i)?r=Xb:-1!==["crossentropy","ce"].indexOf(i)&&(r=tv):this.lossFunctions[e]===Gb?-1!==["accuracy","acc"].indexOf(i)?r=nv:-1!==["crossentropy","ce"].indexOf(i)&&(r=av):-1!==["accuracy","acc"].indexOf(i)?r=Qb:-1!==["crossentropy","ce"].indexOf(i)&&(r=rv),-1!==["accuracy","acc"].indexOf(i)?a="acc":-1!==["crossentropy","ce"].indexOf(i)&&(a="ce"),s=r,n=""+a}else{const e=iv(i);s=e,n=""+ov(i)}let t;Zg(n,()=>{t=s}),a(e,n,t)}})(r[e])}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){null!=this.collectedTrainableWeights&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(e,t,n={}){const r=null==n.batchSize?32:n.batchSize;_v(r);const a=this.standardizeUserDataXY(e,t,!0,r);try{const s=a[0].concat(a[1]);this.makeTestFunction();const i=this.testFunction;return dg(this.testLoop(i,s,r,n.verbose,n.steps))}finally{Pv(a[0],e),Pv(a[1],t)}}async evaluateDataset(e,t){return this.makeTestFunction(),async function(e,t,n){const r=null!=(n=n||{}).batches,a=e.testFunction;let s=[];if(n.verbose>0)throw new ug("Verbose mode is not implemented yet.");va.assert(!r||n.batches>0&&Number.isInteger(n.batches),()=>"Test loop expects `batches` to be a positive integer, but received "+JSON.stringify(n.batches));const i="function"==typeof t.next?t:await t.iterator();let o=0,u=0;for(;!r||u<n.batches;){const t=await i.next();if(s=ru(()=>{if(t.value){const{xs:n,ys:r}=Nv(e,t.value),i=n.concat(r),l=ru(()=>a(i));if(au(i),0===u)for(let e=0;e<l.length;++e)s.push(Ah(0));const c=i[0].shape[0];for(let e=0;e<l.length;++e){const t=l[e],n=s[e];s[e]=ru(()=>gu(s[e],vu(c,t))),u>0&&au(n)}au(l),o+=c,++u}return s}),t.done){r&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${n.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let e=0;e<s.length;++e){const t=s[e];s[e]=bu(s[e],o),au(t)}return dg(s)}(this,e,t)}checkNumSamples(e,t,n,r="steps"){let a;if(null!=n){if(a=null,null!=t)throw new og(`If ${r} is set, batchSize must be null or undefined.Got batchSize = ${t}`)}else{if(null==e)throw new og(`Either the input data should have a defined shape, or ${r} shoud be specified.`);a=Array.isArray(e)?e[0].shape[0]:e.shape[0]}return a}execute(e,t){if(Array.isArray(t)&&0===t.length)throw new og("`outputs` is an empty Array, which is not allowed.");const n=Array.isArray(t),r=n?t:[t],a=this.retrieveSymbolicTensors(r),s=new gv;if(e instanceof Ya&&(e=[e]),Array.isArray(e)){if(e.length!==this.inputs.length)throw new og(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let t=0;t<this.inputs.length;++t)s.add(this.inputs[t],e[t])}else for(const t of this.inputs){const n=e[t.name];if(null==n)throw new og("No value is provided for the model's input "+t.name);s.add(t,n)}const i=vv(a,s);return n?i:i[0]}retrieveSymbolicTensors(e){const t=cg(null,e.length);let n=e.length;for(const r of this.layers){const a=Array.isArray(r.output)?r.output:[r.output],s=a.map(e=>e.name);for(let r=0;r<e.length;++r){const i=s.indexOf(e[r]);if(-1!==i&&(t[r]=a[i],n--),0===n)break}if(0===n)break}if(n>0){const n=[];throw t.forEach((t,r)=>{null==t&&n.push(e[r])}),new og("Cannot find SymbolicTensors for output name(s): "+JSON.stringify(n))}return t}predictLoop(e,t=32,n=!1){return ru(()=>{const r=this.checkNumSamples(e);if(n)throw new ug("Verbose predictLoop() is not implemented yet.");const a=Dv(r,t),s=this.outputs.map(e=>[]);for(let t=0;t<a.length;++t){ru(()=>{const n=a[t][0],r=a[t][1],s=Fv(e,n,r),i=[];if(Array.isArray(s))for(let e=0;e<s.length;++e)i.push({key:this.inputs[e],value:s[e]});else i.push({key:this.inputs[0],value:s});const o=new gv(i);return vv(this.outputs,o)}).forEach((e,t)=>s[t].push(e))}return dg(s.map(e=>Xu(e,0)))})}predict(e,t={}){const n=Lv(e);Uv(n,this.inputNames,this.feedInputShapes,!1);try{const r=null==t.batchSize?32:t.batchSize;return _v(r),this.predictLoop(n,r)}finally{Pv(n,e)}}predictOnBatch(e){Uv(e,this.inputNames,this.feedInputShapes,!0);const t=(Array.isArray(e)?e[0]:e).shape[0];return this.predictLoop(e,t)}standardizeUserDataXY(e,t,n=!0,r){if(null==this.optimizer_)throw new ig("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");const a=[];for(let e=0;e<this.feedOutputShapes.length;++e){const t=this.feedOutputShapes[e];this.feedLossFns[e]===Gb?a.push(t.slice(0,t.length-1).concat([1])):a.push(t)}if(function(e,t,n){const r=wg(e.map(e=>e.shape[0]));r.sort();const a=wg(t.map(e=>e.shape[0]));if(a.sort(),r.length>1)throw new og("All input Tensors (x) should have the same number of samples. Got array shapes: "+JSON.stringify(e.map(e=>e.shape)));if(a.length>1)throw new og("All target Tensors (y) should have the same number of samples. Got array shapes: "+JSON.stringify(t.map(e=>e.shape)));if(r.length>0&&a.length>0&&!va.arraysEqual(r,a))throw new og(`Input Tensors should have the same number of samples as target Tensors. Found ${r[0]} input sample(s) and ${a[0]} target sample(s).`)}(e=zv(e,this.feedInputNames,this.feedInputShapes,!1,"input"),t=zv(t,this.feedOutputNames,a,!1,"target")),function(e,t,n){const r=[Ub,Hb,jb];for(let a=0;a<e.length;++a){const s=e[a],i=t[a],o=n[a];if(null!=i){if(i===jb&&1===s.shape[s.shape.length-1])throw new og(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(-1!==r.indexOf(i)){const e=s.shape.slice(1),t=o.slice(1);for(let n=0;n<e.length;++n){const r=e[n],a=t[n];if(null!=a&&r!==a)throw new og(`A target Tensor with shape ${s.shape} was passed for an output of shape ${o}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}(t,this.feedLossFns,this.feedOutputShapes),this.stateful&&null!=r&&r>0&&e[0].shape[0]%r!=0)throw new og(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${r}. Found: ${e[0].shape[0]} sample(s).`);return[e,t]}async standardizeUserData(e,t,n,r,a=!0,s){const[i,o]=this.standardizeUserDataXY(e,t,a,s);if(null!=n)throw new Error("sample weight is not supported yet.");let u=null;if(null!=r){const e=Iv(r,this.outputNames);u=[];for(let t=0;t<e.length;++t)u.push(await Av(o[t],null,e[t]))}return[i,o,u]}testLoop(e,t,n,r=0,a){return ru(()=>{const s=this.checkNumSamples(t,n,a,"steps"),i=[];if(r>0)throw new ug("Verbose mode is not implemented yet.");if(null!=a)throw new ug("steps mode in testLoop() is not implemented yet");{const r=Dv(s,n),a=Yh(iy(0,s));for(let n=0;n<r.length;++n){const s=r[n][0],o=r[n][1],u=ly(a,s,o-s),l=Mv(t,u),c=e(l);if(0===n)for(let e=0;e<c.length;++e)i.push(Ah(0));for(let e=0;e<c.length;++e){const t=c[e];i[e]=gu(i[e],vu(o-s,t))}}for(let e=0;e<i.length;++e)i[e]=bu(i[e],s)}return i})}getDedupedMetricsNames(){const e=this.metricsNames,t=[];for(let n=0;n<e.length;++n){const r=e[n];let a=r;if(pg(e,r)>1){a+="_"+pg(e.slice(0,n),r)}t.push(a)}return t}makeTrainFunction(){return e=>{const t=[],n=e.slice(0,this.inputs.length),r=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),a=e.slice(this.inputs.length+this.outputs.length,this.inputs.length+2*this.outputs.length),s=[],i=this.collectedTrainableWeights.map(e=>e.read());return[this.optimizer_.minimize(()=>{const e=[];for(let t=0;t<this.inputs.length;++t)e.push({key:this.inputs[t],value:n[t]});const i=new gv(e),o=vv(this.outputs,i,{training:!0});let u;for(let e=0;e<this.lossFunctions.length;++e){let n=(0,this.lossFunctions[e])(r[e],o[e]);null!=a[e]&&(n=Ev(n,a[e]));const s=Lc(n);t.push(s),u=0===e?n:gu(u,n)}for(let e=0;e<this.metricsTensors.length;++e){let n;if(this.outputs.length>1&&e<this.outputs.length)n=t[e];else{const t=this.metricsTensors[e][0],a=this.metricsTensors[e][1];n=Lc(t(r[a],o[a]))}su(n),s.push(n)}return u=Lc(u),this.calculateLosses().forEach(e=>{u=gu(u,e)}),u},!0,i)].concat(s)}}makeTestFunction(){this.testFunction=e=>ru(()=>{const t=[];let n;const r=e.slice(0,this.inputs.length),a=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=[];for(let e=0;e<this.inputs.length;++e)s.push({key:this.inputs[e],value:r[e]});const i=new gv(s),o=vv(this.outputs,i);for(let e=0;e<this.lossFunctions.length;++e){const r=this.lossFunctions[e],s=Lc(r(a[e],o[e]));n=0===e?s:gu(n,s),t.push(n)}for(let e=0;e<this.metricsTensors.length;++e){const n=this.metricsTensors[e][0],r=this.metricsTensors[e][1],s=Lc(n(a[r],o[r]));t.push(s)}return t})}async fit(e,t,n={}){return Ov(this,e,t,n)}async fitDataset(e,t){return Tv(this,e,t)}async trainOnBatch(e,t){const n=await this.standardizeUserData(e,t),r=n[0],a=n[1],s=this.makeTrainFunction()(r.concat(a)),i=[];for(const e of s){const t=await e.data();i.push(t[0])}return au(s),Pv(n[0],e),Pv(n[1],t),dg(i)}getNamedWeights(e){const t=[],n=null!=e&&e.trainableOnly,r=n?this.trainableWeights:this.weights,a=this.getWeights(n);for(let e=0;e<r.length;++e)n&&!r[e].trainable||t.push({name:r[e].originalName,tensor:a[e]});return t}set stopTraining(e){this.stopTraining_=e}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(e){this.optimizer_!==e&&(this.optimizer_=e,this.isOptimizerOwned=!1)}dispose(){const e=super.dispose();if(0===e.refCountAfterDispose&&null!=this.optimizer&&this.isOptimizerOwned){const t=tu().numTensors;this.optimizer_.dispose(),e.numDisposedVariables+=t-tu().numTensors}return e}getLossIdentifiers(){let e;if("string"==typeof this.loss)e=mg(this.loss);else if(Array.isArray(this.loss)){for(const e of this.loss)if("string"!=typeof e)throw new Error("Serialization of non-string loss is not supported.");e=this.loss.map(e=>mg(e))}else{const t=Object.keys(this.loss);e={};const n=this.loss;for(const r of t){if("string"!=typeof n[r])throw new Error("Serialization of non-string loss is not supported.");e[r]=mg(n[r])}}return e}getMetricIdentifiers(){if("string"==typeof this.metrics||"function"==typeof this.metrics)return[mg(ov(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(e=>mg(ov(e)));{const e={};for(const t in this.metrics)e[t]=mg(ov(this.metrics[t]));return e}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(e){if(null!=e.weighted_metrics)throw new Error("Loading weight_metrics is not supported yet.");if(null!=e.loss_weights)throw new Error("Loading loss_weights is not supported yet.");if(null!=e.sample_weight_mode)throw new Error("Loading sample_weight_mode is not supported yet.");const t=Bb(fv(e.optimizer_config));let n,r;if("string"==typeof e.loss)n=gg(e.loss);else if(Array.isArray(e.loss))n=e.loss.map(e=>gg(e));else if(null!=e.loss){n={};for(const t in e.loss)n[t]=gg(e.loss[t])}if(Array.isArray(e.metrics))r=e.metrics.map(e=>gg(e));else if(null!=e.metrics){r={};for(const t in e.metrics)r[t]=gg(e.metrics[t])}this.compile({loss:n,metrics:r,optimizer:t})}async save(e,t){if("string"==typeof e){const t=Ci.getSaveHandlers(e);if(0===t.length)throw new og(`Cannot find any save handlers for URL '${e}'`);if(t.length>1)throw new og(`Found more than one (${t.length}) save handlers for URL '${e}'`);e=t[0]}if(null==e.save)throw new og("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");const n=await Ci.encodeWeights(this.getNamedWeights(t)),r={modelTopology:this.toJSON(null,!1),format:"layers-model",generatedBy:"TensorFlow.js tfjs-layers v"+mv,convertedBy:null};if(null!=t&&t.includeOptimizer&&null!=this.optimizer){r.trainingConfig=this.getTrainingConfig();const e="optimizer",{data:t,specs:a}=await Ci.encodeWeights(await this.optimizer.getWeights(),e);n.specs.push(...a),n.data=Ci.concatenateArrayBuffers([n.data,t])}if(null!=this.userDefinedMetadata){const e=!0;uv(this.userDefinedMetadata,this.name,e),r.userDefinedMetadata=this.userDefinedMetadata}return r.weightData=n.data,r.weightSpecs=n.specs,e.save(r)}setUserDefinedMetadata(e){uv(e,this.name),this.userDefinedMetadata=e}getUserDefinedMetadata(){return this.userDefinedMetadata}};Wv.className="Model",Fo.registerClass(Wv);var Vv=class extends Wv{};async function jv(e,t){"modelTopology"in e||(e={modelTopology:e});let n=(e=e).modelTopology;null!=n.model_config&&(n=n.model_config);const r=Bb(fv(n),t);if(null!=e.weightsManifest){const t=await Ci.loadWeights(e.weightsManifest,e.pathPrefix,r.weights.map(e=>e.originalName)),n={};for(const e of r.weights)n[e.originalName]=t[e.originalName];r.loadWeights(n),au(t)}return r}async function Gv(e,t){if(null==t&&(t={}),"string"==typeof e){const n=Ci.getLoadHandlers(e,t);if(0===n.length)n.push(Ci.browserHTTPRequest(e,t));else if(n.length>1)throw new og(`Found more than one (${n.length}) load handlers for URL '${e}'`);e=n[0]}return async function(e,t,n){null==n&&(n={});if(null==e.load)throw new og("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");const r=await e.load();let a=r.modelTopology;null!=a.model_config&&(a=a.model_config);const s=null==n.strict||n.strict,i=null!=r.weightData&&null!=r.weightSpecs&&s,o=Bb(fv(a),t,i),u=r.trainingConfig;null!=u&&o.loadTrainingConfig(u);null!=r.userDefinedMetadata&&o.setUserDefinedMetadata(r.userDefinedMetadata);if(null!=r.weightData){if(null==r.weightSpecs)throw new og("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");const{modelWeights:e,optimizerWeights:t}=function(e,t){const n=Ci.decodeWeights(e,t),r={},a=[];return t.forEach(e=>{"optimizer"===e.group?a.push({name:e.name,tensor:n[e.name]}):r[e.name]=n[e.name]}),{modelWeights:r,optimizerWeights:a}}(r.weightData,r.weightSpecs);o.loadWeights(e,s),null!=o.optimizer&&t.length>0&&await o.optimizer.setWeights(t),au(e),au(t.map(e=>e.tensor))}return o}(e,void 0,t)}Vv.className="Functional",Fo.registerClass(Vv);var Hv=class extends Wv{constructor(e){if(super({inputs:[],outputs:[]}),e=e||{},this.trainable=!0,this.built=!1,this.name=null!=e.name?e.name:cb("sequential_"),null!=e.layers)for(const t of e.layers)this.add(t)}checkShape(e){if(e.inboundNodes[0].outputTensors[0].shape.some(e=>e<0))throw new og(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)}add(e){const t=e instanceof Hv||e instanceof Wv;let n;if(t){if(n=e,1!==n.outputs.length)throw new og("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(1!==n.inputs.length)throw new og("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(0===this.outputs.length){if(0===e.inboundNodes.length){if(null==e.batchInputShape)throw new og("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");const t=Cb({batchShape:e.batchInputShape,dtype:e.dtype,name:e.name+"_input"});e.apply(t)}if(t)this.outputs=n.outputs,this.inputs=n.inputs;else{if(1!==e.inboundNodes.length)throw new og(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${e.name} which has ${e.inboundNodes.length} pre-existing inbound connections.`);if(1!==e.inboundNodes[0].outputTensors.length)throw new og("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[e.inboundNodes[0].outputTensors[0]],this.inputs=function e(t,n,r){if((null==n||null!=r&&r>0)&&(n=t.sourceLayer,r=t.nodeIndex),0===n.inboundNodes.length)return[t];{const t=n.inboundNodes[r];if(0===t.inboundLayers.length)return t.inputTensors;{const n=[];for(let r=0;r<t.inboundLayers.length;r++){const a=e(t.inputTensors[r],t.inboundLayers[r],t.nodeIndices[r]);for(const e of a)-1===n.indexOf(e)&&n.push(e)}return n}}}(this.outputs[0])}this.inboundNodes=[],new kb({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:cg(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(e=>e.shape),outputShapes:this.outputs[0].shape})}else{const t=e.apply(this.outputs[0]);if(Array.isArray(t))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[t],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(e),this.built=!1}pop(){if(0===this.layers.length)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),0===this.layers.length)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{const e=this.layers.length-1;this.layers[e].outboundNodes=[],this.outputs=[this.layers[e].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(e,t){return null==this.model&&this.build(),this.model.call(e,t)}build(e){if(fb(e),0===this.inputs.length||0===this.outputs.length)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Wv({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(e,t,n=console.log){this.built||this.build(),super.summary(e,t,n)}setWeights(e){null==this.model&&this.build(),this.model.setWeights(e)}evaluate(e,t,n={}){if(!this.built)throw new ig("The model needs to be compiled before being used.");return this.model.evaluate(e,t,n)}async evaluateDataset(e,t){if(!this.built)throw new ig("The model needs to be compiled before being used.");return this.model.evaluateDataset(e,t)}predict(e,t={}){return null==this.model&&this.build(),this.model.predict(e,t)}predictOnBatch(e){return null==this.model&&this.build(),this.model.predictOnBatch(e)}compile(e){this.build(),this.model.compile(e),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return null==this.model?void 0:this.model.optimizer}set optimizer(e){this.model.optimizer=e}async fit(e,t,n={}){if(!this.built)throw new ig("The model needs to be compiled before being used.");return this.model.fit(e,t,n)}async fitDataset(e,t){if(!this.built)throw new ig("The model needs to be compiled before being used.");return this.model.fitDataset(e,t)}async trainOnBatch(e,t){return this.model.trainOnBatch(e,t)}static fromConfig(e,t,n={},r=!1){let a,s={};if(t instanceof Array){if(null==t[0].className||"Merge"===t[0].className)throw new og("Legacy serialization format not supported yet.");a=t}else va.assert(null!=t.layers,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),a=t.layers,delete t.layers,s=t;const i=new e(s);if(!(i instanceof Hv))throw new ug("Sequential.fromConfig called on non-Sequential input: "+i);for(const e of a){const t=Bb(e,void 0,r);r&&t.setFastWeightInitDuringBuild(!0),i.add(t)}return i}set stopTraining(e){if(null==this.model)throw new og("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=e}get stopTraining(){if(null==this.model)throw new og("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){const e=[];for(const t of this.layers){const n={};n.className=t.getClassName(),n.config=t.getConfig(),e.push(n)}return{name:this.name,layers:e}}};function qv(e){return new Wv(e)}function Kv(e){return new Hv(e)}function Yv(e,t){return null==t&&(t={}),Gv(e,t)}function Xv(e){return Cb(e)}function Qv(e,t){Pb.registerCallbackConstructor(e,t)}Hv.className="Sequential",Fo.registerClass(Hv);var Zv=class extends Fo.Serializable{getConfig(){return{}}},Jv=class extends Zv{apply(e,t=1){return function(e,t=1){if(1!==t)throw new ug(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Ll(e)}(e,t)}};Jv.className="elu",Fo.registerClass(Jv);var ex=class extends Zv{apply(e){return Eh(e)}};ex.className="selu",Fo.registerClass(ex);var tx=class extends Zv{apply(e){return gh(e)}};tx.className="relu",Fo.registerClass(tx);var nx=class extends Zv{apply(e){return ru(()=>Uc(6,gh(e)))}};nx.className="relu6",Fo.registerClass(nx);var rx=class extends Zv{apply(e){return e}};rx.className="linear",Fo.registerClass(rx);var ax=class extends Zv{apply(e){return Qu(e)}};ax.className="sigmoid",Fo.registerClass(ax);var sx=class extends Zv{apply(e){return function(e){return ru(()=>{const t=gu(.5,vu(.2,e));return cl(t,0,1)})}(e)}};sx.className="hardSigmoid",Fo.registerClass(sx);var ix=class extends Zv{apply(e){return dc(e)}};ix.className="softplus",Fo.registerClass(ix);var ox=class extends Zv{apply(e){return function(e){return ru(()=>bu(e,gu(xu(e),1)))}(e)}};ox.className="softsign",Fo.registerClass(ox);var ux=class extends Zv{apply(e){return Ju(e)}};ux.className="tanh",Fo.registerClass(ux);var lx=class extends Zv{apply(e,t=-1){return Lh(e,t)}};lx.className="softmax",Fo.registerClass(lx);var cx=class extends Zv{apply(e,t=-1){return bc(e,t)}};cx.className="logSoftmax",Fo.registerClass(cx);var hx=class extends Zv{apply(e,t=1){return ru(()=>vu(Qu(vu(e,t)),e))}};hx.className="swish",Fo.registerClass(hx);var px=class extends Zv{apply(e){return ru(()=>vu(e,Ju(dc(e))))}};function dx(e){return e.getClassName()}function fx(e,t={}){return vg(e,Fo.SerializationMap.getMap().classNameMap,t,"activation")}function mx(e){if(null==e){const e={className:"linear",config:{}};return fx(e)}if("string"==typeof e){const t={};return t.className=e,t.config={},fx(t)}return e instanceof Zv?e:fx(e)}function gx(e){if(null!=e&&"object"!=typeof e)throw new Error("Argument to L1L2 regularizer's constructor is expected to be an object, but received: "+e)}px.className="mish",Fo.registerClass(px);var yx=class extends Fo.Serializable{},bx=class extends yx{constructor(e){super(),gx(e),this.l1=null==e||null==e.l1?.01:e.l1,this.l2=null==e||null==e.l2?.01:e.l2,this.hasL1=0!==this.l1,this.hasL2=0!==this.l2}apply(e){return ru(()=>{let t=Pc([1]);return this.hasL1&&(t=gu(t,yc(vu(this.l1,xu(e))))),this.hasL2&&(t=gu(t,yc(vu(this.l2,by(e))))),qu(t,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(e,t){return new e({l1:t.l1,l2:t.l2})}};bx.className="L1L2",Fo.registerClass(bx);var vx={l1l2:"L1L2"};function xx(e){return bg(e)}function wx(e,t={}){return vg(e,Fo.SerializationMap.getMap().classNameMap,t,"regularizer")}function kx(e){if(null==e)return null;if("string"==typeof e){return wx({className:e in vx?vx[e]:e,config:{}})}return e instanceof yx?e:wx(e)}var Sx=class extends Ib{constructor(e){super(null==e?{}:e),this.supportsMasking=!0,null!=e&&(this.maxValue=e.maxValue)}call(e,t){e=db(e);let n=gh(e);return null!=this.maxValue&&(n=cl(n,0,this.maxValue)),n}computeOutputShape(e){return e}getConfig(){const e={maxValue:this.maxValue},t=super.getConfig();return Object.assign(e,t),e}};Sx.className="ReLU",Fo.registerClass(Sx);var Ix=class extends Ib{constructor(e){super(null==e?{}:e),this.DEFAULT_ALPHA=.3,null==e&&(e={}),this.alpha=null==e.alpha?this.DEFAULT_ALPHA:e.alpha}call(e,t){const n=db(e);return Zl(n,this.alpha)}computeOutputShape(e){return e}getConfig(){const e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Ix.className="LeakyReLU",Fo.registerClass(Ix);var Ax=class extends Ib{constructor(e){if(super(null==e?{}:e),this.DEFAULT_ALPHA_INITIALIZER="zeros",null==e&&(e={}),this.supportsMasking=!0,this.alphaInitializer=jy(e.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=kx(e.alphaRegularizer),this.alphaConstraint=Lg(e.alphaConstraint),null==e.sharedAxes)this.sharedAxes=null;else if(Array.isArray(e.sharedAxes))this.sharedAxes=e.sharedAxes;else{if("number"!=typeof e.sharedAxes)throw new og("Expected sharedAxes to be a number or an array of numbers, but got "+e.sharedAxes);this.sharedAxes=[e.sharedAxes]}}build(e){const t=(e=fb(e)).slice(1);if(null!=this.sharedAxes)for(const e of this.sharedAxes)t[e-1]=1;this.alpha=this.addWeight("alpha",t,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);const n={};if(null!=this.sharedAxes)for(let t=1;t<e.length;++t)n[t]=e[t];this.inputSpec=[new vb({ndim:e.length,axes:n})],this.built=!0}call(e,t){return e=db(e),sh(e,this.alpha.read())}getConfig(){const e={alphaInitializer:Vy(this.alphaInitializer),alphaRegularizer:xx(this.alphaRegularizer),alphaConstraint:Dg(this.alphaConstraint),sharedAxes:this.sharedAxes},t=super.getConfig();return Object.assign(e,t),e}};Ax.className="PReLU",Fo.registerClass(Ax);var Ex=class extends Ib{constructor(e){if(super(null==e?{}:e),this.DEFAULT_ALPHA=1,null==e&&(e={}),null!=e.alpha&&e.alpha!==this.DEFAULT_ALPHA)throw new ug(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);this.alpha=null==e.alpha?this.DEFAULT_ALPHA:e.alpha}call(e,t){const n=db(e);return Ll(n)}computeOutputShape(e){return e}getConfig(){const e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Ex.className="ELU",Fo.registerClass(Ex);var Nx=class extends Ib{constructor(e){super(null==e?{}:e),this.DEFAULT_THETA=1,null==e&&(e={}),this.theta=null==e.theta?this.DEFAULT_THETA:e.theta}call(e,t){const n=db(e);return vu(n,Ai(Hl(n,this.theta),"float32"))}computeOutputShape(e){return e}getConfig(){const e={theta:this.theta},t=super.getConfig();return Object.assign(e,t),e}};Nx.className="ThresholdedReLU",Fo.registerClass(Nx);var Cx=class extends Ib{constructor(e){super(null==e?{}:e),this.DEFAULT_AXIS=1,null==e&&(e={}),this.softmax=(new lx).apply,this.axis=null==e.axis?this.DEFAULT_AXIS:e.axis}call(e,t){const n=db(e);return this.softmax(n,this.axis)}computeOutputShape(e){return e}getConfig(){const e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};function Tx(e,t,n){if("number"==typeof e)return cg(e,t);if(e.length!==t)throw new og(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);for(let a=0;a<t;++a){const s=e[a];if((r=s)!==parseInt(r.toString(),10))throw new og(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${s}`)}return e;var r}function Rx(e,t,n,r,a=1){if(null==e)return e;let s;return s="same"===n?e:e-(t+(t-1)*(a-1))+1,Math.floor((s+r-1)/r)}function _x(e,t,n,r){if(null==e)return null;if("valid"===r)e=e*t+sy([n-t,0]);else{if("same"!==r)throw new og(`Unsupport padding mode: ${r}.`);e*=t}return e}function Fx(e,t){return ru(()=>(Kg(t),"channelsFirst"===t?qi(e,[0,2,3,1]):e))}function Mx(e,t){return ru(()=>(Kg(t),"channelsFirst"===t?qi(e,[0,2,3,4,1]):e))}function Dx(e,t,n,r=[1,1],a="valid",s,i,o=null){return ru(()=>{if(null==s&&(s="channelsLast"),Kg(s),3!==e.rank&&4!==e.rank)throw new og(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);if(3!==t.rank&&4!==t.rank)throw new og(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);let u=Fx(e,s);if("causal"===a)throw new ug("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return u=bp.conv2d({x:u,filter:t,strides:r,pad:"same"===a?"same":"valid",dilations:i,dataFormat:"NHWC",bias:n,activation:o}),"channelsFirst"===s&&(u=qi(u,[0,3,1,2])),u})}Cx.className="Softmax",Fo.registerClass(Cx);var Ox=class extends Ib{constructor(e,t){if(super(t),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",Ox.verifyArgs(t),this.rank=e,Ag(this.rank,"rank"),1!==this.rank&&2!==this.rank&&3!==this.rank)throw new ug(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=Tx(t.kernelSize,e,"kernelSize"),this.strides=Tx(null==t.strides?1:t.strides,e,"strides"),this.padding=null==t.padding?"valid":t.padding,Yg(this.padding),this.dataFormat=null==t.dataFormat?"channelsLast":t.dataFormat,Kg(this.dataFormat),this.activation=mx(t.activation),this.useBias=null==t.useBias||t.useBias,this.biasInitializer=jy(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=Lg(t.biasConstraint),this.biasRegularizer=kx(t.biasRegularizer),this.activityRegularizer=kx(t.activityRegularizer),this.dilationRate=Tx(null==t.dilationRate?1:t.dilationRate,e,"dilationRate"),1===this.rank&&Array.isArray(this.dilationRate)&&1!==this.dilationRate.length)throw new og("dilationRate must be a number or an array of a single number for 1D convolution, but received "+JSON.stringify(this.dilationRate));if(2===this.rank){if("number"==typeof this.dilationRate)this.dilationRate=[this.dilationRate,this.dilationRate];else if(2!==this.dilationRate.length)throw new og("dilationRate must be a number or array of two numbers for 2D convolution, but received "+JSON.stringify(this.dilationRate))}else if(3===this.rank)if("number"==typeof this.dilationRate)this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(3!==this.dilationRate.length)throw new og("dilationRate must be a number or array of three numbers for 3D convolution, but received "+JSON.stringify(this.dilationRate))}static verifyArgs(e){if(hg("kernelSize"in e,"required key 'kernelSize' not in config"),"number"!=typeof e.kernelSize&&!Ig(e.kernelSize,"number",1,3))throw new og(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)}getConfig(){const e={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:dx(this.activation),useBias:this.useBias,biasInitializer:Vy(this.biasInitializer),biasRegularizer:xx(this.biasRegularizer),activityRegularizer:xx(this.activityRegularizer),biasConstraint:Dg(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}},Lx=class extends Ox{constructor(e,t){super(e,t),this.kernel=null,Lx.verifyArgs(t),this.filters=t.filters,Ag(this.filters,"filters"),this.kernelInitializer=jy(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=Lg(t.kernelConstraint),this.kernelRegularizer=kx(t.kernelRegularizer)}build(e){e=fb(e);const t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t])throw new og("The channel dimension of the input should be defined. Found "+e[t]);const n=e[t],r=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight("kernel",r,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[t]:n}}],this.built=!0}call(e,t){return ru(()=>{let t;e=db(e);const n=null==this.bias?null:this.bias.read(),r=Eg(this.activation.getClassName());if(null!=r&&2===this.rank)t=Dx(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate,r);else{if(1===this.rank)t=function(e,t,n,r=1,a="valid",s,i=1){return ru(()=>{if(null==s&&(s="channelsLast"),Kg(s),3!==e.shape.length)throw new og(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);if(3!==t.shape.length)throw new og(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(null!=n&&1!==n.shape.length)throw new og(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if("channelsFirst"===s&&(e=qi(e,[0,2,1])),"causal"===a)throw new ug("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let o=gl(e,t,r,"same"===a?"same":"valid","NWC",i);return null!=n&&(o=xy(o,n)),o})}(e,this.kernel.read(),n,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(2===this.rank)t=Dx(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate);else{if(3!==this.rank)throw new ug("convolutions greater than 3D are not implemented yet.");t=function(e,t,n,r=[1,1,1],a="valid",s,i){return ru(()=>{if(null==s&&(s="channelsLast"),Kg(s),4!==e.rank&&5!==e.rank)throw new og(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);if(4!==t.rank&&5!==t.rank)throw new og(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);let o=Mx(e,s);if("causal"===a)throw new ug("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return o=vl(o,t,r,"same"===a?"same":"valid","NDHWC",i),null!=n&&(o=xy(o,n)),"channelsFirst"===s&&(o=qi(o,[0,4,1,2,3])),o})}(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate)}null!=this.activation&&(t=this.activation.apply(t))}return t})}computeOutputShape(e){e=fb(e);const t=[],n="channelsLast"===this.dataFormat?e.slice(1,e.length-1):e.slice(2);for(let e=0;e<n.length;++e){const r=Rx(n[e],this.kernelSize[e],this.padding,this.strides[e],"number"==typeof this.dilationRate?this.dilationRate:this.dilationRate[e]);t.push(r)}let r=[e[0]];return"channelsLast"===this.dataFormat?(r=r.concat(t),r.push(this.filters)):(r.push(this.filters),r=r.concat(t)),r}getConfig(){const e={filters:this.filters,kernelInitializer:Vy(this.kernelInitializer),kernelRegularizer:xx(this.kernelRegularizer),kernelConstraint:Dg(this.kernelConstraint)},t=super.getConfig();return Object.assign(e,t),e}static verifyArgs(e){if(!("filters"in e)||"number"!=typeof e.filters||e.filters<1)throw new og("Convolution layer expected config.filters to be a 'number' > 0 but got "+JSON.stringify(e.filters))}},Px=class extends Lx{constructor(e){super(2,e),Px.verifyArgs(e)}getConfig(){const e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if("number"!=typeof e.kernelSize&&!Ig(e.kernelSize,"number",1,2))throw new og(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)}};Px.className="Conv2D",Fo.registerClass(Px);var $x=class extends Lx{constructor(e){super(3,e),$x.verifyArgs(e)}getConfig(){const e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if("number"!=typeof e.kernelSize&&(!Array.isArray(e.kernelSize)||1!==e.kernelSize.length&&3!==e.kernelSize.length))throw new og(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)}};$x.className="Conv3D",Fo.registerClass($x);var Bx=class extends Px{constructor(e){if(super(e),this.inputSpec=[new vb({ndim:4})],"same"!==this.padding&&"valid"!==this.padding)throw new og("Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode "+this.padding)}build(e){if(4!==(e=fb(e)).length)throw new og("Input should have rank 4; Received input shape: "+JSON.stringify(e));const t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t])throw new og("The channel dimension of the inputs should be defined. Found `None`.");const n=e[t],r=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",r,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new vb({ndim:4,axes:{[t]:n}})],this.built=!0}call(e,t){return ru(()=>{let t=db(e);if(4!==t.shape.length)throw new og("Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-"+t.shape.length);const n=t.shape,r=n[0];let a,s;"channelsFirst"===this.dataFormat?(a=2,s=3):(a=1,s=2);const i=n[a],o=n[s],u=this.kernelSize[0],l=this.kernelSize[1],c=this.strides[0],h=this.strides[1],p=[r,_x(i,c,u,this.padding),_x(o,h,l,this.padding),this.filters];"channelsLast"!==this.dataFormat&&(t=qi(t,[0,2,3,1]));let d=bl(t,this.kernel.read(),p,this.strides,this.padding);return"channelsLast"!==this.dataFormat&&(d=qi(d,[0,3,1,2])),null!=this.bias&&(d=xy(d,this.bias.read(),this.dataFormat)),null!=this.activation&&(d=this.activation.apply(d)),d})}computeOutputShape(e){const t=(e=fb(e)).slice();let n,r,a;"channelsFirst"===this.dataFormat?(n=1,r=2,a=3):(n=3,r=1,a=2);const s=this.kernelSize[0],i=this.kernelSize[1],o=this.strides[0],u=this.strides[1];return t[n]=this.filters,t[r]=_x(t[r],o,s,this.padding),t[a]=_x(t[a],u,i,this.padding),t}getConfig(){const e=super.getConfig();return delete e.dilationRate,e}};Bx.className="Conv2DTranspose",Fo.registerClass(Bx);var zx=class extends $x{constructor(e){if(super(e),this.inputSpec=[new vb({ndim:5})],"same"!==this.padding&&"valid"!==this.padding)throw new og("Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode "+this.padding)}build(e){if(5!==(e=fb(e)).length)throw new og("Input should have rank 5; Received input shape: "+JSON.stringify(e));const t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t])throw new og("The channel dimension of the inputs should be defined. Found `None`.");const n=e[t],r=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",r,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new vb({ndim:5,axes:{[t]:n}})],this.built=!0}call(e,t){return ru(()=>{let t=db(e);if(5!==t.shape.length)throw new og("Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-"+t.shape.length);const n=t.shape,r=n[0];let a,s,i;"channelsFirst"===this.dataFormat?(i=2,a=3,s=4):(i=1,a=2,s=3);const o=n[i],u=n[a],l=n[s],c=this.kernelSize[0],h=this.kernelSize[1],p=this.kernelSize[2],d=this.strides[0],f=this.strides[1],m=this.strides[2],g=[r,_x(o,d,c,this.padding),_x(u,f,h,this.padding),_x(l,m,p,this.padding),this.filters];"channelsLast"!==this.dataFormat&&(t=qi(t,[0,2,3,4,1]));let y=wl(t,this.kernel.read(),g,this.strides,this.padding);return"channelsLast"!==this.dataFormat&&(y=qi(y,[0,4,1,2,3])),null!==this.bias&&(y=xy(y,this.bias.read(),this.dataFormat)),null!==this.activation&&(y=this.activation.apply(y)),y})}computeOutputShape(e){const t=(e=fb(e)).slice();let n,r,a,s;"channelsFirst"===this.dataFormat?(n=1,r=2,a=3,s=4):(n=4,r=1,a=2,s=3);const i=this.kernelSize[0],o=this.kernelSize[1],u=this.kernelSize[2],l=this.strides[0],c=this.strides[1],h=this.strides[2];return t[n]=this.filters,t[r]=_x(t[r],l,i,this.padding),t[a]=_x(t[a],c,o,this.padding),t[s]=_x(t[s],h,u,this.padding),t}getConfig(){const e=super.getConfig();return delete e.dilationRate,e}};zx.className="Conv3DTranspose",Fo.registerClass(zx);var Ux=class extends Lx{constructor(e,t){if(super(e,t),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,null==t.filters)throw new og("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(null!=t.kernelInitializer||null!=t.kernelRegularizer||null!=t.kernelConstraint)throw new og("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(null!=t.padding&&"same"!==t.padding&&"valid"!==t.padding)throw new og(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);this.depthMultiplier=null==t.depthMultiplier?1:t.depthMultiplier,this.depthwiseInitializer=jy(t.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=kx(t.depthwiseRegularizer),this.depthwiseConstraint=Lg(t.depthwiseConstraint),this.pointwiseInitializer=jy(t.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=kx(t.pointwiseRegularizer),this.pointwiseConstraint=Lg(t.pointwiseConstraint)}build(e){if((e=fb(e)).length<this.rank+2)throw new og(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);const t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t]||e[t]<0)throw new og("The channel dimension of the inputs should be defined, but found "+JSON.stringify(e[t]));const n=e[t],r=this.kernelSize.concat([n,this.depthMultiplier]),a=[];for(let e=0;e<this.rank;++e)a.push(1);a.push(n*this.depthMultiplier,this.filters);this.depthwiseKernel=this.addWeight("depthwise_kernel",r,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",a,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,!0,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.inputSpec=[new vb({ndim:this.rank+2,axes:{[t]:n}})],this.built=!0}call(e,t){return ru(()=>{let t;if(e=db(e),1===this.rank)throw new ug("1D separable convolution is not implemented yet.");return 2===this.rank&&("channelsFirst"===this.dataFormat&&(e=qi(e,[0,2,3,1])),t=Nh(e,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(t=xy(t,this.bias.read(),this.dataFormat)),null!=this.activation&&(t=this.activation.apply(t)),"channelsFirst"===this.dataFormat&&(t=qi(t,[0,3,1,2])),t})}getConfig(){const e=super.getConfig();return delete e.rank,delete e.kernelInitializer,delete e.kernelRegularizer,delete e.kernelConstraint,e.depthwiseInitializer=Vy(this.depthwiseInitializer),e.pointwiseInitializer=Vy(this.pointwiseInitializer),e.depthwiseRegularizer=xx(this.depthwiseRegularizer),e.pointwiseRegularizer=xx(this.pointwiseRegularizer),e.depthwiseConstraint=Dg(this.depthwiseConstraint),e.pointwiseConstraint=Dg(this.pointwiseConstraint),e}};Ux.className="SeparableConv";var Wx=class extends Ux{constructor(e){super(2,e)}};Wx.className="SeparableConv2D",Fo.registerClass(Wx);var Vx=class extends Lx{constructor(e){super(1,e),Vx.verifyArgs(e),this.inputSpec=[{ndim:3}]}getConfig(){const e=super.getConfig();return delete e.rank,delete e.dataFormat,e}static verifyArgs(e){if("number"!=typeof e.kernelSize&&!Ig(e.kernelSize,"number",1,1))throw new og(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)}};Vx.className="Conv1D",Fo.registerClass(Vx);var jx=class extends Ib{constructor(e){super(e),"number"==typeof e.cropping?this.cropping=[[e.cropping,e.cropping],[e.cropping,e.cropping]]:"number"==typeof e.cropping[0]?this.cropping=[[e.cropping[0],e.cropping[0]],[e.cropping[1],e.cropping[1]]]:this.cropping=e.cropping,this.dataFormat=void 0===e.dataFormat?"channelsLast":e.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(e){return"channelsFirst"===this.dataFormat?[e[0],e[1],e[2]-this.cropping[0][0]-this.cropping[0][1],e[3]-this.cropping[1][0]-this.cropping[1][1]]:[e[0],e[1]-this.cropping[0][0]-this.cropping[0][1],e[2]-this.cropping[1][0]-this.cropping[1][1],e[3]]}call(e,t){return ru(()=>{if(e=db(e),"channelsLast"===this.dataFormat){const t=hy(e,this.cropping[0][0],e.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return hy(t,this.cropping[1][0],e.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}{const t=hy(e,this.cropping[0][0],e.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return hy(t,this.cropping[1][0],e.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){const e={cropping:this.cropping,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};jx.className="Cropping2D",Fo.registerClass(jx);var Gx=class extends Ib{constructor(e){var t;super(e),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=null==e.size?this.DEFAULT_SIZE:e.size,this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Kg(this.dataFormat),this.interpolation=null==e.interpolation?"nearest":e.interpolation,t=this.interpolation,Sg(Vg,"InterpolationFormat",t)}computeOutputShape(e){if("channelsFirst"===this.dataFormat){const t=null==e[2]?null:this.size[0]*e[2],n=null==e[3]?null:this.size[1]*e[3];return[e[0],e[1],t,n]}{const t=null==e[1]?null:this.size[0]*e[1],n=null==e[2]?null:this.size[1]*e[2];return[e[0],t,n,e[3]]}}call(e,t){return ru(()=>{let t=db(e);const n=t.shape;if("channelsFirst"===this.dataFormat){t=qi(t,[0,2,3,1]);const e=this.size[0]*n[2],r=this.size[1]*n[3],a="nearest"===this.interpolation?Nd.resizeNearestNeighbor(t,[e,r]):Nd.resizeBilinear(t,[e,r]);return qi(a,[0,3,1,2])}{const e=this.size[0]*n[1],r=this.size[1]*n[2];return"nearest"===this.interpolation?Nd.resizeNearestNeighbor(t,[e,r]):Nd.resizeBilinear(t,[e,r])}})}getConfig(){const e={size:this.size,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};Gx.className="UpSampling2D",Fo.registerClass(Gx);var Hx=class extends Ox{constructor(e){super(2,e),this.depthwiseKernel=null,this.depthMultiplier=null==e.depthMultiplier?1:e.depthMultiplier,this.depthwiseInitializer=jy(e.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=Lg(e.depthwiseConstraint),this.depthwiseRegularizer=kx(e.depthwiseRegularizer)}build(e){if((e=fb(e)).length<4)throw new og(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);const t="channelsFirst"===this.dataFormat?1:3;if(null==e[t]||e[t]<0)throw new og(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);const n=e[t],r=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",r,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return ru(()=>{let t=function(e,t,n=[1,1],r="valid",a,s){return ru(()=>{null==a&&(a="channelsLast"),Kg(a);let i=Fx(e,a);if(4!==e.rank)throw new og(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);if(4!==t.rank)throw new og(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=Nl(i,t,n,"same"===r?"same":"valid","NHWC",s),"channelsFirst"===a&&(i=qi(i,[0,3,1,2])),i})}(e=db(e),this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(t=xy(t,this.bias.read(),this.dataFormat)),null!=this.activation&&(t=this.activation.apply(t)),t})}computeOutputShape(e){e=fb(e);const t="channelsFirst"===this.dataFormat?e[2]:e[1],n="channelsFirst"===this.dataFormat?e[3]:e[2],r="channelsFirst"===this.dataFormat?e[1]*this.depthMultiplier:e[3]*this.depthMultiplier,a=Rx(t,this.kernelSize[0],this.padding,this.strides[0]),s=Rx(n,this.kernelSize[1],this.padding,this.strides[1]);return"channelsFirst"===this.dataFormat?[e[0],r,a,s]:[e[0],a,s,r]}getConfig(){const e=super.getConfig();return e.depthMultiplier=this.depthMultiplier,e.depthwiseInitializer=Vy(this.depthwiseInitializer),e.depthwiseRegularizer=xx(this.depthwiseRegularizer),e.depthwiseConstraint=Dg(this.depthwiseRegularizer),e}};function qx(e,t,n,r){if(Array.isArray(e)){if(null!=t||null!=n)throw new og("When inputs is an array, neither initialState or constants should be provided");null!=r&&(n=e.slice(e.length-r,e.length),e=e.slice(0,e.length-r)),e.length>1&&(t=e.slice(1,e.length)),e=e[0]}function a(e){return null==e||Array.isArray(e)?e:[e]}return{inputs:e,initialState:t=a(t),constants:n=a(n)}}function Kx(e,t,n,r=!1,a,s,i=!1,o=!1){return ru(()=>{const u=t.shape.length;if(u<3)throw new og(`Input should be at least 3D, but is ${u}D.`);const l=[1,0].concat(iy(2,u));if(t=qi(t,l),null!=s)throw new ug("The rnn() functoin of the deeplearn.js backend does not support constants yet.");i&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),null!=a&&((a=Ai(Ai(a,"bool"),"float32")).rank===u-1&&(a=Bl(a,-1)),a=qi(a,l)),r&&(t=bh(t,0),null!=a&&(a=bh(a,0)));const c=[];let h,p=n;const d=t.shape[0],f=ap(t);let m,g;null!=a&&(m=ap(a));for(let t=0;t<d;++t){const n=f[t],r=ru(()=>e(n,p));if(null==a)h=r[0],p=r[1];else{const e=ru(()=>{const e=m[t],n=gc(Yc(e),e);return{output:gu(vu(r[0],e),vu(p[0],n)),newStates:p.map((t,a)=>gu(vu(r[1][a],e),vu(t,n)))}});h=e.output,p=e.newStates}o&&c.push(h)}if(o){g=Gh(c,1)}return[h,g,p]})}Hx.className="DepthwiseConv2D",Fo.registerClass(Hx);var Yx=class extends Ib{constructor(e){let t;if(super(e),null==e.cell)throw new og("cell property is missing for the constructor of RNN.");if(t=Array.isArray(e.cell)?new rw({cells:e.cell}):e.cell,null==t.stateSize)throw new og("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=t,this.returnSequences=null!=e.returnSequences&&e.returnSequences,this.returnState=null!=e.returnState&&e.returnState,this.goBackwards=null!=e.goBackwards&&e.goBackwards,this._stateful=null!=e.stateful&&e.stateful,this.unroll=null!=e.unroll&&e.unroll,this.supportsMasking=!0,this.inputSpec=[new vb({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(null==this.states_){return iy(0,Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1).map(e=>null)}return this.states_}setStates(e){this.states_=e}computeOutputShape(e){hb(e)&&(e=e[0]),e=e;let t=this.cell.stateSize;Array.isArray(t)||(t=[t]);const n=t[0];let r;if(r=this.returnSequences?[e[0],e[1],n]:[e[0],n],this.returnState){const n=[];for(const r of t)n.push([e[0],r]);return[r].concat(n)}return r}computeMask(e,t){return ru(()=>{Array.isArray(t)&&(t=t[0]);const e=this.returnSequences?t:null;if(this.returnState){const t=this.states.map(e=>null);return[e].concat(t)}return e})}get states(){if(null==this.states_){const e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,t=[];for(let n=0;n<e;++n)t.push(null);return t}return this.states_}set states(e){this.states_=e}build(e){if(null!=this.numConstants)throw new ug("Constants support is not implemented in RNN yet.");hb(e)&&(e=e[0]),e=e;const t=this.stateful?e[0]:null,n=e.slice(2);this.inputSpec[0]=new vb({shape:[t,null,...n]});const r=[e[0]].concat(e.slice(2));let a;if(this.cell.build(r),a=Array.isArray(this.cell.stateSize)?this.cell.stateSize:[this.cell.stateSize],null!=this.stateSpec){if(!va.arraysEqual(this.stateSpec.map(e=>e.shape[e.shape.length-1]),a))throw new og(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=a.map(e=>new vb({shape:[null,e]}));this.stateful&&this.resetStates()}resetStates(e,t=!1){ru(()=>{if(!this.stateful)throw new sg("Cannot call resetStates() on an RNN Layer that is not stateful.");const n=this.inputSpec[0].shape[0];if(null==n)throw new og("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(null==this.states_)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(e=>Pc([n,e])):this.states_=[Pc([n,this.cell.stateSize])];else if(null==e)au(this.states_),null!=this.keptStates&&(au(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(e=>Pc([n,e])):this.states_[0]=Pc([n,this.cell.stateSize]);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new og(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);!0===t?this.keptStates.push(this.states_.slice()):au(this.states_);for(let t=0;t<this.states_.length;++t){const r=e[t],a=Array.isArray(this.cell.stateSize)?this.cell.stateSize[t]:this.cell.stateSize,s=[n,a];if(!va.arraysEqual(r.shape,s))throw new og(`State ${t} is incompatible with layer ${this.name}: expected shape=${s}, received shape=${r.shape}`);this.states_[t]=r}}this.states_=this.states_.map(e=>su(e.clone()))})}apply(e,t){let n=null==t?null:t.initialState,r=null==t?null:t.constants;null==t&&(t={});const a=qx(e,n,r,this.numConstants);e=a.inputs,n=a.initialState,r=a.constants;let s=[],i=[];if(null!=n){t.initialState=n,s=s.concat(n),this.stateSpec=[];for(const e of n)this.stateSpec.push(new vb({shape:e.shape}));i=i.concat(this.stateSpec)}null!=r&&(t.constants=r,s=s.concat(r),this.numConstants=r.length);if(s[0]instanceof xb){const n=[e].concat(s),r=this.inputSpec.concat(i),a=this.inputSpec;this.inputSpec=r;const o=super.apply(n,t);return this.inputSpec=a,o}return super.apply(e,t)}call(e,t){return ru(()=>{const n=null==t?null:t.mask,r=null==t?null:t.training;let a=null==t?null:t.initialState;e=db(e),null==a&&(a=this.stateful?this.states_:this.getInitialState(e));const s=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(a.length!==s)throw new og(`RNN Layer has ${s} state(s) but was passed ${a.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");const i={training:r},o=Kx((e,t)=>{const n=this.cell.call([e].concat(t),i);return[n[0],n.slice(1)]},e,a,this.goBackwards,n,null,this.unroll,this.returnSequences),u=o[0],l=o[1],c=o[2];this.stateful&&this.resetStates(c,r);const h=this.returnSequences?l:u;return this.returnState?[h].concat(c):h})}getInitialState(e){return ru(()=>{let t=Pc(e.shape);return t=yc(t,[1,2]),t=uy(t),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(e=>e>1?fy(t,[1,e]):t):this.cell.stateSize>1?[fy(t,[1,this.cell.stateSize])]:[t]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),null!=this.cell&&this.cell.setFastWeightInitDuringBuild(e)}getConfig(){const e=super.getConfig(),t={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};null!=this.numConstants&&(t.numConstants=this.numConstants);const n=this.cell.getConfig();return this.getClassName()===Yx.className&&(t.cell={className:this.cell.getClassName(),config:n}),Object.assign({},n,e,t)}static fromConfig(e,t,n={}){const r=Bb(t.cell,n);return new e(Object.assign(t,{cell:r}))}};Yx.className="RNN",Fo.registerClass(Yx);var Xx=class extends Ib{},Qx=class extends Xx{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Ag(this.units,"units"),this.activation=mx(null==e.activation?this.DEFAULT_ACTIVATION:e.activation),this.useBias=null==e.useBias||e.useBias,this.kernelInitializer=jy(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=jy(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=jy(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=kx(e.kernelRegularizer),this.recurrentRegularizer=kx(e.recurrentRegularizer),this.biasRegularizer=kx(e.biasRegularizer),this.kernelConstraint=Lg(e.kernelConstraint),this.recurrentConstraint=Lg(e.recurrentConstraint),this.biasConstraint=Lg(e.biasConstraint),this.dropout=ay([1,sy([0,null==e.dropout?0:e.dropout])]),this.recurrentDropout=ay([1,sy([0,null==e.recurrentDropout?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=fb(e),this.kernel=this.addWeight("kernel",[e[e.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return ru(()=>{if(2!==(e=e).length)throw new og(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);let n=e[1];e=e[0];const r=null!=t.training&&t.training;let a;0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=aw({ones:()=>Yc(e),rate:this.dropout,training:r,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=aw({ones:()=>Yc(n),rate:this.recurrentDropout,training:r,dropoutFunc:this.dropoutFunc}));const s=this.dropoutMask,i=this.recurrentDropoutMask;a=gy(null!=s?vu(e,s):e,this.kernel.read()),null!=this.bias&&(a=xy(a,this.bias.read())),null!=i&&(n=vu(n,i));let o=gu(a,gy(n,this.recurrentKernel.read()));return null!=this.activation&&(o=this.activation.apply(o)),[o,o]})}getConfig(){const e=super.getConfig(),t={units:this.units,activation:dx(this.activation),useBias:this.useBias,kernelInitializer:Vy(this.kernelInitializer),recurrentInitializer:Vy(this.recurrentInitializer),biasInitializer:Vy(this.biasInitializer),kernelRegularizer:xx(this.kernelRegularizer),recurrentRegularizer:xx(this.recurrentRegularizer),biasRegularizer:xx(this.biasRegularizer),activityRegularizer:xx(this.activityRegularizer),kernelConstraint:Dg(this.kernelConstraint),recurrentConstraint:Dg(this.recurrentConstraint),biasConstraint:Dg(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign({},e,t)}};Qx.className="SimpleRNNCell",Fo.registerClass(Qx);var Zx=class extends Yx{constructor(e){e.cell=new Qx(e),super(e)}call(e,t){return ru(()=>{null!=this.cell.dropoutMask&&(au(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(au(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);const n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}static fromConfig(e,t){return new e(t)}};Zx.className="SimpleRNN",Fo.registerClass(Zx);var Jx=class extends Xx{constructor(e){if(super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.resetAfter)throw new og("GRUCell does not support reset_after parameter set to true.");this.units=e.units,Ag(this.units,"units"),this.activation=mx(void 0===e.activation?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=mx(void 0===e.recurrentActivation?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=null==e.useBias||e.useBias,this.kernelInitializer=jy(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=jy(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=jy(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=kx(e.kernelRegularizer),this.recurrentRegularizer=kx(e.recurrentRegularizer),this.biasRegularizer=kx(e.biasRegularizer),this.kernelConstraint=Lg(e.kernelConstraint),this.recurrentConstraint=Lg(e.recurrentConstraint),this.biasConstraint=Lg(e.biasConstraint),this.dropout=ay([1,sy([0,null==e.dropout?0:e.dropout])]),this.recurrentDropout=ay([1,sy([0,null==e.recurrentDropout?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){const t=(e=fb(e))[e.length-1];this.kernel=this.addWeight("kernel",[t,3*this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,3*this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[3*this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return ru(()=>{if(2!==(e=e).length)throw new og(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);const n=null!=t.training&&t.training;let r=e[1];e=e[0],0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=aw({ones:()=>Yc(e),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=aw({ones:()=>Yc(r),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));const a=this.dropoutMask,s=this.recurrentDropoutMask;let i,o,u;0<this.dropout&&this.dropout<1&&(e=vu(e,a[0]));let l=gy(e,this.kernel.read());this.useBias&&(l=xy(l,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(r=vu(r,s[0]));const c=this.recurrentKernel.read(),[h,p]=zh(c,[2*this.units,this.units],c.rank-1),d=gy(r,h),[f,m,g]=zh(l,3,l.rank-1),[y,b]=zh(d,2,d.rank-1);i=this.recurrentActivation.apply(gu(f,y)),o=this.recurrentActivation.apply(gu(m,b));const v=gy(vu(o,r),p);u=this.activation.apply(gu(g,v));const x=gu(vu(i,r),vu(gu(1,pc(i)),u));return[x,x]})}getConfig(){const e=super.getConfig(),t={units:this.units,activation:dx(this.activation),recurrentActivation:dx(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Vy(this.kernelInitializer),recurrentInitializer:Vy(this.recurrentInitializer),biasInitializer:Vy(this.biasInitializer),kernelRegularizer:xx(this.kernelRegularizer),recurrentRegularizer:xx(this.recurrentRegularizer),biasRegularizer:xx(this.biasRegularizer),activityRegularizer:xx(this.activityRegularizer),kernelConstraint:Dg(this.kernelConstraint),recurrentConstraint:Dg(this.recurrentConstraint),biasConstraint:Dg(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign({},e,t)}};Jx.className="GRUCell",Fo.registerClass(Jx);var ew=class extends Yx{constructor(e){0===e.implementation&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new Jx(e),super(e)}call(e,t){return ru(()=>{null!=this.cell.dropoutMask&&(au(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(au(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);const n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}static fromConfig(e,t){return 0===t.implmentation&&(t.implementation=1),new e(t)}};ew.className="GRU",Fo.registerClass(ew);var tw=class extends Xx{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Ag(this.units,"units"),this.activation=mx(void 0===e.activation?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=mx(void 0===e.recurrentActivation?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=null==e.useBias||e.useBias,this.kernelInitializer=jy(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=jy(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=jy(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=e.unitForgetBias,this.kernelRegularizer=kx(e.kernelRegularizer),this.recurrentRegularizer=kx(e.recurrentRegularizer),this.biasRegularizer=kx(e.biasRegularizer),this.kernelConstraint=Lg(e.kernelConstraint),this.recurrentConstraint=Lg(e.recurrentConstraint),this.biasConstraint=Lg(e.biasConstraint),this.dropout=ay([1,sy([0,null==e.dropout?0:e.dropout])]),this.recurrentDropout=ay([1,sy([0,null==e.recurrentDropout?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){var t;const n=(e=fb(e))[e.length-1];let r;if(this.kernel=this.addWeight("kernel",[n,4*this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,4*this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){if(this.unitForgetBias){const e=this.biasInitializer,n=this.units;r=new((t=class extends Ay{apply(t,r){const a=e.apply([n]),s=(new Ny).apply([n]),i=e.apply([2*n]);return dy(dy(a,s),i)}}).className="CustomInit",t)}else r=this.biasInitializer;this.bias=this.addWeight("bias",[4*this.units],null,r,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(e,t){return ru(()=>{const n=null!=t.training&&t.training;if(3!==(e=e).length)throw new og(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let r=e[1];const a=e[2];e=e[0],0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=aw({ones:()=>Yc(e),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=aw({ones:()=>Yc(r),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));const s=this.dropoutMask,i=this.recurrentDropoutMask;let o,u,l,c;0<this.dropout&&this.dropout<1&&(e=vu(e,s[0]));let h=gy(e,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(r=vu(r,i[0])),h=gu(h,gy(r,this.recurrentKernel.read())),this.useBias&&(h=xy(h,this.bias.read()));const[p,d,f,m]=zh(h,4,h.rank-1);o=this.recurrentActivation.apply(p),u=this.recurrentActivation.apply(d),l=gu(vu(u,a),vu(o,this.activation.apply(f))),c=this.recurrentActivation.apply(m);const g=vu(c,this.activation.apply(l));return[g,g,l]})}getConfig(){const e=super.getConfig(),t={units:this.units,activation:dx(this.activation),recurrentActivation:dx(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Vy(this.kernelInitializer),recurrentInitializer:Vy(this.recurrentInitializer),biasInitializer:Vy(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:xx(this.kernelRegularizer),recurrentRegularizer:xx(this.recurrentRegularizer),biasRegularizer:xx(this.biasRegularizer),activityRegularizer:xx(this.activityRegularizer),kernelConstraint:Dg(this.kernelConstraint),recurrentConstraint:Dg(this.recurrentConstraint),biasConstraint:Dg(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign({},e,t)}};tw.className="LSTMCell",Fo.registerClass(tw);var nw=class extends Yx{constructor(e){0===e.implementation&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new tw(e),super(e)}call(e,t){return ru(()=>{null!=this.cell.dropoutMask&&(au(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(au(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);const n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}static fromConfig(e,t){return 0===t.implmentation&&(t.implementation=1),new e(t)}};nw.className="LSTM",Fo.registerClass(nw);var rw=class extends Xx{constructor(e){super(e),this.cells=e.cells}get stateSize(){const e=[];for(const t of this.cells.slice().reverse())Array.isArray(t.stateSize)?e.push(...t.stateSize):e.push(t.stateSize);return e}call(e,t){return ru(()=>{let n=(e=e).slice(1);const r=[];for(const e of this.cells.slice().reverse())Array.isArray(e.stateSize)?r.push(n.splice(0,e.stateSize.length)):r.push(n.splice(0,1));r.reverse();const a=[];let s;for(let i=0;i<this.cells.length;++i){const o=this.cells[i];n=r[i],s=0===i?[e[0]].concat(n):[s[0]].concat(n),s=o.call(s,t),a.push(s.slice(1))}n=[];for(const e of a.slice().reverse())n.push(...e);return[s[0]].concat(n)})}build(e){let t;hb(e)&&(e=e[0]),e=e,this.cells.forEach((n,r)=>{Zg("RNNCell_"+r,()=>{n.build(e),t=Array.isArray(n.stateSize)?n.stateSize[0]:n.stateSize,e=[e[0],t]})}),this.built=!0}getConfig(){const e=super.getConfig(),t={cells:this.cells.map(e=>({className:e.getClassName(),config:e.getConfig()}))};return Object.assign({},e,t)}static fromConfig(e,t,n={}){const r=[];for(const e of t.cells)r.push(Bb(e,n));return new e({cells:r})}get trainableWeights(){if(!this.trainable)return[];const e=[];for(const t of this.cells)e.push(...t.trainableWeights);return e}get nonTrainableWeights(){const e=[];for(const t of this.cells)e.push(...t.nonTrainableWeights);if(!this.trainable){const t=[];for(const e of this.cells)t.push(...e.trainableWeights);return t.concat(e)}return e}getWeights(){const e=[];for(const t of this.cells)e.push(...t.weights);return yb(e)}setWeights(e){const t=[];for(const n of this.cells){const r=n.weights.length,a=e.splice(r);for(let e=0;e<n.weights.length;++e)t.push([n.weights[e],a[e]])}bb(t)}};function aw(e){const{ones:t,rate:n,training:r=!1,count:a=1,dropoutFunc:s}=e,i=()=>null!=s?s(t(),n):wy(t(),n),o=()=>ky(i,t,r);if(!a||a<=1)return su(o().clone());return Array(a).fill(void 0).map(o).map(e=>su(e.clone()))}rw.className="StackedRNNCells",Fo.registerClass(rw);var sw=class extends Yx{constructor(e){if(e.unroll)throw new ug("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(e.cell))throw new ug("It is not possible at the moment to stack convolutional cells.");super(e),this.inputSpec=[new vb({ndim:5})]}call(e,t){return ru(()=>{if(null!=this.cell.dropoutMask&&(au(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(au(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),t&&t.constants)throw new og("ConvRNN2D cell does not support constants");const n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}computeOutputShape(e){let t=this.computeSingleOutputShape(e);return this.returnSequences||(t=[t[0],...t.slice(2)]),this.returnState&&(t=[t,...Array(2).fill([e[0],...t.slice(-3)])]),t}getInitialState(e){return ru(()=>{const{stateSize:t}=this.cell,n=e.shape,r=this.computeSingleOutputShape(n),a=Pc([r[0],...r.slice(2)]);return Array.isArray(t)?Array(t.length).fill(a):[a]})}resetStates(e,t=!1){ru(()=>{if(!this.stateful)throw new sg("Cannot call resetStates() on an RNN Layer that is not stateful.");const n=this.inputSpec[0].shape,r=this.computeSingleOutputShape(n),a=[r[0],...r.slice(2)];if(null==n[0])throw new og("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(null==this.getStates())Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>Pc(a)):this.states_=[Pc(a)];else if(null==e)au(this.states_),null!=this.keptStates&&(au(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>Pc(a)):this.states_[0]=Pc(a);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new og(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t?this.keptStates.push(this.states_.slice()):au(this.states_);for(let t=0;t<this.states_.length;++t){const n=e[t],r=a;if(!va.arraysEqual(n.shape,r))throw new og(`State ${t} is incompatible with layer ${this.name}: expected shape=${r}, received shape=${n.shape}`);this.states_[t]=n}}this.states_=this.states_.map(e=>su(e.clone()))})}computeSingleOutputShape(e){const{dataFormat:t,filters:n,kernelSize:r,padding:a,strides:s,dilationRate:i}=this.cell,o="channelsFirst"===t,u=e[o?3:2],l=e[o?4:3],c=Rx(u,r[0],a,s[0],i[0]),h=Rx(l,r[1],a,s[1],i[1]);return[...e.slice(0,2),...o?[n,c,h]:[c,h,n]]}};sw.className="ConvRNN2D";var iw=class extends tw{constructor(e){const{filters:t,kernelSize:n,strides:r,padding:a,dataFormat:s,dilationRate:i}=e;super(Object.assign({},e,{units:t})),this.filters=t,Ag(this.filters,"filters"),this.kernelSize=Tx(n,2,"kernelSize"),this.kernelSize.forEach(e=>Ag(e,"kernelSize")),this.strides=Tx(r||1,2,"strides"),this.strides.forEach(e=>Ag(e,"strides")),this.padding=a||"valid",Yg(this.padding),this.dataFormat=s||"channelsLast",Kg(this.dataFormat),this.dilationRate=Tx(i||1,2,"dilationRate"),this.dilationRate.forEach(e=>Ag(e,"dilationRate"))}build(e){var t;e=fb(e);const n="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[n])throw new og("The channel dimension of the input should be defined. Found "+e[n]);const r=e[n],a=this.kernelSize.concat([r,4*this.filters]);this.kernel=this.addWeight("kernel",a,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);const s=this.kernelSize.concat([this.filters,4*this.filters]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",s,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let e;if(this.unitForgetBias){const n=this.biasInitializer,r=this.filters;e=new((t=class extends Ay{apply(e,t){return py([n.apply([r]),$c([r]),n.apply([2*r])])}}).className="CustomInit",t)}else e=this.biasInitializer;this.bias=this.addWeight("bias",[4*this.filters],null,e,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(e,t){return ru(()=>{if(3!==e.length)throw new og(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);const n=t.training||!1,r=e[0],a=e[1],s=e[2];0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=aw({ones:()=>Yc(r),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));const i=this.dropoutMask,o=(e,t,n)=>t&&t[n]?vu(t[n],e):e;let u=o(r,i,0),l=o(r,i,1),c=o(r,i,2),h=o(r,i,3);0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=aw({ones:()=>Yc(a),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));const p=this.recurrentDropoutMask;let d=o(a,p,0),f=o(a,p,1),m=o(a,p,2),g=o(a,p,3);const[y,b,v,x]=zh(this.kernel.read(),4,3),[w,k,S,I]=this.useBias?zh(this.bias.read(),4):[null,null,null,null];u=this.inputConv(u,y,w,this.padding),l=this.inputConv(l,b,k,this.padding),c=this.inputConv(c,v,S,this.padding),h=this.inputConv(h,x,I,this.padding);const[A,E,N,C]=zh(this.recurrentKernel.read(),4,3);d=this.recurrentConv(d,A),f=this.recurrentConv(f,E),m=this.recurrentConv(m,N),g=this.recurrentConv(g,C);const T=this.recurrentActivation.apply(gu(u,d)),R=this.recurrentActivation.apply(gu(l,f)),_=gu(vu(R,s),vu(T,this.activation.apply(gu(c,m)))),F=vu(this.recurrentActivation.apply(gu(h,g)),this.activation.apply(_));return[F,F,_]})}getConfig(){const e=function(e,t){var n={};for(var r in e)Object.prototype.hasOwnProperty.call(e,r)&&t.indexOf(r)<0&&(n[r]=e[r]);if(null!=e&&"function"==typeof Object.getOwnPropertySymbols){var a=0;for(r=Object.getOwnPropertySymbols(e);a<r.length;a++)t.indexOf(r[a])<0&&Object.prototype.propertyIsEnumerable.call(e,r[a])&&(n[r[a]]=e[r[a]])}return n}(super.getConfig(),["units"]),t={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign({},e,t)}inputConv(e,t,n,r){const a=ml(e,t,this.strides,r||"valid","channelsFirst"===this.dataFormat?"NCHW":"NHWC",this.dilationRate);return n?xy(a,n,this.dataFormat):a}recurrentConv(e,t){return ml(e,t,1,"same","channelsFirst"===this.dataFormat?"NCHW":"NHWC")}};iw.className="ConvLSTM2DCell",Fo.registerClass(iw);var ow=class extends sw{constructor(e){const t=new iw(e);super(Object.assign({},e,{cell:t}))}static fromConfig(e,t){return new e(t)}};ow.className="ConvLSTM2D",Fo.registerClass(ow);var uw=class extends Ib{constructor(e){super(e),this.rate=Math.max(Math.min(e.rate,1),0),this.noiseShape=e.noiseShape,this.seed=e.seed,this.supportsMasking=!0}getNoiseShape(e){if(null==this.noiseShape)return this.noiseShape;const t=e.shape,n=[];for(let e=0;e<this.noiseShape.length;++e)n.push(null==this.noiseShape[e]?t[e]:this.noiseShape[e]);return n}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e);if(0<this.rate&&this.rate<1){const e=null!=t.training&&t.training,r=this.getNoiseShape(n);return ky(()=>wy(n,this.rate,r,this.seed),()=>n,e)}return e})}getConfig(){const e={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},t=super.getConfig();return Object.assign(e,t),e}dispose(){return super.dispose()}};uw.className="Dropout",Fo.registerClass(uw);var lw=class extends uw{constructor(e){super(e),this.inputSpec=[{ndim:3}]}getNoiseShape(e){const t=e.shape;return[t[0],1,t[2]]}};lw.className="SpatialDropout1D",Fo.registerClass(lw);var cw=class extends Ib{constructor(e){if(super(e),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",null==e.batchInputShape&&null==e.inputShape&&null!=e.inputDim){let t=null;null!=e.batchSize&&(t=e.batchSize),this.batchInputShape=[t,e.inputDim]}this.units=e.units,Ag(this.units,"units"),this.activation=mx(e.activation),null!=e.useBias&&(this.useBias=e.useBias),this.kernelInitializer=jy(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=jy(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=Lg(e.kernelConstraint),this.biasConstraint=Lg(e.biasConstraint),this.kernelRegularizer=kx(e.kernelRegularizer),this.biasRegularizer=kx(e.biasRegularizer),this.activityRegularizer=kx(e.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(e){const t=(e=fb(e))[e.length-1];null==this.kernel&&(this.kernel=this.addWeight("kernel",[t,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:t}}],this.built=!0}computeOutputShape(e){const t=(e=fb(e)).slice();return t[t.length-1]=this.units,t}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e),r=Eg(this.activation.getClassName());let a;return null!=r?a=gy(n,this.kernel.read(),r,this.bias?this.bias.read():null):(a=gy(n,this.kernel.read()),null!=this.bias&&(a=xy(a,this.bias.read())),null!=this.activation&&(a=this.activation.apply(a))),a})}getConfig(){const e={units:this.units,activation:dx(this.activation),useBias:this.useBias,kernelInitializer:Vy(this.kernelInitializer),biasInitializer:Vy(this.biasInitializer),kernelRegularizer:xx(this.kernelRegularizer),biasRegularizer:xx(this.biasRegularizer),activityRegularizer:xx(this.activityRegularizer),kernelConstraint:Dg(this.kernelConstraint),biasConstraint:Dg(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}};cw.className="Dense",Fo.registerClass(cw);var hw=class extends Ib{constructor(e){super(e=e||{}),this.inputSpec=[{minNDim:3}],this.dataFormat=e.dataFormat}computeOutputShape(e){e=fb(e);for(const t of e.slice(1))if(null==t)throw new og(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[e[0],ry(e,1)]}call(e,t){return ru(()=>{this.invokeCallHook(e,t);let n=db(e);if("channelsFirst"===this.dataFormat&&n.rank>1){const e=[0];for(let t=2;t<n.rank;++t)e.push(t);e.push(1),n=qi(n,e)}return function(e){if(e.rank<=1)throw new og(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);const t=[e.shape[0],ry(e.shape,1)];return qu(e,t)}(n)})}getConfig(){const e={};null!=this.dataFormat&&(e.dataFormat=this.dataFormat);const t=super.getConfig();return Object.assign(e,t),e}};hw.className="Flatten",Fo.registerClass(hw);var pw=class extends Ib{constructor(e){super(e),this.supportsMasking=!0,this.activation=mx(e.activation)}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e);return this.activation.apply(n)})}getConfig(){const e={activation:dx(this.activation)},t=super.getConfig();return Object.assign(e,t),e}};pw.className="Activation",Fo.registerClass(pw);var dw=class extends Ib{constructor(e){super(e),this.n=e.n,this.inputSpec=[{ndim:2}]}computeOutputShape(e){return[e[0],this.n,e[1]]}call(e,t){return ru(()=>{return e=db(e),t=e,n=this.n,ru(()=>{if(2!==t.shape.length)throw new og(`repeat() expects a rank-2 tensor, but received a rank-${t.shape.length} tensor.`);return fy(uy(t,1),[1,n,1])});var t,n})}getConfig(){const e={n:this.n},t=super.getConfig();return Object.assign(e,t),e}};dw.className="RepeatVector",Fo.registerClass(dw);var fw=class extends Ib{constructor(e){super(e),this.targetShape=e.targetShape;for(let e=0;e<this.targetShape.length;++e)this.isUnknown(this.targetShape[e])&&(this.targetShape[e]=null)}isUnknown(e){return e<0||null==e}fixUnknownDimension(e,t){const n="Total size of new array must be unchanged.",r=t.slice();let a=1,s=null;for(let e=0;e<r.length;++e){const t=r[e];if(this.isUnknown(t)){if(null!==s)throw new og("Can only specifiy one unknown dimension.");s=e}else a*=t}const i=ry(e);if(null!==s){if(0===a||i%a!=0)throw new og(n);r[s]=i/a}else if(i!==a)throw new og(n);return r}computeOutputShape(e){let t=!1;for(let n=0;n<e.length;++n)if(this.isUnknown(e[n])){t=!0;break}return t?e.slice(0,1).concat(this.targetShape):e.slice(0,1).concat(this.fixUnknownDimension(e.slice(1),this.targetShape))}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e),r=n.shape,a=r.slice(0,1).concat(this.fixUnknownDimension(r.slice(1),this.targetShape));return qu(n,a)})}getConfig(){const e={targetShape:this.targetShape},t=super.getConfig();return Object.assign(e,t),e}};fw.className="Reshape",Fo.registerClass(fw);var mw=class extends Ib{constructor(e){if(super(e),null==e.dims)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(e.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);const t=iy(1,e.dims.length+1);if(!va.arraysEqual(e.dims.slice().sort(),t))throw new Error("Invalid permutation `dims`: "+JSON.stringify(e.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=e.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new vb({ndim:this.dims.length+1})]}computeOutputShape(e){const t=(e=fb(e)).slice();return this.dims.forEach((n,r)=>{t[r+1]=e[n]}),t}call(e,t){return qi(db(e),this.dimsIncludingBatch)}getConfig(){const e={dims:this.dims},t=super.getConfig();return Object.assign(e,t),e}};mw.className="Permute",Fo.registerClass(mw);var gw=class extends Ib{constructor(e){super(null==e?{}:e),this.supportsMasking=!0,this.maskValue=null!=e?null==e.maskValue?0:e.maskValue:0}computeOutputShape(e){return e}getConfig(){const e=super.getConfig(),t={maskValue:this.maskValue};return Object.assign(t,e),t}computeMask(e,t){const n=db(e);return Au(Kc(n,this.maskValue),-1)}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e),r=Au(Kc(n,this.maskValue),-1,!0);return vu(n,Ai(r,n.dtype))})}};gw.className="Masking",Fo.registerClass(gw);var yw=class extends Ib{constructor(e){if(super(e),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",null==e.batchInputShape&&null==e.inputShape){let t=null;null!=e.batchSize&&(t=e.batchSize),null==e.inputLength?this.batchInputShape=[t,null]:this.batchInputShape=[t].concat(fg(e.inputLength))}this.inputDim=e.inputDim,Ag(this.inputDim,"inputDim"),this.outputDim=e.outputDim,Ag(this.outputDim,"outputDim"),this.embeddingsInitializer=jy(e.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=kx(e.embeddingsRegularizer),this.activityRegularizer=kx(e.activityRegularizer),this.embeddingsConstraint=Lg(e.embeddingsConstraint),this.maskZero=e.maskZero,this.supportsMasking=e.maskZero,this.inputLength=e.inputLength}build(e){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(e){}computeMask(e,t){return ru(()=>this.maskZero?(e=db(e),Kc(e,Fl(e))):null)}computeOutputShape(e){if(e=fb(e),null==this.inputLength)return[...e,this.outputDim];const t=fg(this.inputLength);if(t.length!==e.length-1)throw new og(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);{let n=0;for(let r=0;r<t.length;++r){const a=t[r],s=e[r+1];if(null!=a&&null!=s&&a!==s)throw new og(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);null==a&&(t[n]=s),n++}}return[e[0],...t,this.outputDim]}call(e,t){return ru(()=>{this.invokeCallHook(e,t);let n=db(e);"int32"!==n.dtype&&(n=oy(n,"int32"));const r=yy(this.embeddings.read(),qu(n,[n.size]));return qu(r,fb(this.computeOutputShape(n.shape)))})}getConfig(){const e={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Vy(this.embeddingsInitializer),embeddingsRegularizer:xx(this.embeddingsRegularizer),activityRegularizer:xx(this.activityRegularizer),embeddingsConstraint:Dg(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},t=super.getConfig();return Object.assign(e,t),e}};yw.className="Embedding",Fo.registerClass(yw);var bw=class extends Ib{constructor(e){super(e||{}),this.supportsMasking=!0}mergeFunction(e){throw new ug}computeElementwiseOpOutputShape(e,t){if(null==e||null==t)return null;if(e.length<t.length)return this.computeElementwiseOpOutputShape(t,e);if(0===t.length)return e;const n=e.slice(0,e.length-t.length);for(let r=0;r<t.length;++r){const a=e[e.length-t.length+r],s=t[r];if(null==a||null==s||a<0||s<0)n.push(null);else if(1===a)n.push(s);else if(1===s)n.push(a);else{if(a!==s)throw new og("Operands could not be broadcast together with shapes "+JSON.stringify(e)+" "+JSON.stringify(t));n.push(a)}}return n}build(e){if(Array.isArray(e)&&!Array.isArray(e[0])&&(e=[fb(e)]),(e=e).length<2)throw new og(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);let t=[];for(const n of e)null!=n&&null!==n[0]&&t.push(n[0]);if(t=wg(t),t.length>1)throw new og(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);let n=null==e[0]?null:e[0].slice(1);for(let t=1;t<e.length;++t){const r=null==e[t]?null:e[t].slice(1);n=this.computeElementwiseOpOutputShape(n,r)}const r=e.map(e=>e.length);-1===e.indexOf(null)&&1===wg(r).length?this.reshapeRequired=!1:this.reshapeRequired=!0}call(e,t){return ru(()=>{if(e=e,this.reshapeRequired){const t=[],n=e.map(e=>e.rank);if(-1===n.indexOf(null)){const r=sy(n);for(let n of e){const e=n.rank;for(let t=0;t<r-e;++t)n=uy(n,1);t.push(n)}return this.mergeFunction(t)}{let n=!1;for(const r of e){const e=r.rank;if(null==e){const e=r.shape,a=e[0],s=e.slice(1).concat([a]);let i=qu(r,[a].concat(ry(e.slice(1))));i=qi(i,[1,0]),i=qu(i,s),t.push(i),n=!0}else if(e>1){const a=iy(1,e).concat([0]);t.push(qi(r,a)),n=!0}else t.push(r)}let r=this.mergeFunction(t);const a=r.rank;if(n)if(null==a){const e=r.shape,t=e[e.length-1],n=[t].concat(e.slice(0,e.length-1));r=qu(qi(qu(r,[-1,t]),[1,0]),n)}else if(a>1){const e=[a-1].concat(iy(0,a-1));r=qi(r,e)}return r}}return this.mergeFunction(e)})}computeOutputShape(e){let t;t=null==(e=e)[0]?null:e[0].slice(1);for(let n=1;n<e.length;++n){const r=null==e[n]?null:e[n].slice(1);t=this.computeElementwiseOpOutputShape(t,r)}let n=[];for(const t of e)null!=t&&null!==t[0]&&n.push(t[0]);return n=wg(n),t=1===n.length?n.concat(t):[null].concat(t),t}computeMask(e,t){return ru(()=>{if(null==t)return null;if(!Array.isArray(t))throw new og("`mask` should be an Array");if(!Array.isArray(e))throw new og("`inputs` should be an Array");if(t.length!==e.length)throw new og(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);if(t.every(e=>null==e))return null;let n=(t=t.map(e=>null==e?e:Bl(e,0)))[0];for(let e=1;e<t.length-1;++e)n=Cc(n,t[e]);return n})}},vw=class extends bw{constructor(e){super(e)}mergeFunction(e){return ru(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=gu(t,e[n]);return t})}};vw.className="Add",Fo.registerClass(vw);var xw=class extends bw{constructor(e){super(e)}mergeFunction(e){return ru(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=vu(t,e[n]);return t})}};xw.className="Multiply",Fo.registerClass(xw);var ww=class extends bw{constructor(e){super(e)}mergeFunction(e){return ru(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=gu(t,e[n]);return vu(1/e.length,t)})}};ww.className="Average",Fo.registerClass(ww);var kw=class extends bw{constructor(e){super(e)}mergeFunction(e){return ru(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=Oc(t,e[n]);return t})}};kw.className="Maximum",Fo.registerClass(kw);var Sw=class extends bw{constructor(e){super(e)}mergeFunction(e){return ru(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=Uc(t,e[n]);return t})}};Sw.className="Minimum",Fo.registerClass(Sw);var Iw=class extends bw{constructor(e){super(e),this.DEFAULT_AXIS=-1,null==e&&(e={}),this.axis=null==e.axis?this.DEFAULT_AXIS:e.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){if(!Array.isArray(e)||!Array.isArray(e[0])||1===e.length)throw new og("A `Concatenate` layer should be called on a list of at least 2 inputs");e=e;let t=!0;for(const n of e)if(null!=n){t=!1;break}if(t)return;const n=[];for(let t=0;t<e.length;++t){const r=e[t].slice();r.splice(this.axis,1);let a=!1;for(const e of n)if(va.arraysEqual(e,r)){a=!0;break}a||n.push(r)}if(n.length>1)throw new og("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(e))}mergeFunction(e){return ru(()=>py(e,this.axis))}computeOutputShape(e){if(!Array.isArray(e)||!Array.isArray(e[0]))throw new og("A `Concatenate` layer should be called on a list of inputs.");const t=e,n=t[0].slice(),r=this.axis<0?n.length+this.axis:this.axis;for(const e of t.slice(1)){if(null==n[r]||null==e[r]){n[r]=null;break}n[r]+=e[r]}return n}computeMask(e,t){if(null==t)return null;if(!Array.isArray(t))throw new og("`mask` should be an array for Concatenate");if(!Array.isArray(e))throw new og("`inputs` should be an array for Concatenate");if(t.length!==e.length)throw new og(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);return ru(()=>{let n=!0;if(t.forEach(e=>{null==e||(n=!1)}),n)return null;const r=[];for(let n=0;n<e.length;++n)null==t[n]?r.push(Ai(Yc(e[n]),"bool")):t[n].rank<e[n].rank?r.push(Bl(t[n],-1)):r.push(t[n]);const a=Xu(r,this.axis);return Iu(a,-1,!1)})}getConfig(){const e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};function Aw(e,t){for(;e<0;)e+=t;return e}Iw.className="Concatenate",Fo.registerClass(Iw);var Ew=class extends bw{constructor(e){super(e),this.axes=e.axes,this.normalize=null!=e.normalize&&e.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){va.assert(Array.isArray(e)&&2===e.length&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");const t=e[0],n=e[1];if(t.length>3||n.length>3)throw new ug("Dot layer does not support tensors of 4D or higher rank yet.");const r=this.interpretAxes(t,n);if(t[r[0]]!==n[r[1]])throw new og(`Dimension incompatibility: ${t[r[0]]} !== ${n[r[1]]}`)}mergeFunction(e){if(2!==e.length)throw new og(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);let t,n=e[0],r=e[1];return t=Array.isArray(this.axes)?this.axes.map((t,n)=>Aw(t,e[n].shape.length)):[Aw(this.axes,n.shape.length),Aw(this.axes,r.shape.length)],this.normalize&&(n=zb(n,t[0]),r=zb(r,t[1])),function(e,t,n){if(e.shape.length>3||t.shape.length>3)throw new ug("batchDot is not implemented for tensors of 4D or higher rank yet");if(va.assert(e.shape.length>=2,()=>"batchDot requires the rank of x to be >= 2, but got "+e.shape.length),va.assert(e.shape.length>=2,()=>"batchDot requires the rank of y to be >= 2, but got "+t.shape.length),"number"==typeof n&&(n=[n,n]),"complex64"===e.dtype||"complex64"===t.dtype)throw new ug("batchDot is not implemented for complex64-type Tensors yet.");const r=e.shape.length,a=t.shape.length;null==n&&(n=[r-1,a-2]);const s=n;return ru(()=>{let n,i;if(r>a){n=r-a;const e=[];for(let t=0;t<n;++t)e.push(1);t=qu(t,t.shape.concat(e))}else if(a>r){n=a-r;const t=[];for(let e=0;e<n;++e)t.push(1);e=qu(e,e.shape.concat(t))}else n=0;if(2===e.shape.length&&2===t.shape.length)i=s[0]===s[1]?yc(vu(e,t),s[0]):yc(vu(qi(e,[1,0]),t),s[1]);else{const n=s[0]!==e.shape.length-1,r=s[1]===t.shape.length-1;i=Gi(e,t,n,r)}if(n>0){let e;e=r>a?r+a-3:r-1;const t=[];for(let r=e;r<e+n;++r)t.push(r);i=jh(i,t)}return 1===i.shape.length&&(i=Bl(i,1)),i})}(n,r,t)}interpretAxes(e,t){let n;return n=Array.isArray(this.axes)?this.axes:[Aw(this.axes,e.length),Aw(this.axes,t.length)],n}computeOutputShape(e){va.assert(Array.isArray(e)&&2===e.length&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");const t=e[0].slice(),n=e[1].slice();if(t.length>3||n.length>3)throw new ug("Dot layer does not support tensors of 4D or higher rank yet.");const r=this.interpretAxes(t,n);t.splice(r[0],1),n.splice(r[1],1),n.splice(0,1);const a=t.concat(n);return 1===a.length&&a.push(1),a}computeMask(e,t){return null}getConfig(){const e={axes:this.axes,normalize:this.normalize},t=super.getConfig();return Object.assign(e,t),e}};Ew.className="Dot",Fo.registerClass(Ew);var Nw=class extends Ib{constructor(e){super(e),this.supportsMasking=!0,this.stddev=e.stddev}computeOutputShape(e){return e}getConfig(){const e=super.getConfig(),t={stddev:this.stddev};return Object.assign(t,e),t}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e);return ky(()=>gu(my(n.shape,0,this.stddev),n),()=>n,t.training||!1)})}};Nw.className="GaussianNoise",Fo.registerClass(Nw);var Cw=class extends Ib{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate}computeOutputShape(e){return e}getConfig(){const e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return ru(()=>{this.invokeCallHook(e,t);const n=db(e);if(this.rate>0&&this.rate<1){return ky(()=>{const e=Math.sqrt(this.rate/(1-this.rate));return vu(n,my(n.shape,1,e))},()=>n,t.training||!1)}return n})}};Cw.className="GaussianDropout",Fo.registerClass(Cw);var Tw=class extends Ib{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate,this.noiseShape=e.noiseShape}_getNoiseShape(e){return this.noiseShape||db(e).shape}computeOutputShape(e){return e}getConfig(){const e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return ru(()=>{if(this.rate<1&&this.rate>0){const n=this._getNoiseShape(e);return ky(()=>{const t=db(e),r=-1.7580993408473766;let a=ql(ph(n),this.rate);a=oy(a,"float32");const s=((1-this.rate)*(1+this.rate*r**2))**-.5,i=-s*r*this.rate,o=gu(vu(t,a),vu(gu(a,-1),r));return gu(vu(o,s),i)},()=>db(e),t.training||!1)}return e})}};function Rw(e,t,n,r,a,s=.001){let i;if(2===e.rank)i=rl(e,t,n,r,a,s);else if(3===e.rank)i=al(e,t,n,r,a,s);else{if(4!==e.rank)throw new ug(`batchNormalization is not implemented for array of rank ${e.rank} yet`);i=sl(e,t,n,r,a,s)}return i}function _w(e,t,n,r,a=.001){return va.arraysEqual(r.slice().sort(),iy(0,e.rank-1))?function(e,t,n,r,a=.001){return ru(()=>{const s=Gc(e,r),i=s.mean,o=s.variance;return[Rw(e,i,o,n,t,a),i,o]})}(e,t,n,r,a):function(e,t,n,r,a=.001){return ru(()=>{const s=Gc(e,r),i=s.mean,o=s.variance,u=[];for(const t of iy(0,e.rank))-1!==r.indexOf(t)?u.push(1):u.push(e.shape[t]);const l=qu(i,u),c=qu(o,u),h=null==t?null:qu(t,u),p=null==n?null:qu(n,u);return[Rw(e,l,c,p,h,a),i,o]})}(e,t,n,r,a)}Tw.className="AlphaDropout",Fo.registerClass(Tw);var Fw=class extends Ib{constructor(e){null==e&&(e={}),super(e),this.supportsMasking=!0,this.axis=null==e.axis?-1:e.axis,this.momentum=null==e.momentum?.99:e.momentum,this.epsilon=null==e.epsilon?.001:e.epsilon,this.center=null==e.center||e.center,this.scale=null==e.scale||e.scale,this.betaInitializer=jy(e.betaInitializer||"zeros"),this.gammaInitializer=jy(e.gammaInitializer||"ones"),this.movingMeanInitializer=jy(e.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=jy(e.movingVarianceInitializer||"ones"),this.betaConstraint=Lg(e.betaConstraint),this.gammaConstraint=Lg(e.gammaConstraint),this.betaRegularizer=kx(e.betaRegularizer),this.gammaRegularizer=kx(e.gammaRegularizer)}build(e){e=fb(e);const t=this.axis>=0?this.axis:this.axis+e.length,n=e[t];if(null==n)throw new og(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);this.inputSpec=[new vb({ndim:e.length,axes:{[t]:n}})];const r=[n];this.scale&&(this.gamma=this.addWeight("gamma",r,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",r,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",r,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",r,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(e,t){return ru(()=>{const n=null!=t.training&&t.training,r=db(e),a=r.shape,s=a.length,i=iy(0,s),o=this.axis>=0?this.axis:this.axis+s;i.splice(o,1);const u=cg(1,s);u[o]=a[o];const l=i.slice();l.sort();const c=!va.arraysEqual(l,iy(0,s).slice(0,s-1));if(!n)return(()=>{if(c){const e=qu(this.movingMean.read(),u),t=qu(this.movingVariance.read(),u),n=this.center?qu(this.beta.read(),u):null,a=this.scale?qu(this.gamma.read(),u):null;return Rw(r,e,t,n,a,this.epsilon)}return Rw(r,this.movingMean.read(),this.movingVariance.read(),null==this.beta?null:this.beta.read(),null==this.gamma?null:this.gamma.read(),this.epsilon)})();const[h,p,d]=_w(r,this.gamma.read(),this.beta.read(),i,this.epsilon),f=(e,t,n)=>{ru(()=>{const r=1-n,a=e.read(),s=vu(gc(a,t),r);e.write(gc(a,s))})};return(()=>{f(this.movingMean,p,this.momentum),f(this.movingVariance,d,this.momentum)})(),h})}getConfig(){const e={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Vy(this.betaInitializer),gammaInitializer:Vy(this.gammaInitializer),movingMeanInitializer:Vy(this.movingMeanInitializer),movingVarianceInitializer:Vy(this.movingVarianceInitializer),betaRegularizer:xx(this.betaRegularizer),gammaRegularizer:xx(this.gammaRegularizer),betaConstraint:Dg(this.betaConstraint),gammaConstraint:Dg(this.gammaConstraint)},t=super.getConfig();return Object.assign(e,t),e}};Fw.className="BatchNormalization",Fo.registerClass(Fw);var Mw=class extends Ib{constructor(e){if(null==e&&(e={}),super(e),this.axis=null==e.axis?-1:e.axis,"number"==typeof this.axis){if(!Number.isInteger(this.axis))throw new Error("Expected axis to be an integer, but received "+this.axis)}else{if(!Array.isArray(this.axis))throw new Error("Expected axis to be an integer or an array of integers, but received "+JSON.stringify(this.axis));for(const e of this.axis)if(!Number.isInteger(e))throw new Error("Expected axis to be an array of integers, but received "+JSON.stringify(this.axis))}this.epsilon=null==e.epsilon?.001:e.epsilon,this.center=null==e.center||e.center,this.scale=null==e.scale||e.scale,this.betaInitializer=jy(e.betaInitializer||"zeros"),this.gammaInitializer=jy(e.gammaInitializer||"ones"),this.betaRegularizer=kx(e.betaRegularizer),this.gammaRegularizer=kx(e.gammaRegularizer),this.supportsMasking=!0}build(e){const t=(e=fb(e)).length;"number"==typeof this.axis&&(this.axis=[this.axis]);for(let e=0;e<this.axis.length;++e)this.axis[e]<0&&(this.axis[e]+=t);for(const e of this.axis)if(e<0||e>=t)throw new Error("Invalid axis: "+e);if(this.axis.length!==wg(this.axis).length)throw new Error("Found duplicate axes in: "+this.axis);const n=this.axis.map(t=>e[t]);this.scale?this.gamma=this.addWeight("gamma",n,"float32",this.gammaInitializer,this.gammaRegularizer,!0):this.gamma=null,this.center?this.beta=this.addWeight("beta",n,"float32",this.betaInitializer,this.betaRegularizer,!0):this.beta=null,this.built=!0}call(e,t){const n=db(e),r=n.shape,a=r.length;return ru(()=>{let{mean:e,variance:t}=Gc(n,this.axis,!0);const s=cg(1,a);for(const e of this.axis)s[e]=r[e];const i=e=>null!=e&&e.shape.length!==a?qu(e,s):e;let o=i(this.gamma.read()),u=i(this.beta.read());const l=[],c=[];for(let e=0;e<a;++e)-1!==this.axis.indexOf(e)?(l.push(r[e]),c.push(1)):(l.push(1),c.push(r[e]));return e=Ul(e,l),t=Ul(t,l),o=Ul(o,c),u=Ul(u,c),Rw(n,e,t,u,o,this.epsilon)})}getConfig(){const e={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Vy(this.betaInitializer),gammaInitializer:Vy(this.gammaInitializer),betaRegularizer:xx(this.betaRegularizer),gammaRegularizer:xx(this.gammaRegularizer)},t=super.getConfig();return Object.assign(e,t),e}};Mw.className="LayerNormalization",Fo.registerClass(Mw);var Dw=class extends Ib{constructor(e){if(null==e&&(e={}),super(e),this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,null==e.padding)this.padding=[[1,1],[1,1]];else if("number"==typeof e.padding)this.padding=[[e.padding,e.padding],[e.padding,e.padding]];else{if(e.padding=e.padding,2!==e.padding.length)throw new og(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);let t,n;if("number"==typeof e.padding[0])t=[e.padding[0],e.padding[0]],n=[e.padding[1],e.padding[1]];else{if(e.padding=e.padding,2!==e.padding[0].length)throw new og(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);if(t=e.padding[0],2!==e.padding[1].length)throw new og(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);n=e.padding[1]}this.padding=[t,n]}this.inputSpec=[new vb({ndim:4})]}computeOutputShape(e){let t,n;return e=fb(e),"channelsFirst"===this.dataFormat?(t=null!=e[2]&&e[2]>=0?e[2]+this.padding[0][0]+this.padding[0][1]:null,n=null!=e[3]&&e[3]>=0?e[3]+this.padding[1][0]+this.padding[1][1]:null,[e[0],e[1],t,n]):(t=null!=e[1]&&e[1]>=0?e[1]+this.padding[0][0]+this.padding[0][1]:null,n=null!=e[2]&&e[2]>=0?e[2]+this.padding[1][0]+this.padding[1][1]:null,[e[0],t,n,e[3]])}call(e,t){return ru(()=>{return t=db(e),n=this.padding,r=this.dataFormat,ru(()=>{if(4!==t.rank)throw new og(`temporalPadding expects input tensor to be 4-D, but received a ${t.rank}-D tensor.`);if(null==n&&(n=[[1,1],[1,1]]),2!==n.length||2!==n[0].length||2!==n[1].length)throw new og("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(null==r&&(r="channelsLast"),"channelsLast"!==r&&"channelsFirst"!==r)throw new og(`Unknown data format: ${r}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let e;return e="channelsFirst"===r?[[0,0],[0,0],n[0],n[1]]:[[0,0],n[0],n[1],[0,0]],Qc(t,e)});var t,n,r})}getConfig(){const e={padding:this.padding,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};function Ow(e,t,n,r,a,s){return ru(()=>{let i;Kg(a),Xg(s),Yg(r),null==n&&(n=[1,1]),null==r&&(r="valid"),null==a&&(a="channelsLast"),null==s&&(s="max"),e=Fx(e,a);const o="same"===r?"same":"valid";return i="max"===s?Fc(e,t,n,o):Ku(e,t,n,o),"channelsFirst"===a&&(i=qi(i,[0,3,1,2])),i})}function Lw(e,t,n,r,a,s){return ru(()=>{let i;Kg(a),Xg(s),Yg(r),null==n&&(n=[1,1,1]),null==r&&(r="valid"),null==a&&(a="channelsLast"),null==s&&(s="max"),e=Mx(e,a);const o="same"===r?"same":"valid";return i="max"===s?Mc(e,t,n,o):Yu(e,t,n,o),"channelsFirst"===a&&(i=qi(i,[0,4,1,2,3])),i})}Dw.className="ZeroPadding2D",Fo.registerClass(Dw);var Pw=class extends Ib{constructor(e){if(null==e.poolSize&&(e.poolSize=2),super(e),"number"==typeof e.poolSize)this.poolSize=[e.poolSize];else{if(!Array.isArray(e.poolSize)||1!==e.poolSize.length||"number"!=typeof e.poolSize[0])throw new og("poolSize for 1D convolutional layer must be a number or an Array of a single number, but received "+JSON.stringify(e.poolSize));this.poolSize=e.poolSize}if(Ag(this.poolSize,"poolSize"),null==e.strides)this.strides=this.poolSize;else if("number"==typeof e.strides)this.strides=[e.strides];else{if(!Array.isArray(e.strides)||1!==e.strides.length||"number"!=typeof e.strides[0])throw new og("strides for 1D convolutional layer must be a number or an Array of a single number, but received "+JSON.stringify(e.strides));this.strides=e.strides}Ag(this.strides,"strides"),this.padding=null==e.padding?"valid":e.padding,Yg(this.padding),this.inputSpec=[new vb({ndim:3})]}computeOutputShape(e){const t=Rx((e=fb(e))[1],this.poolSize[0],this.padding,this.strides[0]);return[e[0],t,e[2]]}call(e,t){return ru(()=>{this.invokeCallHook(e,t),e=uy(db(e),2);const n=this.poolingFunction(db(e),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return jh(n,[2])})}getConfig(){const e={poolSize:this.poolSize,padding:this.padding,strides:this.strides},t=super.getConfig();return Object.assign(e,t),e}},$w=class extends Pw{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Kg(a),Yg(r),Ow(e,t,n,r,a,"max")}};$w.className="MaxPooling1D",Fo.registerClass($w);var Bw=class extends Pw{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Kg(a),Yg(r),Ow(e,t,n,r,a,"avg")}};Bw.className="AveragePooling1D",Fo.registerClass(Bw);var zw=class extends Ib{constructor(e){if(null==e.poolSize&&(e.poolSize=[2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize],null==e.strides)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(2!==e.strides.length)throw new og(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides];Ag(this.poolSize,"poolSize"),Ag(this.strides,"strides"),this.padding=null==e.padding?"valid":e.padding,this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Kg(this.dataFormat),Yg(this.padding),this.inputSpec=[new vb({ndim:4})]}computeOutputShape(e){e=fb(e);let t="channelsFirst"===this.dataFormat?e[2]:e[1],n="channelsFirst"===this.dataFormat?e[3]:e[2];return t=Rx(t,this.poolSize[0],this.padding,this.strides[0]),n=Rx(n,this.poolSize[1],this.padding,this.strides[1]),"channelsFirst"===this.dataFormat?[e[0],e[1],t,n]:[e[0],t,n,e[3]]}call(e,t){return ru(()=>(this.invokeCallHook(e,t),this.poolingFunction(db(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){const e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},Uw=class extends zw{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Kg(a),Yg(r),Ow(e,t,n,r,a,"max")}};Uw.className="MaxPooling2D",Fo.registerClass(Uw);var Ww=class extends zw{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Kg(a),Yg(r),Ow(e,t,n,r,a,"avg")}};Ww.className="AveragePooling2D",Fo.registerClass(Ww);var Vw=class extends Ib{constructor(e){if(null==e.poolSize&&(e.poolSize=[2,2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize,e.poolSize],null==e.strides)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(3!==e.strides.length)throw new og(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides,e.strides];Ag(this.poolSize,"poolSize"),Ag(this.strides,"strides"),this.padding=null==e.padding?"valid":e.padding,this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Kg(this.dataFormat),Yg(this.padding),this.inputSpec=[new vb({ndim:5})]}computeOutputShape(e){e=fb(e);let t="channelsFirst"===this.dataFormat?e[2]:e[1],n="channelsFirst"===this.dataFormat?e[3]:e[2],r="channelsFirst"===this.dataFormat?e[4]:e[3];return t=Rx(t,this.poolSize[0],this.padding,this.strides[0]),n=Rx(n,this.poolSize[1],this.padding,this.strides[1]),r=Rx(r,this.poolSize[2],this.padding,this.strides[2]),"channelsFirst"===this.dataFormat?[e[0],e[1],t,n,r]:[e[0],t,n,r,e[4]]}call(e,t){return ru(()=>(this.invokeCallHook(e,t),this.poolingFunction(db(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){const e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},jw=class extends Vw{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Kg(a),Yg(r),Lw(e,t,n,r,a,"max")}};jw.className="MaxPooling3D",Fo.registerClass(jw);var Gw=class extends Vw{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Kg(a),Yg(r),Lw(e,t,n,r,a,"avg")}};Gw.className="AveragePooling3D",Fo.registerClass(Gw);var Hw=class extends Ib{constructor(e){super(e),this.inputSpec=[new vb({ndim:3})]}computeOutputShape(e){return[e[0],e[2]]}call(e,t){throw new ug}},qw=class extends Hw{constructor(e){super(e||{})}call(e,t){return ru(()=>{const t=db(e);return Lc(t,1)})}};qw.className="GlobalAveragePooling1D",Fo.registerClass(qw);var Kw=class extends Hw{constructor(e){super(e||{})}call(e,t){return ru(()=>{const t=db(e);return mc(t,1)})}};Kw.className="GlobalMaxPooling1D",Fo.registerClass(Kw);var Yw=class extends Ib{constructor(e){super(e),this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Kg(this.dataFormat),this.inputSpec=[new vb({ndim:4})]}computeOutputShape(e){return e=e,"channelsLast"===this.dataFormat?[e[0],e[3]]:[e[0],e[1]]}call(e,t){throw new ug}getConfig(){const e={dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},Xw=class extends Yw{call(e,t){return ru(()=>{const t=db(e);return"channelsLast"===this.dataFormat?Lc(t,[1,2]):Lc(t,[2,3])})}};Xw.className="GlobalAveragePooling2D",Fo.registerClass(Xw);var Qw=class extends Yw{call(e,t){return ru(()=>{const t=db(e);return"channelsLast"===this.dataFormat?mc(t,[1,2]):mc(t,[2,3])})}};Qw.className="GlobalMaxPooling2D",Fo.registerClass(Qw);var Zw=class extends Ib{constructor(e){super(e),this.layer=e.layer}build(e){this.built=!0}get trainable(){return null!=this.layer&&this.layer.trainable}set trainable(e){null!=this.layer&&(this.layer.trainable=e)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(e){this.layer.setWeights(e)}getConfig(){const e={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},t=super.getConfig();return Object.assign(e,t),e}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),null!=this.layer&&this.layer.setFastWeightInitDuringBuild(e)}static fromConfig(e,t,n={}){const r=Bb(t.layer,n);delete t.layer;const a={layer:r};return Object.assign(a,t),new e(a)}},Jw=class extends Zw{constructor(e){super(e),this.supportsMasking=!0}build(e){if((e=fb(e)).length<3)throw new og("TimeDistributed layer expects an input shape >= 3D, but received input shape "+JSON.stringify(e));this.inputSpec=[{shape:e}];const t=[e[0]].concat(e.slice(2));this.layer.built||(this.layer.build(t),this.layer.built=!0),super.build(e)}computeOutputShape(e){const t=[(e=fb(e))[0]].concat(e.slice(2)),n=this.layer.computeOutputShape(t),r=e[1];return[n[0],r].concat(n.slice(1))}call(e,t){return ru(()=>Kx((e,n)=>[db(this.layer.call(e,t)),[]],e=db(e),[],!1,null,null,!1,!0)[1])}};Jw.className="TimeDistributed",Fo.registerClass(Jw);var ek=class extends Zw{constructor(e){super(e);const t=e.layer.getConfig(),n={};n.className=e.layer.getClassName(),n.config=t,this.forwardLayer=Bb(n),t.goBackwards=!0!==t.goBackwards;const r={};var a;if(r.className=e.layer.getClassName(),r.config=t,this.backwardLayer=Bb(r),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=void 0===e.mergeMode?"concat":e.mergeMode,a=this.mergeMode,Sg(Hg,"BidirectionalMergeMode",a),e.weights)throw new ug("weights support is not implemented for Bidirectional layer yet.");this._stateful=e.layer.stateful,this.returnSequences=e.layer.returnSequences,this.returnState=e.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=e.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(e){this._trainable=e,null!=this.forwardLayer&&(this.forwardLayer.trainable=e),null!=this.backwardLayer&&(this.backwardLayer.trainable=e)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(e){const t=e.length,n=Math.floor(t/2);this.forwardLayer.setWeights(e.slice(0,n)),this.backwardLayer.setWeights(e.slice(n))}computeOutputShape(e){let t,n,r,a=this.forwardLayer.computeOutputShape(e);return Array.isArray(a)&&Array.isArray(a[0])||(a=[a]),a=a,this.returnState?(r=a.slice(1),t=a[0]):t=a[0],t=t,"concat"===this.mergeMode?(t[t.length-1]*=2,n=[t]):n=null==this.mergeMode?[t,t.slice()]:[t],this.returnState?null==this.mergeMode?n.concat(r).concat(r.slice()):[t].concat(r).concat(r.slice()):dg(n)}apply(e,t){let n=null==t?null:t.initialState,r=null==t?null:t.constants;null==t&&(t={});const a=qx(e,n,r,this.numConstants);if(e=a.inputs,n=a.initialState,r=a.constants,Array.isArray(e)&&(n=e.slice(1),e=e[0]),(null==n||0===n.length)&&null==r)return super.apply(e,t);const s=[],i=[];if(null!=n){const e=n.length;if(e%2>0)throw new og("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");t.initialState=n,s.push(...n);const r=n.map(e=>new vb({shape:e.shape}));this.forwardLayer.stateSpec=r.slice(0,e/2),this.backwardLayer.stateSpec=r.slice(e/2),i.push(...r)}if(null!=r)throw new ug("Support for constants in Bidirectional layers is not implemented yet.");const o=s[0]instanceof xb;for(const e of s)if(e instanceof xb!==o)throw new og("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(o){const n=[e].concat(s),r=this.inputSpec.concat(i),a=this.inputSpec;this.inputSpec=r;const o=super.apply(n,t);return this.inputSpec=a,o}return super.apply(e,t)}call(e,t){return ru(()=>{const n=t.initialState;let r,a,s,i;if(null==n)r=this.forwardLayer.call(e,t),a=this.backwardLayer.call(e,t);else{const s=n.slice(0,n.length/2),i=n.slice(n.length/2);r=this.forwardLayer.call(e,Object.assign(t,{initialState:s})),a=this.backwardLayer.call(e,Object.assign(t,{initialState:i}))}return this.returnState&&(Array.isArray(r)&&(s=r.slice(1).concat(a.slice(1))),r=r[0],a=a[0]),this.returnSequences&&(a=bh(a,1)),"concat"===this.mergeMode?i=py([r,a]):"sum"===this.mergeMode?i=gu(r,a):"ave"===this.mergeMode?i=vu(.5,gu(r,a)):"mul"===this.mergeMode?i=vu(r,a):null==this.mergeMode&&(i=[r,a]),this.returnState?null==this.mergeMode?i.concat(s):[i].concat(s):i})}resetStates(e){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(e){Zg(this.forwardLayer.name,()=>{this.forwardLayer.build(e)}),Zg(this.backwardLayer.name,()=>{this.backwardLayer.build(e)}),this.built=!0}computeMask(e,t){let n;if(Array.isArray(t)&&(t=t[0]),n=this.returnSequences?null==this.mergeMode?[t,t]:t:null==this.mergeMode?[null,null]:null,this.returnState){const e=this.forwardLayer.states.map(e=>null);return Array.isArray(n)?n.concat(e).concat(e):[n].concat(e).concat(e)}return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),null!=this.forwardLayer&&this.forwardLayer.setFastWeightInitDuringBuild(e),null!=this.backwardLayer&&this.backwardLayer.setFastWeightInitDuringBuild(e)}getConfig(){const e={mergeMode:this.mergeMode},t=super.getConfig();return Object.assign(e,t),e}static fromConfig(e,t){const n=Bb(t.layer);if(delete t.layer,null!=t.numConstants)throw new ug("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");const r=t;return r.layer=n,new e(r)}};function tk(e){return new Nb(e)}function nk(e){return new Ex(e)}function rk(e){return new Sx(e)}function ak(e){return new Ix(e)}function sk(e){return new Ax(e)}function ik(e){return new Cx(e)}function ok(e){return new Nx(e)}function uk(e){return new Vx(e)}function lk(e){return new Px(e)}function ck(e){return new Bx(e)}function hk(e){return new $x(e)}function pk(e){return new zx(e)}function dk(e){return new Wx(e)}function fk(e){return new jx(e)}function mk(e){return new Gx(e)}function gk(e){return new Hx(e)}function yk(e){return new pw(e)}function bk(e){return new cw(e)}function vk(e){return new uw(e)}function xk(e){return new lw(e)}function wk(e){return new hw(e)}function kk(e){return new dw(e)}function Sk(e){return new fw(e)}function Ik(e){return new mw(e)}function Ak(e){return new yw(e)}function Ek(e){return new vw(e)}function Nk(e){return new ww(e)}function Ck(e){return new Iw(e)}function Tk(e){return new kw(e)}function Rk(e){return new Sw(e)}function _k(e){return new xw(e)}function Fk(e){return new Ew(e)}function Mk(e){return new Fw(e)}function Dk(e){return new Mw(e)}function Ok(e){return new Dw(e)}function Lk(e){return new Bw(e)}function Pk(e){return Lk(e)}function $k(e){return Lk(e)}function Bk(e){return new Ww(e)}function zk(e){return Bk(e)}function Uk(e){return Bk(e)}function Wk(e){return new Gw(e)}function Vk(e){return Wk(e)}function jk(e){return Wk(e)}function Gk(e){return new qw(e)}function Hk(e){return new Xw(e)}function qk(e){return new Kw(e)}function Kk(e){return new Qw(e)}function Yk(e){return new $w(e)}function Xk(e){return new Uw(e)}function Qk(e){return new jw(e)}function Zk(e){return new ew(e)}function Jk(e){return new Jx(e)}function eS(e){return new nw(e)}function tS(e){return new tw(e)}function nS(e){return new Zx(e)}function rS(e){return new Qx(e)}function aS(e){return new ow(e)}function sS(e){return new iw(e)}function iS(e){return new Yx(e)}function oS(e){return new rw(e)}function uS(e){return new ek(e)}function lS(e){return new Jw(e)}ek.className="Bidirectional",Fo.registerClass(ek);var cS=qk,hS=Kk,pS=Yk,dS=Xk;function fS(e){return new Nw(e)}function mS(e){return new Cw(e)}function gS(e){return new Tw(e)}function yS(e){return new gw(e)}var bS={};function vS(e,t){return Xb(e,t)}function xS(e,t){return tv(e,t)}function wS(e,t){return nv(e,t)}function kS(e,t){return Qb(e,t)}function SS(e,t){return rv(e,t)}function IS(e,t){return Jb(e,t)}function AS(e,t){return ev(e,t)}function ES(e,t){return qb(e,t)}function NS(e,t){return Wb(e,t)}function CS(e,t){return Vb(e,t)}function TS(e,t){return Vb(e,t)}function RS(e,t){return Vb(e,t)}function _S(e,t){return Ub(e,t)}function FS(e,t){return Ub(e,t)}function MS(e,t){return Ub(e,t)}w(bS,{MAPE:()=>TS,MSE:()=>FS,binaryAccuracy:()=>vS,binaryCrossentropy:()=>xS,categoricalAccuracy:()=>kS,categoricalCrossentropy:()=>SS,cosineProximity:()=>ES,mape:()=>RS,meanAbsoluteError:()=>NS,meanAbsolutePercentageError:()=>CS,meanSquaredError:()=>_S,mse:()=>MS,precision:()=>IS,recall:()=>AS,sparseCategoricalAccuracy:()=>wS});var DS={};w(DS,{modelFromJSON:()=>jv});var OS={};function LS(e){return new bx(e)}function PS(e){return gx(t=e),new bx({l1:null!=t?t.l1:null,l2:0});var t}function $S(e){return gx(t=e),new bx({l2:null!=t?t.l2:null,l1:0});var t}w(OS,{l1:()=>PS,l1l2:()=>LS,l2:()=>$S});var BS=class extends _b{constructor(){super(...arguments),this.model=null}setModel(e){if(!(e instanceof Wv))throw new Error("model must be a LayersModel, not some other Container");this.model=e}};function zS(e,t){return e<t}function US(e,t){return e>t}var WS=class extends BS{constructor(e){if(super(),null==e&&(e={}),e.restoreBestWeights)throw new ug("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=e.monitor||"val_loss",this.minDelta=Math.abs(e.minDelta||0),this.patience=e.patience||0,this.verbose=e.verbose||0,this.mode=e.mode||"auto",this.baseline=e.baseline,-1===["auto","min","max"].indexOf(this.mode)&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),"min"===this.mode?this.monitorFunc=zS:"max"===this.mode||-1!==this.monitor.indexOf("acc")?this.monitorFunc=US:this.monitorFunc=zS,this.monitorFunc===zS&&(this.minDelta*=-1)}async onTrainBegin(e){this.wait=0,this.stoppedEpoch=0,null!=this.baseline?this.best=this.baseline:this.best=this.monitorFunc===zS?1/0:-1/0}async onEpochEnd(e,t){await Tb(t);const n=this.getMonitorValue(t);null!=n&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=e,this.model.stopTraining=!0)))}async onTrainEnd(e){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(e){null==e&&(e={});const t=e[this.monitor];return null==t&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`),t}};var VS,jS,GS,HS,qS,KS={earlyStopping:function(e){return new WS(e)}};Ye().registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,e=>{e&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")}),(jS=VS||(VS={}))[jS.DT_INVALID=0]="DT_INVALID",jS[jS.DT_FLOAT=1]="DT_FLOAT",jS[jS.DT_DOUBLE=2]="DT_DOUBLE",jS[jS.DT_INT32=3]="DT_INT32",jS[jS.DT_UINT8=4]="DT_UINT8",jS[jS.DT_INT16=5]="DT_INT16",jS[jS.DT_INT8=6]="DT_INT8",jS[jS.DT_STRING=7]="DT_STRING",jS[jS.DT_COMPLEX64=8]="DT_COMPLEX64",jS[jS.DT_INT64=9]="DT_INT64",jS[jS.DT_BOOL=10]="DT_BOOL",jS[jS.DT_QINT8=11]="DT_QINT8",jS[jS.DT_QUINT8=12]="DT_QUINT8",jS[jS.DT_QINT32=13]="DT_QINT32",jS[jS.DT_BFLOAT16=14]="DT_BFLOAT16",jS[jS.DT_QINT16=15]="DT_QINT16",jS[jS.DT_QUINT16=16]="DT_QUINT16",jS[jS.DT_UINT16=17]="DT_UINT16",jS[jS.DT_COMPLEX128=18]="DT_COMPLEX128",jS[jS.DT_HALF=19]="DT_HALF",jS[jS.DT_RESOURCE=20]="DT_RESOURCE",jS[jS.DT_VARIANT=21]="DT_VARIANT",jS[jS.DT_UINT32=22]="DT_UINT32",jS[jS.DT_UINT64=23]="DT_UINT64",jS[jS.DT_FLOAT_REF=101]="DT_FLOAT_REF",jS[jS.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",jS[jS.DT_INT32_REF=103]="DT_INT32_REF",jS[jS.DT_UINT8_REF=104]="DT_UINT8_REF",jS[jS.DT_INT16_REF=105]="DT_INT16_REF",jS[jS.DT_INT8_REF=106]="DT_INT8_REF",jS[jS.DT_STRING_REF=107]="DT_STRING_REF",jS[jS.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",jS[jS.DT_INT64_REF=109]="DT_INT64_REF",jS[jS.DT_BOOL_REF=110]="DT_BOOL_REF",jS[jS.DT_QINT8_REF=111]="DT_QINT8_REF",jS[jS.DT_QUINT8_REF=112]="DT_QUINT8_REF",jS[jS.DT_QINT32_REF=113]="DT_QINT32_REF",jS[jS.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",jS[jS.DT_QINT16_REF=115]="DT_QINT16_REF",jS[jS.DT_QUINT16_REF=116]="DT_QUINT16_REF",jS[jS.DT_UINT16_REF=117]="DT_UINT16_REF",jS[jS.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",jS[jS.DT_HALF_REF=119]="DT_HALF_REF",jS[jS.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",jS[jS.DT_VARIANT_REF=121]="DT_VARIANT_REF",jS[jS.DT_UINT32_REF=122]="DT_UINT32_REF",jS[jS.DT_UINT64_REF=123]="DT_UINT64_REF",HS=GS||(GS={}),(qS=HS.CheckpointFormatVersion||(HS.CheckpointFormatVersion={}))[qS.LEGACY=0]="LEGACY",qS[qS.V1=1]="V1",qS[qS.V2=2]="V2";var YS={};function XS(e,t){const n={tfOpName:e,category:"custom",inputs:[],attrs:[],customExecutor:t};YS[e]=n}function QS(e){return YS[e]}function ZS(e){delete YS[e]}function JS(e,t,n,r,a){const s=t.inputParams[e];if(s&&void 0!==s.inputIndexStart){const e=s.inputIndexStart,i=0===s.inputIndexEnd?void 0:void 0===s.inputIndexEnd?e+1:s.inputIndexEnd;if("tensor"===s.type)return eI(t.inputNames[s.inputIndexStart],n,r,a);if("tensors"===s.type){return t.inputNames.slice(e,i).map(e=>eI(e,n,r,a))}const o=eI(t.inputNames.slice(e)[0],n,r,a),u=o.dataSync();return"number"===s.type?u[0]:va.toNestedArray(o.shape,u)}const i=t.attrParams[e];return i&&i.value}function eI(e,t,n,r){const[a,s]=rI(e);if(null!=r){const e=r.getHashTableHandleByName(a);if(null!=e)return e}const i=n.currentContextIds.find(e=>!!t[nI(a,e)]);return void 0!==i?t[nI(a,i)][s]:void 0}function tI(e,t){const[n,r,a]=rI(e);return[nI(n,t&&t.currentContextId),r,a]}function nI(e,t){return t?`${e}-${t}`:e}function rI(e){const t=e.split(":");if(1===t.length)return[e,0,void 0];const n=t[0],r=3===t.length?t[1]:void 0;return[n,Number(t[t.length-1]),r]}function aI(e,t,n){let r=JS("pad",e,t,n);if("explicit"===r){r=JS("explicitPaddings",e,t,n);const a=[[0,0],[0,0],[0,0],[0,0]];for(let e=0;e<4;e++)a[e][0]=r[2*e],a[e][1]=r[2*e+1];return a}return r}function sI(e){return e.kept?e:Ei(e)}var iI={};w(iI,{json:()=>oI});var oI=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],uI={};w(uI,{json:()=>lI});var lI=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Prod",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axes",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],cI={};w(cI,{json:()=>hI});var hI=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]}],pI={};w(pI,{json:()=>dI});var dI=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number"}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}],fI={};w(fI,{json:()=>mI});var mI=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}],gI={};w(gI,{json:()=>yI});var yI=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],bI={};w(bI,{json:()=>vI});var vI=[{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}],xI={};w(xI,{json:()=>wI});var wI=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}],kI={};w(kI,{json:()=>SI});var SI=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]}],II={};w(II,{json:()=>AI});var AI=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]}],EI={};w(EI,{json:()=>NI});var NI=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],CI={};w(CI,{json:()=>TI});var TI=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]}],RI={};w(RI,{json:()=>_I});var _I=[{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"SparseToDense",category:"normalization",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!0,notSupported:!0}]}],FI={};w(FI,{json:()=>MI});var MI=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}],DI={};w(DI,{json:()=>OI});var OI=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]}],LI={};w(LI,{json:()=>PI});var PI=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}],$I={};w($I,{json:()=>BI});var BI=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}],zI={};w(zI,{json:()=>UI});var UI=[{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}],WI={};w(WI,{json:()=>VI});var VI=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}],jI=class{static get Instance(){return this._instance||(this._instance=new this)}constructor(){const e=[].concat(...[iI,uI,cI,pI,fI,gI,bI,xI,kI,II,EI,CI,RI,FI,DI,LI,$I,zI,WI].map(e=>e.json));this.opMappers=e.reduce((e,t)=>(e[t.tfOpName]=t,e),{})}transformGraph(e,t={}){const n=e.node,r=[],a=[],s=[],i=n.reduce((e,t)=>(e[t.name]=this.mapNode(t),t.op.startsWith("Placeholder")?r.push(e[t.name]):"Const"===t.op?a.push(e[t.name]):null!=t.input&&0!==t.input.length||s.push(e[t.name]),e),{});let o=[];const u=[];let l={},c={};null!=t&&(l=this.mapSignatureEntries(t.inputs),c=this.mapSignatureEntries(t.outputs));const h=Object.keys(i);h.forEach(e=>{const t=i[e];t.inputNames.forEach((e,n)=>{const[r,,a]=tI(e),s=i[r];if(null!=s.outputs){const e=s.outputs.indexOf(a);if(-1!==e){const a=`${r}:${e}`;t.inputNames[n]=a}}t.inputs.push(s),s.children.push(t)})}),0===Object.keys(c).length?h.forEach(e=>{const t=i[e];0===t.children.length&&u.push(t)}):Object.keys(c).forEach(e=>{const[t]=tI(e),n=i[t];null!=n&&(n.signatureKey=c[e],u.push(n))}),Object.keys(l).length>0?Object.keys(l).forEach(e=>{const[t]=tI(e),n=i[t];n&&(n.signatureKey=l[e],o.push(n))}):o=r;let p={};null!=e.library&&null!=e.library.function&&(p=e.library.function.reduce((e,t)=>(e[t.signature.name]=this.mapFunction(t),e),{}));const d={nodes:i,inputs:o,outputs:u,weights:a,placeholders:r,signature:t,functions:p};return s.length>0&&(d.initNodes=s),d}mapSignatureEntries(e){return Object.keys(e||{}).reduce((t,n)=>(t[e[n].name]=n,t),{})}mapNode(e){const t=QS(e.op)||this.opMappers[e.op]||{};null==e.attr&&(e.attr={});const n={name:e.name,op:e.op,category:t.category,inputNames:(e.input||[]).map(e=>e.startsWith("^")?e.substr(1):e),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:e.attr,outputs:t.outputs};return null!=t.inputs&&(n.inputParams=t.inputs.reduce((e,t)=>(e[t.name]={type:t.type,inputIndexStart:t.start,inputIndexEnd:t.end},e),{})),null!=t.attrs&&(n.attrParams=t.attrs.reduce((t,n)=>{const r=n.type;let a=void 0;switch(n.type){case"string":a=HI(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=HI(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"string[]":a=nA(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=nA(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"number":a=KI(e.attr,n.tfName,n.defaultValue||0),void 0===a&&n.tfDeprecatedName&&(a=KI(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"number[]":a=tA(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=tA(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"bool":a=qI(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=qI(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"bool[]":a=aA(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=aA(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"shape":a=eA(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=eA(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"shape[]":a=rA(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=rA(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"dtype":a=QI(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=QI(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"dtype[]":a=ZI(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=ZI(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"func":a=XI(e.attr,n.tfName,n.defaultValue),void 0===a&&n.tfDeprecatedName&&(a=XI(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${n.type} for op: ${e.op}`)}return t[n.name]={value:a,type:r},t},{})),n}mapFunction(e){const t=e.nodeDef,n=[];let r={};null!=t&&(r=t.reduce((e,t)=>(e[t.name]=this.mapNode(t),"Const"===t.op&&n.push(e[t.name]),e),{}));const a=[],s=[];e.signature.inputArg.forEach(e=>{const[t]=tI(e.name),n={name:t,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:YI(e.type),type:"dtype"}},children:[]};n.signatureKey=e.name,a.push(n),r[t]=n});Object.keys(r).forEach(e=>{const t=r[e];t.inputNames.forEach((e,n)=>{const[a,,s]=tI(e),i=r[a];if(null!=i.outputs){const e=i.outputs.indexOf(s);if(-1!==e){const r=`${a}:${e}`;t.inputNames[n]=r}}t.inputs.push(i),i.children.push(t)})});const i=e.ret;e.signature.outputArg.forEach(e=>{const[t,n]=tI(i[e.name]),a=r[t];null!=a&&(a.defaultOutput=n,s.push(a))});const o=this.mapArgsToSignature(e);return{nodes:r,inputs:a,outputs:s,weights:n,placeholders:[],signature:o}}mapArgsToSignature(e){return{methodName:e.signature.name,inputs:e.signature.inputArg.reduce((e,t)=>(e[t.name]=this.mapArgToTensorInfo(t),e),{}),outputs:e.signature.outputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n,e.ret),t),{})}}mapArgToTensorInfo(e,t){let n=e.name;return null!=t&&(n=t[n]),{name:n,dtype:e.type}}};function GI(e,t){const n=Array.isArray(e)?String.fromCharCode.apply(null,e):function(e){const t=Ye().global;if(void 0!==t.atob)return t.atob(e);if(void 0!==i)return new i(e,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}(e);return t?n:n.toLowerCase()}function HI(e,t,n,r=!1){const a=e[t];return null!=a?GI(a.s,r):n}function qI(e,t,n){const r=e[t];return r?r.b:n}function KI(e,t,n){const r=e[t]||{},a=null!=r.i?r.i:null!=r.f?r.f:n;return"number"==typeof a?a:parseInt(a,10)}function YI(e){switch("string"==typeof e&&(e=VS[e]),e){case VS.DT_FLOAT:case VS.DT_HALF:return"float32";case VS.DT_INT32:case VS.DT_INT64:case VS.DT_INT8:case VS.DT_UINT8:return"int32";case VS.DT_BOOL:return"bool";case VS.DT_DOUBLE:return"float32";case VS.DT_STRING:return"string";default:return null}}function XI(e,t,n){const r=e[t];return r&&r.func?r.func.name:n}function QI(e,t,n){const r=e[t];return r&&r.type?YI(r.type):n}function ZI(e,t,n){const r=e[t];return r&&r.list&&r.list.type?r.list.type.map(e=>YI(e)):n}function JI(e){if(!e.unknownRank)return null!=e.dim?e.dim.map(e=>"number"==typeof e.size?e.size:parseInt(e.size,10)):[]}function eA(e,t,n){const r=e[t];return r&&r.shape?JI(r.shape):n}function tA(e,t,n){const r=e[t];return r?((r.list.f&&r.list.f.length?r.list.f:r.list.i)||[]).map(e=>"number"==typeof e?e:parseInt(e,10)):n}function nA(e,t,n,r=!1){const a=e[t];return a&&a.list&&a.list.s?a.list.s.map(e=>GI(e,r)):n}function rA(e,t,n){const r=e[t];return r&&r.list&&r.list.shape?r.list.shape.map(e=>JI(e)):n}function aA(e,t,n){const r=e[t];return r&&r.list&&r.list.b?r.list.b:n}function sA(e,t,n=""){if("number"!=typeof e&&"number"!=typeof t){va.assert(e.length===t.length,()=>n+` Shapes ${e} and ${t} must match`);for(let r=0;r<e.length;r++){const a=e[r],s=t[r];va.assert(a<0||s<0||a===s,()=>n+` Shapes ${e} and ${t} must match`)}}}function iA(e){return"number"!=typeof e&&!e.some(e=>e<0)}function oA(e,t,n){let r=uA(e,n);const a=!iA(r);if(a&&0===t.length)throw new Error("Tried to calculate elements of an empty list with non-fully-defined elementShape: "+r);if(a&&t.forEach(e=>{r=uA(e.shape,r)}),!iA(r))throw new Error("Non-fully-defined elementShape: "+r);return r}function uA(e,t){if("number"==typeof e)return t;if("number"==typeof t)return e;if(e.length!==t.length)throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);const n=[];for(let r=0;r<e.length;++r){const a=e[r],s=t[r];if(a>=0&&s>=0&&a!==s)throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);n[r]=a>=0?a:s}return n}var lA=class{constructor(e,t,n,r=-1){this.tensors=e,this.elementShape=t,this.elementDtype=n,null!=e&&e.forEach(e=>{if(n!==e.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${e.dtype}`);sA(t,e.shape,"TensorList shape mismatch: "),su(e)}),this.idTensor=Ah(0),this.maxNumElements=r,su(this.idTensor)}get id(){return this.idTensor.id}copy(){return new lA([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(e){this.tensors.forEach(t=>{null!=e&&e.has(t.id)||t.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(e,t,n=-1){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(-1!==n&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);sA(e,this.elementShape,"TensorList shape mismatch: ");const r=oA(this.elementShape,this.tensors,e);return ru(()=>{const e=this.tensors.map(e=>qu(e,r));return Gh(e,0)})}popBack(e,t){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(0===this.size())throw new Error("Trying to pop from an empty list.");const n=oA(this.elementShape,this.tensors,e),r=this.tensors.pop();return sA(r.shape,e,"TensorList shape mismatch: "),qu(r,n)}pushBack(e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(sA(e.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");su(e),this.tensors.push(e)}resize(e){if(e<0)throw new Error("TensorListResize expects size to be non-negative. Got: "+e);if(-1!==this.maxNumElements&&e>this.maxNumElements)throw new Error(`TensorListResize input size ${e} is greater maxNumElement ${this.maxNumElements}.`);this.tensors.length=e}getItem(e,t,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(e<0||e>this.tensors.length)throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);if(null==this.tensors[e])throw new Error(`element at index ${e} is null.`);sA(this.tensors[e].shape,t,"TensorList shape mismatch: ");const r=oA(this.elementShape,this.tensors,t);return qu(this.tensors[e],r)}setItem(e,t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(e<0||-1!==this.maxNumElements&&e>=this.maxNumElements)throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);sA(this.elementShape,t.shape,"TensorList shape mismatch: "),su(t),this.tensors[e]=t}gather(e,t,n){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);sA(this.elementShape,n,"TensorList shape mismatch: "),e=e.slice(0,this.size());const r=oA(this.elementShape,this.tensors,n);return 0===e.length?Ls([],[0].concat(r)):ru(()=>{const t=e.map(e=>qu(this.tensors[e],r));return Gh(t,0)})}concat(e,t){if(e&&e!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);sA(this.elementShape,t,"TensorList shape mismatch: ");const n=oA(this.elementShape,this.tensors,t);return 0===this.size()?Ls([],[0].concat(n)):ru(()=>{const e=this.tensors.map(e=>qu(e,n));return Xu(e,0)})}};var cA=async(e,t,n)=>{switch(e.op){case"If":case"StatelessIf":{const r=JS("thenBranch",e,t,n),a=JS("elseBranch",e,t,n),s=JS("cond",e,t,n),i=JS("args",e,t,n);return(await s.data())[0]?n.functionMap[r].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap):n.functionMap[a].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap)}case"While":case"StatelessWhile":{const r=JS("body",e,t,n),a=JS("cond",e,t,n),s=JS("args",e,t,n),i=await n.functionMap[a].executeFunctionAsync(s,n.tensorArrayMap,n.tensorListMap),o=s.map(e=>e.id);let u=await i[0].data();i.forEach(e=>{e.kept||-1!==o.indexOf(e.id)||e.dispose()});let l=s;for(;u[0];){const e=l;l=await n.functionMap[r].executeFunctionAsync(l,n.tensorArrayMap,n.tensorListMap);const t=l.map(e=>e.id);e.forEach(e=>{e.kept||-1!==o.indexOf(e.id)||-1!==t.indexOf(e.id)||e.dispose()});const s=await n.functionMap[a].executeFunctionAsync(l,n.tensorArrayMap,n.tensorListMap);u=await s[0].data(),s.forEach(e=>{e.kept||-1!==o.indexOf(e.id)||-1!==t.indexOf(e.id)||e.dispose()})}return l}case"LoopCond":return[sI(JS("pred",e,t,n))];case"Switch":{const r=JS("pred",e,t,n);let a=JS("data",e,t,n);return a.kept||(a=sI(a)),(await r.data())[0]?[void 0,a]:[a,void 0]}case"Merge":{const r=e.inputNames.find(e=>void 0!==eI(e,t,n));if(r){return[sI(eI(r,t,n))]}return}case"Enter":{const r=JS("frameName",e,t,n),a=JS("tensor",e,t,n);return n.enterFrame(r),[sI(a)]}case"Exit":{const r=JS("tensor",e,t,n);return n.exitFrame(),[sI(r)]}case"NextIteration":{const r=JS("tensor",e,t,n);return n.nextIteration(),[sI(r)]}case"TensorArrayV3":{const r=JS("size",e,t,n),a=JS("dtype",e,t,n),s=JS("elementShape",e,t,n),i=JS("dynamicSize",e,t,n),o=JS("clearAfterRead",e,t,n),u=JS("identicalElementShapes",e,t,n),l=JS("name",e,t,n),c=new class{constructor(e,t,n,r,a,s,i){this.name=e,this.dtype=t,this.maxSize=n,this.elementShape=r,this.identicalElementShapes=a,this.dynamicSize=s,this.clearAfterRead=i,this.tensors=[],this.closed_=!1,this.idTensor=Ah(0),su(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(e){this.tensors.forEach(t=>{null!=e&&e.has(t.tensor.id)||t.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||e>=this.size())throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);const t=this.tensors[e];if(t.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(t.cleared=!0),t.read=!0,t.tensor}readMany(e){return e.map(e=>this.read(e))}write(e,t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||!this.dynamicSize&&e>=this.maxSize)throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);const n=this.tensors[e]||{};if(t.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},\n          because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);if(0!==this.size()||null!=this.elementShape&&0!==this.elementShape.length||(this.elementShape=t.shape),sA(this.elementShape,t.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${e}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);n.tensor=t,su(t),n.written=!0,this.tensors[e]=n}writeMany(e,t){if(e.length!==t.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);e.forEach((e,n)=>this.write(e,t[n]))}gather(e,t){if(t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);if(e)e=e.slice(0,this.size());else{e=[];for(let t=0;t<this.size();t++)e.push(t)}if(0===e.length)return Ls([],[0].concat(this.elementShape));const n=this.readMany(e);return sA(this.elementShape,n[0].shape,"TensorArray shape mismatch: "),Gh(n,0)}concat(e){if(e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);if(0===this.size())return Ls([],[0].concat(this.elementShape));const t=[];for(let e=0;e<this.size();e++)t.push(e);const n=this.readMany(t);return sA(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),Xu(n,0)}scatter(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);if(e.length!==t.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);const n=Math.max(...e);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(e,ap(t,0))}split(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);let n=0;const r=e.map(e=>(n+=e,n));if(n!==t.shape[0])throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${n}, and tensor's shape is: ${t.shape}`);if(!this.dynamicSize&&e.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);const a=0===n?0:t.size/n,s=[];ru(()=>{t=qu(t,[1,n,a]);for(let n=0;n<e.length;++n){const i=[0,0===n?0:r[n-1],0],o=[1,e[n],a];s[n]=qu(Zu(t,i,o),this.elementShape)}return s});const i=[];for(let t=0;t<e.length;t++)i[t]=t;this.writeMany(i,s)}}(l,a,r,s,u,i,o);return n.addTensorArray(c),[c.idTensor,Ah(1)]}case"TensorArrayWriteV3":{const r=JS("tensorArrayId",e,t,n),a=JS("index",e,t,n),s=JS("tensor",e,t,n),i=n.getTensorArray(r.id);return i.write(a,s),[i.idTensor]}case"TensorArrayReadV3":{const r=JS("tensorArrayId",e,t,n),a=JS("index",e,t,n);return[n.getTensorArray(r.id).read(a)]}case"TensorArrayGatherV3":{const r=JS("tensorArrayId",e,t,n),a=JS("indices",e,t,n),s=JS("dtype",e,t,n);return[n.getTensorArray(r.id).gather(a,s)]}case"TensorArrayScatterV3":{const r=JS("tensorArrayId",e,t,n),a=JS("indices",e,t,n),s=JS("tensor",e,t,n),i=n.getTensorArray(r.id);return i.scatter(a,s),[i.idTensor]}case"TensorArrayConcatV3":{const r=JS("tensorArrayId",e,t,n),a=n.getTensorArray(r.id),s=JS("dtype",e,t,n);return[a.concat(s)]}case"TensorArraySplitV3":{const r=JS("tensorArrayId",e,t,n),a=JS("tensor",e,t,n),s=JS("lengths",e,t,n),i=n.getTensorArray(r.id);return i.split(s,a),[i.idTensor]}case"TensorArraySizeV3":{const r=JS("tensorArrayId",e,t,n);return[Ah(n.getTensorArray(r.id).size(),"int32")]}case"TensorArrayCloseV3":{const r=JS("tensorArrayId",e,t,n),a=n.getTensorArray(r.id);return a.clearAndClose(),[a.idTensor]}case"TensorListSetItem":{const r=JS("tensorListId",e,t,n),a=JS("index",e,t,n),s=JS("tensor",e,t,n),i=n.getTensorList(r.id);return i.setItem(a,s),[i.idTensor]}case"TensorListGetItem":{const r=JS("tensorListId",e,t,n),a=JS("index",e,t,n),s=JS("elementShape",e,t,n),i=JS("elementDType",e,t,n);return[n.getTensorList(r.id).getItem(a,s,i)]}case"TensorListScatterV2":case"TensorListScatter":{const r=JS("indices",e,t,n),a=function(e,t,n,r){if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);const a=Math.max(...t);if(null!=r&&-1!==r&&a>=r)throw new Error(`Max index must be < array size (${a}  vs. ${r})`);const s=new lA([],n,e.dtype,r),i=ap(e,0);return t.forEach((e,t)=>{s.setItem(e,i[t])}),s}(JS("tensor",e,t,n),r,JS("elementShape",e,t,n),JS("numElements",e,t,n));return n.addTensorList(a),[a.idTensor]}case"TensorListReserve":case"EmptyTensorList":{const r=JS("elementShape",e,t,n),a=JS("elementDType",e,t,n);let s;s="TensorListReserve"===e.op?"numElements":"maxNumElements";const i=function(e,t,n){return new lA([],e,t,n)}(r,a,JS(s,e,t,n));return n.addTensorList(i),[i.idTensor]}case"TensorListGather":{const r=JS("tensorListId",e,t,n),a=JS("indices",e,t,n),s=JS("elementShape",e,t,n),i=JS("elementDType",e,t,n);return[n.getTensorList(r.id).gather(a,i,s)]}case"TensorListStack":{const r=JS("tensorListId",e,t,n),a=JS("elementShape",e,t,n),s=JS("elementDType",e,t,n),i=JS("numElements",e,t,n);return[n.getTensorList(r.id).stack(a,s,i)]}case"TensorListFromTensor":{const r=function(e,t,n){const r=e.dtype;if(e.shape.length<1)throw new Error("Tensor must be at least a vector, but saw shape: "+e.shape);if(e.dtype!==n)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${n}`);sA(e.shape.slice(1),t,"TensorList shape mismatch: ");const a=ap(e);return new lA(a,t,r)}(JS("tensor",e,t,n),JS("elementShape",e,t,n),JS("elementDType",e,t,n));return n.addTensorList(r),[r.idTensor]}case"TensorListConcat":{const r=JS("tensorListId",e,t,n),a=n.getTensorList(r.id),s=JS("dtype",e,t,n),i=JS("elementShape",e,t,n);return[a.concat(s,i)]}case"TensorListPushBack":{const r=JS("tensorListId",e,t,n),a=JS("tensor",e,t,n),s=n.getTensorList(r.id);return s.pushBack(a),[s.idTensor]}case"TensorListPopBack":{const r=JS("tensorListId",e,t,n),a=JS("elementShape",e,t,n),s=JS("elementDType",e,t,n);return[n.getTensorList(r.id).popBack(a,s)]}case"TensorListSplit":{const r=JS("tensor",e,t,n),a=JS("elementShape",e,t,n),s=function(e,t,n){let r=0;const a=t.map(e=>(r+=e,r));if(r!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${r}, and tensor's shape is: ${e.shape}`);const s=uA(e.shape.slice(1),n),i=0===r?0:e.size/r,o=ru(()=>{const n=[];e=qu(e,[1,r,i]);for(let r=0;r<t.length;++r){const o=[0,0===r?0:a[r-1],0],u=[1,t[r],i];n[r]=qu(Zu(e,o,u),s)}return e.dispose(),n}),u=new lA([],n,e.dtype,t.length);for(let e=0;e<o.length;e++)u.setItem(e,o[e]);return u}(r,JS("lengths",e,t,n),a);return n.addTensorList(s),[s.idTensor]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function hA(e,t,n){const[r,a]=JS("fusedOps",e,t,n),s="biasadd"===r,i=!s,o="prelu"===a,u="fusedbatchnorm"===r,l=JS("numArgs",e,t,n);if(s){if(o&&2!==l)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&s&&1!==l)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(u)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");const c=JS("strides",e,t,n),h=aI(e,t,n),p=JS("dataFormat",e,t,n).toUpperCase(),d=JS("dilations",e,t,n);let[f,m]=JS("args",e,t,n);i&&(m=f,f=void 0);return{stride:c,pad:h,dataFormat:p,dilations:d,biasArg:f,preluArg:m,activationFunc:a,leakyreluAlpha:JS("leakyreluAlpha",e,t,n)}}function pA(e,t,n){return{boxes:JS("boxes",e,t,n),scores:JS("scores",e,t,n),maxOutputSize:JS("maxOutputSize",e,t,n),iouThreshold:JS("iouThreshold",e,t,n),scoreThreshold:JS("scoreThreshold",e,t,n),softNmsSigma:JS("softNmsSigma",e,t,n)}}var dA=async(e,t,n,r)=>{switch(e.op){case"HashTable":case"HashTableV2":{const a=JS("keyDType",e,t,n),s=JS("valueDType",e,t,n),i=new class{constructor(e,t){this.keyDType=e,this.valueDType=t,this.handle=Ah(0),this.tensorMap=new Map,su(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return Ah(this.size(),"int32")}async import(e,t){this.checkKeyAndValueTensor(e,t);const n=await e.data();return this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),ru(()=>{const e=ap(t),r=n.length,a=e.length;va.assert(r===a,()=>`The number of elements doesn't match, keys has ${r} elements, the values has ${a} elements.`);for(let t=0;t<r;t++){const r=n[t],a=e[t];su(a),this.tensorMap.set(r,a)}return this.handle})}async find(e,t){this.checkKeyAndValueTensor(e,t);const n=await e.data();return ru(()=>{const e=[];for(let r=0;r<n.length;r++){const a=n[r],s=this.findWithDefault(a,t);e.push(s)}return Gh(e)})}findWithDefault(e,t){const n=this.tensorMap.get(e);return null!=n?n:t}checkKeyAndValueTensor(e,t){if(e.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);if(t.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)}}(a,s);return r.addHashTable(e.name,i),[i.handle]}case"LookupTableImport":case"LookupTableImportV2":{const a=JS("tableHandle",e,t,n,r),s=JS("keys",e,t,n),i=JS("values",e,t,n),o=r.getHashTableById(a.id);return[await o.import(s,i)]}case"LookupTableFind":case"LookupTableFindV2":{const a=JS("tableHandle",e,t,n,r),s=JS("keys",e,t,n),i=JS("defaultValue",e,t,n),o=r.getHashTableById(a.id);return[await o.find(s,i)]}case"LookupTableSize":case"LookupTableSizeV2":{const a=JS("tableHandle",e,t,n,r);return[r.getHashTableById(a.id).tensorSize()]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function fA(e,t,n,r){const a=((e,t,n)=>{switch(e.category){case"arithmetic":return ru(()=>((e,t,n)=>{switch(e.op){case"BiasAdd":case"AddV2":case"Add":return[gu(JS("a",e,t,n),JS("b",e,t,n))];case"AddN":return[Su(JS("tensors",e,t,n))];case"FloorMod":case"Mod":return[Vc(JS("a",e,t,n),JS("b",e,t,n))];case"Mul":return[vu(JS("a",e,t,n),JS("b",e,t,n))];case"RealDiv":case"Div":return[bu(JS("a",e,t,n),JS("b",e,t,n))];case"DivNoNan":return[Ml(JS("a",e,t,n),JS("b",e,t,n))];case"FloorDiv":return[yu(JS("a",e,t,n),JS("b",e,t,n))];case"Sub":return[gc(JS("a",e,t,n),JS("b",e,t,n))];case"Minimum":return[Uc(JS("a",e,t,n),JS("b",e,t,n))];case"Maximum":return[Oc(JS("a",e,t,n),JS("b",e,t,n))];case"Pow":return[ah(JS("a",e,t,n),JS("b",e,t,n))];case"SquaredDifference":return[Vh(JS("a",e,t,n),JS("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"basic_math":return ru(()=>((e,t,n)=>{switch(e.op){case"Abs":case"ComplexAbs":return[xu(JS("x",e,t,n))];case"Acos":return[wu(JS("x",e,t,n))];case"Acosh":return[ku(JS("x",e,t,n))];case"Asin":return[Cu(JS("x",e,t,n))];case"Asinh":return[Tu(JS("x",e,t,n))];case"Atan":return[Ru(JS("x",e,t,n))];case"Atan2":return[_u(JS("x",e,t,n),JS("y",e,t,n))];case"Atanh":return[Fu(JS("x",e,t,n))];case"Ceil":return[ll(JS("x",e,t,n))];case"Complex":return[Ds(JS("real",e,t,n),JS("imag",e,t,n))];case"Cos":return[kl(JS("x",e,t,n))];case"Cosh":return[Sl(JS("x",e,t,n))];case"Elu":return[Ll(JS("x",e,t,n))];case"Erf":return[Pl(JS("x",e,t,n))];case"Exp":return[$l(JS("x",e,t,n))];case"Expm1":return[zl(JS("x",e,t,n))];case"Floor":return[jl(JS("x",e,t,n))];case"Log":return[rc(JS("x",e,t,n))];case"Log1p":return[ac(JS("x",e,t,n))];case"Imag":return[Kl(JS("x",e,t,n))];case"Neg":return[pc(JS("x",e,t,n))];case"Reciprocal":return[mh(JS("x",e,t,n))];case"Real":return[fh(JS("x",e,t,n))];case"Relu":return[gh(JS("x",e,t,n))];case"Round":return[Sh(JS("x",e,t,n))];case"Selu":return[Eh(JS("x",e,t,n))];case"Sigmoid":return[Qu(JS("x",e,t,n))];case"Sin":return[Rh(JS("x",e,t,n))];case"Sign":return[Th(JS("x",e,t,n))];case"Sinh":return[_h(JS("x",e,t,n))];case"Softplus":return[dc(JS("x",e,t,n))];case"Sqrt":return[Wh(JS("x",e,t,n))];case"Square":return[jc(JS("x",e,t,n))];case"Tanh":return[Ju(JS("x",e,t,n))];case"Tan":return[Kh(JS("x",e,t,n))];case"ClipByValue":return[cl(JS("x",e,t,n),JS("clipValueMin",e,t,n),JS("clipValueMax",e,t,n))];case"Relu6":return[yh(JS("x",e,t,n))];case"Rsqrt":return[Ih(eI(e.inputNames[0],t,n))];case"Prod":return[ih(JS("x",e,t,n),JS("axes",e,t,n))];case"LeakyRelu":return[Zl(JS("x",e,t,n),JS("alpha",e,t,n))];case"Prelu":return[sh(JS("x",e,t,n),JS("alpha",e,t,n))];case"IsNan":return[Ql(eI(e.inputNames[0],t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"control":return cA(e,t,n);case"convolution":return ru(()=>((e,t,n)=>{switch(e.op){case"Conv1D":{const r=JS("stride",e,t,n),a=JS("pad",e,t,n),s=JS("dataFormat",e,t,n).toUpperCase(),i=JS("dilation",e,t,n);return[gl(JS("x",e,t,n),JS("filter",e,t,n),r,a,s,i)]}case"Conv2D":{const r=JS("strides",e,t,n),a=aI(e,t,n),s=JS("dataFormat",e,t,n).toUpperCase(),i=JS("dilations",e,t,n);return[ml(JS("x",e,t,n),JS("filter",e,t,n),[r[1],r[2]],a,s,[i[1],i[2]])]}case"_FusedConv2D":{const{stride:r,pad:a,dataFormat:s,dilations:i,biasArg:o,preluArg:u,activationFunc:l,leakyreluAlpha:c}=hA(e,t,n);return[bp.conv2d({x:JS("x",e,t,n),filter:JS("filter",e,t,n),strides:[r[1],r[2]],pad:a,dataFormat:s,dilations:[i[1],i[2]],bias:o,activation:l,preluActivationWeights:u,leakyreluAlpha:c})]}case"FusedDepthwiseConv2dNative":{const{stride:r,pad:a,dataFormat:s,dilations:i,biasArg:o,preluArg:u,activationFunc:l,leakyreluAlpha:c}=hA(e,t,n);return[bp.depthwiseConv2d({x:JS("x",e,t,n),filter:JS("filter",e,t,n),strides:[r[1],r[2]],pad:a,dataFormat:s,dilations:[i[1],i[2]],bias:o,activation:l,preluActivationWeights:u,leakyreluAlpha:c})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{const r=JS("outputShape",e,t,n),a=JS("strides",e,t,n),s=aI(e,t,n);return[bl(JS("x",e,t,n),JS("filter",e,t,n),r,[a[1],a[2]],s)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{const r=JS("strides",e,t,n),a=aI(e,t,n),s=JS("dilations",e,t,n),i=JS("dataFormat",e,t,n).toUpperCase();return[Nl(JS("input",e,t,n),JS("filter",e,t,n),[r[1],r[2]],a,i,[s[1],s[2]])]}case"Conv3D":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("dataFormat",e,t,n).toUpperCase(),i=JS("dilations",e,t,n);return[vl(JS("x",e,t,n),JS("filter",e,t,n),[r[1],r[2],r[3]],a,s,[i[1],i[2],i[3]])]}case"AvgPool":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("kernelSize",e,t,n);return[Ku(JS("x",e,t,n),[s[1],s[2]],[r[1],r[2]],a)]}case"MaxPool":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("kernelSize",e,t,n);return[Fc(JS("x",e,t,n),[s[1],s[2]],[r[1],r[2]],a)]}case"MaxPoolWithArgmax":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("kernelSize",e,t,n),i=JS("includeBatchInIndex",e,t,n),{result:o,indexes:u}=Dc(JS("x",e,t,n),[s[1],s[2]],[r[1],r[2]],a,i);return[o,u]}case"AvgPool3D":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("kernelSize",e,t,n);return[Yu(JS("x",e,t,n),[s[1],s[2],s[3]],[r[1],r[2],r[3]],a)]}case"MaxPool3D":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("kernelSize",e,t,n);return[Mc(JS("x",e,t,n),[s[1],s[2],s[3]],[r[1],r[2],r[3]],a)]}case"Dilation2D":{const r=JS("strides",e,t,n),a=JS("pad",e,t,n),s=JS("dilations",e,t,n),i=r[1],o=r[2],u=s[1],l=s[2];return[Tl(JS("x",e,t,n),JS("filter",e,t,n),[i,o],a,[u,l],"NHWC")]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"creation":return ru(()=>((e,t,n)=>{switch(e.op){case"Fill":{const r=JS("shape",e,t,n),a=JS("dtype",e,t,n);return[Vl(r,JS("value",e,t,n),a)]}case"LinSpace":return[tc(JS("start",e,t,n),JS("stop",e,t,n),JS("num",e,t,n))];case"Multinomial":{const r=JS("logits",e,t,n),a=JS("numSamples",e,t,n),s=JS("seed",e,t,n);return[qc(r,a,s)]}case"OneHot":{const r=JS("indices",e,t,n),a=JS("depth",e,t,n),s=JS("onValue",e,t,n),i=JS("offValue",e,t,n);return[Hi(r,a,s,i)]}case"Ones":return[$c(JS("shape",e,t,n),JS("dtype",e,t,n))];case"OnesLike":return[Yc(JS("x",e,t,n))];case"RandomUniform":return[ph(JS("shape",e,t,n),JS("minval",e,t,n),JS("maxval",e,t,n),JS("dtype",e,t,n))];case"Range":return[dh(JS("start",e,t,n),JS("stop",e,t,n),JS("step",e,t,n),JS("dtype",e,t,n))];case"TruncatedNormal":{const r=JS("shape",e,t,n),a=JS("mean",e,t,n),s=JS("stdDev",e,t,n),i=JS("seed",e,t,n);return[tp(r,a,s,JS("dtype",e,t,n),i)]}case"Zeros":return[Pc(JS("shape",e,t,n),JS("dtype",e,t,n))];case"ZerosLike":return[Fl(JS("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"dynamic":return(async(e,t,n)=>{switch(e.op){case"NonMaxSuppressionV5":{const{boxes:r,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o,softNmsSigma:u}=pA(e,t,n),l=await Nd.nonMaxSuppressionWithScoreAsync(r,a,s,i,o,u);return[l.selectedIndices,l.selectedScores]}case"NonMaxSuppressionV4":{const{boxes:r,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o}=pA(e,t,n),u=JS("padToMaxOutputSize",e,t,n),l=await Nd.nonMaxSuppressionPaddedAsync(r,a,s,i,o,u);return[l.selectedIndices,l.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{const{boxes:r,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o}=pA(e,t,n);return[await Nd.nonMaxSuppressionAsync(r,a,s,i,o)]}case"Where":{const r=Ai(JS("condition",e,t,n),"bool"),a=[await op(r)];return r.dispose(),a}case"ListDiff":return Ch(JS("x",e,t,n),JS("y",e,t,n));default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n);case"evaluation":return ru(()=>((e,t,n)=>{switch(e.op){case"TopKV2":{const r=JS("x",e,t,n),a=JS("k",e,t,n),s=JS("sorted",e,t,n),i=ep(r,a,s);return[i.values,i.indices]}case"Unique":{const r=JS("x",e,t,n),a=np(r);return[a.values,a.indices]}case"UniqueV2":{const r=JS("x",e,t,n),a=JS("axis",e,t,n),s=np(r,a);return[s.values,s.indices]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"image":return ru(()=>((e,t,n)=>{switch(e.op){case"ResizeBilinear":{const r=JS("images",e,t,n),a=JS("size",e,t,n),s=JS("alignCorners",e,t,n),i=JS("halfPixelCenters",e,t,n);return[Nd.resizeBilinear(r,[a[0],a[1]],s,i)]}case"ResizeNearestNeighbor":{const r=JS("images",e,t,n),a=JS("size",e,t,n),s=JS("alignCorners",e,t,n),i=JS("halfPixelCenters",e,t,n);return[Nd.resizeNearestNeighbor(r,[a[0],a[1]],s,i)]}case"CropAndResize":{const r=JS("image",e,t,n),a=JS("boxes",e,t,n),s=JS("boxInd",e,t,n),i=JS("cropSize",e,t,n),o=JS("method",e,t,n),u=JS("extrapolationValue",e,t,n);return[Nd.cropAndResize(r,a,s,i,o,u)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"graph":return ru(()=>((e,t,n)=>{switch(e.op){case"Const":return t[e.name];case"PlaceholderWithDefault":const r=JS("default",e,t,n);return[eI(e.name,t,n)||r];case"Placeholder":return[eI(e.name,t,n)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":return[sI(JS("x",e,t,n))];case"IdentityN":return JS("x",e,t,n).map(e=>sI(e));case"Snapshot":return[sI(JS("x",e,t,n))];case"Shape":return[Yh(JS("x",e,t,n).shape,"int32")];case"ShapeN":return JS("x",e,t,n).map(e=>Yh(e.shape));case"Size":return[Ah(JS("x",e,t,n).size,"int32")];case"Rank":return[Ah(JS("x",e,t,n).rank,"int32")];case"NoOp":return[Ah(1)];case"Print":const a=JS("x",e,t,n),s=JS("data",e,t,n),i=JS("message",e,t,n),o=JS("summarize",e,t,n);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(i);for(let e=0;e<s.length;e++)console.log(Array.prototype.slice.call(s[e].dataSync()).slice(0,o));return[a];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"logical":return ru(()=>((e,t,n)=>{switch(e.op){case"Equal":return[Rl(JS("a",e,t,n),JS("b",e,t,n))];case"NotEqual":return[Kc(JS("a",e,t,n),JS("b",e,t,n))];case"Greater":return[Hl(JS("a",e,t,n),JS("b",e,t,n))];case"GreaterEqual":return[ql(JS("a",e,t,n),JS("b",e,t,n))];case"Less":return[Jl(JS("a",e,t,n),JS("b",e,t,n))];case"LessEqual":return[ec(JS("a",e,t,n),JS("b",e,t,n))];case"LogicalAnd":return[Cc(JS("a",e,t,n),JS("b",e,t,n))];case"LogicalNot":return[Tc(JS("a",e,t,n))];case"LogicalOr":return[Rc(JS("a",e,t,n),JS("b",e,t,n))];case"Select":case"SelectV2":return[_l(JS("condition",e,t,n),JS("a",e,t,n),JS("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"matrices":return ru(()=>((e,t,n)=>{switch(e.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[Gi(JS("a",e,t,n),JS("b",e,t,n),JS("transposeA",e,t,n),JS("transposeB",e,t,n))];case"Einsum":return[Ol(JS("equation",e,t,n),...JS("tensors",e,t,n))];case"Transpose":return[qi(JS("x",e,t,n),JS("perm",e,t,n))];case"_FusedMatMul":const[r,a]=JS("fusedOps",e,t,n),s="biasadd"===r,i="prelu"===a,o=JS("numArgs",e,t,n),u=JS("leakyreluAlpha",e,t,n);if(s){if(i&&2!==o)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!i&&1!==o)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}const[l,c]=JS("args",e,t,n);return[bp.matMul({a:JS("a",e,t,n),b:JS("b",e,t,n),transposeA:JS("transposeA",e,t,n),transposeB:JS("transposeB",e,t,n),bias:l,activation:a,preluActivationWeights:c,leakyreluAlpha:u})];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"normalization":return ru(()=>((e,t,n)=>{switch(e.op){case"FusedBatchNorm":case"FusedBatchNormV2":case"FusedBatchNormV3":return[nl(JS("x",e,t,n),JS("mean",e,t,n),JS("variance",e,t,n),JS("offset",e,t,n),JS("scale",e,t,n),JS("epsilon",e,t,n))];case"LRN":return[nc(JS("x",e,t,n),JS("radius",e,t,n),JS("bias",e,t,n),JS("alpha",e,t,n),JS("beta",e,t,n))];case"Softmax":return[Lh(JS("x",e,t,n))];case"LogSoftmax":return[bc(JS("x",e,t,n))];case"SparseToDense":return[pp(JS("sparseIndices",e,t,n),JS("outputShape",e,t,n),JS("sparseValues",e,t,n),JS("defaultValue",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"reduction":return ru(()=>((e,t,n)=>{switch(e.op){case"Max":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[mc(JS("x",e,t,n),r,a)]}case"Mean":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[Lc(JS("x",e,t,n),r,a)]}case"Min":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[zc(JS("x",e,t,n),r,a)]}case"Sum":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[yc(JS("x",e,t,n),r,a)]}case"All":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[Iu(JS("x",e,t,n),r,a)]}case"Any":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[Au(JS("x",e,t,n),r,a)]}case"ArgMax":{const r=JS("axis",e,t,n);return[Eu(JS("x",e,t,n),r)]}case"ArgMin":{const r=JS("axis",e,t,n);return[Nu(JS("x",e,t,n),r)]}case"Prod":{const r=JS("axis",e,t,n),a=JS("keepDims",e,t,n);return[ih(JS("x",e,t,n),r,a)]}case"Cumsum":{const r=JS("axis",e,t,n),a=JS("exclusive",e,t,n),s=JS("reverse",e,t,n);return[Il(JS("x",e,t,n),r,a,s)]}case"Bincount":const r=JS("x",e,t,n),a=JS("weights",e,t,n),s=JS("size",e,t,n);return[il(r,a,s)];case"DenseBincount":{const r=JS("x",e,t,n),a=JS("weights",e,t,n),s=JS("size",e,t,n),i=JS("binaryOutput",e,t,n);return[Al(r,a,s,i)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"slice_join":return ru(()=>((e,t,n)=>{switch(e.op){case"ConcatV2":case"Concat":{const r=JS("n",e,t,n),a=JS("axis",e,t,n);let s=JS("tensors",e,t,n);return s=s.slice(0,r),[Xu(s,a)]}case"Gather":{const r=JS("x",e,t,n),a=JS("indices",e,t,n);return[Gl(r,Ai(a,"int32"),0)]}case"GatherV2":{const r=JS("axis",e,t,n),a=JS("batchDims",e,t,n),s=JS("x",e,t,n),i=JS("indices",e,t,n);return[Gl(s,Ai(i,"int32"),r,a)]}case"Reverse":{const r=JS("dims",e,t,n),a=[];for(let e=0;e<r.length;e++)r[e]&&a.push(e);const s=JS("x",e,t,n);return[bh(s,a)]}case"ReverseV2":{const r=JS("axis",e,t,n),a=JS("x",e,t,n);return[bh(a,r)]}case"Slice":{const r=JS("begin",e,t,n),a=JS("size",e,t,n);return[Zu(JS("x",e,t,n),r,a)]}case"StridedSlice":{const r=JS("begin",e,t,n),a=JS("end",e,t,n),s=JS("strides",e,t,n),i=JS("beginMask",e,t,n),o=JS("endMask",e,t,n),u=JS("ellipsisMask",e,t,n),l=JS("newAxisMask",e,t,n),c=JS("shrinkAxisMask",e,t,n),h=JS("x",e,t,n);return[qh(h,r,a,s,i,o,u,l,c)]}case"Pack":return ru(()=>{const r=JS("axis",e,t,n),a=JS("tensors",e,t,n),s=a[0].shape,i=jh(a[0]).shape,o=a.map(e=>{const t=va.arraysEqual(e.shape,s);if(!t&&!va.arraysEqual(jh(e).shape,i))throw new Error("the input tensors shape does not match");return t?e:qu(e,s)});return[Gh(o,r)]});case"Unpack":{const r=JS("axis",e,t,n),a=JS("tensor",e,t,n);return ap(a,r)}case"Tile":{const r=JS("reps",e,t,n);return[Ul(JS("x",e,t,n),r)]}case"Split":case"SplitV":{const r=JS("axis",e,t,n),a=JS("numOrSizeSplits",e,t,n),s=JS("x",e,t,n);return zh(s,a,r)}case"ScatterNd":{const r=JS("indices",e,t,n),a=JS("values",e,t,n),s=JS("shape",e,t,n);return[hp(r,a,s)]}case"GatherNd":{const r=JS("x",e,t,n),a=JS("indices",e,t,n);return[dp(r,a)]}case"SparseToDense":{const r=JS("sparseIndices",e,t,n),a=JS("outputShape",e,t,n),s=JS("sparseValues",e,t,n),i=JS("defaultValue",e,t,n);return[pp(r,s,a,s.dtype===i.dtype?i:Ai(i,s.dtype))]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"sparse":return ru(()=>((e,t,n)=>{switch(e.op){case"SparseFillEmptyRows":{const{outputIndices:r,outputValues:a,emptyRowIndicator:s,reverseIndexMap:i}=Rd.sparseFillEmptyRows(JS("indices",e,t,n),JS("values",e,t,n),JS("denseShape",e,t,n),JS("defaultValue",e,t,n));return[r,a,s,i]}case"SparseReshape":{const{outputIndices:r,outputShape:a}=Rd.sparseReshape(JS("inputIndices",e,t,n),JS("inputShape",e,t,n),JS("newShape",e,t,n));return[r,a]}case"SparseSegmentMean":return[Rd.sparseSegmentMean(JS("data",e,t,n),JS("indices",e,t,n),JS("segmentIds",e,t,n))];case"SparseSegmentSum":return[Rd.sparseSegmentSum(JS("data",e,t,n),JS("indices",e,t,n),JS("segmentIds",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"spectral":return ru(()=>((e,t,n)=>{switch(e.op){case"FFT":return[Ph(JS("x",e,t,n))];case"IFFT":return[$h(JS("x",e,t,n))];case"RFFT":return[Uh(JS("x",e,t,n))];case"IRFFT":return[Bh(JS("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"string":return ru(()=>((e,t,n)=>{switch(e.op){case"StringNGrams":{const{nGrams:r,nGramsSplits:a}=_d.stringNGrams(JS("data",e,t,n),JS("dataSplits",e,t,n),JS("separator",e,t,n),JS("nGramWidths",e,t,n),JS("leftPad",e,t,n),JS("rightPad",e,t,n),JS("padWidth",e,t,n),JS("preserveShortSequences",e,t,n));return[r,a]}case"StringSplit":{const{indices:r,values:a,shape:s}=_d.stringSplit(JS("input",e,t,n),JS("delimiter",e,t,n),JS("skipEmpty",e,t,n));return[r,a,s]}case"StringToHashBucketFast":return[_d.stringToHashBucketFast(JS("input",e,t,n),JS("numBuckets",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"transformation":return ru(()=>((e,t,n)=>{switch(e.op){case"Cast":return[Ai(JS("x",e,t,n),JS("dtype",e,t,n))];case"ExpandDims":{const r=JS("axis",e,t,n);return[Bl(JS("x",e,t,n),r)]}case"Squeeze":{const r=JS("axis",e,t,n);return[jh(JS("x",e,t,n),r)]}case"Reshape":return[qu(JS("x",e,t,n),JS("shape",e,t,n))];case"MirrorPad":return[Wc(JS("x",e,t,n),JS("padding",e,t,n),JS("mode",e,t,n))];case"PadV2":case"Pad":return[Qc(JS("x",e,t,n),JS("padding",e,t,n),JS("constantValue",e,t,n))];case"SpaceToBatchND":{const r=JS("blockShape",e,t,n),a=JS("paddings",e,t,n);return[nh(JS("x",e,t,n),r,a)]}case"BatchToSpaceND":{const r=JS("blockShape",e,t,n),a=JS("crops",e,t,n);return[tl(JS("x",e,t,n),r,a)]}case"DepthToSpace":{const r=JS("blockSize",e,t,n),a=JS("dataFormat",e,t,n).toUpperCase();return[El(JS("x",e,t,n),r,a)]}case"BroadcastTo":return[ul(JS("x",e,t,n),JS("shape",e,t,n))];case"BroadcastArgs":return[ol(JS("s0",e,t,n),JS("s1",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"hash_table":return dA(e,t,n,r);case"custom":const a=QS(e.op);if(a&&a.customExecutor)return a.customExecutor(new class{constructor(e,t,n){this.node=e,this.tensorMap=t,this.context=n,this.inputs=[],this.attrs={},this.inputs=e.inputNames.map(e=>this.getInput(e)),null!=e.rawAttrs&&(this.attrs=Object.keys(e.rawAttrs).reduce((e,t)=>(e[t]=this.getAttr(t),e),{}))}getInput(e){return eI(e,this.tensorMap,this.context)}getAttr(e,t){const n=this.node.rawAttrs[e];if(null!=n.tensor)return eI(e,this.tensorMap,this.context);if(null!=n.i||null!=n.f)return KI(this.node.rawAttrs,e,t);if(null!=n.s)return HI(this.node.rawAttrs,e,t);if(null!=n.b)return qI(this.node.rawAttrs,e,t);if(null!=n.shape)return eA(this.node.rawAttrs,e,t);if(null!=n.type)return QI(this.node.rawAttrs,e,t);if(null!=n.list){if(null!=n.list.i||null!=n.list.f)return tA(this.node.rawAttrs,e,t);if(null!=n.list.s)return nA(this.node.rawAttrs,e,t);if(null!=n.list.shape)return rA(this.node.rawAttrs,e,t);if(null!=n.list.b)return aA(this.node.rawAttrs,e,t);if(null!=n.list.type)return ZI(this.node.rawAttrs,e,t)}return t}}(e,t,n));throw TypeError(`Custom op ${e.op} is not registered.`);default:throw TypeError(`Unknown op '${e.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(e,t,n);return va.isPromise(a)?a.then(e=>[].concat(e)):[].concat(a)}var mA=class{constructor(e={},t={},n={},r={}){this.weightMap=e,this.tensorArrayMap=t,this.tensorListMap=n,this.functionMap=r,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(e,t){return{id:e,frameName:t,iterationId:0}}set currentContext(e){this.contexts!==e&&(this.contexts=e,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){const e=[];for(let t=0;t<this.contexts.length-1;t++){const n=this.contexts.slice(0,this.contexts.length-t);e.push(this.contextIdforContexts(n))}e.push(""),this._currentContextIds=e}contextIdforContexts(e){return e?e.map(e=>0===e.id&&0===e.iterationId?"":`${e.frameName}-${e.iterationId}`).join("/"):""}enterFrame(e){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,e)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(!(this.contexts&&this.contexts.length>1))throw new Error("Cannot exit frame, the context is empty");this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift()}nextIteration(){if(!(this.contexts&&this.contexts.length>0))throw new Error("Cannot increase frame iteration, the context is empty");{this.contexts=this.contexts.slice(),this.lastId++;const e=Object.assign({},this.contexts[this.contexts.length-1]);e.iterationId+=1,e.id=this.lastId,this.contexts.splice(-1,1,e),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}}getWeight(e){return this.weightMap[e]}addTensorArray(e){this.tensorArrayMap[e.id]=e}getTensorArray(e){return this.tensorArrayMap[e]}addTensorList(e){this.tensorListMap[e.id]=e}getTensorList(e){return this.tensorListMap[e]}dispose(e){for(const t in this.tensorArrayMap)this.tensorArrayMap[t].clearAndClose(e);for(const t in this.tensorListMap)this.tensorListMap[t].clearAndClose(e)}};function gA(e,t,n,r){const a=new Set,s=[];let i=null,o=null;const u=new Set,l=Object.keys(e).map(e=>rI(e)[0]);let c=[];null!=r&&(c=r.map(e=>rI(e.name)[0]));const h=[...t];for(;h.length>0;){const e=h.pop();(xA(e)||wA(e)||kA(e))&&null==i&&(i=e,o=i.children.map(e=>e.name).filter(e=>a.has(e))),a.add(e.name),null==n[e.name]&&(-1===l.indexOf(e.name)&&-1===c.indexOf(e.name)&&(0!==e.inputs.length?e.inputs.forEach(e=>{u.has(e.name)||(u.add(e.name),h.push(e))}):s.push(e.name)))}return{inputs:e,outputs:t,usedNodes:a,missingInputs:s,dynamicNode:i,syncInputs:o}}var yA=["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"],bA=["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"],vA=["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"];function xA(e){return yA.indexOf(e.op)>=0}function wA(e){return bA.indexOf(e.op)>=0}function kA(e){return vA.indexOf(e.op)>=0}var SA=class{constructor(e,t){this.graph=e,this.parent=t,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=",",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=e.outputs,this._inputs=e.inputs,this._initNodes=e.initNodes,this._signature=e.signature,this._functions=e.functions,null!=e.functions&&Object.keys(e.functions).forEach(t=>{this._functionExecutorMap[t]=new SA(e.functions[t],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(e){const t=Object.keys(e).map(t=>e[t].map(e=>e.id));this._weightIds=[].concat(...t),this._weightMap=e}set resourceManager(e){this._resourceManager=e}get inputs(){return this._inputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(e=>e.signatureKey||e.name)}get outputNodes(){return this._outputs.map(e=>{const t=e.signatureKey||e.name;return e.defaultOutput?`${t}:${e.defaultOutput}`:t})}get functions(){return Object.keys(this._functions).reduce((e,t)=>(e[t]=this._functions[t].signature,e),{})}getCompilationKey(e,t){const n=e.map(e=>e.name).sort(),r=t.map(e=>e.name).sort();return n.join(this.SEPERATOR)+"--"+r.join(this.SEPERATOR)}compile(e,t){const n=gA(e,t,this.weightMap,this._initNodes),{missingInputs:r,dynamicNode:a,syncInputs:s}=n;if(null!=a)throw new Error(`This execution contains the node '${a.name}', which has the dynamic op '${a.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${s}]`);if(r.length>0){const n=t.map(e=>e.name),a=Object.keys(e);throw new Error(`Cannot compute the outputs [${n}] from the provided inputs [${a}]. Missing the following inputs: [${r}]`)}return function(e,t,n){const{usedNodes:r,inputs:a}=n,s=[],i=Object.keys(a).map(e=>rI(e)[0]).map(t=>e.nodes[t]),o=e.initNodes;i.forEach(e=>{r.has(e.name)&&s.push(e)}),e.weights.forEach(e=>{r.has(e.name)&&s.push(e)}),null!=o&&o.forEach(e=>{r.has(e.name)&&s.push(e)});const u=new Set,l=[];for(;s.length>0;){const e=s.pop();u.add(e.name),t[e.name]||l.push(e),e.children.forEach(e=>{!u.has(e.name)&&r.has(e.name)&&e.inputs.every(e=>u.has(e.name))&&s.push(e)})}return l}(this.graph,this.weightMap,n)}execute(e,t){e=this.mapInputs(e);const n=Object.keys(e).sort();this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t);const r=n.map(e=>this.graph.nodes[rI(e)[0]]),a=t.map(e=>rI(e)[0]);let s=a.map(e=>this.graph.nodes[e]);this.resetIntermediateTensors(),0===s.length&&(s=this._outputs);const i=this.getCompilationKey(r,s);let o=this.compiledMap.get(i);null==o&&(o=this.compile(e,s),this.compiledMap.set(i,o));const u={},l={};return ru(()=>{const n=new mA(this.weightMap,u,l,this.functionExecutorMap),r=Object.assign({},this.weightMap);Object.keys(e).forEach(t=>{const[n,a]=rI(t),s=[];s[a]=e[t],r[n]=s});const s=this.getFrozenTensorIds(r),i={};for(let e=0;e<o.length;e++){const t=o[e];if(!r[t.name]){const e=fA(t,r,n,this._resourceManager);if(va.isPromise(e))throw new Error(`The execution of the op '${t.op}' returned a promise. Please use model.executeAsync() instead.`);r[t.name]=e,this.checkTensorForDisposal(t.name,t,r,n,s,a,i)}}return null==this.parent&&n.dispose(s),t.map(e=>eI(e,r,n))})}getFrozenTensorIds(e){const t=[].concat.apply([],Object.keys(e).map(t=>e[t]).map(e=>e.map(e=>e.id)));return new Set(t)}checkTensorForDisposal(e,t,n,r,a,s,i){"control"!==t.category&&-1===s.indexOf(e)&&(n[e].forEach(e=>{null!=e&&(i[e.id]=(i[e.id]||0)+t.children.length)}),t.inputs.forEach(e=>{if("control"!==e.category){const s=function(e,t,n){return t[nI(e,n.currentContextId)]}(e.name,n,r);null!=s&&s.forEach(e=>{if(e&&!e.kept&&!a.has(e.id)){const n=i[e.id];if(1===n){if(this.keepTensorForDebug){const[n,a]=tI(t.name,r);this.intermediateTensors[n]||(this.intermediateTensors[n]=[]),this.intermediateTensors[n][a]=e}else e.dispose();delete i[e.id]}else null!=n&&i[e.id]--}})}}))}async executeAsync(e,t){return this._executeAsync(e,t)}disposeIntermediateTensors(){this.intermediateTensors&&(Object.keys(this.intermediateTensors).forEach(e=>this.intermediateTensors[e].forEach(e=>e.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){this.tensorsMap&&Object.keys(this.tensorsMap).forEach(e=>{this.tensorsMap[e].forEach(e=>{!e||e.kept||e.isDisposed||this.keepIds.has(e.id)||e.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(const e in this.intermediateTensors)this.intermediateTensors[e].forEach(e=>e.dispose()),delete this.intermediateTensors[e]}async _executeAsync(e,t,n=!1,r={},a={}){n||(e=this.mapInputs(e),this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t));try{this.keepTensorForDebug=Ye().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(e){console.warn(e.message)}this.resetIntermediateTensors();const s=new mA(this.weightMap,r,a,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(e,s,t,n);const i=t.map(e=>eI(e,this.tensorsMap,s)),o=i.map(e=>e.id),u=Object.keys(e).map(t=>e[t].id);return this.keepIds=new Set([...o,...u,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),null==this.parent&&s.dispose(this.keepIds),i}async executeFunctionAsync(e,t,n){const r=e.reduce((e,t,n)=>(e[this.inputs[n].name]=t,e),{});return this._executeAsync(r,this.outputNodes,!0,t,n)}async executeWithControlFlow(e,t,n,r){const a=Object.keys(e),s=a.map(e=>this.graph.nodes[rI(e)[0]]),i=n.map(e=>rI(e)[0]);let o=i.map(e=>this.graph.nodes[e]);0===o.length&&(o=this._outputs);const{usedNodes:u,missingInputs:l,dynamicNode:c,syncInputs:h}=gA(e,o,this.weightMap,this._initNodes),p=[...s,...this.graph.weights,...this._initNodes||[]].map(e=>({node:e,contexts:t.currentContext})),d=Object.assign({},this.weightMap);Object.keys(e).forEach(t=>{const[n,r]=rI(t),a=[];a[r]=e[t],d[n]=a});const f={},m=this.getFrozenTensorIds(d),g={};for(;p.length>0;){const e=this.processStack(s,p,t,d,g,m,i,f,u);await Promise.all(e)}null!=c||r||console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");const y=o.filter(e=>!xA(e)&&!eI(e.name,d,t)).map(e=>e.name);if(y.length>0){let e="";throw null!=c&&(e=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${h}]`),new Error(`Cannot compute the outputs [${y}] from the provided inputs [${a}]. Consider providing the following inputs: [${l}]. ${e}`)}return d}processStack(e,t,n,r,a,s,i,o,u){const l=[];for(;t.length>0;){const e=t.pop();n.currentContext=e.contexts;let c="";if("Enter"===e.node.op&&JS("isConstant",e.node,r,n)&&([c]=tI(e.node.name,n)),null==r[e.node.name]){const h=fA(e.node,r,n,this._resourceManager);c||([c]=tI(e.node.name,n));const p=n.currentContext;va.isPromise(h)?l.push(h.then(l=>(r[c]=l,n.currentContext=p,this.checkTensorForDisposal(c,e.node,r,n,s,i,o),this.processChildNodes(e.node,t,n,r,a,u),l))):(r[c]=h,this.checkTensorForDisposal(c,e.node,r,n,s,i,o),this.processChildNodes(e.node,t,n,r,a,u))}else this.processChildNodes(e.node,t,n,r,a,u)}return l}processChildNodes(e,t,n,r,a,s){e.children.forEach(e=>{const[i]=tI(e.name,n);!a[i]&&s.has(e.name)&&("Merge"===e.op?e.inputNames.some(e=>!!eI(e,r,n))&&(a[i]=!0,t.push({contexts:n.currentContext,node:e})):e.inputNames.every(e=>!!eI(e,r,n))&&(a[i]=!0,t.push({contexts:n.currentContext,node:e})))})}dispose(){Object.keys(this.weightMap).forEach(e=>this.weightMap[e].forEach(e=>e.dispose()))}checkInputShapeAndType(e){Object.keys(e).forEach(t=>{const n=e[t],[r]=rI(t),a=this.graph.nodes[r];if(a.attrParams.shape&&a.attrParams.shape.value){const e=a.attrParams.shape.value,t=e.length===n.shape.length&&n.shape.every((t,n)=>-1===e[n]||e[n]===t);va.assert(t,()=>`The shape of dict['${a.name}'] provided in model.execute(dict) must be [${e}], but was [${n.shape}]`)}a.attrParams.dtype&&a.attrParams.dtype.value&&va.assert(n.dtype===a.attrParams.dtype.value,()=>`The dtype of dict['${a.name}'] provided in model.execute(dict) must be ${a.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(e){const t={};for(const n in e)if(null!=this._signature&&null!=this._signature.inputs&&null!=this._signature.inputs[n]){t[this._signature.inputs[n].name]=e[n]}else t[n]=e[n];return t}checkInputs(e){const t=Object.keys(e).filter(e=>{const[t]=rI(e);return null==this.graph.nodes[t]});if(t.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)}mapOutputs(e){return e.map(e=>{if(null!=this._signature&&null!=this._signature.outputs&&null!=this._signature.outputs[e]){return this._signature.outputs[e].name}return e},{})}checkOutputs(e){e.forEach(e=>{const[t]=rI(e);if(!this.graph.nodes[t])throw new Error(`The output '${e}' is not found in the graph`)})}},IA=class{constructor(e,t={}){this.modelUrl=e,this.loadOptions=t,this.version="n/a",null==t&&(this.loadOptions={}),this.resourceManager=new class{constructor(e={},t={}){this.hashTableNameToHandle=e,this.hashTableMap=t}addHashTable(e,t){this.hashTableNameToHandle[e]=t.handle,this.hashTableMap[t.id]=t}getHashTableHandleByName(e){return this.hashTableNameToHandle[e]}getHashTableById(e){return this.hashTableMap[e]}dispose(){for(const e in this.hashTableMap)this.hashTableMap[e].clearAndClose(),delete this.hashTableMap[e];for(const e in this.hashTableNameToHandle)this.hashTableNameToHandle[e].dispose(),delete this.hashTableNameToHandle[e]}}}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}findIOHandler(){const e=this.modelUrl;if(null!=e.load)this.handler=e;else if(null!=this.loadOptions.requestInit)this.handler=Ci.browserHTTPRequest(e,this.loadOptions);else{const t=Ci.getLoadHandlers(e,this.loadOptions);if(0===t.length)t.push(Ci.browserHTTPRequest(e,this.loadOptions));else if(t.length>1)throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);this.handler=t[0]}}async load(){if(this.findIOHandler(),null==this.handler.load)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");const e=await this.handler.load();return this.loadSync(e)}loadSync(e){this.artifacts=e;const t=this.artifacts.modelTopology;let n;n=null!=this.artifacts.userDefinedMetadata&&null!=this.artifacts.userDefinedMetadata.signature?this.artifacts.userDefinedMetadata.signature:this.artifacts.signature,this.signature=n,this.version=`${t.versions.producer}.${t.versions.minConsumer}`;const r=Ci.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new SA(jI.Instance.transformGraph(t,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(r),this.executor.resourceManager=this.resourceManager,null!=e.modelInitializer&&null!=e.modelInitializer.node){const t=jI.Instance.transformGraph(e.modelInitializer);this.initializer=new SA(t),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializer.executeAsync({},[])}return!0}async save(e,t){if("string"==typeof e){const t=Ci.getSaveHandlers(e);if(0===t.length)throw new Error(`Cannot find any save handlers for URL '${e}'`);if(t.length>1)throw new Error(`Found more than one (${t.length}) save handlers for URL '${e}'`);e=t[0]}if(null==e.save)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return e.save(this.artifacts)}predict(e,t){return this.execute(e,this.outputNodes)}normalizeInputs(e){if(!(e instanceof Ya||Array.isArray(e)))return e;if((e=Array.isArray(e)?e:[e]).length!==this.inputNodes.length)throw new Error(`Input tensor count mismatch,the graph model has ${this.inputNodes.length} placeholders, while there are ${e.length} input tensors.`);return this.inputNodes.reduce((t,n,r)=>(t[n]=e[r],t),{})}normalizeOutputs(e){return e=e||this.outputNodes,Array.isArray(e)?e:[e]}execute(e,t){e=this.normalizeInputs(e),t=this.normalizeOutputs(t);const n=this.executor.execute(e,t);return n.length>1?n:n[0]}async executeAsync(e,t){e=this.normalizeInputs(e),t=this.normalizeOutputs(t);const n=await this.executor.executeAsync(e,t);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(e){return Object.keys(e).reduce((t,n)=>(t[n]=[e[n]],t),{})}dispose(){this.executor.dispose(),this.initializer&&this.initializer.dispose(),this.resourceManager.dispose()}};async function AA(e,t={}){if(null==e)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");null==t&&(t={}),t.fromTFHub&&null==e.load&&(e.endsWith("/")||(e+="/"),e+="model.json?tfjs-format=file");const n=new IA(e,t);return await n.load(),n}var EA="3.13.0",NA={};w(NA,{CSVDataset:()=>yE,Dataset:()=>iE,FileDataSource:()=>RE,TextLineDataset:()=>hE,URLDataSource:()=>_E,array:()=>uE,csv:()=>FE,func:()=>ME,generator:()=>DE,microphone:()=>LE,version_data:()=>PE,webcam:()=>OE,zip:()=>lE});var CA=k(W()),TA=k(W());function RA(e,t,n=new Map,r=new Set){if(null==e)return null;if("function"==typeof Blob&&e instanceof Blob)return e.slice();if(r.has(e))throw new Error("Circular references are not supported.");if(n.has(e))return n.get(e);const a=t(e);if(a.recurse&&null!==a.value)throw new Error("A deep map function may not return both a value and recurse=true.");if(a.recurse){if(DA(e)){const a=Array.isArray(e)?[]:{};r.add(e);for(const s in e){const i=RA(e[s],t,n,r);a[s]=i}return r.delete(e),e.__proto__&&(a.__proto__=e.__proto__),a}throw new Error("Can't recurse into non-iterable type: "+e)}return n.set(e,a.value),a.value}function _A(e,t=FA){return function e(t,n,r=new Set){const a=t[0];if(r.has(a))throw new Error("Circular references are not supported.");const s=n(t);if(s.recurse&&null!==s.value)throw new Error("A deep zip function may not return both a value and recurse=true.");if(s.recurse){if(DA(a)){const s=Array.isArray(a)?[]:{};r.add(a);for(const i in a){const a=t.map(e=>e[i]),o=e(a,n,r);s[i]=o}return r.delete(a),s}throw new Error("Can't recurse into non-iterable type: "+a)}return s.value}(e,t)}function FA(e){return null===e?null:DA(e[0])?{value:null,recurse:!0}:{value:e,recurse:!1}}async function MA(e,t){const n=new Map;RA(e,t,n);for(const e of Array.from(n.keys())){const t=n.get(e);if(va.isPromise(t)){const r=await t;n.set(e,r)}}return RA(e,t,n)}function DA(e){let t=!1;if(Ye().get("IS_BROWSER"))t=e instanceof TextDecoder;else{const{StringDecoder:n}=V();t=e instanceof n}return null!=e&&!ArrayBuffer.isView(e)&&(Array.isArray(e)||"object"==typeof e&&!(e instanceof Ya)&&!(e instanceof Promise)&&!t)}function OA(e){return RA(e,LA)}function LA(e){return e instanceof Ya?{value:e.clone(),recurse:!1}:DA(e)?{value:null,recurse:!0}:{value:e,recurse:!1}}var PA=class{constructor(e){if(this.capacity=e,this.begin=0,this.end=0,null==e)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(e<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(e),this.doubledCapacity=2*e}wrap(e){for(;e<0;)e+=this.doubledCapacity;return e%this.doubledCapacity}get(e){if(e<0)throw new RangeError("Can't get item at a negative index.");return this.data[e%this.capacity]}set(e,t){if(e<0)throw new RangeError("Can't set item at a negative index.");this.data[e%this.capacity]=t}length(){let e=this.end-this.begin;return e<0&&(e=this.doubledCapacity+e),e}isFull(){return this.length()===this.capacity}isEmpty(){return 0===this.length()}push(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,e),this.end=this.wrap(this.end+1)}pushAll(e){for(const t of e)this.push(t)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);const e=this.get(this.end);return this.set(this.end,void 0),e}unshift(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,e)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");const e=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),e}shuffleExcise(e){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");const t=this.wrap(this.begin+e),n=this.get(t);return this.set(t,this.pop()),n}},$A=class extends PA{constructor(){super($A.INITIAL_CAPACITY)}isFull(){return!1}push(e){super.isFull()&&this.expand(),super.push(e)}unshift(e){super.isFull()&&this.expand(),super.unshift(e)}expand(){const e=2*this.capacity,t=new Array(e),n=this.length();for(let e=0;e<n;e++)t[e]=this.get(this.wrap(this.begin+e));this.data=t,this.capacity=e,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};function BA(e){return new jA(e)}function zA(e){return new GA(e)}$A.INITIAL_CAPACITY=32;var UA,WA,VA=class{async toArray(){const e=[];let t=await this.next();for(;!t.done;)e.push(t.value),t=await this.next();return e}async toArrayForTest(){const e=this.prefetch(100),t=[];let n=await e.next();for(;!n.done;)t.push(n.value),n=await e.next();return t}async resolveFully(){let e=await this.next();for(;!e.done;)e=await this.next()}async resolveWhile(e){let t=await this.next(),n=e(t.value);for(;!t.done&&n;)t=await this.next(),n=e(t.value)}handleErrors(e){return new ZA(this,e)}filter(e){return new XA(this,e)}map(e){return new QA(this,e)}mapAsync(e){return new JA(this,e)}serialMapAsync(e){return new JA(this,e).serial()}flatmap(e){return new tE(this,e)}async forEachAsync(e){return this.map(e).resolveFully()}async serialForEach(e){return this.serialMapAsync(e).resolveWhile(e=>!0===e)}rowMajorBatch(e,t=!0){return new YA(this,e,t)}columnMajorBatch(e,t=!0,n=FA){return this.rowMajorBatch(e,t).map(e=>_A(e,n))}concatenate(e,t){return new nE(BA([this,e]),t)}take(e){return e<0||null==e?this:new KA(this,e)}skip(e){return e<0||null==e?this:new qA(this,e)}prefetch(e){return new aE(this,e)}shuffle(e,t){return new sE(this,e,t)}serial(){return new HA(this)}},jA=class extends VA{constructor(e){super(),this.items=e,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};const e=this.items[this.trav];return this.trav++,{value:OA(e),done:!1}}},GA=class extends VA{constructor(e){super(),this.nextFn=e}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(e){throw e.message="Error thrown while iterating through a dataset: "+e.message,e}}},HA=class extends VA{constructor(e){super(),this.upstream=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return this.upstream.summary()+" -> Serial"}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},qA=class extends VA{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return this.upstream.summary()+" -> Skip"}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){const e=await this.upstream.next();if(e.done)return e;au(e.value)}return this.upstream.next()}},KA=class extends VA{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0}summary(){return this.upstream.summary()+" -> Take"}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},YA=class extends VA{constructor(e,t,n=!0){super(),this.upstream=e,this.batchSize=t,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return this.upstream.summary()+" -> RowMajorBatch"}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){const e=[];for(;e.length<this.batchSize;){const t=await this.upstream.next();if(t.done)return this.enableSmallLastBatch&&e.length>0?{value:e,done:!1}:{value:null,done:!0};e.push(t.value)}return{value:e,done:!1}}},XA=class extends VA{constructor(e,t){super(),this.upstream=e,this.predicate=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return this.upstream.summary()+" -> Filter"}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){const e=await this.upstream.next();if(e.done||this.predicate(e.value))return e;au(e.value)}}},QA=class extends VA{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return this.upstream.summary()+" -> Map"}async next(){const e=await this.upstream.next();if(e.done)return{value:null,done:!0};const t=us.getTensorsInContainer(e.value),n=this.transform(e.value),r=us.getTensorsInContainer(n);for(const e of t)us.isTensorInList(e,r)||e.dispose();return{value:n,done:!1}}},ZA=class extends VA{constructor(e,t){super(),this.upstream=e,this.handler=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return this.upstream.summary()+" -> handleErrors"}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(e){if(!this.handler(e))return{value:null,done:!0}}}},JA=class extends VA{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return this.upstream.summary()+" -> AsyncMap"}async next(){const e=await this.upstream.next();if(e.done)return{value:null,done:!0};const t=us.getTensorsInContainer(e.value),n=await this.transform(e.value),r=us.getTensorsInContainer(n);for(const e of t)us.isTensorInList(e,r)||e.dispose();return{value:n,done:!1}}},eE=class extends VA{constructor(){super(),this.outputQueue=new $A,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;0===this.outputQueue.length();)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},tE=class extends eE{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return this.upstream.summary()+" -> Flatmap"}async pump(){const e=await this.upstream.next();if(e.done)return!1;const t=us.getTensorsInContainer(e.value),n=this.transform(e.value),r=us.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(const e of t)us.isTensorInList(e,r)||e.dispose();return!0}},nE=class extends VA{constructor(e,t){super(),this.baseErrorHandler=t,this.lastRead=null,this.iterator=null,this.moreIterators=e}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(e){if(await e,null==this.iterator){const e=await this.moreIterators.next();if(e.done)return{value:null,done:!0};this.iterator=e.value,null!=this.baseErrorHandler&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}const t=await this.iterator.next();return t.done?(this.iterator=null,this.readFromChain(e)):t}};(WA=UA||(UA={}))[WA.FAIL=0]="FAIL",WA[WA.SHORTEST=1]="SHORTEST",WA[WA.LONGEST=2]="LONGEST";var rE=class extends VA{constructor(e,t=UA.FAIL){super(),this.iterators=e,this.mismatchMode=t,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(e){await e;let t=0,n=0;const r=await MA(this.iterators,(function(e){if(e instanceof VA){return{value:e.next().then(e=>(t++,e.done&&n++,e.value)),recurse:!1}}return{value:null,recurse:!0}}));if(t===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case UA.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case UA.SHORTEST:return{value:null,done:!0}}return this.count++,{value:r,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},aE=class extends VA{constructor(e,t){super(),this.upstream=e,this.bufferSize=t,this.buffer=new PA(t)}summary(){return this.upstream.summary()+" -> Prefetch"}refill(){for(;!this.buffer.isFull();){const e=this.upstream.next();this.buffer.push(e)}}next(){return this.refill(),this.buffer.shift()}},sE=class extends aE{constructor(e,t,n){super(e,t),this.upstream=e,this.windowSize=t,this.upstreamExhausted=!1,this.random=TA.alea(n||va.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(e){return Math.floor(this.random()*e)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){const e=this.chooseIndex(),t=await this.buffer.shuffleExcise(e);if(!t.done)return this.refill(),t;this.upstreamExhausted=!0}return{value:null,done:!0}}},iE=class{constructor(){this.size=null}batch(e,t=!0){const n=this;let r;return va.assert(e>0,()=>"batchSize needs to be positive, but it is\n      "+e),r=this.size===1/0||null==this.size?this.size:t?Math.ceil(this.size/e):Math.floor(this.size/e),oE(async()=>(await n.iterator()).columnMajorBatch(e,t,cE),r)}concatenate(e){const t=this;let n;return n=this.size===1/0||e.size===1/0?1/0:null!=this.size&&null!=e.size?this.size+e.size:null,oE(async()=>(await t.iterator()).concatenate(await e.iterator()),n)}filter(e){const t=this;let n;return n=this.size===1/0?1/0:null,oE(async()=>(await t.iterator()).filter(t=>ru(()=>e(t))),n)}async forEachAsync(e){return(await this.iterator()).forEachAsync(e)}map(e){const t=this;return oE(async()=>(await t.iterator()).map(t=>ru(()=>e(t))),this.size)}mapAsync(e){const t=this;return oE(async()=>(await t.iterator()).mapAsync(e),this.size)}prefetch(e){if(null==e)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");const t=this;return oE(async()=>(await t.iterator()).prefetch(e),this.size)}repeat(e){const t=this;let n;return n=null!=this.size&&e>0?this.size*e:0===e?0:null!=this.size&&(void 0===e||e<0)?1/0:null,oE(async()=>{const n=zA(async()=>({value:await t.iterator(),done:!1}));return r=n.take(e),new nE(r,a);var r,a},n)}skip(e){const t=this;let n;return n=null!=this.size&&e>=0&&this.size>=e?this.size-e:null!=this.size&&(this.size<e||void 0===e||e<0)?0:null,oE(async()=>(await t.iterator()).skip(e),n)}shuffle(e,t,n=!0){if(null==e||e<0)throw null==this.size?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);const r=this,a=CA.alea(t||va.now().toString());return oE(async()=>{let t=a.int32();return n&&(t+=a.int32()),(await r.iterator()).shuffle(e,t.toString())},this.size)}take(e){const t=this;let n;return n=null!=this.size&&this.size>e?e:null!=this.size&&this.size<=e?this.size:null,oE(async()=>(await t.iterator()).take(e),n)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};function oE(e,t=null){return new class extends iE{constructor(){super(...arguments),this.size=t}async iterator(){return e()}}}function uE(e){return oE(async()=>BA(e),e.length)}function lE(e){if(!DA(e))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(e))for(let n=0;n<e.length;n++)t=null==t?e[n].size:Math.min(t,e[n].size);else if(e instanceof Object)for(const n in e)t=null==t?e[n].size:Math.min(t,e[n].size);return oE(async()=>function(e,t=UA.FAIL){return new rE(e,t)}(await MA(e,e=>{if(e instanceof iE)return{value:e.iterator(),recurse:!1};if(DA(e))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")}),UA.SHORTEST),t)}function cE(e){if(null===e)return null;const t=e[0];if(null==(n=t)||null===(r=n)||"object"!=typeof r&&"function"!=typeof r||Array.isArray(n)||"object"==typeof n&&n instanceof Ya||va.isTypedArray(n)){return{value:function(e){if(0===e.length)throw new Error("Can't make a batch of zero elements.");return e[0]instanceof Ya?Gh(e):Ls(e)}(e),recurse:!1}}var n,r;return{value:null,recurse:!0}}iE.MAX_BUFFER_SIZE=1e4;var hE=class extends iE{constructor(e){super(),this.input=e}async iterator(){return(await this.input.iterator()).decodeUTF8().split("\n").map(e=>(e.endsWith("\r")&&(e=e.slice(0,-1)),e))}},pE=Symbol("out"),dE=Symbol("field"),fE=Symbol("quote"),mE=Symbol("quoteafterquote"),gE=Symbol("quoteinquote"),yE=class extends iE{constructor(e,t){super(),this.input=e,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new hE(e),t||(t={}),this.hasHeader=!1!==t.hasHeader,this.fullColumnNames=t.columnNames,this.columnConfigs=t.columnConfigs,this.configuredColumnsOnly=t.configuredColumnsOnly,t.delimWhitespace?(va.assert(null==t.delimiter,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=t.delimiter?t.delimiter:","}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){const e=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!e)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&e&&va.assert(e.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+e.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=e);const t=this.fullColumnNames.reduce((e,t)=>(e[t]=e[t]+1||1,e),{}),n=Object.keys(t).filter(e=>t[e]>1);if(va.assert(0===n.length,()=>"Duplicate column names found: "+n.toString()),this.columnConfigs)for(const e of Object.keys(this.columnConfigs)){if(-1===this.fullColumnNames.indexOf(e))throw new Error('The key "'+e+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){const e=await this.base.iterator(),t=await e.next();if(t.done)throw new Error("No data was found for CSV parsing.");const n=t.value;return this.parseRow(n,!1)}return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let e=await this.base.iterator();return this.hasHeader&&(e=e.skip(1)),e.map(e=>this.makeDataElement(e))}makeDataElement(e){const t=this.parseRow(e),n={},r={};for(let a=0;a<this.fullColumnNames.length;a++){const s=this.fullColumnNames[a],i=this.columnConfigs?this.columnConfigs[s]:null;if(!this.configuredColumnsOnly||i){const o=t[a];let u=null;if(""===o)if(i&&void 0!==i.default)u=i.default;else{if(i&&(i.required||i.isLabel))throw new Error(`Required column ${s} is empty in this line: ${e}`);u=void 0}else{const e=Number(o);if(isNaN(e))u=i&&"bool"===i.dtype?this.getBoolean(o):o;else if(i&&i.dtype)switch(i.dtype){case"float32":u=e;break;case"int32":u=Math.floor(e);break;case"bool":u=this.getBoolean(o);break;default:u=e}else u=e}i&&i.isLabel?r[s]=u:n[s]=u}}return 0===Object.keys(r).length?n:{xs:n,ys:r}}getBoolean(e){return"1"===e||"true"===e.toLowerCase()?1:0}parseRow(e,t=!0){const n=[];let r=0;const a=e.length;let s=pE;for(let t=0;t<a;t++)switch(s){case pE:switch(e.charAt(t)){case'"':r=t+1,s=fE;break;case this.delimiter:if(r=t+1," "===this.delimiter&&this.delimWhitespace)break;n.push(""),s=pE;break;default:s=dE,r=t}break;case dE:switch(e.charAt(t)){case this.delimiter:n.push(e.substring(r,t)),s=pE,r=t+1}break;case fE:switch(e.charAt(t)){case'"':s=mE}break;case mE:switch(e.charAt(t)){case this.delimiter:n.push(e.substring(r,t-1)),s=pE,r=t+1;break;case'"':s=fE;break;default:s=gE}break;case gE:switch(e.charAt(t)){case'"':s=fE}}if(s===mE?n.push(e.substring(r,a-1)):n.push(e.substring(r)),t&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}},bE=class extends VA{constructor(e){super(),this.microphoneConfig=e,this.isClosed=!1,this.fftSize=e.fftSize||1024;const t=Math.log2(this.fftSize);if(this.fftSize<0||t<4||t>14||!Number.isInteger(t))throw new Error("Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got "+this.fftSize);if(this.numFrames=e.numFramesPerSpectrogram||43,this.sampleRateHz=e.sampleRateHz,this.columnTruncateLength=e.columnTruncateLength||this.fftSize,this.audioTrackConstraints=e.audioTrackConstraints,this.smoothingTimeConstant=e.smoothingTimeConstant||0,this.includeSpectrogram=!1!==e.includeSpectrogram,this.includeWaveform=!0===e.includeWaveform,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(e={}){if(Ye().get("IS_NODE"))throw new Error("microphone API is only supported in browser environment.");const t=new bE(e);return await t.start(),t}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:null==this.audioTrackConstraints||this.audioTrackConstraints,video:!1})}catch(e){throw new Error("Error thrown while initializing video stream: "+e.message)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");const e=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new e,this.sampleRateHz){if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`)}else this.sampleRateHz=this.audioContext.sampleRate;const t=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=2*this.fftSize,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,t.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let e,t;const n=await this.getAudioData();if(this.includeSpectrogram){const t=this.flattenQueue(n.freqDataQueue);e=this.getTensorFromAudioDataArray(t,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){const e=this.flattenQueue(n.timeDataQueue);t=this.getTensorFromAudioDataArray(e,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:e,waveform:t},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){const e=[],t=[];let n=0;return new Promise(r=>{const a=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&r({freqDataQueue:e,timeDataQueue:t}),e.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),t.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(a),r({freqDataQueue:e,timeDataQueue:t}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),null!=this.stream&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(e){const t=e[0].length,n=new Float32Array(e.length*t);return e.forEach((e,r)=>n.set(e,r*t)),n}getTensorFromAudioDataArray(e,t){const n=new Float32Array(va.sizeFromShape(t));return n.set(e,n.length-e.length),Ls(n,t)}},vE=class extends VA{constructor(e,t){if(super(),this.webcamVideoElement=e,this.webcamConfig=t,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Yh([0],"int32"),this.webcamConfig.centerCrop){const e=1*this.webcamConfig.resizeWidth/this.webcamVideoElement.width,t=1*this.webcamConfig.resizeHeight/this.webcamVideoElement.height,n=(1-e)/2,r=(1-t)/2,a=n+e,s=t+r;this.cropBox=Xh([r,n,s,a],[1,4])}else this.cropBox=Xh([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(e,t={}){if(Ye().get("IS_NODE"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!e){if(e=document.createElement("video"),!t.resizeWidth||!t.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");e.width=t.resizeWidth,e.height=t.resizeHeight}const n=new vE(e,t);return await n.start(),n}async start(){this.webcamConfig.facingMode&&va.assert("user"===this.webcamConfig.facingMode||"environment"===this.webcamConfig.facingMode,()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(e){throw e.message="Error thrown while initializing video stream: "+e.message,e}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(e){console.log(e),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(e=>{this.webcamVideoElement.onloadedmetadata=()=>{e()}})}async next(){if(this.isClosed)return{value:null,done:!0};let e;try{e=eo.fromPixels(this.webcamVideoElement)}catch(e){throw new Error("Error thrown converting video to pixels: "+JSON.stringify(e))}if(!this.resize)return{value:e,done:!1};try{return{value:this.cropAndResizeFrame(e),done:!1}}catch(e){throw new Error("Error thrown cropping the video: "+e.message)}finally{e.dispose()}}needToResize(){return!(!this.webcamConfig.resizeWidth||!this.webcamConfig.resizeHeight||this.webcamVideoElement.width===this.webcamConfig.resizeWidth&&this.webcamVideoElement.height===this.webcamConfig.resizeHeight)}cropAndResizeFrame(e){return ru(()=>{const t=Bl(Ai(e,"float32"),0);let n;n=Nd.cropAndResize(t,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");const r=n.shape;return qu(n,r.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}},xE=class{},wE=class extends VA{split(e){return new kE(this,e)}},kE=class extends wE{constructor(e,t){super(),this.upstream=e,this.impl=new SE(e,t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},SE=class extends eE{constructor(e,t){super(),this.upstream=e,this.separator=t,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){const e=await this.upstream.next();if(e.done)return""!==this.carryover&&(this.outputQueue.push(this.carryover),this.carryover="",!0);const t=e.value.split(this.separator);t[0]=this.carryover+t[0];for(const e of t.slice(0,-1))this.outputQueue.push(e);return this.carryover=t[t.length-1],!0}},IE=class extends VA{decodeUTF8(){return new AE(this)}},AE=class extends wE{constructor(e){super(),this.upstream=e,this.impl=new EE(e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},EE=class extends eE{constructor(e){if(super(),this.upstream=e,Ye().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{const{StringDecoder:e}=V();this.decoder=new e("utf8")}}summary(){return this.upstream.summary()+" -> Utf8"}async pump(){const e=await this.upstream.next();let t,n;return!e.done&&(t=e.value,n=Ye().get("IS_BROWSER")?this.decoder.decode(t,{stream:!0}):this.decoder.write(i.from(t.buffer)),this.outputQueue.push(n),!0)}},NE=class extends IE{constructor(e,t={}){super(),this.file=e,this.options=t,va.assert(e instanceof Uint8Array||!!Ye().get("IS_BROWSER")&&(e instanceof File||e instanceof Blob),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=t.offset||0,this.chunkSize=t.chunkSize||1048576}summary(){return"FileChunks "+this.file}async next(){if(this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size))return{value:null,done:!0};const e=new Promise((e,t)=>{const n=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,n)));else{const r=new FileReader;r.onload=n=>{let a=r.result;if(a instanceof ArrayBuffer&&(a=new Uint8Array(a)),!(a instanceof Uint8Array))return t(new TypeError("FileReader returned unknown type."));e(a)},r.onabort=e=>t(new Error("Aborted")),r.onerror=e=>t(new Error(e.type));const a=this.file.slice(this.offset,n);r.readAsArrayBuffer(a)}this.offset=n});return{value:await e,done:!1}}};var CE=e=>({method:e.method,headers:e.headers,body:e.body,mode:e.mode,credentials:e.credentials,cache:e.cache,redirect:e.redirect,referrer:e.referrer,integrity:e.integrity});function TE(e){return"string"==typeof e&&"file://"===e.substr(0,7)}var RE=class extends xE{constructor(e,t={}){super(),this.input=e,this.options=t}async iterator(){if(TE(this.input)&&Ye().get("IS_NODE")){const e=j();this.input=e.readFileSync(this.input.substr(7))}return new NE(this.input,this.options)}},_E=class extends xE{constructor(e,t={}){super(),this.url=e,this.fileOptions=t}async iterator(){return TE(this.url)?new RE(this.url,this.fileOptions).iterator():async function(e,t={},n){let r,a;"string"==typeof e?r=e:(r=e.url,a=CE(e));const s=await(n||va.fetch)(r,a);if(s.ok){const e=new Uint8Array(await s.arrayBuffer());return new NE(e,t)}throw new Error(s.statusText)}(this.url,this.fileOptions)}};function FE(e,t={}){return new yE(new _E(e),t)}function ME(e){const t=zA(e);return oE(async()=>t)}function DE(e){return oE(async()=>{const t=await e();return zA(()=>t.next())})}async function OE(e,t){return vE.create(e,t)}async function LE(e){return bE.create(e)}var PE="3.13.0";function $E(e,t){Array.isArray(e)||(e=[e]),e.forEach(e=>{null!=e&&va.assert("complex64"!==e.dtype,()=>t+" does not support complex64 tensors in the CPU backend.")})}var BE=jf.whereImpl,zE=class extends Z{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new Q(this,eu())}nextDataId(){return zE.nextDataId++}write(e,t,n){this.firstUse&&(this.firstUse=!1,Ye().get("IS_NODE")&&jd.warn("\n============================\nHi there 👋. Looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, which binds to TensorFlow C++, by running npm i @tensorflow/tfjs-node, or npm i @tensorflow/tfjs-node-gpu if you have CUDA. Then call require('@tensorflow/tfjs-node'); (-gpu suffix for CUDA) at the start of your program. Visit https://github.com/tensorflow/tfjs-node for more details.\n============================"));const r={id:this.nextDataId()};return this.data.set(r,{values:e,dtype:n,refCount:1}),r}makeTensorInfo(e,t,n){let r;if("string"===t&&null!=n&&n.length>0&&va.isString(n[0])){const a=n.map(e=>va.encodeString(e));r=this.write(a,e,t)}else r=this.write(n,e,t);return{dataId:r,shape:e,dtype:t}}refCount(e){if(this.data.has(e)){return this.data.get(e).refCount}return 0}incRef(e){this.data.get(e).refCount++}decRef(e){if(this.data.has(e)){this.data.get(e).refCount--}}move(e,t,n,r,a){this.data.set(e,{values:t,dtype:r,refCount:a})}numDataIds(){return this.data.numDataIds()}async read(e){return this.readSync(e)}readSync(e){const{dtype:t,complexTensorInfos:n}=this.data.get(e);if("complex64"===t){const e=this.readSync(n.real.dataId),t=this.readSync(n.imag.dataId);return jd.mergeRealAndImagArrays(e,t)}return this.data.get(e).values}bufferSync(e){const t=this.readSync(e.dataId);let n=t;if("string"===e.dtype)try{n=t.map(e=>va.decodeString(e))}catch(e){throw new Error("Failed to decode encoded string bytes into utf-8")}return Ii(e.shape,e.dtype,n)}makeOutput(e,t,n){const r=this.write(e,t,n);return eu().makeTensorFromDataId(r,t,n,this)}disposeData(e,t=!1){if(this.data.has(e)){if(this.data.get(e).refCount--,!t&&this.data.get(e).refCount>0)return!1;const{complexTensorInfos:n}=this.data.get(e);null!=n&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(e)}return!0}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}async time(e){const t=va.now();e();return{kernelMs:va.now()-t}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(e){$E([e],"where");const t=this.readSync(e.dataId);return BE(e.shape,t)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};zE.nextDataId=0;var UE={};function WE(e){const t=new Float32Array(e.length);for(let n=0;n<e.length;++n)t[n]=Math.abs(e[n]);return t}w(UE,{addImpl:()=>nN,bincountImpl:()=>iN,bincountReduceImpl:()=>oN,ceilImpl:()=>hN,concatImpl:()=>fN,equalImpl:()=>mN,expImpl:()=>bN,expm1Impl:()=>wN,floorImpl:()=>IN,gatherNdImpl:()=>NN,gatherV2Impl:()=>CN,greaterEqualImpl:()=>FN,greaterImpl:()=>TN,lessEqualImpl:()=>$N,lessImpl:()=>ON,linSpaceImpl:()=>UN,logImpl:()=>WN,maxImpl:()=>GN,maximumImpl:()=>HN,minimumImpl:()=>YN,multiplyImpl:()=>ZN,negImpl:()=>nC,notEqualImpl:()=>aC,prodImpl:()=>cC,rangeImpl:()=>pC,rsqrtImpl:()=>dC,sigmoidImpl:()=>gC,simpleAbsImpl:()=>WE,sliceImpl:()=>vC,sparseFillEmptyRowsImpl:()=>kC,sparseReshapeImpl:()=>SC,sparseSegmentReductionImpl:()=>IC,sqrtImpl:()=>AC,squaredDifferenceImpl:()=>CC,stridedSliceImpl:()=>_C,stringNGramsImpl:()=>FC,stringSplitImpl:()=>DC,stringToHashBucketFastImpl:()=>OC,subImpl:()=>LC,tileImpl:()=>zC,topKImpl:()=>VC,transposeImpl:()=>oC,uniqueImpl:()=>jC});var VE={kernelName:et,backendName:"cpu",kernelFunc:e=>{const{x:t}=e.inputs,n=e.backend;$E(t,"abs");let r=new Float32Array(va.sizeFromShape(t.shape));return r=WE(n.data.get(t.dataId).values),n.makeOutput(r,t.shape,t.dtype)}};function jE(e){return(t,n,r,a,s)=>{const i=jd.assertAndGetBroadcastShape(t,n),o=i.length,u=va.computeStrides(i),l=va.sizeFromShape(i),c=va.getTypedArrayFromDType(s,l),h=t.length,p=n.length,d=va.computeStrides(t),f=va.computeStrides(n),m=jd.getBroadcastDims(t,i),g=jd.getBroadcastDims(n,i);if(m.length+g.length===0)for(let t=0;t<c.length;++t)c[t]=e(r[t%r.length],a[t%a.length]);else for(let t=0;t<c.length;++t){const n=va.indexToLoc(t,o,u),s=n.slice(-h);m.forEach(e=>s[e]=0);const i=va.locToIndex(s,h,d),l=n.slice(-p);g.forEach(e=>l[e]=0);const y=va.locToIndex(l,p,f);c[t]=e(r[i],a[y])}return[c,i]}}function GE(e){const{inputs:t,backend:n}=e,{real:r,imag:a}=t,s=n.data.get(r.dataId).values,i=n.data.get(a.dataId).values,o=n.makeTensorInfo(r.shape,"complex64");return n.data.get(o.dataId).complexTensorInfos={real:n.makeTensorInfo(r.shape,"float32",s),imag:n.makeTensorInfo(a.shape,"float32",i)},o}var HE={kernelName:Et,backendName:"cpu",kernelFunc:GE};function qE(e,t,n="float32"){if("complex64"===n){return GE({inputs:{real:qE(e,t,"float32"),imag:qE(e,t,"float32")},backend:e})}const r=va.makeZerosTypedArray(va.sizeFromShape(t),n);return e.makeTensorInfo(t,n,r)}function KE(e){const{inputs:t,backend:n}=e,{x:r}=t;return n.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var YE={kernelName:fn,backendName:"cpu",kernelFunc:KE};function XE(e){const{inputs:t,backend:n}=e,{input:r}=t,a=n.data.get(r.dataId).complexTensorInfos.real,s=n.data.get(a.dataId).values;return n.makeTensorInfo(a.shape,a.dtype,s)}var QE={kernelName:ir,backendName:"cpu",kernelFunc:XE};function ZE(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dtype:s}=r;if("complex64"===s){if("complex64"===a.dtype)return KE({inputs:{x:a},backend:n});const e=qE(n,a.shape,a.dtype),t=ZE({inputs:{x:a},backend:n,attrs:{dtype:"float32"}}),r=GE({inputs:{real:t,imag:e},backend:n});return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),r}if("complex64"===a.dtype){const e=XE({inputs:{input:a},backend:n}),t=ZE({inputs:{x:e},backend:n,attrs:{dtype:s}});return n.disposeIntermediateTensorInfo(e),t}if(!va.hasEncodingLoss(a.dtype,s)){const e=KE({inputs:{x:a},backend:n});return{dataId:e.dataId,shape:e.shape,dtype:s}}if("int32"===s){const e=n.data.get(a.dataId).values,t=Int32Array.from(e);return n.makeTensorInfo(a.shape,"int32",t)}if("bool"===s){const e=n.data.get(a.dataId).values,t=va.toTypedArray([0],a.dtype),[r,s]=jE((e,t)=>e!==t?1:0)(a.shape,[],e,t,"bool");return n.makeTensorInfo(s,"bool",r)}throw new Error(`Error in Cast: failed to cast ${a.dtype} to ${s}`)}var JE={kernelName:St,backendName:"cpu",kernelFunc:ZE};function eN(e,t,n,r){return null==n?({inputs:n,backend:a})=>{const{a:s,b:i}=n,o=a;$E([s,i],e);const u=o.data.get(s.dataId).values,l=o.data.get(i.dataId).values,c="string"===s.dtype?jd.fromUint8ToStringArray(u):u,h="string"===s.dtype?jd.fromUint8ToStringArray(l):l,p=r||s.dtype,[d,f]=t(s.shape,i.shape,c,h,p);return o.makeTensorInfo(f,p,d)}:({inputs:e,backend:a})=>{const{a:s,b:i}=e,o=a;if("complex64"===s.dtype||"complex64"===i.dtype){const e=ZE({inputs:{x:s},backend:o,attrs:{dtype:"complex64"}}),t=o.data.get(e.dataId),r=t.complexTensorInfos.real,a=t.complexTensorInfos.imag,u=o.data.get(r.dataId).values,l=o.data.get(a.dataId).values,c=ZE({inputs:{x:i},backend:o,attrs:{dtype:"complex64"}}),h=o.data.get(c.dataId),p=h.complexTensorInfos.real,d=h.complexTensorInfos.imag,f=o.data.get(p.dataId).values,m=o.data.get(d.dataId).values,[g,y,b]=n(s.shape,i.shape,u,l,f,m),v=o.makeTensorInfo(b,"float32",g),x=o.makeTensorInfo(b,"float32",y),w=GE({inputs:{real:v,imag:x},backend:o});return o.disposeIntermediateTensorInfo(e),o.disposeIntermediateTensorInfo(c),o.disposeIntermediateTensorInfo(v),o.disposeIntermediateTensorInfo(x),w}{const e=o.data.get(s.dataId).values,n=o.data.get(i.dataId).values,a=r||s.dtype,[u,l]=t(s.shape,i.shape,e,n,a);return o.makeTensorInfo(l,a,u)}}}function tN(e){return(t,n,r,a,s,i)=>{const o=jd.assertAndGetBroadcastShape(t,n),u=va.sizeFromShape(o),l=o.length,c=va.computeStrides(o),h=va.getTypedArrayFromDType("float32",u),p=va.getTypedArrayFromDType("float32",u),d=jd.getBroadcastDims(t,o),f=jd.getBroadcastDims(n,o),m=jd.mergeRealAndImagArrays(r,a),g=jd.mergeRealAndImagArrays(s,i),y=t.length,b=va.computeStrides(t),v=n.length,x=va.computeStrides(n);if(d.length+f.length===0)for(let t=0;t<h.length;t++){const n=t%m.length,r=t%g.length,a=e(m[2*n],m[2*n+1],g[2*r],g[2*r+1]);h[t]=a.real,p[t]=a.imag}else for(let t=0;t<h.length;t++){const n=va.indexToLoc(t,l,c),r=n.slice(-y);d.forEach(e=>r[e]=0);const a=va.locToIndex(r,y,b),s=n.slice(-v);f.forEach(e=>s[e]=0);const i=va.locToIndex(s,v,x),o=e(m[2*a],m[2*a+1],g[2*i],g[2*i+1]);h[t]=o.real,p[t]=o.imag}return[h,p,o]}}var nN=jE((e,t)=>e+t),rN=tN((e,t,n,r)=>({real:e+n,imag:t+r})),aN=eN(rt,nN,rN),sN={kernelName:rt,backendName:"cpu",kernelFunc:aN};function iN(e,t,n,r,a){const s=va.sizeFromShape(r),i=va.makeZerosTypedArray(a,n);for(let n=0;n<e.length;n++){const r=e[n];if(r<0)throw new Error("Input x must be non-negative!");r>=a||(i[r]+=s>0?t[n]:1)}return i}function oN(e,t,n,r=!1){const a=e.shape[0],s=e.shape[1],i=Ii([a,n],t.dtype);for(let o=0;o<a;o++)for(let a=0;a<s;a++){const s=e.get(o,a);if(s<0)throw new Error("Input x must be non-negative!");s>=n||(r?i.set(1,o,s):t.size>0?i.set(i.get(o,s)+t.get(o,a),o,s):i.set(i.get(o,s)+1,o,s))}return i}function uN(e){return(t,n,r)=>{const a=va.getTypedArrayFromDType(n,t.length);for(let n=0;n<t.length;++n)a[n]=e(t[n],r);return a}}function lN(e,t,n){return({inputs:r,attrs:a,backend:s})=>{const{x:i}=r;if($E(i,e),"string"===i.dtype||"string"===n)throw new Error("unaryKernelFunc does not support string input/output");const o=s,u=o.data.get(i.dataId).values,l=va.sizeFromShape(i.shape),c=n||i.dtype,h=va.getArrayFromDType(c,l);for(let e=0;e<l;++e)h[e]=t(u[e],a);return o.makeTensorInfo(i.shape,c,h)}}function cN(e,t,n){return({inputs:r,attrs:a,backend:s})=>{const{x:i}=r;if($E(i,e),"string"===i.dtype||"string"===n)throw new Error("unaryKernelFunc does not support string input/output");const o=s,u=o.data.get(i.dataId).values,l=n||i.dtype,c=t(u,l,a);return o.makeTensorInfo(i.shape,l,c)}}var hN=uN(e=>Math.ceil(e)),pN=cN(It,hN),dN={kernelName:It,backendName:"cpu",kernelFunc:pN};function fN(e,t,n,r){const a=va.getArrayFromDType(n,va.sizeFromShape(t));if(r&&"string"!==n){let t=0;e.forEach(e=>{const n=va.sizeFromShape(e.shape);a.set(e.vals,t),t+=n})}else{let r=0;e.forEach(e=>{const s="string"===n?jd.fromUint8ToStringArray(e.vals):e.vals;let i=0;for(let n=0;n<e.shape[0];++n){const o=n*t[1]+r;for(let t=0;t<e.shape[1];++t)a[o+t]=s[i++]}r+=e.shape[1]})}return a}var mN=jE((e,t)=>e===t?1:0),gN=eN(Jt,mN,null,"bool"),yN={kernelName:Jt,backendName:"cpu",kernelFunc:gN},bN=uN(e=>Math.exp(e)),vN=cN(en,bN,"float32"),xN={kernelName:en,backendName:"cpu",kernelFunc:vN},wN=uN(e=>Math.expm1(e)),kN=cN(nn,wN),SN={kernelName:nn,backendName:"cpu",kernelFunc:kN},IN=uN(e=>Math.floor(e)),AN=cN(on,IN),EN={kernelName:on,backendName:"cpu",kernelFunc:AN};function NN(e,t,n,r,a,s,i,o,u){const l=Ii([r,s],n);for(let n=0;n<r;n++){const r=[];let c=0;for(let t=0;t<a;t++){const s=e[n*a+t];c+=s*i[t],r.push(s)}if(c<0||c>=u/s)throw new Error(`Invalid indices: ${r} does not index into ${o}`);for(let e=0;e<s;e++)l.values[n*s+e]=t.get(...t.indexToLoc(c*s+e))}return l}function CN(e,t,n){const r=Ii(n,e.dtype);for(let n=0;n<r.size;++n){const a=r.indexToLoc(n).slice(),s=a[0],i=a[2],o=t.locToIndex([s,i]);a[2]=t.values[o];const u=e.locToIndex(a);0<=u&&u<e.values.length&&(r.values[n]=e.values[u])}return r}var TN=jE((e,t)=>e>t?1:0),RN=eN(pn,TN,null,"bool"),_N={kernelName:pn,backendName:"cpu",kernelFunc:RN},FN=jE((e,t)=>e>=t?1:0),MN=eN(dn,FN,null,"bool"),DN={kernelName:dn,backendName:"cpu",kernelFunc:MN},ON=jE((e,t)=>e<t?1:0),LN=eN(wn,ON,null,"bool"),PN={kernelName:wn,backendName:"cpu",kernelFunc:LN},$N=jE((e,t)=>e<=t?1:0),BN=eN(kn,$N,null,"bool"),zN={kernelName:kn,backendName:"cpu",kernelFunc:BN};function UN(e,t,n){const r=(t-e)/(n-1),a=va.makeZerosTypedArray(n,"float32");a[0]=e;for(let e=1;e<a.length;e++)a[e]=a[e-1]+r;return a}var WN=uN(e=>Math.log(e)),VN=cN(In,WN),jN={kernelName:In,backendName:"cpu",kernelFunc:VN};function GN(e,t,n,r){const a=va.getTypedArrayFromDType(r,va.sizeFromShape(n));for(let n=0;n<a.length;++n){const r=n*t;let s=e[r];for(let n=0;n<t;++n){const t=e[r+n];(Number.isNaN(t)||t>s)&&(s=t)}a[n]=s}return a}var HN=jE((e,t)=>Math.max(e,t)),qN=eN(Mn,HN),KN={kernelName:Mn,backendName:"cpu",kernelFunc:qN},YN=jE((e,t)=>Math.min(e,t)),XN=eN(Un,YN),QN={kernelName:Un,backendName:"cpu",kernelFunc:XN},ZN=jE((e,t)=>e*t),JN=tN((e,t,n,r)=>({real:e*n-t*r,imag:e*r+t*n})),eC=eN(Gn,ZN,JN),tC={kernelName:Gn,backendName:"cpu",kernelFunc:eC};function nC(e,t,n){const r=va.createScalarValue(-1,n);return ZN([],t,r,e,n)}var rC={kernelName:Hn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r}=t;$E(r,"neg");const a=n.data.get(r.dataId).values,[s,i]=nC(a,r.shape,r.dtype);return n.makeTensorInfo(i,r.dtype,s)}},aC=jE((e,t)=>e!==t?1:0),sC=eN(qn,aC,null,"bool"),iC={kernelName:qn,backendName:"cpu",kernelFunc:sC};function oC(e,t,n,r,a){const s=t.length,i=va.sizeFromShape(t),o=va.computeStrides(t),u=va.computeStrides(a),l=va.getTypedArrayFromDType(n,va.sizeFromShape(a));for(let t=0;t<i;++t){const n=va.indexToLoc(t,s,o),a=new Array(n.length);for(let e=0;e<a.length;e++)a[e]=n[r[e]];l[va.locToIndex(a,s,u)]=e[t]}return l}function uC(e){const{inputs:t,attrs:n,backend:r}=e,{x:a}=t,{perm:s}=n;$E(a,"transpose");const i=a.shape.length,o=new Array(i);for(let e=0;e<o.length;e++)o[e]=a.shape[s[e]];const u=oC(r.data.get(a.dataId).values,a.shape,a.dtype,s,o);return{dataId:r.write(u,o,a.dtype),shape:o,dtype:a.dtype}}var lC={kernelName:Yr,backendName:"cpu",kernelFunc:uC};function cC(e,t,n,r){const[a,s]=jd.computeOutAndReduceShapes(e,r),i=cs(t,"int32"),o=va.makeZerosTypedArray(va.sizeFromShape(a),i),u=va.sizeFromShape(s);for(let e=0;e<o.length;++e){const t=e*u;let r=1;for(let e=0;e<u;++e)r*=n[t+e];o[e]=r}return{outVals:o,outShape:a,outDtype:i}}var hC={kernelName:ar,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;$E(a,"prod");const o=a.shape.length,u=va.parseAxisParam(s,a.shape),l=jd.getAxesPermutation(u,o);let c=u,h=a;const p=[];null!=l&&(h=uC({inputs:{x:a},backend:n,attrs:{perm:l}}),p.push(h),c=jd.getInnerMostAxes(c.length,o));const d=n.data.get(h.dataId).values,{outVals:f,outShape:m,outDtype:g}=cC(h.shape,h.dtype,d,c);let y=m;return i&&(y=jd.expandShapeToKeepDim(m,u)),p.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(y,g,f)}};function pC(e,t,n,r){if(e===t||e<t&&n<0||t<e&&n>1)return va.makeZerosTypedArray(0,r);const a=Math.abs(Math.ceil((t-e)/n)),s=va.makeZerosTypedArray(a,r);t<e&&1===n&&(n=-1),s[0]=e;for(let e=1;e<s.length;e++)s[e]=s[e-1]+n;return s}var dC=uN(e=>1/Math.sqrt(e)),fC=cN(yr,dC),mC={kernelName:yr,backendName:"cpu",kernelFunc:fC},gC=uN(e=>1/(1+Math.exp(-e))),yC=lN(Ar,e=>1/(1+Math.exp(-e))),bC={kernelName:Ar,backendName:"cpu",kernelFunc:yC};function vC(e,t,n,r,a){const s=fo.isSliceContinous(r,t,n),i=va.sizeFromShape(n),o=va.computeStrides(r);if(s){const n=fo.computeFlatOffset(t,o);return"string"===a?e.slice(n,n+i):e.subarray(n,n+i)}const u=Ii(r,a,"string"===a?jd.fromUint8ToStringArray(e):e),l=Ii(n,a);for(let e=0;e<l.size;++e){const n=l.indexToLoc(e),r=n.map((e,n)=>e+t[n]);l.set(u.get(...r),...n)}return"string"===a?jd.fromStringArrayToUint8(l.values):l.values}function xC(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,size:i}=r;$E(a,"slice");const[o,u]=fo.parseSliceParams(a,s,i);fo.assertParamsValid(a,o,u);const l=vC(n.data.get(a.dataId).values,o,u,a.shape,a.dtype);return n.makeTensorInfo(u,a.dtype,l)}var wC={kernelName:wr,backendName:"cpu",kernelFunc:xC};function kC(e,t,n,r,a,s,i){const o=t[0],u=s[0],l=new Array(u),c=new Array(o),h=t[1];if(0===u){if(0!==o)throw new Error(jd.getSparseFillEmptyRowsIndicesDenseShapeMismatch(o));return[va.getArrayFromDType(n,0),[0,h],va.getArrayFromDType(a,0),l,c]}let p=!0,d=0;const f=new Array(u).fill(0);for(let t=0;t<o;++t){const n=e[t*h];if(n<0)throw new Error(jd.getSparseFillEmptyRowsNegativeIndexErrorMessage(t,n));if(n>=u)throw new Error(jd.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(t,n,u));++f[n],p=p&&n>=d,d=n}let m=!0;for(let e=0;e<u;++e){const t=0===f[e];l[e]=t,m=m&&!t,f[e]=Math.max(f[e],1),e>0&&(f[e]+=f[e-1])}if(m&&p){const t=e,n=r;for(let e=0;e<o;++e)c[e]=e;return[t,[o,h],n,l,c]}{const t=f[u-1],s=va.getArrayFromDType(n,t*h),p=va.getArrayFromDType(a,t),d=new Array(u).fill(0);for(let t=0;t<o;++t){const n=e[t*h],a=d[n],i=(0===n?0:f[n-1])+a;d[n]++;for(let n=0;n<h;++n)s[i*h+n]=e[t*h+n];p[i]=r[t],c[t]=i}for(let e=0;e<u;++e){if(0===d[e]){const t=0===e?0:f[e-1];s[t*h+0]=e;for(let e=1;e<h;++e)s[t*h+e]=0;p[t]=i}}return[s,[t,h],p,l,c]}}function SC(e,t,n,r,a){const s=va.sizeFromShape(r),i=t[0],o=a.length,u=[];let l=1,c=-1;for(let e=0;e<o;++e){const t=a[e];if(-1===t){if(-1!==c)throw new Error(jd.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(c,e));c=e,u.push(1)}else{if(t<0)throw new Error(jd.getSparseReshapeNegativeOutputDimErrorMessage(e,t));l*=t,u.push(t)}}if(-1!==c){if(l<=0)throw new Error(jd.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());const e=Math.trunc(s/l);if(l*e!==s)throw new Error(jd.getSparseReshapeInputOutputMultipleErrorMessage(r,u));u[c]=e}if(va.sizeFromShape(u)!==s)throw new Error(jd.getSparseReshapeInputOutputMismatchErrorMessage(r,u));const h=r.length,p=[];if(h>0){p[h-1]=1;for(let e=h-2;e>=0;--e)p[e]=p[e+1]*r[e+1]}const d=[];if(o>0){d[o-1]=1;for(let e=o-2;e>=0;--e)d[e]=d[e+1]*u[e+1]}const f=va.getArrayFromDType(n,i*o);for(let t=0;t<i;++t){let n=0;for(let r=0;r<h;++r)n+=e[t*h+r]*p[r];for(let e=0;e<o;++e)f[t*o+e]=Math.trunc(n/d[e]),n%=d[e]}return[f,[i,o],u]}function IC(e,t,n,r,a,s=!1,i=0){const o=r.length,u=[t[0],e.length/t[0]],l=u[1],c=o>0?a[o-1]+1:0;if(c<0)throw new Error(jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());const h=t.slice();h[0]=c;const p=h.reduce((e,t)=>e*t,1),d=va.getArrayFromDType(n,p);if(0===o)return c>0&&d.fill(i),[d,h];if(c<=0)throw new Error(jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let f=0,m=1,g=0,y=a[f];for(;;){let t=0;if(m<o){if(t=a[m],y===t){++m;continue}if(y>=t)throw new Error(jd.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(y<0||y>=c)throw new Error(jd.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y,c));y>g&&d.fill(i,g*l,y*l);for(let t=f;t<m;++t){const n=r[t];if(n<0||n>=u[0])throw new Error(jd.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(t,r[t],u[0]));for(let t=0;t<l;t++)d[y*l+t]+=e[n*l+t]}if(s)for(let e=0;e<l;e++)d[y*l+e]/=m-f;if(f=m,++m,g=y+1,y=t,m>o)break}return g<c&&d.fill(i,g*l,c*l),[d,h]}var AC=uN(e=>Math.sqrt(e)),EC=lN(Nr,e=>Math.sqrt(e)),NC={kernelName:Nr,backendName:"cpu",kernelFunc:EC},CC=jE((e,t)=>{const n=e-t;return n*n}),TC=eN(Pr,CC),RC={kernelName:Pr,backendName:"cpu",kernelFunc:TC};function _C(e,t,n,r){const a=Ii(e,t.dtype);for(let e=0;e<a.size;e++){const s=a.indexToLoc(e),i=new Array(s.length);for(let e=0;e<i.length;e++)i[e]=s[e]*n[e]+r[e];a.set(t.get(...i),...s)}return a}function FC(e,t,n,r,a,s,i,o){return new class{constructor(e,t,n,r,a,s){this.separator=va.encodeString(e),this.nGramWidths=t,this.leftPad=va.encodeString(n),this.rightPad=va.encodeString(r),this.padWidth=a,this.preserveShort=s}getPadWidth(e){return Math.min(this.padWidth<0?e-1:this.padWidth,e-1)}getNumNGrams(e,t){const n=this.getPadWidth(t);return Math.max(0,e+2*n-t+1)}createNGrams(e,t,n,r,a,s){for(let i=0;i<a;++i){const o=this.getPadWidth(s),u=Math.max(0,o-i),l=Math.max(0,o-(a-(i+1))),c=s-(u+l),h=t+(u>0?0:i-o);let p=0;p+=u*this.leftPad.length;for(let t=0;t<c;++t)p+=e[h+t].length;p+=l*this.rightPad.length;p+=(u+l+c-1)*this.separator.length,n[r+i]=new Uint8Array(p);const d=n[r+i];let f=0;const m=e=>e.forEach(e=>d[f++]=e);for(let e=0;e<u;++e)m(this.leftPad),m(this.separator);for(let t=0;t<c-1;++t)m(e[h+t]),m(this.separator);if(c>0){m(e[h+c-1]);for(let e=0;e<l;++e)m(this.separator),m(this.rightPad)}else{for(let e=0;e<l-1;++e)m(this.rightPad),m(this.separator);m(this.rightPad)}}}compute(e,t){const n=e.length,r=t.length;if(r>0){let e=t[0];if(0!==e)throw new Error("First split value must be 0, got "+e);for(let a=1;a<r;++a){let r=t[a]>=e;if(r=r&&t[a]<=n,!r)throw new Error(`Invalid split value ${t[a]}, must be in [${e}, ${n}]`);e=t[a]}if(e!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${e}`)}const a=r-1,s=va.getArrayFromDType("int32",r);if(0===n||0===r){const e=new Array(n);for(let e=0;e<=a;++e)s[e]=0;return[e,s]}s[0]=0;for(let e=1;e<=a;++e){const n=t[e]-t[e-1];let r=0;this.nGramWidths.forEach(e=>{r+=this.getNumNGrams(n,e)}),this.preserveShort&&n>0&&0===r&&(r=1),s[e]=s[e-1]+r}const i=new Array(s[a]);for(let n=0;n<a;++n){const r=t[n];let a=s[n];if(this.nGramWidths.forEach(s=>{const o=t[n+1]-t[n],u=this.getNumNGrams(o,s);this.createNGrams(e,r,i,a,u,s),a+=u}),this.preserveShort&&a===s[n]){const s=t[n+1]-t[n];if(0===s)continue;const o=s+2*this.padWidth,u=1;this.createNGrams(e,r,i,a,u,o)}}return[i,s]}}(n,r,a,s,i,o).compute(e,t)}function MC(e,t,n,r){if(!e.length)return;if(0===t.length){for(let t=0;t<e.length;++t)r.push(e.subarray(t,t+1));return}if(1===t.length){const a=t[0];let s=e.indexOf(a);for(;-1!==s;){const t=e.subarray(0,s);n&&0===t.length||r.push(t),s=(e=e.subarray(s+1)).indexOf(a)}return void(n&&0===e.length||r.push(e))}let a=0;for(let s=0;s<e.length+1;s++)if(s===e.length||-1!==t.indexOf(e[s])){const t=e.subarray(a,s);n&&0===t.length||r.push(t),a=s+1}}function DC(e,t,n){const r=e.length,a=[];let s=0,i=0;const o=new Array(r);for(let u=0;u<r;++u){const r=a.length;MC(e[u],t,n,a);const l=a.length-r;o[u]=l,s+=l,i=Math.max(i,l)}const u=va.getArrayFromDType("int32",2*s),l=new Array(s),c=[r,i];let h=0;for(let e=0;e<r;++e)for(let t=0;t<o[e];++t)u[2*h]=e,u[2*h+1]=t,l[h]=a[h],++h;return[u,l,c]}function OC(e,t){const n=va.getArrayFromDType("int32",e.length);for(let r=0;r<e.length;++r)n[r]=va.fingerPrint64(e[r]).modulo(t).getLowBitsUnsigned();return n}var LC=jE((e,t)=>e-t),PC=tN((e,t,n,r)=>({real:e-n,imag:t-r})),$C=eN(Vr,LC,PC),BC={kernelName:Vr,backendName:"cpu",kernelFunc:$C};function zC(e,t){const n=new Array(e.rank);for(let r=0;r<n.length;r++)n[r]=e.shape[r]*t[r];const r=Ii(n,e.dtype);for(let t=0;t<r.values.length;++t){const n=r.indexToLoc(t),a=new Array(e.rank);for(let t=0;t<a.length;t++)a[t]=n[t]%e.shape[t];const s=e.locToIndex(a);r.values[t]=e.values[s]}return r}var UC=(e,t)=>{const n=t.value-e.value;return 0===n?e.index-t.index:n};function WC(e,t,n=0,r=e.length-1){for(;r>n;){if(r-n>600){const a=r-n+1,s=t-n+1,i=Math.log(a),o=.5*Math.exp(2*i/3),u=.5*Math.sqrt(i*o*(a-o)/a)*Math.sign(s-a/2);WC(e,t,Math.max(n,Math.floor(t-s*o/a+u)),Math.min(r,Math.floor(t+(a-s)*o/a+u)))}const a=e[t];let s=n,i=r;for(va.swap(e,n,t),UC(e[r],a)>0&&va.swap(e,n,r);s<i;){for(va.swap(e,s,i),s++,i--;UC(e[s],a)<0;)s+=1;for(;UC(e[i],a)>0;)i-=1}0===UC(e[n],a)?va.swap(e,n,i):(i+=1,va.swap(e,i,r)),i<=t&&(n=i+1),t<=i&&(r=i-1)}}function VC(e,t,n,r,a){const s=t[t.length-1],[i,o]=[e.length/s,s],u=va.getTypedArrayFromDType(n,i*r),l=va.getTypedArrayFromDType("int32",i*r);for(let t=0;t<i;t++){const n=t*o,s=e.subarray(n,n+o);let i=new Array(s.length);s.forEach((e,t)=>i[t]={value:e,index:t}),r<i.length&&(WC(i,r),i=i.slice(0,r)),a&&i.sort(UC);const c=t*r,h=u.subarray(c,c+r),p=l.subarray(c,c+r);for(let e=0;e<r;e++)h[e]=i[e].value,p[e]=i[e].index}const c=t.slice();return c[c.length-1]=r,[Ii(c,n,u),Ii(c,"int32",l)]}function jC(e,t,n,r){const a=va.parseAxisParam(t,n)[0],s=[1,n[0],1];for(let e=0;e<a;e++)s[0]*=n[e];s[1]=n[a];for(let e=a+1;e<n.length;e++)s[2]*=n[e];const i={},o=new Int32Array(n[a]),u=new Ha(s,r,e),l=[],c=1===s[0]&&1===s[2];for(let t=0;t<n[a];t++){let n;if(c)n=e[t].toString();else{const e=[];for(let n=0;n<s[0];n++)for(let r=0;r<s[2];r++)e.push(u.get(n,t,r));n=e.join(",")}if(void 0!==i[n])o[t]=i[n];else{const e=Object.keys(i).length;i[n]=e,o[t]=e,l.push(t)}}const h=s.slice();h[1]=Object.keys(i).length;const p=new Ha(h,r);l.forEach((e,t)=>{for(let n=0;n<s[0];n++)for(let r=0;r<s[2];r++)p.set(u.get(n,e,r),n,t,r)});const d=n.slice();return d[a]=h[1],{outputValues:p.values,outputShape:d,indices:o}}du("cpu",()=>new zE,1);var GC=lN(Xt,e=>e>=0?e:Math.exp(e)-1),HC={kernelName:Xt,backendName:"cpu",kernelFunc:GC};function qC(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{alpha:s}=r;$E([a],"leakyRelu");const i=va.sizeFromShape(a.shape),o=n.data.get(a.dataId).values,u=va.getTypedArrayFromDType("float32",i);for(let e=0;e<o.length;e++)u[e]=o[e]<0?s*o[e]:o[e];return n.makeTensorInfo(a.shape,"float32",u)}var KC={kernelName:xn,backendName:"cpu",kernelFunc:qC},YC=jE((e,t)=>e<0?t*e:e);function XC(e){const{inputs:t,backend:n}=e,{x:r,alpha:a}=t;$E([r,a],"prelu");const s=n.data.get(r.dataId).values,i=n.data.get(a.dataId).values,[o,u]=YC(r.shape,a.shape,s,i,"float32");return n.makeTensorInfo(u,"float32",o)}var QC={kernelName:rr,backendName:"cpu",kernelFunc:XC},ZC=lN(ur,e=>Math.max(0,e)),JC={kernelName:ur,backendName:"cpu",kernelFunc:ZC},eT=lN(fr,e=>Math.min(Math.max(0,e),6)),tT={kernelName:fr,backendName:"cpu",kernelFunc:eT};function nT(e,t,n,r,a){if("linear"===n)return KE({inputs:{x:t},backend:e});if("relu"===n)return ZC({inputs:{x:t},backend:e});if("elu"===n)return GC({inputs:{x:t},backend:e});if("relu6"===n)return eT({inputs:{x:t},backend:e});if("prelu"===n)return XC({inputs:{x:t,alpha:r},backend:e});if("leakyrelu"===n)return qC({inputs:{x:t},backend:e,attrs:{alpha:a}});if("sigmoid"===n)return yC({inputs:{x:t},backend:e});throw new Error(`Activation ${n} has not been implemented for the CPU backend.`)}function rT(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{shape:s}=r,i=va.sizeFromShape(a.shape),o=va.inferFromImplicitShape(s,i),u=va.sizeFromShape(o);va.assert(i===u,()=>`The new shape (${o}) has ${u} elements and the old shape (${a.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),n.incRef(a.dataId);const l=n.data.get(a.dataId);if(null!=l.complexTensorInfos){const e=l.complexTensorInfos.real,t=l.complexTensorInfos.imag;e.shape=o,t.shape=o}return{dataId:a.dataId,shape:o,dtype:a.dtype}}var aT={kernelName:lr,backendName:"cpu",kernelFunc:rT};function sT(e){const{inputs:t,backend:n,attrs:r}=e,{a:a,b:s}=t,{transposeA:i,transposeB:o}=r;$E([a,s],"matMul");const u=a.shape.length,l=s.shape.length,c=i?a.shape[u-2]:a.shape[u-1],h=o?s.shape[l-1]:s.shape[l-2],p=i?a.shape[u-1]:a.shape[u-2],d=o?s.shape[l-2]:s.shape[l-1],f=a.shape.slice(0,-2),m=s.shape.slice(0,-2),g=va.sizeFromShape(f),y=va.sizeFromShape(m),b=Yi.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)).concat([p,d]);va.assert(c===h,()=>`Error in matMul: inner shapes (${c}) and (${h}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);const v=o?[y,d,h]:[y,h,d],x=rT({inputs:{x:a},backend:n,attrs:{shape:i?[g,c,p]:[g,p,c]}}),w=rT({inputs:{x:s},backend:n,attrs:{shape:v}}),k=i?x.shape[1]:x.shape[2],S=i?x.shape[2]:x.shape[1],I=o?w.shape[1]:w.shape[2],A=Math.max(g,y),E=n.data.get(x.dataId).values,N=n.data.get(w.dataId).values,C=va.computeStrides(x.shape),T=va.computeStrides(w.shape),[R,_,F]=i?[C[0],1,C[1]]:[C[0],C[1],1],[M,D,O]=o?[1,T[1],T[0]]:[T[1],1,T[0]],L=S*I,P=Ii([A,S,I],x.dtype),$=P.values,B=n.blockSize;for(let e=0;e<A;e++)for(let t=0;t<S;t+=B)for(let n=0;n<I;n+=B)for(let r=0;r<k;r+=B){const a=Math.min(t+B,S),s=Math.min(n+B,I),i=Math.min(r+B,k);for(let o=t;o<a;o++)for(let t=n;t<s;t++){let n=0;for(let a=r;a<i;a++){const r=Math.min(e,g-1)*R,s=Math.min(e,y-1)*O;n+=E[r+o*_+a*F]*N[a*M+t*D+s]}$[e*L+(o*I+t)]+=n}}return n.disposeIntermediateTensorInfo(x),n.disposeIntermediateTensorInfo(w),n.makeTensorInfo(b,P.dtype,P.values)}var iT={kernelName:bt,backendName:"cpu",kernelFunc:sT};var oT={kernelName:ra,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{a:a,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:h}=r;let p,d,f;const m=[];p=sT({inputs:{a:a,b:s},attrs:{transposeA:u,transposeB:l},backend:n}),i&&(d=aN({inputs:{a:p,b:i},backend:n}),m.push(p),p=d),c&&(f=nT(n,p,c,o,h),m.push(p),p=f);for(const e of m)n.disposeIntermediateTensorInfo(e);return p}},uT=lN(tt,e=>Math.acos(e)),lT={kernelName:tt,backendName:"cpu",kernelFunc:uT},cT=lN(nt,e=>Math.acosh(e)),hT={kernelName:nt,backendName:"cpu",kernelFunc:cT};var pT={kernelName:at,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,r=t;$E(t,"addN");const a=r.map(e=>n.data.get(e.dataId).values),s=Ii(r[0].shape,r[0].dtype),i=s.values;for(let e=0;e<r.length;e++){const t=a[e];for(let e=0;e<i.length;e++)i[e]+=t[e]}return n.makeTensorInfo(s.shape,s.dtype,s.values)}};var dT={kernelName:st,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;$E(a,"all");const o=va.parseAxisParam(s,a.shape);let u=o;const l=jd.getAxesPermutation(u,a.shape.length);let c=a;null!=l&&(c=uC({inputs:{x:a},backend:n,attrs:{perm:l}}),u=jd.getInnerMostAxes(u.length,a.shape.length)),jd.assertAxesAreInnerMostDims("all",u,c.shape.length);const[h,p]=jd.computeOutAndReduceShapes(c.shape,u),d=va.sizeFromShape(p),f=va.makeZerosTypedArray(va.sizeFromShape(h),c.dtype),m=n.data.get(c.dataId).values;for(let e=0;e<f.length;++e){const t=e*d;let n=m[t];for(let e=0;e<d;++e){const r=m[t+e];n=n&&r}f[e]=n}null!=l&&n.disposeIntermediateTensorInfo(c);const g=n.makeTensorInfo(h,c.dtype,f);if(i){const e=rT({inputs:{x:g},backend:n,attrs:{shape:jd.expandShapeToKeepDim(h,o)}});return n.disposeIntermediateTensorInfo(g),e}return g}};var fT={kernelName:it,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;$E(a,"any");const o=va.parseAxisParam(s,a.shape);let u=o;const l=jd.getAxesPermutation(u,a.shape.length);let c=a;null!=l&&(c=uC({inputs:{x:a},backend:n,attrs:{perm:l}}),u=jd.getInnerMostAxes(u.length,a.shape.length)),jd.assertAxesAreInnerMostDims("any",u,c.shape.length);const[h,p]=jd.computeOutAndReduceShapes(c.shape,u),d=va.sizeFromShape(p),f=va.makeZerosTypedArray(va.sizeFromShape(h),c.dtype),m=n.data.get(c.dataId).values;for(let e=0;e<f.length;++e){const t=e*d;let n=m[t];for(let e=0;e<d;++e){const r=m[t+e];n=n||r}f[e]=n}null!=l&&n.disposeIntermediateTensorInfo(c);const g=n.makeTensorInfo(h,c.dtype,f);if(i){const e=rT({inputs:{x:g},backend:n,attrs:{shape:jd.expandShapeToKeepDim(h,o)}});return n.disposeIntermediateTensorInfo(g),e}return g}};var mT={kernelName:ot,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r;$E(a,"argMax");let i=va.parseAxisParam(s,a.shape);const o=jd.getAxesPermutation(i,a.shape.length);let u=a;const l=[];null!=o&&(u=uC({inputs:{x:a},backend:n,attrs:{perm:o}}),l.push(u),i=jd.getInnerMostAxes(i.length,u.shape.length)),i=[i[0]],jd.assertAxesAreInnerMostDims("argMax",i,u.shape.length);const[c,h]=jd.computeOutAndReduceShapes(u.shape,i),p=va.sizeFromShape(c),d=va.makeZerosTypedArray(p,"int32"),f=va.sizeFromShape(h),m=n.data.get(u.dataId).values;for(let e=0;e<d.length;++e){const t=e*f;let n=m[t],r=0;for(let e=0;e<f;++e){const a=m[t+e];a>n&&(n=a,r=e)}d[e]=r}return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(c,"int32",d)}};var gT={kernelName:ut,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r;$E(a,"argMin");let i=va.parseAxisParam(s,a.shape);const o=jd.getAxesPermutation(i,a.shape.length);let u=a;const l=[];null!=o&&(u=uC({inputs:{x:a},backend:n,attrs:{perm:o}}),l.push(u),i=jd.getInnerMostAxes(i.length,u.shape.length)),i=[i[0]],jd.assertAxesAreInnerMostDims("argMin",i,u.shape.length);const[c,h]=jd.computeOutAndReduceShapes(u.shape,i),p=va.sizeFromShape(c),d=va.makeZerosTypedArray(p,"int32"),f=va.sizeFromShape(h),m=n.data.get(u.dataId).values;for(let e=0;e<d.length;++e){const t=e*f;let n=m[t],r=0;for(let e=0;e<f;++e){const a=m[t+e];a<n&&(n=a,r=e)}d[e]=r}return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(c,"int32",d)}},yT=lN(lt,e=>Math.asin(e)),bT={kernelName:lt,backendName:"cpu",kernelFunc:yT},vT=lN(ct,e=>Math.asinh(e)),xT={kernelName:ct,backendName:"cpu",kernelFunc:vT},wT=lN(ht,e=>Math.atan(e)),kT={kernelName:ht,backendName:"cpu",kernelFunc:wT},ST=jE((e,t)=>Math.atan2(e,t)),IT=eN(dt,ST),AT={kernelName:dt,backendName:"cpu",kernelFunc:IT},ET=lN(pt,e=>Math.atanh(e)),NT={kernelName:pt,backendName:"cpu",kernelFunc:ET};function CT(e,t,n,r,a,s){const i=a.strideHeight,o=a.strideWidth,u=a.dilationHeight,l=a.dilationWidth,c=a.effectiveFilterHeight,h=a.effectiveFilterWidth,p=a.padInfo.top,d=a.padInfo.left,f="max"===s?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,m=Ii(a.outShape,n),g=m.values,y=a.outShape[1]*a.outShape[2]*a.outShape[3],b=a.outShape[2]*a.outShape[3],v=a.outShape[3];for(let t=0;t<a.batchSize;++t){const n=t*y,m=t*r[0];for(let t=0;t<a.inChannels;++t)for(let y=0;y<a.outHeight;++y){const x=y*i-p,w=Math.max(0,x),k=Math.min(a.inHeight,c+x),S=n+y*b;for(let n=0;n<a.outWidth;++n){const i=n*o-d,c=Math.max(0,i),p=Math.min(a.inWidth,h+i);let y=f,b=0,x=0;for(let n=w;n<k;n+=u){const a=m+n*r[1];for(let n=c;n<p;n+=l){const i=e[a+n*r[2]+t];"max"===s&&i>y?y=i:"avg"===s&&(b+=i,x++)}if(isNaN(y))break}g[S+n*v+t]="avg"===s?b/x:y}}}return m}function TT(e,t,n,r,a=!1,s=!1){const i=Ii(r.outShape,"int32"),o=r.strideHeight,u=r.strideWidth,l=r.dilationHeight,c=r.dilationWidth,h=r.effectiveFilterHeight,p=r.effectiveFilterWidth,d=r.padInfo.top,f=r.padInfo.left,m=Ii(t,n,e);for(let e=0;e<r.batchSize;++e)for(let t=0;t<r.inChannels;++t)for(let n=0;n<r.outHeight;++n){const g=n*o-d;let y=g;for(;y<0;)y+=l;const b=Math.min(r.inHeight,h+g);for(let o=0;o<r.outWidth;++o){const h=o*u-f;let d=h;for(;d<0;)d+=c;const v=Math.min(r.inWidth,p+h);let x=Number.NEGATIVE_INFINITY,w=-1;for(let n=y;n<b;n+=l){const i=n-g;for(let o=d;o<v;o+=c){const u=o-h,l=m.get(e,n,o,t);l>x&&(x=l,w=a?s?((e*r.inHeight+n)*r.inWidth+o)*r.inChannels+t:(n*r.inWidth+o)*r.inChannels+t:i*p+u)}}i.set(w,e,n,o,t)}}return i}function RT(e,t,n,r,a,s){const i=a.strideDepth,o=a.strideHeight,u=a.strideWidth,l=a.dilationDepth,c=a.dilationHeight,h=a.dilationWidth,p=a.effectiveFilterDepth,d=a.effectiveFilterHeight,f=a.effectiveFilterWidth,m=a.padInfo.front,g=a.padInfo.top,y=a.padInfo.left,b="max"===s?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,v=Ii(a.outShape,n),x=v.values,w=a.outShape[1]*a.outShape[2]*a.outShape[3]*a.outShape[4],k=a.outShape[2]*a.outShape[3]*a.outShape[4],S=a.outShape[3]*a.outShape[4],I=a.outShape[4];for(let t=0;t<a.batchSize;++t){const n=t*w,v=t*r[0];for(let t=0;t<a.inChannels;++t)for(let w=0;w<a.outDepth;++w){const A=w*i-m;let E=A;for(;E<0;)E+=l;const N=Math.min(a.inDepth,p+A),C=n+w*k;for(let n=0;n<a.outHeight;++n){const i=n*o-g;let p=i;for(;p<0;)p+=c;const m=Math.min(a.inHeight,d+i),w=C+n*S;for(let n=0;n<a.outWidth;++n){const i=n*u-y;let o=i;for(;o<0;)o+=h;const d=Math.min(a.inWidth,f+i),g=w+n*I;let k=b,S=0,A=0;for(let n=E;n<N;n+=l){const a=v+n*r[1];for(let n=p;n<m;n+=c){const i=a+n*r[2];for(let n=o;n<d;n+=h){const a=e[i+n*r[3]+t];if("max"===s&&a>k?k=a:"avg"===s&&(S+=a,A++),isNaN(k))break}if(isNaN(k))break}if(isNaN(k))break}x[g+t]="avg"===s?S/A:k}}}}return v}var _T={kernelName:ft,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t;$E(a,"avgPool");const{filterSize:s,strides:i,pad:o,dimRoundingMode:u}=r;va.assert(jd.eitherStridesOrDilationsAreOne(i,1),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);const l=jd.computePool2DInfo(a.shape,s,i,1,o,u);let c;if(1===l.filterWidth&&1===l.filterHeight&&va.arraysEqual(l.inShape,l.outShape))c=KE({inputs:{x:a},backend:n});else{const e=n.data.get(a.dataId).values,t=va.computeStrides(a.shape),r=CT(e,a.shape,a.dtype,t,l,"avg");c=n.makeTensorInfo(l.outShape,a.dtype,r.values)}return c}};var FT={kernelName:gt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:u,dataFormat:l}=r;$E(a,"avgPool3d");const c=jd.computePool3DInfo(a.shape,s,i,1,o,u,l),h=RT(n.data.get(a.dataId).values,a.shape,a.dtype,va.computeStrides(a.shape),c,"avg");return n.makeTensorInfo(h.shape,"float32",h.values)}};var MT={kernelName:yt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:u,dimRoundingMode:l}=r;$E([a,s],"avgPool3DGrad");const c=jd.computePool3DInfo(s.shape,i,o,1,u,l),h=c.strideDepth,p=c.strideHeight,d=c.strideWidth,f=c.filterDepth,m=c.filterHeight,g=c.filterWidth,y=c.dilationDepth,b=c.dilationHeight,v=c.dilationWidth,x=c.effectiveFilterDepth,w=c.effectiveFilterHeight,k=c.effectiveFilterWidth,S=x-1-c.padInfo.front,I=k-1-c.padInfo.left,A=w-1-c.padInfo.top,E=Ii(s.shape,"float32"),N=1/(f*m*g),C=n.bufferSync(a);for(let e=0;e<c.batchSize;++e)for(let t=0;t<c.inChannels;++t)for(let n=0;n<c.inDepth;++n)for(let r=0;r<c.inHeight;++r)for(let a=0;a<c.inWidth;++a){const s=n-S,i=r-A,o=a-I;let u=0;for(let n=0;n<x;n+=y){const r=(s+n)/h;if(!(r<0||r>=c.outDepth||Math.floor(r)!==r))for(let n=0;n<w;n+=b){const a=(i+n)/p;if(!(a<0||a>=c.outHeight||Math.floor(a)!==a))for(let n=0;n<k;n+=v){const s=(o+n)/d;if(s<0||s>=c.outWidth||Math.floor(s)!==s)continue;u+=C.get(e,r,a,s,t)}}}E.set(u*N,e,n,r,a,t)}return n.makeTensorInfo(E.shape,E.dtype,E.values)}};var DT={kernelName:mt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s;$E([a,s],"avgPoolGrad");const{filterSize:o,strides:u,pad:l}=r,c=jd.computePool2DInfo(i.shape,o,u,1,l),h=c.strideHeight,p=c.strideWidth,d=c.filterHeight,f=c.filterWidth,m=c.dilationHeight,g=c.dilationWidth,y=c.effectiveFilterHeight,b=c.effectiveFilterWidth,v=b-1-c.padInfo.left,x=y-1-c.padInfo.top,w=Ii(i.shape,"float32"),k=1/(d*f),S=n.data.get(a.dataId).values,I=Ii(a.shape,"float32",S);for(let e=0;e<c.batchSize;++e)for(let t=0;t<c.inChannels;++t)for(let n=0;n<c.inHeight;++n)for(let r=0;r<c.inWidth;++r){const a=n-x,s=r-v;let i=0;for(let n=0;n<y;n+=m){const r=(a+n)/h;if(!(r<0||r>=c.outHeight||Math.floor(r)!==r))for(let n=0;n<b;n+=g){const a=(s+n)/p;if(a<0||a>=c.outWidth||Math.floor(a)!==a)continue;i+=I.get(e,r,a,t)}}w.set(i*k,e,n,r,t)}return n.makeTensorInfo(w.shape,w.dtype,w.values)}};var OT={kernelName:ln,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,scale:s,offset:i,mean:o,variance:u}=t;va.assert(o.shape.length===u.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),va.assert(null==i||o.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),va.assert(null==s||o.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),$E([a,o,u,s,i],"batchNorm");let{varianceEpsilon:l}=r;null==l&&(l=.001);const c=n.data.get(a.dataId).values,h=n.data.get(o.dataId).values,p=n.data.get(u.dataId).values,d=s?n.data.get(s.dataId).values:new Float32Array([1]),f=i?n.data.get(i.dataId).values:new Float32Array([0]),m=new Float32Array(c.length),g=f.length,y=d.length,b=p.length,v=h.length;let x=0,w=0,k=0,S=0;for(let e=0;e<c.length;++e)m[e]=f[x++]+(c[e]-h[w++])*d[k++]/Math.sqrt(p[S++]+l),x>=g&&(x=0),w>=v&&(w=0),k>=y&&(k=0),S>=b&&(S=0);return n.makeTensorInfo(a.shape,a.dtype,m)}};var LT={kernelName:vt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,crops:i}=r;$E([a],"batchToSpaceND");const o=s.reduce((e,t)=>e*t),u=jd.getReshaped(a.shape,s,o),l=jd.getPermuted(u.length,s.length),c=jd.getReshapedPermuted(a.shape,s,o),h=jd.getSliceBeginCoords(i,s.length),p=jd.getSliceSize(c,i,s.length),d=rT({inputs:{x:a},backend:n,attrs:{shape:u}}),f=uC({inputs:{x:d},backend:n,attrs:{perm:l}}),m=rT({inputs:{x:f},backend:n,attrs:{shape:c}}),g=xC({inputs:{x:m},backend:n,attrs:{begin:h,size:p}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),g}};var PT={kernelName:xt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i}=r,o=iN(n.data.get(a.dataId).values,n.data.get(s.dataId).values,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,o)}};var $T={kernelName:kt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{s0:r,s1:a}=t,s=n.data.get(r.dataId).values,i=n.data.get(a.dataId).values,o=jd.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}},BT=lN(At,(e,t)=>{const n=t;return e>n.clipValueMax?n.clipValueMax:e<n.clipValueMin?n.clipValueMin:e}),zT={kernelName:At,backendName:"cpu",kernelFunc:BT},UT={kernelName:Nt,backendName:"cpu",kernelFunc:e=>{const{x:t}=e.inputs,n=e.backend,r=new Float32Array(va.sizeFromShape(t.shape)),a=n.data.get(t.dataId),s=a.complexTensorInfos.real,i=a.complexTensorInfos.imag,o=n.data.get(s.dataId).values,u=n.data.get(i.dataId).values;for(let e=0;e<o.length;e++){const t=o[e],n=u[e];r[e]=Math.hypot(t,n)}return n.makeOutput(r,t.shape,"float32")}};function WT(e){const{inputs:t,backend:n}=e,{input:r}=t,a=n.data.get(r.dataId).complexTensorInfos.imag,s=n.data.get(a.dataId).values;return n.makeTensorInfo(a.shape,a.dtype,s)}var VT={kernelName:gn,backendName:"cpu",kernelFunc:WT};function jT(e){const{inputs:t,backend:n,attrs:r}=e,{axis:a}=r,s=va.parseAxisParam(a,t[0].shape)[0];let i=jd.computeOutShape(t.map(e=>e.shape),s);if(0===va.sizeFromShape(i))return n.makeTensorInfo(i,t[0].dtype,[]);const o=t.filter(e=>va.sizeFromShape(e.shape)>0);if(1===o.length)return KE({inputs:{x:o[0]},backend:n});const u=o.map(e=>e.shape);if(jd.assertParamsConsistent(u,s),"complex64"===o[0].dtype){const e=o.map(e=>XE({inputs:{input:e},backend:n})),t=o.map(e=>WT({inputs:{input:e},backend:n})),r=jT({inputs:e,backend:n,attrs:{axis:s}}),a=jT({inputs:t,backend:n,attrs:{axis:s}}),i=GE({inputs:{real:r,imag:a},backend:n});return e.forEach(e=>n.disposeIntermediateTensorInfo(e)),t.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(a),i}const l=o.map(e=>{const t=va.sizeFromShape(e.shape.slice(s));return rT({inputs:{x:e},backend:n,attrs:{shape:[-1,t]}})}),c=l.map(e=>({vals:n.data.get(e.dataId).values,shape:e.shape}));i=jd.computeOutShape(l.map(e=>e.shape),1);const h=1===l[0].shape[0],p=fN(c,i,t[0].dtype,h),d=jd.computeOutShape(o.map(e=>e.shape),s),f=n.makeTensorInfo(d,t[0].dtype,p);return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),f}var GT={kernelName:Ct,backendName:"cpu",kernelFunc:jT};function HT(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dataFormat:u,dilations:l,dimRoundingMode:c}=r;$E([a,s],"conv2d");const h=jd.convertConv2DDataFormat(u),p=jd.computeConv2DInfo(a.shape,s.shape,i,l,o,c,!1,h),d=p.filterHeight,f=p.filterWidth,m=p.dilationHeight,g=p.dilationWidth,y=p.padInfo.left,b=p.padInfo.top,v="channelsLast"===p.dataFormat,x=new Ha(p.outShape,a.dtype),w=va.computeStrides(a.shape),k=va.computeStrides(s.shape),S=w[0],I=v?w[1]:w[2],A=v?w[2]:1,E=v?1:w[1],N=x.strides[0],C=v?x.strides[1]:x.strides[2],T=v?x.strides[2]:1,R=v?1:x.strides[1],_=n.data.get(a.dataId).values,F=n.data.get(s.dataId).values,M=x.values;for(let e=0;e<p.batchSize;++e){const t=e*S,n=e*N;for(let e=0;e<p.outHeight;++e){const r=n+e*C,a=e*p.strideHeight-b;for(let e=0;e<d;++e){const n=a+e*m;if(n<0||n>=p.inHeight)continue;const s=e*k[0],i=t+n*I;for(let e=0;e<p.outWidth;++e){const t=r+e*T,n=e*p.strideWidth-y;for(let e=0;e<f;++e){const r=n+e*g;if(r<0||r>=p.inWidth)continue;const a=i+r*A;let o=s+e*k[1];for(let e=0;e<p.inChannels;++e){const n=_[a+e*E];for(let e=0;e<p.outChannels;++e)M[t+e*R]+=n*F[o+e];o+=p.outChannels}}}}}}return n.makeTensorInfo(x.shape,x.dtype,M)}var qT={kernelName:Tt,backendName:"cpu",kernelFunc:HT};var KT={kernelName:Rt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,dataFormat:u,dimRoundingMode:l,filterShape:c}=r;$E([a,s],"conv2dBackpropFilter");const h=jd.convertConv2DDataFormat(u),p=jd.computeConv2DInfo(a.shape,c,i,1,o,l,!1,h),{strideHeight:d,strideWidth:f,filterHeight:m,filterWidth:g}=p,y="channelsLast"===p.dataFormat,b=new Ha(p.filterShape,"float32"),v=p.padInfo.left,x=p.padInfo.top,w=n.data.get(a.dataId).values,k=n.data.get(s.dataId).values,S=new Ha(a.shape,a.dtype,w),I=new Ha(s.shape,s.dtype,k);for(let e=0;e<m;++e){const t=Math.max(0,Math.ceil((x-e)/d)),n=Math.min(p.outHeight,(p.inHeight+x-e)/d);for(let r=0;r<g;++r){const a=Math.max(0,Math.ceil((v-r)/f)),s=Math.min(p.outWidth,(p.inWidth+v-r)/f);for(let i=0;i<p.inChannels;++i)for(let o=0;o<p.outChannels;++o){let u=0;for(let l=0;l<p.batchSize;++l)for(let c=t;c<n;++c){const t=e+c*d-x;for(let e=a;e<s;++e){const n=r+e*f-v;u+=y?S.get(l,t,n,i)*I.get(l,c,e,o):S.get(l,i,t,n)*I.get(l,o,c,e)}}b.set(u,e,r,i,o)}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}};var YT={kernelName:_t,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:u,dataFormat:l,dimRoundingMode:c}=r;$E([a,s],"conv2dBackpropInput");const h=va.computeStrides(s.shape),p=va.computeStrides(a.shape);let d=jd.convertConv2DDataFormat(l);const f=jd.computeConv2DInfo(i,s.shape,o,1,u,c,!1,d),m=new Ha(f.inShape,"float32"),g=m.values,y=n.data.get(a.dataId).values,b=n.data.get(s.dataId).values,[v,x,w]=h,{batchSize:k,filterHeight:S,filterWidth:I,inChannels:A,inHeight:E,inWidth:N,outChannels:C,outHeight:T,outWidth:R,strideHeight:_,strideWidth:F}=f;d=f.dataFormat;const M=S-1-f.padInfo.top,D=I-1-f.padInfo.left,O="channelsLast"===d,L=m.strides[0],P=O?m.strides[1]:m.strides[2],$=O?m.strides[2]:1,B=O?1:m.strides[1],z=p[0],U=O?p[1]:p[2],W=O?p[2]:1,V=O?1:p[1];for(let e=0;e<k;++e)for(let t=0;t<A;++t)for(let n=0;n<E;++n){const r=n-M,a=Math.max(0,Math.ceil(r/_)),s=Math.min(T,(S+r)/_);for(let i=0;i<N;++i){const o=i-D,u=Math.max(0,Math.ceil(o/F)),l=Math.min(R,(I+o)/F);let c=0;for(let n=a;n<s;++n){const a=n*_-r;for(let r=u;r<l;++r){const s=z*e+U*n+W*r,i=v*(S-1-a)+x*(I-1-(r*F-o))+w*t;for(let e=0;e<C;++e){c+=y[s+V*e]*b[i+e]}}}g[L*e+P*n+$*i+B*t]=c}}return n.makeTensorInfo(m.shape,m.dtype,m.values)}};var XT={kernelName:Ft,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:u}=r;$E([a,s],"conv3d");const l=jd.computeConv3DInfo(a.shape,s.shape,i,u,o),{filterDepth:c,filterHeight:h,filterWidth:p,dilationDepth:d,dilationHeight:f,dilationWidth:m,padInfo:g}=l,y=g.front,b=g.left,v=g.top,x=new Ha(l.outShape,a.dtype),w=n.data.get(a.dataId).values,k=n.data.get(s.dataId).values,S=x.values,I=va.computeStrides(a.shape),A=va.computeStrides(s.shape);for(let e=0;e<l.batchSize;++e){const t=e*I[0],n=e*x.strides[0];for(let e=0;e<l.outDepth;++e){const r=n+e*x.strides[1],a=e*l.strideDepth-y;for(let e=0;e<c;++e){const n=a+e*d;if(n<0||n>=l.inDepth)continue;const s=e*A[0],i=t+n*I[1];for(let e=0;e<l.outHeight;++e){const t=r+e*x.strides[2],n=e*l.strideHeight-v;for(let e=0;e<h;++e){const r=n+e*f;if(r<0||r>=l.inHeight)continue;const a=s+e*A[1],o=i+r*I[2];for(let e=0;e<l.outWidth;++e){const n=t+e*l.outChannels,r=e*l.strideWidth-b;for(let e=0;e<p;++e){const t=r+e*m;if(t<0||t>=l.inWidth)continue;const s=a+e*A[2],i=o+t*l.inChannels;let u=s;for(let e=0;e<l.inChannels;++e){const t=w[i+e];for(let e=0;e<l.outChannels;++e)S[n+e]+=t*k[u+e];u+=l.outChannels}}}}}}}}return n.makeTensorInfo(x.shape,x.dtype,x.values)}};var QT={kernelName:Mt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:u}=r;$E([a,s],"conv3dBackpropFilterV2");const l=va.computeStrides(a.shape),c=va.computeStrides(s.shape),h=jd.computeConv3DInfo(a.shape,u,i,1,o),p=h.strideDepth,d=h.strideHeight,f=h.strideWidth,m=h.filterDepth,g=h.filterHeight,y=h.filterWidth,b=new Ha(h.filterShape,"float32"),v=b.values,[x,w,k,S]=b.strides,I=n.data.get(s.dataId).values,[A,E,N,C]=c,T=n.data.get(a.dataId).values,[R,_,F,M]=l,D=h.padInfo.front,O=h.padInfo.left,L=h.padInfo.top;for(let e=0;e<m;++e){const t=Math.max(0,Math.ceil((D-e)/p)),n=Math.min(h.outDepth,(h.inDepth+D-e)/p),r=e*x;for(let a=0;a<g;++a){const s=Math.max(0,Math.ceil((L-a)/d)),i=Math.min(h.outHeight,(h.inHeight+L-a)/d),o=a*w+r;for(let r=0;r<y;++r){const u=Math.max(0,Math.ceil((O-r)/f)),l=Math.min(h.outWidth,(h.inWidth+O-r)/f),c=r*k+o;for(let o=0;o<h.inChannels;++o){const m=o*S+c;for(let c=0;c<h.outChannels;++c){let g=0;for(let m=0;m<h.batchSize;++m){const h=m*R,y=m*A;for(let m=t;m<n;++m){const t=(e+m*p-D)*_+h,n=m*E+y;for(let e=s;e<i;++e){const s=(a+e*d-L)*F+t,i=e*N+n;for(let e=u;e<l;++e){const t=e*C+i;g+=T[(r+e*f-O)*M+s+o]*I[t+c]}}}}v[m+c]=g}}}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}};var ZT={kernelName:Dt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:u}=r;$E([a],"conv3dBackpropInputV2");const l=va.computeStrides(a.shape),c=va.computeStrides(s.shape),h=jd.computeConv3DInfo(u,s.shape,o,1,i),p=new Ha(h.inShape,"float32"),d=p.values,[f,m,g,y]=p.strides,b=n.data.get(a.dataId).values,[v,x,w,k]=l,S=n.data.get(s.dataId).values,[I,A,E,N]=c,{batchSize:C,filterDepth:T,filterHeight:R,filterWidth:_,inChannels:F,inDepth:M,inHeight:D,inWidth:O,outChannels:L,outDepth:P,outHeight:$,outWidth:B,strideDepth:z,strideHeight:U,strideWidth:W}=h,V=T-1-h.padInfo.front,j=R-1-h.padInfo.top,G=_-1-h.padInfo.left;for(let e=0;e<C;++e)for(let t=0;t<F;++t)for(let n=0;n<M;++n){const r=n-V,a=Math.max(0,Math.ceil(r/z)),s=Math.min(P,(T+r)/z);for(let i=0;i<D;++i){const o=i-j,u=Math.max(0,Math.ceil(o/U)),l=Math.min($,(R+o)/U);for(let c=0;c<O;++c){const h=c-G,p=Math.max(0,Math.ceil(h/W)),C=Math.min(B,(_+h)/W);let F=0;for(let n=a;n<s;++n){const a=n*z-r;for(let r=u;r<l;++r){const s=r*U-o;for(let i=p;i<C;++i){const o=v*e+x*n+w*r+k*i,u=I*(T-1-a)+A*(R-1-s)+E*(_-1-(i*W-h))+N*t;for(let e=0;e<L;++e){F+=b[o+e]*S[u+e]}}}}d[f*e+m*n+g*i+y*c+t]=F}}}return n.makeTensorInfo(p.shape,p.dtype,p.values)}},JT=lN(Ot,e=>Math.cos(e)),eR={kernelName:Ot,backendName:"cpu",kernelFunc:JT},tR=lN(Lt,e=>Math.cosh(e)),nR={kernelName:Lt,backendName:"cpu",kernelFunc:tR};var rR={kernelName:$t,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:u,extrapolationValue:l}=r,[c,h,p,d]=a.shape,f=s.shape[0],[m,g]=o,y=Ii([f,m,g,d],"float32"),b=n.data.get(s.dataId).values,v=n.data.get(i.dataId).values,x=n.data.get(a.dataId).values,w=va.computeStrides(a.shape),k=va.computeStrides(y.shape);for(let e=0;e<f;e++){const t=4*e,n=b[t],r=b[t+1],a=b[t+2],s=b[t+3],i=v[e];if(i>=c)continue;const o=m>1?(a-n)*(h-1)/(m-1):0,f=g>1?(s-r)*(p-1)/(g-1):0;for(let t=0;t<m;t++){const c=m>1?n*(h-1)+t*o:.5*(n+a)*(h-1);if(c<0||c>h-1)for(let n=0;n<g;n++)for(let r=0;r<d;r++){const a=r+n*k[2]+t*k[1]+e*k[0];y.values[a]=l}else if("bilinear"===u){const n=Math.floor(c),a=Math.ceil(c),o=c-n;for(let u=0;u<g;u++){const c=g>1?r*(p-1)+u*f:.5*(r+s)*(p-1);if(c<0||c>p-1){for(let n=0;n<d;n++){const r=n+u*k[2]+t*k[1]+e*k[0];y.values[r]=l}continue}const h=Math.floor(c),m=Math.ceil(c),b=c-h;for(let r=0;r<d;r++){let s=r+h*w[2]+n*w[1]+i*w[0];const l=x[s];s=r+m*w[2]+n*w[1]+i*w[0];const c=x[s];s=r+h*w[2]+a*w[1]+i*w[0];const p=x[s];s=r+m*w[2]+a*w[1]+i*w[0];const d=l+(c-l)*b,f=p+(x[s]-p)*b;s=r+u*k[2]+t*k[1]+e*k[0],y.values[s]=d+(f-d)*o}}}else for(let n=0;n<g;++n){const a=g>1?r*(p-1)+n*f:.5*(r+s)*(p-1);if(a<0||a>p-1){for(let r=0;r<d;r++){const a=r+n*k[2]+t*k[1]+e*k[0];y.values[a]=l}continue}const o=Math.round(a),u=Math.round(c);for(let r=0;r<d;r++){const a=r+o*w[2]+u*w[1]+i*w[0],s=r+n*k[2]+t*k[1]+e*k[0];y.values[s]=x[a]}}}}return n.makeTensorInfo(y.shape,y.dtype,y.values)}};var aR={kernelName:Pt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r;$E(a,"cumsum");const u=jd.getAxesPermutation([s],a.shape.length);let l=a;null!=u&&(l=uC({inputs:{x:a},backend:n,attrs:{perm:u}}));const c=jd.getInnerMostAxes(1,a.shape.length)[0];if(c!==l.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${l.shape.length-1} but got axis=${c}`);const h=cs(l.dtype,"int32"),p=va.makeZerosTypedArray(va.sizeFromShape(l.shape),h),d=n.data.get(l.dataId).values,f=l.shape[l.shape.length-1],m=o?(e,t)=>e+f-t-1:(e,t)=>e+t;for(let e=0;e<d.length;e+=f)for(let t=0;t<f;t++){const n=m(e,t);if(0===t)p[n]=i?0:d[n];else{const r=m(e,t-1);p[n]=i?d[r]+p[r]:d[n]+p[r]}}const g=n.makeTensorInfo(l.shape,h,p);if(null!=u){const e=uC({inputs:{x:g},backend:n,attrs:{perm:jd.getUndoAxesPermutation(u)}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(l),e}return g}};var sR={kernelName:Bt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i,binaryOutput:o}=r;if(1===a.shape.length){const e=iN(n.data.get(a.dataId).values,n.data.get(s.dataId).values,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,e)}if(2===a.shape.length){const e=oN(n.bufferSync(a),n.bufferSync(s),i,o);return n.makeTensorInfo(e.shape,s.dtype,e.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)}};var iR={kernelName:zt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockSize:s,dataFormat:i}=r;va.assert("NHWC"===i,()=>"Only NHWC dataFormat supported on CPU for depthToSpace. Got "+i);const o=a.shape[0],u=a.shape[1],l=a.shape[2],c=a.shape[3],h=u*s,p=l*s,d=c/(s*s),f=n.data.get(a.dataId).values,m=new Float32Array(o*h*p*d);let g=0;for(let e=0;e<o;++e)for(let t=0;t<h;++t){const n=Math.floor(t/s),r=t%s;for(let t=0;t<p;++t){const a=Math.floor(t/s),i=(r*s+t%s)*d;for(let t=0;t<d;++t){const r=t+i+c*(a+l*(n+u*e));m[g++]=f[r]}}}return n.makeTensorInfo([o,h,p,d],a.dtype,m)}};function oR(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:u,dimRoundingMode:l}=r;$E([a,s],"depthwiseConv2DNative");const c=va.computeStrides(a.shape),h=va.computeStrides(s.shape);let p=u;null==p&&(p=[1,1]),va.assert(jd.eitherStridesOrDilationsAreOne(i,p),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${p}'`);const d=jd.computeConv2DInfo(a.shape,s.shape,i,p,o,l,!0),{filterHeight:f,filterWidth:m,dilationHeight:g,dilationWidth:y,padInfo:b}=d,v=b.left,x=b.top,w=d.outChannels/d.inChannels,k=new Ha(d.outShape,a.dtype),S=n.data.get(a.dataId).values,I=n.data.get(s.dataId).values,A=k.values;for(let e=0;e<d.batchSize;++e){const t=e*c[0],n=e*k.strides[0];for(let e=0;e<d.outHeight;++e){const r=n+e*k.strides[1],a=e*d.strideHeight-x;for(let e=0;e<f;++e){const n=a+e*g;if(n<0||n>=d.inHeight)continue;const s=e*h[0],i=t+n*c[1];for(let e=0;e<d.outWidth;++e){const t=r+e*k.strides[2],n=e*d.strideWidth-v;for(let e=0;e<m;++e){const r=n+e*y;if(r<0||r>=d.inWidth)continue;const a=s+e*h[1],o=i+r*d.inChannels;let u=t,l=a;for(let e=0;e<d.inChannels;++e){const t=S[o+e];for(let e=0;e<w;++e)A[u+e]+=t*I[l+e];u+=w,l+=w}}}}}}return n.makeTensorInfo(k.shape,k.dtype,k.values)}var uR={kernelName:Ut,backendName:"cpu",kernelFunc:oR};var lR={kernelName:Wt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,dilations:o,pad:u,dimRoundingMode:l,filterShape:c}=r;$E([a,s],"depthwiseConv2dNativeBackpropFilter");const h=jd.computeConv2DInfo(a.shape,c,i,o,u,l,!0),{strideHeight:p,strideWidth:d,filterHeight:f,filterWidth:m}=h,g=new Ha(h.filterShape,"float32"),y=h.padInfo.left,b=h.padInfo.top,v=h.outChannels/h.inChannels,x=n.data.get(a.dataId).values,w=new Ha(a.shape,a.dtype,x),k=n.data.get(s.dataId).values,S=new Ha(s.shape,s.dtype,k);for(let e=0;e<f;++e){const t=Math.max(0,Math.ceil((b-e)/p)),n=Math.min(h.outHeight,(h.inHeight+b-e)/p);for(let r=0;r<m;++r){const a=Math.max(0,Math.ceil((y-r)/d)),s=Math.min(h.outWidth,(h.inWidth+y-r)/d);for(let i=0;i<h.outChannels;++i){const o=Math.trunc(i/v),u=i%v;let l=0;for(let u=0;u<h.batchSize;++u)for(let c=t;c<n;++c){const t=e+c*p-b;for(let e=a;e<s;++e){const n=r+e*d-y;l+=w.get(u,t,n,o)*S.get(u,c,e,i)}}g.set(l,e,r,o,u)}}}return n.makeTensorInfo(g.shape,g.dtype,g.values)}};var cR={kernelName:Vt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{strides:i,dilations:o,pad:u,dimRoundingMode:l,inputShape:c}=r;$E([a,s],"depthwiseConv2DNativeBackpropInput");const h=va.computeStrides(a.shape),p=va.computeStrides(s.shape),d=jd.computeConv2DInfo(c,s.shape,i,o,u,l,!0),f=new Ha(d.inShape,"float32"),m=f.values,[g,y,b]=f.strides,v=n.data.get(a.dataId).values,[x,w,k]=h,S=n.data.get(s.dataId).values,[I,A,E]=p,{batchSize:N,filterHeight:C,filterWidth:T,inChannels:R,inHeight:_,inWidth:F,outChannels:M,outHeight:D,outWidth:O,strideHeight:L,strideWidth:P}=d,$=C-1-d.padInfo.top,B=T-1-d.padInfo.left,z=M/R;for(let e=0;e<N;++e)for(let t=0;t<R;++t)for(let n=0;n<_;++n){const r=n-$,a=Math.max(0,Math.ceil(r/L)),s=Math.min(D,(C+r)/L);for(let i=0;i<F;++i){const o=i-B,u=Math.max(0,Math.ceil(o/P)),l=Math.min(O,(T+o)/P);let c=0;for(let n=a;n<s;++n){const a=n*L-r;for(let r=u;r<l;++r){const s=x*e+w*n+k*r,i=I*(C-1-a)+A*(T-1-(r*P-o))+E*t;for(let e=0;e<z;++e){c+=v[s+(t*z+e)]*S[i+e]}}}m[g*e+y*n+b*i+t]=c}}return n.makeTensorInfo(f.shape,f.dtype,f.values)}};var hR={kernelName:jt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r}=t,a=va.sizeFromShape(r.shape),s=n.data.get(r.dataId).values,i=Ii([a,a],r.dtype),o=i.values;for(let e=0;e<s.length;e++)o[e*a+e]=s[e];const u=[...r.shape,...r.shape];return n.makeTensorInfo(u,i.dtype,i.values)}},pR={kernelName:Gt,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{const{x:r,filter:a}=e,{strides:s,pad:i,dilations:o}=n,u=t,l=u.data.get(r.dataId).values,c=r.shape.length,h=u.data.get(a.dataId).values,p=a.shape.length,{batchSize:d,inHeight:f,inWidth:m,inChannels:g,outHeight:y,outWidth:b,padInfo:v,strideHeight:x,strideWidth:w,filterHeight:k,filterWidth:S,dilationHeight:I,dilationWidth:A,outShape:E}=jd.computeDilation2DInfo(r.shape,a.shape,s,i,"NHWC",o),N=va.sizeFromShape(E),C=E.length,T=va.getArrayFromDType(r.dtype,N);for(let e=0;e<d;++e)for(let t=0;t<y;++t){const n=t*x-v.top;for(let s=0;s<b;++s){const i=s*w-v.left;for(let o=0;o<g;++o){let u=Number.MIN_SAFE_INTEGER;for(let t=0;t<k;++t){const s=n+t*I;if(s>=0&&s<f)for(let n=0;n<S;++n){const d=i+n*A;if(d>=0&&d<m){const i=va.locToIndex([e,s,d,o],c,va.computeStrides(r.shape)),f=va.locToIndex([t,n,o],p,va.computeStrides(a.shape)),m=l[i]+h[f];m>u&&(u=m)}}}T[va.locToIndex([e,t,s,o],C,va.computeStrides(E))]=u}}}return{dataId:u.write(va.toTypedArray(T,r.dtype),E,r.dtype),shape:E,dtype:r.dtype}}},dR={kernelName:qt,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{const{x:r,filter:a,dy:s}=e,{strides:i,pad:o,dilations:u}=n,l=t,c=va.toNestedArray(r.shape,l.data.get(r.dataId).values),h=va.toNestedArray(a.shape,l.data.get(a.dataId).values),{batchSize:p,inHeight:d,inWidth:f,inChannels:m,outHeight:g,outWidth:y,padInfo:b,strideHeight:v,strideWidth:x,filterHeight:w,filterWidth:k,dilationHeight:S,dilationWidth:I,outShape:A}=jd.computeDilation2DInfo(r.shape,a.shape,i,o,"NHWC",u);va.assert(s.rank===A.length,()=>`Error in ${qt}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);const E=va.toNestedArray(A,l.data.get(s.dataId).values),N=va.makeZerosNestedTypedArray(a.shape,a.dtype);for(let e=0;e<p;++e)for(let t=0;t<g;++t){const n=t*v-b.top;for(let r=0;r<y;++r){const a=r*x-b.left;for(let s=0;s<m;++s){let i=Number.MIN_SAFE_INTEGER,o=0,u=0;for(let t=0;t<w;++t){const r=n+t*S;if(r>=0&&r<d)for(let n=0;n<k;++n){const l=a+n*I;if(l>=0&&l<f){const a=c[e][r][l][s]+h[t][n][s];a>i&&(i=a,o=t,u=n)}}}N[o][u][s]+=E[e][t][r][s]}}}return{dataId:l.write(va.toTypedArray(N,r.dtype),a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},fR={kernelName:Ht,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{const{x:r,filter:a,dy:s}=e,{strides:i,pad:o,dilations:u}=n,l=t,c=va.toNestedArray(r.shape,l.data.get(r.dataId).values),h=va.toNestedArray(a.shape,l.data.get(a.dataId).values),{batchSize:p,inHeight:d,inWidth:f,inChannels:m,outHeight:g,outWidth:y,padInfo:b,strideHeight:v,strideWidth:x,filterHeight:w,filterWidth:k,dilationHeight:S,dilationWidth:I,outShape:A}=jd.computeDilation2DInfo(r.shape,a.shape,i,o,"NHWC",u);va.assert(s.rank===A.length,()=>`Error in ${Ht}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);const E=va.toNestedArray(A,l.data.get(s.dataId).values),N=va.makeZerosNestedTypedArray(r.shape,r.dtype);for(let e=0;e<p;++e)for(let t=0;t<g;++t){const n=t*v-b.top;for(let r=0;r<y;++r){const a=r*x-b.left;for(let s=0;s<m;++s){let i=Number.MIN_SAFE_INTEGER,o=n<0?0:n,u=a<0?0:a;for(let t=0;t<w;++t){const r=n+t*S;if(r>=0&&r<d)for(let n=0;n<k;++n){const l=a+n*I;if(l>=0&&l<f){const a=c[e][r][l][s]+h[t][n][s];a>i&&(i=a,o=r,u=l)}}}N[e][o][u][s]+=E[e][t][r][s]}}}return{dataId:l.write(va.toTypedArray(N,r.dtype),r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}};function mR(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;let o;$E(a,"sum"),o="bool"===a.dtype?ZE({inputs:{x:a},backend:n,attrs:{dtype:"int32"}}):KE({inputs:{x:a},backend:n});const u=o.shape.length,l=va.parseAxisParam(s,o.shape),c=jd.getAxesPermutation(l,u);let h=l,p=o;null!=c&&(p=uC({inputs:{x:o},backend:n,attrs:{perm:c}}),h=jd.getInnerMostAxes(h.length,u)),jd.assertAxesAreInnerMostDims("sum",h,p.shape.length);const[d,f]=jd.computeOutAndReduceShapes(p.shape,h);let m=qE(n,d,jd.upcastType(p.dtype,"int32"));const g=va.sizeFromShape(f),y=n.data.get(m.dataId).values,b=n.data.get(p.dataId).values;for(let e=0;e<y.length;++e){const t=e*g;let n=0;for(let e=0;e<g;++e)n+=b[t+e];y[e]=n}if(i){const e=m;m=rT({inputs:{x:m},backend:n,attrs:{shape:jd.expandShapeToKeepDim(m.shape,l)}}),n.disposeIntermediateTensorInfo(e)}return n.disposeIntermediateTensorInfo(o),null!=c&&n.disposeIntermediateTensorInfo(p),m}var gR={kernelName:Cr,backendName:"cpu",kernelFunc:mR};var yR={kernelName:Yt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{equation:a}=r,s=t,{allDims:i,summedDims:o,idDims:u}=jd.decodeEinsumEquation(a,s.length);jd.checkEinsumDimSizes(i.length,u,s);const{path:l,steps:c}=jd.getEinsumComputePath(o,u),h=c.length;let p=null,d=i.length;const f=[];for(let e=0;e<h;++e){for(const t of c[e]){const{permutationIndices:e,expandDims:r}=jd.getEinsumPermutation(d,u[t]);let a;jd.isIdentityPermutation(e)?a=s[t]:(a=uC({inputs:{x:s[t]},backend:n,attrs:{perm:e}}),f.push(a));const i=a.shape.slice();for(let e=0;e<r.length;++e)i.splice(r[e],0,1);va.arraysEqual(a.shape,i)||(a=rT({inputs:{x:a},backend:n,attrs:{shape:i}}),f.push(a)),null===p?p=a:(p=eC({inputs:{a:a,b:p},backend:n}),f.push(p))}e<h-1&&(l[e]>=0&&(p=mR({inputs:{x:p},backend:n,attrs:{axis:l[e]-(i.length-d),keepDims:!1}}),f.push(p)),d--)}for(const e of f)e!==p&&n.disposeIntermediateTensorInfo(e);return p}};var bR={kernelName:Qt,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{dy:r,y:a}=t;$E([r,a],"eluGrad");const s=new Float32Array(va.sizeFromShape(a.shape)),i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values;for(let e=0;e<i.length;++e){const t=i[e];s[e]=t>=1?o[e]:o[e]*(t+1)}return n.makeTensorInfo(a.shape,"float32",s)}},vR=jd.ERF_P,xR=jd.ERF_A1,wR=jd.ERF_A2,kR=jd.ERF_A3,SR=jd.ERF_A4,IR=jd.ERF_A5,AR=lN(Zt,e=>{const t=Math.sign(e),n=Math.abs(e),r=1/(1+vR*n);return t*(1-((((IR*r+SR)*r+kR)*r+wR)*r+xR)*r*Math.exp(-n*n))}),ER={kernelName:Zt,backendName:"cpu",kernelFunc:AR};function NR(e){const{inputs:t,backend:n,attrs:r}=e,{input:a}=t,{dim:s}=r,i=a.shape.length,o=a.shape.slice();let u=s;return s<0&&(va.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+s+1),o.splice(u,0,1),rT({inputs:{x:a},backend:n,attrs:{shape:o}})}var CR={kernelName:tn,backendName:"cpu",kernelFunc:NR},TR=jE((e,t)=>e/t),RR=eN(Kt,TR),_R={kernelName:Kt,backendName:"cpu",kernelFunc:RR};function FR(e,t,n){const r=e.shape,a=r[0],s=r[1],i=n.data.get(e.dataId),o=i.complexTensorInfos.real,u=i.complexTensorInfos.imag,l=[a,s],c=va.sizeFromShape(l),h=va.getTypedArrayFromDType("float32",c),p=va.getTypedArrayFromDType("float32",c);for(let e=0;e<a;e++){const r=xC({inputs:{x:o},backend:n,attrs:{begin:[e,0],size:[1,s]}}),a=xC({inputs:{x:u},backend:n,attrs:{begin:[e,0],size:[1,s]}}),i=GE({inputs:{real:r,imag:a},backend:n}),{real:l,imag:c}=MR(i,t,n),d=jd.mergeRealAndImagArrays(l,c);for(let t=0;t<s;t++){const n=jd.getComplexWithIndex(d,t);h[e*s+t]=n.real,p[e*s+t]=n.imag}n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(i)}const d=n.makeTensorInfo(l,"float32",h),f=n.makeTensorInfo(l,"float32",p),m=GE({inputs:{real:d,imag:f},backend:n});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(f),m}function MR(e,t,n){const r=va.sizeFromShape(e.shape),a=n.data.get(e.dataId),s=n.data.get(a.complexTensorInfos.real.dataId).values,i=n.data.get(a.complexTensorInfos.imag.dataId).values;if(0==((o=r)&o-1)){const a=function e(t,n,r,a,s){if(1===r)return{real:t,imag:n};const i=jd.mergeRealAndImagArrays(t,n),o=r/2,u=jd.complexWithEvenIndex(i),l=u.real,c=u.imag,h=[l.length],p=s.makeTensorInfo(h,"float32",l),d=s.makeTensorInfo(h,"float32",c),f=GE({inputs:{real:p,imag:d},backend:s}),m=jd.complexWithOddIndex(i),g=m.real,y=m.imag,b=[g.length],v=s.makeTensorInfo(b,"float32",g),x=s.makeTensorInfo(b,"float32",y),w=GE({inputs:{real:v,imag:x},backend:s}),k=e(l,c,o,a,s),S=k.real,I=k.imag,A=[S.length],E=s.makeTensorInfo(A,"float32",S),N=s.makeTensorInfo(A,"float32",I),C=GE({inputs:{real:E,imag:N},backend:s}),T=e(g,y,o,a,s),R=T.real,_=T.imag,F=[R.length],M=s.makeTensorInfo(F,"float32",R),D=s.makeTensorInfo(F,"float32",_),O=GE({inputs:{real:M,imag:D},backend:s}),L=jd.exponents(r,a),P=[L.real.length],$=s.makeTensorInfo(P,"float32",L.real),B=s.makeTensorInfo(P,"float32",L.imag),z=GE({inputs:{real:$,imag:B},backend:s}),U=eC({inputs:{a:z,b:O},backend:s}),W=aN({inputs:{a:C,b:U},backend:s}),V=$C({inputs:{a:C,b:U},backend:s}),j=XE({inputs:{input:W},backend:s}),G=XE({inputs:{input:V},backend:s}),H=WT({inputs:{input:W},backend:s}),q=WT({inputs:{input:V},backend:s}),K=jT({inputs:[j,G],backend:s,attrs:{axis:0}}),Y=jT({inputs:[H,q],backend:s,attrs:{axis:0}}),X=s.data.get(K.dataId).values,Q=s.data.get(Y.dataId).values;return s.disposeIntermediateTensorInfo(p),s.disposeIntermediateTensorInfo(d),s.disposeIntermediateTensorInfo(f),s.disposeIntermediateTensorInfo(v),s.disposeIntermediateTensorInfo(x),s.disposeIntermediateTensorInfo(w),s.disposeIntermediateTensorInfo(E),s.disposeIntermediateTensorInfo(N),s.disposeIntermediateTensorInfo(C),s.disposeIntermediateTensorInfo(M),s.disposeIntermediateTensorInfo(D),s.disposeIntermediateTensorInfo(O),s.disposeIntermediateTensorInfo($),s.disposeIntermediateTensorInfo(B),s.disposeIntermediateTensorInfo(z),s.disposeIntermediateTensorInfo(U),s.disposeIntermediateTensorInfo(W),s.disposeIntermediateTensorInfo(V),s.disposeIntermediateTensorInfo(j),s.disposeIntermediateTensorInfo(H),s.disposeIntermediateTensorInfo(G),s.disposeIntermediateTensorInfo(q),s.disposeIntermediateTensorInfo(K),s.disposeIntermediateTensorInfo(Y),{real:X,imag:Q}}(s,i,r,t,n),o=[e.shape[0],e.shape[1]];if(t){const e=n.makeTensorInfo(o,"float32",a.real),t=n.makeTensorInfo(o,"float32",a.imag),s=n.makeTensorInfo([],"float32",va.createScalarValue(r,"float32")),i=KE({inputs:{x:s},backend:n}),u=_R.kernelFunc({inputs:{a:e,b:s},backend:n}),l=_R.kernelFunc({inputs:{a:t,b:i},backend:n}),c=n.data.get(u.dataId).values,h=n.data.get(l.dataId).values;return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(l),{real:c,imag:h}}return a}{const e=function(e,t,n){const r=new Float32Array(2*t);for(let a=0;a<t;a++){let s=0,i=0;for(let r=0;r<t;r++){const o=jd.exponent(a*r,t,n),u=jd.getComplexWithIndex(e,r);s+=u.real*o.real-u.imag*o.imag,i+=u.real*o.imag+u.imag*o.real}n&&(s/=t,i/=t),jd.assignToTypedArray(r,s,i,a)}return r}(jd.mergeRealAndImagArrays(s,i),r,t);return jd.splitRealAndImagArrays(e)}var o}var DR={kernelName:rn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{input:r}=t,a=va.sizeFromShape(r.shape),s=r.shape[r.shape.length-1],i=rT({inputs:{x:r},backend:n,attrs:{shape:[a/s,s]}}),o=FR(i,!1,n),u=rT({inputs:{x:o},backend:n,attrs:{shape:r.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),u}};function OR(e){const{backend:t,attrs:n}=e,{shape:r,value:a,dtype:s}=n,i=s||va.inferDtype(a),o=va.getArrayFromDType(i,va.sizeFromShape(r));return function(e,t,n){e.fill(t)}(o,a),t.makeTensorInfo(r,i,o)}var LR={kernelName:an,backendName:"cpu",kernelFunc:OR};var PR={kernelName:sn,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{const{image:r}=e,a=n,s=va.getTypedArrayFromDType(r.dtype,va.sizeFromShape(r.shape)),[i,o,u,l]=r.shape,c=a.data.get(r.dataId).values;for(let e=0;e<i;e++){const t=e*u*o*l;for(let e=0;e<o;e++){const n=e*(u*l);for(let e=0;e<u;e++){const r=e*l;for(let a=0;a<l;a++){const i=Math.round(u-e-1),o=t+n+r+a;let h=c[o];if(i>=0&&i<u){h=c[t+n+i*l+a]}s[o]=h}}}}return{dataId:a.write(s,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},$R=jE((e,t)=>Math.floor(e/t)),BR=eN(un,$R,null,"int32"),zR={kernelName:un,backendName:"cpu",kernelFunc:BR};var UR={kernelName:aa,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:u,pad:l,dataFormat:c,dilations:h,dimRoundingMode:p,activation:d,leakyreluAlpha:f}=r;let m=HT({inputs:{x:a,filter:s},backend:n,attrs:{strides:u,pad:l,dataFormat:c,dilations:h,dimRoundingMode:p}});if(i){const e=m;m=aN({inputs:{a:m,b:i},backend:n}),n.disposeIntermediateTensorInfo(e)}if(d){const e=m;m=nT(n,m,d,o,f),n.disposeIntermediateTensorInfo(e)}return m}};var WR={kernelName:sa,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:u,pad:l,dataFormat:c,dilations:h,dimRoundingMode:p,activation:d,leakyreluAlpha:f}=r;let m=oR({inputs:{x:a,filter:s},backend:n,attrs:{strides:u,pad:l,dataFormat:c,dilations:h,dimRoundingMode:p}});if(i){const e=m;m=aN({inputs:{a:m,b:i},backend:n}),n.disposeIntermediateTensorInfo(e)}if(d){const e=m;m=nT(n,m,d,o,f),n.disposeIntermediateTensorInfo(e)}return m}};var VR={kernelName:hn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{params:r,indices:a}=t,s=va.sizeFromShape(r.shape),i=a.shape,o=i[i.length-1],[u,l,c,h]=jd.prepareAndValidate(r,a);if(0===l)return n.makeTensorInfo(u,r.dtype,[]);const p=NN(n.data.get(a.dataId).values,n.bufferSync(r),r.dtype,l,o,c,h,r.shape,s);return n.makeTensorInfo(u,r.dtype,p.values)}};var jR={kernelName:cn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=r;$E([a,s],"gatherV2");const u=va.parseAxisParam(i,a.shape)[0],l=n.data.get(s.dataId).values,c=a.shape[u];for(let e=0;e<l.length;++e){const t=l[e];va.assert(t<=c-1&&t>=0,()=>`GatherV2: the index value ${t} is not in [0, ${c-1}]`)}let h=o;null==o&&(h=0);const p=va.sizeFromShape(s.shape),d=jd.segment_util.collectGatherOpShapeInfo(a,s,u,h),f=rT({inputs:{x:a},backend:n,attrs:{shape:[d.batchSize,d.outerSize,d.dimSize,d.sliceSize]}}),m=rT({inputs:{x:s},backend:n,attrs:{shape:[d.batchSize,p/d.batchSize]}}),g=[d.batchSize,d.outerSize,p/d.batchSize,d.sliceSize],y=n.bufferSync(m),b=CN(n.bufferSync(f),y,g);return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),n.makeTensorInfo(d.outputShape,b.dtype,b.values)}};var GR={kernelName:mn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{input:r}=t,a=va.sizeFromShape(r.shape),s=r.shape[r.shape.length-1],i=rT({inputs:{x:r},backend:n,attrs:{shape:[a/s,s]}}),o=FR(i,!0,n),u=rT({inputs:{x:o},backend:n,attrs:{shape:r.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),u}},HR=lN(yn,e=>Number.isFinite(e)?1:0,"bool"),qR={kernelName:yn,backendName:"cpu",kernelFunc:HR},KR=lN(bn,e=>Math.abs(e)===1/0?1:0,"bool"),YR={kernelName:bn,backendName:"cpu",kernelFunc:KR},XR=lN(vn,e=>Number.isNaN(e)?1:0,"bool"),QR={kernelName:vn,backendName:"cpu",kernelFunc:XR};var ZR={kernelName:Sn,backendName:"cpu",kernelFunc:function(e){const{backend:t,attrs:n}=e,{start:r,stop:a,num:s}=n,i=UN(r,a,s);return t.makeTensorInfo([i.length],"float32",i)}},JR=lN(An,e=>Math.log1p(e)),e_={kernelName:An,backendName:"cpu",kernelFunc:JR},t_=jE((e,t)=>e&&t),n_=eN(En,t_,null,"bool"),r_={kernelName:En,backendName:"cpu",kernelFunc:n_},a_=lN(Nn,e=>e?0:1,"bool"),s_={kernelName:Nn,backendName:"cpu",kernelFunc:a_},i_=jE((e,t)=>e||t),o_=eN(Cn,i_,null,"bool"),u_={kernelName:Cn,backendName:"cpu",kernelFunc:o_};var l_={kernelName:Rn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:u}=r;$E(a,"LRN");const l=a.shape[3],c=l-1,h=n.data.get(a.dataId).values,p=va.sizeFromShape(a.shape),d=new Float32Array(p);function f(e){const t=e%l;let n=e-t+Math.max(0,t-s);const r=e-t+Math.min(t+s,c);let a=0;for(;n<=r;n++){const e=h[n];a+=e*e}return a}for(let e=0;e<p;e++){const t=f(e),n=h[e]*Math.pow(i+o*t,-u);d[e]=n}return n.makeTensorInfo(a.shape,a.dtype,d)}};var c_={kernelName:_n,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:u,alpha:l,beta:c}=r;$E(i,"LRNGrad");const h=va.sizeFromShape(i.shape),p=i.shape[3],d=n.data.get(i.dataId).values,f=n.data.get(a.dataId).values,m=n.data.get(s.dataId).values,g=new Float32Array(h),y=h;for(let e=0;e<y;e++){const t=e%p,n=e-t+Math.max(0,t-o),r=e-t+Math.min(p,t+o+1);let a=0;for(let e=n;e<r;e++)a+=Math.pow(f[e],2);a=l*a+u;for(let t=n;t<r;t++){let n=-2*l*c*f[t]*m[e]/a;e===t&&(n+=Math.pow(a,-c)),n*=d[e],g[t]+=n}}return n.makeTensorInfo(i.shape,a.dtype,g)}};function h_(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=r,o=n;let u=a.shape;const l=u.length,c=va.parseAxisParam(s,u);let h=c;const p=jd.getAxesPermutation(h,l);let d=o.data.get(a.dataId).values;if(null!=p){const e=new Array(l);for(let t=0;t<e.length;t++)e[t]=u[p[t]];d=oC(d,u,a.dtype,p,e),h=jd.getInnerMostAxes(h.length,l),u=e}$E(a,"max"),jd.assertAxesAreInnerMostDims("max",h,l);const[f,m]=jd.computeOutAndReduceShapes(u,h),g=GN(d,va.sizeFromShape(m),f,a.dtype),y=o.write(g,f,a.dtype);let b=f;if(i){b=jd.expandShapeToKeepDim(f,c)}return{dataId:y,shape:b,dtype:a.dtype}}var p_={kernelName:Fn,backendName:"cpu",kernelFunc:h_};var d_={kernelName:Dn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t;$E(a,"maxPool");const{filterSize:s,strides:i,pad:o,dimRoundingMode:u}=r;va.assert(jd.eitherStridesOrDilationsAreOne(i,1),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);const l=jd.computePool2DInfo(a.shape,s,i,1,o,u);let c;if(1===l.filterWidth&&1===l.filterHeight&&va.arraysEqual(l.inShape,l.outShape))c=KE({inputs:{x:a},backend:n});else{const e=n.data.get(a.dataId).values,t=va.computeStrides(a.shape),r=CT(e,a.shape,a.dtype,t,l,"max");c=n.makeTensorInfo(l.outShape,a.dtype,r.values)}return c}};var f_={kernelName:Ln,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:u,dataFormat:l}=r;$E(a,"maxPool3d");const c=jd.computePool3DInfo(a.shape,s,i,1,o,u,l),h=RT(n.data.get(a.dataId).values,a.shape,a.dtype,va.computeStrides(a.shape),c,"max");return n.makeTensorInfo(h.shape,"float32",h.values)}};var m_={kernelName:Pn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:u,dimRoundingMode:l}=r;$E([a,s],"maxPool3DGrad");const c=jd.computePool3DInfo(s.shape,i,o,1,u,l),h=function(e,t){const n=Ii(t.outShape,"int32"),r=t.strideDepth,a=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,o=t.dilationHeight,u=t.dilationWidth,l=t.effectiveFilterDepth,c=t.effectiveFilterHeight,h=t.effectiveFilterWidth,p=t.padInfo.front,d=t.padInfo.top,f=t.padInfo.left;for(let m=0;m<t.batchSize;++m)for(let g=0;g<t.inChannels;++g)for(let y=0;y<t.outDepth;++y){const b=y*r-p;let v=b;for(;v<0;)v+=i;const x=Math.min(t.inDepth,l+b);for(let r=0;r<t.outHeight;++r){const l=r*a-d;let p=l;for(;p<0;)p+=o;const w=Math.min(t.inHeight,c+l);for(let a=0;a<t.outWidth;++a){const d=a*s-f;let k=d;for(;k<0;)k+=u;const S=Math.min(t.inWidth,h+d);let I=Number.NEGATIVE_INFINITY,A=-1;for(let t=v;t<x;t+=i){const n=t-b;for(let r=p;r<w;r+=o){const a=r-l;for(let s=k;s<S;s+=u){const i=s-d,o=e.get(m,t,r,s,g);o>=I&&(I=o,A=n*c*h+a*c+i)}}}n.set(A,m,y,r,a,g)}}}return n}(n.bufferSync(s),c),p=c.strideDepth,d=c.strideHeight,f=c.strideWidth,m=c.dilationDepth,g=c.dilationHeight,y=c.dilationWidth,b=c.effectiveFilterDepth,v=c.effectiveFilterHeight,x=c.effectiveFilterWidth,w=b-1-c.padInfo.front,k=x-1-c.padInfo.left,S=v-1-c.padInfo.top,I=Ii(s.shape,"float32"),A=n.bufferSync(a);for(let e=0;e<c.batchSize;++e)for(let t=0;t<c.inChannels;++t)for(let n=0;n<c.inDepth;++n)for(let r=0;r<c.inHeight;++r)for(let a=0;a<c.inWidth;++a){const s=n-w,i=r-S,o=a-k;let u=0;for(let n=0;n<b;n+=m){const r=(s+n)/p;if(!(r<0||r>=c.outDepth||Math.floor(r)!==r))for(let a=0;a<v;a+=g){const s=(i+a)/d;if(!(s<0||s>=c.outHeight||Math.floor(s)!==s))for(let i=0;i<x;i+=y){const l=(o+i)/f;if(l<0||l>=c.outWidth||Math.floor(l)!==l)continue;const p=b*v*x-1-h.get(e,r,s,l,t)===n*v*x+a*x+i?1:0;if(0===p)continue;u+=A.get(e,r,s,l,t)*p}}}I.set(u,e,n,r,a,t)}return n.makeTensorInfo(I.shape,I.dtype,I.values)}};var g_={kernelName:On,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s,output:i}=t,o=s;$E([s,i],"maxPoolGrad");const{filterSize:u,strides:l,pad:c,dimRoundingMode:h}=r,p=jd.computePool2DInfo(o.shape,u,l,1,c,h),d=n.data.get(o.dataId).values,f=Ii(p.outShape,o.dtype,TT(d,o.shape,o.dtype,p).values),m=p.strideHeight,g=p.strideWidth,y=p.dilationHeight,b=p.dilationWidth,v=p.effectiveFilterHeight,x=p.effectiveFilterWidth,w=x-1-p.padInfo.left,k=v-1-p.padInfo.top,S=Ii(o.shape,"float32"),I=n.data.get(a.dataId).values,A=Ii(a.shape,"float32",I);for(let e=0;e<p.batchSize;++e)for(let t=0;t<p.inChannels;++t)for(let n=0;n<p.inHeight;++n)for(let r=0;r<p.inWidth;++r){const a=n-k,s=r-w;let i=0;for(let n=0;n<v;n+=y){const r=(a+n)/m;if(!(r<0||r>=p.outHeight||Math.floor(r)!==r))for(let a=0;a<x;a+=b){const o=(s+a)/g;if(o<0||o>=p.outWidth||Math.floor(o)!==o)continue;const u=v*x-1-f.get(e,r,o,t)===n*x+a?1:0;if(0===u)continue;i+=A.get(e,r,o,t)*u}}S.set(i,e,n,r,t)}return n.makeTensorInfo(S.shape,S.dtype,S.values)}};var y_={kernelName:$n,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{const{x:r}=e,{filterSize:a,strides:s,pad:i,includeBatchInIndex:o}=t,u=n;$E(r,"MaxPoolWithArgmax");const l=u.data.get(r.dataId).values,c=jd.computePool2DInfo(r.shape,a,s,[1,1],i),[h,p]=function(e,t,n,r,a){const s=CT(e,0,n,va.computeStrides(t),a,"max"),i=TT(e,t,n,a,!0,r);return[s.values,i.values]}(l,r.shape,r.dtype,o,c),d=u.write(h,c.outShape,r.dtype),f=u.write(p,c.outShape,r.dtype);return[{dataId:d,shape:c.outShape,dtype:r.dtype},{dataId:f,shape:c.outShape,dtype:"int32"}]}};var b_={kernelName:Bn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=va.parseAxisParam(s,a.shape),u=jd.computeOutAndReduceShapes(a.shape,o)[1],l=va.sizeFromShape(u),c=[],h=n.makeTensorInfo([],"float32",new Float32Array([l]));c.push(h);const p=ZE({inputs:{x:a},backend:n,attrs:{dtype:"float32"}});c.push(p);const d=RR({inputs:{a:p,b:h},backend:n});c.push(d);const f=mR({inputs:{x:d},backend:n,attrs:{axis:s,keepDims:i}});return c.forEach(e=>n.disposeIntermediateTensorInfo(e)),f}};var v_={kernelName:zn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;$E(a,"min");const o=va.parseAxisParam(s,a.shape);let u=o;const l=jd.getAxesPermutation(u,a.shape.length);let c=a;null!=l&&(c=uC({inputs:{x:a},backend:n,attrs:{perm:l}}),u=jd.getInnerMostAxes(u.length,a.shape.length)),jd.assertAxesAreInnerMostDims("min",u,c.shape.length);const[h,p]=jd.computeOutAndReduceShapes(c.shape,u),d=va.sizeFromShape(p),f=va.makeZerosTypedArray(va.sizeFromShape(h),c.dtype),m=n.data.get(c.dataId).values;for(let e=0;e<f.length;++e){const t=e*d;let n=m[t];for(let e=0;e<d;++e){const r=m[t+e];(Number.isNaN(r)||r<n)&&(n=r)}f[e]=n}null!=l&&n.disposeIntermediateTensorInfo(c);const g=n.makeTensorInfo(h,c.dtype,f);if(i){const e=rT({inputs:{x:g},backend:n,attrs:{shape:jd.expandShapeToKeepDim(h,o)}});return n.disposeIntermediateTensorInfo(g),e}return g}};var x_={kernelName:Wn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{paddings:s,mode:i}=r;$E(a,"mirrorPad");const o=s.map((e,t)=>e[0]+a.shape[t]+e[1]),u=s.map(e=>e[0]),l=s.map((e,t)=>e[0]+a.shape[t]),c="reflect"===i?0:1,h=n.data.get(a.dataId).values,p=a.shape.length,d=va.computeStrides(a.shape),f=va.sizeFromShape(o),m=o.length,g=va.computeStrides(o),y=va.getTypedArrayFromDType(a.dtype,f);for(let e=0;e<f;e++){let t=va.indexToLoc(e,m,g);for(let e=0;e<m;e++)t[e]<u[e]?t[e]=2*u[e]-t[e]-c:t[e]>=l[e]&&(t[e]=2*(l[e]-1)-t[e]+c);t=t.map((e,t)=>e-u[t]);const n=va.locToIndex(t,p,d);y[e]=h[n]}return{dataId:n.write(y,o,a.dtype),shape:o,dtype:a.dtype}}},w_=jE((e,t)=>{const n=e%t;return e<0&&t<0||e>=0&&t>=0?n:(n+t)%t}),k_=eN(Vn,w_),S_={kernelName:Vn,backendName:"cpu",kernelFunc:k_},I_=k(D());function A_(e){const{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{dim:s}=r,i=a.shape.length;let o=s;if(-1===o&&(o=i-1),o!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${o}`);const u=va.parseAxisParam([o],a.shape),l=h_({inputs:{x:a},backend:n,attrs:{reductionIndices:u,keepDims:!1}}),c=jd.expandShapeToKeepDim(l.shape,u),h=rT({inputs:{x:l},backend:n,attrs:{shape:c}}),p=$C({inputs:{a:a,b:h},backend:n}),d=vN({inputs:{x:p},backend:n}),f=mR({inputs:{x:d},backend:n,attrs:{axis:u,keepDims:!1}}),m=rT({inputs:{x:f},backend:n,attrs:{shape:c}}),g=RR({inputs:{a:d,b:m},backend:n});return n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),g}var E_={kernelName:_r,backendName:"cpu",kernelFunc:A_};var N_={kernelName:jn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=r;$E(a,"multinomial");const u=o?a:A_({inputs:{logits:a},backend:n,attrs:{dim:-1}}),l=u.shape[0],c=u.shape[1],h=n.data.get(u.dataId).values,p=[l,s],d=va.makeZerosTypedArray(va.sizeFromShape(p),"int32");for(let e=0;e<l;++e){const t=e*c,n=new Float32Array(c-1);n[0]=h[t];for(let e=1;e<n.length;++e)n[e]=n[e-1]+h[t+e];const r=I_.alea(i.toString()),a=e*s;for(let e=0;e<s;++e){const t=r();d[a+e]=n.length;for(let r=0;r<n.length;r++)if(t<n[r]){d[a+e]=r;break}}}return o||n.disposeIntermediateTensorInfo(u),n.makeTensorInfo(p,"int32",d)}},C_=jf.nonMaxSuppressionV3Impl;var T_={kernelName:Kn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:u}=r;$E(a,"NonMaxSuppression");const l=n.data.get(a.dataId).values,c=n.data.get(s.dataId).values,{selectedIndices:h}=C_(l,c,i,o,u);return n.makeTensorInfo([h.length],"int32",new Int32Array(h))}},R_=jf.nonMaxSuppressionV4Impl;var __={kernelName:Yn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:u,padToMaxOutputSize:l}=r;$E(a,"NonMaxSuppressionPadded");const c=n.data.get(a.dataId).values,h=n.data.get(s.dataId).values,{selectedIndices:p,validOutputs:d}=R_(c,h,i,o,u,l);return[n.makeTensorInfo([p.length],"int32",new Int32Array(p)),n.makeTensorInfo([],"int32",new Int32Array([d]))]}},F_=jf.nonMaxSuppressionV5Impl;var M_={kernelName:Xn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:u,softNmsSigma:l}=r;$E(a,"NonMaxSuppressionWithScore");const c=n.data.get(a.dataId).values,h=n.data.get(s.dataId).values,p=i,d=o,f=u,m=l,{selectedIndices:g,selectedScores:y}=F_(c,h,p,d,f,m);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}};var D_={kernelName:Zn,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{indices:a}=t,{depth:s,onValue:i,offValue:o}=r;$E(a,"oneHot");const u=va.sizeFromShape(a.shape),l=new Float32Array(u*s);l.fill(o);const c=n.data.get(a.dataId).values;for(let e=0;e<u;++e)c[e]>=0&&c[e]<s&&(l[e*s+c[e]]=i);return n.makeTensorInfo([...a.shape,s],"int32",l)}};function O_(e){const{inputs:t,backend:n}=e,{x:r}=t;if("string"===r.dtype)throw new Error("zerosLike is not supported for string tensors");if("complex64"===r.dtype){const e=XE({inputs:{input:r},backend:n}),t=O_({inputs:{x:e},backend:n}),a=WT({inputs:{input:r},backend:n}),s=O_({inputs:{x:a},backend:n}),i=GE({inputs:{real:t,imag:s},backend:n});return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(s),i}return OR({backend:n,attrs:{shape:r.shape,value:0,dtype:r.dtype}})}var L_={kernelName:Jr,backendName:"cpu",kernelFunc:O_};var P_={kernelName:Qn,backendName:"cpu",kernelFunc:function e(t){const{inputs:n,backend:r}=t,{x:a}=n;if("string"===a.dtype)throw new Error("onesLike is not supported for string tensors");if("complex64"===a.dtype){const t=XE({inputs:{input:a},backend:r}),n=e({inputs:{x:t},backend:r}),s=WT({inputs:{input:a},backend:r}),i=O_({inputs:{x:s},backend:r}),o=GE({inputs:{real:n,imag:i},backend:r});return r.disposeIntermediateTensorInfo(t),r.disposeIntermediateTensorInfo(n),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}return OR({backend:r,attrs:{shape:a.shape,value:1,dtype:a.dtype}})}};function $_(e){const{inputs:t,backend:n,attrs:r}=e,{axis:a}=r;if(1===t.length)return NR({inputs:{input:t[0]},backend:n,attrs:{dim:a}});const s=t[0].shape,i=t[0].dtype;t.forEach(e=>{va.assertShapesMatch(s,e.shape,"All tensors passed to stack must have matching shapes"),va.assert(i===e.dtype,()=>"All tensors passed to stack must have matching dtypes")});const o=[],u=jT({inputs:t.map(e=>{const t=NR({inputs:{input:e},backend:n,attrs:{dim:a}});return o.push(t),t}),backend:n,attrs:{axis:a}});return o.forEach(e=>n.disposeIntermediateTensorInfo(e)),u}var B_={kernelName:Jn,backendName:"cpu",kernelFunc:$_};var z_={kernelName:er,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{paddings:s,constantValue:i}=r;$E(a,"pad");const o=s.map((e,t)=>e[0]+a.shape[t]+e[1]),u=s.map(e=>e[0]),l=n.data.get(a.dataId).values,c=va.sizeFromShape(a.shape),h=a.shape.length,p=va.computeStrides(a.shape),d=va.sizeFromShape(o),f=o.length,m=va.computeStrides(o),g=va.getTypedArrayFromDType(a.dtype,d);0!==i&&g.fill(i);for(let e=0;e<c;e++){const t=va.indexToLoc(e,h,p).map((e,t)=>e+u[t]);g[va.locToIndex(t,f,m)]=l[e]}return{dataId:n.write(g,o,a.dtype),shape:o,dtype:a.dtype}}},U_=jE((e,t)=>Math.pow(e,t)),W_=eN(nr,U_),V_={kernelName:nr,backendName:"cpu",kernelFunc:W_};var j_={kernelName:sr,backendName:"cpu",kernelFunc:function(e){const{backend:t,attrs:n}=e,{start:r,stop:a,dtype:s,step:i}=n,o=pC(r,a,i,s);return t.makeTensorInfo([o.length],s,o)}},G_=lN(or,e=>1/e),H_={kernelName:or,backendName:"cpu",kernelFunc:G_};var q_={kernelName:pr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r;$E(a,"resizeBilinear");const u=va.computeStrides(a.shape),[l,c]=o,[h,p,d,f]=a.shape,m=n.data.get(a.dataId).values,g=new Float32Array(va.sizeFromShape([h,l,c,f])),y=[s&&l>1?p-1:p,s&&c>1?d-1:d],b=[s&&l>1?l-1:l,s&&c>1?c-1:c];let v=0;const x=y[0]/b[0],w=y[1]/b[1];for(let e=0;e<h;e++)for(let t=0;t<l;t++){let n;n=i?x*(t+.5)-.5:x*t;const r=Math.max(0,Math.floor(n)),a=n-r,s=Math.min(p-1,Math.ceil(n)),o=e*u[0]+r*u[1],l=e*u[0]+s*u[1];for(let e=0;e<c;e++){let t;t=i?w*(e+.5)-.5:w*e;const n=Math.max(0,Math.floor(t)),r=t-n,s=Math.min(d-1,Math.ceil(t)),c=o+n*u[2],h=l+n*u[2],p=o+s*u[2],y=l+s*u[2];for(let e=0;e<f;e++){const t=m[c+e],n=m[h+e],s=t+(m[p+e]-t)*r,i=s+(n+(m[y+e]-n)*r-s)*a;g[v++]=i}}}return n.makeTensorInfo([h,l,c,f],"float32",g)}};var K_={kernelName:dr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r;$E([s,a],"resizeBilinearGrad");const o=va.computeStrides(a.shape),[u,l,c,h]=a.shape,[,p,d]=s.shape,f=new Float32Array(u*l*c*h),m=[i&&p>1?l-1:l,i&&d>1?c-1:c],g=[i&&p>1?p-1:p,i&&d>1?d-1:d],y=m[0]/g[0],b=m[1]/g[1],v=n.data.get(s.dataId).values;let x=0;for(let e=0;e<u;e++){const t=e*o[0];for(let e=0;e<p;e++){const n=e*y,r=Math.floor(n),a=Math.min(Math.ceil(n),l-1),s=t+r*o[1],i=t+a*o[1],u=n-r,p=1-u;for(let e=0;e<d;e++){const t=e*b,n=Math.floor(t),r=Math.min(Math.ceil(t),c-1),a=t-n,l=1-a,d=s+n*o[2],m=s+r*o[2],g=i+n*o[2],y=i+r*o[2],w=p*l,k=p*a,S=u*l,I=u*a;for(let e=0;e<h;e++){const t=v[x++];f[d+e]+=t*w,f[m+e]+=t*k,f[g+e]+=t*S,f[y+e]+=t*I}}}}return n.makeTensorInfo([u,c,l,h],"float32",f)}};var Y_={kernelName:cr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r;$E(a,"resizeNearestNeighbor");const u=va.computeStrides(a.shape),[l,c]=o,[h,p,d,f]=a.shape,m=n.data.get(a.dataId).values,g=new Float32Array(h*l*c*f),y=[s&&l>1?p-1:p,s&&c>1?d-1:d],b=[s&&l>1?l-1:l,s&&c>1?c-1:c],v=y[0]/b[0],x=y[1]/b[1];let w=0;for(let e=0;e<h;e++){const t=e*u[0];for(let e=0;e<l;e++){const n=i?v*(e+.5):v*e;let r=Math.min(p-1,s?Math.round(n):Math.floor(n));i&&(r=Math.max(0,r));const a=t+r*u[1];for(let e=0;e<c;e++){const t=i?x*(e+.5):x*e;let n=Math.min(d-1,s?Math.round(t):Math.floor(t));i&&(n=Math.max(0,n));const r=a+n*u[2];for(let e=0;e<f;e++){const t=m[r+e];g[w++]=t}}}}return n.makeTensorInfo([h,l,c,f],a.dtype,g)}};var X_={kernelName:hr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r;$E([s,a],"resizeNearestNeighborGrad");const o=va.computeStrides(a.shape),u=va.computeStrides(s.shape),[l,c,h,p]=a.shape,[,d,f]=s.shape,m=new Float32Array(l*c*h*p),g=n.data.get(s.dataId).values,y=[i&&d>1?c-1:c,i&&f>1?h-1:h],b=[i&&d>1?d-1:d,i&&f>1?f-1:f],v=y[0]/b[0],x=y[1]/b[1],w=1/v,k=1/x,S=2*Math.ceil(w)+2,I=2*Math.ceil(k)+2;for(let e=0;e<l;e++){const t=e*o[0];for(let e=0;e<c;e++){const n=t+e*o[1],r=Math.floor(e*w),a=Math.floor(r-S/2);for(let r=0;r<h;r++){const s=n+r*o[2],l=Math.floor(r*k),y=Math.floor(l-I/2);for(let n=0;n<p;n++){let o=0;for(let s=0;s<S;s++){const l=s+a;if(l<0||l>=d)continue;const p=t+l*u[1],m=l*v;if(e===Math.min(c-1,i?Math.round(m):Math.floor(m)))for(let e=0;e<I;e++){const t=e+y;if(t<0||t>=f)continue;const a=p+t*u[2],s=t*x;r===Math.min(h-1,i?Math.round(s):Math.floor(s))&&(o+=g[a+n])}}m[s+n]=o}}}}return n.makeTensorInfo(a.shape,a.dtype,m)}};var Q_={kernelName:mr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dims:s}=r;$E(a,"reverse");const i=a.shape.length,o=va.parseAxisParam(s,a.shape);if(0===i)return KE({inputs:{x:a},backend:n});const u=new Ha(a.shape,a.dtype),l=n.bufferSync(a);for(let e=0;e<u.size;e++){const t=u.indexToLoc(e),n=t.slice();o.forEach(e=>n[e]=a.shape[e]-1-n[e]),u.set(l.get(...n),...t)}return n.makeTensorInfo(u.shape,u.dtype,u.values)}},Z_={kernelName:na,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{const{image:r}=e,{radians:a,fillValue:s,center:i}=t,o=n,u=va.getTypedArrayFromDType(r.dtype,va.sizeFromShape(r.shape)),[l,c,h,p]=r.shape,[d,f]=jd.getImageCenter(i,c,h),m=Math.sin(a),g=Math.cos(a),y=o.data.get(r.dataId).values;for(let e=0;e<l;e++){const t=e*h*c*p;for(let e=0;e<c;e++){const n=e*(h*p);for(let r=0;r<h;r++){const a=r*p;for(let i=0;i<p;i++){const o=[l,e,r,i],b=o[2],v=o[1];let x=(b-d)*g-(v-f)*m,w=(b-d)*m+(v-f)*g;x=Math.round(x+d),w=Math.round(w+f);let k=s;if("number"!=typeof s&&(k=3===i?255:s[i]),x>=0&&x<h&&w>=0&&w<c){k=y[t+w*(h*p)+x*p+i]}u[t+n+a+i]=k}}}}return{dataId:o.write(u,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},J_=lN(gr,e=>{const t=Math.floor(e);return e-t<.5?Math.floor(e):e-t>.5?Math.ceil(e):t%2==0?t:t+1}),eF={kernelName:gr,backendName:"cpu",kernelFunc:J_};function tF(e,t,n,r,a,s,i,o,u,l){const c=[r/a,a],h=e.values,p=t.values;if(0===r)return Ii(n,t.dtype);const d=Ii(c,t.dtype);d.values.fill(u);for(let e=0;e<s;e++){const s=[];let u=0;for(let t=0;t<i;t++){const n=h[e*i+t];s.push(n),u+=n*o[t]}if(u<0||u>=r/a)throw new Error(`Invalid indices: ${s} does not index into ${n}`);for(let n=0;n<a;n++)l?d.values[u*a+n]+=p[e*a+n]:d.values[u*a+n]=0===t.rank?p[0]:p[e*a+n]}return d}var nF={kernelName:br,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{indices:a,updates:s}=t,{shape:i}=r,{sliceRank:o,numUpdates:u,sliceSize:l,strides:c,outputSize:h}=jd.calculateShapes(s,a,i),p=tF(n.bufferSync(a),n.bufferSync(s),i,h,l,u,o,c,0,!0);return n.makeTensorInfo(i,p.dtype,p.values)}};var rF={kernelName:vr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{condition:r,t:a,e:s}=t;$E([r,a,s],"select");const i=r.shape.length,o=n.data.get(r.dataId).values,u=n.data.get(a.dataId).values,l=n.data.get(s.dataId).values,c=cs(a.dtype,s.dtype),h=va.makeZerosTypedArray(va.sizeFromShape(a.shape),c);let p=0;const d=0===i||i>1||1===a.shape.length?1:va.sizeFromShape(a.shape.slice(1));for(let e=0;e<o.length;e++)for(let t=0;t<d;t++)1===o[e]?h[p++]=u[e]:h[p++]=l[e];return n.makeTensorInfo(a.shape,c,h)}},aF=jd.SELU_SCALEALPHA,sF=jd.SELU_SCALE,iF=lN(xr,e=>e>=0?sF*e:aF*(Math.exp(e)-1)),oF={kernelName:xr,backendName:"cpu",kernelFunc:iF},uF=lN(Ir,e=>e<0?-1:e>0?1:0),lF={kernelName:Ir,backendName:"cpu",kernelFunc:uF},cF=lN(kr,e=>Math.sin(e)),hF={kernelName:kr,backendName:"cpu",kernelFunc:cF},pF=lN(Sr,e=>Math.sinh(e)),dF={kernelName:Sr,backendName:"cpu",kernelFunc:pF},fF=Math.log(1.1920928955078125e-7)+2,mF=lN(Er,e=>{const t=e>-fF,n=e<fF,r=Math.exp(e);let a;return a=n?r:t?e:Math.log(1+r),a}),gF={kernelName:Er,backendName:"cpu",kernelFunc:mF};var yF={kernelName:Tr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,paddings:i}=r;$E([a],"spaceToBatchND");const o=va.sizeFromShape(s),u=[[0,0]];u.push(...i);for(let e=1+s.length;e<a.shape.length;++e)u.push([0,0]);const l=z_.kernelFunc({inputs:{x:a},backend:n,attrs:{paddings:u,constantValue:0}}),c=jd.getReshaped(l.shape,s,o,!1),h=jd.getPermuted(c.length,s.length,!1),p=jd.getReshapedPermuted(l.shape,s,o,!1),d=rT({inputs:{x:l},backend:n,attrs:{shape:c}}),f=uC({inputs:{x:d},backend:n,attrs:{perm:h}}),m=rT({inputs:{x:f},backend:n,attrs:{shape:p}});return n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(f),m}};var bF={kernelName:Fr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{indices:r,values:a,denseShape:s,defaultValue:i}=t;if(1!==s.shape.length)throw new Error("Dense shape must be a vector, saw:\n        "+s.shape);if(2!==r.shape.length)throw new Error("Indices must be a matrix, saw:\n        "+r.shape);if(1!==a.shape.length)throw new Error("Values must be a vector, saw:\n        "+a.shape);if(0!==i.shape.length)throw new Error("Default value must be a scalar, saw:\n        "+i.shape);const o=n.data.get(r.dataId).values,u=n.data.get(a.dataId).values,l=n.data.get(s.dataId).values,c=n.data.get(i.dataId).values[0],[h,p,d,f,m]=kC(o,r.shape,r.dtype,u,a.dtype,l,c);return[n.makeTensorInfo(p,r.dtype,h),n.makeTensorInfo([p[0]],a.dtype,d),n.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(e=>Number(e)))),n.makeTensorInfo([m.length],r.dtype,new Int32Array(m))]}};var vF={kernelName:Mr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{inputIndices:r,inputShape:a,newShape:s}=t;if(2!==r.shape.length)throw new Error("Input indices should be a matrix but received shape\n        "+r.shape);if(1!==a.shape.length)throw new Error("Input shape should be a vector but received shape\n        "+a.shape);if(1!==s.shape.length)throw new Error("Target shape should be a vector but received shape "+s.shape);const i=Array.from(n.data.get(a.dataId).values),o=n.data.get(r.dataId).values,u=Array.from(n.data.get(s.dataId).values),[l,c,h]=SC(o,r.shape,r.dtype,i,u);return[n.makeTensorInfo(c,r.dtype,l),n.makeTensorInfo([h.length],s.dtype,new Int32Array(h))]}};var xF={kernelName:Dr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error("Indices should be a vector but received shape\n          "+a.shape);if(1!==s.shape.length)throw new Error("Segment ids should be a vector but received shape\n          "+s.shape);if(a.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");const i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values,u=n.data.get(s.dataId).values,[l,c]=IC(i,r.shape,r.dtype,o,u,!0);return n.makeTensorInfo(c,r.dtype,l)}};var wF={kernelName:Or,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error("Indices should be a vector but received shape\n         "+a.shape);if(1!==s.shape.length)throw new Error("Segment ids should be a vector but received shape\n         "+s.shape);if(a.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");const i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values,u=n.data.get(s.dataId).values,[l,c]=IC(i,r.shape,r.dtype,o,u);return n.makeTensorInfo(c,r.dtype,l)}};var kF={kernelName:Lr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=t,{outputShape:o}=r,{sliceRank:u,numUpdates:l,sliceSize:c,strides:h,outputSize:p}=jd.calculateShapes(s,a,o),d=tF(n.bufferSync(a),n.bufferSync(s),o,p,c,l,u,h,n.data.get(i.dataId).values[0],!1);return n.makeTensorInfo(o,d.dtype,d.values)}};var SF={kernelName:Rr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=r,o=va.parseAxisParam(i,a.shape)[0],u=jd.prepareSplitSize(a,s,o),l=new Array(a.shape.length).fill(0),c=a.shape.slice();return u.map(e=>{const t=[...c];t[o]=e;const r=xC({inputs:{x:a},backend:n,attrs:{begin:l,size:t}});return l[o]+=e,r})}},IF={kernelName:$r,backendName:"cpu",kernelFunc:({inputs:e,backend:t})=>{const{x:n}=e,r=t;$E(n,"square");const a=r.data.get(n.dataId).values,s=new Float32Array(a.length);for(let e=0;e<a.length;++e){const t=a[e];s[e]=t*t}return{dataId:r.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},AF=lN(ea,(e,t)=>{const n=t;return isNaN(e)?NaN:e>0?1:n.alpha}),EF={kernelName:ea,backendName:"cpu",kernelFunc:AF};var NF={kernelName:Br,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,end:i,strides:o,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:h,shrinkAxisMask:p}=r;$E(a,"stridedSlice");const{finalShapeSparse:d,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:b,end:v,strides:x}=fo.sliceInfo(a.shape,s,i,o,u,l,c,h,p);let w;if(m)w=rT({inputs:{x:a},backend:n,attrs:{shape:f}});else if(g||y){va.assert(a.shape.length>=1,()=>"Input must have rank at least 1, got: "+a.shape.length);const e=fo.computeOutShape(b,v,x),t=xC({inputs:{x:a},backend:n,attrs:{begin:b,size:e}});w=rT({inputs:{x:t},backend:n,attrs:{shape:f}}),n.disposeIntermediateTensorInfo(t)}else{const e=_C(d,n.bufferSync(a),x,b);w=n.makeTensorInfo(f,e.dtype,e.values)}return w}};var CF={kernelName:zr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:u,preserveShortSequences:l}=r,{data:c,dataSplits:h}=t,p=n.data.get(c.dataId).values,d=n.data.get(h.dataId).values,[f,m]=FC(p,d,a,s,i,o,u,l);return[n.makeTensorInfo([f.length],"string",f),n.makeTensorInfo(h.shape,"int32",m)]}};var TF={kernelName:Ur,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{skipEmpty:a}=r,{input:s,delimiter:i}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(1!==s.shape.length)throw new Error("Input must be a vector, got shape: "+s.shape);if(0!==i.shape.length)throw new Error("Delimiter must be a scalar, got shape: "+i.shape);const o=n.data.get(s.dataId).values,u=n.data.get(i.dataId).values[0],[l,c,h]=DC(o,u,a),p=c.length;return[n.makeTensorInfo([p,2],"int32",l),n.makeTensorInfo([p],"string",c),n.makeTensorInfo([2],"int32",new Int32Array(h))]}};var RF={kernelName:Wr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{numBuckets:a}=r,{input:s}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(a<=0)throw new Error("Number of buckets must be at least 1");const i=OC(n.data.get(s.dataId).values,a);return n.makeTensorInfo(s.shape,"int32",i)}},_F=lN(jr,e=>Math.tan(e)),FF={kernelName:jr,backendName:"cpu",kernelFunc:_F},MF=lN(Gr,e=>Math.tanh(e));function DF(e,t,n){switch(n){case"reflect":return function(e,t){let n=e;if(n<0)if(t<=1)n=0;else{const e=2*t;n<e&&(n=e*Math.trunc(-n/e)+n),n=n<-t?n+e:-n-1}else if(n>t-1)if(t<=1)n=0;else{const e=2*t;n-=e*Math.trunc(n/e),n>=t&&(n=e-n-1)}return va.clamp(0,n,t-1)}(e,t);case"wrap":return function(e,t){let n=e;if(n<0)if(t<=1)n=0;else{const e=t-1;n+=t*(Math.trunc(-n/e)+1)}else if(n>t-1)if(t<=1)n=0;else{const e=t-1;n-=t*Math.trunc(n/e)}return va.clamp(0,n,t-1)}(e,t);case"nearest":return function(e,t){return va.clamp(0,e,t-1)}(e,t);case"constant":default:return function(e,t){return e}(e)}}function OF(e,t,n,r,a,s,i,o,u,l,c){return 0<=o&&o<t&&0<=u&&u<n?e[i*r+o*a+u*s+l]:c}function LF(e,t,n,r,a,s,i,o,u,l,c){return OF(e,t,n,r,a,s,i,Math.round(o),Math.round(u),l,c)}function PF(e,t,n,r,a,s,i,o,u,l,c){const h=Math.floor(o),p=Math.floor(u),d=h+1,f=p+1;return(d-o)*((f-u)*OF(e,t,n,r,a,s,i,h,p,l,c)+(u-p)*OF(e,t,n,r,a,s,i,h,f,l,c))+(o-h)*((f-u)*OF(e,t,n,r,a,s,i,d,p,l,c)+(u-p)*OF(e,t,n,r,a,s,i,d,f,l,c))}var $F=[oT,VE,lT,hT,sN,pT,dT,fT,mT,gT,bT,xT,kT,AT,NT,_T,FT,MT,DT,iT,OT,LT,PT,$T,JE,dN,zT,HE,UT,GT,qT,KT,YT,XT,QT,ZT,eR,nR,rR,aR,sR,iR,uR,lR,cR,hR,pR,dR,fR,yR,HC,bR,yN,ER,xN,CR,SN,DR,LR,PR,EN,zR,UR,WR,VR,jR,_N,DN,YE,GR,VT,qR,YR,QR,KC,PN,zN,ZR,jN,e_,r_,s_,u_,l_,c_,p_,KN,d_,f_,m_,g_,y_,b_,v_,QN,x_,S_,N_,tC,rC,T_,__,M_,iC,D_,P_,B_,z_,V_,QC,hC,j_,QE,_R,H_,JC,tT,aT,q_,K_,Y_,X_,Q_,Z_,eF,mC,nF,rF,oF,bC,lF,hF,dF,wC,E_,gF,yF,bF,vF,xF,wF,kF,SF,NC,IF,RC,EF,NF,CF,TF,RF,BC,gR,FF,{kernelName:Gr,backendName:"cpu",kernelFunc:MF},{kernelName:Hr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reps:s}=r;$E(a,"tile");const i=zC(n.bufferSync(a),s);return n.makeTensorInfo(i.shape,i.dtype,i.values)}},{kernelName:qr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{k:s,sorted:i}=r;$E(a,"topk");const o=n.data.get(a.dataId).values,[u,l]=VC(o,a.shape,a.dtype,s,i);return[n.makeTensorInfo(u.shape,u.dtype,u.values),n.makeTensorInfo(l.shape,l.dtype,l.values)]}},{kernelName:Kr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:u,outputShape:l}=n,[c,h,p,d]=a.shape,[f,m]=null!=l?l:[h,p],g=[c,f,m,d],y=va.computeStrides(a.shape),b=y[0],v=y[1],x=y[2],w=va.getTypedArrayFromDType(a.dtype,va.sizeFromShape(g));w.fill(u);const k=r.data.get(a.dataId).values,S=r.data.get(s.dataId).values;for(let e=0;e<c;++e){const t=1===s.shape[0]?S:S.subarray(8*e,8*e+8);for(let n=0;n<f;++n)for(let r=0;r<m;++r)for(let a=0;a<d;++a){let s;const l=t[6]*r+t[7]*n+1;if(0===l)continue;const c=(t[0]*r+t[1]*n+t[2])/l,d=(t[3]*r+t[4]*n+t[5])/l,f=DF(c,p,o),m=DF(d,h,o);switch(i){case"nearest":s=LF(k,h,p,b,v,x,e,m,f,a,u);break;case"bilinear":s=PF(k,h,p,b,v,x,e,m,f,a,u);break;default:throw new Error("Error in Transform: Expect 'nearest' or 'bilinear', but got "+i)}w[e*b+n*v+r*x+a]=s}return r.makeTensorInfo(g,a.dtype,w)}return{dataId:r.write(w,g,a.dtype),shape:a.shape,dtype:a.dtype}}},lC,{kernelName:Xr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{axis:a}=n,{x:s}=t;$E(s,"unique");const i=r.data.get(s.dataId).values,{outputValues:o,outputShape:u,indices:l}=jC(i,a,s.shape,s.dtype);return[r.makeTensorInfo(u,s.dtype,o),r.makeTensorInfo([l.length],"int32",l)]}},{kernelName:Qr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{value:a}=t;let{axis:s}=r;s<0&&(s+=a.shape.length);const i=a.shape.length,o=a.shape[s],u=new Array(i-1);let l=0;for(let e=0;e<i;e++)e!==s&&(u[l++]=a.shape[e]);const c=new Array(i).fill(0),h=a.shape.slice();h[s]=1;const p=new Array(o);for(let e=0;e<p.length;e++){c[s]=e;const t=xC({inputs:{x:a},backend:n,attrs:{begin:c,size:h}});p[e]=rT({inputs:{x:t},backend:n,attrs:{shape:u}}),n.disposeIntermediateTensorInfo(t)}return p}},{kernelName:Zr,backendName:"cpu",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,segmentIds:s}=t,{numSegments:i}=r;$E(a,"unsortedSegmentSum");const o=[],u=[],l=a.shape.length-s.shape.length;let c=s;for(let e=0;e<l;++e){const t=NR({inputs:{input:c},backend:n,attrs:{dim:e+1}});c=t,u.push(t)}for(let e=0;e<i;++e){const t=va.createScalarValue(e,"int32"),r=n.makeTensorInfo([],"int32",t),s=gN({inputs:{a:r,b:c},backend:n}),i=ZE({inputs:{x:s},backend:n,attrs:{dtype:"float32"}}),l=eC({inputs:{a:i,b:a},backend:n}),h=mR({inputs:{x:l},backend:n,attrs:{axis:0,keepDims:!1}});o.push(h),u.push(r),u.push(s),u.push(i),u.push(l),u.push(h)}const h=$_({inputs:o,backend:n,attrs:{axis:0}});return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),h}},L_];for(const e of $F)da(e);var BF={};w(BF,{assertNotComplex:()=>KM,bindCanvasToFramebuffer:()=>IM,bindColorTextureToFramebuffer:()=>AM,bindTextureToProgramUniformSampler:()=>SM,bindTextureUnit:()=>vM,bindVertexBufferToProgramAttribute:()=>bM,callAndCheck:()=>eM,canBeRepresented:()=>tM,createFragmentShader:()=>sM,createFramebuffer:()=>yM,createProgram:()=>lM,createStaticIndexBuffer:()=>dM,createStaticVertexBuffer:()=>pM,createTexture:()=>mM,createVertexShader:()=>aM,getBatchDim:()=>_M,getExtensionOrThrow:()=>rM,getFramebufferErrorMessage:()=>CM,getMaxTexturesInShader:()=>zM,getNumChannels:()=>fM,getProgramUniformLocation:()=>kM,getProgramUniformLocationOrThrow:()=>wM,getRowsCols:()=>FM,getShapeAs3D:()=>MM,getTextureShapeFromLogicalShape:()=>DM,getWebGLDisjointQueryTimerVersion:()=>UM,getWebGLErrorMessage:()=>nM,getWebGLMaxTextureSize:()=>PM,hasExtension:()=>WM,isCapableOfRenderingToFloatTexture:()=>jM,isDownloadFloatTextureEnabled:()=>GM,isReshapeFree:()=>LM,isWebGLFenceEnabled:()=>qM,isWebGLVersionEnabled:()=>VM,linkProgram:()=>cM,resetMaxTextureSize:()=>$M,resetMaxTexturesInShader:()=>BM,unbindColorTextureFromFramebuffer:()=>EM,unbindTextureUnit:()=>xM,validateFramebuffer:()=>NM,validateProgram:()=>hM,validateTextureSize:()=>gM});var zF,UF,WF,VF,jF,GF,HF={},qF={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function KF(e,t){HF[e]=t}function YF(e,t){if(!(e in HF)){const n=function(e,t){if(1!==e&&2!==e)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");const n=null==t?function(e){if("undefined"!=typeof OffscreenCanvas&&2===e)return new OffscreenCanvas(300,150);if("undefined"!=typeof document)return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}(e):t;if(n.addEventListener("webglcontextlost",t=>{t.preventDefault(),delete HF[e]},!1),1===e)return n.getContext("webgl",qF)||n.getContext("experimental-webgl",qF);return n.getContext("webgl2",qF)}(e,t);if(null===n)return console.log("Could not get context for WebGL version",e),null;HF[e]=n}const n=HF[e];return null==n||n.isContextLost()?(delete HF[e],YF(e)):(n.disable(n.DEPTH_TEST),n.disable(n.STENCIL_TEST),n.disable(n.BLEND),n.disable(n.DITHER),n.disable(n.POLYGON_OFFSET_FILL),n.disable(n.SAMPLE_COVERAGE),n.enable(n.SCISSOR_TEST),n.enable(n.CULL_FACE),n.cullFace(n.BACK),HF[e])}function XF(e,t){return[t,e]}function QF(e){const t=va.sizeFromShape(e),n=Math.ceil(t/4);return va.sizeToSquarishShape(n)}function ZF(e,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(e/2))]}function JF(e,t){const n=e;let r,a,s,i,o,u,l,c,h,p;return 2===Ye().getNumber("WEBGL_VERSION")?(r=n.R32F,a=n.R16F,s=n.RGBA16F,i=n.RGBA32F,o=n.RED,l=4,c=1,h=n.HALF_FLOAT,p=n.FLOAT,u=n.RGBA8):(r=e.RGBA,a=e.RGBA,s=e.RGBA,i=n.RGBA,o=e.RGBA,l=4,c=4,h=null!=t?t.HALF_FLOAT_OES:null,p=e.FLOAT,u=e.RGBA),{internalFormatFloat:r,internalFormatHalfFloat:a,internalFormatPackedHalfFloat:s,internalFormatPackedFloat:i,textureFormatFloat:o,downloadTextureFormat:u,downloadUnpackNumChannels:l,defaultNumChannels:c,textureTypeHalfFloat:h,textureTypeFloat:p}}function eM(e,t){const n=t();return Ye().getBool("DEBUG")&&function(e){const t=e.getError();if(t!==e.NO_ERROR)throw new Error("WebGL Error: "+nM(e,t))}(e),n}(UF=zF||(zF={}))[UF.DENSE=0]="DENSE",UF[UF.SHARED_BATCH=1]="SHARED_BATCH",(VF=WF||(WF={}))[VF.RENDER=0]="RENDER",VF[VF.UPLOAD=1]="UPLOAD",VF[VF.PIXELS=2]="PIXELS",VF[VF.DOWNLOAD=3]="DOWNLOAD",(GF=jF||(jF={}))[GF.UNPACKED_FLOAT16=0]="UNPACKED_FLOAT16",GF[GF.UNPACKED_FLOAT32=1]="UNPACKED_FLOAT32",GF[GF.PACKED_4X1_UNSIGNED_BYTE=2]="PACKED_4X1_UNSIGNED_BYTE",GF[GF.PACKED_2X2_FLOAT32=3]="PACKED_2X2_FLOAT32",GF[GF.PACKED_2X2_FLOAT16=4]="PACKED_2X2_FLOAT16";function tM(e){return!!(Ye().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||0===e||5.96e-8<Math.abs(e)&&Math.abs(e)<65504)}function nM(e,t){switch(t){case e.NO_ERROR:return"NO_ERROR";case e.INVALID_ENUM:return"INVALID_ENUM";case e.INVALID_VALUE:return"INVALID_VALUE";case e.INVALID_OPERATION:return"INVALID_OPERATION";case e.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case e.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case e.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return"Unknown error code "+t}}function rM(e,t){return TM(e,()=>e.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function aM(e,t){const n=TM(e,()=>e.createShader(e.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(eM(e,()=>e.shaderSource(n,t)),eM(e,()=>e.compileShader(n)),!1===e.getShaderParameter(n,e.COMPILE_STATUS))throw console.log(e.getShaderInfoLog(n)),new Error("Failed to compile vertex shader.");return n}function sM(e,t){const n=TM(e,()=>e.createShader(e.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(eM(e,()=>e.shaderSource(n,t)),eM(e,()=>e.compileShader(n)),!1===e.getShaderParameter(n,e.COMPILE_STATUS))throw function(e,t){const n=uM.exec(t);if(null==n)return console.log("Couldn't parse line number in error: "+t),void console.log(e);const r=+n[1],a=e.split("\n"),s=a.length.toString().length+2,i=a.map((e,t)=>va.rightPad((t+1).toString(),s)+e);let o=0;for(let e=0;e<i.length;e++)o=Math.max(i[e].length,o);const u=i.slice(0,r-1),l=i.slice(r-1,r),c=i.slice(r);console.log(u.join("\n")),console.log(t.split("\n")[0]),console.log("%c "+va.rightPad(l[0],o),"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(c.join("\n"))}(t,e.getShaderInfoLog(n)),new Error("Failed to compile fragment shader.");return n}var iM,oM,uM=/ERROR: [0-9]+:([0-9]+):/g;function lM(e){return TM(e,()=>e.createProgram(),"Unable to create WebGLProgram.")}function cM(e,t){if(eM(e,()=>e.linkProgram(t)),!1===e.getProgramParameter(t,e.LINK_STATUS))throw console.log(e.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function hM(e,t){if(eM(e,()=>e.validateProgram(t)),!1===e.getProgramParameter(t,e.VALIDATE_STATUS))throw console.log(e.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function pM(e,t){const n=TM(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return eM(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),eM(e,()=>e.bufferData(e.ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function dM(e,t){const n=TM(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return eM(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,n)),eM(e,()=>e.bufferData(e.ELEMENT_ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function fM(){return 2===Ye().getNumber("WEBGL_VERSION")?1:4}function mM(e){return TM(e,()=>e.createTexture(),"Unable to create WebGLTexture.")}function gM(e,t){const n=Ye().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(e<=0||t<=0){throw new Error("Requested texture size "+`[${e}x${t}]`+" is invalid.")}if(e>n||t>n){throw new Error("Requested texture size "+`[${e}x${t}]`+" greater than WebGL maximum on this browser / GPU "+`[${n}x${n}]`+".")}}function yM(e){return TM(e,()=>e.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function bM(e,t,n,r,a,s,i){const o=e.getAttribLocation(t,n);return-1!==o&&(eM(e,()=>e.bindBuffer(e.ARRAY_BUFFER,r)),eM(e,()=>e.vertexAttribPointer(o,a,e.FLOAT,!1,s,i)),eM(e,()=>e.enableVertexAttribArray(o)),!0)}function vM(e,t,n){RM(e,n),eM(e,()=>e.activeTexture(e.TEXTURE0+n)),eM(e,()=>e.bindTexture(e.TEXTURE_2D,t))}function xM(e,t){RM(e,t),eM(e,()=>e.activeTexture(e.TEXTURE0+t)),eM(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function wM(e,t,n){return TM(e,()=>e.getUniformLocation(t,n),'uniform "'+n+'" not present in program.')}function kM(e,t,n){return e.getUniformLocation(t,n)}function SM(e,t,n,r){eM(e,()=>vM(e,t,r)),eM(e,()=>e.uniform1i(n,r))}function IM(e){eM(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),eM(e,()=>e.viewport(0,0,e.canvas.width,e.canvas.height)),eM(e,()=>e.scissor(0,0,e.canvas.width,e.canvas.height))}function AM(e,t,n){eM(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,n)),eM(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,t,0))}function EM(e,t){eM(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,t)),eM(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,null,0))}function NM(e){const t=e.checkFramebufferStatus(e.FRAMEBUFFER);if(t!==e.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+CM(e,t))}function CM(e,t){switch(t){case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case e.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return"unknown error "+t}}function TM(e,t,n){const r=eM(e,()=>t());if(null==r)throw new Error(n);return r}function RM(e,t){const n=e.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,r=t+e.TEXTURE0;if(r<e.TEXTURE0||r>n){throw new Error(`textureUnit must be in ${`[gl.TEXTURE0, gl.TEXTURE${n}]`}.`)}}function _M(e,t=2){return va.sizeFromShape(e.slice(0,e.length-t))}function FM(e){if(0===e.length)throw Error("Cannot get rows and columns of an empty shape array.");return[e.length>1?e[e.length-2]:1,e[e.length-1]]}function MM(e){let t=[1,1,1];return 0===e.length||1===e.length&&1===e[0]||(t=[_M(e),...FM(e)]),t}function DM(e,t=!1){let n=Ye().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(t&&(n*=2,1===(e=e.map((t,n)=>n>=e.length-2?va.nearestLargerEven(e[n]):e[n])).length&&(e=[2,e[0]])),2!==e.length){const t=va.squeezeShape(e);e=t.newShape}let r=va.sizeFromShape(e);if(e.length<=1&&r<=n)return[1,r];if(2===e.length&&e[0]<=n&&e[1]<=n)return e;if(3===e.length&&e[0]*e[1]<=n&&e[2]<=n)return[e[0]*e[1],e[2]];if(3===e.length&&e[0]<=n&&e[1]*e[2]<=n)return[e[0],e[1]*e[2]];if(4===e.length&&e[0]*e[1]*e[2]<=n&&e[3]<=n)return[e[0]*e[1]*e[2],e[3]];if(4===e.length&&e[0]<=n&&e[1]*e[2]*e[3]<=n)return[e[0],e[1]*e[2]*e[3]];if(t){const t=_M(e);let n=2,a=2;return e.length&&([n,a]=FM(e)),r=t*(n/2)*(a/2),va.sizeToSquarishShape(r).map(e=>2*e)}return va.sizeToSquarishShape(r)}function OM(e){return e%2==0}function LM(e,t){if(e=e.slice(-2),t=t.slice(-2),va.arraysEqual(e,t))return!0;if(!e.length||!t.length)return!0;if(0===e[0]||0===e[1]||0===t[0]||0===t[1])return!0;if(e.length!==t.length){const n=e.slice(-1)[0],r=t.slice(-1)[0];if(n===r)return!0;if(OM(n)&&OM(r)&&(1===e[0]||1===t[0]))return!0}return e[1]===t[1]&&OM(e[0])&&OM(t[0])}function PM(e){if(null==iM){const t=YF(e);iM=t.getParameter(t.MAX_TEXTURE_SIZE)}return iM}function $M(){iM=null}function BM(){oM=null}function zM(e){if(null==oM){const t=YF(e);oM=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,oM)}function UM(e){if(0===e)return 0;let t;const n=YF(e);return t=WM(n,"EXT_disjoint_timer_query_webgl2")&&2===e?2:WM(n,"EXT_disjoint_timer_query")?1:0,t}function WM(e,t){return null!=e.getExtension(t)}function VM(e){try{if(null!=YF(e))return!0}catch(e){return console.log("Error when getting WebGL context: ",e),!1}return!1}function jM(e){if(0===e)return!1;const t=YF(e);if(1===e){if(!WM(t,"OES_texture_float"))return!1}else if(!WM(t,"EXT_color_buffer_float"))return!1;return HM(t)}function GM(e){if(0===e)return!1;const t=YF(e);if(1!==e){if(WM(t,"EXT_color_buffer_float"))return HM(t);const e="EXT_color_buffer_half_float";if(WM(t,e)){const n=t.getExtension(e);return function(e,t){const n=JF(e,t),r=e.createTexture();e.bindTexture(e.TEXTURE_2D,r);e.texImage2D(e.TEXTURE_2D,0,n.internalFormatHalfFloat,1,1,0,n.textureFormatFloat,n.textureTypeHalfFloat,null);const a=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,a),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,r,0);const s=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(r),e.deleteFramebuffer(a),s}(t,n)}return!1}if(!WM(t,"OES_texture_float"))return!1;if(!WM(t,"WEBGL_color_buffer_float"))return!1;return HM(t)}function HM(e){const t=JF(e),n=e.createTexture();e.bindTexture(e.TEXTURE_2D,n);e.texImage2D(e.TEXTURE_2D,0,t.internalFormatFloat,1,1,0,t.textureFormatFloat,t.textureTypeFloat,null);const r=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,r),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,n,0);const a=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(n),e.deleteFramebuffer(r),a}function qM(e){if(2!==e)return!1;return null!=YF(e).fenceSync}function KM(e,t){Array.isArray(e)||(e=[e]),e.forEach(e=>{null!=e&&va.assert("complex64"!==e.dtype,()=>t+" does not support complex64 tensors in the WebGL backend.")})}var YM=Ye();function XM(){let e,t,n,r,a,s,i,o,u,l;return 2===Ye().getNumber("WEBGL_VERSION")?(e="#version 300 es",t="in",n="out",r="in",a="texture",s="outputColor",i="out vec4 outputColor;",o="\n      bool isnan_custom(float val) {\n        return (val > 0.0 || val < 0.0) ? false : val != 0.0;\n      }\n\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan_custom(val.x),\n          isnan_custom(val.y), isnan_custom(val.z), isnan_custom(val.w));\n      }\n\n      #define isnan(value) isnan_custom(value)\n    ",u="",l="\n      #define round(value) newRound(value)\n      int newRound(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 newRound(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    "):(e="",t="attribute",n="varying",r="varying",a="texture2D",s="gl_FragColor",i="",o="\n      #define isnan(value) isnan_custom(value)\n      bool isnan_custom(float val) {\n        return (val > 0. || val < 1. || val == 0.) ? false : true;\n      }\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));\n      }\n    ",u="\n      uniform float INFINITY;\n\n      bool isinf(float val) {\n        return abs(val) == INFINITY;\n      }\n      bvec4 isinf(vec4 val) {\n        return equal(abs(val), vec4(INFINITY));\n      }\n    ",l="\n      int round(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 round(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    "),{version:e,attribute:t,varyingVs:n,varyingFs:r,texture2D:a,output:s,defineOutput:i,defineSpecialNaN:o,defineSpecialInf:u,defineRound:l}}function QM(e,t,n="index"){const r=va.computeStrides(t);return r.map((t,a)=>`${`int ${e[a]} = ${n} / ${t}`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${t}`:`index -= ${e[a]} * ${t}`};`).join("")}function ZM(e,t,n="index"){const r=va.computeStrides(t);return r.map((t,a)=>`${`int ${e[a]} = ${n} / outShapeStrides[${a}]`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * outShapeStrides[${a}]`:`index -= ${e[a]} * outShapeStrides[${a}]`};`).join("")}function JM(e,t,n="index"){const r=function(e,t){const n=e.length,r=e.map(e=>`${t}[${e}]`),a=new Array(n-1);a[n-2]=r[n-1];for(let e=n-3;e>=0;--e)a[e]=`(${a[e+1]} * ${r[e+1]})`;return a}(e.map((e,t)=>t),t);return r.map((t,a)=>`${`int ${e[a]} = ${n} / ${r[a]}`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${r[a]}`:`index -= ${e[a]} * ${r[a]}`};`).join("")}function eD(e){const t=va.computeStrides(e).map(e=>e.toString());return`\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;\n  }\n`}YM.registerFlag("HAS_WEBGL",()=>YM.getNumber("WEBGL_VERSION")>0),YM.registerFlag("WEBGL_VERSION",()=>VM(2)?2:VM(1)?1:0),YM.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1),YM.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>2===YM.get("WEBGL_VERSION")),YM.registerFlag("WEBGL_CPU_FORWARD",()=>!0),YM.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1),YM.registerFlag("WEBGL_PACK",()=>YM.getBool("HAS_WEBGL")),YM.registerFlag("WEBGL_PACK_NORMALIZATION",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_CLIP",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_PACK_REDUCE",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_LAZILY_UNPACK",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_CONV_IM2COL",()=>YM.getBool("WEBGL_PACK")),YM.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>PM(YM.getNumber("WEBGL_VERSION"))),YM.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>zM(YM.getNumber("WEBGL_VERSION"))),YM.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{const e=YM.getNumber("WEBGL_VERSION");return 0===e?0:UM(e)}),YM.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>YM.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!Ss.isMobile()),YM.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>jM(YM.getNumber("WEBGL_VERSION"))),YM.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>!YM.getBool("WEBGL_FORCE_F16_TEXTURES")&&YM.getBool("WEBGL_RENDER_FLOAT32_CAPABLE")),YM.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>GM(YM.getNumber("WEBGL_VERSION"))),YM.registerFlag("WEBGL_FENCE_API_ENABLED",()=>qM(YM.getNumber("WEBGL_VERSION"))),YM.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>YM.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0),YM.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,e=>{if(e<0&&-1!==e)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)}),YM.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>Ss.isMobile()?1:-1,e=>{if(e<0&&-1!==e)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)}),YM.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128),YM.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1),YM.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5),YM.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128);var tD="\n  const float FLOAT_MAX = 1.70141184e38;\n  const float FLOAT_MIN = 1.17549435e-38;\n\n  lowp vec4 encode_float(highp float v) {\n    if (isnan(v)) {\n      return vec4(255, 255, 255, 255);\n    }\n\n    highp float av = abs(v);\n\n    if(av < FLOAT_MIN) {\n      return vec4(0.0, 0.0, 0.0, 0.0);\n    } else if(v > FLOAT_MAX) {\n      return vec4(0.0, 0.0, 128.0, 127.0) / 255.0;\n    } else if(v < -FLOAT_MAX) {\n      return vec4(0.0, 0.0,  128.0, 255.0) / 255.0;\n    }\n\n    highp vec4 c = vec4(0,0,0,0);\n\n    highp float e = floor(log2(av));\n    highp float m = exp2(fract(log2(av))) - 1.0;\n\n    c[2] = floor(128.0 * m);\n    m -= c[2] / 128.0;\n    c[1] = floor(32768.0 * m);\n    m -= c[1] / 32768.0;\n    c[0] = floor(8388608.0 * m);\n\n    highp float ebias = e + 127.0;\n    c[3] = floor(ebias / 2.0);\n    ebias -= c[3] * 2.0;\n    c[2] += floor(ebias) * 128.0;\n\n    c[3] += 128.0 * step(0.0, -v);\n\n    return c / 255.0;\n  }\n",{getBroadcastDims:nD}=jd;function rD(e,t,n){const r=[];if(e.forEach(e=>{const t=va.sizeFromShape(e.shapeInfo.logicalShape);if(e.shapeInfo.isUniform?r.push(`uniform float ${e.name}${t>1?`[${t}]`:""};`):(r.push(`uniform sampler2D ${e.name};`),r.push(`uniform int offset${e.name};`)),n.enableShapeUniforms){const{uniformShape:t}=fD(n.packedInputs,e.shapeInfo.logicalShape,e.shapeInfo.texShape);switch(t.length){case 1:r.push(`uniform int ${e.name}Shape;`);break;case 2:r.push(`uniform ivec2 ${e.name}Shape;`);break;case 3:r.push(`uniform ivec3 ${e.name}Shape;`);break;case 4:r.push(`uniform ivec4 ${e.name}Shape;`)}r.push(`uniform ivec2 ${e.name}TexShape;`)}}),n.enableShapeUniforms){switch(t.logicalShape.length){case 1:r.push("uniform int outShape;");break;case 2:r.push("uniform ivec2 outShape;"),r.push("uniform int outShapeStrides;");break;case 3:r.push("uniform ivec3 outShape;"),r.push("uniform ivec2 outShapeStrides;");break;case 4:r.push("uniform ivec4 outShape;"),r.push("uniform ivec3 outShapeStrides;")}r.push("uniform ivec2 outTexShape;")}n.customUniforms&&n.customUniforms.forEach(e=>{r.push(`uniform ${e.type} ${e.name}${e.arrayIndex?`[${e.arrayIndex}]`:""};`)});const a=r.join("\n"),s=e.map(e=>function(e,t,n=!1,r){let a="";a+=n?sD(e,r):aD(e,r);const s=e.shapeInfo.logicalShape,i=t.logicalShape;s.length<=i.length&&(a+=n?function(e,t){const n=e.name,r=n.charAt(0).toUpperCase()+n.slice(1),a="get"+r+"AtOutCoords",s=e.shapeInfo.logicalShape.length,i=t.logicalShape.length,o=nD(e.shapeInfo.logicalShape,t.logicalShape),u=dD(i),l=i-s;let c;const h=["x","y","z","w","u","v"];c=0===s?"":i<2&&o.length>=1?"coords = 0;":o.map(e=>`coords.${h[e+l]} = 0;`).join("\n");let p="";p=i<2&&s>0?"coords":e.shapeInfo.logicalShape.map((e,t)=>"coords."+h[t+l]).join(", ");let d="return outputValue;";const f=1===va.sizeFromShape(e.shapeInfo.logicalShape),m=1===va.sizeFromShape(t.logicalShape);if(1!==s||f||m){if(f&&!m)d=1===i?"\n        return vec4(outputValue.x, outputValue.x, 0., 0.);\n      ":"\n        return vec4(outputValue.x);\n      ";else if(o.length){const e=s-2,t=s-1;o.indexOf(e)>-1&&o.indexOf(t)>-1?d="return vec4(outputValue.x);":o.indexOf(e)>-1?d="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":o.indexOf(t)>-1&&(d="return vec4(outputValue.xx, outputValue.zz);")}}else d="\n      return vec4(outputValue.xy, outputValue.xy);\n    ";return`\n    vec4 ${a}() {\n      ${u} coords = getOutputCoords();\n      ${c}\n      vec4 outputValue = get${r}(${p});\n      ${d}\n    }\n  `}(e,t):function(e,t){const n=e.name,r=n.charAt(0).toUpperCase()+n.slice(1),a="get"+r+"AtOutCoords",s=t.texShape,i=e.shapeInfo.texShape,o=e.shapeInfo.logicalShape.length,u=t.logicalShape.length;if(!e.shapeInfo.isUniform&&o===u&&null==e.shapeInfo.flatOffset&&va.arraysEqual(i,s))return`\n      float ${a}() {\n        return sampleTexture(${n}, resultUV);\n      }\n    `;const l=dD(u),c=nD(e.shapeInfo.logicalShape,t.logicalShape),h=u-o;let p;const d=["x","y","z","w","u","v"];p=0===o?"":u<2&&c.length>=1?"coords = 0;":c.map(e=>`coords.${d[e+h]} = 0;`).join("\n");let f="";f=u<2&&o>0?"coords":e.shapeInfo.logicalShape.map((e,t)=>"coords."+d[t+h]).join(", ");return`\n    float ${a}() {\n      ${l} coords = getOutputCoords();\n      ${p}\n      return get${r}(${f});\n    }\n  `}(e,t));return a}(e,t,n.packedInputs,n.enableShapeUniforms)).join("\n"),i=t.texShape,o=XM(),u=function(e){return`\n    float sampleTexture(sampler2D textureSampler, vec2 uv) {\n      return ${e.texture2D}(textureSampler, uv).r;\n    }\n  `}(o);let l,c,h=function(e){return`${e.version}\n    precision highp float;\n    precision highp int;\n    precision highp sampler2D;\n    ${e.varyingFs} vec2 resultUV;\n    ${e.defineOutput}\n    const vec2 halfCR = vec2(0.5, 0.5);\n\n    struct ivec5\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n    };\n\n    struct ivec6\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n      int v;\n    };\n\n    uniform float NAN;\n    ${e.defineSpecialNaN}\n    ${e.defineSpecialInf}\n    ${e.defineRound}\n\n    int imod(int x, int y) {\n      return x - y * (x / y);\n    }\n\n    int idiv(int a, int b, float sign) {\n      int res = a / b;\n      int mod = imod(a, b);\n      if (sign < 0. && mod != 0) {\n        res -= 1;\n      }\n      return res;\n    }\n\n    //Based on the work of Dave Hoskins\n    //https://www.shadertoy.com/view/4djSRW\n    #define HASHSCALE1 443.8975\n    float random(float seed){\n      vec2 p = resultUV * seed;\n      vec3 p3  = fract(vec3(p.xyx) * HASHSCALE1);\n      p3 += dot(p3, p3.yzx + 19.19);\n      return fract((p3.x + p3.y) * p3.z);\n    }\n\n    ${iD}\n    ${oD}\n    ${uD}\n  `}(o);t.isPacked?(l=function(e,t,n){switch(e.length){case 0:return cD();case 1:return function(e,t,n){const r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(1===r[0])return n?"\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));\n      }\n    ":`\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ${r[1]}.0);\n      }\n    `;if(1===r[1])return n?"\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));\n      }\n    ":`\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ${r[0]}.0);\n      }\n    `;if(n)return"\n    int getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);\n    }\n  ";return`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      return 2 * (resTexRC.x * ${r[1]} + resTexRC.y);\n    }\n  `}(0,t,n);case 2:return function(e,t,n){const r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(va.arraysEqual(e,t))return n?"\n      ivec2 getOutputCoords() {\n        ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n        return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        return 2 * ivec2(resultUV.yx * vec2(${r[0]}, ${r[1]}));\n      }\n    `;const a=Math.ceil(e[1]/2);if(n)return"\n    ivec2 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec2(r, c);\n    }\n  ";return`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec2(r, c);\n    }\n  `}(e,t,n);case 3:return function(e,t,n){if(n)return"\n    ivec3 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec3(b, r, c);\n    }\n  ";const r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],a=Math.ceil(e[2]/2),s=a*Math.ceil(e[1]/2);return`\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `}(e,t,n);default:return function(e,t,n){if(n)return"\n    ivec4 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int texelsInLogicalRow = int(ceil(float(outShape[3]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatchN = texelsInBatch * outShape[1];\n\n      int b2 = index / texelsInBatchN;\n      index -= b2 * texelsInBatchN;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec4(b2, b, r, c);\n    }\n  ";const r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],a=Math.ceil(e[e.length-1]/2),s=a*Math.ceil(e[e.length-2]/2);let i=s,o="",u="b, r, c";for(let t=2;t<e.length-1;t++)i*=e[e.length-t-1],o=`\n      int b${t} = index / ${i};\n      index -= b${t} * ${i};\n    `+o,u=`b${t}, `+u;return`\n    ivec${e.length} getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n\n      ${o}\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec${e.length}(${u});\n    }\n  `}(e,t,n)}}(t.logicalShape,i,n.enableShapeUniforms),c=function(e){return`\n    void setOutput(vec4 val) {\n      ${e.output} = val;\n    }\n  `}(o)):(l=function(e,t,n){switch(e.length){case 0:return cD();case 1:return function(e,t,n){if(1===t[0])return n?"\n      int getOutputCoords() {\n        return int(resultUV.x * float(outTexShape[1]));\n      }\n    ":`\n      int getOutputCoords() {\n        return int(resultUV.x * ${t[1]}.0);\n      }\n    `;if(1===t[1])return n?"\n      int getOutputCoords() {\n        return int(resultUV.y * float(outTexShape[0]));\n      }\n    ":`\n      int getOutputCoords() {\n        return int(resultUV.y * ${t[0]}.0);\n      }\n    `;if(n)return"\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      return resTexRC.x * outTexShape[1] + resTexRC.y;\n    }\n  ";return`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      return resTexRC.x * ${t[1]} + resTexRC.y;\n    }\n  `}(0,t,n);case 2:return function(e,t,n){if(va.arraysEqual(e,t))return n?"\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));\n      }\n    `;if(1===e[1])return n?"\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `;if(1===e[0])return n?"\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(0, index);\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `;if(n)return"\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      int r = index / outShape[1];\n      int c = index - r * outShape[1];\n      return ivec2(r, c);\n    }\n  ";return`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      int r = index / ${e[1]};\n      int c = index - r * ${e[1]};\n      return ivec2(r, c);\n    }\n  `}(e,t,n);case 3:return function(e,t,n){if(n){return`\n  ivec3 getOutputCoords() {\n    ivec2 resTexRC = ivec2(resultUV.yx *\n                           vec2(outTexShape[0], outTexShape[1]));\n    int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n    ${ZM(["r","c","d"],e)}\n    return ivec3(r, c, d);\n  }\n`}const r=QM(["r","c","d"],e);return`\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${r}\n      return ivec3(r, c, d);\n    }\n  `}(e,t,n);case 4:return function(e,t,n){if(n){return`\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      ${ZM(["r","c","d","d2"],e)}\n      return ivec4(r, c, d, d2);\n    }\n  `}const r=QM(["r","c","d","d2"],e);return`\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${r}\n      return ivec4(r, c, d, d2);\n    }\n  `}(e,t,n);case 5:return function(e,t){const n=QM(["r","c","d","d2","d3"],e);return`\n    ivec5 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},\n                             ${t[1]}));\n\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${n}\n\n      ivec5 outShape = ivec5(r, c, d, d2, d3);\n      return outShape;\n    }\n  `}(e,t);case 6:return function(e,t){const n=QM(["r","c","d","d2","d3","d4"],e);return`\n    ivec6 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${n}\n\n      ivec6 result = ivec6(r, c, d, d2, d3, d4);\n      return result;\n    }\n  `}(e,t);default:throw new Error(e.length+"-D output sampling is not yet supported")}}(t.logicalShape,i,n.enableShapeUniforms),c=function(e){return`\n    void setOutput(float val) {\n      ${e.output} = vec4(val, 0, 0, 0);\n    }\n  `}(o)),n.packedInputs&&(h+=lD);return[h,u,c,a,l,s,n.userCode].join("\n")}function aD(e,t=!1){const n=e.shapeInfo.logicalShape;switch(n.length){case 0:return function(e,t){const n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`float ${r}() {return ${n};}`;const[a,s]=e.shapeInfo.texShape;if(1===a&&1===s)return`\n      float ${r}() {\n        return sampleTexture(${n}, halfCR);\n      }\n    `;const i=hD(n);if(t)return`\n    float ${r}() {\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], ${i});\n      return sampleTexture(${n}, uv);\n    }\n  `;const[o,u]=e.shapeInfo.texShape;return`\n    float ${r}() {\n      vec2 uv = uvFromFlat(${o}, ${u}, ${i});\n      return sampleTexture(${n}, uv);\n    }\n  `}(e,t);case 1:return function(e,t){const n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`\n      float ${r}(int index) {\n        ${pD(e)}\n      }\n    `;const a=e.shapeInfo.texShape,s=a[0],i=a[1];if(1===i&&1===s)return`\n      float ${r}(int index) {\n        return sampleTexture(${n}, halfCR);\n      }\n    `;const o=hD(n);if(1===i)return t?`\n      float ${r}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / float(${n}TexShape[0]));\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n      float ${r}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / ${s}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(1===s)return t?`\n      float ${r}(int index) {\n        vec2 uv = vec2((float(index + ${o}) + 0.5) / float(${n}TexShape[1]), 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n      float ${r}(int index) {\n        vec2 uv = vec2((float(index + ${o}) + 0.5) / ${i}.0, 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(t)return`\n    float ${r}(int index) {\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${o});\n      return sampleTexture(${n}, uv);\n    }\n  `;return`\n    float ${r}(int index) {\n      vec2 uv = uvFromFlat(${s}, ${i}, index + ${o});\n      return sampleTexture(${n}, uv);\n    }\n  `}(e,t);case 2:return function(e,t){const n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=e.shapeInfo.texShape;if(null!=s&&va.arraysEqual(n,s)){if(t)return`\n      float ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `;const e=s[0],n=s[1];return`\n    float ${a}(int row, int col) {\n      vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}.0, ${e}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `}const{newShape:i,keptDims:o}=va.squeezeShape(n),u=i;if(u.length<n.length){const n=mD(e,u),r=["row","col"];return`\n      ${aD(n,t)}\n      float ${a}(int row, int col) {\n        return ${a}(${gD(r,o)});\n      }\n    `}if(e.shapeInfo.isUniform)return`\n      float ${a}(int row, int col) {\n        int index = round(dot(vec2(row, col), vec2(${n[1]}, 1)));\n        ${pD(e)}\n      }\n    `;const l=s[0],c=s[1],h=hD(r);if(1===c)return t?`\n      float ${a}(int row, int col) {\n        float index = dot(vec3(row, col, ${h}), vec3(${r}Shape[1], 1, 1));\n        vec2 uv = vec2(0.5, (index + 0.5) / float(${r}TexShape[0]));\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n    float ${a}(int row, int col) {\n      float index = dot(vec3(row, col, ${h}), vec3(${n[1]}, 1, 1));\n      vec2 uv = vec2(0.5, (index + 0.5) / ${l}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `;if(1===l)return t?`\n      float ${a}(int row, int col) {\n        float index = dot(vec3(row, col, ${h}), vec3(${r}Shape[1], 1, 1));\n        vec2 uv = vec2((index + 0.5) / float(${r}TexShape[1]), 0.5);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n    float ${a}(int row, int col) {\n      float index = dot(vec3(row, col, ${h}), vec3(${n[1]}, 1, 1));\n      vec2 uv = vec2((index + 0.5) / ${c}.0, 0.5);\n      return sampleTexture(${r}, uv);\n    }\n  `;if(t)return`\n      float ${a}(int row, int col) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${r}Shape[1] + col + ${h};\n        vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);\n        return sampleTexture(${r}, uv);\n      }\n    `;return`\n  float ${a}(int row, int col) {\n    // Explicitly use integer operations as dot() only works on floats.\n    int index = row * ${n[1]} + col + ${h};\n    vec2 uv = uvFromFlat(${l}, ${c}, index);\n    return sampleTexture(${r}, uv);\n  }\n`}(e,t);case 3:return function(e,t){const n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=n[1]*n[2],i=n[2],{newShape:o,keptDims:u}=va.squeezeShape(n),l=o;if(l.length<n.length){const n=mD(e,l),r=["row","col","depth"];return`\n        ${aD(n,t)}\n        float ${a}(int row, int col, int depth) {\n          return ${a}(${gD(r,u)});\n        }\n      `}if(e.shapeInfo.isUniform)return`\n      float ${a}(int row, int col, int depth) {\n        int index = round(dot(vec3(row, col, depth),\n                          vec3(${s}, ${i}, 1)));\n        ${pD(e)}\n      }\n    `;const c=e.shapeInfo.texShape,h=c[0],p=c[1],d=e.shapeInfo.flatOffset;if(p===s&&null==d)return t?`\n      float ${a}(int row, int col, int depth) {\n        int stride1 = ${r}Shape[2];\n        float texR = float(row);\n        float texC = dot(vec2(col, depth), vec2(stride1, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n        float ${a}(int row, int col, int depth) {\n          float texR = float(row);\n          float texC = dot(vec2(col, depth), vec2(${i}, 1));\n          vec2 uv = (vec2(texC, texR) + halfCR) /\n                     vec2(${p}.0, ${h}.0);\n          return sampleTexture(${r}, uv);\n        }\n      `;if(p===i&&null==d)return t?`\n      float ${a}(int row, int col, int depth) {\n        float texR = dot(vec2(row, col), vec2(${r}Shape[1], 1));\n        float texC = float(depth);\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n    float ${a}(int row, int col, int depth) {\n      float texR = dot(vec2(row, col), vec2(${n[1]}, 1));\n      float texC = float(depth);\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${p}.0, ${h}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `;const f=hD(r);if(t)return`\n    float ${a}(int row, int col, int depth) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int stride0 = ${r}Shape[1] * ${r}Shape[2];\n      int stride1 = ${r}Shape[2];\n      int index = row * ${s} + col * ${i} + depth + ${f};\n      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);\n      return sampleTexture(${r}, uv);\n    }\n    `;return`\n      float ${a}(int row, int col, int depth) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${s} + col * ${i} + depth + ${f};\n        vec2 uv = uvFromFlat(${h}, ${p}, index);\n        return sampleTexture(${r}, uv);\n      }\n  `}(e,t);case 4:return function(e,t){const n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=n[3],i=n[2]*s,o=n[1]*i,{newShape:u,keptDims:l}=va.squeezeShape(n);if(u.length<n.length){const n=mD(e,u),r=["row","col","depth","depth2"];return`\n      ${aD(n,t)}\n      float ${a}(int row, int col, int depth, int depth2) {\n        return ${a}(${gD(r,l)});\n      }\n    `}if(e.shapeInfo.isUniform)return`\n      float ${a}(int row, int col, int depth, int depth2) {\n        int index = round(dot(vec4(row, col, depth, depth2),\n                          vec4(${o}, ${i}, ${s}, 1)));\n        ${pD(e)}\n      }\n    `;const c=e.shapeInfo.flatOffset,h=e.shapeInfo.texShape,p=h[0],d=h[1],f=`int stride2 = ${r}Shape[3];`,m=`int stride1 = ${r}Shape[2] * stride2;`,g=`int stride0 = ${r}Shape[1] * stride1;`;if(d===o&&null==c)return t?`\n      float ${a}(int row, int col, int depth, int depth2) {\n        ${f}\n        ${m}\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(stride1, stride2, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(${i}, ${s}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${d}.0, ${p}.0);\n        return sampleTexture(${r}, uv);\n      }\n    `;if(d===s&&null==c)return t?`\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${r}Shape[1] * ${r}Shape[2], ${r}Shape[2], 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${n[1]*n[2]}, ${n[2]}, 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${d}.0, ${p}.0);\n        return sampleTexture(${r}, uv);\n      }\n    `;const y=hD(r);if(t)return`\n    float ${a}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      ${f}\n      ${m}\n      ${g}\n      int index = row * stride0 + col * stride1 +\n          depth * stride2 + depth2;\n      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index + ${y});\n      return sampleTexture(${r}, uv);\n    }\n  `;return`\n    float ${a}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${o} + col * ${i} +\n          depth * ${s} + depth2;\n      vec2 uv = uvFromFlat(${p}, ${d}, index + ${y});\n      return sampleTexture(${r}, uv);\n    }\n  `}(e,t);case 5:return function(e){const t=e.shapeInfo.logicalShape,n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),a=t[4],s=t[3]*a,i=t[2]*s,o=t[1]*i,{newShape:u,keptDims:l}=va.squeezeShape(t);if(u.length<t.length){const t=mD(e,u),n=["row","col","depth","depth2","depth3"];return`\n      ${aD(t)}\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        return ${r}(${gD(n,l)});\n      }\n    `}if(e.shapeInfo.isUniform)return`\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        float index = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${o}, ${i}, ${s}, ${a})) +\n          depth3;\n        ${pD(e)}\n      }\n    `;const c=e.shapeInfo.flatOffset,h=e.shapeInfo.texShape,p=h[0],d=h[1];if(d===o&&null==c)return`\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n                         vec4(${i}, ${s}, ${a}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${d}.0, ${p}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(d===a&&null==c)return`\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        float texR = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]},\n               ${t[2]*t[3]}, ${t[3]}, 1));\n        int texC = depth3;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${d}.0, ${p}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;const f=hD(n);return`\n    float ${r}(int row, int col, int depth, int depth2, int depth3) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${o} + col * ${i} + depth * ${s} +\n          depth2 * ${a} + depth3 + ${f};\n      vec2 uv = uvFromFlat(${p}, ${d}, index);\n      return sampleTexture(${n}, uv);\n    }\n  `}(e);case 6:return function(e){const t=e.shapeInfo.logicalShape,n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),{newShape:a,keptDims:s}=va.squeezeShape(t);if(a.length<t.length){const t=mD(e,a),n=["row","col","depth","depth2","depth3","depth4"];return`\n      ${aD(t)}\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        return ${r}(${gD(n,s)});\n      }\n    `}const i=t[5],o=t[4]*i,u=t[3]*o,l=t[2]*u,c=t[1]*l;if(e.shapeInfo.isUniform)return`\n      float ${r}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n        int index = round(dot(\n          vec4(row, col, depth, depth2),\n          vec4(${c}, ${l}, ${u}, ${o})) +\n          dot(\n            vec2(depth3, depth4),\n            vec2(${i}, 1)));\n        ${pD(e)}\n      }\n    `;const h=e.shapeInfo.flatOffset,p=e.shapeInfo.texShape,d=p[0],f=p[1];if(f===c&&null==h)return`\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n          vec4(${l}, ${u}, ${o}, ${i})) +\n               float(depth4);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${f}.0, ${d}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(f===i&&null==h)return`\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        float texR = dot(vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]*t[4]},\n               ${t[2]*t[3]*t[4]},\n               ${t[3]*t[4]},\n               ${t[4]})) + float(depth3);\n        int texC = depth4;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${f}.0, ${d}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;const m=hD(n);return`\n    float ${r}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${c} + col * ${l} + depth * ${u} +\n          depth2 * ${o} + depth3 * ${i} + depth4 + ${m};\n      vec2 uv = uvFromFlat(${d}, ${f}, index);\n      return sampleTexture(${n}, uv);\n    }\n  `}(e);default:throw new Error(n.length+"-D input sampling is not yet supported")}}function sD(e,t){switch(e.shapeInfo.logicalShape.length){case 0:return function(e){const t=e.name,n="get"+t.charAt(0).toUpperCase()+t.slice(1),r=XM();return`\n    vec4 ${n}() {\n      return ${r.texture2D}(${t}, halfCR);\n    }\n  `}(e);case 1:return function(e,t){const n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),a=e.shapeInfo.texShape,s=XM();if(t)return`\n    vec4 ${r}(int index) {\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      vec2 uv = packedUVfrom1D(\n        packedTexShape[0], packedTexShape[1], index);\n      return ${s.texture2D}(${n}, uv);\n    }\n  `;const i=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)];return`\n    vec4 ${r}(int index) {\n      vec2 uv = packedUVfrom1D(\n        ${i[0]}, ${i[1]}, index);\n      return ${s.texture2D}(${n}, uv);\n    }\n  `}(e,t);case 2:return function(e,t){const n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=e.shapeInfo.texShape,i=s[0],o=s[1],u=XM();if(null!=s&&va.arraysEqual(n,s))return t?`\n      vec4 ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n\n        return ${u.texture2D}(${r}, uv);\n      }\n    `:`\n      vec4 ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${o}.0, ${i}.0);\n\n        return ${u.texture2D}(${r}, uv);\n      }\n    `;if(t)return`\n    vec4 ${a}(int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${r}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);\n      return ${u.texture2D}(${r}, uv);\n    }\n  `;const l=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],c=Math.ceil(n[1]/2);return`\n    vec4 ${a}(int row, int col) {\n      vec2 uv = packedUVfrom2D(${c}, ${l[0]}, ${l[1]}, row, col);\n      return ${u.texture2D}(${r}, uv);\n    }\n  `}(e,t);case 3:return function(e,t){const n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=e.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(1===n[0]){const r=n.slice(1),s=[1,2],i=mD(e,r),o=["b","row","col"];return`\n        ${sD(i,t)}\n        vec4 ${a}(int b, int row, int col) {\n          return ${a}(${gD(o,s)});\n        }\n      `}const o=XM();if(t)return`\n    vec4 ${a}(int b, int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${r}Shape[2]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${r}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom3D(\n        packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);\n      return ${o.texture2D}(${r}, uv);\n    }\n  `;const u=i[0],l=i[1],c=Math.ceil(n[2]/2),h=c*Math.ceil(n[1]/2);return`\n    vec4 ${a}(int b, int row, int col) {\n      vec2 uv = packedUVfrom3D(\n        ${u}, ${l}, ${h}, ${c}, b, row, col);\n      return ${o.texture2D}(${r}, uv);\n    }\n  `}(e,t);default:return function(e,t){const n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),a=XM();if(t)return`\n    vec4 ${r}(int b2, int b, int row, int col) {\n      int valuesPerRow = int(ceil(float(${n}Shape[3]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[2]) / 2.0));\n      int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);\n      texelsInBatch *= ${n}Shape[1];\n      index = b2 * texelsInBatch + index;\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      int texR = index / packedTexShape[1];\n      int texC = index - texR * packedTexShape[1];\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${a.texture2D}(${n}, uv);\n    }\n  `;const s=e.shapeInfo.logicalShape,i=s.length,o=e.shapeInfo.texShape,u=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)],l=u[0],c=u[1],h=Math.ceil(s[i-1]/2);let p=h*Math.ceil(s[i-2]/2),d="int b, int row, int col",f=`b * ${p} + (row / 2) * ${h} + (col / 2)`;for(let e=2;e<i-1;e++)d=`int b${e}, `+d,p*=s[i-e-1],f=`b${e} * ${p} + `+f;return`\n    vec4 ${r}(${d}) {\n      int index = ${f};\n      int texR = index / ${c};\n      int texC = index - texR * ${c};\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${c}, ${l});\n      return ${a.texture2D}(${n}, uv);\n    }\n  `}(e,t)}}var iD="\nvec2 uvFromFlat(int texNumR, int texNumC, int index) {\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\nvec2 packedUVfrom1D(int texNumR, int texNumC, int index) {\n  int texelIndex = index / 2;\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",oD="\nvec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,\n  int texNumC, int row, int col) {\n  int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",uD="\nvec2 packedUVfrom3D(int texNumR, int texNumC,\n    int texelsInBatch, int texelsInLogicalRow, int b,\n    int row, int col) {\n  int index = b * texelsInBatch + (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",lD="\n  float getChannel(vec4 frag, vec2 innerDims) {\n    vec2 modCoord = mod(innerDims, 2.);\n    return modCoord.x == 0. ?\n      (modCoord.y == 0. ? frag.r : frag.g) :\n      (modCoord.y == 0. ? frag.b : frag.a);\n  }\n  float getChannel(vec4 frag, int dim) {\n    float modCoord = mod(float(dim), 2.);\n    return modCoord == 0. ? frag.r : frag.g;\n  }\n";function cD(){return"\n    int getOutputCoords() {\n      return 0;\n    }\n  "}function hD(e){return"offset"+e}function pD(e){const t=e.name,n=va.sizeFromShape(e.shapeInfo.logicalShape);return n<2?`return ${t};`:`\n    for (int i = 0; i < ${n}; i++) {\n      if (i == index) {\n        return ${t}[i];\n      }\n    }\n  `}function dD(e){if(e<=1)return"int";if(2===e)return"ivec2";if(3===e)return"ivec3";if(4===e)return"ivec4";if(5===e)return"ivec5";if(6===e)return"ivec6";throw Error(`GPU for rank ${e} is not yet supported`)}function fD(e,t,n){const{newShape:r,keptDims:a}=va.squeezeShape(t),s=t.length,i=e&&3===s&&1===t[0],o=i?t.slice(1):r,u=!e&&s>1&&!va.arraysEqual(t,n)&&r.length<s||i;return{useSqueezeShape:u,uniformShape:u?o:t,keptDims:a}}function mD(e,t){const n=JSON.parse(JSON.stringify(e));return n.shapeInfo.logicalShape=t,n}function gD(e,t){return t.map(t=>e[t]).join(", ")}function yD(e,t){if(e.length!==t.length)throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);e.forEach((e,n)=>{const r=e.logicalShape,a=t[n],s=a.shape;if(!va.arraysEqual(r,s))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${r} and ${s} must match`);if(e.isUniform&&a.isUniform)return;const i=e.texShape,o=a.isUniform?null:a.texData.texShape;if(!va.arraysEqual(i,o))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${i} and ${o} must match`)})}function bD(e){return Ye().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&e<=4}var vD={};function xD(e){const t=XM();return aM(e,`${t.version}\n    precision highp float;\n    ${t.attribute} vec3 clipSpacePos;\n    ${t.attribute} vec2 uv;\n    ${t.varyingVs} vec2 resultUV;\n\n    void main() {\n      gl_Position = vec4(clipSpacePos, 1);\n      resultUV = uv;\n    }`)}function wD(e){return pM(e,new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]))}function kD(e){return dM(e,new Uint16Array([0,1,2,2,1,3]))}function SD(e,t,n,r,a,s){gM(t,n);const i=mM(e),o=e.TEXTURE_2D;return eM(e,()=>e.bindTexture(o,i)),eM(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_S,e.CLAMP_TO_EDGE)),eM(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_T,e.CLAMP_TO_EDGE)),eM(e,()=>e.texParameteri(o,e.TEXTURE_MIN_FILTER,e.NEAREST)),eM(e,()=>e.texParameteri(o,e.TEXTURE_MAG_FILTER,e.NEAREST)),1===Ye().getNumber("WEBGL_VERSION")?eM(e,()=>e.texImage2D(o,0,r,t,n,0,a,s,null)):eM(e,()=>e.texStorage2D(o,1,r,t,n)),eM(e,()=>e.bindTexture(e.TEXTURE_2D,null)),{texture:i,texShape:[n,t]}}function ID(e){return e.internalFormatFloat}function AD(e,t,n,r){const[a,s]=XF(t,n);return SD(e,a,s,ID(r),r.textureFormatFloat,e.FLOAT)}function ED(e){return e.internalFormatHalfFloat}function ND(e,t,n,r){const[a,s]=XF(t,n);return SD(e,a,s,ED(r),r.textureFormatFloat,r.textureTypeHalfFloat)}function CD(e){return e.downloadTextureFormat}function TD(e,t,n,r){const[a,s]=XF(t,n);return SD(e,a,s,CD(r),e.RGBA,e.UNSIGNED_BYTE)}function RD(e){return e.internalFormatPackedFloat}function _D(e,t,n,r){const[a,s]=ZF(t,n);return SD(e,a,s,RD(r),e.RGBA,e.FLOAT)}function FD(e){return e.internalFormatPackedHalfFloat}function MD(e,t,n,r){const[a,s]=ZF(t,n);return SD(e,a,s,FD(r),e.RGBA,r.textureTypeHalfFloat)}function DD(e,t,n){eM(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n));return bM(e,t,"clipSpacePos",n,3,20,0)&&bM(e,t,"uv",n,2,20,12)}function OD(e,t,n,r,a,s){let i,o,u;eM(e,()=>e.bindTexture(e.TEXTURE_2D,t)),a instanceof Uint8Array?(i=new Uint8Array(n*r*4),o=e.UNSIGNED_BYTE,u=e.RGBA):(i=new Float32Array(n*r*4),o=e.FLOAT,u=s.internalFormatPackedFloat),i.set(a),2===Ye().getNumber("WEBGL_VERSION")?eM(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n,r,e.RGBA,o,i)):eM(e,()=>e.texImage2D(e.TEXTURE_2D,0,u,n,r,0,e.RGBA,o,i)),eM(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function LD(e,t,n){eM(e,()=>e.bindTexture(e.TEXTURE_2D,t)),n.data instanceof Uint8Array?2===Ye().getNumber("WEBGL_VERSION")?eM(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n.width,n.height,e.RGBA,e.UNSIGNED_BYTE,n.data)):eM(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,n.width,n.height,0,e.RGBA,e.UNSIGNED_BYTE,n.data)):2===Ye().getNumber("WEBGL_VERSION")?eM(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,e.RGBA,e.UNSIGNED_BYTE,n)):eM(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,e.RGBA,e.UNSIGNED_BYTE,n)),eM(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function PD(e,t,n,r){const a=e.createBuffer();eM(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,a));const s=16*t*n;return eM(e,()=>e.bufferData(e.PIXEL_PACK_BUFFER,s,e.STREAM_READ)),eM(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,0)),eM(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,null)),a}function $D(e,t,n){const r=e,a=new Float32Array(n);return r.bindBuffer(r.PIXEL_PACK_BUFFER,t),r.getBufferSubData(r.PIXEL_PACK_BUFFER,0,a),r.bindBuffer(r.PIXEL_PACK_BUFFER,null),a}function BD(e,t,n,r){const[a,s]=XF(t,n),i=new Uint8Array(t*n*4);return eM(e,()=>e.readPixels(0,0,a,s,r.downloadTextureFormat,e.UNSIGNED_BYTE,i)),new Float32Array(i.buffer)}function zD(e,t,n,r,a,s,i,o){const u=e,l=new Float32Array(function(e,t){const[n,r]=ZF(e,t);return n*r*4}(s,i));return u.bindBuffer(u.PIXEL_PACK_BUFFER,t),u.getBufferSubData(u.PIXEL_PACK_BUFFER,0,l),u.bindBuffer(u.PIXEL_PACK_BUFFER,null),l}function UD(e,t,n){const r=new Float32Array(t*n*4);return eM(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,r)),r}w(vD,{bindVertexProgramAttributeStreams:()=>DD,createBufferFromOutputTexture:()=>PD,createFloat16MatrixTexture:()=>ND,createFloat16PackedMatrixTexture:()=>MD,createFloat32MatrixTexture:()=>AD,createIndexBuffer:()=>kD,createPackedMatrixTexture:()=>_D,createUnsignedBytesMatrixTexture:()=>TD,createVertexBuffer:()=>wD,createVertexShader:()=>xD,downloadByteEncodedFloatMatrixFromOutputTexture:()=>BD,downloadFloat32MatrixFromBuffer:()=>$D,downloadMatrixFromPackedOutputTexture:()=>UD,downloadPackedMatrixFromBuffer:()=>zD,getInternalFormatForFloat16MatrixTexture:()=>ED,getInternalFormatForFloat16PackedMatrixTexture:()=>FD,getInternalFormatForFloat32MatrixTexture:()=>ID,getInternalFormatForPackedMatrixTexture:()=>RD,getInternalFormatForUnsignedBytesMatrixTexture:()=>CD,uploadDenseMatrixToTexture:()=>OD,uploadPixelDataToTexture:()=>LD});var WD=class{constructor(e){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];const t=Ye().getNumber("WEBGL_VERSION");null!=e?(this.gl=e,KF(t,e)):this.gl=YF(t);let n="WEBGL_color_buffer_float";if(1===Ye().getNumber("WEBGL_VERSION")){const e="OES_texture_float",t="OES_texture_half_float";if(this.textureFloatExtension=rM(this.gl,e),WM(this.gl,t))this.textureHalfFloatExtension=rM(this.gl,t);else if(Ye().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(n),WM(this.gl,"EXT_color_buffer_half_float"))this.colorBufferHalfFloatExtension=rM(this.gl,"EXT_color_buffer_half_float");else if(Ye().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(n="EXT_color_buffer_float",WM(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else{if(!WM(this.gl,"EXT_color_buffer_half_float"))throw new Error("GL context does not support color renderable floats");this.colorBufferHalfFloatExtension=this.gl.getExtension("EXT_color_buffer_half_float")}this.vertexBuffer=wD(this.gl),this.indexBuffer=kD(this.gl),this.framebuffer=yM(this.gl),this.textureConfig=JF(this.gl,this.textureHalfFloatExtension)}get debug(){return Ye().getBool("DEBUG")}dispose(){if(this.disposed)return;null!=this.program&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),null!=this.outputTexture&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");const e=this.gl;eM(e,()=>e.finish()),eM(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),eM(e,()=>e.deleteFramebuffer(this.framebuffer)),eM(e,()=>e.bindBuffer(e.ARRAY_BUFFER,null)),eM(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,null)),eM(e,()=>e.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(e,t){return this.throwIfDisposed(),AD(this.gl,e,t,this.textureConfig)}createFloat16MatrixTexture(e,t){return this.throwIfDisposed(),ND(this.gl,e,t,this.textureConfig)}createUnsignedBytesMatrixTexture(e,t){return this.throwIfDisposed(),TD(this.gl,e,t,this.textureConfig)}uploadPixelDataToTexture(e,t){this.throwIfDisposed(),LD(this.gl,e,t)}uploadDenseMatrixToTexture(e,t,n,r){this.throwIfDisposed(),OD(this.gl,e,t,n,r,this.textureConfig)}createFloat16PackedMatrixTexture(e,t){return this.throwIfDisposed(),MD(this.gl,e,t,this.textureConfig)}createPackedMatrixTexture(e,t){return this.throwIfDisposed(),_D(this.gl,e,t,this.textureConfig)}deleteMatrixTexture(e){this.throwIfDisposed(),this.outputTexture===e&&(EM(this.gl,this.framebuffer),this.outputTexture=null),eM(this.gl,()=>this.gl.deleteTexture(e))}downloadByteEncodedFloatMatrixFromOutputTexture(e,t,n){return this.downloadMatrixDriver(e,()=>BD(this.gl,t,n,this.textureConfig))}downloadPackedMatrixFromBuffer(e,t,n,r,a,s){return zD(this.gl,e,0,0,0,a,s)}downloadFloat32MatrixFromBuffer(e,t){return $D(this.gl,e,t)}createBufferFromTexture(e,t,n){this.bindTextureToFrameBuffer(e);const r=PD(this.gl,t,n);return this.unbindTextureToFrameBuffer(),r}createAndWaitForFence(){const e=this.createFence(this.gl);return this.pollFence(e)}createFence(e){let t,n;if(Ye().getBool("WEBGL_FENCE_API_ENABLED")){const r=e,a=r.fenceSync(r.SYNC_GPU_COMMANDS_COMPLETE,0);e.flush(),n=()=>{const e=r.clientWaitSync(a,0,0);return e===r.ALREADY_SIGNALED||e===r.CONDITION_SATISFIED},t=a}else Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(t=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(t,Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):n=()=>!0;return{query:t,isFencePassed:n}}downloadMatrixFromPackedTexture(e,t,n){return this.downloadMatrixDriver(e,()=>UD(this.gl,t,n))}createProgram(e){this.throwIfDisposed();const t=this.gl;null==this.vertexShader&&(this.vertexShader=xD(t));const n=lM(t);return eM(t,()=>t.attachShader(n,this.vertexShader)),eM(t,()=>t.attachShader(n,e)),cM(t,n),this.debug&&hM(t,n),this.vertexAttrsAreBound||(this.setProgram(n),this.vertexAttrsAreBound=DD(t,this.program,this.vertexBuffer)),n}deleteProgram(e){this.throwIfDisposed(),e===this.program&&(this.program=null),null!=e&&eM(this.gl,()=>this.gl.deleteProgram(e))}setProgram(e){this.throwIfDisposed(),this.program=e,null!=this.program&&this.debug&&hM(this.gl,this.program),eM(this.gl,()=>this.gl.useProgram(e))}getUniformLocation(e,t,n=!0){return this.throwIfDisposed(),n?wM(this.gl,e,t):kM(this.gl,e,t)}getAttributeLocation(e,t){return this.throwIfDisposed(),eM(this.gl,()=>this.gl.getAttribLocation(e,t))}getUniformLocationNoThrow(e,t){return this.throwIfDisposed(),this.gl.getUniformLocation(e,t)}setInputMatrixTexture(e,t,n){this.throwIfDisposed(),this.throwIfNoProgram(),SM(this.gl,e,t,n)}setOutputMatrixTexture(e,t,n){this.setOutputMatrixTextureDriver(e,n,t)}setOutputPackedMatrixTexture(e,t,n){this.throwIfDisposed();const[r,a]=ZF(t,n);this.setOutputMatrixTextureDriver(e,r,a)}setOutputMatrixWriteRegion(e,t,n,r){this.setOutputMatrixWriteRegionDriver(n,e,r,t)}setOutputPackedMatrixWriteRegion(e,t,n,r){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){null!=this.program&&hM(this.gl,this.program),NM(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();const e=this.gl;this.debug&&this.debugValidate(),eM(e,()=>e.drawElements(e.TRIANGLES,6,e.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),eM(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return null==this.disjointQueryTimerExtension&&(this.disjointQueryTimerExtension=rM(this.gl,2===Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(2===Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")){const e=this.gl,t=this.getQueryTimerExtensionWebGL2(),n=e.createQuery();return e.beginQuery(t.TIME_ELAPSED_EXT,n),n}const e=this.getQueryTimerExtensionWebGL1(),t=e.createQueryEXT();return e.beginQueryEXT(e.TIME_ELAPSED_EXT,t),t}endQuery(){if(2===Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")){const e=this.gl,t=this.getQueryTimerExtensionWebGL2();return void e.endQuery(t.TIME_ELAPSED_EXT)}const e=this.getQueryTimerExtensionWebGL1();e.endQueryEXT(e.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(e){return await va.repeatedTry(()=>this.disposed||this.isQueryAvailable(e,Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(e,Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(e,t){if(0===t)return null;if(2===t){const t=this.gl;return t.getQueryParameter(e,t.QUERY_RESULT)/1e6}{const t=this.getQueryTimerExtensionWebGL1();return t.getQueryObjectEXT(e,t.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(e,t){if(0===t)return!0;if(2===t){const t=this.gl,n=this.getQueryTimerExtensionWebGL2(),r=t.getQueryParameter(e,t.QUERY_RESULT_AVAILABLE);return null==this.disjoint&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),r&&!this.disjoint}{const t=this.getQueryTimerExtensionWebGL1(),n=t.getQueryObjectEXT(e,t.QUERY_RESULT_AVAILABLE_EXT);return null==this.disjoint&&(this.disjoint=this.gl.getParameter(t.GPU_DISJOINT_EXT)),n&&!this.disjoint}}pollFence(e){return new Promise(t=>{this.addItemToPoll(()=>e.isFencePassed(),()=>t())})}pollItems(){const e=function(e){let t=0;for(;t<e.length;++t){if(!e[t]())break}return t-1}(this.itemsToPoll.map(e=>e.isDoneFn));for(let t=0;t<=e;++t){const{resolveFn:e}=this.itemsToPoll[t];e()}this.itemsToPoll=this.itemsToPoll.slice(e+1)}addItemToPoll(e,t){this.itemsToPoll.push({isDoneFn:e,resolveFn:t}),this.itemsToPoll.length>1||va.repeatedTry(()=>(this.pollItems(),0===this.itemsToPoll.length))}bindTextureToFrameBuffer(e){this.throwIfDisposed(),AM(this.gl,e,this.framebuffer),this.debug&&NM(this.gl)}unbindTextureToFrameBuffer(){null!=this.outputTexture?(AM(this.gl,this.outputTexture,this.framebuffer),this.debug&&NM(this.gl)):EM(this.gl,this.framebuffer)}downloadMatrixDriver(e,t){this.bindTextureToFrameBuffer(e);const n=t();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(e,t,n){this.throwIfDisposed();const r=this.gl;AM(r,e,this.framebuffer),this.debug&&NM(r),this.outputTexture=e,eM(r,()=>r.viewport(0,0,t,n)),eM(r,()=>r.scissor(0,0,t,n))}setOutputMatrixWriteRegionDriver(e,t,n,r){this.throwIfDisposed(),eM(this.gl,()=>this.gl.scissor(e,t,n,r))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(null==this.program)throw new Error("No GPU program is currently set.")}};var{addImpl:VD,bincountImpl:jD,bincountReduceImpl:GD,ceilImpl:HD,concatImpl:qD,equalImpl:KD,expImpl:YD,expm1Impl:XD,floorImpl:QD,gatherNdImpl:ZD,gatherV2Impl:JD,greaterImpl:eO,greaterEqualImpl:tO,lessImpl:nO,lessEqualImpl:rO,linSpaceImpl:aO,logImpl:sO,maxImpl:iO,maximumImpl:oO,minimumImpl:uO,multiplyImpl:lO,negImpl:cO,notEqualImpl:hO,prodImpl:pO,rangeImpl:dO,rsqrtImpl:fO,sigmoidImpl:mO,simpleAbsImpl:gO,sliceImpl:yO,sparseFillEmptyRowsImpl:bO,sparseReshapeImpl:vO,sparseSegmentReductionImpl:xO,sqrtImpl:wO,stridedSliceImpl:kO,stringNGramsImpl:SO,stringSplitImpl:IO,stringToHashBucketFastImpl:AO,subImpl:EO,tileImpl:NO,topKImpl:CO,transposeImpl:TO,uniqueImpl:RO}=UE;function _O(e,t){return["x","y","z","w","u","v"].slice(0,t).map(t=>`${e}.${t}`)}function FO(e,t){return 1===t?[e]:_O(e,t)}var MO=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length);let n="";for(let e=0;e<4;e++){let t="thisRC = rc;";e%2==1&&(t+="thisRC.z += 1;"),e>1&&(t+="thisRC.y += 1;"),n+=`\n        ${t}\n        ${e>0?"if(thisRC.y < rows && thisRC.z < cols){":""}\n          int flatIndex = getFlatIndex(thisRC);\n\n          ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);\n          vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));\n\n          result[${e}] =\n            getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);\n        ${e>0?"}":""}\n      `}var r,a;this.userCode=`\n      ${r=t,a=this.enableShapeUniforms,`\n    ivec3 inputCoordsFromReshapedOutCoords(int index) {\n      ${a?JM(["r","c","d"],"inputShape"):QM(["r","c","d"],r)}\n      return ivec3(r, c, d);\n    }\n  `}\n      ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":eD(e)}\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0.);\n\n        ivec3 thisRC;\n        int rows = ${this.enableShapeUniforms?"outShape[1]":e[1]};\n        int cols = ${this.enableShapeUniforms?"outShape[2]":e[2]};\n\n        ${n}\n\n        setOutput(result);\n      }\n    `}};function DO(e,t,n,r,a){const s=function(e,t){switch(e){case jF.PACKED_2X2_FLOAT32:return RD(t);case jF.PACKED_2X2_FLOAT16:return FD(t);case jF.UNPACKED_FLOAT32:return ID(t);case jF.UNPACKED_FLOAT16:return ED(t);case jF.PACKED_4X1_UNSIGNED_BYTE:return CD(t);default:throw new Error("Unknown physical texture type "+e)}}(t,r);let i;if(a){const[t,n]=ZF(e[0],e[1]);i=t*n}else{const[t,n]=XF(e[0],e[1]);i=t*n}return i*function(e,t){const n=e;if(t===n.R32F)return 4;if(t===n.R16F)return 2;if(t===n.RGBA32F)return 16;if(t===e.RGBA)return 16;if(t===n.RGBA16F)return 8;if(t===n.RGBA8)return 4;throw new Error("Unknown internal format "+t)}(n,s)}function OO(e,t){if(e===WF.UPLOAD)return jF.PACKED_2X2_FLOAT32;if(e===WF.RENDER||null==e)return function(e){return Ye().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?e?jF.PACKED_2X2_FLOAT32:jF.UNPACKED_FLOAT32:e?jF.PACKED_2X2_FLOAT16:jF.UNPACKED_FLOAT16}(t);if(e===WF.DOWNLOAD||e===WF.PIXELS)return jF.PACKED_4X1_UNSIGNED_BYTE;throw new Error("Unknown logical texture type "+e)}function LO(e,t,n){return`${e[0]}_${e[1]}_${t}_${n}`}var PO=class{constructor(e,t){this.variableNames=["A"],this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length),this.userCode=`\n      float unaryOperation(float x) {\n        ${t}\n      }\n\n      void main() {\n        float x = getAAtOutCoords();\n        float y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `}},$O="return abs(x);",BO="return x;",zO=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length),this.userCode=`\n      vec4 unaryOperation(vec4 x) {\n        ${t}\n      }\n\n      void main() {\n        vec4 x = getAAtOutCoords();\n        vec4 y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `}},UO=jf.whereImpl,WO={};var VO=Ye().getNumber("CPU_HANDOFF_SIZE_THRESHOLD");var jO=class extends Z{constructor(e){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!Ye().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let t;if(null!=e){if(e instanceof WD)t=e;else{const n=YF(Ye().getNumber("WEBGL_VERSION"),e);t=new WD(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{const e=YF(Ye().getNumber("WEBGL_VERSION"));t=new WD(e),this.binaryCache=((n=Ye().getNumber("WEBGL_VERSION"))in WO||(WO[n]={}),WO[n]),this.gpgpuCreatedLocally=!0}var n;this.gpgpu=t,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new class{constructor(e){this.gpgpu=e,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(e,t,n){const r=OO(t,n),a=LO(e,r,n);a in this.freeTextures||(this.freeTextures[a]=[]),a in this.usedTextures||(this.usedTextures[a]=[]);const s=DO(e,r,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[a].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=s,this.log();const e=this.freeTextures[a].shift();return this.usedTextures[a].push(e),e}let i;return r===jF.PACKED_2X2_FLOAT32?i=this.gpgpu.createPackedMatrixTexture(e[0],e[1]):r===jF.PACKED_2X2_FLOAT16?i=this.gpgpu.createFloat16PackedMatrixTexture(e[0],e[1]):r===jF.UNPACKED_FLOAT32?i=this.gpgpu.createFloat32MatrixTexture(e[0],e[1]):r===jF.UNPACKED_FLOAT16?i=this.gpgpu.createFloat16MatrixTexture(e[0],e[1]):r===jF.PACKED_4X1_UNSIGNED_BYTE&&(i=this.gpgpu.createUnsignedBytesMatrixTexture(e[0],e[1])),this.usedTextures[a].push(i),this.numUsedTextures++,this._numBytesAllocated+=s,this.log(),i}releaseTexture(e,t,n,r){if(null==this.freeTextures)return;const a=OO(n,r),s=LO(t,a,r);s in this.freeTextures||(this.freeTextures[s]=[]);const i=DO(t,a,this.gpgpu.gl,this.gpgpu.textureConfig,r),o=Ye().get("WEBGL_DELETE_TEXTURE_THRESHOLD");-1!==o&&this._numBytesAllocated>o?(this.gpgpu.deleteMatrixTexture(e.texture),this._numBytesAllocated-=i):(this.freeTextures[s].push(e),this.numFreeTextures++,this._numBytesFree+=i),this.numUsedTextures--;const u=this.usedTextures[s],l=u.indexOf(e);if(l<0)throw new Error("Cannot release a texture that was never provided by this texture manager");u.splice(l,1),this.log()}log(){if(!this.logEnabled)return;const e=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${e})`);const t=this._numBytesFree/this._numBytesAllocated;console.log("Bytes allocated: "+this._numBytesAllocated),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(null!=this.freeTextures){for(const e in this.freeTextures)this.freeTextures[e].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});for(const e in this.usedTextures)this.usedTextures[e].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}}(this.gpgpu),this.numMBBeforeWarning=null==Ye().global.screen?1024:Ye().global.screen.height*Ye().global.screen.width*window.devicePixelRatio*600/1024/1024,this.texData=new Q(this,eu())}nextDataId(){return jO.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}write(e,t,n){if((Ye().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||Ye().getBool("DEBUG"))&&this.checkNumericalProblems(e),"complex64"===n&&null!=e)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");const r={id:this.nextDataId()};return this.texData.set(r,{shape:t,dtype:n,values:e,usage:WF.UPLOAD,refCount:1}),r}refCount(e){if(this.texData.has(e)){return this.texData.get(e).refCount}return 0}incRef(e){this.texData.get(e).refCount++}decRef(e){if(this.texData.has(e)){this.texData.get(e).refCount--}}move(e,t,n,r,a){if(Ye().getBool("DEBUG")&&this.checkNumericalProblems(t),"complex64"===r)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(e,{shape:n,dtype:r,values:t,usage:WF.UPLOAD,refCount:a})}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}readSync(e){const t=this.texData.get(e),{values:n,dtype:r,complexTensorInfos:a,slice:s,shape:i,isPacked:o}=t;if(null!=s){let t;t=o?new zO(i,BO):new PO(i,BO);const n=this.runWebGLProgram(t,[{dataId:e,shape:i,dtype:r}],r),a=this.readSync(n.dataId);return this.disposeIntermediateTensorInfo(n),a}if(null!=n)return this.convertAndCacheOnCPU(e);if("string"===r)return n;const u=null!=this.activeTimers;let l,c;if(u&&(l=va.now()),"complex64"===r){const e=this.readSync(a.real.dataId),t=this.readSync(a.imag.dataId);c=jd.mergeRealAndImagArrays(e,t)}else c=this.getValuesFromTexture(e);return u&&(this.downloadWaitMs+=va.now()-l),this.convertAndCacheOnCPU(e,c)}async read(e){if(this.pendingRead.has(e)){const t=this.pendingRead.get(e);return new Promise(e=>t.push(e))}const t=this.texData.get(e),{values:n,shape:r,slice:a,dtype:s,complexTensorInfos:i,isPacked:o}=t;if(null!=a){let t;t=o?new zO(r,BO):new PO(r,BO);const n=this.runWebGLProgram(t,[{dataId:e,shape:r,dtype:s}],s),a=this.read(n.dataId);return this.disposeIntermediateTensorInfo(n),a}if(null!=n)return this.convertAndCacheOnCPU(e);if(Ye().getBool("DEBUG")&&!Ye().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&2===Ye().getNumber("WEBGL_VERSION"))throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let u,l,c=null;if("complex64"!==s&&Ye().get("WEBGL_BUFFER_SUPPORTED")){u=this.decode(e);const t=this.texData.get(u.dataId);c=this.gpgpu.createBufferFromTexture(t.texture.texture,...QF(r))}if(this.pendingRead.set(e,[]),"complex64"!==s&&await this.gpgpu.createAndWaitForFence(),"complex64"===s){const e=await Promise.all([this.read(i.real.dataId),this.read(i.imag.dataId)]),t=e[0],n=e[1];l=jd.mergeRealAndImagArrays(t,n)}else if(null==c)l=this.getValuesFromTexture(e);else{const e=va.sizeFromShape(r);l=this.gpgpu.downloadFloat32MatrixFromBuffer(c,e)}if(null!=u&&this.disposeIntermediateTensorInfo(u),null!=c){const e=this.gpgpu.gl;eM(e,()=>e.deleteBuffer(c))}const h=this.convertAndCacheOnCPU(e,l),p=this.pendingRead.get(e);return this.pendingRead.delete(e),p.forEach(e=>e(h)),this.pendingDisposal.has(e)&&(this.pendingDisposal.delete(e),this.disposeData(e)&&eu().removeDataId(e,this),this.pendingDeletes--),h}readToGPU(e,t={}){const n=this.texData.get(e),{values:r,shape:a,slice:s,dtype:i,isPacked:o,texture:u}=n;if("complex64"===i)throw new Error("Does not support reading texture for complex64 dtype.");if(null!=s){let n;n=o?new zO(a,BO):new PO(a,BO);const r=this.runWebGLProgram(n,[{dataId:e,shape:a,dtype:i}],i),s=this.readToGPU(r,t);return this.disposeIntermediateTensorInfo(r),s}if(null==u)throw null!=r?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");const l=this.decode(e,t.customTexShape),c=eu().makeTensorFromDataId(l.dataId,l.shape,l.dtype),h=this.texData.get(l.dataId);return Object.assign({tensorRef:c},h.texture)}bufferSync(e){const t=this.readSync(e.dataId);let n=t;if("string"===e.dtype)try{n=t.map(e=>va.decodeString(e))}catch(e){throw new Error("Failed to decode encoded string bytes into utf-8")}return Ii(e.shape,e.dtype,n)}checkNumericalProblems(e){if(null!=e)for(let t=0;t<e.length;t++){const n=e[t];if(!tM(n)){if(Ye().getBool("WEBGL_RENDER_FLOAT32_CAPABLE"))throw Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`);throw Error(`The value ${n} cannot be represented on this device.`)}}}getValuesFromTexture(e){const{shape:t,dtype:n,isPacked:r}=this.texData.get(e),a=va.sizeFromShape(t);if(Ye().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){const n=this.decode(e),r=this.texData.get(n.dataId),s=this.gpgpu.downloadMatrixFromPackedTexture(r.texture.texture,...QF(t)).subarray(0,a);return this.disposeIntermediateTensorInfo(n),s}const s=Ye().getBool("WEBGL_PACK")&&!0===r,i=s?MM(t):t,o=s?new class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=WF.DOWNLOAD;const t=XM();this.outputShape=e,this.userCode=`\n      ${tD}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));\n        ${t.output} = encode_float(x);\n      }\n    `}}(i):new class{constructor(e){this.variableNames=["A"],this.outTexUsage=WF.DOWNLOAD;const t=XM();this.outputShape=e,this.userCode=`\n      ${tD}\n\n      void main() {\n        float x = getAAtOutCoords();\n        ${t.output} = encode_float(x);\n      }\n    `}}(i),u=this.runWebGLProgram(o,[{shape:i,dtype:n,dataId:e}],"float32"),l=this.texData.get(u.dataId),c=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(l.texture.texture,l.texShape[0],l.texShape[1]).subarray(0,a);return this.disposeIntermediateTensorInfo(u),c}timerAvailable(){return Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(e){const t=this.activeTimers,n=[];let r=!1;null==this.programTimersStack?(this.programTimersStack=n,r=!0):this.activeTimers.push(n),this.activeTimers=n,e();const a=va.flatten(this.activeTimers.map(e=>e.query)).filter(e=>null!=e),s=va.flatten(this.activeTimers.map(e=>e.name)).filter(e=>null!=e);this.activeTimers=t,r&&(this.programTimersStack=null);const i={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){const e=await Promise.all(a);i.kernelMs=va.sum(e),i.getExtraProfileInfo=()=>e.map((e,t)=>({name:s[t],ms:e})).map(e=>`${e.name}: ${e.ms}`).join(", ")}else i.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,i})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:va.now(),endMs:null}}endTimer(e){return Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),e):(e.endMs=va.now(),e)}async getQueryTime(e){if(Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(e);const t=e;return t.endMs-t.startMs}disposeData(e,t=!1){if(this.pendingDisposal.has(e))return!1;if(!this.texData.has(e))return!0;if(t?this.texData.get(e).refCount=0:this.texData.get(e).refCount--,!t&&this.texData.get(e).refCount>0)return!1;if(this.pendingRead.has(e))return this.pendingDisposal.add(e),this.pendingDeletes++,!1;this.releaseGPUData(e);const{complexTensorInfos:n}=this.texData.get(e);return null!=n&&(this.disposeData(n.real.dataId,t),this.disposeData(n.imag.dataId,t)),this.texData.delete(e),!0}releaseGPUData(e){const{texture:t,dtype:n,texShape:r,usage:a,isPacked:s,slice:i}=this.texData.get(e),o=i&&i.origDataId||e,u=this.dataRefCount.get(o);u>1?this.dataRefCount.set(o,u-1):(this.dataRefCount.delete(o),null!=t&&(this.numBytesInGPU-=this.computeBytes(r,n),this.textureManager.releaseTexture(t,r,a,s)));const l=this.texData.get(e);l.texture=null,l.texShape=null,l.isPacked=!1,l.slice=null}getTexture(e){return this.uploadToGPU(e),this.texData.get(e).texture.texture}getDataInfo(e){return this.texData.get(e)}shouldExecuteOnCPU(e,t=VO){return Ye().getBool("WEBGL_CPU_FORWARD")&&e.every(e=>null==this.texData.get(e.dataId).texture&&va.sizeFromShape(e.shape)<t)}getGPGPUContext(){return this.gpgpu}where(e){jd.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");const t=e.dataSync();return UO(e.shape,t)}packedUnaryOp(e,t,n){const r=new zO(e.shape,t),a=this.compileAndRun(r,[e],n);return eu().makeTensorFromDataId(a.dataId,a.shape,a.dtype)}abs(e){if(this.shouldExecuteOnCPU([e])&&"complex64"!==e.dtype){const t=gO(this.texData.get(e.dataId).values);return this.makeOutput(e.shape,e.dtype,t)}if(Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(e,$O,e.dtype);const t=new PO(e.shape,$O),n=this.compileAndRun(t,[e]);return eu().makeTensorFromDataId(n.dataId,n.shape,n.dtype)}makeTensorInfo(e,t,n){let r;if("string"===t&&null!=n&&n.length>0&&va.isString(n[0])){const a=n.map(e=>va.encodeString(e));r=this.write(a,e,t)}else r=this.write(n,e,t);return this.texData.get(r).usage=null,{dataId:r,shape:e,dtype:t}}makeOutput(e,t,n){const{dataId:r}=this.makeTensorInfo(e,t,n);return eu().makeTensorFromDataId(r,e,t,this)}unpackTensor(e){const t=new class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length);const t=e.length,n=FO("rc",t),r=dD(t),a=function(e,t){if(1===e)return"rc";let n="";for(let r=0;r<e;r++)n+=t[r],r<e-1&&(n+=",");return n}(t,n),s=n.slice(-2),i=t<=1?"rc":`vec2(${s.join(",")})`;this.userCode=`\n      void main() {\n        ${r} rc = getOutputCoords();\n        vec4 packedInput = getA(${a});\n\n        setOutput(getChannel(packedInput, ${i}));\n      }\n    `}}(e.shape);return this.runWebGLProgram(t,[e],e.dtype)}packTensor(e){const t=new class{constructor(e){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.enableShapeUniforms=bD(this.outputShape.length),0===this.rank)this.userCode="\n        void main() {\n          setOutput(vec4(getA(), 0., 0., 0.));\n        }\n      ";else{const e=FO("rc",this.rank),t=dD(this.rank),n=this.getOutOfBoundsCondition(e),r=this.getSetup(e),a=this.getOutput(e);this.userCode=`\n        void main() {\n          ${t} rc = getOutputCoords();\n\n          if(${n}) {\n            setOutput(vec4(0));\n          } else {\n            ${r}\n\n            setOutput(vec4(${a}));\n          }\n        }\n      `}}getSourceCoordsArr(e){const t=[];for(let n=0;n<=1;n++)for(let r=0;r<=1;r++){let a=`${0===n?"r":"rp1"}, ${0===r?"c":"cp1"}`;for(let t=2;t<this.rank;t++)a=e[e.length-1-t]+","+a;t.push(a)}return t}getOutOfBoundsCondition(e){if(1===this.rank)return"rc > "+(this.enableShapeUniforms?"outShape":this.outputShape[0]);let t="";for(let n=this.rank-2;n<this.rank;n++)t+=`${e[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(t+="||");return t}getSetup(e){if(1===this.rank)return"";const t=e.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],r=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`\n      int r = ${t[0]};\n      int c = ${t[1]};\n      int rp1 = r + 1;\n      int cp1 = c + 1;\n\n      bool cEdge = cp1 >= ${n};\n      bool rEdge = rp1 >= ${r};\n    `}getOutput(e){const t=this.getSourceCoordsArr(e);if(1===this.rank){return`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`}return`getA(${t[0]}),\n            cEdge ? 0. : getA(${t[1]}),\n            rEdge ? 0. : getA(${t[2]}),\n            rEdge || cEdge ? 0. : getA(${t[3]})`}}(e.shape);return this.runWebGLProgram(t,[e],e.dtype,null,!0)}packedReshape(e,t){const n=[_M(e.shape),...FM(e.shape)],r={dtype:e.dtype,shape:n,dataId:e.dataId},a=[_M(t),...FM(t)],s=new MO(a,n),i=[n],o=this.runWebGLProgram(s,[r],e.dtype,i,!0);return{dataId:o.dataId,shape:t,dtype:o.dtype}}decode(e,t){const n=this.texData.get(e),{isPacked:r,shape:a,dtype:s}=n;if(null!=t){const e=va.sizeFromShape(a),n=t[0]*t[1]*4;va.assert(e<=n,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}const i=MM(a);let o;o=r?new class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=zF.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];const t=XM();this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length),this.userCode=`\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?ZM(["r","c","d"],e):QM(["r","c","d"],e)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));\n        }\n\n        ${t.output} = result;\n      }\n    `}}(i):new class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=zF.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];const t=XM();this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length),this.userCode=`\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?ZM(["r","c","d"],e):QM(["r","c","d"],e)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getA(rc.x, rc.y, rc.z);\n        }\n\n        ${t.output} = result;\n      }\n    `}}(i);const u=[null!=t?t:QF(i)];return{dtype:s,shape:a,dataId:this.runWebGLProgram(o,[{shape:i,dtype:s,dataId:e}],s,u,!0,t).dataId}}runWebGLProgram(e,t,n,r,a=!1,s){const i=this.makeTensorInfo(e.outputShape,n),o=this.texData.get(i.dataId);if(e.packedOutput&&(o.isPacked=!0),e.outPackingScheme===zF.DENSE){const t=null!=s?s:QF(e.outputShape);o.texShape=t.map(e=>2*e)}if(null!=e.outTexUsage&&(o.usage=e.outTexUsage),0===va.sizeFromShape(i.shape))return o.values=va.getTypedArrayFromDType(i.dtype,0),i;const u=[],l=t.map(t=>{if("complex64"===t.dtype)throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let n=this.texData.get(t.dataId);if(null==n.texture){if(!e.packedInputs&&va.sizeFromShape(t.shape)<=Ye().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:t.shape,texData:null,isUniform:!0,uniformValues:n.values};e.packedInputs&&(n.isPacked=!0,n.shape=t.shape)}if(this.uploadToGPU(t.dataId),!!n.isPacked!=!!e.packedInputs)t=n.isPacked?this.unpackTensor(t):this.packTensor(t),u.push(t),n=this.texData.get(t.dataId);else if(n.isPacked&&!LM(n.shape,t.shape)){const e=t,r=t.shape;t.shape=n.shape,t=this.packedReshape(t,r),u.push(t),n=this.texData.get(t.dataId),e.shape=r}return{shape:t.shape,texData:n,isUniform:!1}});this.uploadToGPU(i.dataId);const c={shape:i.shape,texData:o,isUniform:!1},h=function(e,t,n){let r="";t.concat(n).forEach(t=>{const a=null!=t.texData&&null!=t.texData.slice&&t.texData.slice.flatOffset>0;if(e.enableShapeUniforms&&!t.isUniform){const s=t.texData.texShape,{useSqueezeShape:i,uniformShape:o,keptDims:u}=fD(e.packedInputs,t.shape,s);let l="",c="",h="";if(1===o.length&&e.packedInputs){const e=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];l=`${e[0]>1}_${e[1]>1}`}else if(2!==o.length||e.packedInputs){if(o.length>2&&!e.packedInputs){const e=va.computeStrides(o);h=`${e[0]===s[1]}_${e[e.length-1]===s[1]}`}}else c=`${o[0]>1}_${o[1]>1}`;const p=t.shape.length,d=2===o.length&&va.arraysEqual(t.shape,s),f=1===va.sizeFromShape(t.shape),m=jd.getBroadcastDims(t.shape,n.shape),g=!e.packedInputs&&p===n.shape.length&&va.arraysEqual(s,n.texData.texShape),y=e.packedInputs||o.length>2?"":`${s[0]>1}_${s[1]>1}`;r+=`${p}_${g}_${i?u:""}_${o.length}_${f}_${m}_${d}_${l}_${c}_${h}_${y}_${a}`}else{const e=t.isUniform?"uniform":t.texData.texShape;r+=`${t.shape}_${e}_${a}`}});const a=e.userCode;let s=e.constructor.name;return s+="_"+r+"_"+a+Ye().getNumber("WEBGL_VERSION"),s}(e,l,c),p=this.getAndSaveBinary(h,()=>function(e,t,n,r){const a=n.map((e,n)=>{const r={logicalShape:e.shape,texShape:e.isUniform?null:e.texData.texShape,isUniform:e.isUniform,isPacked:!e.isUniform&&e.texData.isPacked,flatOffset:null};return null!=e.texData&&null!=e.texData.slice&&e.texData.slice.flatOffset>0&&(r.flatOffset=e.texData.slice.flatOffset),{name:t.variableNames[n],shapeInfo:r}}),s=a.map(e=>e.shapeInfo),i={logicalShape:r.shape,texShape:r.texData.texShape,isUniform:!1,isPacked:r.texData.isPacked,flatOffset:null},o=rD(a,i,t),u=sM(e.gl,o),l=e.createProgram(u);let c=null;const h=e.getUniformLocation(l,"NAN",!1);1===Ye().getNumber("WEBGL_VERSION")&&(c=e.getUniformLocation(l,"INFINITY",!1));const p={},d={},f={};for(let n=0;n<t.variableNames.length;n++){const r=t.variableNames[n];p[r]=e.getUniformLocation(l,r,!1),p["offset"+r]=e.getUniformLocation(l,"offset"+r,!1),t.enableShapeUniforms&&(d[r+"Shape"]=e.getUniformLocation(l,r+"Shape",!1),f[r+"TexShape"]=e.getUniformLocation(l,r+"TexShape",!1))}let m,g,y;t.enableShapeUniforms&&(m=e.getUniformLocation(l,"outShape",!1),y=e.getUniformLocation(l,"outShapeStrides",!1),g=e.getUniformLocation(l,"outTexShape",!1));const b=[];return t.customUniforms&&t.customUniforms.forEach((t,n)=>{b[n]=e.getUniformLocation(l,t.name,!1)}),{program:t,fragmentShader:u,source:o,webGLProgram:l,uniformLocations:p,customUniformLocations:b,inShapeInfos:s,outShapeInfo:i,infLoc:c,nanLoc:h,inShapesLocations:d,inTexShapesLocations:f,outShapeLocation:m,outShapeStridesLocation:y,outTexShapeLocation:g}}(this.gpgpu,e,l,c)),d=null!=this.activeTimers;let f;d&&(f=this.startTimer()),function(e,t,n,r,a){t.program.enableShapeUniforms||(yD(t.inShapeInfos,n),yD([t.outShapeInfo],[r]));const s=r.texData.texture,i=r.texData.texShape;r.texData.isPacked?e.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):e.setOutputMatrixTexture(s.texture,i[0],i[1]),e.setProgram(t.webGLProgram),1===Ye().getNumber("WEBGL_VERSION")&&null!==t.infLoc&&e.gl.uniform1f(t.infLoc,1/0),null!==t.nanLoc&&e.gl.uniform1f(t.nanLoc,NaN),n.forEach((n,r)=>{const a=t.program.variableNames[r],s=t.uniformLocations[a],i=t.uniformLocations["offset"+a],o=t.inShapesLocations[a+"Shape"],u=t.inTexShapesLocations[a+"TexShape"];if(o){const{uniformShape:r}=fD(t.program.packedInputs,n.shape,n.texData.texShape);switch(r.length){case 1:e.gl.uniform1iv(o,new Int32Array(r));break;case 2:e.gl.uniform2iv(o,new Int32Array(r));break;case 3:e.gl.uniform3iv(o,new Int32Array(r));break;case 4:e.gl.uniform4iv(o,new Int32Array(r))}}if(u&&e.gl.uniform2i(u,n.texData.texShape[0],n.texData.texShape[1]),null!=s)if(n.isUniform)if(va.sizeFromShape(n.shape)<2)e.gl.uniform1f(s,n.uniformValues[0]);else{let t=n.uniformValues;t instanceof Float32Array||(t=new Float32Array(t)),e.gl.uniform1fv(s,t)}else null!=n.texData.slice&&null!=i&&e.gl.uniform1i(i,n.texData.slice.flatOffset),e.setInputMatrixTexture(n.texData.texture.texture,s,r)});const o=t.outShapeLocation;if(o)switch(r.shape.length){case 1:e.gl.uniform1iv(o,new Int32Array(r.shape));break;case 2:e.gl.uniform2iv(o,new Int32Array(r.shape));break;case 3:e.gl.uniform3iv(o,new Int32Array(r.shape));break;case 4:e.gl.uniform4iv(o,new Int32Array(r.shape))}if(t.outShapeStridesLocation){const n=va.computeStrides(r.shape);switch(r.shape.length){case 2:e.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(n));break;case 3:e.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(n));break;case 4:e.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(n))}}t.outTexShapeLocation&&e.gl.uniform2i(t.outTexShapeLocation,r.texData.texShape[0],r.texData.texShape[1]),t.program.customUniforms&&a&&t.program.customUniforms.forEach((n,r)=>{const s=t.customUniformLocations[r],i=a[r];if("float"===n.type)e.gl.uniform1fv(s,i);else if("vec2"===n.type)e.gl.uniform2fv(s,i);else if("vec3"===n.type)e.gl.uniform3fv(s,i);else if("vec4"===n.type)e.gl.uniform4fv(s,i);else if("int"===n.type)e.gl.uniform1iv(s,i);else if("ivec2"===n.type)e.gl.uniform2iv(s,i);else if("ivec3"===n.type)e.gl.uniform3iv(s,i);else{if("ivec4"!==n.type)throw Error(`uniform type ${n.type} is not supported yet.`);e.gl.uniform4iv(s,i)}}),e.executeProgram()}(this.gpgpu,p,l,c,r),u.forEach(e=>this.disposeIntermediateTensorInfo(e)),d&&(f=this.endTimer(f),this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(f)}));const m=Ye().get("WEBGL_FLUSH_THRESHOLD");if(m>0){const e=va.now();e-this.lastGlFlushTime>m&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=e)}if(!Ye().getBool("WEBGL_LAZILY_UNPACK")&&o.isPacked&&!1===a){const e=this.unpackTensor(i);return this.disposeIntermediateTensorInfo(i),e}return i}compileAndRun(e,t,n,r,a=!1){n=n||t[0].dtype;return this.runWebGLProgram(e,t,n,r,a)}getAndSaveBinary(e,t){return e in this.binaryCache||(this.binaryCache[e]=t()),this.binaryCache[e]}getTextureManager(){return this.textureManager}dispose(){if(!this.disposed){if(!Ye().getBool("IS_TEST")){Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]})}this.textureManager.dispose(),null!=this.canvas&&"undefined"!=typeof HTMLCanvasElement&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0}}floatPrecision(){return null==this.floatPrecisionValue&&(this.floatPrecisionValue=ru(()=>{if(!Ye().get("WEBGL_RENDER_FLOAT32_ENABLED")){const e=Ye().getBool("DEBUG");Ye().set("DEBUG",!1);const t=this.abs(Ah(1e-8)).dataSync()[0];if(Ye().set("DEBUG",e),t>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return 32===this.floatPrecision()?1e-7:1e-4}uploadToGPU(e){const t=this.texData.get(e),{shape:n,dtype:r,values:a,texture:s,usage:i,isPacked:o}=t;if(null!=s)return;const u=null!=this.activeTimers;let l;u&&(l=va.now());let c=t.texShape;if(null==c&&(c=DM(n,o),t.texShape=c),null!=a){const e=MM(n);let s,i=c[1],h=c[0];const p=a instanceof Uint8Array||a instanceof Uint8ClampedArray;!o&&p||([i,h]=ZF(c[0],c[1])),s=o?new class{constructor(e,t=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];const n=XM();this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length);let r="",a="result";t&&(a="floor(result * 255. + 0.5)");for(let t=0;t<=1;t++)for(let a=0;a<=1;a++){const s=2*t+a;r+=`\n          localCoords = coords;\n          if(localCoords[2] + ${a} < ${this.enableShapeUniforms?"outShape[2]":""+e[2]}) {\n          localCoords[2] += ${a};\n          if (localCoords[1] + ${t} < ${this.enableShapeUniforms?"outShape[1]":""+e[1]}) {\n            localCoords[1] += ${t};\n\n            flatIndex = getFlatIndex(localCoords);\n            offset = imod(flatIndex, 4);\n\n            flatIndex = idiv(flatIndex, 4, 1.);\n\n            int r = flatIndex / texShape[1];\n            int c = imod(flatIndex, texShape[1]);\n            vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n            values = ${n.texture2D}(A, uv);\n\n            if (offset == 0) {\n              result[${s}] = values[0];\n            } else if (offset == 1) {\n              result[${s}] = values[1];\n            } else if (offset == 2) {\n              result[${s}] = values[2];\n            } else {\n              result[${s}] = values[3];\n            }\n          }\n        }\n        `}this.userCode=`\n        ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":eD(e)}\n\n        void main() {\n          ivec3 coords = getOutputCoords();\n\n          vec4 result = vec4(0.);\n          int flatIndex, r, c, offset;\n          ivec3 localCoords;\n          vec2 uv;\n          vec4 values;\n\n          ${r}\n\n          ${n.output} = ${a};\n        }\n    `}}(e,p):new class{constructor(e,t=!1){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];const n=XM();this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length);let r="result";t&&(r="floor(result * 255. + 0.5)"),this.userCode=`\n      ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":eD(e)}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n\n        int flatIndex = getFlatIndex(coords);\n        int offset = imod(flatIndex, 4);\n\n        flatIndex = idiv(flatIndex, 4, 1.);\n\n        int r = flatIndex / texShape[1];\n        int c = imod(flatIndex, texShape[1]);\n        vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n        vec4 values = ${n.texture2D}(A, uv);\n\n        float result;\n\n        if(offset == 0) {\n          result = values[0];\n        } else if(offset == 1) {\n          result = values[1];\n        } else if(offset == 2) {\n          result = values[2];\n        } else {\n          result = values[3];\n        }\n\n        ${n.output} = vec4(${r}, 0., 0., 0.);\n      }\n    `}}(e,p);const d=p?[h,i]:c,f=this.makeTensorInfo(d,r),m=this.texData.get(f.dataId);m.usage=p?WF.PIXELS:WF.UPLOAD,m.texShape=d,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(f.dataId),i,h,a);const g=[[h,i]],y=!0,b=this.runWebGLProgram(s,[f],r,g,y),v=this.texData.get(b.dataId);t.texture=v.texture,t.texShape=v.texShape,t.isPacked=v.isPacked,t.usage=v.usage,this.disposeIntermediateTensorInfo(f),this.texData.delete(b.dataId),t.values=null,u&&(this.uploadWaitMs+=va.now()-l)}else{const e=this.acquireTexture(c,i,r,o);t.texture=e}}convertAndCacheOnCPU(e,t){const n=this.texData.get(e),{dtype:r}=n;return this.releaseGPUData(e),null!=t&&(n.values=function(e,t){if("float32"===t||"complex64"===t)return e;if("int32"===t||"bool"===t){const n="int32"===t?new Int32Array(e.length):new Uint8Array(e.length);for(let t=0;t<n.length;++t)n[t]=Math.round(e[t]);return n}throw new Error("Unknown dtype "+t)}(t,r)),n.values}acquireTexture(e,t,n,r){if(this.numBytesInGPU+=this.computeBytes(e,n),!this.warnedAboutMemory&&this.numBytesInGPU>1024*this.numMBBeforeWarning*1024){const e=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${e} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(e,t,r)}computeBytes(e,t){return e[0]*e[1]*va.bytesPerElement(t)}};jO.nextDataId=0;var GO="3.13.0";function HO(){Ye().set("WEBGL_FORCE_F16_TEXTURES",!0)}Ss.isBrowser()&&du("webgl",()=>new jO,2);var qO={forceHalfFloat:HO},KO=class{constructor(e,t,n){this.variableNames=["A","B"],this.outputShape=jd.assertAndGetBroadcastShape(t,n),this.enableShapeUniforms=bD(this.outputShape.length),this.userCode=`\n      float binaryOperation(float a, float b) {\n        ${e}\n      }\n\n      void main() {\n        float a = getAAtOutCoords();\n        float b = getBAtOutCoords();\n        setOutput(binaryOperation(a, b));\n      }\n    `}},YO=class{constructor(e,t,n,r=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=jd.assertAndGetBroadcastShape(t,n);const a=this.outputShape.length;this.enableShapeUniforms=bD(a);let s="";if(r)if(0===a||1===va.sizeFromShape(this.outputShape))s="\n          result.y = 0.;\n          result.z = 0.;\n          result.w = 0.;\n        ";else{if(s=`\n          ${dD(a)} coords = getOutputCoords();\n        `,1===a)this.enableShapeUniforms?s+="\n            result.y = (coords + 1) >= outShape ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          ":s+=`\n            result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `;else{const e=FO("coords",a);this.enableShapeUniforms?s+=`\n            bool nextRowOutOfBounds =\n              (${e[a-2]} + 1) >= outShape[${a} - 2];\n            bool nextColOutOfBounds =\n              (${e[a-1]} + 1) >= outShape[${a} - 1];\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `:s+=`\n            bool nextRowOutOfBounds =\n              (${e[a-2]} + 1) >= ${this.outputShape[a-2]};\n            bool nextColOutOfBounds =\n              (${e[a-1]} + 1) >= ${this.outputShape[a-1]};\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `}}this.userCode=`\n      vec4 binaryOperation(vec4 a, vec4 b) {\n        ${e}\n      }\n\n      void main() {\n        vec4 a = getAAtOutCoords();\n        vec4 b = getBAtOutCoords();\n\n        vec4 result = binaryOperation(a, b);\n        ${s}\n\n        setOutput(result);\n      }\n    `}};function XO(e){const{inputs:t,backend:n}=e,{x:r}=t;return n.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var QO={kernelName:fn,backendName:"webgl",kernelFunc:XO};function ZO(e){const{inputs:t,backend:n}=e,{real:r,imag:a}=t,s=n.makeTensorInfo(r.shape,"complex64"),i=n.texData.get(s.dataId),o=XO({inputs:{x:r},backend:n}),u=XO({inputs:{x:a},backend:n});return i.complexTensorInfos={real:o,imag:u},s}var JO={kernelName:Et,backendName:"webgl",kernelFunc:ZO},eL="return (a < 0.) ? b * a : a;",tL="\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n";var nL={kernelName:xn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{alpha:s}=r,i=n.makeTensorInfo([],"float32",va.createScalarValue(s,"float32")),o=Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new YO(tL,a.shape,i.shape):new KO(eL,a.shape,i.shape),u=n.runWebGLProgram(o,[a,i],"float32");return n.disposeIntermediateTensorInfo(i),u}},rL="return (a < 0.) ? b * a : a;",aL="\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n";var sL={kernelName:rr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r,alpha:a}=t,s=Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new YO(aL,r.shape,a.shape):new KO(rL,r.shape,a.shape);return n.runWebGLProgram(s,[r,a],"float32")}};function iL({opSnippet:e,packedOpSnippet:t,cpuKernelImpl:n,dtype:r}){return({inputs:a,backend:s})=>{const{x:i}=a,o=s,u=r||i.dtype;if(o.shouldExecuteOnCPU([i])&&null!=n){const e=o.texData.get(i.dataId),t=n(e.values,u);return o.makeTensorInfo(i.shape,u,t)}let l;return l=Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&null!=t?new zO(i.shape,t):new PO(i.shape,e),o.runWebGLProgram(l,[i],u)}}function oL({opSnippet:e,packedOpSnippet:t,checkOutOfBounds:n=!1,supportsComplex:r=!1,cpuKernelImpl:a,dtype:s}){return({inputs:i,backend:o})=>{const{a:u,b:l}=i,c=o;if(r&&"complex64"===u.dtype){const t=c.texData.get(u.dataId),n=c.texData.get(l.dataId),[r,a]=[[t.complexTensorInfos.real,n.complexTensorInfos.real],[t.complexTensorInfos.imag,n.complexTensorInfos.imag]].map(t=>{const[n,r]=t,a={dataId:n.dataId,dtype:n.dtype,shape:u.shape},s={dataId:r.dataId,dtype:r.dtype,shape:l.shape},i=new KO(e,u.shape,l.shape);return c.runWebGLProgram(i,[a,s],cs(n.dtype,r.dtype))}),s=ZO({inputs:{real:r,imag:a},backend:c});return c.disposeIntermediateTensorInfo(r),c.disposeIntermediateTensorInfo(a),s}const h=s||cs(u.dtype,l.dtype);if(("string"===u.dtype||"string"===l.dtype||c.shouldExecuteOnCPU([u,l]))&&null!=a){const e=c.texData.get(u.dataId).values,t=c.texData.get(l.dataId).values,n="string"===u.dtype?jd.fromUint8ToStringArray(e):e,r="string"===u.dtype?jd.fromUint8ToStringArray(t):t,[s,i]=a(u.shape,l.shape,n,r,h),o=c.makeTensorInfo(i,h);return c.texData.get(o.dataId).values=s,o}let p;return p=Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&null!=t?new YO(t,u.shape,l.shape,n):new KO(e,u.shape,l.shape),c.runWebGLProgram(p,[u,l],h)}}function uL(e,t=!1){if("linear"===e)return"return x;";if("relu"===e)return t?"\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n":"if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : x;\n";if("elu"===e)return t?"\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n":"return (x >= 0.0) ? x : (exp(x) - 1.0);";if("relu6"===e)return t?"\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n":"if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n";if("prelu"===e)return t?aL:rL;if("leakyrelu"===e)return t?tL:eL;if("sigmoid"===e)return"return 1.0 / (1.0 + exp(-1.0 * x));";throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)}var lL=class{constructor(e,t,n,r=!1,a=!1,s=!1,i=null,o=!1,u=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=bD(this.outputShape.length);const l=r?e[1]:e[2],c=Math.ceil(l/2),h=r?"i * 2, rc.y":"rc.y, i * 2",p=a?"rc.z, i * 2":"i * 2, rc.z",d=r?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],f=a?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"];let m="",g="";i&&(m=o?`vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${i}\n        }`:u?`vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${i}\n        }`:`vec4 activation(vec4 x) {\n          ${i}\n        }`,g="result = activation(result);");const y=s?"result += getBiasAtOutCoords();":"";s&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),u&&this.variableNames.push("leakyreluAlpha");let b="rc.x",v="rc.x";e[0]<t[0]?b=`int(min(float(rc.x), ${e[0]-1}.))`:t[0]<e[0]&&(v=`int(min(float(rc.x), ${t[0]-1}.))`),this.userCode=`\n      ${m}\n      // Don't use uniform for sharedDimensionPacked for performance.\n      const float sharedDimension = ${c}.0;\n\n      vec4 dot2x2ARowBCol(ivec3 rc) {\n        vec4 result = vec4(0);\n        for (int i = 0; i < ${c}; i++) {\n          int batchA = ${b};\n          int batchB = ${v};\n          vec4 a = getMatrixA(batchA, ${h});\n          vec4 b = getMatrixB(batchB, ${p});\n\n          // These swizzled products need to be separately added.\n          // See: https://github.com/tensorflow/tfjs/issues/1735\n          result += (${d[0]} * ${f[0]});\n          result += (${d[1]} * ${f[1]});\n        }\n        return result;\n      }\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n        vec4 result = dot2x2ARowBCol(rc);\n\n        ${y}\n\n        ${g}\n\n        setOutput(result);\n      }\n    `}},cL="return areal * breal - aimag * bimag;",hL="return areal * bimag + aimag * breal;",pL=class{constructor(e,t,n){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=jd.assertAndGetBroadcastShape(t,n),this.userCode=`\n      float binaryOpComplex(\n          float areal, float aimag, float breal, float bimag) {\n        ${e}\n      }\n\n      void main() {\n        float areal = getARealAtOutCoords();\n        float aimag = getAImagAtOutCoords();\n        float breal = getBRealAtOutCoords();\n        float bimag = getBImagAtOutCoords();\n        setOutput(binaryOpComplex(areal, aimag, breal, bimag));\n      }\n    `}},dL="return a * b;";function fL(e){const{inputs:t,backend:n}=e,{a:r,b:a}=t,s=jd.upcastType(r.dtype,a.dtype);if("complex64"===r.dtype){const e=n.texData.get(r.dataId),t=n.texData.get(a.dataId),s=new pL(cL,r.shape,a.shape),i=new pL(hL,r.shape,a.shape),o=[{dataId:e.complexTensorInfos.real.dataId,dtype:e.complexTensorInfos.real.dtype,shape:r.shape},{dataId:e.complexTensorInfos.imag.dataId,dtype:e.complexTensorInfos.imag.dtype,shape:r.shape},{dataId:t.complexTensorInfos.real.dataId,dtype:t.complexTensorInfos.real.dtype,shape:a.shape},{dataId:t.complexTensorInfos.imag.dataId,dtype:t.complexTensorInfos.imag.dtype,shape:a.shape}],u=n.runWebGLProgram(s,o,"float32"),l=n.runWebGLProgram(i,o,"float32"),c=ZO({inputs:{real:u,imag:l},backend:n});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(l),c}if(n.shouldExecuteOnCPU([r,a])){const e=n.texData.get(r.dataId),t=n.texData.get(a.dataId),[i,o]=lO(r.shape,a.shape,e.values,t.values,s),u=n.makeTensorInfo(o,s);return n.texData.get(u.dataId).values=i,u}let i;return i=Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new YO(dL,r.shape,a.shape):new KO(dL,r.shape,a.shape),n.runWebGLProgram(i,[r,a],s)}var mL={kernelName:Gn,backendName:"webgl",kernelFunc:fL};function gL(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{shape:s}=r,i=n,o=va.sizeFromShape(a.shape),u=va.inferFromImplicitShape(s,o),l=va.sizeFromShape(u);va.assert(o===l,()=>`The new shape (${u}) has ${l} elements and the old shape (${a.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`);const c=i.texData.get(a.dataId);return!c.isPacked||LM(a.shape,u)||null!==c.texture&&LM(c.shape,u)?(i.incRef(a.dataId),{dataId:a.dataId,shape:u,dtype:a.dtype}):function(e,t,n){const r=[_M(e.shape),...FM(e.shape)],a={dtype:e.dtype,shape:r,dataId:e.dataId},s=[_M(t),...FM(t)],i=new MO(s,r),o=[r],u=n.runWebGLProgram(i,[a],e.dtype,o,!0);return{dataId:u.dataId,shape:t,dtype:u.dtype}}(a,u,i)}var yL={kernelName:lr,backendName:"webgl",kernelFunc:gL},bL=class{constructor(e,t){this.variableNames=["x"];const{windowSize:n,batchSize:r,inSize:a,outSize:s}=e;this.outputShape=[r,s];const i=4*Math.floor(n/4),o=n%4;let u="sumValue += dot(values, ones);";if(null!=t){const e=1/t;u=`sumValue += dot(values * ${va.isInt(e)?e.toPrecision(2):e}, ones);`}let l="";a%n>0&&(l=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return 0.0;\n        }\n      `),this.userCode=`\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${l}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${i}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${u}\n        }\n\n        int inIdx = inOffset + ${i};\n        if (${1===o}) {\n          vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);\n\n          ${u}\n        } else if (${2===o}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1), 0.0, 0.0);\n\n          ${u}\n        } else if (${3===o}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2), 0.0);\n\n          ${u}\n        }\n        setOutput(sumValue);\n      }\n    `}},vL=class{constructor(e,t){this.variableNames=["x"];const{windowSize:n,batchSize:r,inSize:a,outSize:s}=e;this.outputShape=[r,s];let i="0.0",o="";"prod"===t?i="1.0":"min"===t?(i="1.0 / 1e-20",o="min"):"max"===t&&(i="-1.0 / 1e-20",o="max");let u=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;"sum"===t?u="sumValue":"prod"===t?u="prodValue":"all"===t?u="allValue":"any"===t&&(u="anyValue");const l=4*Math.floor(n/4),c=n%4;let h=`\n      if (${"sum"===t}) {\n        sumValue += dot(values, ones);\n      } else if (${"prod"===t}) {\n        vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);\n        prodValue *= tmp[0] * tmp[1];\n      } else {\n        minMaxValue = ${o}(values, minMaxValue);\n        if (${"min"===t} || ${"max"===t}) {\n          minMaxValue = ${o}(values, minMaxValue);\n          bvec4 isNaN = isnan(values);\n          if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {\n            minMaxValue = vec4(NAN);\n          }\n        }\n      }\n    `,p="vec4";"all"===t?(i="1.0",h="\n        bool reducedAllValue = all(values);\n        float floatedReducedAllValue = float(reducedAllValue);\n        allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);\n      ",p="bvec4"):"any"===t&&(i="0.0",h="\n        bool reducedAnyValue = any(values);\n        float floatedReducedAnyValue = float(reducedAnyValue);\n        anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);\n      ",p="bvec4");let d="";a%n>0&&(d=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return initializationValue;\n        }\n      `),this.userCode=`\n      const float initializationValue = ${i};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${d}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        vec4 minMaxValue = vec4(${i});\n        float prodValue = 1.0;\n        float sumValue = 0.0;\n        float allValue = 1.0;\n        float anyValue = 0.0;\n\n        for (int i = 0; i < ${l}; i += 4) {\n          int inIdx = inOffset + i;\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${h}\n        }\n\n        int inIdx = inOffset + ${l};\n        if (${1===c}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          ${h}\n        } else if (${2===c}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          ${h}\n        } else if (${3===c}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          ${h}\n        }\n        setOutput(${u});\n      }\n    `}};function xL(e,t,n,r){const a=function(e){const t=[];for(;0===t.length||1!==t[t.length-1].outSize;){const n=t.length?t[t.length-1].outSize:e[1],r=jd.computeOptimalWindowSize(n);t.push({inSize:n,windowSize:r,outSize:Math.ceil(n/r)})}return t}(e.shape);let s=e;for(let i=0;i<a.length;i++){const{inSize:o,windowSize:u,outSize:l}=a[i];let c,h;c="mean"===n?0===i?new bL({windowSize:u,inSize:o,batchSize:e.shape[0],outSize:l},o):new bL({windowSize:u,inSize:o,batchSize:e.shape[0],outSize:l}):new vL({windowSize:u,inSize:o,batchSize:e.shape[0],outSize:l},n),h=s,s=r.runWebGLProgram(c,[s],t),h.dataId!==e.dataId&&r.disposeIntermediateTensorInfo(h)}return s}function wL(e,t,n){const r=Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;const n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);const r=dD(this.rank),a=_O("rc",this.rank),s=new Array(this.rank);for(let e=0;e<t.length;e++)s[t[e]]=a[e];const i=`vec2(${s.slice(-2).join()})`,o=`++${a[this.rank-1]} < ${n[this.rank-1]}`,u=`getChannel(getA(${s.join()}), ${i})`;this.userCode=`\n    void main() {\n      ${r} rc = getOutputCoords();\n      vec4 result = vec4(0.);\n      result[0] = ${u};\n      if(${o}) {\n        result[1] = ${u};\n      }\n      --${a[this.rank-1]};\n      if(++${a[this.rank-2]} < ${n[this.rank-2]}) {\n        result[2] = ${u};\n        if(${o}) {\n          result[3] = ${u};\n        }\n      }\n      setOutput(result);\n    }\n    `}}(e.shape,t):new class{constructor(e,t){this.variableNames=["A"];const n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];this.outputShape=n,this.rank=n.length;const r=dD(this.rank),a=function(e){const t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);const n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],r=new Array(t);for(let t=0;t<e.length;t++)r[e[t]]=n[t];return r.join()}(t);this.userCode=`\n    void main() {\n      ${r} resRC = getOutputCoords();\n      setOutput(getA(${a}));\n    }\n    `}}(e.shape,t);return n.runWebGLProgram(r,[e],e.dtype)}function kL(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;return function(e,t,n,r){const a=t,s=e.shape.length,i=va.parseAxisParam(a,e.shape);let o=i;const u=jd.getAxesPermutation(o,s),l=null!=u;let c=e;l&&(c=wL(e,u,r),o=jd.getInnerMostAxes(o.length,s)),jd.assertAxesAreInnerMostDims("sum",o,s);const[h,p]=jd.computeOutAndReduceShapes(c.shape,o);let d=h;n&&(d=jd.expandShapeToKeepDim(h,i));const f=va.sizeFromShape(p),m=gL({inputs:{x:c},attrs:{shape:[va.sizeFromShape(e.shape)/f,f]},backend:r}),g=xL(m,hs(e.dtype),"sum",r),y=gL({inputs:{x:g},attrs:{shape:d},backend:r});return r.disposeIntermediateTensorInfo(m),r.disposeIntermediateTensorInfo(g),l&&r.disposeIntermediateTensorInfo(c),y}(a,s,i,n)}var SL={kernelName:Cr,backendName:"webgl",kernelFunc:kL};function IL(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{perm:s}=r,i=n,o=a.shape.length,u=new Array(o);for(let e=0;e<u.length;e++)u[e]=a.shape[s[e]];let l;if(i.shouldExecuteOnCPU([a])){const e=i.texData.get(a.dataId).values,t=TO(e,a.shape,a.dtype,s,u);l=i.makeTensorInfo(u,a.dtype);i.texData.get(l.dataId).values=t}else l=wL(a,s,i);return l}var AL={kernelName:Yr,backendName:"webgl",kernelFunc:IL};function EL({a:e,b:t,transposeA:n,transposeB:r,backend:a,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:o=0,activation:u=null}){const l=e.shape.length,c=t.shape.length,h=n?e.shape[l-2]:e.shape[l-1],p=r?t.shape[c-1]:t.shape[c-2],d=n?e.shape[l-1]:e.shape[l-2],f=r?t.shape[c-2]:t.shape[c-1],m=e.shape.slice(0,-2),g=t.shape.slice(0,-2),y=va.sizeFromShape(m),b=va.sizeFromShape(g),v=Yi.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([d,f]);va.assert(h===p,()=>`Error in matMul: inner shapes (${h}) and (${p}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${n} and transposeB=${r} must match.`);const x=n?[y,h,d]:[y,d,h],w=r?[b,f,p]:[b,p,f],k=gL({inputs:{x:e},backend:a,attrs:{shape:x}}),S=gL({inputs:{x:t},backend:a,attrs:{shape:w}}),I=[k,S],A=Math.max(y,b),E=n?k.shape[1]:k.shape[2],N=null!=s,C=null!=i,T="leakyrelu"===u,R=null!=u?uL(u,!0):null;let _;if((1===d||1===f)&&E>1e3&&!1===(N||C||T||null!=R)){let e=k,t=S;n&&(e=IL({inputs:{x:k},backend:a,attrs:{perm:[0,2,1]}}),I.push(e)),r&&(t=IL({inputs:{x:S},backend:a,attrs:{perm:[0,2,1]}}),I.push(t));const s=1===f;let i=e;1!==f&&(i=gL({inputs:{x:e},backend:a,attrs:{shape:[A,E,1]}}),I.push(i));const o=1===f?2:1;let u=t;s&&(u=gL({inputs:{x:t},backend:a,attrs:{shape:[A,1,E]}}),I.push(u));const l=fL({inputs:{a:i,b:u},backend:a});_=kL({inputs:{x:l},backend:a,attrs:{axis:o,keepDims:!0}}),I.push(l)}else{const u=cs(e.dtype,t.dtype),l=new lL(x,w,[A,d,f],n,r,N,R,C,T),c=[k,S];if(null!=s&&c.push(s),C&&c.push(i),T){const e=a.makeTensorInfo([],"float32",va.createScalarValue(o,"float32"));c.push(e),I.push(e)}_=a.runWebGLProgram(l,c,u)}const F=gL({inputs:{x:_},backend:a,attrs:{shape:v}});I.push(_);for(const e of I)a.disposeIntermediateTensorInfo(e);return F}var NL={kernelName:ra,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{a:a,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:h}=r;return EL({a:a,b:s,transposeA:u,transposeB:l,backend:n,bias:i,preluActivationWeights:o,leakyreluAlpha:h,activation:c})}},CL="return abs(x);";var TL={kernelName:et,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r}=t;if(n.shouldExecuteOnCPU([r])&&"complex64"!==r.dtype){const e=n.texData.get(r.dataId),t=gO(e.values);return n.makeTensorInfo(r.shape,r.dtype,t)}let a;return a=Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS")?new zO(r.shape,CL):new PO(r.shape,CL),n.runWebGLProgram(a,[r],r.dtype)}},RL=iL({opSnippet:"if (isnan(x)) return x;\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return acos(x);\n"}),_L={kernelName:tt,backendName:"webgl",kernelFunc:RL},FL=iL({opSnippet:"if (isnan(x)) return x;\n  if (x < 1.0) return NAN;\nreturn log(x + sqrt(x * x - 1.0));"}),ML={kernelName:nt,backendName:"webgl",kernelFunc:FL},DL="return a + b;",OL=oL({opSnippet:DL,packedOpSnippet:DL,supportsComplex:!0,cpuKernelImpl:VD}),LL={kernelName:rt,backendName:"webgl",kernelFunc:OL};var PL={kernelName:at,backendName:"webgl",kernelFunc:function e(t){const{inputs:n,backend:r}=t,a=n;if(1===a.length)return XO({inputs:{x:a[0]},backend:r});if(a.length>Ye().get("WEBGL_MAX_TEXTURES_IN_SHADER")){const t=Math.floor(a.length/2),n=e({inputs:a.slice(0,t),backend:r}),s=e({inputs:a.slice(t),backend:r});return e({inputs:[n,s],backend:r})}const s=a.map(e=>e.dtype).reduce((e,t)=>cs(e,t)),i=a.map(e=>e.shape),o=Ye().getBool("WEBGL_PACK")?new class{constructor(e,t){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.variableNames=t.map((e,t)=>"T"+t);const n=[];this.variableNames.forEach(e=>{n.push(`vec4 v${e} = get${e}AtOutCoords();`)});const r=this.variableNames.map(e=>"v"+e).join(" + ");this.userCode=`\n      void main() {\n        ${n.join("\n        ")}\n\n        vec4 result = ${r};\n        setOutput(result);\n      }\n    `}}(a[0].shape,i):new class{constructor(e,t){this.outputShape=[],this.outputShape=e,this.variableNames=t.map((e,t)=>"T"+t);const n=[];this.variableNames.forEach(e=>{n.push(`float v${e} = get${e}AtOutCoords();`)});const r=this.variableNames.map(e=>"v"+e).join(" + ");this.userCode=`\n      void main() {\n        ${n.join("\n        ")}\n\n        float result = ${r};\n        setOutput(result);\n      }\n    `}}(a[0].shape,i);return r.runWebGLProgram(o,a,s)}};var $L={kernelName:st,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,u=va.parseAxisParam(s,a.shape);let l=u;const c=jd.getAxesPermutation(l,o);let h=a;null!=c&&(h=IL({inputs:{x:a},backend:n,attrs:{perm:c}}),l=jd.getInnerMostAxes(l.length,o)),jd.assertAxesAreInnerMostDims("all",l,o);const[p,d]=jd.computeOutAndReduceShapes(h.shape,l),f=gL({inputs:{x:h},backend:n,attrs:{shape:[-1,va.sizeFromShape(d)]}}),m=xL(f,f.dtype,"all",n);let g;if(i){g=gL({inputs:{x:m},backend:n,attrs:{shape:jd.expandShapeToKeepDim(p,u)}})}else g=gL({inputs:{x:m},backend:n,attrs:{shape:p}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),null!=c&&n.disposeIntermediateTensorInfo(h),g}};var BL={kernelName:it,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,u=va.parseAxisParam(s,a.shape);let l=u;const c=jd.getAxesPermutation(l,o);let h=a;null!=c&&(h=IL({inputs:{x:a},backend:n,attrs:{perm:c}}),l=jd.getInnerMostAxes(l.length,o)),jd.assertAxesAreInnerMostDims("any",l,o);const[p,d]=jd.computeOutAndReduceShapes(h.shape,l),f=gL({inputs:{x:h},backend:n,attrs:{shape:[-1,va.sizeFromShape(d)]}}),m=xL(f,f.dtype,"any",n);let g;if(i){g=gL({inputs:{x:m},backend:n,attrs:{shape:jd.expandShapeToKeepDim(p,u)}})}else g=gL({inputs:{x:m},backend:n,attrs:{shape:p}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),null!=c&&n.disposeIntermediateTensorInfo(h),g}};function zL(e,t,n,r=null){const a=null!=r?r.shape:t.shape,s=a[a.length-1],i=jd.computeOptimalWindowSize(s),o=new class{constructor(e,t,n,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,va.assert(e.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);const a=e[e.length-1],s=Math.ceil(a/t);this.outputShape=e.slice(0,-1),s>1&&this.outputShape.push(s),r||this.variableNames.push("bestIndicesA");const i=this.outputShape,o=i.length,u=dD(o),l=FO("coords",o);let c,h;if(1===s){h=o+1;const e=dD(h);c=`\n        ${e} sourceLocR = ${e}(${l.join()}, 0);\n        ++${l[o-1]};\n        ${e} sourceLocG = ${e}(${l.join()}, 0);\n        ++${l[o-2]};\n        ${e} sourceLocA = ${e}(${l.join()}, 0);\n        --${l[o-1]};\n        ${e} sourceLocB = ${e}(${l.join()}, 0);\n        --${l[o-2]};`}else h=o,c=`\n        ${u} sourceLocR = coords;\n        ++${l[o-1]};\n        ${u} sourceLocG = coords;\n        ++${l[o-2]};\n        ${u} sourceLocA = coords;\n        --${l[o-1]};\n        ${u} sourceLocB = coords;\n        --${l[o-2]};`;const p=["x","y","z","w","u","v"].slice(0,h),d="."+p[h-1],f=p.map(e=>"int "+e),m=FO("sourceLocR",h-1).concat("inIdx.r"),g=FO("sourceLocG",h-1).concat("inIdx.g"),y=FO("sourceLocB",h-1).concat("inIdx.b"),b=FO("sourceLocA",h-1).concat("inIdx.a"),v="max"===n?"greaterThan":"lessThan",x=r?"":`\n          inIdx = round(vec4(getBestIndicesAChannel(${m.join()}),\n                             getBestIndicesAChannel(${g.join()}),\n                             getBestIndicesAChannel(${y.join()}),\n                             getBestIndicesAChannel(${b.join()})));`,w=`vec4(\n            getAChannel(${m.join()}),\n            hasNextCol ? getAChannel(${g.join()}) : 0.,\n            hasNextRow ? getAChannel(${y.join()}) : 0.,\n            hasNextRow && hasNextCol ? getAChannel(${b.join()}) : 0.)`,k=r?"":`\n      float getBestIndicesAChannel(${f.join()}) {\n        return getChannel(getBestIndicesA(${p.join()}),\n                                          vec2(${p.slice(-2).join()}));\n      }`;this.userCode=`\n      float getAChannel(${f.join()}) {\n        return getChannel(getA(${p.join()}),\n                               vec2(${p.slice(-2).join()}));\n      }\n      ${k}\n      void main() {\n        ${u} coords = getOutputCoords();\n        bool hasNextCol = ${l[o-1]} < ${i[o-1]-1};\n        bool hasNextRow = ${l[o-2]} < ${i[o-2]-1};\n        ${c}\n        ivec4 srcIdx = ivec4(sourceLocR${d}, sourceLocG${d},\n          sourceLocB${d}, sourceLocA${d}) * ${t};\n        ivec4 inIdx = srcIdx;\n        vec4 bestIndex = vec4(inIdx);\n        vec4 bestValue = ${w};\n\n        for (int i = 0; i < ${t}; i++) {\n          inIdx = srcIdx;\n          ${x}\n          vec4 candidate = ${w};\n          bvec4 nan = isnan(candidate);\n          bvec4 replace = bvec4(\n            vec4(${v}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));\n\n          bestValue = vec4(replace.x  ? candidate.x : bestValue.x,\n                           replace.y  ? candidate.y : bestValue.y,\n                           replace.z  ? candidate.z : bestValue.z,\n                           replace.w  ? candidate.w : bestValue.w);\n          bestIndex = mix(bestIndex, vec4(inIdx), vec4(replace));\n          srcIdx++;\n        }\n        setOutput(bestIndex);\n      }\n    `}}(a,i,n,null==r),u=null==r?[t]:[t,r],l=e.runWebGLProgram(o,u,"int32");if(l.shape.length===t.shape.length){const r=zL(e,t,n,l);return e.disposeIntermediateTensorInfo(l),r}return l}function UL(e,t,n,r){const a=[n];if(jd.assertAxesAreInnerMostDims("arg"+r.charAt(0).toUpperCase()+r.slice(1),a,t.shape.length),!Ye().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){const n=[],s=e.texData.get(t.dataId);let i=t;null!==s&&s.isPacked&&(i=e.unpackTensor(t),n.push(i));const[o,u]=jd.computeOutAndReduceShapes(i.shape,a),l=va.sizeFromShape(u),c=gL({inputs:{x:i},backend:e,attrs:{shape:[-1,l]}});n.push(c);const h=function e(t,n,r,a=null){let s=n.shape[0],i=n.shape[1];null!=a&&(s=a.shape[0],i=a.shape[1]);const o=jd.computeOptimalWindowSize(i),u={windowSize:o,inSize:i,batchSize:s,outSize:Math.ceil(i/o)},l=new class{constructor(e,t,n){this.variableNames=["A"];const{windowSize:r,batchSize:a,outSize:s}=e;n||this.variableNames.push("bestIndicesA"),this.outputShape=[a,s];const i="max"===t?">":"<",o=n?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${r};\n\n        int bestIndex = inOffset;\n        float bestValue = getA(batch, bestIndex);\n\n        for (int i = 0; i < ${r}; i++) {\n          int inIdx = ${o};\n          float candidate = getA(batch, inIdx);\n          if (candidate ${i} bestValue) {\n            bestValue = candidate;\n            bestIndex = inIdx;\n          }\n        }\n        setOutput(float(bestIndex));\n      }\n    `}}(u,r,null==a),c=[n];null!=a&&c.push(a);const h=t.runWebGLProgram(l,c,"int32");if(1===h.shape[1])return h;const p=e(t,n,r,h);return t.disposeIntermediateTensorInfo(h),p}(e,c,r);n.push(h);const p=gL({inputs:{x:h},backend:e,attrs:{shape:o}});return n.forEach(t=>e.disposeIntermediateTensorInfo(t)),p}return zL(e,t,r)}var WL={kernelName:ot,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r;let i=va.parseAxisParam(s,a.shape);const o=jd.getAxesPermutation(i,a.shape.length);let u=a;const l=[];null!=o&&(u=IL({inputs:{x:a},backend:n,attrs:{perm:o}}),l.push(u),i=jd.getInnerMostAxes(i.length,u.shape.length)),jd.assertAxesAreInnerMostDims("argMax",[i[0]],u.shape.length);const c=UL(n,u,i[0],"max");return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),c}};var VL={kernelName:ut,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r;let i=va.parseAxisParam(s,a.shape);const o=jd.getAxesPermutation(i,a.shape.length);let u=a;const l=[];null!=o&&(u=IL({inputs:{x:a},backend:n,attrs:{perm:o}}),l.push(u),i=jd.getInnerMostAxes(i.length,u.shape.length)),jd.assertAxesAreInnerMostDims("argMin",[i[0]],u.shape.length);const c=UL(n,u,i[0],"min");return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),c}},jL=iL({opSnippet:"if (isnan(x)) return x;\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return asin(x);\n"}),GL={kernelName:lt,backendName:"webgl",kernelFunc:jL},HL=iL({opSnippet:"if (isnan(x)) return x;return log(x + sqrt(x * x + 1.0));"}),qL={kernelName:ct,backendName:"webgl",kernelFunc:HL},KL=iL({opSnippet:"if (isnan(x)) return x;\n  return atan(x);\n"}),YL={kernelName:ht,backendName:"webgl",kernelFunc:KL},XL=oL({opSnippet:"\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n\n  return atan(a, b);\n",packedOpSnippet:"\n  vec4 result = atan(a, b);\n  vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n"}),QL={kernelName:dt,backendName:"webgl",kernelFunc:XL},ZL=iL({opSnippet:"if (isnan(x)) return x;\n  if ((x < -1.0) || (x > 1.0)) return NAN;\nreturn (log(1.0 + x) - log(1.0 - x)) / 2.0;"}),JL={kernelName:pt,backendName:"webgl",kernelFunc:ZL},eP=class{constructor(e,t,n,r=!1,a=!1){if(this.variableNames=["x"],"avg"===t&&n)throw new Error("Cannot compute positions for average pool.");const s=e.filterWidth,i=e.strideHeight,o=e.strideWidth,u=e.dilationHeight,l=e.dilationWidth,c=e.effectiveFilterHeight,h=e.effectiveFilterWidth,p=e.padInfo.top,d=e.padInfo.left;this.outputShape=e.outShape;const f="avg"===t,m=`((batch  * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,g=`(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`;let y="0.0";if(f||(y="-1.0 / 1e-20"),n){const t=">=";return void(this.userCode=`\n        const ivec2 strides = ivec2(${i}, ${o});\n        const ivec2 pads = ivec2(${p}, ${d});\n\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int batch = coords[0];\n          int d = coords[3];\n\n          ivec2 xRCCorner = coords.yz * strides - pads;\n          int xRCorner = xRCCorner.x;\n          int xCCorner = xRCCorner.y;\n\n          // max/min x(?, ?, d) to get y(yR, yC, d).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n          float avgValue = 0.0;\n\n          for (int wR = 0; wR < ${c};\n              wR += ${u}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${h};\n                wC += ${l}) {\n              int xC = xCCorner + wC;\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              float value = getX(batch, xR, xC, d);\n\n              // If a min / max value has already been found, use it. If not,\n              // use the current value.\n              float currMinMaxValue = mix(\n                  value, minMaxValue, minMaxValueFound);\n              if (value ${t} currMinMaxValue) {\n                minMaxValue = value;\n                minMaxValueFound = 1.0;\n                minMaxPosition = ${r?a?m:g:`wR * ${h} + wC`};\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `)}let b=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;"avg"===t&&(b="avgValue / count");const v=4*Math.floor(s/4),x=s%4,w=`\n      if (${f}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = max(values, minMaxValue);\n      }\n    `;this.userCode=`\n      const ivec2 strides = ivec2(${i}, ${o});\n      const ivec2 pads = ivec2(${p}, ${d});\n      const float initializationValue = ${y};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xR, int xC, int d) {\n        if (xC < 0 || xC >= ${e.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xR, xC, d);\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d = coords[3];\n\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // max/min x(?, ?, d) to get y(yR, yC, d).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${y});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wR = 0; wR < ${c};\n            wR += ${u}) {\n          int xR = xRCorner + wR;\n\n          if (xR < 0 || xR >= ${e.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${v}; wC += 4) {\n            int xC = xCCorner + wC * ${l};\n\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${l}, d),\n              getValue(batch, xR, xC + 2 * ${l}, d),\n              getValue(batch, xR, xC + 3 * ${l}, d)\n            );\n\n            ${w}\n          }\n\n          int xC = xCCorner + ${v};\n          if (${1===x}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              initializationValue,\n              initializationValue,\n              initializationValue\n            );\n\n            ${w}\n          } else if (${2===x}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${l}, d),\n              initializationValue,\n              initializationValue\n            );\n\n            ${w}\n          } else if (${3===x}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${l}, d),\n              getValue(batch, xR, xC + 2 * ${l}, d),\n              initializationValue\n            );\n\n            ${w}\n          }\n        }\n        setOutput(${b});\n      }\n    `}},tP=class{constructor(e,t,n,r=!1,a=!1){if(this.variableNames=["x"],"avg"===t&&n)throw new Error("Cannot compute positions for average pool.");const s=e.filterWidth,i=e.strideDepth,o=e.strideHeight,u=e.strideWidth,l=e.dilationDepth,c=e.dilationHeight,h=e.dilationWidth,p=e.effectiveFilterDepth,d=e.effectiveFilterHeight,f=e.effectiveFilterWidth,m=e.padInfo.front,g=e.padInfo.top,y=e.padInfo.left;this.outputShape=e.outShape;const b="avg"===t;let v="0.0";if(b||(v="-1.0 / 1e-20"),n){const t=">=";return void(this.userCode=`\n        const ivec3 strides =\n            ivec3(${i}, ${o}, ${u});\n        const ivec3 pads = ivec3(${m}, ${g}, ${y});\n\n        void main() {\n          ivec5 coords = getOutputCoords();\n          int batch = coords.x;\n          int ch = coords.u;\n\n          ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n          int xDCorner = xCorner.x;\n          int xRCorner = xCorner.y;\n          int xCCorner = xCorner.z;\n\n          // max/min x(?, ?, ?, ch) to get y(yD, yR, yC, ch).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n\n          for (int wD = 0; wD < ${p};\n              wD += ${l}) {\n            int xD = xDCorner + wD;\n\n            if (xD < 0 || xD >= ${e.inDepth}) {\n              continue;\n            }\n\n            for (int wR = 0; wR < ${d};\n                wR += ${c}) {\n              int xR = xRCorner + wR;\n\n              if (xR < 0 || xR >= ${e.inHeight}) {\n                continue;\n              }\n\n              for (int wC = 0; wC < ${f};\n                  wC += ${h}) {\n                int xC = xCCorner + wC;\n\n                if (xC < 0 || xC >= ${e.inWidth}) {\n                  continue;\n                }\n\n                float value = getX(batch, xD, xR, xC, ch);\n\n                // If a min / max value has already been found, use it. If not,\n                // use the current value.\n                float currMinMaxValue = mix(\n                    value, minMaxValue, minMaxValueFound);\n                if (value ${t} currMinMaxValue) {\n                  minMaxValue = value;\n                  minMaxValueFound = 1.0;\n                  minMaxPosition = ${r?a?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${d} * ${f} +\n                      wR * ${f} + wC`};\n                }\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `)}let x=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;"avg"===t&&(x="avgValue / count");const w=4*Math.floor(s/4),k=s%4,S=`\n      if (${b}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = max(values, minMaxValue);\n      }\n    `;this.userCode=`\n      const ivec3 strides =\n        ivec3(${i}, ${o}, ${u});\n      const ivec3 pads = ivec3(${m}, ${g}, ${y});\n      const float initializationValue = ${v};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xD, int xR, int xC, int ch) {\n        if (xC < 0 || xC >= ${e.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xD, xR, xC, ch);\n      }\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xDCorner = xCorner.x;\n        int xRCorner = xCorner.y;\n        int xCCorner = xCorner.z;\n\n        // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${v});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wD = 0; wD < ${p};\n            wD += ${l}) {\n          int xD = xDCorner + wD;\n\n          if (xD < 0 || xD >= ${e.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${d};\n            wR += ${c}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${w}; wC += 4) {\n              int xC = xCCorner + wC * ${h};\n\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${h}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${h}, ch),\n                getValue(batch, xD, xR, xC + 3 * ${h}, ch)\n              );\n\n              ${S}\n            }\n\n            int xC = xCCorner + ${w};\n            if (${1===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                initializationValue,\n                initializationValue,\n                initializationValue\n              );\n\n              ${S}\n            } else if (${2===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${h}, ch),\n                initializationValue,\n                initializationValue\n              );\n\n              ${S}\n            } else if (${3===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${h}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${h}, ch),\n                initializationValue\n              );\n\n              ${S}\n            }\n          }\n          setOutput(${x});\n        }\n      }\n    `}};var nP={kernelName:ft,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t;KM(a,"avgPool");const{filterSize:s,strides:i,pad:o,dimRoundingMode:u}=r;va.assert(jd.eitherStridesOrDilationsAreOne(i,1),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);const l=jd.computePool2DInfo(a.shape,s,i,1,o,u);if(1===l.filterWidth&&1===l.filterHeight&&va.arraysEqual(l.inShape,l.outShape))return XO({inputs:{x:a},backend:n});const c=new eP(l,"avg",!1);return n.runWebGLProgram(c,[a],"float32")}};var rP={kernelName:gt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:u,dataFormat:l}=r,c=jd.computePool3DInfo(a.shape,s,i,[1,1,1],o,u,l),h=new tP(c,"avg",!1);return n.runWebGLProgram(h,[a],"float32")}};var aP={kernelName:yt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s,{filterSize:o,strides:u,pad:l,dimRoundingMode:c}=r,h=jd.computePool3DInfo(i.shape,o,u,[1,1,1],l,c),p=new class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;const t=e.filterDepth,n=e.filterHeight,r=e.filterWidth,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,u=e.dilationHeight,l=e.dilationWidth,c=e.effectiveFilterDepth,h=e.effectiveFilterHeight,p=e.effectiveFilterWidth,d=c-1-e.padInfo.front,f=h-1-e.padInfo.top,m=p-1-e.padInfo.left,g=1/(t*n*r);this.userCode=`\n      const ivec3 pads = ivec3(${d}, ${f}, ${m});\n      const float avgMultiplier = float(${g});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, d) with pos mask(:, :, :, ch) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${c};\n            wD += ${o}) {\n          float dyD = float(dyDCorner + wD) / ${a}.0;\n\n          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${h};\n              wR += ${u}) {\n            float dyR = float(dyRCorner + wR) / ${s}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${p};\n                wC += ${l}) {\n              float dyC = float(dyCCorner + wC) / ${i}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n\n              dotProd += dyValue * avgMultiplier;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(h);return n.runWebGLProgram(p,[a],i.dtype)}};var sP={kernelName:mt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s;KM([a,s],"avgPoolGrad");const{filterSize:o,strides:u,pad:l}=r,c=jd.computePool2DInfo(i.shape,o,u,1,l),h=new class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;const t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,a=e.strideWidth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterHeight,u=e.effectiveFilterWidth,l=o-1-e.padInfo.top,c=u-1-e.padInfo.left,h=1/(t*n);this.userCode=`\n      const ivec2 pads = ivec2(${l}, ${c});\n      const float avgMultiplier = float(${h});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${o};\n            wR += ${s}) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${u};\n            wC+= ${i}) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n\n            dotProd += dyValue * avgMultiplier;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(c);return n.runWebGLProgram(h,[a],i.dtype)}};var iP={kernelName:bt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{a:a,b:s}=t,{transposeA:i,transposeB:o}=r;return EL({a:a,b:s,transposeA:i,transposeB:o,backend:n})}},oP={kernelName:ln,backendName:"webgl",kernelFunc:({inputs:e,backend:t,attrs:n})=>{const{x:r,mean:a,variance:s,offset:i,scale:o}=e;va.assert(a.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),va.assert(null==i||a.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),va.assert(null==o||a.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:u}=n;null==u&&(u=.001);const l=[r,a,s];let c=null;null!=i&&(c=i.shape,l.push(i));let h=null;null!=o&&(h=o.shape,l.push(o));const p=Ye().getBool("WEBGL_PACK_NORMALIZATION")?new class{constructor(e,t,n,r,a,s){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],jd.assertAndGetBroadcastShape(e,t),jd.assertAndGetBroadcastShape(e,n);let i="vec4(0.0)";null!=r&&(jd.assertAndGetBroadcastShape(e,r),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="vec4(1.0)";null!=a&&(jd.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`\n      void main() {\n        vec4 offset = ${i};\n        vec4 scale = ${o};\n\n        vec4 x = getXAtOutCoords();\n        vec4 mean = getMeanAtOutCoords();\n        vec4 variance = getVarianceAtOutCoords();\n\n        vec4 inv = scale * inversesqrt(variance + vec4(${s}));\n\n        setOutput((x - mean) * inv + offset);\n      }\n    `}}(r.shape,a.shape,s.shape,c,h,u):new class{constructor(e,t,n,r,a,s){this.outputShape=[],this.variableNames=["x","mean","variance"],jd.assertAndGetBroadcastShape(e,t),jd.assertAndGetBroadcastShape(e,n);let i="0.0";null!=r&&(jd.assertAndGetBroadcastShape(e,r),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="1.0";null!=a&&(jd.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`\n      void main() {\n        float x = getXAtOutCoords();\n        float mean = getMeanAtOutCoords();\n        float variance = getVarianceAtOutCoords();\n        float offset = ${i};\n        float scale = ${o};\n        float inv = scale * inversesqrt(variance + float(${s}));\n        setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));\n      }\n    `}}(r.shape,a.shape,s.shape,c,h,u);return t.runWebGLProgram(p,l,l[0].dtype)}},uP=["x","y","z","w","u","v"];function lP(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,size:i}=r,[o,u]=fo.parseSliceParams(a,s,i);if(fo.assertParamsValid(a,o,u),0===va.sizeFromShape(u))return n.makeTensorInfo(u,a.dtype,[]);if(n.shouldExecuteOnCPU([a])||"string"===a.dtype){const e=n.texData.get(a.dataId),t=yO(e.values,o,u,a.shape,a.dtype);return n.makeTensorInfo(u,a.dtype,t)}const{isPacked:l}=n.texData.get(a.dataId),c=fo.isSliceContinous(a.shape,o,u);if(l||!c){const e=Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new class{constructor(e){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];const t=dD(this.rank),n=FO("coords",this.rank),r=FO("sourceLoc",this.rank),a=1===this.rank?"sourceLoc":`vec2(${r.slice(-2).join()})`,s=`getChannel(getSource(${r.join()}), ${a})`,i=`\n      result.x = ${s};\n      if (++${n[this.rank-1]} < ${e[this.rank-1]}) {\n        ++${r[this.rank-1]};\n        result.y = ${s};\n        --${r[this.rank-1]};\n      }\n    `,o=1===this.rank?"":`\n      --${n[this.rank-1]};\n      if (++${n[this.rank-2]} < ${e[this.rank-2]}) {\n        ++${r[this.rank-2]};\n        result.z = ${s};\n        if (++${n[this.rank-1]} < ${e[this.rank-1]}) {\n          ++${r[this.rank-1]};\n          result.w = ${s};\n        }\n      }\n    `,u=this.rank<=4?`sourceLoc = coords +\n            ${t}(${e.map((e,t)=>`start[${t}]`).join()});`:e.map((e,t)=>`${r[t]} = ${n[t]} + start[${t}];`).join("\n");this.userCode=`\n      void main() {\n        ${t} coords = getOutputCoords();\n        ${t} sourceLoc;\n        ${u}\n        vec4 result = vec4(0.);\n        ${i}\n        ${o}\n        setOutput(result);\n      }\n    `}}(u):new class{constructor(e){this.variableNames=["source"],this.outputShape=e,this.rank=e.length;const t=dD(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];const n=function(e){if(1===e)return"sourceLoc";if(e<=6)return uP.slice(0,e).map(e=>"sourceLoc."+e).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}(this.rank);let r;r=`\n        ${t} sourceLoc;\n        ${t} coords = getOutputCoords();\n        ${e.map((e,t)=>`sourceLoc.${uP[t]} = start[${t}] + coords.${uP[t]};`).join("\n")}\n      `,this.userCode=`\n      void main() {\n        ${r}\n        setOutput(getSource(${n}));\n      }\n    `}}(u),t=[o];return n.runWebGLProgram(e,[a],a.dtype,t)}return n.uploadToGPU(a.dataId),function(e,t,n,r){const a=r.texData.get(e.dataId),s=r.makeTensorInfo(n,e.dtype),i=r.texData.get(s.dataId);Object.assign(i,a),i.refCount=1,i.shape=n,i.dtype=e.dtype;let o=fo.computeFlatOffset(t,va.computeStrides(e.shape));a.slice&&(o+=a.slice.flatOffset),i.slice={flatOffset:o,origDataId:a.slice&&a.slice.origDataId||e.dataId};const u=r.dataRefCount.get(i.slice.origDataId)||1;return r.dataRefCount.set(i.slice.origDataId,u+1),s}(a,o,u,n)}var cP={kernelName:wr,backendName:"webgl",kernelFunc:lP},hP={kernelName:vt,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,crops:i}=r;va.assert(a.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");const o=s.reduce((e,t)=>e*t),u=jd.getReshaped(a.shape,s,o),l=jd.getPermuted(u.length,s.length),c=jd.getReshapedPermuted(a.shape,s,o),h=jd.getSliceBeginCoords(i,s.length),p=jd.getSliceSize(c,i,s.length),d=[],f=gL({inputs:{x:a},backend:n,attrs:{shape:u}}),m=IL({inputs:{x:f},backend:n,attrs:{perm:l}}),g=gL({inputs:{x:m},backend:n,attrs:{shape:c}}),y=lP({inputs:{x:g},backend:n,attrs:{begin:h,size:p}});return d.push(f),d.push(m),d.push(g),d.forEach(e=>n.disposeIntermediateTensorInfo(e)),y}};var pP={kernelName:xt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i}=r,o=n.readSync(a.dataId),u=n.readSync(s.dataId),l=jD(o,u,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,l)}};var dP={kernelName:kt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{s0:r,s1:a}=t,s=n.readSync(r.dataId),i=n.readSync(a.dataId),o=jd.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}},fP=oL({opSnippet:"return float(a != b);",cpuKernelImpl:hO,dtype:"bool"}),mP={kernelName:qn,backendName:"webgl",kernelFunc:fP};function gP(e){const{inputs:t,backend:n}=e,{input:r}=t;return XO({inputs:{x:n.texData.get(r.dataId).complexTensorInfos.real},backend:n})}var yP={kernelName:ir,backendName:"webgl",kernelFunc:gP};var bP={kernelName:St,backendName:"webgl",kernelFunc:function e(t){const{inputs:n,backend:r,attrs:a}=t,{x:s}=n,{dtype:i}=a;if("complex64"===i){if("complex64"===s.dtype)return XO({inputs:{x:s},backend:r});const t=Pc(s.shape),n=e({inputs:{x:s},backend:r,attrs:{dtype:"float32"}}),a=ZO({inputs:{real:n,imag:t},backend:r});return t.dispose(),r.disposeIntermediateTensorInfo(n),a}if("complex64"===s.dtype){const t=gP({inputs:{input:s},backend:r}),n=e({inputs:{x:t},backend:r,attrs:{dtype:i}});return r.disposeIntermediateTensorInfo(t),n}if(!va.hasEncodingLoss(s.dtype,i)){const e=XO({inputs:{x:s},backend:r});return{dataId:e.dataId,shape:e.shape,dtype:i}}if("int32"===i)return function(e,t){const n=new PO(e.shape,"return float(int(x));"),r=t.runWebGLProgram(n,[e],"int32");return{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}(s,r);if("bool"===i){const e=r.makeTensorInfo([],"bool",va.getTypedArrayFromDType("bool",1)),t=fP({inputs:{a:s,b:e},backend:r});return r.disposeIntermediateTensorInfo(e),t}throw new Error(`Error in Cast: failed to cast ${s.dtype} to ${i}`)}},vP="return ceil(x);",xP=iL({opSnippet:vP,packedOpSnippet:vP,cpuKernelImpl:HD}),wP={kernelName:It,backendName:"webgl",kernelFunc:xP};var kP={kernelName:At,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{clipValueMin:s,clipValueMax:i}=r;let o;o=Ye().getBool("WEBGL_PACK_CLIP")?new class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode="\n      void main() {\n        vec4 value = getAAtOutCoords();\n\n        if (any(isnan(value))) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, vec4(minVal), vec4(maxVal)));\n      }\n    "}}(a.shape):new class{constructor(e){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode="\n\n      void main() {\n        float value = getAAtOutCoords();\n        if (isnan(value)) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, minVal, maxVal));\n      }\n    "}}(a.shape);const u=[[s],[i]];return n.runWebGLProgram(o,[a],a.dtype,u)}};function SP(e,t){return{dataId:t.dataId,dtype:t.dtype,shape:e.shape}}var IP={kernelName:Nt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r}=t,a=n.texData.get(r.dataId),s=new class{constructor(e){this.variableNames=["real","imag"],this.outputShape=e,this.userCode="\n      void main() {\n        float re = abs(getRealAtOutCoords());\n        float im = abs(getImagAtOutCoords());\n        float mx = max(re, im);\n\n        // sadly the length function in glsl is not underflow-safe\n        // (at least not on Intel GPUs). So the safe solution is\n        // to ensure underflow-safety in all cases.\n        setOutput(\n          mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))\n        );\n      }\n    "}}(r.shape),i=[SP(r,a.complexTensorInfos.real),SP(r,a.complexTensorInfos.imag)];return n.runWebGLProgram(s,i,i[0].dtype)}};function AP(e,t,n){const r=e.indexOf(t);return e.map((e,t)=>t===r?`${e} - ${n}`:e).join()}function EP(e){const{inputs:t,backend:n}=e,{input:r}=t;return XO({inputs:{x:n.texData.get(r.dataId).complexTensorInfos.imag},backend:n})}var NP={kernelName:gn,backendName:"webgl",kernelFunc:EP};function CP(e,t,n){const r=e[0].dtype;if("complex64"===r){const r=e.map(e=>gP({inputs:{input:e},backend:n})),a=e.map(e=>EP({inputs:{input:e},backend:n})),s=CP(r,t,n),i=CP(a,t,n),o=ZO({inputs:{real:s,imag:i},backend:n});return r.forEach(e=>n.disposeIntermediateTensorInfo(e)),a.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),o}let a=n.shouldExecuteOnCPU(e);if("string"===r&&(a=!0),a){const a=e.map(e=>{const r=va.sizeFromShape(e.shape.slice(t));return gL({inputs:{x:e},backend:n,attrs:{shape:[-1,r]}})}),s=a.map(e=>({vals:n.readSync(e.dataId),shape:e.shape})),i=jd.computeOutShape(a.map(e=>e.shape),1),o=1===a[0].shape[0],u=qD(s,i,r,o),l=jd.computeOutShape(e.map(e=>e.shape),t),c=n.makeTensorInfo(l,r,u);return a.forEach(e=>n.disposeIntermediateTensorInfo(e)),c}if(e.length>Ye().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER")){const r=Math.floor(e.length/2),a=CP(e.slice(0,r),t,n),s=CP(e.slice(r),t,n),i=CP([a,s],t,n);return n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(s),i}if(Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&e[0].shape.length>1){const a=new class{constructor(e,t){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=jd.computeOutShape(e,t);const n=this.outputShape,r=n.length,a=dD(r),s=FO("coords",r),i=["x","y","z","w","u","v"].slice(0,r);this.variableNames=e.map((e,t)=>"T"+t);const o=new Array(e.length-1);o[0]=e[0][t];for(let n=1;n<o.length;n++)o[n]=o[n-1]+e[n][t];const u=i[t],l=i.slice(-2),c=i.join();let h=`if (${u} < ${o[0]}) {\n        return getChannel(\n            getT0(${c}), vec2(${l.join()}));\n        }`;for(let e=1;e<o.length;e++){const t=o[e-1];h+=`\n        if (${u} < ${o[e]}  && ${u} >= ${o[e-1]}) {\n          return getChannel(\n            getT${e}(${AP(i,u,t)}),\n            vec2(${AP(l,u,t)}));\n        }`}const p=o.length,d=o[o.length-1];h+=`\n        return getChannel(\n          getT${p}(${AP(i,u,d)}),\n          vec2(${AP(l,u,d)}));`,this.userCode=`\n      float getValue(${i.map(e=>"int "+e)}) {\n        ${h}\n      }\n\n      void main() {\n        ${a} coords = getOutputCoords();\n        vec4 result = vec4(getValue(${s}), 0., 0., 0.);\n\n        ${s[r-1]} = ${s[r-1]} + 1;\n        if (${s[r-1]} < ${n[r-1]}) {\n          result.g = getValue(${s});\n        }\n\n        ${s[r-2]} = ${s[r-2]} + 1;\n        if (${s[r-2]} < ${n[r-2]}) {\n          result.a = getValue(${s});\n        }\n\n        ${s[r-1]} = ${s[r-1]} - 1;\n        if (${s[r-2]} < ${n[r-2]} &&\n            ${s[r-1]} < ${n[r-1]}) {\n          result.b = getValue(${s});\n        }\n        setOutput(result);\n      }\n    `}}(e.map(e=>e.shape),t);return n.runWebGLProgram(a,e,r)}const{tensors2D:s,outShape:i}=function(e,t,n){const r=jd.computeOutShape(e.map(e=>e.shape),t);return{tensors2D:e.map(e=>gL({inputs:{x:e},attrs:{shape:[-1,va.sizeFromShape(e.shape.slice(t))]},backend:n})),outShape:r}}(e,t,n),o=new class{constructor(e){this.outputShape=[],this.outputShape=jd.computeOutShape(e,1),this.variableNames=e.map((e,t)=>"T"+t);const t=new Array(e.length-1);t[0]=e[0][1];for(let n=1;n<t.length;n++)t[n]=t[n-1]+e[n][1];const n=[`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];for(let e=1;e<t.length;e++){const r=t[e-1];n.push(`else if (yC < ${t[e]}) setOutput(getT${e}(yR, yC-${r}));`)}const r=t.length,a=t[t.length-1];n.push(`else setOutput(getT${r}(yR, yC-${a}));`),this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int yR = coords.x;\n        int yC = coords.y;\n\n        ${n.join("\n        ")}\n      }\n    `}}(s.map(e=>e.shape)),u=n.runWebGLProgram(o,s,r);s.forEach(e=>n.disposeIntermediateTensorInfo(e));const l=gL({inputs:{x:u},attrs:{shape:i},backend:n});return n.disposeIntermediateTensorInfo(u),l}function TP(e){const{inputs:t,backend:n,attrs:r}=e,{axis:a}=r,s=va.parseAxisParam(a,t[0].shape)[0],i=jd.computeOutShape(t.map(e=>e.shape),s);if(0===va.sizeFromShape(i))return n.makeTensorInfo(i,t[0].dtype,[]);const o=t.filter(e=>va.sizeFromShape(e.shape)>0);if(1===o.length)return XO({inputs:{x:o[0]},backend:n});const u=o.map(e=>e.shape);return jd.assertParamsConsistent(u,s),CP(o,s,n)}var RP={kernelName:Ct,backendName:"webgl",kernelFunc:TP},_P=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.outputShape=e.outShape;const s=e.padInfo.top,i=e.padInfo.left,o=e.strideHeight,u=e.strideWidth,l=e.dilationHeight,c=e.dilationWidth,h=e.filterHeight,p=e.filterWidth,d=4*Math.floor(e.inChannels/4),f=e.inChannels%4,m="channelsLast"===e.dataFormat,g=m?1:2,y=m?2:3,b=m?3:1;let v="",x="";n&&(v=r?`float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:a?`float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:`\n          float activation(float x) {\n            ${n}\n          }\n        `,x="result = activation(result);");const w=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n      ${v}\n\n      const ivec2 strides = ivec2(${o}, ${u});\n      const ivec2 pads = ivec2(${s}, ${i});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d2 = coords[${b}];\n\n        ivec2 xRCCorner =\n            ivec2(coords[${g}], coords[${y}]) * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${h}; wR++) {\n          int xR = xRCorner + wR * ${l};\n\n          if (xR < 0 || xR >= ${e.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${p}; wC++) {\n            int xC = xCCorner + wC * ${c};\n\n            if (xC < 0 || xC >= ${e.inWidth}) {\n              continue;\n            }\n\n            for (int d1 = 0; d1 < ${d}; d1 += 4) {\n              vec4 wValues = vec4(\n                getW(wR, wC, d1, d2),\n                getW(wR, wC, d1 + 1, d2),\n                getW(wR, wC, d1 + 2, d2),\n                getW(wR, wC, d1 + 3, d2)\n              );\n\n              if (${m}) {\n                vec4 xValues = vec4(\n                  getX(batch, xR, xC, d1),\n                  getX(batch, xR, xC, d1 + 1),\n                  getX(batch, xR, xC, d1 + 2),\n                  getX(batch, xR, xC, d1 + 3)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec4 xValues = vec4(\n                  getX(batch, d1, xR, xC),\n                  getX(batch, d1 + 1, xR, xC),\n                  getX(batch, d1 + 2, xR, xC),\n                  getX(batch, d1 + 3, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n\n            if (${1===f}) {\n\n              if (${m}) {\n                dotProd +=\n                    getX(batch, xR, xC, ${d}) *\n                    getW(wR, wC, ${d}, d2);\n              } else {\n                dotProd +=\n                    getX(batch, ${d}, xR, xC) *\n                    getW(wR, wC, ${d}, d2);\n              }\n\n            } else if (${2===f}) {\n              vec2 wValues = vec2(\n                getW(wR, wC, ${d}, d2),\n                getW(wR, wC, ${d} + 1, d2)\n              );\n\n              if (${m}) {\n                vec2 xValues = vec2(\n                  getX(batch, xR, xC, ${d}),\n                  getX(batch, xR, xC, ${d} + 1)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec2 xValues = vec2(\n                  getX(batch, ${d}, xR, xC),\n                  getX(batch, ${d} + 1, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            } else if (${3===f}) {\n              vec3 wValues = vec3(\n                getW(wR, wC, ${d}, d2),\n                getW(wR, wC, ${d} + 1, d2),\n                getW(wR, wC, ${d} + 2, d2)\n              );\n\n              if (${m}) {\n                vec3 xValues = vec3(\n                  getX(batch, xR, xC, ${d}),\n                  getX(batch, xR, xC, ${d} + 1),\n                  getX(batch, xR, xC, ${d} + 2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec3 xValues = vec3(\n                  getX(batch, ${d}, xR, xC),\n                  getX(batch, ${d} + 1, xR, xC),\n                  getX(batch, ${d} + 2, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            }\n          }\n        }\n\n        float result = dotProd;\n        ${w}\n        ${x}\n        setOutput(result);\n      }\n    `}};function FP({x:e,filter:t,convInfo:n,backend:r,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){const u=e.shape,l=r.texData.get(e.dataId),c=n.inChannels,h=u[0]*u[1]*u[2],p=n.outChannels,d="channelsLast"===n.dataFormat;let f;const m=[];if(!((1===h||1===p)&&c>1e3)&&l.isPacked&&d&&null!=l.texture&&u[2]%2!=0&&va.arraysEqual(l.shape.slice(-3),u.slice(-3))){const c=u[0]*u[1]*(u[2]+1),h={dataId:e.dataId,shape:[1,c,n.inChannels],dtype:e.dtype},p=l.shape;l.shape=l.shape.slice(),l.shape[l.shape.length-2]++,va.assert(LM(l.shape,h.shape),()=>`packed reshape ${l.shape} to ${h.shape} isn't free`);const d=gL({inputs:{x:t},backend:r,attrs:{shape:[1,n.inChannels,n.outChannels]}});m.push(d);const g=EL({a:h,b:d,backend:r,transposeA:!1,transposeB:!1,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),y=r.texData.get(g.dataId);va.assert(y.isPacked,()=>"batchMatMul result is expected to be packed"),l.shape=p,y.shape=n.outShape,f=XO({inputs:{x:g},backend:r}),f.shape=n.outShape,m.push(g)}else{const l=gL({inputs:{x:e},backend:r,attrs:{shape:[1,d?u[0]*u[1]*u[2]:u[0]*u[2]*u[3],n.inChannels]}}),c=gL({inputs:{x:t},backend:r,attrs:{shape:[1,n.inChannels,n.outChannels]}}),h=EL({a:l,b:c,transposeA:!1,transposeB:!1,backend:r,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i});f=gL({inputs:{x:h},backend:r,attrs:{shape:n.outShape}}),m.push(l),m.push(c),m.push(h)}for(const e of m)r.disposeIntermediateTensorInfo(e);return f}function MP({x:e,filter:t,convInfo:n,backend:r,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){const{filterWidth:u,filterHeight:l,inChannels:c,outWidth:h,outHeight:p,dataFormat:d}=n,f="channelsLast"===d,m=u*l*c,g=p*h,y=[m,g],b=[],v=gL({inputs:{x:e},backend:r,attrs:{shape:e.shape.slice(1)}}),x=gL({inputs:{x:t},backend:r,attrs:{shape:[1,m,va.sizeFromShape(t.shape)/m]}});b.push(v),b.push(x);const w=new class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=e,this.enableShapeUniforms=bD(this.outputShape.length);const{dataFormat:n}=t,r=XM(),a="channelsLast"===n,s=a?0:1,i=a?1:2,o=this.enableShapeUniforms?"if(blockIndex < outShape[1] && pos < outShape[0]) {":`if(blockIndex < ${e[1]} && pos < ${e[0]}) {`;let u="";for(let e=0;e<=1;e++)for(let t=0;t<=1;t++)u+=`\n          blockIndex = rc.y + ${t};\n          pos = rc.x + ${e};\n\n          ${o}\n            offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];\n            d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);\n\n            if(d0 < inputShape[${s}] && d0 >= 0) {\n              // Use custom imod instead mod. On Intel GPU, mod may generate\n              // unexpected value.\n              // https://github.com/tensorflow/tfjs/issues/5447\n              offsetX = imod(blockIndex, outWidth) * stride[1] - pad[1];\n              d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /\n                  inChannels);\n\n              if(d1 < inputShape[${i}] && d1 >= 0) {\n\n                ch = imod(pos, inChannels);\n\n                if (${a}) {\n                  innerDims = vec2(d1, ch);\n                  result[${2*e+t}] = getChannel(\n                    getA(d0, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                } else {\n                  innerDims = vec2(d0, d1);\n                  result[${2*e+t}] = getChannel(\n                    getA(ch, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                }\n              }\n            }\n          }\n        `;this.userCode=`\n      void main() {\n        ivec2 rc = getOutputCoords();\n\n        vec4 result = vec4(0);\n\n        int blockIndex, pos, offsetY, d0, offsetX, d1, ch;\n        vec2 innerDims;\n\n        ${u}\n\n        ${r.output} = result;\n      }\n    `}}(y,n),k=[v.shape,[n.padInfo.top,n.padInfo.left],[n.strideHeight,n.strideWidth],[n.dilationHeight,n.dilationWidth],[n.inChannels],[n.filterWidth*n.inChannels],[n.outWidth]],S=r.runWebGLProgram(w,[v],"float32",k),I=gL({inputs:{x:S},backend:r,attrs:{shape:[1,y[0],y[1]]}});b.push(S),b.push(I);const A=null!=a,E=null!=s,N="leakyrelu"===o,C=o?uL(o,!0):null,T=new lL(I.shape,x.shape,[1,g,n.outChannels],!0,!1,A,C,E,N),R=[I,x];if(a&&R.push(a),E&&R.push(s),N){const e=r.makeTensorInfo([],"float32",va.createScalarValue(i,"float32"));R.push(e),b.push(e)}const _=r.runWebGLProgram(T,R,"float32"),F=gL({inputs:{x:_},backend:r,attrs:{shape:f?[1,p,h,n.outChannels]:[1,n.outChannels,p,h]}});b.push(_);for(const e of b)r.disposeIntermediateTensorInfo(e);return F}var DP={kernelName:Tt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dataFormat:u,dilations:l,dimRoundingMode:c}=r,h=jd.convertConv2DDataFormat(u),p=jd.computeConv2DInfo(a.shape,s.shape,i,l,o,c,!1,h);let d;if(1!==p.filterHeight||1!==p.filterWidth||1!==p.dilationHeight||1!==p.dilationWidth||1!==p.strideHeight||1!==p.strideWidth||"SAME"!==p.padInfo.type&&"VALID"!==p.padInfo.type)if(Ye().getBool("WEBGL_CONV_IM2COL")&&1===a.shape[0])d=MP({x:a,filter:s,convInfo:p,backend:n});else{const e=new _P(p);d=n.runWebGLProgram(e,[a,s],"float32")}else d=FP({x:a,filter:s,convInfo:p,backend:n});const f=gL({inputs:{x:d},backend:n,attrs:{shape:p.outShape}});return n.disposeIntermediateTensorInfo(d),f}};var OP={kernelName:Rt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,dataFormat:u,dimRoundingMode:l,filterShape:c}=r,h=jd.convertConv2DDataFormat(u),p=jd.computeConv2DInfo(a.shape,c,i,1,o,l,!1,h),d=new class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;const t=e.strideHeight,n=e.strideWidth,r=e.padInfo.top,a=e.padInfo.left,s="channelsLast"===e.dataFormat;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int d2 = coords.w;\n\n        // Convolve x(?, ?, d1) with dy(:, :, d2) to get dw(wR, wC, d1, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yR = 0; yR < ${e.outHeight}; yR++) {\n            int xR = wR + yR * ${t} - ${r};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${e.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${a};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              if (${s}) {\n                float dyValue = getDy(b, yR, yC, d2);\n                float xValue = getX(b, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              } else {\n                float dyValue = getDy(b, d2, yR, yC);\n                float xValue = getX(b, d1, xR, xC);\n                dotProd += (xValue * dyValue);\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(p);return n.runWebGLProgram(d,[a,s],"float32")}};var LP={kernelName:_t,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:u,dataFormat:l,dimRoundingMode:c}=r,h=jd.convertConv2DDataFormat(l),p=jd.computeConv2DInfo(i,s.shape,o,1,u,c,!1,h),d=new class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;const t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,a=e.strideWidth,s="channelsLast"===e.dataFormat,i=t-1-e.padInfo.top,o=n-1-e.padInfo.left,u=s?1:2,l=s?2:3,c=s?3:1;this.userCode=`\n      const ivec2 pads = ivec2(${i}, ${o});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[${c}];\n\n        ivec2 dyCorner = ivec2(coords[${u}], coords[${l}]) - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            for (int d2 = 0; d2 < ${e.outChannels}; d2++) {\n\n              if (${s}) {\n                float xValue = getDy(batch, idyR, idyC, d2);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              } else {\n                float xValue = getDy(batch, d2, idyR, idyC);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(p);return n.runWebGLProgram(d,[a,s],"float32")}};var PP={kernelName:Ft,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:u}=r,l=jd.computeConv3DInfo(a.shape,s.shape,i,u,o),c=new class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;const t=e.padInfo.front,n=e.padInfo.top,r=e.padInfo.left,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,u=e.dilationHeight,l=e.dilationWidth,c=e.filterDepth,h=e.filterHeight,p=e.filterWidth,d=4*Math.floor(e.inChannels/4),f=e.inChannels%4;this.userCode=`\n      const ivec3 strides = ivec3(${a}, ${s}, ${i});\n      const ivec3 pads = ivec3(${t}, ${n}, ${r});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d2 = coords.u;\n\n        ivec3 xFRCCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xFCorner = xFRCCorner.x;\n        int xRCorner = xFRCCorner.y;\n        int xCCorner = xFRCCorner.z;\n\n        // Convolve x(?, ?, ?, d1) with w(:, :, :, d1, d2) to get\n        // y(yF, yR, yC, d2). ? = to be determined. : = across all\n        // values in that axis.\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${c}; wF++) {\n          int xF = xFCorner + wF * ${o};\n\n          if (xF < 0 || xF >= ${e.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${h}; wR++) {\n            int xR = xRCorner + wR * ${u};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${p}; wC++) {\n              int xC = xCCorner + wC * ${l};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              for (int d1 = 0; d1 < ${d}; d1 += 4) {\n                vec4 xValues = vec4(\n                  getX(batch, xF, xR, xC, d1),\n                  getX(batch, xF, xR, xC, d1 + 1),\n                  getX(batch, xF, xR, xC, d1 + 2),\n                  getX(batch, xF, xR, xC, d1 + 3)\n                );\n                vec4 wValues = vec4(\n                  getW(wF, wR, wC, d1, d2),\n                  getW(wF, wR, wC, d1 + 1, d2),\n                  getW(wF, wR, wC, d1 + 2, d2),\n                  getW(wF, wR, wC, d1 + 3, d2)\n                );\n\n                dotProd += dot(xValues, wValues);\n              }\n\n              if (${1===f}) {\n                dotProd +=\n                  getX(batch, xF, xR, xC, ${d}) *\n                  getW(wF, wR, wC, ${d}, d2);\n              } else if (${2===f}) {\n                vec2 xValues = vec2(\n                  getX(batch, xF, xR, xC, ${d}),\n                  getX(batch, xF, xR, xC, ${d} + 1)\n                );\n                vec2 wValues = vec2(\n                  getW(wF, wR, wC, ${d}, d2),\n                  getW(wF, wR, wC, ${d} + 1, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else if (${3===f}) {\n                vec3 xValues = vec3(\n                  getX(batch, xF, xR, xC, ${d}),\n                  getX(batch, xF, xR, xC, ${d} + 1),\n                  getX(batch, xF, xR, xC, ${d} + 2)\n                );\n                vec3 wValues = vec3(\n                  getW(wF, wR, wC, ${d}, d2),\n                  getW(wF, wR, wC, ${d} + 1, d2),\n                  getW(wF, wR, wC, ${d} + 2, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(l);return n.runWebGLProgram(c,[a,s],"float32")}};var $P={kernelName:Mt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:u}=r,l=jd.computeConv3DInfo(a.shape,u,i,1,o),c=new class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;const t=e.strideDepth,n=e.strideHeight,r=e.strideWidth,a=e.padInfo.front,s=e.padInfo.top,i=e.padInfo.left;this.userCode=`\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int wF = coords.x;\n        int wR = coords.y;\n        int wC = coords.z;\n        int d1 = coords.w;\n        int d2 = coords.u;\n\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yF = 0; yF < ${e.outDepth}; yF++) {\n            int xF = wF + yF * ${t} - ${a};\n\n            if (xF < 0 || xF >= ${e.inDepth}) {\n              continue;\n            }\n\n            for (int yR = 0; yR < ${e.outHeight}; yR++) {\n              int xR = wR + yR * ${n} - ${s};\n\n              if (xR < 0 || xR >= ${e.inHeight}) {\n                continue;\n              }\n\n              for (int yC = 0; yC < ${e.outWidth}; yC++) {\n                int xC = wC + yC * ${r} - ${i};\n\n                if (xC < 0 || xC >= ${e.inWidth}) {\n                  continue;\n                }\n\n                float dyValue = getDy(b, yF, yR, yC, d2);\n                float xValue = getX(b, xF, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(l);return n.runWebGLProgram(c,[a,s],"float32")}};var BP={kernelName:Dt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:u}=r,l=jd.computeConv3DInfo(u,s.shape,o,1,i),c=new class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;const t=e.filterDepth,n=e.filterHeight,r=e.filterWidth,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=t-1-e.padInfo.front,u=n-1-e.padInfo.top,l=r-1-e.padInfo.left;this.userCode=`\n      const ivec3 pads = ivec3(${o}, ${u}, ${l});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.u;\n\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyFCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${t}; wF++) {\n          float dyF = float(dyFCorner + wF) / ${a}.0;\n\n          if (dyF < 0.0 || dyF >= ${e.outDepth}.0 || fract(dyF) > 0.0) {\n            continue;\n          }\n          int idyF = int(dyF);\n\n          int wFPerm = ${t} - 1 - wF;\n\n          for (int wR = 0; wR < ${n}; wR++) {\n            float dyR = float(dyRCorner + wR) / ${s}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n              fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            int wRPerm = ${n} - 1 - wR;\n\n            for (int wC = 0; wC < ${r}; wC++) {\n              float dyC = float(dyCCorner + wC) / ${i}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              int wCPerm = ${r} - 1 - wC;\n\n              for (int d2 = 0; d2 < ${e.outChannels}; d2++) {\n                float xValue = getDy(batch, idyF, idyR, idyC, d2);\n                float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(l);return n.runWebGLProgram(c,[a,s],"float32")}},zP=iL({opSnippet:"if (isnan(x)) return x;\n  return cos(x);\n"}),UP={kernelName:Ot,backendName:"webgl",kernelFunc:zP},WP=iL({opSnippet:"\n  float e2x = exp(-x);\n  return (e2x + 1.0 / e2x) / 2.0;\n"}),VP={kernelName:Lt,backendName:"webgl",kernelFunc:WP},jP={kernelName:$t,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n,attrs:r}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:u,extrapolationValue:l}=r,c=new class{constructor(e,t,n,r,a){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];const[s,i,o,u]=e,[l]=t,[c,h]=n;this.outputShape=[l,c,h,u];const p="bilinear"===r?1:0,[d,f]=[i-1+".0",o-1+".0"],[m,g,y]=c>1?[""+(i-1)/(c-1),"(y2-y1) * height_ratio",`y1*${d} + float(y)*(height_scale)`]:["0.0","0.0","0.5 * (y1+y2) * "+d],[b,v,x]=h>1?[""+(o-1)/(h-1),"(x2-x1) * width_ratio",`x1*${f} + float(x)*(width_scale)`]:["0.0","0.0","0.5 * (x1+x2) * "+f];this.userCode=`\n      const float height_ratio = float(${m});\n      const float width_ratio = float(${b});\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int y = coords[1];\n        int x = coords[2];\n        int d = coords[3];\n\n        // get box vals\n        float y1 = getBoxes(b,0);\n        float x1 = getBoxes(b,1);\n        float y2 = getBoxes(b,2);\n        float x2 = getBoxes(b,3);\n\n        // get image in batch index\n        int bInd = round(getBoxInd(b));\n        if(bInd < 0 || bInd >= ${s}) {\n          return;\n        }\n\n        float height_scale = ${g};\n        float width_scale = ${v};\n\n        float in_y = ${y};\n        if( in_y < 0.0 || in_y > ${d} ) {\n          setOutput(float(${a}));\n          return;\n        }\n        float in_x = ${x};\n        if( in_x < 0.0 || in_x > ${f} ) {\n          setOutput(float(${a}));\n          return;\n        }\n\n        vec2 sourceFracIndexCR = vec2(in_x,in_y);\n        if(${p} == 1) {\n          // Compute the four integer indices.\n          ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);\n          ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));\n\n          float topLeft = getImage(b, sourceFloorCR.y, sourceFloorCR.x, d);\n          float bottomLeft = getImage(b, sourceCeilCR.y, sourceFloorCR.x, d);\n          float topRight = getImage(b, sourceFloorCR.y, sourceCeilCR.x, d);\n          float bottomRight = getImage(b, sourceCeilCR.y, sourceCeilCR.x, d);\n\n          vec2 fracCR = sourceFracIndexCR - vec2(sourceFloorCR);\n\n          float top = topLeft + (topRight - topLeft) * fracCR.x;\n          float bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;\n          float newValue = top + (bottom - top) * fracCR.y;\n          setOutput(newValue);\n        } else {\n          // Compute the coordinators of nearest neighbor point.\n          ivec2 sourceNearestCR = ivec2(floor(\n            sourceFracIndexCR + vec2(0.5,0.5)));\n          float newValue = getImage(b, sourceNearestCR.y, sourceNearestCR.x, d);\n          setOutput(newValue);\n        }\n      }\n    `}}(a.shape,s.shape,o,u,l);return n.runWebGLProgram(c,[a,s,i],"float32")}},GP=class{constructor(e,t,n){this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}],this.outputShape=e;const r=e.length,a=t?"0.0":`getX(${HP(r,"coords")})`,s=e[e.length-1];let i="",o="";t?(i=n?"end != "+(s-1):"end != 0",o=n?"end + 1":"end - 1"):(i=n?"end + pow2 < "+s:"end >= pow2",o=n?"end + pow2":"end - pow2"),this.userCode=`\n      void main() {\n        ${dD(r)} coords = getOutputCoords();\n        int end = ${qP(r,"coords")};\n        float val = ${a};\n        int pow2 = int(pow(2.0, index));\n        if (${i}) {\n          int idx = ${o};\n          ${qP(r,"coords")} = idx;\n          val += getX(${HP(r,"coords")});\n        }\n        setOutput(val);\n      }\n    `}};function HP(e,t){if(1===e)return""+t;if(2===e)return`${t}.x, ${t}.y`;if(3===e)return`${t}.x, ${t}.y, ${t}.z`;if(4===e)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw Error(`Cumulative sum for rank ${e} is not yet supported`)}function qP(e,t){if(1===e)return""+t;if(2===e)return t+".y";if(3===e)return t+".z";if(4===e)return t+".w";throw Error(`Cumulative sum for rank ${e} is not yet supported`)}var KP={kernelName:Pt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r,u=a.shape.length,l=jd.getAxesPermutation([s],u);let c=a;null!=l&&(c=IL({inputs:{x:a},backend:n,attrs:{perm:l}}));const h=jd.getInnerMostAxes(1,u)[0];if(h!==u-1)throw new Error(`WebGL cumsum shader expects an inner-most axis=${a.shape.length-1} but got axis=${s}`);const p=c.shape[h];let d=XO({inputs:{x:c},backend:n});for(let e=0;e<=Math.ceil(Math.log2(p))-1;e++){const t=new GP(c.shape,!1,o),r=[[e]],a=d;d=n.runWebGLProgram(t,[d],d.dtype,r),n.disposeIntermediateTensorInfo(a)}if(i){const e=new GP(c.shape,i,o),t=d;d=n.runWebGLProgram(e,[d],d.dtype),n.disposeIntermediateTensorInfo(t)}if(null!=l){const e=IL({inputs:{x:d},backend:n,attrs:{perm:jd.getUndoAxesPermutation(l)}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),e}return d}};var YP={kernelName:Bt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i,binaryOutput:o}=r;if(1===a.shape.length){const e=n.readSync(a.dataId),t=n.readSync(s.dataId),r=jD(e,t,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,r)}if(2===a.shape.length){const e=n.bufferSync(a),t=n.bufferSync(s),r=GD(e,t,i,o);return n.makeTensorInfo(r.shape,s.dtype,r.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)}};var XP={kernelName:zt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockSize:s,dataFormat:i}=r,o=a.shape[0],u=("NHWC"===i?a.shape[1]:a.shape[2])*s,l=("NHWC"===i?a.shape[2]:a.shape[3])*s,c=("NHWC"===i?a.shape[3]:a.shape[1])/(s*s),h="NHWC"===i?[o,u,l,c]:[o,c,u,l],p=new class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=[],this.outputShape=e,this.blockSize=t,this.dataFormat=n,this.userCode=`\n    void main() {\n      ivec4 coords = getOutputCoords();\n      int b = coords[0];\n      int h = ${this.getHeightCoordString()};\n      int w = ${this.getWidthCoordString()};\n      int d = ${this.getDepthCoordString()};\n\n      int in_h = h / ${t};\n      int offset_h = imod(h, ${t});\n      int in_w = w / ${t};\n      int offset_w = imod(w, ${t});\n      int offset_d = (offset_h * ${t} + offset_w) *\n        ${this.getOutputDepthSize()};\n      int in_d = d + offset_d;\n\n      float result = ${this.getInputSamplingString()};\n      setOutput(result);\n    }\n  `}getHeightCoordString(){return"NHWC"===this.dataFormat?"coords[1]":"coords[2]"}getWidthCoordString(){return"NHWC"===this.dataFormat?"coords[2]":"coords[3]"}getDepthCoordString(){return"NHWC"===this.dataFormat?"coords[3]":"coords[1]"}getOutputDepthSize(){return"NHWC"===this.dataFormat?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return"NHWC"===this.dataFormat?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}}(h,s,i);return n.runWebGLProgram(p,[a],a.dtype)}},QP=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=bD(this.outputShape.length);const s=e.filterHeight,i=e.filterWidth,o=e.outChannels/e.inChannels;let u="",l="";n&&(u=r?`float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:a?`float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:`\n          float activation(float x) {\n            ${n}\n          }\n        `,l="result = activation(result);");const c=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n      ${u}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${o};\n        int q = d2 - d1 * ${o};\n\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.\n        for (int wR = 0; wR < ${s}; wR++) {\n          int xR = xRCorner + wR * dilations[0];\n\n          if (xR < 0 || xR >= inDims[0]) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${i}; wC++) {\n            int xC = xCCorner + wC * dilations[1];\n\n            if (xC < 0 || xC >= inDims[1]) {\n              continue;\n            }\n\n            float xVal = getX(batch, xR, xC, d1);\n            float wVal = getW(wR, wC, d1, q);\n            dotProd += xVal * wVal;\n          }\n        }\n\n        float result = dotProd;\n        ${c}\n        ${l}\n        setOutput(result);\n      }\n    `}},ZP=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=bD(this.outputShape.length);const s=e.outChannels/e.inChannels,i=e.padInfo.left,o=e.strideWidth,u=e.dilationWidth,l=e.filterHeight,c=e.filterWidth,h=c;let p="\n      int xR; int xC; int xCOffset;\n      vec4 wTexel; vec4 previous; vec4 final;";for(let e=0;e<c;e++)p+=`\n          vec4 xTexelC${2*e};\n          int xTexelC${2*e}Ready;\n          vec4 xTexelC${2*e+1};\n          int xTexelC${2*e+1}Ready;\n          vec4 xC${e};`;p+=`\n    for (int r = 0; r < ${l}; r++) {\n      `;for(let e=0;e<c;e++)p+=`\n          xTexelC${2*e} = vec4(0.0);\n          xTexelC${2*e}Ready = 0;\n          xTexelC${2*e+1} = vec4(0.0);\n          xTexelC${2*e+1}Ready = 0;\n          xC${e} = vec4(0.0);`;p+="\n        xR = xRCorner + r * dilations[0];\n        if (xR >=0 && xR < inDims[0]) {\n      ";for(let e=0;e<(h+1)/2;e++){const t=2*e;if(p+=`\n          xC = xCCorner + ${t*u};\n          `,1===o){if(t<c&&(i%2==1?(p+=`\n                xCOffset = xC + 1;\n                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xCOffset, d1);\n\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n              `,p+=1===u&&t>0?`\n                xC${t} = vec4(xTexelC${t-2}.zw, xTexelC${t}.xy);\n                `:`\n                  xCOffset = xC + 1 - 2;\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                    previous = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      previous.zw = vec2(0.0);\n                    }\n\n                    xC${t} = vec4(previous.zw, xTexelC${t}.xy);\n                  } else {\n                    xC${t} = vec4(0.0, 0.0, xTexelC${t}.xy);\n                  }\n                  `):p+=`\n                if (xC >= 0 && xC < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                xC${t} = xTexelC${t};\n                `,t+1<c)){const e=i%2==0?va.nearestLargerEven(u):u;u%2==0&&i%2==1||u%2!=0&&i%2!=1?(p+=`\n                  xCOffset = xC + imod(pads[1], 2) + ${e};\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                    xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      xTexelC${t+1}.zw = vec2(0.0);\n                    }\n                    xTexelC${t+1}Ready = 1;\n                  }\n                  `,u>1&&(p+=`\n                    xCOffset -= 2;\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                      xTexelC${t} = getX(batch, xR, xCOffset, d1);\n                      xTexelC${t}Ready = 1;\n                    }\n                    `),p+=`\n                  xC${t+1} = vec4(xTexelC${t}.zw, xTexelC${t+1}.xy);\n                  `):p+=1===e?`\n                    xC${t+1} = xTexelC${t};\n                    `:`\n                    xCOffset = xC + ${e};\n\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                      xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n                      if (xCOffset + 1 >= inDims[1]) {\n                        xTexelC${t+1}.zw = vec2(0.0);\n                      }\n                      xTexelC${t+1}Ready = 1;\n                    }\n\n                    xC${t+1} = xTexelC${t+1};\n                    `}}else t<c&&(i%2==1?(p+=`\n                xCOffset = xC + 1 - strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xCOffset, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${t+1}Ready == 0) {\n                  xTexelC${t+1} = getX(batch, xR, xC + 1, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xC + 2 >= inDims[1]) {\n                    xTexelC${t+1}.zw = vec2(0.0);\n                  }\n                  xTexelC${t+1}Ready = 1;\n                }\n\n                xC${t} = vec4(xTexelC${t}.zw, xTexelC${t+1}.zw);\n              `,t+1<c&&(p+=`\n                  final = vec4(0.0);\n                  xCOffset = xC + 1 + strides[1];\n                  if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                    final = getX(batch, xR, xCOffset, d1);\n                  }\n                  xC${t+1} = vec4(xTexelC${t+1}.xy, final.xy);\n                `)):(p+=`\n                if(xC >= 0 && xC < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                xCOffset = xC + strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                  xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t+1}.zw = vec2(0.);\n                  }\n                  xTexelC${t+1}Ready = 1;\n                }\n\n                xC${t} = vec4(\n                  xTexelC${t}.xy, xTexelC${t+1}.xy);\n              `,t+1<c&&(p+=`\n                  xC${t+1} = vec4(xTexelC${t}.zw, xTexelC${t+1}.zw);\n                `)));t<c&&(p+=`\n            wTexel = getW(r, ${t}, d1, q);\n            dotProd += xC${t} * vec4(wTexel.xz, wTexel.xz);\n          `,t+1<c&&(p+=`\n              wTexel = getW(r, ${t+1}, d1, q);\n              dotProd += xC${t+1} * vec4(wTexel.xz, wTexel.xz);\n            `))}p+="\n    }\n  ",p+="\n      }\n    ";let d="",f="";n&&(d=r?`vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:a?`vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:`vec4 activation(vec4 x) {\n          ${n}\n        }`,f="result = activation(result);");const m=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n      ${d}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${s};\n        int q = d2 - d1 * ${s};\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n        vec4 dotProd = vec4(0.000000000000001);\n\n        ${p}\n\n        vec4 result = dotProd - vec4(0.000000000000001);\n        ${m}\n        ${f}\n        setOutput(result);\n      }\n    `}};var JP={kernelName:Ut,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:u,dimRoundingMode:l}=r;let c=u;null==c&&(c=[1,1]),va.assert(jd.eitherStridesOrDilationsAreOne(i,c),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${c}'`);const h=jd.computeConv2DInfo(a.shape,s.shape,i,c,o,l,!0);let p;p=Ye().getBool("WEBGL_PACK_DEPTHWISECONV")&&h.strideWidth<=2&&h.outChannels/h.inChannels==1?new ZP(h):new QP(h);const d=[[h.padInfo.top,h.padInfo.left],[h.strideHeight,h.strideWidth],[h.dilationHeight,h.dilationWidth],[h.inHeight,h.inWidth]];return n.runWebGLProgram(p,[a,s],"float32",d)}};var e$={kernelName:Wt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,dilations:o,pad:u,dimRoundingMode:l,filterShape:c}=r,h=jd.computeConv2DInfo(a.shape,c,i,o,u,l,!0),p=new class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;const t=e.strideHeight,n=e.strideWidth,r=e.padInfo.top,a=e.padInfo.left,s=e.outChannels/e.inChannels;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int dm = coords.w;\n        int d2 = d1 * ${s} + dm;\n\n        float dotProd = 0.0;\n\n        // TO DO: Vec4 over the batch size\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yR = 0; yR < ${e.outHeight}; yR++) {\n            int xR = wR + yR * ${t} - ${r};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${e.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${a};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              float dyValue = getDy(b, yR, yC, d2);\n              float xValue = getX(b, xR, xC, d1);\n              dotProd += (xValue * dyValue);\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(h);return n.runWebGLProgram(p,[a,s],"float32")}};var t$={kernelName:Vt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{strides:i,dilations:o,pad:u,dimRoundingMode:l,inputShape:c}=r,h=jd.computeConv2DInfo(c,s.shape,i,o,u,l,!0),p=new class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;const t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,a=e.strideWidth,s=t-1-e.padInfo.top,i=n-1-e.padInfo.left,o=e.outChannels/e.inChannels;this.userCode=`\n      const ivec2 pads = ivec2(${s}, ${i});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[3];\n        ivec2 dyCorner = coords.yz - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        float dotProd = 0.0;\n\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            // TO DO: Vec4 over the channelMul\n            for (int dm = 0; dm < ${o}; dm++) {\n              int d2 = d1 * ${o} + dm;\n              float xValue = getDy(batch, idyR, idyC, d2);\n              float wValue = getW(wRPerm, wCPerm, d1, dm);\n              dotProd += xValue * wValue;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(h);return n.runWebGLProgram(p,[a,s],"float32")}};var n$={kernelName:jt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r}=t,a=[...r.shape,...r.shape],s=va.sizeFromShape(r.shape),i=gL({inputs:{x:r},backend:n,attrs:{shape:[s]}}),o=new class{constructor(e){this.variableNames=["X"],this.outputShape=[e,e],this.userCode="\n      void main() {\n          ivec2 coords = getOutputCoords();\n          float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;\n          setOutput(val);\n      }\n    "}}(s),u=n.runWebGLProgram(o,[i],i.dtype),l=gL({inputs:{x:u},backend:n,attrs:{shape:a}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(u),l}};var r$={kernelName:Gt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:u}=r,l=jd.computeDilation2DInfo(a.shape,s.shape,i,o,"NHWC",u);let c;const h=new class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;const{inHeight:t,inWidth:n,padInfo:r,strideHeight:a,strideWidth:s,filterHeight:i,filterWidth:o,dilationHeight:u,dilationWidth:l}=e,{top:c,left:h}=r;this.userCode=`\n      const ivec2 strides = ivec2(${a}, ${s});\n      const ivec2 pads = ivec2(${c}, ${h});\n      const float neg_infinity = -3.4e38;\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.w;\n        ivec2 outTopLeftCorner =\n            coords.yz * strides - pads;\n        int hBeg = outTopLeftCorner.x;\n        int wBeg = outTopLeftCorner.y;\n\n        float curVal = neg_infinity;\n        for (int h = 0; h < ${i}; h++) {\n          int hIn = hBeg + h * ${u};\n\n          if (hIn >= 0 && hIn < ${t}) {\n            for (int w = 0; w < ${o}; w++) {\n              int wIn = wBeg + w * ${l};\n\n              if (wIn >= 0 && wIn < ${n}) {\n                float xVal = getX(batch, hIn, wIn, d1);\n                float wVal = getW(h, w, d1);\n\n                float val = xVal + wVal;\n                if (val > curVal) {\n                  curVal = val;\n                }\n              }\n            }\n          }\n        }\n\n        float result = curVal;\n        setOutput(result);\n      }\n    `}}(l);c=n.runWebGLProgram(h,[a,s],"float32");const p=gL({inputs:{x:c},backend:n,attrs:{shape:l.outShape}});return n.disposeIntermediateTensorInfo(c),p}};var a$={kernelName:Yt,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{equation:a}=r,s=t,{allDims:i,summedDims:o,idDims:u}=jd.decodeEinsumEquation(a,s.length);jd.checkEinsumDimSizes(i.length,u,s);const{path:l,steps:c}=jd.getEinsumComputePath(o,u),h=c.length;let p=null,d=i.length;const f=[];for(let e=0;e<h;++e){for(const t of c[e]){const{permutationIndices:e,expandDims:r}=jd.getEinsumPermutation(d,u[t]);let a;jd.isIdentityPermutation(e)?a=s[t]:(a=IL({inputs:{x:s[t]},backend:n,attrs:{perm:e}}),f.push(a));const i=a.shape.slice();for(let e=0;e<r.length;++e)i.splice(r[e],0,1);va.arraysEqual(a.shape,i)||(a=gL({inputs:{x:a},backend:n,attrs:{shape:i}}),f.push(a)),null===p?p=a:(p=fL({inputs:{a:a,b:p},backend:n}),f.push(p))}e<h-1&&(l[e]>=0&&(p=kL({inputs:{x:p},backend:n,attrs:{axis:l[e]-(i.length-d),keepDims:!1}}),f.push(p)),d--)}for(const e of f)e!==p&&n.disposeIntermediateTensorInfo(e);return p}},s$=iL({opSnippet:"return (x >= 0.0) ? x : (exp(x) - 1.0);",packedOpSnippet:"\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n"}),i$={kernelName:Xt,backendName:"webgl",kernelFunc:s$},o$={kernelName:Qt,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n}=e,{dy:r,y:a}=t,s=Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new YO("\n  vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));\n  return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));\n",r.shape,a.shape):new KO("return (b >= 1.0) ? a : a * (b + 1.0);",r.shape,a.shape);return n.runWebGLProgram(s,[r,a],r.dtype)}},u$=oL({opSnippet:"return float(a == b);",packedOpSnippet:"\n  return vec4(equal(a, b));\n",dtype:"bool",cpuKernelImpl:KD}),l$={kernelName:Jt,backendName:"webgl",kernelFunc:u$},c$=iL({opSnippet:`\n  // Error function is calculated approximately with elementary function.\n  // See "Handbook of Mathematical Functions with Formulas,\n  // Graphs, and Mathematical Tables", Abramowitz and Stegun.\n  float p = ${jd.ERF_P};\n  float a1 = ${jd.ERF_A1};\n  float a2 = ${jd.ERF_A2};\n  float a3 = ${jd.ERF_A3};\n  float a4 = ${jd.ERF_A4};\n  float a5 = ${jd.ERF_A5};\n\n  float sign = sign(x);\n  x = abs(x);\n  float t = 1.0 / (1.0 + p * x);\n  return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));\n`}),h$={kernelName:Zt,backendName:"webgl",kernelFunc:c$},p$=iL({opSnippet:"if (isnan(x)) return x;\n  return exp(x);\n",packedOpSnippet:"\n  vec4 result = exp(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n",cpuKernelImpl:YD,dtype:"float32"}),d$={kernelName:en,backendName:"webgl",kernelFunc:p$};function f$(e){const{inputs:t,attrs:n,backend:r}=e,{dim:a}=n,{input:s}=t,i=s.shape.length,o=s.shape.slice();let u=a;return a<0&&(va.assert(-(i+1)<=a,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+a+1),o.splice(u,0,1),gL({inputs:{x:s},backend:r,attrs:{shape:o}})}var m$={kernelName:tn,backendName:"webgl",kernelFunc:f$},g$="return exp(x) - 1.0;",y$=iL({opSnippet:g$,packedOpSnippet:g$,cpuKernelImpl:XD}),b$={kernelName:nn,backendName:"webgl",kernelFunc:y$},v$=class{constructor(e,t,n){this.variableNames=["real","imag"];const r=t[1];this.outputShape=t;const a=n?"2.0 * "+Math.PI:"-2.0 * "+Math.PI,s=n?r+".0":"1.0";let i;if("real"===e)i="return real * expR - imag * expI;";else{if("imag"!==e)throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);i="return real * expI + imag * expR;"}this.userCode=`\n      const float exponentMultiplier = ${a};\n\n      float unaryOpComplex(float real, float expR, float imag, float expI) {\n        ${i}\n      }\n\n      float mulMatDFT(int batch, int index) {\n        float indexRatio = float(index) / float(${r});\n        float exponentMultiplierTimesIndexRatio =\n            exponentMultiplier * indexRatio;\n\n        float result = 0.0;\n\n        for (int i = 0; i < ${r}; i++) {\n          // x = (-2|2 * PI / N) * index * i;\n          float x = exponentMultiplierTimesIndexRatio * float(i);\n          float expR = cos(x);\n          float expI = sin(x);\n          float real = getReal(batch, i);\n          float imag = getImag(batch, i);\n\n          result +=\n              unaryOpComplex(real, expR, imag, expI) / ${s};\n        }\n\n        return result;\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        setOutput(mulMatDFT(coords[0], coords[1]));\n      }\n    `}};function x$(e,t,n){const r=n.texData.get(e.dataId),a=va.sizeFromShape(e.shape),s=e.shape[e.shape.length-1],i=gL({inputs:{x:e},backend:n,attrs:{shape:[a/s,s]}}),o=i.shape,u=new v$("real",o,t),l=new v$("imag",o,t),c=[{dataId:r.complexTensorInfos.real.dataId,dtype:r.complexTensorInfos.real.dtype,shape:o},{dataId:r.complexTensorInfos.imag.dataId,dtype:r.complexTensorInfos.imag.dtype,shape:o}],h=n.runWebGLProgram(u,c,"float32"),p=n.runWebGLProgram(l,c,"float32"),d=ZO({inputs:{real:h,imag:p},backend:n});n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(p);const f=gL({inputs:{x:d},backend:n,attrs:{shape:e.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(d),f}var w$={kernelName:rn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{input:r}=t;return x$(r,!1,n)}};function k$(e){const{backend:t,attrs:n}=e,{shape:r,value:a}=n;let{dtype:s}=n;if(s=s||va.inferDtype(a),"string"===s){const e=va.getArrayFromDType(s,va.sizeFromShape(r));return e.fill(a),t.makeTensorInfo(r,s,e)}{const e=new class{constructor(e,t){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=e,this.userCode="\n      void main() {\n        // Input can be obtained from uniform value.\n        setOutput(value);\n      }\n    "}}(r,a),n=[[a]];return t.runWebGLProgram(e,[],s,n)}}var S$,I$={kernelName:an,backendName:"webgl",kernelFunc:k$},A$={kernelName:sn,backendName:"webgl",kernelFunc:({inputs:e,backend:t})=>{const{image:n}=e,r=t,a=new class{constructor(e){this.variableNames=["Image"],this.outputShape=[];const t=e[2];this.outputShape=e,this.userCode=`\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n\n          int coordX = ${t} - x - 1;\n          float outputValue;\n          if(coordX >= 0 && coordX < ${t}) {\n            outputValue = getImage(coords[0], coords[1], coordX, coords[3]);\n          } else {\n            outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `}}(n.shape);return r.runWebGLProgram(a,[n],n.dtype)}},E$="return floor(x);",N$=iL({opSnippet:E$,packedOpSnippet:E$,cpuKernelImpl:QD}),C$={kernelName:on,backendName:"webgl",kernelFunc:N$},T$=oL({opSnippet:"\n  float s = sign(a) * sign(b);\n  int ia = round(a);\n  int ib = round(b);\n  if (ib != 0) {\n    // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n    return float(idiv(ia, ib, s));\n  } else {\n    return NAN;\n  }\n",packedOpSnippet:"\n  ivec4 ia = round(a);\n  ivec4 ib = round(b);\n  bvec4 cond = notEqual(ib, ivec4(0));\n  ivec4 result = ivec4(0);\n  vec4 s = sign(a) * sign(b);\n\n  // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n  if (cond[0]) {\n    result[0] = idiv(ia[0], ib[0], s[0]);\n  }\n  if (cond[1]) {\n    result[1] = idiv(ia[1], ib[1], s[1]);\n  }\n  if (cond[2]) {\n    result[2] = idiv(ia[2], ib[2], s[2]);\n  }\n  if (cond[3]) {\n    result[3] = idiv(ia[3], ib[3], s[3]);\n  }\n  return vec4(result);\n",dtype:"int32"}),R$={kernelName:un,backendName:"webgl",kernelFunc:T$},_$={kernelName:ta,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e;let{pixels:a}=t;const{numChannels:s}=r,i="undefined"!=typeof HTMLVideoElement&&a instanceof HTMLVideoElement,o="undefined"!=typeof HTMLImageElement&&a instanceof HTMLImageElement,[u,l]=i?[a.videoWidth,a.videoHeight]:[a.width,a.height],c=[l,u],h=[l,u,s];(o||i)&&(null==S$&&(S$=document.createElement("canvas").getContext("2d")),S$.canvas.width=u,S$.canvas.height=l,S$.drawImage(a,0,0,u,l),a=S$.canvas);const p=n.makeTensorInfo(c,"int32");n.texData.get(p.dataId).usage=WF.PIXELS,n.gpgpu.uploadPixelDataToTexture(n.getTexture(p.dataId),a);const d=Ye().getBool("WEBGL_PACK")?new class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;const t=XM(),[n,r]=e;this.outputShape=e,this.userCode=`\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n\n        vec4 result = vec4(0.);\n\n        for(int row=0; row<=1; row++) {\n          for(int col=0; col<=1; col++) {\n            texC = coords[1] + row;\n            depth = coords[2] + col;\n\n            vec2 uv = (vec2(texC, texR) + halfCR) /\n                       vec2(${r}.0, ${n}.0);\n            vec4 values = ${t.texture2D}(A, uv);\n            float value;\n            if (depth == 0) {\n              value = values.r;\n            } else if (depth == 1) {\n              value = values.g;\n            } else if (depth == 2) {\n              value = values.b;\n            } else if (depth == 3) {\n              value = values.a;\n            }\n\n            result[row * 2 + col] = floor(value * 255.0 + 0.5);\n          }\n        }\n\n        ${t.output} = result;\n      }\n    `}}(h):new class{constructor(e){this.variableNames=["A"];const t=XM(),[n,r]=e;this.outputShape=e,this.userCode=`\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}.0, ${n}.0);\n\n        vec4 values = ${t.texture2D}(A, uv);\n        float value;\n        if (depth == 0) {\n          value = values.r;\n        } else if (depth == 1) {\n          value = values.g;\n        } else if (depth == 2) {\n          value = values.b;\n        } else if (depth == 3) {\n          value = values.a;\n        }\n\n        setOutput(floor(value * 255.0 + 0.5));\n      }\n    `}}(h),f=n.runWebGLProgram(d,[p],"int32");return n.disposeData(p.dataId),f}};var F$={kernelName:aa,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:u,pad:l,dataFormat:c,dilations:h,dimRoundingMode:p,activation:d,leakyreluAlpha:f}=r,m=jd.convertConv2DDataFormat(c),g=jd.computeConv2DInfo(a.shape,s.shape,u,h,l,p,!1,m);let y;const b=[];if(1!==g.filterHeight||1!==g.filterWidth||1!==g.dilationHeight||1!==g.dilationWidth||1!==g.strideHeight||1!==g.strideWidth||"SAME"!==g.padInfo.type&&"VALID"!==g.padInfo.type)if(Ye().getBool("WEBGL_CONV_IM2COL")&&1===a.shape[0])y=MP({x:a,filter:s,convInfo:g,backend:n,bias:i,activation:d,preluActivationWeights:o,leakyreluAlpha:f});else{const e=null!=i,t=null!=o,r="leakyrelu"===d,u=d?uL(d,!1):null,l=new _P(g,e,u,t,r),c=[a,s];if(i&&c.push(i),o&&c.push(o),r){const e=n.makeTensorInfo([],"float32",va.createScalarValue(f,"float32"));c.push(e),b.push(e)}y=n.runWebGLProgram(l,c,"float32")}else y=FP({x:a,filter:s,convInfo:g,backend:n,bias:i,activation:d,preluActivationWeights:o,leakyreluAlpha:f});const v=gL({inputs:{x:y},backend:n,attrs:{shape:g.outShape}});return b.push(y),b.forEach(e=>n.disposeIntermediateTensorInfo(e)),v}};var M$={kernelName:sa,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:u,pad:l,dilations:c,dimRoundingMode:h,activation:p,leakyreluAlpha:d}=r,f=[];let m=c;null==m&&(m=[1,1]),va.assert(jd.eitherStridesOrDilationsAreOne(u,m),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${u} and dilations '${m}'`);const g=jd.computeConv2DInfo(a.shape,s.shape,u,m,l,h,!0),y=Ye().getBool("WEBGL_PACK_DEPTHWISECONV")&&g.strideWidth<=2&&g.outChannels/g.inChannels==1,b=p?uL(p,y):null,v=[a,s],x=null!=i,w=null!=o,k="leakyrelu"===p;if(x&&v.push(i),w&&v.push(o),k){const e=n.makeTensorInfo([],"float32",va.createScalarValue(d,"float32"));v.push(e),f.push(e)}let S;S=y?new ZP(g,x,b,w,k):new QP(g,x,b,w,k);const I=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],A=n.runWebGLProgram(S,v,"float32",I);return f.forEach(e=>n.disposeIntermediateTensorInfo(e)),A}};var D$={kernelName:hn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{params:r,indices:a}=t,s=a.shape,i=s[s.length-1],o=va.sizeFromShape(r.shape),[u,l,c,h]=jd.prepareAndValidate(r,a),p=gL({inputs:{x:a},backend:n,attrs:{shape:[l,i]}}),d=gL({inputs:{x:r},backend:n,attrs:{shape:[va.sizeFromShape(r.shape)/c,c]}});if(n.shouldExecuteOnCPU([r,a])||"string"===r.dtype){const e=n.readSync(a.dataId),t=n.bufferSync(r),s=ZD(e,t,r.dtype,l,i,c,h,r.shape,o);return n.makeTensorInfo(u,r.dtype,s.values)}const f=new class{constructor(e,t,n){this.sliceDim=e,this.strides=t,this.variableNames=["x","indices"],this.outputShape=n;const r=dD(t.length),a=dD(n.length),s=this.sliceDim>1?"strides[j]":"strides";this.userCode=`\n        ${r} strides = ${r}(${this.strides});\n         void main() {\n          ${a} coords = getOutputCoords();\n          int flattenIndex = 0;\n          for (int j = 0; j < ${this.sliceDim}; j++) {\n            int index = round(getIndices(coords[0], j));\n            flattenIndex += index * ${s};\n          }\n          setOutput(getX(flattenIndex, coords[1]));\n        }\n      `}}(i,h,[l,c]),m=n.runWebGLProgram(f,[d,p],d.dtype),g=gL({inputs:{x:m},backend:n,attrs:{shape:u}});return n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(m),g}};function O$(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=r,u=va.parseAxisParam(i,a.shape)[0];if(Ye().get("DEBUG")){const e=n.readSync(s.dataId),t=a.shape[u];for(let n=0;n<e.length;++n){const r=e[n];va.assert(r<=t-1&&r>=0,()=>`GatherV2: the index value ${r} is not in [0, ${t-1}]`)}}const l=jd.segment_util.collectGatherOpShapeInfo(a,s,u,o),c=va.sizeFromShape(s.shape),h=[],p=gL({inputs:{x:a},backend:n,attrs:{shape:[l.batchSize,l.outerSize,l.dimSize,l.sliceSize]}}),d=gL({inputs:{x:s},backend:n,attrs:{shape:[l.batchSize,c/l.batchSize]}});h.push(p),h.push(d);const f=[l.batchSize,l.outerSize,c/l.batchSize,l.sliceSize];if(n.shouldExecuteOnCPU([a,s])||"string"===a.dtype){const e=n.bufferSync(d),t=n.bufferSync(p),r=JD(t,e,f);return h.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(l.outputShape,r.dtype,r.values)}const m=new class{constructor(e,t){this.variableNames=["A","indices"],this.outputShape=t,this.rank=t.length;const n=dD(this.rank),r=function(e,t){const n=["resRC.x","resRC.y","resRC.z","resRC.w"],r=[];for(let t=0;t<e.length;t++)2===t?r.push("index"):r.push(""+n[t]);return r.join()}(e);this.userCode=`\n      void main() {\n        ${n} resRC = getOutputCoords();\n        int index = int(getIndices(resRC.x, resRC.z));\n        float inBounds = (index >= 0) && (index < ${e[2]}) ? 1.0 : 0.0;\n        setOutput(inBounds * getA(${r}));\n      }\n    `}}(p.shape,f),g=n.runWebGLProgram(m,[p,d],p.dtype);h.push(g);const y=gL({inputs:{x:g},backend:n,attrs:{shape:l.outputShape}});return h.forEach(e=>n.disposeIntermediateTensorInfo(e)),y}var L$={kernelName:cn,backendName:"webgl",kernelFunc:O$},P$=oL({opSnippet:"return float(a > b);",packedOpSnippet:"\n  return vec4(greaterThan(a, b));\n",cpuKernelImpl:eO,dtype:"bool"}),$$={kernelName:pn,backendName:"webgl",kernelFunc:P$},B$=oL({opSnippet:"return float(a >= b);",packedOpSnippet:"\n  return vec4(greaterThanEqual(a, b));\n",dtype:"bool",cpuKernelImpl:tO}),z$={kernelName:dn,backendName:"webgl",kernelFunc:B$};var U$={kernelName:mn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{input:r}=t;return x$(r,!0,n)}},W$=iL({opSnippet:"return float(!isnan(x) && !isinf(x));",dtype:"bool"}),V$={kernelName:yn,backendName:"webgl",kernelFunc:W$},j$=iL({opSnippet:"return float(isinf(x));",dtype:"bool"}),G$={kernelName:bn,backendName:"webgl",kernelFunc:j$},H$=iL({opSnippet:"return float(isnan(x));",dtype:"bool"}),q$={kernelName:vn,backendName:"webgl",kernelFunc:H$},K$=oL({opSnippet:"return float(a < b);",packedOpSnippet:"\n  return vec4(lessThan(a, b));\n",cpuKernelImpl:nO,dtype:"bool"}),Y$={kernelName:wn,backendName:"webgl",kernelFunc:K$},X$=oL({opSnippet:"return float(a <= b);",packedOpSnippet:"\n  return vec4(lessThanEqual(a, b));\n",cpuKernelImpl:rO,dtype:"bool"}),Q$={kernelName:kn,backendName:"webgl",kernelFunc:X$};var Z$={kernelName:Sn,backendName:"webgl",kernelFunc:function(e){const{backend:t,attrs:n}=e,{start:r,stop:a,num:s}=n,i=aO(r,a,s);return t.makeTensorInfo([i.length],"float32",i)}},J$=iL({opSnippet:"if (isnan(x)) return x;\n  return x < 0.0 ? 0./0. : log(x);\n",packedOpSnippet:"\n  vec4 result = log(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);\n  result.g = isNaN.g ? x.g : (x.g < 0.0 ? 0./0. : result.g);\n  result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);\n  result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);\n  return result;\n",cpuKernelImpl:sO}),eB={kernelName:In,backendName:"webgl",kernelFunc:J$},tB=iL({opSnippet:"if (isnan(x)) return x;\n  return log(1.0 + x);\n"}),nB={kernelName:An,backendName:"webgl",kernelFunc:tB},rB=oL({opSnippet:"return float(a >= 1.0 && b >= 1.0);",packedOpSnippet:"\n  return vec4(\n    vec4(greaterThanEqual(a, vec4(1.0))) *\n    vec4(greaterThanEqual(b, vec4(1.0))));\n",dtype:"bool"}),aB={kernelName:En,backendName:"webgl",kernelFunc:rB},sB=iL({opSnippet:"return float(!(x >= 1.0));"}),iB={kernelName:Nn,backendName:"webgl",kernelFunc:sB},oB=oL({opSnippet:"return float(a >= 1.0 || b >= 1.0);",packedOpSnippet:"\n  return min(\n    vec4(greaterThanEqual(a, vec4(1.0))) +\n    vec4(greaterThanEqual(b, vec4(1.0))),\n    vec4(1.0));\n",dtype:"bool"}),uB={kernelName:Cn,backendName:"webgl",kernelFunc:oB},lB={kernelName:Rn,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:u}=r,l=Ye().getBool("WEBGL_PACK_NORMALIZATION")?new class{constructor(e,t,n,r,a){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;const s=t,i=e[3]-1;let o;this.outputShape=e;const u=`float(${n}) + float(${r}) * sum`;o=.5===a?`inversesqrt(${u})`:1===a?`1.0/(${u})`:`exp(log(${u}) * float(-${a}));`,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords.x;\n        int r = coords.y;\n        int c = coords.z;\n        int d = coords.w;\n\n        bool hasNextCol = d < ${this.outputShape[3]};\n        bool hasNextRow = c < ${this.outputShape[2]};\n\n        vec4 sum = vec4(0.);\n        vec4 xFragAtOutputCoords = getX(b, r, c, d);\n\n        vec4 xAtOutputCoords = vec4(\n          getChannel(xFragAtOutputCoords, vec2(c, d)),\n          hasNextCol ?\n            getChannel(xFragAtOutputCoords, vec2(c, d + 1)) : 0.0,\n          hasNextRow ?\n            getChannel(xFragAtOutputCoords , vec2(c + 1, d)) : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0\n        );\n\n        int firstChannel = d - ${s};\n        vec2 cache = vec2(0.);\n        if(firstChannel >= 0){\n          vec4 firstChannelFrag = getX(b, r, c, firstChannel);\n          cache.x = getChannel(firstChannelFrag, vec2(c, firstChannel));\n            if(hasNextRow){\n              cache.y = getChannel(firstChannelFrag, vec2(c + 1, firstChannel));\n            }\n        }\n\n        ivec2 depth = ivec2(d, d + 1);\n        for (int j = - ${s}; j <= ${s}; j++) {\n          ivec2 idx = depth + j;\n          bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));\n          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${i}));\n\n          bool depthInRange = aboveLowerBound.x && belowUpperBound.x;\n          bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;\n\n          if(depthInRange || depthPlusOneInRange){\n            vec4 z = vec4(0.);\n            vec4 xFragAtCurrentDepth;\n            z.xz = cache.xy;\n            if(depthPlusOneInRange && hasNextCol){\n              xFragAtCurrentDepth = idx.y != d ?\n                getX(b, r, c, idx.y) : xFragAtOutputCoords;\n              z.y = getChannel(xFragAtCurrentDepth, vec2(c, idx.y));\n              if(hasNextRow){\n                z.w = getChannel(xFragAtCurrentDepth, vec2(c + 1, idx.y));\n              }\n            }\n            cache.xy = z.yw;\n            sum += z * z;\n          }\n        }\n        vec4 result = xAtOutputCoords * ${o};\n        setOutput(result);\n      }\n    `}}(a.shape,s,i,o,u):new class{constructor(e,t,n,r,a){this.variableNames=["x"],this.outputShape=[];const s=t,i=e[3]-1;let o;this.outputShape=e;const u=`float(${n}) + float(${r}) * sum`;o=.5===a?`inversesqrt(${u})`:1===a?`1.0/(${u})`:`exp(log(${u}) * float(-${a}));`,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n        int d = coords[3];\n        float x = getX(b, r, c, d);\n        float sum = 0.0;\n        for (int j = -${s}; j <= ${s}; j++) {\n          int idx = d + j;\n          if (idx >= 0 && idx <=  ${i}) {\n            float z = getX(b, r, c, idx);\n            sum += z * z;\n          }\n        }\n        float val = x * ${o};\n        setOutput(val);\n      }\n    `}}(a.shape,s,i,o,u);return n.runWebGLProgram(l,[a],a.dtype)}},cB={kernelName:_n,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n,attrs:r}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:u,alpha:l,beta:c}=r,h=new class{constructor(e,t,n,r,a){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=e,this.depth=e[3],this.depthRadius=t,this.bias=n,this.alpha=r,this.beta=a,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n\n        float result = 0.0;\n        for (int d = 0; d < ${this.depth}; ++d) {\n          int depthBegin = int(max(0.0, float(d - ${t})));\n          int depthEnd = int(min(float(${this.depth}),\n              float(d + ${t} + 1)));\n\n          const int MIN_DEPTH_BEGIN = 0;\n          const int MAX_DEPTH_END = ${this.depth};\n\n          float norm = 0.0;\n          for (int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k) {\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd) {\n              norm += getInputImage(b, r, c, k) * getInputImage(b, r, c, k);\n            }\n            else {\n              break;\n            }\n          }\n\n          norm = float(${r}) * norm + float(${n});\n\n          for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd){\n              float dyi = -2.0 * float(${r})\n                * float(${a})\n                * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)\n                / norm;\n              if (k == d) {\n                dyi += pow(norm, -1.0 * ${a});\n              }\n              if (k == coords[3]) {\n                dyi *= getDy(b, r, c, d);\n                result += dyi;\n              }\n            }\n            else {\n              break;\n            }\n          }\n      }\n      setOutput(result);\n      }\n    `}}(a.shape,o,u,l,c);return n.runWebGLProgram(h,[a,s,i],a.dtype)}};function hB(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=r,o=a.shape.length,u=va.parseAxisParam(s,a.shape);let l=u;const c=jd.getAxesPermutation(l,o),h=null!=c,p=n.shouldExecuteOnCPU([a]);let d=a;if(h){if(p){const e=n.texData.get(d.dataId).values,t=new Array(o);for(let e=0;e<t.length;e++)t[e]=a.shape[c[e]];const r=TO(e,a.shape,a.dtype,c,t);d=n.makeTensorInfo(t,a.dtype);n.texData.get(d.dataId).values=r}else d=wL(a,c,n);l=jd.getInnerMostAxes(l.length,o)}jd.assertAxesAreInnerMostDims("max",l,o);const[f,m]=jd.computeOutAndReduceShapes(d.shape,l);let g,y=f;if(i&&(y=jd.expandShapeToKeepDim(f,u)),p){const e=n.texData.get(d.dataId).values,t=iO(e,va.sizeFromShape(m),y,a.dtype);g=n.makeTensorInfo(y,a.dtype);n.texData.get(g.dataId).values=t}else g=function(e,t,n,r){const a=va.sizeFromShape(t),s=gL({inputs:{x:e},attrs:{shape:[va.sizeFromShape(e.shape)/a,a]},backend:r}),i=xL(s,e.dtype,"max",r),o=gL({inputs:{x:i},attrs:{shape:n},backend:r});return r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}(d,m,y,n);return h&&n.disposeIntermediateTensorInfo(d),g}var pB={kernelName:Fn,backendName:"webgl",kernelFunc:hB},dB=oL({opSnippet:"\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n\n  return max(a, b);\n",packedOpSnippet:"\n  vec4 result = vec4(max(a, b));\n  vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n",cpuKernelImpl:oO}),fB={kernelName:Mn,backendName:"webgl",kernelFunc:dB};var mB={kernelName:Dn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t;KM(a,"maxPool");const{filterSize:s,strides:i,pad:o,dimRoundingMode:u}=r;va.assert(jd.eitherStridesOrDilationsAreOne(i,1),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);const l=jd.computePool2DInfo(a.shape,s,i,1,o,u);if(1===l.filterWidth&&1===l.filterHeight&&va.arraysEqual(l.inShape,l.outShape))return XO({inputs:{x:a},backend:n});const c=new eP(l,"max",!1);return n.runWebGLProgram(c,[a],a.dtype)}};var gB={kernelName:Ln,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dataFormat:u,dimRoundingMode:l}=r,c=jd.computePool3DInfo(a.shape,s,i,[1,1,1],o,l,u),h=new tP(c,"max",!1);return n.runWebGLProgram(h,[a],a.dtype)}};var yB={kernelName:Pn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s,{filterSize:o,strides:u,pad:l,dimRoundingMode:c}=r,h=jd.computePool3DInfo(i.shape,o,u,[1,1,1],l,c),p=new tP(h,"max",!0),d=n.runWebGLProgram(p,[i],i.dtype),f=new class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;const t=e.strideDepth,n=e.strideHeight,r=e.strideWidth,a=e.dilationDepth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterDepth,u=e.effectiveFilterHeight,l=e.effectiveFilterWidth,c=o-1-e.padInfo.front,h=u-1-e.padInfo.top,p=l-1-e.padInfo.left,d=o*u*l-1;this.userCode=`\n      const ivec3 pads = ivec3(${c}, ${h}, ${p});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, ch) with pos mask(:, :, :, d) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${o};\n           wD += ${a}) {\n          float dyD = float(dyDCorner + wD) / ${t}.0;\n\n          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${u};\n              wR += ${s}) {\n            float dyR = float(dyRCorner + wR) / ${n}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${l};\n                wC += ${i}) {\n              float dyC = float(dyCCorner + wC) / ${r}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n              int maxPosValue = ${d} -\n                  int(getMaxPos(batch, idyD, idyR, idyC, ch));\n\n              // Get the current value, check it against the value from the\n              // position matrix.\n              int curPosValue =\n                  wD * ${u} * ${l} +\n                  wR * ${l} + wC;\n              float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n              dotProd += dyValue * mask;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(h),m=n.runWebGLProgram(f,[a,d],i.dtype);return n.disposeIntermediateTensorInfo(d),m}};var bB={kernelName:On,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s,output:i}=t,o=s;KM([s,i],"maxPoolGrad");const{filterSize:u,strides:l,pad:c,dimRoundingMode:h}=r,p=jd.computePool2DInfo(o.shape,u,l,1,c,h),d=new eP(p,"max",!0),f=n.runWebGLProgram(d,[o],o.dtype),m=new class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;const t=e.strideHeight,n=e.strideWidth,r=e.dilationHeight,a=e.effectiveFilterHeight,s=e.effectiveFilterWidth,i=a-1-e.padInfo.top,o=s-1-e.padInfo.left,u=a*s-1;this.userCode=`\n      const ivec2 pads = ivec2(${i}, ${o});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${a};\n          wR += ${r}) {\n          float dyR = float(dyRCorner + wR) / ${t}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${s}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${n}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n            int maxPosValue = ${u} - int(getMaxPos(b, idyR, idyC, d));\n\n            // Get the current value, check it against the value from the\n            // position matrix.\n            int curPosValue = wR * ${s} + wC;\n            float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n            dotProd += dyValue * mask;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}}(p),g=n.runWebGLProgram(m,[a,f],o.dtype);return n.disposeIntermediateTensorInfo(f),g}};var vB={kernelName:$n,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{const{x:r}=e,{filterSize:a,strides:s,pad:i,includeBatchInIndex:o}=t,u=n;va.assert(4===r.shape.length,()=>`Error in maxPool: input must be rank 4 but got rank ${r.shape.length}.`);const l=[1,1];va.assert(jd.eitherStridesOrDilationsAreOne(s,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${l}'`);const c=jd.computePool2DInfo(r.shape,a,s,l,i),[h,p]=function(e,t,n,r){let a=new eP(n,"max",!1);const s=r.runWebGLProgram(a,[e],"float32");return a=new eP(n,"max",!0,!0,t),[s,r.runWebGLProgram(a,[e],"float32")]}(r,o,c,u);return[h,p]}};var xB={kernelName:Bn,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{const{x:r}=e,{keepDims:a,axis:s}=t,i=n,o=r.shape.length,u=va.parseAxisParam(s,r.shape);let l=u;const c=jd.getAxesPermutation(l,o),h=null!=c,p=i.shouldExecuteOnCPU([r]),d=[];let f=r;if(h){if(p){const e=i.texData.get(f.dataId).values,t=new Array(o);for(let e=0;e<t.length;e++)t[e]=r.shape[c[e]];const n=TO(e,r.shape,r.dtype,c,t);f=i.makeTensorInfo(t,r.dtype);i.texData.get(f.dataId).values=n}else f=wL(r,c,i);d.push(f),l=jd.getInnerMostAxes(l.length,o)}jd.assertAxesAreInnerMostDims("sum",l,o);const[m,g]=jd.computeOutAndReduceShapes(f.shape,l);let y=m;a&&(y=jd.expandShapeToKeepDim(m,u));const b=function(e,t,n,r){const a=va.sizeFromShape(t),s=gL({inputs:{x:e},attrs:{shape:[va.sizeFromShape(e.shape)/a,a]},backend:r}),i=xL(s,"float32","mean",r),o=gL({inputs:{x:i},attrs:{shape:n},backend:r});return r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}(f,g,y,i);for(const e of d)i.disposeIntermediateTensorInfo(e);return b}};var wB={kernelName:zn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,u=va.parseAxisParam(s,a.shape);let l=u;const c=jd.getAxesPermutation(l,o);let h=a;null!=c&&(h=IL({inputs:{x:a},backend:n,attrs:{perm:c}}),l=jd.getInnerMostAxes(l.length,a.shape.length)),jd.assertAxesAreInnerMostDims("min",l,o);const[p,d]=jd.computeOutAndReduceShapes(h.shape,l),f=gL({inputs:{x:h},backend:n,attrs:{shape:[-1,va.sizeFromShape(d)]}}),m=xL(f,f.dtype,"min",n);let g;if(i){g=gL({inputs:{x:m},backend:n,attrs:{shape:jd.expandShapeToKeepDim(p,u)}})}else g=gL({inputs:{x:m},backend:n,attrs:{shape:p}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),null!=c&&n.disposeIntermediateTensorInfo(h),g}},kB=oL({opSnippet:"\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n\n  return min(a, b);\n",packedOpSnippet:"\n  vec4 result = vec4(min(a, b));\n  vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n",cpuKernelImpl:uO}),SB={kernelName:Un,backendName:"webgl",kernelFunc:kB},IB={kernelName:Wn,backendName:"webgl",kernelFunc:({inputs:e,backend:t,attrs:n})=>{const{x:r}=e,{paddings:a,mode:s}=n,i=Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);const r=e.length,a=dD(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=FO("rc",r),u=FO("source",r),l=`${o[r-1]} < ${this.outputShape[r-1]}`,c=1===r?"source":`vec2(${u.slice(-2).join()})`,h="reflect"===n?0:1;let p="";if(1===r){const e=`\n        ${a} source = rc;\n        if (source < start) {\n          source = start * 2 - source - ${h};\n        } else if (source >= end) {\n          source = (end - 1) * 2 - source + ${h};\n        }\n        source -= start;\n      `;p=`\n        ${a} rc = outputLoc;\n        ${e}\n        result[0] = getChannel(getX(${u.join()}), ${c});\n        ${o[r-1]} += 1;\n        if(${l}) {\n          ${e}\n          result[1] = getChannel(getX(${u.join()}), ${c});\n        }\n      `}else{const e=`\n        ${a} source = rc;\n        ${a} lt = ${a}(lessThan(source, start));\n        ${a} gte = ${a}(greaterThanEqual(source, end));\n        ${a} orig = 1 - (lt + gte);\n        source = orig * source +\n                lt * (start * 2 - source - ${h}) +\n                gte * ((end - 1) * 2 - source + ${h});\n        source -= start;\n      `;p=`\n        ${a} rc = outputLoc;\n        ${e}\n        result[0] = getChannel(getX(${u.join()}), ${c});\n        ${o[r-1]} += 1;\n        if(${l}) {\n          ${e}\n          result[1] = getChannel(getX(${u.join()}), ${c});\n        }\n        rc = outputLoc;\n        ${o[r-2]} += 1;\n        if(${o[r-2]} < ${this.outputShape[r-2]}) {\n          ${e}\n          result[2] = getChannel(getX(${u.join()}), ${c});\n          ${o[r-1]} += 1;\n          if(${l}) {\n            ${e}\n            result[3] = getChannel(getX(${u.join()}), ${c});\n          }\n        }\n      `}this.userCode=`\n      const ${a} start = ${a}(${s});\n      const ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${p}\n        setOutput(result);\n      }\n    `}}(r.shape,a,s):new class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);const r=e.length,a=dD(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,r),u="reflect"===n?0:1;this.userCode=1!==r?`\n      ${a} start = ${a}(${s});\n      ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outC = getOutputCoords();\n        for (int i = 0; i < ${r}; i++) {\n          if (outC[i] < start[i]) {\n            outC[i] = start[i] * 2 - outC[i] - ${u};\n          } else if(outC[i] >= end[i]) {\n            outC[i] = (end[i] - 1) * 2 - outC[i] + ${u};\n          }\n        }\n        ${a} coords = outC - start;\n        setOutput(getX(${o}));\n      }\n    `:`\n        int start = ${s};\n        int end = ${i};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start) {\n            outC = start * 2 - outC - ${u};\n          } else if(outC >= end) {\n            outC = (end - 1) * 2 - outC + ${u};\n          }\n          setOutput(getX(outC - start));\n        }\n      `}}(r.shape,a,s);return t.runWebGLProgram(i,[r],r.dtype)}},AB=oL({opSnippet:"if (b == 0.0) return NAN;\n  return mod(a, b);",packedOpSnippet:"\n  vec4 result = mod(a, b);\n  vec4 isNaN = vec4(equal(b, vec4(0.0)));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n"}),EB={kernelName:Vn,backendName:"webgl",kernelFunc:AB},NB=oL({opSnippet:"\nif (a == b) {\n  return 1.0;\n};\nreturn a / b;",packedOpSnippet:"\n  // vec4 one = vec4(equal(a, b));\n  // return one + (vec4(1.0) - one) * a / b;\n  vec4 result = a / b;\n  if(a.x == b.x) {\n    result.x = 1.;\n  }\n  if(a.y == b.y) {\n    result.y = 1.;\n  }\n  if(a.z == b.z) {\n    result.z = 1.;\n  }\n  if(a.w == b.w) {\n    result.w = 1.;\n  }\n\n  return result;\n",checkOutOfBounds:!0}),CB={kernelName:Kt,backendName:"webgl",kernelFunc:NB},TB="return a - b;",RB=oL({opSnippet:TB,packedOpSnippet:TB,supportsComplex:!0,cpuKernelImpl:EO}),_B={kernelName:Vr,backendName:"webgl",kernelFunc:RB};function FB(e){const{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{dim:s}=r,i=va.parseAxisParam([s],a.shape),o=hB({inputs:{x:a},backend:n,attrs:{reductionIndices:i,keepDims:!1}}),u=jd.expandShapeToKeepDim(o.shape,i),l=gL({inputs:{x:o},backend:n,attrs:{shape:u}}),c=RB({inputs:{a:a,b:l},backend:n}),h=p$({inputs:{x:c},backend:n}),p=kL({inputs:{x:h},backend:n,attrs:{axis:i,keepDims:!1}}),d=gL({inputs:{x:p},backend:n,attrs:{shape:u}}),f=NB({inputs:{a:h,b:d},backend:n});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),f}var MB={kernelName:_r,backendName:"webgl",kernelFunc:FB};var DB={kernelName:jn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=r,u=o?a:FB({inputs:{logits:a},backend:n,attrs:{dim:a.shape.length-1}}),l=u.shape[0],c=u.shape[1],h=new class{constructor(e,t,n){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[e,n],this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n\n        float r = random(seed);\n        float cdf = 0.0;\n\n        for (int i = 0; i < ${t-1}; i++) {\n          cdf += getProbs(batch, i);\n\n          if (r < cdf) {\n            setOutput(float(i));\n            return;\n          }\n        }\n\n        // If no other event happened, last event happened.\n        setOutput(float(${t-1}));\n      }\n    `}}(l,c,s),p=[[i]],d=n.runWebGLProgram(h,[u],"int32",p);return o||n.disposeIntermediateTensorInfo(u),d}};var OB={kernelName:Hn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r}=t;if(n.shouldExecuteOnCPU([r])){const e=n.texData.get(r.dataId),[t,a]=cO(e.values,r.shape,r.dtype);return n.makeTensorInfo(a,r.dtype,t)}let a;return a=Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS")?new zO(r.shape,"\n  vec4 result = -x;\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"):new PO(r.shape,"if (isnan(x)) return x;\n  return -x;\n"),n.runWebGLProgram(a,[r],r.dtype)}},LB=jf.nonMaxSuppressionV3Impl;var PB={kernelName:Kn,backendName:"webgl",kernelFunc:function(e){jd.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");const{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:u}=r,l=n.readSync(a.dataId),c=n.readSync(s.dataId),{selectedIndices:h}=LB(l,c,i,o,u);return n.makeTensorInfo([h.length],"int32",new Int32Array(h))}},$B=jf.nonMaxSuppressionV4Impl;var BB={kernelName:Yn,backendName:"webgl",kernelFunc:function(e){jd.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");const{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:u,padToMaxOutputSize:l}=r,c=n.readSync(a.dataId),h=n.readSync(s.dataId),{selectedIndices:p,validOutputs:d}=$B(c,h,i,o,u,l);return[n.makeTensorInfo([p.length],"int32",new Int32Array(p)),n.makeTensorInfo([],"int32",new Int32Array([d]))]}},zB=jf.nonMaxSuppressionV5Impl;var UB={kernelName:Xn,backendName:"webgl",kernelFunc:function(e){jd.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");const{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:u,softNmsSigma:l}=r,c=n.readSync(a.dataId),h=n.readSync(s.dataId),p=i,d=o,f=u,m=l,{selectedIndices:g,selectedScores:y}=zB(c,h,p,d,f,m);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}},WB={kernelName:Zn,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n,attrs:r}=e,{indices:a}=t,{depth:s,onValue:i,offValue:o}=r,u=va.sizeFromShape(a.shape),l=new class{constructor(e,t,n,r){this.variableNames=["indices"],this.outputShape=[e,t],this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int index = round(getIndices(coords.x));\n        setOutput(mix(float(${r}), float(${n}),\n                      float(index == coords.y)));\n      }\n    `}}(u,s,i,o),c=gL({inputs:{x:a},backend:n,attrs:{shape:[u]}}),h=n.runWebGLProgram(l,[c],a.dtype);n.disposeIntermediateTensorInfo(c);const p=gL({inputs:{x:h},backend:n,attrs:{shape:[...a.shape,s]}});return n.disposeIntermediateTensorInfo(h),p}};function VB(e){const{inputs:t,backend:n}=e,{x:r}=t;if("complex64"===r.dtype){const e=gP({inputs:{input:r},backend:n}),t=VB({inputs:{x:e},backend:n}),a=EP({inputs:{input:r},backend:n}),s=VB({inputs:{x:a},backend:n}),i=ZO({inputs:{real:t,imag:s},backend:n});return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(s),i}return k$({attrs:{shape:r.shape,dtype:r.dtype,value:"string"===r.dtype?"":0},backend:n})}var jB={kernelName:Jr,backendName:"webgl",kernelFunc:VB};var GB={kernelName:Qn,backendName:"webgl",kernelFunc:function e(t){const{inputs:n,backend:r}=t,{x:a}=n;if("string"===a.dtype)throw new Error("onesLike is not supported under string dtype");if("complex64"===a.dtype){const t=gP({inputs:{input:a},backend:r}),n=e({inputs:{x:t},backend:r}),s=EP({inputs:{input:a},backend:r}),i=VB({inputs:{x:s},backend:r}),o=ZO({inputs:{real:n,imag:i},backend:r});return r.disposeIntermediateTensorInfo(t),r.disposeIntermediateTensorInfo(n),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}return k$({attrs:{shape:a.shape,dtype:a.dtype,value:1},backend:r})}};var HB={kernelName:Jn,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{axis:a}=r;if(1===t.length)return f$({inputs:{input:t[0]},backend:n,attrs:{dim:a}});const s=t[0].shape,i=t[0].dtype;t.forEach(e=>{va.assertShapesMatch(s,e.shape,"All tensors passed to stack must have matching shapes"),va.assert(i===e.dtype,()=>"All tensors passed to stack must have matching dtypes")});const o=[],u=TP({inputs:t.map(e=>{const t=f$({inputs:{input:e},backend:n,attrs:{dim:a}});return o.push(t),t}),backend:n,attrs:{axis:a}});return o.forEach(e=>n.disposeIntermediateTensorInfo(e)),u}},qB=e=>{const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{paddings:s,constantValue:i}=r;if(0===va.sizeFromShape(a.shape)){return k$({backend:n,attrs:{shape:s.map((e,t)=>e[0]+a.shape[t]+e[1]),value:i,dtype:a.dtype}})}const o=Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);const r=e.length,a=dD(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=FO("rc",r),u=FO("source",r),l=`${o[r-1]} < ${this.outputShape[r-1]}`,c=1===r?"source":`vec2(${u.slice(-2).join()})`,h=[a+" rc = outputLoc;",`${o[r-1]} += 1;\n       if(${l}) {\n      `,1===r?"":`}\n       rc = outputLoc;\n       ${o[r-2]} += 1;\n       if(${o[r-2]} < ${this.outputShape[r-2]}) {`,1===r?"":`  ${o[r-1]} += 1;\n         if(${l}) {`],p=1===r?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))";let d="";for(let e=0,t=1===r?2:4;e<t;e++)d+=`\n        ${h[e]}\n        if (${p}) {\n          result[${e}] = float(value);\n        } else {\n          ${a} source = rc - start;\n          result[${e}] = getChannel(getX(${u.join()}), ${c});\n        }\n      `;d+=1===r?"} ":"}}",this.userCode=`\n      const ${a} start = ${a}(${s});\n      const ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${d}\n        setOutput(result);\n      }\n    `}}(a.shape,s,i):new class{constructor(e,t,n){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);const r=e.length,a=dD(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,r);this.userCode=1!==r?`\n      ${a} start = ${a}(${s});\n      ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outC = getOutputCoords();\n        if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {\n          setOutput(value);\n        } else {\n          ${a} coords = outC - start;\n          setOutput(getX(${o}));\n        }\n      }\n    `:`\n        int start = ${s};\n        int end = ${i};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start || outC >= end) {\n            setOutput(value);\n          } else {\n            setOutput(getX(outC - start));\n          }\n        }\n      `}}(a.shape,s,i),u=[[i]];return n.runWebGLProgram(o,[a],a.dtype,u)},KB={kernelName:er,backendName:"webgl",kernelFunc:qB},YB=oL({opSnippet:"\n  if(a < 0.0 && floor(b) < b){\n    return NAN;\n  }\n  if (b == 0.0) {\n    return 1.0;\n  }\n  return (round(mod(b, 2.0)) != 1) ?\n      pow(abs(a), b) : sign(a) * pow(abs(a), b);\n",packedOpSnippet:"\n  // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.\n  vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));\n  vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);\n  vec4 result = multiplier * pow(abs(a), b);\n\n  // Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS\n  bvec4 isExpZero = equal(b, vec4(0.0));\n  result.r = isExpZero.r ? 1.0 : result.r;\n  result.g = isExpZero.g ? 1.0 : result.g;\n  result.b = isExpZero.b ? 1.0 : result.b;\n  result.a = isExpZero.a ? 1.0 : result.a;\n\n  vec4 isNaN = vec4(lessThan(a, vec4(0.0))) * vec4(lessThan(floor(b), b));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n"}),XB={kernelName:nr,backendName:"webgl",kernelFunc:YB};var QB={kernelName:ar,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,u=[],l=va.parseAxisParam(s,a.shape);let c=l;const h=jd.getAxesPermutation(c,o);let p,d=a;if(null!=h&&(d=IL({inputs:{x:a},backend:n,attrs:{perm:h}}),c=jd.getInnerMostAxes(c.length,o),u.push(d)),jd.assertAxesAreInnerMostDims("prod",c,o),n.shouldExecuteOnCPU([d])){const e=n.texData.get(d.dataId).values,{outVals:t,outShape:r,outDtype:a}=pO(d.shape,d.dtype,e,c);p=n.makeTensorInfo(r,a,t)}else{const[e,t]=jd.computeOutAndReduceShapes(d.shape,c),r=va.sizeFromShape(t),s=gL({inputs:{x:d},backend:n,attrs:{shape:[-1,r]}}),i=xL(s,hs(a.dtype),"prod",n);p=gL({inputs:{x:i},backend:n,attrs:{shape:e}}),u.push(s),u.push(i)}if(i){u.push(p);const e=jd.expandShapeToKeepDim(p.shape,l);p=gL({inputs:{x:p},backend:n,attrs:{shape:e}})}return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),p}},ZB=e=>{const{backend:t,attrs:n}=e,{start:r,stop:a,step:s,dtype:i}=n,o=dO(r,a,s,i);return t.makeTensorInfo([o.length],i,o)},JB={kernelName:sr,backendName:"webgl",kernelFunc:ZB},ez=iL({opSnippet:"return 1.0 / x;"}),tz={kernelName:or,backendName:"webgl",kernelFunc:ez},nz=iL({opSnippet:"if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : x;\n",packedOpSnippet:"\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"}),rz={kernelName:ur,backendName:"webgl",kernelFunc:nz},az=iL({opSnippet:"if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n",packedOpSnippet:"\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"}),sz={kernelName:fr,backendName:"webgl",kernelFunc:az};var iz={kernelName:pr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r,[u,l]=o,c=Ye().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new class{constructor(e,t,n,r,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];const[s,i,o,u]=e;this.outputShape=[s,t,n,u];const l=[r&&t>1?i-1:i,r&&n>1?o-1:o],c=[r&&t>1?t-1:t,r&&n>1?n-1:n];let h;h=a?"(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":"vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]},\n          ${l[1]/c[1]});\n      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,\n                                     ${o}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${h};\n\n        // Compute the four integer indices.\n        ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));\n        ivec3 sourceCeilRC = ivec3(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${u-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        // In parallel, construct four corners for all four components in\n        // packed 2x2 cell.\n        vec4 topLeft = vec4(\n          getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 bottomLeft = vec4(\n          getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 topRight = vec4(\n          getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec4 bottomRight = vec4(\n          getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec3 fracRC = sourceFracIndexRC - vec3(sourceFloorRC);\n\n        vec4 top = mix(topLeft, topRight, fracRC.yyzz);\n        vec4 bottom = mix(bottomLeft, bottomRight, fracRC.yyzz);\n        vec4 newValue = mix(top, bottom, fracRC.x);\n\n        setOutput(newValue);\n      }\n    `}}(a.shape,u,l,s,i):new class{constructor(e,t,n,r,a){this.variableNames=["A"],this.outputShape=[];const[s,i,o,u]=e;this.outputShape=[s,t,n,u];const l=[r&&t>1?i-1:i,r&&n>1?o-1:o],c=[r&&t>1?t-1:t,r&&n>1?n-1:n];let h;h=a?"(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":"vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]});\n      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${h};\n\n        // Compute the four integer indices.\n        ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));\n        ivec2 sourceCeilRC = ivec2(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        float topLeft = getA(b, sourceFloorRC.x, sourceFloorRC.y, d);\n        float bottomLeft = getA(b, sourceCeilRC.x, sourceFloorRC.y, d);\n        float topRight = getA(b, sourceFloorRC.x, sourceCeilRC.y, d);\n        float bottomRight = getA(b, sourceCeilRC.x, sourceCeilRC.y, d);\n\n        vec2 fracRC = sourceFracIndexRC - vec2(sourceFloorRC);\n\n        float top = topLeft + (topRight - topLeft) * fracRC.y;\n        float bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;\n        float newValue = top + (bottom - top) * fracRC.x;\n\n        setOutput(newValue);\n      }\n    `}}(a.shape,u,l,s,i);return n.runWebGLProgram(c,[a],"float32")}};var oz={kernelName:dr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r,o=new class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;const[,r,a]=t,[,s,i]=e,o=[n&&s>1?r-1:r,n&&i>1?a-1:a],u=[n&&s>1?s-1:s,n&&i>1?i-1:i],l=o[0]/u[0],c=o[1]/u[1],h=1/l,p=1/c,d=2*Math.ceil(h)+2,f=2*Math.ceil(p)+2;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${l});\n        const float widthScale = float(${c});\n\n        const float invHeightScale = float(${h});\n        const float invWidthScale = float(${p});\n\n        const int winHeight = int(${d});\n        const int winWidth = int(${f});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(startRLerp - float(winHeight / 2));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(startCLerp - float(winWidth / 2));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${s}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${i}) {\n              continue;\n            }\n\n            float dxR = float(dyR) * heightScale;\n            int topDxRIndex = int(floor(dxR));\n            int bottomDxRIndex = int(min(ceil(dxR), ${r-1}.0));\n            float dxRLerp = dxR - float(topDxRIndex);\n            float inverseDxRLerp = 1.0 - dxRLerp;\n\n            float dxC = float(dyC) * widthScale;\n            int leftDxCIndex = int(floor(dxC));\n            int rightDxCIndex = int(min(ceil(dxC), ${a-1}.0));\n            float dxCLerp = dxC - float(leftDxCIndex);\n            float inverseDxCLerp = 1.0 - dxCLerp;\n\n            if (r == topDxRIndex && c == leftDxCIndex) {\n              // topLeft\n              accumulator +=\n                getDy(b, dyR, dyC, d) * inverseDxRLerp * inverseDxCLerp;\n            }\n\n            if (r == topDxRIndex && c == rightDxCIndex) {\n              // topRight\n              accumulator += getDy(b, dyR, dyC, d) * inverseDxRLerp * dxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == leftDxCIndex) {\n              // bottomLeft\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * inverseDxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == rightDxCIndex) {\n              // bottomRight\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * dxCLerp;\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `}}(s.shape,a.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}};var uz={kernelName:cr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r,[u,l]=o,c=Ye().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new class{constructor(e,t,n,r,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];const[s,i,o,u]=e;this.outputShape=[s,t,n,u];const l=[r&&t>1?i-1:i,r&&n>1?o-1:o],c=[r&&t>1?t-1:t,r&&n>1?n-1:n],h=r?"0.5":"0.0";let p;p=a?"max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":"vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]},\n          ${l[1]/c[1]});\n      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,\n                                     ${o}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${p};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec3 sourceNearestRC = ivec3(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${h})));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${u-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        vec4 newValue = vec4(\n          getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),\n          hasNextCol ? getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d + 1) : 0.0);\n\n        setOutput(newValue);\n      }\n    `}}(a.shape,u,l,s,i):new class{constructor(e,t,n,r,a){this.variableNames=["A"],this.outputShape=[];const[s,i,o,u]=e;this.outputShape=[s,t,n,u];const l=[r&&t>1?i-1:i,r&&n>1?o-1:o],c=[r&&t>1?t-1:t,r&&n>1?n-1:n],h=r?"0.5":"0.0";let p;p=a?"max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":"vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]});\n      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${p};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec2 sourceNearestRC = ivec2(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${h})));\n        float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);\n\n        setOutput(newValue);\n      }\n    `}}(a.shape,u,l,s,i);return n.runWebGLProgram(c,[a],a.dtype)}};var lz={kernelName:hr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r,o=new class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;const[,r,a]=t,[,s,i]=e,o=[n&&s>1?r-1:r,n&&i>1?a-1:a],u=[n&&s>1?s-1:s,n&&i>1?i-1:i],l=o[0]/u[0],c=o[1]/u[1],h=1/l,p=1/c,d=2*Math.ceil(h)+2,f=2*Math.ceil(p)+2;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${l});\n        const float widthScale = float(${c});\n\n        const float invHeightScale = float(${h});\n        const float invWidthScale = float(${p});\n\n        const int winHeight = int(${d});\n        const int winWidth = int(${f});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(floor(startRLerp - float(winHeight / 2)));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(floor(startCLerp - float(winWidth / 2)));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${s}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${i}) {\n              continue;\n            }\n\n            float sourceFracRow =\n              float(${o[0]}) *\n                (float(dyR) / float(${u[0]}));\n\n            float sourceFracCol =\n                float(${o[1]}) *\n                  (float(dyC) / float(${u[1]}));\n\n            int sourceNearestRow = int(min(\n                float(int(${r}) - 1),\n                ${n} ? float(round(sourceFracRow)) :\n                                  float(floor(sourceFracRow))));\n\n            int sourceNearestCol = int(min(\n                float(int(${a}) - 1),\n                ${n} ? float(round(sourceFracCol)) :\n                                  float(floor(sourceFracCol))));\n\n            if (r == sourceNearestRow && c == sourceNearestCol) {\n              accumulator += getDy(b, dyR, dyC, d);\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `}}(s.shape,a.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}};var cz={kernelName:mr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dims:s}=r,i=a.shape.length,o=va.parseAxisParam(s,a.shape);if(0===i)return XO({inputs:{x:a},backend:n});const u=Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new class{constructor(e,t){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;const n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=e;const r=FO("rc",n),a=`${r[n-1]} + 1 < ${this.outputShape[n-1]}`,s=`${r[n-2]} + 1 < ${this.outputShape[n-2]}`,i=dD(n);var o;function u(n){const r=e.map((r,a)=>function(n,r){return-1!==t.indexOf(n)&&1!==e[n]?`${e[n]} - ${r[n]} - 1`:""+r[n]}(a,n));return`getChannel(getX(${r.join(",")}), vec2(${r.slice(-2).join(",")}))`}this.userCode=1===n?`\n        void main(){\n          int rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = getChannel(getX(${e[0]} - rc - 1),\n            ${e[0]} - rc - 1);\n          if(${a}){\n              result.g = getChannel(getX(${e[0]} - (rc  + 1) - 1),\n                ${e[0]} - (rc  + 1) - 1);\n          }\n          setOutput(result);\n        }\n      `:`\n        void main() {\n          ${i} rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = ${o=r.slice(),u(o)};\n          if(${a}){\n            result.g = ${function(e){return e[n-1]="("+e[n-1]+" + 1)",u(e)}(r.slice())};\n          }\n          if(${s}) {\n            result.b = ${function(e){return e[n-2]="("+e[n-2]+" + 1)",u(e)}(r.slice())};\n            if(${a}) {\n              result.a = ${function(e){return e[n-1]="("+e[n-1]+" + 1)",e[n-2]="("+e[n-2]+" + 1)",u(e)}(r.slice())};\n            }\n          }\n          setOutput(result);\n        }\n    `}}(a.shape,o):new class{constructor(e,t){this.variableNames=["x"];const n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=e,1===n)return void(this.userCode=`\n        void main() {\n          int coord = getOutputCoords();\n          setOutput(getX(${e[0]} - coord - 1));\n        }\n      `);const r=e.map((n,r)=>(n=>-1!==t.indexOf(n)&&1!==e[n]?`${e[n]} - coords[${n}] - 1`:`coords[${n}]`)(r)).join(","),a=dD(n);this.userCode=`\n      void main() {\n        ${a} coords = getOutputCoords();\n        setOutput(getX(${r}));\n      }\n    `}}(a.shape,o);return n.runWebGLProgram(u,[a],a.dtype)}},hz={kernelName:na,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{const{image:r}=e,{radians:a,fillValue:s,center:i}=t,o=n,u=new class{constructor(e,t){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];const n=e[1],r=e[2];this.outputShape=e;let a="";a="number"==typeof t?`float outputValue = ${t.toFixed(2)};`:`\n        vec3 fill = vec3(${t.join(",")});\n        float outputValue = fill[coords[3]];`,this.userCode=`\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n          int y = coords[1];\n          float coordXFloat = (float(x) - params[0]) * params[3] -\n            (float(y) - params[1]) * params[2];\n          float coordYFloat = (float(x) - params[0]) * params[2] +\n            (float(y) - params[1]) * params[3];\n          int coordX = int(round(coordXFloat + params[0]));\n          int coordY = int(round(coordYFloat + params[1]));\n          ${a}\n          if(coordX >= 0 && coordX < ${r} && coordY >= 0 && coordY < ${n}) {\n            outputValue = getImage(coords[0], coordY, coordX, coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `}}(r.shape,s),[l,c]=jd.getImageCenter(i,r.shape[1],r.shape[2]),h=[[l,c,Math.sin(a),Math.cos(a)]];return o.runWebGLProgram(u,[r],r.dtype,h)}},pz=iL({opSnippet:"\n  // OpenGL ES does not support round function.\n  // The algorithm is based on banker's rounding.\n  float base = floor(x);\n  if ((x - base) < 0.5) {\n    return floor(x);\n  } else if ((x - base) > 0.5) {\n    return ceil(x);\n  } else {\n    if (mod(base, 2.0) == 0.0) {\n      return base;\n    } else {\n      return base + 1.0;\n    }\n  }\n"}),dz={kernelName:gr,backendName:"webgl",kernelFunc:pz},fz=iL({opSnippet:"return inversesqrt(x);",cpuKernelImpl:fO}),mz={kernelName:yr,backendName:"webgl",kernelFunc:fz},gz=class{constructor(e,t,n,r,a,s,i=!0){this.variableNames=["updates","indices","defaultValue"],this.outputShape=s;const o=dD(a.length),u=dD(s.length);let l="";1===n?l="i":2===n&&(l="i, j");const c=`getIndices(${l})`;let h="";1===r?h="i":2===r&&(h="i, coords[1]");const p=`getUpdates(${h})`,d=t>1?"strides[j]":"strides";this.userCode=`\n        ${o} strides = ${o}(${a});\n\n        void main() {\n          ${u} coords = getOutputCoords();\n          float sum = 0.0;\n          bool found = false;\n          for (int i = 0; i < ${e}; i++) {\n            int flattenedIndex = 0;\n            for (int j = 0; j < ${t}; j++) {\n              int index = round(${c});\n              flattenedIndex += index * ${d};\n            }\n            if (flattenedIndex == coords[0]) {\n              sum += ${p};\n              found = true;\n            }\n          }\n          setOutput(mix(getDefaultValue(), sum, float(found)));\n        }\n      `}};var yz={kernelName:br,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{indices:a,updates:s}=t,{shape:i}=r,{sliceRank:o,numUpdates:u,sliceSize:l,strides:c,outputSize:h}=jd.calculateShapes(s,a,i),p=[h/l,l];if(0===h)return n.makeTensorInfo(i,a.dtype);const d=gL({inputs:{x:a},backend:n,attrs:{shape:[u,o]}}),f=gL({inputs:{x:s},backend:n,attrs:{shape:[u,l]}}),m=n.makeTensorInfo([],"float32",new Float32Array([0])),g=new gz(u,o,d.shape.length,f.shape.length,c,p),y=n.runWebGLProgram(g,[f,d,m],f.dtype),b=gL({inputs:{x:y},backend:n,attrs:{shape:i}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(m),b}};var bz={kernelName:vr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{condition:r,t:a,e:s}=t,i=new class{constructor(e,t,n){let r,a;if(this.variableNames=["c","a","b"],this.outputShape=t,n>4)throw Error(`Where for rank ${n} is not yet supported`);if(1===n)a="resRC",r="resRC";else{const n=["resRC.x","resRC.y","resRC.z","resRC.w"],s=[],i=[];for(let r=0;r<t.length;r++)i.push(""+n[r]),r<e&&s.push(""+n[r]);r=s.join(),a=i.join()}const s=dD(n);this.userCode=`\n      void main() {\n        ${s} resRC = getOutputCoords();\n        float cVal = getC(${r});\n        if (cVal >= 1.0) {\n          setOutput(getA(${a}));\n        } else {\n          setOutput(getB(${a}));\n        }\n      }\n    `}}(r.shape.length,a.shape,a.shape.length);return n.runWebGLProgram(i,[r,a,s],cs(a.dtype,s.dtype))}},vz=iL({opSnippet:`\n  // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.\n  // see: https://arxiv.org/abs/1706.02515\n  float scaleAlpha = ${jd.SELU_SCALEALPHA};\n  float scale = ${jd.SELU_SCALE};\n  return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);\n`}),xz={kernelName:xr,backendName:"webgl",kernelFunc:vz},wz=iL({opSnippet:"if (isnan(x)) return x;\n  return 1.0 / (1.0 + exp(-1.0 * x));\n",packedOpSnippet:"\n  vec4 result = 1.0 / (1.0 + exp(-1.0 * x));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n",cpuKernelImpl:mO}),kz={kernelName:Ar,backendName:"webgl",kernelFunc:wz},Sz=iL({opSnippet:"\n  if (isnan(x)) { return 0.0; }\n  return sign(x);\n"}),Iz={kernelName:Ir,backendName:"webgl",kernelFunc:Sz},Az=iL({opSnippet:"if (isnan(x)) return x;\n  return sin(x);\n"}),Ez={kernelName:kr,backendName:"webgl",kernelFunc:Az},Nz=iL({opSnippet:"\n  float e2x = exp(x);\n  return (e2x - 1.0 / e2x) / 2.0;\n"}),Cz={kernelName:Sr,backendName:"webgl",kernelFunc:Nz},Tz=iL({opSnippet:"\n  float epsilon = 1.1920928955078125e-7;\n  float threshold = log(epsilon) + 2.0;\n\n  bool too_large = x > -threshold;\n  bool too_small = x < threshold;\n\n  float result;\n  float exp_x = exp(x);\n\n  if (too_large){\n    result = x;\n  }\n  else if (too_small){\n    result = exp_x;\n  }\n  else{\n    result = log(exp_x + 1.0);\n  }\n  return result;\n"}),Rz={kernelName:Er,backendName:"webgl",kernelFunc:Tz},_z={kernelName:Tr,backendName:"webgl",kernelFunc:e=>{const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,paddings:i}=r;va.assert(a.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");const o=s.reduce((e,t)=>e*t),u=[[0,0]];u.push(...i);for(let e=1+s.length;e<a.shape.length;++e)u.push([0,0]);const l=[],c=qB({inputs:{x:a},backend:n,attrs:{paddings:u,constantValue:0}}),h=jd.getReshaped(c.shape,s,o,!1),p=jd.getPermuted(h.length,s.length,!1),d=jd.getReshapedPermuted(c.shape,s,o,!1),f=gL({inputs:{x:c},backend:n,attrs:{shape:h}}),m=IL({inputs:{x:f},backend:n,attrs:{perm:p}}),g=gL({inputs:{x:m},backend:n,attrs:{shape:d}});return l.push(c),l.push(f),l.push(m),l.forEach(e=>n.disposeIntermediateTensorInfo(e)),g}};var Fz={kernelName:Fr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{indices:r,values:a,denseShape:s,defaultValue:i}=t;if(1!==s.shape.length)throw new Error("Dense shape must be a vector, saw:\n         "+s.shape);if(2!==r.shape.length)throw new Error("Indices must be a matrix, saw:\n         "+r.shape);if(1!==a.shape.length)throw new Error("Values must be a vector, saw:\n         "+a.shape);if(0!==i.shape.length)throw new Error("Default value must be a scalar, saw:\n        "+i.shape);const o=n.readSync(r.dataId),u=n.readSync(a.dataId),l=n.readSync(s.dataId),c=n.readSync(i.dataId)[0],[h,p,d,f,m]=bO(o,r.shape,r.dtype,u,a.dtype,l,c);return[n.makeTensorInfo(p,r.dtype,h),n.makeTensorInfo([p[0]],a.dtype,d),n.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(e=>Number(e)))),n.makeTensorInfo([m.length],r.dtype,new Int32Array(m))]}};var Mz={kernelName:Mr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{inputIndices:r,inputShape:a,newShape:s}=t;if(2!==r.shape.length)throw new Error("Input indices should be a matrix but received shape "+r.shape);if(1!==a.shape.length)throw new Error("Input shape should be a vector but received shape "+a.shape);if(1!==s.shape.length)throw new Error("Target shape should be a vector but received shape "+s.shape);const i=Array.from(n.readSync(a.dataId)),o=n.readSync(r.dataId),u=Array.from(n.readSync(s.dataId)),[l,c,h]=vO(o,r.shape,r.dtype,i,u);return[n.makeTensorInfo(c,r.dtype,l),n.makeTensorInfo([h.length],s.dtype,new Int32Array(h))]}};var Dz={kernelName:Dr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error("Indices should be a vector but received shape\n              "+a.shape);if(1!==s.shape.length)throw new Error("Segment ids should be a vector but received shape\n              "+s.shape);const i=n.readSync(r.dataId),o=n.readSync(a.dataId),u=n.readSync(s.dataId),[l,c]=xO(i,r.shape,r.dtype,o,u,!0);return n.makeTensorInfo(c,r.dtype,l)}};var Oz={kernelName:Or,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error("Indices should be a vector but received shape\n             "+a.shape);if(1!==s.shape.length)throw new Error("Segment ids should be a vector but received shape\n             "+s.shape);const i=n.readSync(r.dataId),o=n.readSync(a.dataId),u=n.readSync(s.dataId),[l,c]=xO(i,r.shape,r.dtype,o,u);return n.makeTensorInfo(c,r.dtype,l)}};var Lz={kernelName:Lr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=t,{outputShape:o}=r,{sliceRank:u,numUpdates:l,strides:c,outputSize:h}=jd.calculateShapes(s,a,o),p=new gz(l,u,a.shape.length,s.shape.length,c,[h,1],!1),d=n.runWebGLProgram(p,[s,a,i],s.dtype),f=gL({inputs:{x:d},backend:n,attrs:{shape:o}});return n.disposeIntermediateTensorInfo(d),f}};var Pz={kernelName:Rr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=r,o=va.parseAxisParam(i,a.shape)[0],u=jd.prepareSplitSize(a,s,o),l=a.shape.length,c=new Array(l).fill(0),h=a.shape.slice();return u.map(e=>{const t=[...h];t[o]=e;const r=lP({inputs:{x:a},backend:n,attrs:{begin:c,size:t}});return c[o]+=e,r})}},$z="return sqrt(x);",Bz=iL({opSnippet:$z,packedOpSnippet:$z,cpuKernelImpl:wO}),zz={kernelName:Nr,backendName:"webgl",kernelFunc:Bz},Uz=iL({opSnippet:"return x * x;"}),Wz={kernelName:$r,backendName:"webgl",kernelFunc:Uz},Vz=oL({opSnippet:"return (a - b) * (a - b);",packedOpSnippet:"return (a - b) * (a - b);"}),jz={kernelName:Pr,backendName:"webgl",kernelFunc:Vz};var Gz={kernelName:ea,backendName:"webgl",kernelFunc:function({inputs:e,attrs:t,backend:n}){const{x:r}=e,a=`if (isnan(x)) return x;\n    return x > 0.0 ? 1.0 : float(${t.alpha});\n  `,s=new PO(r.shape,a);return n.runWebGLProgram(s,[r],r.dtype)}};var Hz={kernelName:Br,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,end:i,strides:o,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:h,shrinkAxisMask:p}=r,{finalShapeSparse:d,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:b,end:v,strides:x}=fo.sliceInfo(a.shape,s,i,o,u,l,c,h,p);let w;if(m)w=gL({inputs:{x:a},backend:n,attrs:{shape:f}});else if(g||y){va.assert(a.shape.length>=1,()=>"Input must have rank at least 1, got: "+a.shape.length);const e=fo.computeOutShape(b,v,x),t=lP({inputs:{x:a},backend:n,attrs:{begin:b,size:e}});w=gL({inputs:{x:t},backend:n,attrs:{shape:f}}),n.disposeIntermediateTensorInfo(t)}else{if(n.shouldExecuteOnCPU([a])){const e=n.readSync(a.dataId),t=Ii(a.shape,a.dtype,e),r=kO(d,t,x,b);w=n.makeTensorInfo(f,a.dtype,r.values)}else{const e=new class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=n;const r=n.length,a=dD(n.length),s=dD(n.length);let i="";if(1===r)i="coords * strides + begin";else{let e=0;i=n.map((t,r)=>(e++,1===n.length?`coords * strides[${r}] + begin[${r}]`:`coords[${e-1}] * strides[${r}] + begin[${r}]`)).join(",")}this.userCode=`\n      ${a} begin = ${a}(${e});\n      ${a} strides = ${a}(${t});\n\n      void main() {\n        ${s} coords = getOutputCoords();\n        setOutput(getX(${i}));\n      }\n    `}}(b,x,d);w=n.runWebGLProgram(e,[a],a.dtype)}}const k=gL({inputs:{x:w},backend:n,attrs:{shape:f}});return n.disposeIntermediateTensorInfo(w),k}};var qz={kernelName:zr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:u,preserveShortSequences:l}=r,{data:c,dataSplits:h}=t,p=n.readSync(c.dataId),d=n.readSync(h.dataId),[f,m]=SO(p,d,a,s,i,o,u,l);return[n.makeTensorInfo([f.length],"string",f),n.makeTensorInfo(h.shape,"int32",m)]}};var Kz={kernelName:Ur,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{skipEmpty:a}=r,{input:s,delimiter:i}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(1!==s.shape.length)throw new Error("Input must be a vector, got shape: "+s.shape);if(0!==i.shape.length)throw new Error("Delimiter must be a scalar, got shape: "+i.shape);const o=n.readSync(s.dataId),u=n.readSync(i.dataId)[0],[l,c,h]=IO(o,u,a),p=c.length;return[n.makeTensorInfo([p,2],"int32",l),n.makeTensorInfo([p],"string",c),n.makeTensorInfo([2],"int32",new Int32Array(h))]}};var Yz={kernelName:Wr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{numBuckets:a}=r,{input:s}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(a<=0)throw new Error("Number of buckets must be at least 1");const i=n.readSync(s.dataId),o=AO(i,a);return n.makeTensorInfo(s.shape,"int32",o)}},Xz=iL({opSnippet:"return tan(x);"}),Qz={kernelName:jr,backendName:"webgl",kernelFunc:Xz},Zz=iL({opSnippet:"\n  float e2x = exp(-2.0 * abs(x));\n  return sign(x) * (1.0 - e2x) / (1.0 + e2x);\n"}),Jz={kernelName:Gr,backendName:"webgl",kernelFunc:Zz};function eU(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reps:s}=r;if("string"===a.dtype||a.shape.length>5){const e=n.readSync(a.dataId),t="string"===a.dtype?e.map(e=>va.decodeString(e)):e,r=Ii(a.shape,a.dtype,t),i=NO(r,s);return n.makeTensorInfo(i.shape,i.dtype,i.values)}const i=new class{constructor(e,t){this.variableNames=["A"];const n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[r]*t[r];this.outputShape=n,this.rank=n.length;const r=dD(this.rank),a=function(e){const t=e.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(1===t)return`imod(resRC, ${e[0]})`;const n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],r=[];for(let t=0;t<e.length;t++)r.push(`imod(${n[t]}, ${e[t]})`);return r.join()}(e);this.userCode=`\n      void main() {\n        ${r} resRC = getOutputCoords();\n        setOutput(getA(${a}));\n      }\n    `}}(a.shape,s);return n.runWebGLProgram(i,[a],a.dtype)}var tU={kernelName:Hr,backendName:"webgl",kernelFunc:eU},nU=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=e,this.userCode="\n    void main() {\n         // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // The output size is half of the previous size.\n         // If the previous sequence is | | | | _ _ _ _  | | | |  _ _ _ _ (k=4),\n         // we only need to output the indices at positions |, the indices at\n         // positions _ can be thrown away, see Figure5(b) After Phase 2\n         // (Merge phase) in the Bitonic Top K paper referenced above.\n         // For example, the paper shows we only need to output the orange bars.\n         // The output sequence should look like this | | | | | | | |.\n         // Because the sequence is halved, to map the output index back\n         // to the previous sequence to find the corresponding value,\n         // we need to double the index. When we double the index,\n         // we basically interpolate a position, so 2i looks like\n         // | _ | _ | _ | _ | _ | _ | _. We move the | to the first k position\n         // of each 2k positions by - elemIdx % k. E.g. for output at\n         // index 4,5,6,7, we want to get the corresponding element at\n         // original index 8,9,10,11, for output at index 8,9,10,11,\n         // we want to get the corresponding element at original index\n         // 16,17,18,19, so on and so forth.\n\n         int i = elemIdx < k ? elemIdx : (elemIdx * 2 - imod(elemIdx, k));\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + k : int(getIndices(batch, i + k));\n\n         float x0 = getX(batch, i0);\n         float x1 = i1 < n ? getX(batch, i1) : x0;\n\n         setOutput(x0 >= x1 ? float(i0) : float(i1));\n       }\n     "}};function rU(e,t){null!==t&&e.disposeIntermediateTensorInfo(t)}function aU(e){let t=1;for(;t<e;)t*=2;return t}var sU={kernelName:qr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{k:s,sorted:i}=r,o=Ye().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),u=Ye().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),l=a.shape,c=l[l.length-1];if(n.shouldExecuteOnCPU([a])||c<o||s>u){const e=n.readSync(a.dataId),[t,r]=CO(e,l,a.dtype,s,i);return[n.makeTensorInfo(t.shape,t.dtype,t.values),n.makeTensorInfo(r.shape,r.dtype,r.values)]}if(0===s)return l[l.length-1]=0,[n.makeTensorInfo(l,a.dtype,[]),n.makeTensorInfo(l,"int32",[])];if(1===c)return[a,k$({attrs:{shape:l,dtype:"int32",value:0},backend:n})];const h=n.texData.get(a.dataId),p=null!==h&&h.isPacked,d=p?n.unpackTensor(a):a,f=va.sizeFromShape(l)/c,m=gL({inputs:{x:d},attrs:{shape:[f,c]},backend:n});p&&rU(n,d);const g=aU(s),y=aU(c);let b=null;const v=()=>null===b?[m,m]:[m,b],x=(e,t,r)=>{const a=v(),s=new class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=e,this.userCode="\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // We compare elements pair-wise within a group of size 2 * inc.\n         // The comparing rule for each group alternates between ascending\n         // and descending. Within each group, we compare each pair at\n         // positions i and i+inc. To decide whether an element at position i\n         // is x0 or x1, we mod it by 2 * inc, if the result is smaller than\n         // inc, it is in the first half of the group, we denote it as x0,\n         // otherwise we denote it as x1.\n         // For example, as shown in the Bitonic top K paper referenced above,\n         // Figure5(a) shows that element[1] is in the\n         // second half of the group when group size is 2, but it is in the\n         // first half of the group when group size is 4.\n\n         bool isFirstInPair = imod(elemIdx, 2 * inc) < inc;\n         int i = isFirstInPair ? elemIdx : elemIdx - inc;\n\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + inc : int(getIndices(batch, i + inc));\n         float x0 = i0 < n ? getX(batch, i0) : negativeInf;\n         float x1 = i1 < n ? getX(batch, i1) : negativeInf;\n\n         // Denotes which direction indices are in (ascending or descending).\n         bool reverse = imod(elemIdx, 2 * dir) >= dir;\n         bool isGreater = x0 > x1 || (x0 == x1 && i1 > i0);\n         if (reverse == isGreater) { // Elements in opposite order of direction\n           int iTemp = i0;\n           i0 = i1;\n           i1 = iTemp;\n         }\n         if (isFirstInPair) {\n            setOutput(float(i0));\n         } else {\n            setOutput(float(i1));\n         }\n       }\n     "}}(r),i=[[c],[null===b?1:0],[Number.NEGATIVE_INFINITY],[e],[t]],o=b;b=n.runWebGLProgram(s,a,"int32",i),rU(n,o)};for(let e=1;e<g;e*=2){const t=2*e;for(let n=e;n>=1;n/=2)x(t,n,[f,y])}for(let e=y;e>g;e/=2){const t=v(),r=new nU([f,e/2]),a=[[c],[null===b?1:0],[g]],s=b;b=n.runWebGLProgram(r,t,"int32",a),rU(n,s);const i=g/2,o=2*i;for(let e=i;e>=1;e/=2)x(o,e,b.shape)}let w=b;b=lP({inputs:{x:b},backend:n,attrs:{begin:0,size:[f,s]}}),rU(n,w);let k=O$({inputs:{x:m,indices:b},backend:n,attrs:{axis:1,batchDims:1}});rU(n,m);const S=l.slice(0,-1);S.push(s),w=b,b=gL({inputs:{x:b},attrs:{shape:S},backend:n}),rU(n,w);const I=k;return k=gL({inputs:{x:k},attrs:{shape:S},backend:n}),rU(n,I),[k,b]}};var iU={kernelName:Kr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:u,outputShape:l}=r,[c,h,p,d]=a.shape,[f,m]=null!=l?l:[h,p],g=[c,f,m,d],y=new class{constructor(e,t,n,r,a,s){this.variableNames=["Image","Transforms"],this.outputShape=s;const i="nearest"===n?1:2;let o;switch(r){case"constant":o=1;break;case"reflect":o=2;break;case"wrap":o=3;break;case"nearest":o=4;break;default:o=1}this.userCode=`\n            float mapCoord(float outCoord, float len) {\n              float inCoord = outCoord;\n              if(${o} == 2) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    if (inCoord < sz2) {\n                      inCoord = sz2 * float(int(float(-inCoord / sz2))) +\n                      inCoord;\n                    }\n                    inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1.0;\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    inCoord -= sz2 * float(int(float(inCoord / sz2)));\n                    if (inCoord >= len) {\n                      inCoord = sz2 - inCoord - 1.0;\n                    }\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${o} == 3) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord += len * (float(int(float(-inCoord / sz))) + 1.0);\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord -= len * float(int(float(inCoord / sz)));\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${o} == 4) {\n                return clamp(outCoord, 0.0, len - 1.0);\n              } else {\n                return outCoord;\n              }\n            }\n\n            float readWithFillValue(int batch, int coordY, int coordX,\n              int channel) {\n              float outputValue;\n              if (0 <= coordY && coordY < ${e} && 0 <= coordX && coordX < ${t}) {\n                  outputValue = getImage(batch, coordY, coordX, channel);\n              } else {\n                outputValue = float(${a});\n              }\n              return outputValue;\n            }\n\n            void main() {\n              ivec4 coords = getOutputCoords();\n              float outputValue;\n              int batch = coords[0];\n              int x = coords[2];\n              int y = coords[1];\n              int channel = coords[3];\n              float xf = float(x);\n              float yf = float(y);\n              float a1 = getTransforms(batch, 0);\n              float a2 = getTransforms(batch, 1);\n              float a3 = getTransforms(batch, 2);\n              float b1 = getTransforms(batch, 3);\n              float b2 = getTransforms(batch, 4);\n              float b3 = getTransforms(batch, 5);\n              float c1 = getTransforms(batch, 6);\n              float c2 = getTransforms(batch, 7);\n              float projection = c1 * xf + c2 * yf + 1.0;\n              if (projection == 0.0) {\n                outputValue = float(${a});\n              } else {\n                float inX = (a1 * xf + a2 * yf + a3) / projection;\n                float inY = (b1 * xf + b2 * yf + b3) / projection;\n                float mapX = mapCoord(inX, float(${t}));\n                float mapY = mapCoord(inY, float(${e}));\n\n                if (${i} == 1) {\n                  int coordY = int(round(mapY));\n                  int coordX = int(round(mapX));\n                  outputValue = readWithFillValue(batch, coordY, coordX,\n                    channel);\n                } else {\n                  float yFloor = floor(mapY);\n                  float xFloor = floor(mapX);\n                  float yCeil = yFloor + 1.0;\n                  float xCeil = xFloor + 1.0;\n                  float valueYFloor = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yFloor), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yFloor), int(xCeil), channel);\n                  float valueYCeil = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yCeil), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yCeil), int(xCeil), channel);\n                  outputValue = (yCeil - mapY) * valueYFloor +\n                  (mapY - yFloor) * valueYCeil;\n                }\n              }\n              setOutput(outputValue);\n            }\n        `}}(h,p,i,o,u,g);return n.runWebGLProgram(y,[a,s],"float32")}};var oU={kernelName:Xr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{axis:a}=n,{x:s}=t;KM(s,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");const i=r.readSync(s.dataId),{outputValues:o,outputShape:u,indices:l}=RO(i,a,s.shape,s.dtype);return[r.makeTensorInfo(u,s.dtype,o),r.makeTensorInfo([l.length],"int32",l)]}};var uU={kernelName:Qr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{value:a}=t;let{axis:s}=r;s<0&&(s+=a.shape.length);const i=a,o=i.shape.length,u=a.shape[s],l=new Array(o-1);let c=0;for(let e=0;e<o;e++)e!==s&&(l[c++]=i.shape[e]);const h=[],p=new Array(o).fill(0),d=i.shape.slice();d[s]=1;const f=new Array(u);for(let e=0;e<f.length;e++){p[s]=e;const t=lP({inputs:{x:i},backend:n,attrs:{begin:p,size:d}}),r=gL({inputs:{x:t},backend:n,attrs:{shape:l}});f[e]=r,h.push(t)}return h.forEach(e=>n.disposeIntermediateTensorInfo(e)),f}};var lU,cU,hU,pU,dU,fU=[NL,TL,_L,ML,LL,PL,$L,BL,WL,VL,GL,qL,YL,QL,JL,nP,rP,aP,sP,iP,oP,hP,pP,dP,bP,wP,kP,JO,IP,RP,DP,OP,LP,PP,$P,BP,UP,VP,jP,KP,YP,XP,JP,e$,t$,n$,r$,a$,i$,o$,l$,h$,d$,m$,b$,w$,I$,A$,C$,R$,_$,F$,M$,D$,L$,$$,z$,QO,U$,NP,V$,G$,q$,nL,Y$,Q$,Z$,eB,nB,aB,iB,uB,lB,cB,pB,fB,mB,gB,yB,bB,vB,xB,wB,SB,IB,EB,DB,mL,OB,PB,BB,UB,mP,WB,GB,HB,KB,XB,sL,QB,JB,yP,CB,tz,rz,sz,yL,iz,oz,uz,lz,cz,hz,dz,mz,yz,bz,xz,kz,Iz,Ez,Cz,cP,MB,Rz,_z,Fz,Mz,Dz,Oz,Lz,Pz,zz,Wz,jz,Gz,Hz,qz,Kz,Yz,_B,SL,Qz,Jz,tU,sU,iU,AL,oU,uU,{kernelName:Zr,backendName:"webgl",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a,segmentIds:s}=t,{numSegments:i}=r,o=a.shape.length,u=[];let l=0;const c=jd.getAxesPermutation([l],o);let h=a;null!=c&&(h=IL({inputs:{x:a},backend:n,attrs:{perm:c}}),u.push(h),l=jd.getInnerMostAxes(1,o)[0]);const p=jd.segment_util.computeOutShape(h.shape,l,i),d=va.sizeFromShape([h.shape[l]]),f=gL({inputs:{x:h},backend:n,attrs:{shape:[-1,d]}});u.push(f);const m=hs(a.dtype),g=(e,t,r,a,s)=>{const i=e.shape[0],o=e.shape[1],l=jd.segment_util.segOpComputeOptimalWindowSize(o,s),c={windowSize:l,inSize:o,batchSize:i,numSegments:s},h=new class{constructor(e,t){this.variableNames=["x","segmentIds"];const n=e.windowSize,r=e.batchSize,a=e.inSize,s=e.numSegments,i=s*Math.ceil(a/n);this.outputShape=[r,i];const o=4*Math.floor(n/4),u=n%4,l="\n        sumValue += dot(values, segFilter);\n    ";let c="";a%n>0&&(c=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return initializationValue;\n        }\n      `);let h="";a%n>0&&(h=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return -1.0;\n        }\n      `),this.userCode=`\n      const float initializationValue = 0.0;\n\n      float getValue(int batch, int inIdx) {\n        ${c}\n        return getX(batch, inIdx);\n      }\n\n      float getSegmentIdAtIndex(int inIdx) {\n        ${h}\n        return getSegmentIds(inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = int(floor(float(outIdx) / float(\n          ${s})) * float(${n}));\n        int currentSeg = int(mod(float(outIdx), float(${s})));\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${o}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0\n          );\n\n          ${l}\n        }\n\n        int inIdx = inOffset + ${o};\n        if (${1===u}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          int inIdxSeg = int(getSegmentIdAtIndex(inIdx));\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            0,\n            0,\n            0\n          );\n\n          ${l}\n        } else if (${2===u}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n              0,\n              0\n          );\n\n          ${l}\n        } else if (${3===u}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            0\n          );\n\n          ${l}\n        }\n        setOutput(sumValue);\n      }\n    `}}(c,t),p=n.compileAndRun(h,[e,r],a);if(u.push(p),p.shape[1]===s)return p;const d=ZB({backend:n,attrs:{start:0,stop:s,step:1,dtype:"float32"}}),f=eU({inputs:{x:d},backend:n,attrs:{reps:[o/l]}});u.push(d),u.push(f);return g(p,t,f,a,s)},y=gL({inputs:{x:g(f,"unsortedSegmentSum",s,m,i)},backend:n,attrs:{shape:p}});let b=y;if(null!=c){u.push(y);const e=jd.getUndoAxesPermutation(c);b=IL({inputs:{x:b},backend:n,attrs:{perm:e}})}return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),b}},jB];for(const e of fU)da(e);(cU=lU||(lU={}))[cU.float32=0]="float32",cU[cU.int32=1]="int32",cU[cU.bool=2]="bool",cU[cU.string=3]="string",cU[cU.complex64=4]="complex64",(pU=hU||(hU={}))[pU.linear=0]="linear",pU[pU.relu=1]="relu",pU[pU.relu6=2]="relu6",pU[pU.prelu=3]="prelu",pU[pU.leakyrelu=4]="leakyrelu",pU[pU.sigmoid=5]="sigmoid",pU[pU.elu=6]="elu";var mU={kernelName:ra,backendName:"wasm",setupFunc:function(e){dU=e.wasm.cwrap(ra,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{a:a,b:s,bias:i,preluActivationWeights:o}=t;if("float32"!==a.dtype||"float32"!==s.dtype)throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");const{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:h}=r,p=n.dataIdMap.get(a.dataId).id,d=n.dataIdMap.get(s.dataId).id;let f=0;if(null!=i){const e=n.dataIdMap.get(i.dataId);if(1!==e.shape.length)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${e.shape.length}.`);f=e.id}const m=null==o?0:n.dataIdMap.get(o.dataId).id,g=hU[c];if(null==g)throw new Error(c+" activation not yet supported for FusedConv2D in the wasm backend.");const y=u?a.shape[2]:a.shape[1],b=l?s.shape[1]:s.shape[2],v=Yi.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)),x=n.makeOutput([...v,y,b],a.dtype),w=n.dataIdMap.get(x.dataId).id,k=new Uint8Array(new Int32Array(a.shape).buffer),S=new Uint8Array(new Int32Array(s.shape).buffer);return dU(p,k,a.shape.length,d,S,s.shape.length,u,l,g,f,m,h||0,w),x}};function gU(e,t){let n;return{kernelName:e,backendName:"wasm",setupFunc:function(t){n=t.wasm.cwrap(e,null,["number","number","number"])},kernelFunc:function(e){const{backend:r,inputs:{x:a}}=e,s=r.dataIdMap.get(a.dataId).id,i=r.makeOutput(a.shape,t||a.dtype),o=r.dataIdMap.get(i.dataId).id;return 0===va.sizeFromShape(i.shape)||n(s,lU[a.dtype],o),i}}}var yU=gU(et);function bU(e,t,n){let r;return{kernelName:e,backendName:"wasm",setupFunc:function(t){r=t.wasm.cwrap(e,null,["number","array","number","number","array","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:a}=e,{a:s,b:i}=a,o=t.dataIdMap.get(s.dataId).id,u=t.dataIdMap.get(i.dataId).id,l=null!=n?n:s.dtype,c=jd.assertAndGetBroadcastShape(s.shape,i.shape),h=t.makeOutput(c,l);if(0===va.sizeFromShape(c))return h;const p=new Uint8Array(new Int32Array(s.shape).buffer),d=new Uint8Array(new Int32Array(i.shape).buffer),f=t.dataIdMap.get(h.dataId).id;return r(o,p,s.shape.length,u,d,i.shape.length,lU[s.dtype],f),h}}}var vU,xU=bU(rt);var wU={kernelName:at,backendName:"wasm",setupFunc:function(e){vU=e.wasm.cwrap(at,null,["array","number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n}=e,r=n.makeOutput(t[0].shape,t[0].dtype);if(0===va.sizeFromShape(r.shape))return r;const a=t.map(e=>n.dataIdMap.get(e.dataId).id),s=new Uint8Array(new Int32Array(a).buffer),i=n.dataIdMap.get(r.dataId).id;return vU(s,a.length,lU[r.dtype],i),r}};function kU(e){const{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype),a=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(r).set(a),r}var SU,IU={kernelName:fn,backendName:"wasm",kernelFunc:kU};function AU(e){const{inputs:t,backend:n,attrs:r}=e,[a,s]=function(e,t){const n=[],r=[];for(let a=0;a<e.length;++a)1!==e[a]&&n.push(e[a]),1!==e[t[a]]&&r.push(t[a]);for(let e=0;e<r.length;++e){let t=-1;for(let n=0;n<r.length;++n)r[n]>=e&&(-1===t||r[t]>r[n])&&(t=n);r[t]=e}return[n,r]}(t.x.shape,r.perm);let i=!0;for(let e=0;e<s.length;e++)s[e]!==e&&(i=!1);const o=function(e,t){const n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];return n}(t.x.shape,r.perm),u={dataId:t.x.dataId,shape:a,dtype:t.x.dtype};if(i){const e=kU({inputs:t,backend:n});return e.shape=o,e}const l=n.makeOutput(o,u.dtype),c=n.dataIdMap.get(u.dataId).id,h=n.dataIdMap.get(l.dataId).id,p=new Uint8Array(new Int32Array(s).buffer),d=new Uint8Array(new Int32Array(u.shape).buffer);return SU(c,d,u.shape.length,lU[u.dtype],h,p,s.length),l}var EU,NU={kernelName:Yr,backendName:"wasm",kernelFunc:AU,setupFunc:function(e){SU=e.wasm.cwrap(Yr,null,["number","array","number","number","number","array","number"])}};function CU(e,t,n){const r=e.shape,a=e.shape.length,s=va.parseAxisParam(t,r);let i=s;const o=jd.getAxesPermutation(i,a);let u=null,l=!1;if(null!=o){const t=new Array(a);for(let e=0;e<t.length;e++)t[e]=r[o[e]];i=jd.getInnerMostAxes(i.length,a),u=AU({inputs:{x:e},attrs:{perm:o},backend:n});const s=n.dataIdMap.get(e.dataId).id;n.dataIdMap.get(u.dataId).id!==s&&(l=!0)}return{transposed:u,originalAxes:s,axes:i,inputWasTransposed:l}}var TU,RU={kernelName:st,backendName:"wasm",setupFunc:function(e){EU=e.wasm.cwrap(st,null,["number, number, number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n;let o=t.dataIdMap.get(i.dataId).id,u=i;const{transposed:l,axes:c,originalAxes:h,inputWasTransposed:p}=CU(i,a,t);if(p){u=l,o=t.dataIdMap.get(l.dataId).id}const d=u.shape.length;jd.assertAxesAreInnerMostDims("all",c,d);const[f,m]=jd.computeOutAndReduceShapes(u.shape,c),g=va.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(0!==va.sizeFromShape(u.shape)){const e=t.dataIdMap.get(y.dataId).id;EU(o,g,e)}if(p&&t.disposeData(l.dataId),s){const e=jd.expandShapeToKeepDim(y.shape,h);y.shape=e}return y}};var _U,FU={kernelName:it,backendName:"wasm",setupFunc:function(e){TU=e.wasm.cwrap(it,null,["number, number, number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n;let o=t.dataIdMap.get(i.dataId).id,u=i;const{transposed:l,axes:c,originalAxes:h,inputWasTransposed:p}=CU(i,a,t);if(p){u=l,o=t.dataIdMap.get(l.dataId).id}const d=u.shape.length;jd.assertAxesAreInnerMostDims("any",c,d);const[f,m]=jd.computeOutAndReduceShapes(u.shape,c),g=va.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(0!==va.sizeFromShape(u.shape)){const e=t.dataIdMap.get(y.dataId).id;TU(o,g,e)}if(p&&t.disposeData(l.dataId),s){const e=jd.expandShapeToKeepDim(y.shape,h);y.shape=e}return y}};var MU,DU={kernelName:ot,backendName:"wasm",kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a}=r,{x:s}=n,i=t.dataIdMap.get(s.dataId).id;let o=i,u=s;const{transposed:l,axes:c,inputWasTransposed:h}=CU(s,a,t);if(h){const e=t.dataIdMap.get(l.dataId).id;e!==i&&(u=l,o=e)}const p=u.shape.slice(0,-1),d=t.makeOutput(p,"int32"),f=t.dataIdMap.get(d.dataId).id,m=va.sizeFromShape(d.shape),g=u.shape[c[0]];return _U(o,lU[u.dtype],m,g,f),h&&t.disposeData(l.dataId),d},setupFunc:function(e){_U=e.wasm.cwrap(ot,null,["number","number","number","number","number"])}};var OU={kernelName:ft,backendName:"wasm",setupFunc:function(e){MU=e.wasm.cwrap(ft,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,a=t.x,s=r.dataIdMap.get(a.dataId).id,{filterSize:i,strides:o,pad:u,dimRoundingMode:l}=n,c=jd.computePool2DInfo(a.shape,i,o,1,u,l),h=c.filterHeight,p=c.filterWidth,d=c.padInfo.top,f=c.padInfo.right,m=c.padInfo.bottom,g=c.padInfo.left,y=c.strideHeight,b=c.strideWidth,v=c.inChannels;if("channelsLast"!==c.dataFormat)throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);if(1!==c.dilationWidth||1!==c.dilationHeight)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${c.dilationHeight}, ${c.dilationWidth}].`);const x=r.makeOutput(c.outShape,"float32"),w=r.dataIdMap.get(x.dataId).id;return MU(s,a.shape[0],a.shape[1],a.shape[2],h,p,d,f,m,g,y,b,v,w),x}};function LU(e){const{inputs:t,attrs:n}=e,{x:r}=t,{shape:a}=n,s=va.sizeFromShape(r.shape),i=va.inferFromImplicitShape(a,s);return va.assert(s===va.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${r.shape}. New shape and old shape must have the same number of elements.`),e.backend.incRef(r.dataId),{dataId:r.dataId,shape:i,dtype:r.dtype}}var PU,$U={kernelName:lr,backendName:"wasm",kernelFunc:LU};var BU={kernelName:bt,backendName:"wasm",setupFunc:function(e){PU=e.wasm.cwrap(bt,null,["number","array","number","number","array","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{a:a,b:s}=t,{transposeA:i,transposeB:o}=r;if("float32"!==a.dtype||"float32"!==s.dtype)throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");const u=a.shape.length,l=s.shape.length,c=i?a.shape[u-2]:a.shape[u-1],h=o?s.shape[l-1]:s.shape[l-2],p=i?a.shape[u-1]:a.shape[u-2],d=o?s.shape[l-2]:s.shape[l-1],f=a.shape.slice(0,-2),m=s.shape.slice(0,-2),g=va.sizeFromShape(f),y=va.sizeFromShape(m),b=Yi.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)).concat([p,d]);va.assert(c===h,()=>`Error in matMul: inner shapes (${c}) and (${h}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);const v=o?[y,d,h]:[y,h,d],x=LU({inputs:{x:a},backend:n,attrs:{shape:i?[g,c,p]:[g,p,c]}}),w=LU({inputs:{x:s},backend:n,attrs:{shape:v}}),k=n.dataIdMap.get(x.dataId).id,S=n.dataIdMap.get(w.dataId).id,I=i?x.shape[2]:x.shape[1],A=o?w.shape[1]:w.shape[2],E=Math.max(g,y),N=n.makeOutput([E,I,A],x.dtype),C=n.dataIdMap.get(N.dataId).id,T=new Uint8Array(new Int32Array(x.shape).buffer),R=new Uint8Array(new Int32Array(w.shape).buffer);return PU(k,T,x.shape.length,S,R,w.shape.length,i,o,C),n.disposeData(x.dataId),n.disposeData(w.dataId),N.shape=b,N}};function zU(e){const{inputs:{x:t},attrs:{begin:n,size:r},backend:a}=e,[s,i]=fo.parseSliceParams(t,n,r),o=fo.isSliceContinous(t.shape,s,i),u=a.readSync(t.dataId),l=a.makeOutput(i,t.dtype),c=va.computeStrides(t.shape),h=a.dataIdMap.get(l.dataId);if(o){const e=fo.computeFlatOffset(s,c);if("string"===t.dtype)h.stringBytes=u.slice(e,e+va.sizeFromShape(i));else{a.typedArrayFromHeap(l).set(u.subarray(e,e+va.sizeFromShape(i)))}return l}if("string"===t.dtype){const e=vC(u,s,i,t.shape,t.dtype);return h.stringBytes=e,l}const p=a.typedArrayFromHeap(l),d=t.shape.length;if(2===d)!function(e,t,n,r,a){let s=0;const i=r[0],o=r[1],u=i+a[0];for(let r=i;r<u;r++){const i=r*t+o;n.set(e.subarray(i,i+a[1]),s),s+=a[1]}}(u,c[0],p,s,i);else if(3===d)!function(e,t,n,r,a,s){let i=0;const o=a[0],u=a[1],l=a[2],c=o+s[0],h=u+s[1];for(let a=o;a<c;a++)for(let o=u;o<h;o++){const u=a*t+o*n+l;r.set(e.subarray(u,u+s[2]),i),i+=s[2]}}(u,c[0],c[1],p,s,i);else if(4===d)!function(e,t,n,r,a,s,i){let o=0;const u=s[0],l=s[1],c=s[2],h=u+i[0],p=l+i[1],d=c+i[2],f=s[3];for(let s=u;s<h;s++)for(let u=l;u<p;u++)for(let l=c;l<d;l++){const c=s*t+u*n+l*r+f;a.set(e.subarray(c,c+i[3]),o),o+=i[3]}}(u,c[0],c[1],c[2],p,s,i);else{const e=vC(u,s,i,t.shape,t.dtype);p.set(e)}return l}var UU={kernelName:wr,backendName:"wasm",kernelFunc:zU};var WU={kernelName:vt,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,crops:i}=r,o=s.reduce((e,t)=>e*t),u=jd.getReshaped(a.shape,s,o),l=jd.getPermuted(u.length,s.length),c=jd.getReshapedPermuted(a.shape,s,o),h=jd.getSliceBeginCoords(i,s.length),p=jd.getSliceSize(c,i,s.length),d=LU({inputs:{x:a},backend:n,attrs:{shape:u}}),f=AU({inputs:{x:d},backend:n,attrs:{perm:l}}),m=zU({inputs:{x:LU({inputs:{x:f},backend:n,attrs:{shape:c}})},backend:n,attrs:{begin:h,size:p}});return n.disposeData(d.dataId),n.disposeData(f.dataId),n.disposeData(d.dataId),m}};function VU(e){const{inputs:{x:t},attrs:{dtype:n},backend:r}=e,a=r.makeOutput(t.shape,n),s=r.typedArrayFromHeap(t);return r.typedArrayFromHeap(a).set(s),a}var jU,GU={kernelName:St,backendName:"wasm",kernelFunc:VU},HU=gU(It);var qU={kernelName:At,backendName:"wasm",setupFunc:function(e){jU=e.wasm.cwrap(At,null,["number","number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{clipValueMin:s,clipValueMax:i}=r,o=n.dataIdMap.get(a.dataId).id,u=n.makeOutput(a.shape,a.dtype),l=n.dataIdMap.get(u.dataId).id;return jU(o,s,i,l),u}};function KU(e){const{inputs:t,backend:n}=e,r=va.parseAxisParam(e.attrs.axis,t[0].shape)[0];let a=jd.computeOutShape(t.map(e=>e.shape),r);const s=t.filter(e=>va.sizeFromShape(e.shape)>0);if(1===s.length)return kU({inputs:{x:s[0]},backend:n});const i=n.makeOutput(a,t[0].dtype);if(0===va.sizeFromShape(a))return i;const o=s.map(e=>e.shape);if(jd.assertParamsConsistent(o,r),"string"===s[0].dtype){const e=s.map(e=>{const t=va.sizeFromShape(e.shape.slice(r));return LU({inputs:{x:e},backend:n,attrs:{shape:[-1,t]}})}),o=e.map(e=>({vals:n.readSync(e.dataId),shape:e.shape}));a=jd.computeOutShape(e.map(e=>e.shape),1);const u=1===e[0].shape[0],l=fN(o,a,t[0].dtype,u),c=jd.computeOutShape(s.map(e=>e.shape),r);i.shape=c;return n.dataIdMap.get(i.dataId).stringBytes=jd.fromStringArrayToUint8(l),e.forEach(e=>n.disposeData(e.dataId)),i}const u=va.sizeFromShape(s[0].shape.slice(0,r));let l=0;const c=s.map(e=>{const t=va.sizeFromShape(e.shape.slice(r));return l+=t,t}),h=s.map(e=>n.typedArrayFromHeap(e)),p=n.typedArrayFromHeap(i);for(let e=0;e<u;e++){let t=e*l;for(let n=0;n<h.length;n++){const r=c[n],a=e*r,s=h[n].subarray(a,a+r);p.set(s,t),t+=r}}return i}var YU,XU={kernelName:Ct,backendName:"wasm",kernelFunc:KU};var QU,ZU={kernelName:Tt,backendName:"wasm",setupFunc:function(e){YU=e.wasm.cwrap(Tt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s}=t,i=r.dataIdMap.get(a.dataId).id,o=r.dataIdMap.get(s.dataId).id,{strides:u,dilations:l,pad:c,dimRoundingMode:h,dataFormat:p}=n,d=jd.convertConv2DDataFormat(p),f=jd.computeConv2DInfo(a.shape,s.shape,u,l,c,h,!1,d),m=f.filterHeight,g=f.filterWidth,y=f.padInfo.top,b=f.padInfo.right,v=f.padInfo.bottom,x=f.padInfo.left,w=f.dilationHeight,k=f.dilationWidth,S=f.strideHeight,I=f.strideWidth,A=f.inChannels,E=f.outChannels,N="SAME"===f.padInfo.type?1:0;if("channelsLast"!==f.dataFormat)throw new Error(`wasm backend Conv2D does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);const C=r.makeOutput(f.outShape,"float32"),T=r.dataIdMap.get(C.dataId).id;return YU(i,a.shape[0],a.shape[1],a.shape[2],o,m,g,y,b,v,x,N,w,k,S,I,A,E,T),C}};var JU,eW,tW,nW={kernelName:_t,backendName:"wasm",setupFunc:function(e){QU=e.wasm.cwrap(_t,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{dy:a,filter:s}=n,{strides:i,pad:o,dataFormat:u,dimRoundingMode:l,inputShape:c}=r,h=jd.convertConv2DDataFormat(u),p=jd.computeConv2DInfo(c,s.shape,i,1,o,l,!1,h),{batchSize:d,filterHeight:f,filterWidth:m,inChannels:g,inHeight:y,inWidth:b,outChannels:v,outHeight:x,outWidth:w,strideHeight:k,strideWidth:S}=p,I=f-1-p.padInfo.top,A=m-1-p.padInfo.left,E="channelsLast"===p.dataFormat,N=va.computeStrides(p.inShape),C=va.computeStrides(a.shape),[T,R,_]=va.computeStrides(s.shape),F=N[0],M=E?N[1]:N[2],D=E?N[2]:1,O=E?1:N[1],L=C[0],P=E?C[1]:C[2],$=E?C[2]:1,B=E?1:C[1],z=t.makeOutput(p.inShape,"float32"),U=t.dataIdMap.get(z.dataId).id,W=t.dataIdMap.get(a.dataId).id,V=t.dataIdMap.get(s.dataId).id;return QU(W,V,d,f,m,y,b,g,x,w,v,k,S,I,A,T,R,_,F,M,D,O,L,P,$,B,U),z}},rW=gU(Ot),aW=gU(Lt);(eW=JU||(JU={}))[eW.bilinear=0]="bilinear",eW[eW.nearest=1]="nearest";var sW,iW={kernelName:$t,backendName:"wasm",setupFunc:function(e){tW=e.wasm.cwrap($t,null,["number","number","number","number","array","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{method:a,extrapolationValue:s,cropSize:i}=r,{image:o,boxes:u,boxInd:l}=n,c=u.shape[0],[h,p]=i,d=[c,h,p,o.shape[3]];let f,m=t.dataIdMap.get(o.dataId);"float32"!==o.dtype&&(f=VU({backend:t,inputs:{x:o},attrs:{dtype:"float32"}}),m=t.dataIdMap.get(f.dataId));const g=m.id,y=t.dataIdMap.get(u.dataId).id,b=t.dataIdMap.get(l.dataId).id,v=t.makeOutput(d,"float32"),x=t.dataIdMap.get(v.dataId).id,w=new Uint8Array(new Int32Array(o.shape).buffer);return tW(g,y,b,c,w,h,p,JU[a],s,x),null!=f&&t.disposeData(f.dataId),v}};var oW,uW={kernelName:Pt,backendName:"wasm",setupFunc:function(e){sW=e.wasm.cwrap(Pt,null,["number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r,u=a.shape.length;va.assert("float32"===a.dtype||"int32"===a.dtype,()=>`cumsum does not support ${a.dtype} tensors in the WASM backend`);const l=jd.getAxesPermutation([s],u);let c=a;null!==l&&(c=AU({inputs:{x:a},attrs:{perm:l},backend:n}));const h=jd.getInnerMostAxes(1,u)[0];jd.assertAxesAreInnerMostDims("cumsum",[h],u);const p=n.makeOutput(c.shape,c.dtype),d=c.shape[h],f=n.dataIdMap.get(c.dataId).id,m=n.dataIdMap.get(p.dataId).id;sW(f,i?1:0,o?1:0,d,m,lU[a.dtype]);let g=p;if(null!==l){g=AU({inputs:{x:p},attrs:{perm:jd.getUndoAxesPermutation(l)},backend:n}),n.disposeData(c.dataId),n.disposeData(p.dataId)}return g}};var lW,cW={kernelName:zt,backendName:"wasm",setupFunc:function(e){oW=e.wasm.cwrap(zt,null,["number","number","number","array","number","array","array","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{x:a}=n,{blockSize:s,dataFormat:i}=r,o=a.shape[0],u=("NHWC"===i?a.shape[1]:a.shape[2])*s,l=("NHWC"===i?a.shape[2]:a.shape[3])*s,c=("NHWC"===i?a.shape[3]:a.shape[1])/(s*s),h="NHWC"===i?[o,u,l,c]:[o,c,u,l],p=t.makeOutput(h,"float32"),d=t.dataIdMap.get(a.dataId).id,f=new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer),m=new Uint8Array(new Int32Array(h).buffer),g=new Uint8Array(new Int32Array(va.computeStrides(h)).buffer),y=t.dataIdMap.get(p.dataId).id;return oW(d,s,"NHWC"===i?1:0,f,a.shape.length-1,m,g,h.length,y),p}};var hW={kernelName:Ut,backendName:"wasm",setupFunc:function(e){lW=e.wasm.cwrap(Ut,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s}=t,i=r.dataIdMap.get(a.dataId).id,o=r.dataIdMap.get(s.dataId).id,{strides:u,dilations:l,pad:c,dimRoundingMode:h}=n,p=null==l?[1,1]:l,d=jd.computeConv2DInfo(a.shape,s.shape,u,p,c,h,!0),f=d.filterHeight,m=d.filterWidth,g=d.padInfo.top,y=d.padInfo.right,b=d.padInfo.bottom,v=d.padInfo.left,x=d.dilationHeight,w=d.dilationWidth,k=d.strideHeight,S=d.strideWidth,I=d.inChannels,A=d.outChannels,E="SAME"===d.padInfo.type?1:0;if("channelsLast"!==d.dataFormat)throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${d.dataFormat}'. Please use 'channelsLast'.`);const N=r.makeOutput(d.outShape,"float32"),C=r.dataIdMap.get(N.dataId).id;return lW(i,a.shape[0],a.shape[1],a.shape[2],o,f,m,g,y,b,v,E,x,w,k,S,I,A,C),N}},pW=gU(Xt),dW=bU(Jt,0,"bool"),fW=gU(en,"float32");function mW(e){const{inputs:t,attrs:n,backend:r}=e,{input:a}=t,{dim:s}=n,i=a.shape.length,o=a.shape.slice();let u=s;return s<0&&(va.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+s+1),o.splice(u,0,1),LU({inputs:{x:a},backend:r,attrs:{shape:o}})}var gW={kernelName:tn,backendName:"wasm",kernelFunc:mW};function yW(e){const{attrs:{shape:t,value:n,dtype:r},backend:a}=e,s=a.makeOutput(t,r);return a.typedArrayFromHeap(s).fill(n),s}var bW,vW={kernelName:an,backendName:"wasm",kernelFunc:yW};var xW,wW={kernelName:sn,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n}=e,{image:r}=t,a=n.makeOutput(r.shape,r.dtype),s=n.dataIdMap.get(r.dataId).id,i=n.dataIdMap.get(a.dataId).id,[o,u,l,c]=r.shape;return bW(s,o,u,l,c,i),a},setupFunc:function(e){bW=e.wasm.cwrap(sn,null,["number","number","number","number","number","number"])}},kW=gU(on),SW=bU(un);var IW,AW={kernelName:ln,backendName:"wasm",setupFunc:function(e){xW=e.wasm.cwrap(ln,null,["number","number","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{varianceEpsilon:a}=r,{x:s,mean:i,variance:o,offset:u,scale:l}=n,c=t.dataIdMap.get(s.dataId).id,h=t.dataIdMap.get(i.dataId).id,p=t.dataIdMap.get(o.dataId).id,d=null!=u?t.dataIdMap.get(u.dataId).id:0,f=null!=l?t.dataIdMap.get(l.dataId).id:0,m=t.makeOutput(s.shape,s.dtype);if(0===va.sizeFromShape(s.shape))return m;const g=t.dataIdMap.get(m.dataId).id;return xW(c,h,p,d,f,a,g),m}};var EW,NW={kernelName:aa,backendName:"wasm",setupFunc:function(e){IW=e.wasm.cwrap(aa,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:u,pad:l,dilations:c,dataFormat:h,dimRoundingMode:p,activation:d,leakyreluAlpha:f}=n,m=jd.computeConv2DInfo(a.shape,s.shape,u,c,l,p),g=hU[d];if(null==g)throw new Error(d+" activation not yet supported for FusedConv2D in the wasm backend.");const y=r.dataIdMap.get(a.dataId).id,b=r.dataIdMap.get(s.dataId).id,v=m.outChannels;let x=0;if(null!=i){const e=r.dataIdMap.get(i.dataId);if(1!==e.shape.length)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${e.shape.length}.`);if(e.shape[0]!==v)throw new Error(`FusedConv2D bias shape (${e.shape}) does not match the number of output channels (${v})`);x=e.id}const w=m.filterHeight,k=m.filterWidth,S=m.padInfo.top,I=m.padInfo.right,A=m.padInfo.bottom,E=m.padInfo.left,N=m.dilationHeight,C=m.dilationWidth,T=m.strideHeight,R=m.strideWidth,_=m.inChannels,F="SAME"===m.padInfo.type?1:0,M=m.batchSize,D=m.inHeight,O=m.inWidth;if("NHWC"!==h)throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${h}'. Please use 'NHWC'.`);const L=r.makeOutput(m.outShape,"float32"),P=r.dataIdMap.get(L.dataId).id,$=null==o?0:r.dataIdMap.get(o.dataId).id;return IW(y,M,D,O,b,w,k,x,S,I,A,E,F,N,C,T,R,_,v,g,$,f||0,P),L}};var CW,TW={kernelName:sa,backendName:"wasm",setupFunc:function(e){EW=e.wasm.cwrap(sa,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:u,pad:l,dilations:c,dataFormat:h,dimRoundingMode:p,activation:d,leakyreluAlpha:f}=n,m=jd.computeConv2DInfo(a.shape,s.shape,u,c,l,p,!0),g=hU[d];if(null==g)throw new Error(d+" activation not yet supported for FusedDepthwiseConv2D in the wasm backend.");const y=r.dataIdMap.get(a.dataId).id,b=r.dataIdMap.get(s.dataId).id,v=m.outChannels;let x=0;if(null!=i){const e=r.dataIdMap.get(i.dataId);if(1!==e.shape.length)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${e.shape.length}.`);if(e.shape[0]!==v)throw new Error(`FusedDepthwiseConv2D bias shape (${e.shape}) does not match the number of output channels (${v})`);x=e.id}const w=m.filterHeight,k=m.filterWidth,S=m.padInfo.top,I=m.padInfo.right,A=m.padInfo.bottom,E=m.padInfo.left,N=m.dilationHeight,C=m.dilationWidth,T=m.strideHeight,R=m.strideWidth,_=m.inChannels,F="SAME"===m.padInfo.type?1:0,M=m.batchSize,D=m.inHeight,O=m.inWidth;if("NHWC"!==h)throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${h}'. Please use 'NHWC'.`);const L=r.makeOutput(m.outShape,"float32"),P=r.dataIdMap.get(L.dataId).id,$=null==o?0:r.dataIdMap.get(o.dataId).id;return EW(y,M,D,O,b,w,k,x,S,I,A,E,F,N,C,T,R,_,v,g,$,f||0,P),L}};var RW,_W={kernelName:hn,backendName:"wasm",setupFunc:function(e){CW=e.wasm.cwrap(hn,null,["number","number","number","number","number","number","array","number"])},kernelFunc:function(e){const{backend:t,inputs:n}=e,{params:r,indices:a}=n,[s,i,o,u]=oo.prepareAndValidate(r,a),l=t.makeOutput(s,r.dtype);if(0===i)return l;const c=a.shape,h=c[c.length-1],p=t.dataIdMap.get(r.dataId).id,d=t.dataIdMap.get(a.dataId).id,f=new Uint8Array(new Int32Array(u).buffer),m=t.dataIdMap.get(l.dataId).id;return CW(p,lU[r.dtype],d,i,h,o,f,m),l}};var FW,MW={kernelName:cn,backendName:"wasm",setupFunc:function(e){RW=e.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{x:a,indices:s}=n,{axis:i,batchDims:o}=r,u=va.parseAxisParam(i,a.shape)[0],l=t.readSync(s.dataId),c=a.shape[u];for(let e=0;e<l.length;++e){const t=l[e];va.assert(t<=c-1&&t>=0,()=>`GatherV2: the index value ${t} is not in [0, ${c-1}]`)}const h=jd.segment_util.collectGatherOpShapeInfo(a,s,u,o),p=LU({inputs:{x:a},attrs:{shape:[h.batchSize,h.outerSize,h.dimSize,h.sliceSize]},backend:t}),d=va.sizeFromShape(s.shape),f=LU({inputs:{x:s},attrs:{shape:[h.batchSize,d/h.batchSize]},backend:t}),m=[h.batchSize,h.outerSize,d/h.batchSize,h.sliceSize],g=t.makeOutput(m,a.dtype);if(0===va.sizeFromShape(a.shape))return g;const y=p.shape.length-1,b=t.dataIdMap.get(p.dataId).id,v=t.dataIdMap.get(f.dataId).id,x=t.dataIdMap.get(g.dataId).id,w=new Uint8Array(new Int32Array(va.computeStrides(p.shape)).buffer),k=new Uint8Array(new Int32Array(va.computeStrides(m)).buffer);return RW(b,lU[a.dtype],w,y,v,h.batchSize,k,x),t.disposeData(p.dataId),t.disposeData(f.dataId),g.shape=h.outputShape,g}},DW=bU(pn,0,"bool"),OW=bU(dn,0,"bool");var LW,PW={kernelName:xn,backendName:"wasm",setupFunc:function(e){FW=e.wasm.cwrap(xn,null,["number","number","number","number"])},kernelFunc:function(e){const{inputs:{x:t},attrs:{alpha:n},backend:r}=e,a=r.dataIdMap.get(t.dataId).id,s=r.makeOutput(t.shape,"float32");if(0!==va.sizeFromShape(t.shape)){const e=r.dataIdMap.get(s.dataId).id;FW(a,lU[t.dtype],n,e)}return s}},$W=bU(wn,0,"bool"),BW=bU(kn,0,"bool"),zW=gU(In),UW=bU(En,0,"bool");var WW,VW={kernelName:Fn,backendName:"wasm",setupFunc:function(e){LW=e.wasm.cwrap(Fn,null,["number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{reductionIndices:a,keepDims:s}=r,{x:i}=n;let o=t.dataIdMap.get(i.dataId).id,u=i;const{transposed:l,axes:c,originalAxes:h,inputWasTransposed:p}=CU(i,a,t);if(p){u=l,o=t.dataIdMap.get(l.dataId).id}const d=u.shape.length;jd.assertAxesAreInnerMostDims("max",c,d);const[f,m]=jd.computeOutAndReduceShapes(u.shape,c),g=va.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(0!==va.sizeFromShape(u.shape)){const e=t.dataIdMap.get(y.dataId).id;LW(o,lU[i.dtype],g,e)}if(p&&t.disposeData(l.dataId),s){const e=jd.expandShapeToKeepDim(y.shape,h);y.shape=e}return y}},jW=bU(Mn);var GW,HW={kernelName:Dn,backendName:"wasm",setupFunc:function(e){WW=e.wasm.cwrap(Dn,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,a=t.x,s=r.dataIdMap.get(a.dataId).id;va.assert("float32"===a.dtype,()=>`Error in MaxPool: only float32 input is supported. Got ${a.dtype}.`);const{filterSize:i,strides:o,pad:u,dimRoundingMode:l}=n,c=jd.computePool2DInfo(a.shape,i,o,1,u,l),h=c.filterHeight,p=c.filterWidth,d=c.padInfo.top,f=c.padInfo.right,m=c.padInfo.bottom,g=c.padInfo.left,y=c.dilationHeight,b=c.dilationWidth,v=c.strideHeight,x=c.strideWidth,w=c.inChannels,k=c.outChannels;if("channelsLast"!==c.dataFormat)throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);const S=r.makeOutput(c.outShape,"float32"),I=r.dataIdMap.get(S.dataId).id;return WW(s,a.shape[0],a.shape[1],a.shape[2],h,p,d,f,m,g,y,b,v,x,w,k,I),S}};var qW,KW={kernelName:Bn,backendName:"wasm",setupFunc:function(e){GW=e.wasm.cwrap(Bn,null,["number, number, number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id;let u=o,l=i;const{transposed:c,axes:h,originalAxes:p,inputWasTransposed:d}=CU(i,a,t);let f=h;if(d){const e=t.dataIdMap.get(c.dataId).id;e!==o&&(l=c,u=e,f=jd.getInnerMostAxes(f.length,l.shape.length))}jd.assertAxesAreInnerMostDims("mean",f,l.shape.length);const[m,g]=jd.computeOutAndReduceShapes(l.shape,f),y=va.sizeFromShape(g);let b=l;"float32"!==l.dtype&&(b=VU({backend:t,inputs:{x:l},attrs:{dtype:"float32"}}),u=t.dataIdMap.get(b.dataId).id);const v=t.makeOutput(m,"float32");if(0!==va.sizeFromShape(l.shape)){const e=t.dataIdMap.get(v.dataId).id;GW(u,y,e)}if(d&&t.disposeData(c.dataId),s){const e=jd.expandShapeToKeepDim(v.shape,p);v.shape=e}return"float32"!==l.dtype&&t.disposeData(b.dataId),v}};var YW,XW,QW,ZW={kernelName:zn,backendName:"wasm",setupFunc:function(e){qW=e.wasm.cwrap(zn,null,["number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id;let u=o,l=i;const{transposed:c,axes:h,originalAxes:p,inputWasTransposed:d}=CU(i,a,t);if(d){const e=t.dataIdMap.get(c.dataId).id;e!==o&&(l=c,u=e)}const f=l.shape.length;jd.assertAxesAreInnerMostDims("min",h,f);const[m,g]=jd.computeOutAndReduceShapes(l.shape,h),y=va.sizeFromShape(g),b=t.makeOutput(m,l.dtype);if(0!==va.sizeFromShape(l.shape)){const e=t.dataIdMap.get(b.dataId).id;qW(u,lU[i.dtype],y,e)}if(d&&t.disposeData(c.dataId),s){const e=jd.expandShapeToKeepDim(b.shape,p);b.shape=e}return b}},JW=bU(Un);(XW=YW||(YW={}))[XW.reflect=0]="reflect",XW[XW.symmetric=1]="symmetric";var eV,tV={kernelName:Wn,backendName:"wasm",kernelFunc:function(e){const{inputs:{x:t},backend:n,attrs:{paddings:r,mode:a}}=e,s=r.map((e,n)=>e[0]+t.shape[n]+e[1]),i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),u=n.dataIdMap.get(o.dataId).id,l=new Uint8Array(new Int32Array(t.shape).buffer),c=r.map(e=>e[0]),h=r.map(e=>e[1]),p=new Uint8Array(new Int32Array(c).buffer),d=new Uint8Array(new Int32Array(h).buffer);return QW(i,l,t.shape.length,lU[t.dtype],p,d,YW[a],u),o},setupFunc:function(e){QW=e.wasm.cwrap(Wn,null,["number","array","number","number","array","array","number","number"])}},nV=bU(Gn),rV=gU(Hn);function aV(e,t){const n=new Int32Array(e.wasm.HEAPU8.buffer,t,4),r=n[0],a=n[1],s=n[2],i=n[3];return e.wasm._free(t),{pSelectedIndices:r,selectedSize:a,pSelectedScores:s,pValidOutputs:i}}var sV,iV={kernelName:Kn,backendName:"wasm",setupFunc:function(e){eV=e.wasm.cwrap(Kn,"number",["number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i}=r,{boxes:o,scores:u}=n,l=t.dataIdMap.get(o.dataId).id,c=t.dataIdMap.get(u.dataId).id,h=eV(l,c,s,a,i),{pSelectedIndices:p,selectedSize:d,pSelectedScores:f,pValidOutputs:m}=aV(t,h);return t.wasm._free(f),t.wasm._free(m),t.makeOutput([d],"int32",p)}};var oV,uV={kernelName:Yn,backendName:"wasm",setupFunc:function(e){sV=e.wasm.cwrap(Yn,"number",["number","number","number","number","number","bool"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:o}=r,{boxes:u,scores:l}=n,c=t.dataIdMap.get(u.dataId).id,h=t.dataIdMap.get(l.dataId).id,p=sV(c,h,s,a,i,o),{pSelectedIndices:d,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=aV(t,p);return t.wasm._free(m),[t.makeOutput([f],"int32",d),t.makeOutput([],"int32",g)]}};var lV,cV={kernelName:Xn,backendName:"wasm",setupFunc:function(e){oV=e.wasm.cwrap(Xn,"number",["number","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i,softNmsSigma:o}=r,{boxes:u,scores:l}=n,c=t.dataIdMap.get(u.dataId).id,h=t.dataIdMap.get(l.dataId).id,p=oV(c,h,s,a,i,o),{pSelectedIndices:d,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=aV(t,p);return t.wasm._free(g),[t.makeOutput([f],"int32",d),t.makeOutput([f],"float32",m)]}},hV=bU(qn,0,"bool");var pV={kernelName:Zn,backendName:"wasm",setupFunc:function(e){lV=e.wasm.cwrap(Zn,null,["number","number","number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{indices:a}=t,{depth:s,onValue:i,offValue:o}=r,u=n.makeOutput([...a.shape,s],"int32"),l=n.dataIdMap.get(u.dataId).id,c=n.dataIdMap.get(a.dataId).id;return lV(c,s,i,o,l),u}};var dV={kernelName:Qn,backendName:"wasm",kernelFunc:function(e){const{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(r).fill(1),r}};var fV,mV={kernelName:Jn,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{axis:a}=r;if(1===t.length)return mW({inputs:{input:t[0]},backend:n,attrs:{dim:a}});const s=t[0].shape,i=t[0].dtype;t.forEach(e=>{va.assertShapesMatch(s,e.shape,"All tensors passed to stack must have matching shapes"),va.assert(i===e.dtype,()=>"All tensors passed to stack must have matching dtypes")});const o=[],u=KU({inputs:t.map(e=>{const t=mW({inputs:{input:e},backend:n,attrs:{dim:a}});return o.push(t),t}),backend:n,attrs:{axis:a}});return o.forEach(e=>n.disposeData(e.dataId)),u}};var gV,yV={kernelName:er,backendName:"wasm",kernelFunc:function(e){const{inputs:{x:t},backend:n,attrs:{paddings:r,constantValue:a}}=e,s=r.map((e,n)=>e[0]+t.shape[n]+e[1]);if(0===va.sizeFromShape(t.shape))return yW({backend:n,attrs:{shape:s,value:a,dtype:t.dtype}});const i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),u=n.dataIdMap.get(o.dataId).id,l=new Uint8Array(new Int32Array(t.shape).buffer),c=r.map(e=>e[0]),h=r.map(e=>e[1]),p=new Uint8Array(new Int32Array(c).buffer),d=new Uint8Array(new Int32Array(h).buffer);return fV(i,l,t.shape.length,lU[t.dtype],p,d,a,u),o},setupFunc:function(e){fV=e.wasm.cwrap(er,null,["number","array","number","number","array","array","number","number"])}},bV=bU(nr);var vV,xV={kernelName:rr,backendName:"wasm",setupFunc:function(e){gV=e.wasm.cwrap(rr,null,["number","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n}=e,{x:r,alpha:a}=t,s=n.dataIdMap.get(r.dataId).id,i=n.dataIdMap.get(a.dataId).id;let o=s;const u=r;let l=u;"float32"!==u.dtype&&(l=VU({backend:n,inputs:{x:r},attrs:{dtype:"float32"}}),o=n.dataIdMap.get(l.dataId).id);const c=n.makeOutput(r.shape,"float32"),h=n.dataIdMap.get(c.dataId).id;return gV(o,i,h),"float32"!==u.dtype&&n.disposeData(l.dataId),c}};var wV,kV={kernelName:ar,backendName:"wasm",setupFunc:function(e){vV=e.wasm.cwrap(ar,null,["number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id;let u=o,l=i;const{transposed:c,axes:h,originalAxes:p,inputWasTransposed:d}=CU(i,a,t);let f=h;if(d){const e=t.dataIdMap.get(c.dataId).id;e!==o&&(l=c,u=e,f=jd.getInnerMostAxes(f.length,l.shape.length))}jd.assertAxesAreInnerMostDims("prod",f,l.shape.length);const[m,g]=jd.computeOutAndReduceShapes(l.shape,f),y=va.sizeFromShape(g),b=t.makeOutput(m,l.dtype);if(0!==va.sizeFromShape(l.shape)){const e=t.dataIdMap.get(b.dataId).id;vV(u,y,lU[b.dtype],e)}if(d&&t.disposeData(c.dataId),s){const e=jd.expandShapeToKeepDim(b.shape,p);b.shape=e}return b}},SV={kernelName:sr,backendName:"wasm",kernelFunc:e=>{const{backend:t,attrs:n}=e,{start:r,stop:a,step:s,dtype:i}=n,o=pC(r,a,s,i),u=t.makeOutput([o.length],i);return t.typedArrayFromHeap(u).set(o),u}},IV=bU(Kt),AV=gU(ur),EV=gU(fr);var NV,CV={kernelName:pr,backendName:"wasm",setupFunc:function(e){wV=e.wasm.cwrap(pr,null,["number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{images:a}=n,{alignCorners:s,halfPixelCenters:i,size:o}=r,[u,l]=o,[c,h,p,d]=a.shape,f=[c,u,l,d];let m,g=t.dataIdMap.get(a.dataId);"float32"!==g.dtype&&(m=VU({backend:t,inputs:{x:a},attrs:{dtype:"float32"}}),g=t.dataIdMap.get(m.dataId));const y=g.id,b=t.makeOutput(f,"float32");if(0===va.sizeFromShape(a.shape))return b;const v=t.dataIdMap.get(b.dataId).id;return wV(y,c,h,p,d,u,l,s?1:0,i?1:0,v),null!=m&&t.disposeData(m.dataId),b}};var TV,RV={kernelName:mr,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dims:s}=r,i=va.parseAxisParam(s,a.shape);if(0===a.shape.length)return kU({inputs:{x:a},backend:n});const o=n.makeOutput(a.shape,a.dtype),u=n.dataIdMap.get(a.dataId).id,l=n.dataIdMap.get(o.dataId).id,c=new Uint8Array(new Int32Array(i).buffer),h=new Uint8Array(new Int32Array(a.shape).buffer);NV(u,c,i.length,h,a.shape.length,l);const p=LU({inputs:{x:o},attrs:{shape:a.shape},backend:n});return n.disposeData(o.dataId),p},setupFunc:function(e){NV=e.wasm.cwrap(mr,null,["number","array","number","array","number","number"])}};var _V,FV={kernelName:na,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{image:a}=t,{radians:s,fillValue:i,center:o}=r,u=n.makeOutput(a.shape,a.dtype),l=n.dataIdMap.get(a.dataId).id,c=n.dataIdMap.get(u.dataId).id,[h,p,d,f]=a.shape,[m,g]=jd.getImageCenter(o,p,d),y="number"==typeof i?[i,i,i,0===i?0:255]:[...i,255],b=new Uint8Array(new Int32Array(y).buffer);return TV(l,h,p,d,f,s,m,g,b,y.length,c),u},setupFunc:function(e){TV=e.wasm.cwrap(na,null,["number","number","number","number","number","number","number","number","array","number","number"])}},MV=gU(gr),DV=gU(yr);var OV,LV={kernelName:br,backendName:"wasm",setupFunc:function(e){_V=e.wasm.cwrap(br,null,["number","number","number","number","number","number","array","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{indices:a,updates:s}=n,{shape:i}=r,o=t.makeOutput(i,s.dtype);if(0===va.sizeFromShape(i))return o;const{sliceRank:u,numUpdates:l,sliceSize:c,strides:h,outputSize:p}=lo.calculateShapes(s,a,i),d=t.dataIdMap.get(a.dataId).id,f=t.dataIdMap.get(s.dataId).id,m=new Uint8Array(new Int32Array(h).buffer),g=t.dataIdMap.get(o.dataId).id;return _V(d,f,lU[s.dtype],u,l,c,m,p,g),o}};var PV,$V={kernelName:vr,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n}=e,{condition:r,t:a,e:s}=t,i=n.dataIdMap.get(r.dataId).id,o=n.dataIdMap.get(a.dataId).id,u=n.dataIdMap.get(s.dataId).id,l=n.makeOutput(a.shape,a.dtype),c=n.dataIdMap.get(l.dataId).id,h=r.shape.length,p=a.shape.length,d=0===h||h>1||1===p?1:va.sizeFromShape(a.shape.slice(1));return OV(i,o,u,d,c),l},setupFunc:function(e){OV=e.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}};var BV,zV={kernelName:"Sigmoid",backendName:"wasm",setupFunc:function(e){PV=e.wasm.cwrap(Ar,null,["number","number"])},kernelFunc:function(e){const{backend:t,inputs:{x:n}}=e,r=t.dataIdMap.get(n.dataId).id,a=t.makeOutput(n.shape,n.dtype),s=t.dataIdMap.get(a.dataId).id;return 0===va.sizeFromShape(a.shape)||PV(r,s),a}},UV=gU(kr);var WV={kernelName:_r,backendName:"wasm",setupFunc:function(e){BV=e.wasm.cwrap(_r,null,["number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:{logits:n},attrs:{dim:r}}=e,a=t.dataIdMap.get(n.dataId).id,s=t.makeOutput(n.shape,n.dtype),i=t.dataIdMap.get(s.dataId).id,o=n.shape[r],u=va.sizeFromShape(n.shape)/o;return 0===va.sizeFromShape(s.shape)||BV(a,i,o,u),s}};var VV,jV={kernelName:Tr,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,paddings:i}=r,o=va.sizeFromShape(s),u=[[0,0]];u.push(...i);for(let e=1+s.length;e<a.shape.length;++e)u.push([0,0]);const l=yV.kernelFunc({inputs:{x:a},backend:n,attrs:{paddings:u,constantValue:0}}),c=jd.getReshaped(l.shape,s,o,!1),h=jd.getPermuted(c.length,s.length,!1),p=jd.getReshapedPermuted(l.shape,s,o,!1),d=LU({inputs:{x:l},backend:n,attrs:{shape:c}}),f=AU({inputs:{x:d},backend:n,attrs:{perm:h}}),m=LU({inputs:{x:f},backend:n,attrs:{shape:p}});return n.disposeData(l.dataId),n.disposeData(d.dataId),n.disposeData(f.dataId),m}};var GV,HV={kernelName:Fr,backendName:"wasm",setupFunc:function(e){VV=e.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n}=e,{indices:r,values:a,denseShape:s,defaultValue:i}=n,o=r.shape[0],u=r.shape[1],l=t.readSync(s.dataId)[0],c=[o+l,u],h=t.dataIdMap.get(r.dataId).id,p=t.dataIdMap.get(a.dataId).id,d=t.dataIdMap.get(i.dataId).id,f=t.makeOutput(c,r.dtype),m=t.dataIdMap.get(f.dataId).id,g=t.makeOutput(c.slice(0,1),a.dtype),y=t.dataIdMap.get(g.dataId).id,b=t.makeOutput([l],"bool"),v=t.dataIdMap.get(b.dataId).id,x=t.makeOutput([o],r.dtype),w=t.dataIdMap.get(x.dataId).id,k=t.makeOutput([4],"int32"),S=t.dataIdMap.get(k.dataId).id,I=VV(h,p,lU[a.dtype],o,l,u,d,m,y,v,w,S),A=t.readSync(k.dataId);let E;switch(A[0]){case 1:E=jd.getSparseFillEmptyRowsIndicesDenseShapeMismatch(A[1]);break;case 2:E=jd.getSparseFillEmptyRowsNegativeIndexErrorMessage(A[1],A[2]);break;case 3:E=jd.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(A[1],A[2],A[3]);break;default:E=""}if(t.disposeData(k.dataId),E)throw t.disposeData(f.dataId),t.disposeData(g.dataId),t.disposeData(b.dataId),t.disposeData(x.dataId),new Error(E);let N=f,C=g;return I!==c[0]&&(N=zU({inputs:{x:f},attrs:{begin:0,size:[I,u]},backend:t}),C=zU({inputs:{x:g},attrs:{begin:0,size:I},backend:t}),t.disposeData(f.dataId),t.disposeData(g.dataId)),[N,C,b,x]}};var qV,KV={kernelName:Mr,backendName:"wasm",setupFunc:function(e){GV=e.wasm.cwrap(Mr,null,["number","number","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n}=e,{inputIndices:r,inputShape:a,newShape:s}=n;if(2!==r.shape.length)throw new Error("Input indices should be a matrix but received shape\n        "+r.shape);if(1!==a.shape.length)throw new Error("Input shape should be a vector but received shape\n        "+a.shape);if(1!==s.shape.length)throw new Error("Target shape should be a vector but received shape "+s.shape);const i=t.dataIdMap.get(r.dataId).id,o=t.dataIdMap.get(a.dataId).id,u=t.dataIdMap.get(s.dataId).id,l=r.shape[0],c=va.sizeFromShape(s.shape),h=t.makeOutput([l,c],r.dtype),p=t.dataIdMap.get(h.dataId).id,d=t.makeOutput([c],s.dtype),f=t.dataIdMap.get(d.dataId).id,m=t.makeOutput([3],"int32"),g=t.dataIdMap.get(m.dataId).id;GV(i,o,u,l,p,f,g);const y=t.readSync(m.dataId);let b;switch(y[0]){case 0:b=jd.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(y[1],y[2]);break;case 1:b=jd.getSparseReshapeNegativeOutputDimErrorMessage(y[1],y[2]);break;case 2:b=jd.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{const e=Array.from(t.readSync(a.dataId)),n=Array.from(t.readSync(d.dataId));b=jd.getSparseReshapeInputOutputMultipleErrorMessage(e,n);break}case 4:{const e=Array.from(t.readSync(a.dataId)),n=Array.from(t.readSync(d.dataId));b=jd.getSparseReshapeInputOutputMismatchErrorMessage(e,n);break}default:b=""}if(t.disposeData(m.dataId),b)throw t.disposeData(h.dataId),t.disposeData(d.dataId),new Error(b);return[h,d]}};function YV(e){qV=e.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function XV(e,t){const{backend:n,inputs:r}=e,{data:a,indices:s,segmentIds:i}=r,o=s.shape[0],u=n.readSync(i.dataId,o-1,o)[0],l=o>0?u+1:0;if(l<0)throw new Error(jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());const c=a.shape.slice();c[0]=l;const h=n.dataIdMap.get(a.dataId).id,p=n.dataIdMap.get(s.dataId).id,d=n.dataIdMap.get(i.dataId).id,f=n.makeOutput(c,a.dtype),m=n.dataIdMap.get(f.dataId).id,g=n.makeOutput([4],"int32"),y=n.dataIdMap.get(g.dataId).id;qV(h,lU[a.dtype],a.shape[0],p,d,m,y,t,0);const b=n.readSync(g.dataId);let v;switch(b[0]){case 0:v=jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break;case 1:v=jd.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break;case 2:v=jd.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(b[1],b[2]);break;case 3:v=jd.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(b[1],b[2],b[3]);break;default:v=""}if(n.disposeData(g.dataId),v)throw n.disposeData(f.dataId),new Error(v);return f}var QV={kernelName:Dr,backendName:"wasm",setupFunc:YV,kernelFunc:function(e){return XV(e,!0)}};var ZV={kernelName:Or,backendName:"wasm",setupFunc:YV,kernelFunc:function(e){return XV(e,!1)}};var JV,ej={kernelName:Rr,backendName:"wasm",kernelFunc:function(e){const{inputs:t,attrs:n,backend:r}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=n,o=va.parseAxisParam(i,a.shape)[0],u=jd.prepareSplitSize(a,s,o),l=new Array(a.shape.length).fill(0),c=a.shape.slice();return u.map(e=>{const t=[...c];t[o]=e;const n=zU({inputs:{x:a},attrs:{begin:l,size:t},backend:r});return l[o]+=e,n})}},tj=gU(Nr),nj=gU($r),rj=bU(Pr);var aj,sj={kernelName:ea,backendName:"wasm",setupFunc:function(e){JV=e.wasm.cwrap(ea,null,["number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{alpha:a}=r,{x:s}=n,i=t.dataIdMap.get(s.dataId).id,o=t.makeOutput(s.shape,s.dtype),u=t.dataIdMap.get(o.dataId).id;return JV(i,a,lU[s.dtype],u),o}};var ij,oj={kernelName:Br,backendName:"wasm",setupFunc:function(e){aj=e.wasm.cwrap(Br,null,["number","array","number","array","array","array","array","array","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{x:a}=n,{begin:s,end:i,strides:o,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:h,shrinkAxisMask:p}=r,{finalShapeSparse:d,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:b,end:v,strides:x}=fo.sliceInfo(a.shape,s,i,o,u,l,c,h,p);let w;if(m)w=LU({inputs:{x:a},backend:t,attrs:{shape:f}});else if(g||y){va.assert(a.shape.length>=1,()=>"Input must have rank at least 1, got: "+a.shape.length);const e=fo.computeOutShape(b,v,x),n=zU({inputs:{x:a},backend:t,attrs:{begin:b,size:e}});w=LU({inputs:{x:n},backend:t,attrs:{shape:f}}),t.disposeData(n.dataId)}else{const e=t.makeOutput(d,"float32"),n=t.dataIdMap.get(a.dataId).id,r=new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer),s=new Uint8Array(new Int32Array(b).buffer),i=new Uint8Array(new Int32Array(v).buffer),o=new Uint8Array(new Int32Array(x).buffer),u=new Uint8Array(new Int32Array(d).buffer),l=new Uint8Array(new Int32Array(va.computeStrides(d)).buffer),c=t.dataIdMap.get(e.dataId).id;aj(n,r,a.shape.length,s,i,o,u,l,d.length,c),w=LU({inputs:{x:e},backend:t,attrs:{shape:f}}),t.disposeData(e.dataId)}return w}},uj=bU(Vr);var lj,cj,hj={kernelName:Cr,backendName:"wasm",setupFunc:function(e){ij=e.wasm.cwrap(Cr,null,["number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id;let u=o,l=i;const{transposed:c,axes:h,originalAxes:p,inputWasTransposed:d}=CU(i,a,t);let f=h;if(d){const e=t.dataIdMap.get(c.dataId).id;e!==o&&(l=c,u=e,f=jd.getInnerMostAxes(f.length,l.shape.length))}jd.assertAxesAreInnerMostDims("sum",f,l.shape.length);const[m,g]=jd.computeOutAndReduceShapes(l.shape,f),y=va.sizeFromShape(g),b=t.makeOutput(m,l.dtype);if(0!==va.sizeFromShape(l.shape)){const e=t.dataIdMap.get(b.dataId).id;ij(u,y,lU[b.dtype],e)}if(d&&t.disposeData(c.dataId),s){const e=jd.expandShapeToKeepDim(b.shape,p);b.shape=e}return b}},pj=gU(jr),dj=gU(Gr);var fj;var mj=[mU,yU,xU,wU,RU,FU,DU,OU,BU,WU,GU,HU,qU,XU,ZU,nW,rW,aW,iW,uW,cW,hW,pW,dW,fW,gW,vW,wW,kW,SW,AW,NW,TW,_W,MW,DW,OW,IU,PW,$W,BW,zW,UW,VW,jW,HW,KW,ZW,JW,tV,nV,rV,iV,uV,cV,hV,pV,dV,mV,yV,bV,xV,kV,SV,IV,AV,EV,$U,CV,RV,FV,MV,DV,LV,$V,zV,UV,UU,WV,jV,HV,KV,QV,ZV,ej,tj,nj,rj,sj,oj,uj,hj,pj,dj,{kernelName:Hr,backendName:"wasm",setupFunc:function(e){lj=e.wasm.cwrap(Hr,null,["number","array","number","array","number","number"])},kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{x:a}=t,s=n.dataIdMap.get(a.dataId).id,{reps:i}=r,o=new Array(a.shape.length);for(let e=0;e<o.length;e++)o[e]=a.shape[e]*i[e];const u=new Uint8Array(new Int32Array(a.shape).buffer),l=new Uint8Array(new Int32Array(o).buffer),c=n.makeOutput(o,a.dtype),h=n.dataIdMap.get(c.dataId).id;return lj(s,u,a.shape.length,l,o.length,lU[c.dtype],h),c}},{kernelName:qr,backendName:"wasm",setupFunc:function(e){cj=e.wasm.cwrap(qr,null,["number","array","number","number","number","bool","number","number"])},kernelFunc:({inputs:e,backend:t,attrs:n})=>{const{x:r}=e,{k:a,sorted:s}=n,i=t.dataIdMap.get(r.dataId).id,o=new Uint8Array(new Int32Array(r.shape).buffer),u=r.shape.slice();u[u.length-1]=a;const l=t.makeOutput(u,r.dtype),c=t.dataIdMap.get(l.dataId).id,h=t.makeOutput(u,"int32"),p=t.dataIdMap.get(h.dataId).id;return cj(i,o,r.shape.length,lU[r.dtype],a,s,c,p),[l,h]}},{kernelName:Kr,backendName:"wasm",setupFunc:function(e){fj=e.wasm.cwrap(Kr,null,["number","number","bool","number","number","number","number","number","number","array","number","number","number","number","number"])},kernelFunc:function(e){const{backend:t,inputs:n,attrs:r}=e,{image:a,transforms:s}=n,{interpolation:i,fillMode:o,fillValue:u,outputShape:l}=r,[c,h,p,d]=a.shape,[f,m]=null!=l?l:[h,p],g=[c,f,m,d],y=new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer),b=t.makeOutput(g,a.dtype),v=t.dataIdMap.get(b.dataId).id,x=t.dataIdMap.get(a.dataId).id,w=t.dataIdMap.get(s.dataId).id,k="nearest"===i?1:2;let S;switch(o){case"constant":S=1;break;case"reflect":S=2;break;case"wrap":S=3;break;case"nearest":S=4;break;default:S=1}return fj(x,w,s.shape[0]>1,c,f,m,d,p,h,y,a.shape.length-1,k,S,u,v),b}},NU,{kernelName:Qr,backendName:"wasm",kernelFunc:function(e){const{inputs:t,backend:n,attrs:r}=e,{value:a}=t;let{axis:s}=r;s<0&&(s+=a.shape.length);const i=a.shape[s],o=a.shape.length,u=new Array(o-1);let l=0;for(let e=0;e<o;e++)e!==s&&(u[l++]=a.shape[e]);const c=new Array(i),h=new Array(o).fill(0),p=a.shape.slice();p[s]=1;for(let e=0;e<c.length;e++)h[s]=e,c[e]=zU({inputs:{x:a},attrs:{begin:h,size:p},backend:n});return c.map(({dataId:e,dtype:t})=>({dataId:e,dtype:t,shape:u}))}},{kernelName:Jr,backendName:"wasm",kernelFunc:function(e){const{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(r).fill(0),r}}];for(const e of mj)da(e);var gj=Ye();gj.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))),gj.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(gj.get("IS_NODE"))return!1;try{return(new MessageChannel).port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(e){return!1}});var yj=k(Y()),bj=k(X()),vj=class extends Z{constructor(e){super(),this.wasm=e,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(Rj),_j=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new Q(this,eu())}write(e,t,n){const r={id:this.dataIdNextNumber++};return this.move(r,e,t,n,1),r}numDataIds(){return this.dataIdMap.numDataIds()}async time(e){const t=va.now();e();return{kernelMs:va.now()-t}}move(e,t,n,r,a){const s=this.dataIdNextNumber++;if("string"===r){const i=t;return void this.dataIdMap.set(e,{id:s,stringBytes:i,shape:n,dtype:r,memoryOffset:null,refCount:a})}const i=va.sizeFromShape(n),o=i*va.bytesPerElement(r),u=this.wasm._malloc(o);this.dataIdMap.set(e,{id:s,memoryOffset:u,shape:n,dtype:r,refCount:a}),this.wasm.tfjs.registerTensor(s,i,u),null!=t&&this.wasm.HEAPU8.set(new Uint8Array(t.buffer,t.byteOffset,o),u)}async read(e){return this.readSync(e)}readSync(e,t,n){const{memoryOffset:r,dtype:a,shape:s,stringBytes:i}=this.dataIdMap.get(e);if("string"===a)return null!=t&&0!==t||!(null==n||n>=i.length)?i.slice(t,n):i;t=t||0,n=n||va.sizeFromShape(s);const o=va.bytesPerElement(a);return function(e,t){switch(t){case"float32":return new Float32Array(e);case"int32":return new Int32Array(e);case"bool":return new Uint8Array(e);default:throw new Error("Unknown dtype "+t)}}(this.wasm.HEAPU8.slice(r+t*o,r+n*o).buffer,a)}disposeData(e,t=!1){if(this.dataIdMap.has(e)){const n=this.dataIdMap.get(e);if(n.refCount--,!t&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(e)}return!0}refCount(e){if(this.dataIdMap.has(e)){return this.dataIdMap.get(e).refCount}return 0}incRef(e){const t=this.dataIdMap.get(e);null!=t&&t.refCount++}floatPrecision(){return 32}getMemoryOffset(e){return this.dataIdMap.get(e).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(e,t,n){let r;if(null==n)r=this.write(null,e,t);else{const a=this.dataIdNextNumber++;r={id:a},this.dataIdMap.set(r,{id:a,memoryOffset:n,shape:e,dtype:t,refCount:1});const s=va.sizeFromShape(e);this.wasm.tfjs.registerTensor(a,s,n)}return{dataId:r,shape:e,dtype:t}}typedArrayFromHeap({shape:e,dtype:t,dataId:n}){const r=this.wasm.HEAPU8.buffer,{memoryOffset:a}=this.dataIdMap.get(n),s=va.sizeFromShape(e);switch(t){case"float32":return new Float32Array(r,a,s);case"int32":return new Int32Array(r,a,s);case"bool":return new Uint8Array(r,a,s);default:throw new Error("Unknown dtype "+t)}}};function xj(e,t,n){if(null!=Sj)return Sj;let r="tfjs-backend-wasm.wasm";return e&&t?r="tfjs-backend-wasm-threaded-simd.wasm":e&&(r="tfjs-backend-wasm-simd.wasm"),null!=Aj&&null!=Aj[r]?Aj[r]:n+r}async function wj(){const[e,t]=await Promise.all([Ye().getAsync("WASM_HAS_SIMD_SUPPORT"),Ye().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((n,r)=>{const a={};var s;a.locateFile=(n,r)=>{if(n.endsWith(".worker.js")){const e=new Blob(['var Module={};function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;this.alert=threadAlert;Module["instantiateWasm"]=function(info,receiveInstance){var instance=new WebAssembly.Instance(Module["wasmModule"],info);Module["wasmModule"]=null;receiveInstance(instance);return instance.exports};function moduleLoaded(){}this.onmessage=function(e){try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob==="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance;moduleLoaded()})}else if(e.data.cmd==="objectTransfer"){Module["PThread"].receiveObjectTransfer(e.data)}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.threadInfoStruct,0,0);var max=e.data.stackBase;var top=e.data.stackBase+e.data.stackSize;Module["establishStackSpace"](top,max);Module["_emscripten_tls_init"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].setThreadStatus(Module["_pthread_self"](),1);try{var result=Module["invokeEntryPoint"](e.data.start_routine,e.data.arg);if(!Module["getNoExitRuntime"]())Module["PThread"].threadExit(result)}catch(ex){if(ex==="Canceled!"){Module["PThread"].threadCancel()}else if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["getNoExitRuntime"]()){}else{Module["PThread"].threadExit(ex.status)}}else{Module["PThread"].threadExit(-2);throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["PThread"].threadCancel()}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processThreadQueue"){if(Module["_pthread_self"]()){Module["_emscripten_current_thread_process_queued_calls"]()}}else{err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){err("worker.js onmessage() captured an uncaught exception: "+ex);if(ex&&ex.stack)err(ex.stack);throw ex}};if(typeof process==="object"&&typeof process.versions==="object"&&typeof process.versions.node==="string"){self={location:{href:__filename}};var onmessage=this.onmessage;var nodeWorkerThreads=require("worker_threads");global.Worker=nodeWorkerThreads.Worker;var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",function(data){onmessage({data:data})});var nodeFS=require("fs");var nodeRead=function(filename){return nodeFS.readFileSync(filename,"utf8")};function globalEval(x){global.require=require;global.Module=Module;eval.call(null,x)}importScripts=function(f){globalEval(nodeRead(f))};postMessage=function(msg){parentPort.postMessage(msg)};if(typeof performance==="undefined"){performance={now:function(){return Date.now()}}}}'],{type:"application/javascript"});return URL.createObjectURL(e)}return n.endsWith(".wasm")?xj(e,t,null!=Ij?Ij:r):r+n},Nj&&(a.instantiateWasm=(s=xj(e,t,null!=Ij?Ij:""),(e,t)=>(va.fetch(s,{credentials:"same-origin"}).then(n=>{n.ok||e.env.a(`failed to load wasm binary file at '${s}'`),n.arrayBuffer().then(n=>{WebAssembly.instantiate(n,e).then(e=>{t(e.instance,e.module)})})}),{})));let i,o=!1;a.onAbort=()=>{if(o)return;if(Ej)return;Ej=!0;r({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"})},t&&e&&null==Sj?(a.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+yj.default.toString()],{type:"text/javascript"}),i=(0,yj.default)(a)):i=(0,bj.default)(a),i.then(e=>{o=!0,Ej=!1;e.tfjs={init:e.cwrap("init",null,[]),initWithThreadsCount:e.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:e.cwrap("get_threads_count","number",[]),registerTensor:e.cwrap("register_tensor",null,["number","number","number"]),disposeData:e.cwrap("dispose_data",null,["number"]),dispose:e.cwrap("dispose",null,[])},n({wasm:e})})})}var kj=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],Sj=null,Ij=null,Aj={},Ej=!1,Nj=!1;function Cj(e,t=!1){if(Zo("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),Ej)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");Sj=e,Nj=t}function Tj(e,t=!1){if(Ej)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if("string"==typeof e)Ij=e;else{Aj=e;const t=kj.filter(e=>null==Aj[e]);if(t.length>0)throw new Error(`There were no entries found for the following binaries: ${t.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}Nj=t}var Rj=-1,_j=-1;function Fj(e){Rj=e}function Mj(){if(-1===_j)throw new Error("WASM backend not initialized.");return _j}var Dj="3.13.0";du("wasm",async()=>{const{wasm:e}=await wj();return new vj(e)},2);var Oj={tfjs:"3.13.0","tfjs-core":"3.13.0","tfjs-data":"3.13.0","tfjs-layers":"3.13.0","tfjs-converter":"3.13.0","tfjs-backend-cpu":"3.13.0","tfjs-backend-webgl":"3.13.0","tfjs-backend-wasm":"3.13.0"};function Lj(e,t,n=!1){if(e.beginPath(),t.slice(1).forEach(({x:n,y:r},a)=>{const s=t[a];e.moveTo(s.x,s.y),e.lineTo(n,r)}),n){const n=t[t.length-1],r=t[0];if(!n||!r)return;e.moveTo(n.x,n.y),e.lineTo(r.x,r.y)}e.stroke()}p({},{AnchorPosition:()=>kG,DrawBox:()=>EG,DrawBoxOptions:()=>AG,DrawFaceLandmarks:()=>hH,DrawFaceLandmarksOptions:()=>cH,DrawTextField:()=>IG,DrawTextFieldOptions:()=>SG,drawContour:()=>Lj,drawDetections:()=>NG,drawFaceExpressions:()=>oH,drawFaceLandmarks:()=>pH});p({},{computeReshapedDimensions:()=>qj,getCenterPoint:()=>Kj,isDimensions:()=>Hj,isEven:()=>jj,isFloat:()=>Vj,isTensor:()=>$j,isTensor1D:()=>Bj,isTensor2D:()=>zj,isTensor3D:()=>Uj,isTensor4D:()=>Wj,isValidNumber:()=>Xj,isValidProbablitiy:()=>Qj,range:()=>Yj,round:()=>Gj});var Pj=class{constructor(e,t){if(!Xj(e)||!Xj(t))throw new Error("Dimensions.constructor - expected width and height to be valid numbers, instead have "+JSON.stringify({width:e,height:t}));this._width=e,this._height=t}get width(){return this._width}get height(){return this._height}reverse(){return new Pj(1/this.width,1/this.height)}};function $j(e,t){return e instanceof Ya&&e.shape.length===t}function Bj(e){return $j(e,1)}function zj(e){return $j(e,2)}function Uj(e){return $j(e,3)}function Wj(e){return $j(e,4)}function Vj(e){return e%1!=0}function jj(e){return e%2==0}function Gj(e,t=2){const n=10**t;return Math.floor(e*n)/n}function Hj(e){return e&&e.width&&e.height}function qj({width:e,height:t},n){const r=n/Math.max(t,e);return new Pj(Math.round(e*r),Math.round(t*r))}function Kj(e){return e.reduce((e,t)=>e.add(t),new Zj(0,0)).div(new Zj(e.length,e.length))}function Yj(e,t,n){return Array(e).fill(0).map((e,r)=>t+r*n)}function Xj(e){return!!e&&e!==1/0&&e!==-1/0&&!Number.isNaN(e)||0===e}function Qj(e){return Xj(e)&&e>=0&&e<=1}var Zj=class{constructor(e,t){this._x=e,this._y=t}get x(){return this._x}get y(){return this._y}add(e){return new Zj(this.x+e.x,this.y+e.y)}sub(e){return new Zj(this.x-e.x,this.y-e.y)}mul(e){return new Zj(this.x*e.x,this.y*e.y)}div(e){return new Zj(this.x/e.x,this.y/e.y)}abs(){return new Zj(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new Zj(Math.floor(this.x),Math.floor(this.y))}},Jj=class{static isRect(e){return!!e&&[e.x,e.y,e.width,e.height].every(Xj)}static assertIsValidBox(e,t,n=!1){if(!Jj.isRect(e))throw new Error(`${t} - invalid box: ${JSON.stringify(e)}, expected object with properties x, y, width, height`);if(!n&&(e.width<0||e.height<0))throw new Error(`${t} - width (${e.width}) and height (${e.height}) must be positive numbers`)}constructor(e,t=!0){const n=e||{},r=[n.left,n.top,n.right,n.bottom].every(Xj),a=[n.x,n.y,n.width,n.height].every(Xj);if(!a&&!r)throw new Error("Box.constructor - expected box to be IBoundingBox | IRect, instead have "+JSON.stringify(n));const[s,i,o,u]=a?[n.x,n.y,n.width,n.height]:[n.left,n.top,n.right-n.left,n.bottom-n.top];Jj.assertIsValidBox({x:s,y:i,width:o,height:u},"Box.constructor",t),this._x=s,this._y=i,this._width=o,this._height=u}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new Zj(this.left,this.top)}get topRight(){return new Zj(this.right,this.top)}get bottomLeft(){return new Zj(this.left,this.bottom)}get bottomRight(){return new Zj(this.right,this.bottom)}round(){const[e,t,n,r]=[this.x,this.y,this.width,this.height].map(e=>Math.round(e));return new Jj({x:e,y:t,width:n,height:r})}floor(){const[e,t,n,r]=[this.x,this.y,this.width,this.height].map(e=>Math.floor(e));return new Jj({x:e,y:t,width:n,height:r})}toSquare(){let{x:e,y:t,width:n,height:r}=this;const a=Math.abs(n-r);return n<r&&(e-=a/2,n+=a),r<n&&(t-=a/2,r+=a),new Jj({x:e,y:t,width:n,height:r})}rescale(e){const t=Hj(e)?e.width:e,n=Hj(e)?e.height:e;return new Jj({x:this.x*t,y:this.y*n,width:this.width*t,height:this.height*n})}pad(e,t){const[n,r,a,s]=[this.x-e/2,this.y-t/2,this.width+e,this.height+t];return new Jj({x:n,y:r,width:a,height:s})}clipAtImageBorders(e,t){const{x:n,y:r,right:a,bottom:s}=this,i=Math.max(n,0),o=Math.max(r,0),u=a-i,l=s-o,c=Math.min(u,e-i),h=Math.min(l,t-o);return new Jj({x:i,y:o,width:c,height:h}).floor()}shift(e,t){const{width:n,height:r}=this,a=this.x+e,s=this.y+t;return new Jj({x:a,y:s,width:n,height:r})}padAtBorders(e,t){const n=this.width+1,r=this.height+1;let a=n,s=r,i=this.left,o=this.top,u=this.right,l=this.bottom;return u>t&&(a=-u+t+n,u=t),l>e&&(s=-l+e+r,l=e),i<1&&(s=2-i,i=1),o<1&&(s=2-o,o=1),{dy:1,edy:s,dx:1,edx:a,y:o,ey:l,x:i,ex:u,w:n,h:r}}calibrate(e){return new Jj({left:this.left+e.left*this.width,top:this.top+e.top*this.height,right:this.right+e.right*this.width,bottom:this.bottom+e.bottom*this.height}).toSquare().round()}},eG=class extends Jj{constructor(e,t,n,r,a=!1){super({left:e,top:t,right:n,bottom:r},a)}},tG=class{constructor(e,t,n,r,a){this._imageDims=new Pj(a.width,a.height),this._score=e,this._classScore=t,this._className=n,this._box=new Jj(r).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new Jj(this._box).rescale(this.imageDims.reverse())}forSize(e,t){return new tG(this.score,this.classScore,this.className,this.relativeBox,{width:e,height:t})}},nG=class extends tG{constructor(e,t,n){super(e,e,"",t,n)}forSize(e,t){const{score:n,relativeBox:r,imageDims:a}=super.forSize(e,t);return new nG(n,r,a)}};function rG(e,t,n=!0){const r=Math.max(0,Math.min(e.right,t.right)-Math.max(e.left,t.left))*Math.max(0,Math.min(e.bottom,t.bottom)-Math.max(e.top,t.top));return n?r/(e.area+t.area-r):r/Math.min(e.area,t.area)}function aG(e,t){return ru(()=>{const[n,r,a]=t,s=Vl([...e.shape.slice(0,3),1],n,"float32"),i=Vl([...e.shape.slice(0,3),1],r,"float32"),o=Vl([...e.shape.slice(0,3),1],a,"float32"),u=Xu([s,i,o],3);return gc(e,u)})}function sG(e){return 1/(1+Math.exp(-e))}var iG,oG=class extends Jj{constructor(e,t,n,r,a=!1){super({x:e,y:t,width:n,height:r},a)}},uG=class{constructor(e,t,n=new Zj(0,0)){const{width:r,height:a}=t;this._imgDims=new Pj(r,a),this._shift=n,this._positions=e.map(e=>e.mul(new Zj(r,a)).add(n))}get shift(){return new Zj(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(e=>e.sub(this._shift).div(new Zj(this.imageWidth,this.imageHeight)))}forSize(e,t){return new this.constructor(this.relativePositions,{width:e,height:t})}shiftBy(e,t){return new this.constructor(this.relativePositions,this._imgDims,new Zj(e,t))}shiftByPoint(e){return this.shiftBy(e.x,e.y)}align(e,t={}){if(e){const n=e instanceof nG?e.box.floor():new Jj(e);return this.shiftBy(n.x,n.y).align(null,t)}const{useDlibAlignment:n,minBoxPadding:r}={useDlibAlignment:!1,minBoxPadding:.2,...t};return n?this.alignDlib():this.alignMinBbox(r)}alignDlib(){const e=this.getRefPointsForAlignment(),[t,n,r]=e,a=e=>r.sub(e).magnitude(),s=(a(t)+a(n))/2,i=Math.floor(s/.45),o=Kj(e),u=Math.floor(Math.max(0,o.x-.5*i)),l=Math.floor(Math.max(0,o.y-.43*i));return new oG(u,l,Math.min(i,this.imageWidth+u),Math.min(i,this.imageHeight+l))}alignMinBbox(e){const t=function(e){const t=e.map(e=>e.x),n=e.map(e=>e.y),r=t.reduce((e,t)=>t<e?t:e,1/0),a=n.reduce((e,t)=>t<e?t:e,1/0),s=t.reduce((e,t)=>e<t?t:e,0),i=n.reduce((e,t)=>e<t?t:e,0);return new eG(r,a,s,i)}(this.positions);return t.pad(t.width*e,t.height*e)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}},lG=class extends uG{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(Kj)}};function cG(e){return e.detection instanceof nG}function hG(e,t){return{...e,...{detection:t}}}function pG(){const e=window.fetch;if(!e)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D:CanvasRenderingContext2D,Image:HTMLImageElement,ImageData:ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:e,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function dG(){return"object"==typeof s&&void 0!==r&&null!=r.versions&&null!=r.versions.node}function fG(e){let t="";if(!e&&dG())try{e=h("fs")}catch(e){t=e.toString()}return{readFile:e?t=>new Promise((n,r)=>{e.readFile(t,(e,t)=>e?r(e):n(t))}):()=>{throw new Error("readFile - failed to require fs in nodejs environment with error: "+t)}}}function mG(){const e=s.Canvas||s.HTMLCanvasElement,t=s.Image||s.HTMLImageElement,n=s.Video||s.HTMLVideoElement,r=s.fetch,a=fG();return{Canvas:e||class{},CanvasRenderingContext2D:s.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:s.ImageData||class{},Video:s.HTMLVideoElement||class{},createCanvasElement:()=>{if(e)return new e;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},createImageElement:()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},createVideoElement:()=>{if(n)return new n;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},fetch:r,...a}}function gG(){return"object"==typeof window&&"undefined"!=typeof document&&"undefined"!=typeof HTMLImageElement&&"undefined"!=typeof HTMLCanvasElement&&"undefined"!=typeof HTMLVideoElement&&"undefined"!=typeof ImageData&&"undefined"!=typeof CanvasRenderingContext2D}function yG(e){iG=e}function bG(){return gG()?yG(pG()):dG()?yG(mG()):null}var vG={getEnv:function(){if(!iG)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return iG},setEnv:yG,initialize:bG,createBrowserEnv:pG,createFileSystem:fG,createNodejsEnv:mG,monkeyPatch:function(e){if(iG||bG(),!iG)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");const{Canvas:t=iG.Canvas,Image:n=iG.Image}=e;iG.Canvas=t,iG.Image=n,iG.createCanvasElement=e.createCanvasElement||(()=>new t),iG.createImageElement=e.createImageElement||(()=>new n),iG.ImageData=e.ImageData||iG.ImageData,iG.Video=e.Video||iG.Video,iG.fetch=e.fetch||iG.fetch,iG.readFile=e.readFile||iG.readFile},isBrowser:gG,isNodejs:dG};function xG(e){return vG.isNodejs()||"string"!=typeof e?e:document.getElementById(e)}function wG(e){const{Canvas:t,CanvasRenderingContext2D:n}=vG.getEnv();if(e instanceof n)return e;const r=xG(e);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");const a=r.getContext("2d");if(!a)throw new Error("resolveContext2d - canvas 2d context is null");return a}bG();var kG=(e=>(e.TOP_LEFT="TOP_LEFT",e.TOP_RIGHT="TOP_RIGHT",e.BOTTOM_LEFT="BOTTOM_LEFT",e.BOTTOM_RIGHT="BOTTOM_RIGHT",e))(kG||{}),SG=class{constructor(e={}){const{anchorPosition:t,backgroundColor:n,fontColor:r,fontSize:a,fontStyle:s,padding:i}=e;this.anchorPosition=t||"TOP_LEFT",this.backgroundColor=n||"rgba(0, 0, 0, 0.5)",this.fontColor=r||"rgba(255, 255, 255, 1)",this.fontSize=a||14,this.fontStyle=s||"Georgia",this.padding=i||4}},IG=class{constructor(e,t,n={}){this.text="string"==typeof e?[e]:e instanceof IG?e.text:e,this.anchor=t,this.options=new SG(n)}measureWidth(e){const{padding:t}=this.options;return this.text.map(t=>e.measureText(t).width).reduce((e,t)=>e<t?t:e,0)+2*t}measureHeight(){const{fontSize:e,padding:t}=this.options;return this.text.length*e+2*t}getUpperLeft(e,t){const{anchorPosition:n}=this.options,r="BOTTOM_RIGHT"===n||"TOP_RIGHT"===n,a="BOTTOM_LEFT"===n||"BOTTOM_RIGHT"===n,s=this.measureWidth(e),i=this.measureHeight(),o=r?this.anchor.x-s:this.anchor.x,u=a?this.anchor.y-i:this.anchor.y;if(t){const{width:e,height:n}=t;return{x:Math.max(Math.min(o,e-s),0),y:Math.max(Math.min(u,n-i),0)}}return{x:o,y:u}}draw(e){const t=xG(e),n=wG(t),{backgroundColor:r,fontColor:a,fontSize:s,fontStyle:i,padding:o}=this.options;n.font=`${s}px ${i}`;const u=this.measureWidth(n),l=this.measureHeight();n.fillStyle=r;const c=this.getUpperLeft(n,t);n.fillRect(c.x,c.y,u,l),n.fillStyle=a,this.text.forEach((e,t)=>{const r=o+c.x,a=o+c.y+(t+1)*s;n.fillText(e,r,a)})}},AG=class{constructor(e={}){const{boxColor:t,lineWidth:n,label:r,drawLabelOptions:a}=e;this.boxColor=t||"rgba(0, 0, 255, 1)",this.lineWidth=n||2,this.label=r;const s={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new SG({...s,...a})}},EG=class{constructor(e,t={}){this.box=new Jj(e),this.options=new AG(t)}draw(e){const t=wG(e),{boxColor:n,lineWidth:r}=this.options,{x:a,y:s,width:i,height:o}=this.box;t.strokeStyle=n,t.lineWidth=r,t.strokeRect(a,s,i,o);const{label:u}=this.options;u&&new IG([u],{x:a-r/2,y:s},this.options.drawLabelOptions).draw(e)}};function NG(e,t){(Array.isArray(t)?t:[t]).forEach(t=>{const n=t instanceof nG?t.score:cG(t)?t.detection.score:void 0,r=t instanceof nG?t.box:cG(t)?t.detection.box:new Jj(t),a=n?""+Gj(n):void 0;new EG(r,{label:a}).draw(e)})}function CG(e){const{Image:t,Video:n}=vG.getEnv();return e instanceof t&&e.complete||e instanceof n&&e.readyState>=3}function TG(e){const{Image:t,Video:n}=vG.getEnv();return e instanceof t?new Pj(e.naturalWidth,e.naturalHeight):e instanceof n?new Pj(e.videoWidth,e.videoHeight):new Pj(e.width,e.height)}function RG({width:e,height:t}){const{createCanvasElement:n}=vG.getEnv(),r=n();return r.width=e,r.height=t,r}function _G(e,t){const{ImageData:n}=vG.getEnv();if(!(e instanceof n||CG(e)))throw new Error("createCanvasFromMedia - media has not finished loading yet");const{width:r,height:a}=t||TG(e),s=RG({width:r,height:a});return e instanceof n?wG(s).putImageData(e,0,0):wG(s).drawImage(e,0,0,r,a),s}function FG(e){const{Image:t,Canvas:n,Video:r}=vG.getEnv();return e instanceof t||e instanceof n||e instanceof r}var MG=class{constructor(e,t=!1){if(this._imageTensors=[],this._canvases=[],this._treatAsBatchInput=!1,this._inputDimensions=[],this._inputSize=0,!Array.isArray(e))throw new Error("NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have "+e);this._treatAsBatchInput=t,this._batchSize=e.length,e.forEach((e,t)=>{if(Uj(e))return this._imageTensors[t]=e,void(this._inputDimensions[t]=e.shape);if(Wj(e)){const n=e.shape[0];if(1!==n)throw new Error(`NetInput - tf.Tensor4D with batchSize ${n} passed, but not supported in input array`);return this._imageTensors[t]=e,void(this._inputDimensions[t]=e.shape.slice(1))}const n=e instanceof vG.getEnv().Canvas?e:_G(e);this._canvases[t]=n,this._inputDimensions[t]=[n.height,n.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return Yj(this.batchSize,0,1).map((e,t)=>this.getReshapedInputDimensions(t))}getInput(e){return this.canvases[e]||this.imageTensors[e]}getInputDimensions(e){return this._inputDimensions[e]}getInputHeight(e){return this._inputDimensions[e][0]}getInputWidth(e){return this._inputDimensions[e][1]}getReshapedInputDimensions(e){if("number"!=typeof this.inputSize)throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");return qj({width:this.getInputWidth(e),height:this.getInputHeight(e)},this.inputSize)}toBatchTensor(e,t=!0){return this._inputSize=e,ru(()=>{const n=Yj(this.batchSize,0,1).map(n=>{const r=this.getInput(n);if(r instanceof Ya){let n=Wj(r)?r:Bl(r);return n=function(e,t=!1){return ru(()=>{const[n,r]=e.shape.slice(1);if(n===r)return e;const a=Math.abs(n-r),s=Math.round(a*(t?.5:1)),i=n>r?2:1,o=t=>{const n=e.shape.slice();return n[i]=t,Vl(n,0,"float32")},u=o(s),l=a-u.shape[i],c=[t&&l?o(l):null,e,u].filter(e=>!!e).map(e=>Ai(e,"float32"));return Xu(c,i)})}(n,t),n.shape[1]===e&&n.shape[2]===e||(n=Nd.resizeBilinear(n,[e,e],!1,!1)),n.as3D(e,e,3)}if(r instanceof vG.getEnv().Canvas)return eo.fromPixels(function(e,t,n=!1){const{Image:r,Canvas:a}=vG.getEnv();if(!(e instanceof r||e instanceof a))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return RG({width:1,height:1});const s=TG(e),i=t/Math.max(s.height,s.width),o=i*s.width,u=i*s.height,l=RG({width:t,height:t}),c=e instanceof a?e:_G(e),h=Math.abs(o-u)/2,p=n&&o<u?h:0,d=n&&u<o?h:0;return c.width>0&&c.height>0&&wG(l).drawImage(c,p,d,o,u),l}(r,e,t));throw new Error(`toBatchTensor - at batchIdx ${n}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${r}`)});return Gh(n.map(e=>Ai(e,"float32"))).as4D(this.batchSize,e,e,3)})}};async function DG(e){if(e instanceof MG)return e;const t=Array.isArray(e)?e:[e];if(!t.length)throw new Error("toNetInput - empty array passed as input");const n=t=>Array.isArray(e)?` at input index ${t}:`:"",r=t.map(xG);return r.forEach((e,r)=>{if(!FG(e)&&!Uj(e)&&!Wj(e)){if("string"==typeof t[r])throw new Error(`toNetInput -${n(r)} string passed, but could not resolve HTMLElement for element id ${t[r]}`);throw new Error(`toNetInput -${n(r)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`)}if(Wj(e)){const t=e.shape[0];if(1!==t)throw new Error(`toNetInput -${n(r)} tf.Tensor4D with batchSize ${t} passed, but not supported in input array`)}}),await Promise.all(r.map(e=>{return FG(e)&&(t=e,new Promise((e,n)=>{function r(e){e.currentTarget&&(e.currentTarget.removeEventListener("load",a),e.currentTarget.removeEventListener("error",r),n(e))}function a(t){t.currentTarget&&(t.currentTarget.removeEventListener("load",a),t.currentTarget.removeEventListener("error",r),e(t))}(t instanceof vG.getEnv().Canvas||CG(t))&&e(null),t.addEventListener("load",a),t.addEventListener("error",r)}));var t})),new MG(r,Array.isArray(e))}async function OG(e,t){const{Canvas:n}=vG.getEnv();let r=e;if(!(e instanceof n)){const t=await DG(e);if(t.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");const a=t.getInput(0);r=a instanceof n?a:await async function(e,t){const n=t||vG.getEnv().createCanvasElement(),[r,a,s]=e.shape.slice(Wj(e)?1:0),i=ru(()=>e.as3D(r,a,s).toInt());return await eo.toPixels(i,n),i.dispose(),n}(a)}const a=wG(r);return t.map(e=>e instanceof nG?e.forSize(r.width,r.height).box.floor():e).map(e=>e.clipAtImageBorders(r.width,r.height)).map(({x:e,y:t,width:n,height:r})=>{const s=RG({width:n,height:r});return n>0&&r>0&&wG(s).putImageData(a.getImageData(e,t,n,r),0,0),s})}async function LG(e,t){if(!Uj(e)&&!Wj(e))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(Wj(e)&&e.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return ru(()=>{const[n,r,a]=e.shape.slice(Wj(e)?1:0);return t.map(e=>e instanceof nG?e.forSize(r,n).box:e).map(e=>e.clipAtImageBorders(r,n)).map(({x:t,y:s,width:i,height:o})=>Dh(e.as3D(n,r,a),[s,t,0],[o,i,a]))})}async function PG(e){return(await async function(e,t){const{fetch:n}=vG.getEnv(),r=await n(e,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}(e)).json()}function $G(e,t){const n=t+"-weights_manifest.json";if(!e)return{modelBaseUri:"",manifestUri:n};if("/"===e)return{modelBaseUri:"/",manifestUri:"/"+n};const r=e.startsWith("http://")?"http://":e.startsWith("https://")?"https://":"",a=(e=e.replace(r,"")).split("/").filter(e=>e),s=e.endsWith(".json")?a[a.length-1]:n;let i=r+(e.endsWith(".json")?a.slice(0,a.length-1):a).join("/");return i=e.startsWith("/")?"/"+i:i,{modelBaseUri:i,manifestUri:"/"===i?"/"+s:`${i}/${s}`}}var BG=class{constructor(e){this._params=void 0,this._paramMappings=[],this._name=e}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(e){const{obj:t,objProp:n}=this.traversePropertyPath(e);return t[n]}reassignParamFromPath(e,t){const{obj:n,objProp:r}=this.traversePropertyPath(e);n[r].dispose(),n[r]=t}getParamList(){return this._paramMappings.map(({paramPath:e})=>({path:e,tensor:this.getParamFromPath(e)}))}getTrainableParams(){return this.getParamList().filter(e=>e.tensor instanceof Qa)}getFrozenParams(){return this.getParamList().filter(e=>!(e.tensor instanceof Qa))}variable(){this.getFrozenParams().forEach(({path:e,tensor:t})=>{this.reassignParamFromPath(e,t.variable())})}freeze(){this.getTrainableParams().forEach(({path:e,tensor:t})=>{const n=Ls(t.dataSync());t.dispose(),this.reassignParamFromPath(e,n)})}dispose(e=!0){this.getParamList().forEach(t=>{if(e&&t.tensor.isDisposed)throw new Error("param tensor has already been disposed for path "+t.path);t.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:e})=>Array.from(e.dataSync())).reduce((e,t)=>e.concat(t)))}async load(e){e instanceof Float32Array?this.extractWeights(e):await this.loadFromUri(e)}async loadFromUri(e){if(e&&"string"!=typeof e)throw new Error(this._name+".loadFromUri - expected model uri");const t=await async function(e,t){const{manifestUri:n,modelBaseUri:r}=$G(e,t),a=await PG(n);return Ci.loadWeights(a,r)}(e,this.getDefaultModelName());this.loadFromWeightMap(t)}async loadFromDisk(e){if(e&&"string"!=typeof e)throw new Error(this._name+".loadFromDisk - expected model file path");const{readFile:t}=vG.getEnv(),{manifestUri:n,modelBaseUri:r}=$G(e,this.getDefaultModelName()),a=Ci.weightsLoaderFactory(e=>Promise.all(e.map(e=>t(e).then(e=>e.buffer)))),s=JSON.parse((await t(n)).toString()),i=await a(s,r);this.loadFromWeightMap(i)}loadFromWeightMap(e){const{paramMappings:t,params:n}=this.extractParamsFromWeightMap(e);this._paramMappings=t,this._params=n}extractWeights(e){const{paramMappings:t,params:n}=this.extractParams(e);this._paramMappings=t,this._params=n}traversePropertyPath(e){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");const t=e.split("/").reduce((t,n)=>{if(!t.nextObj.hasOwnProperty(n))throw new Error(`traversePropertyPath - object does not have property ${n}, for path ${e}`);return{obj:t.nextObj,objProp:n,nextObj:t.nextObj[n]}},{nextObj:this.params}),{obj:n,objProp:r}=t;if(!(n&&r&&n[r]instanceof Ya))throw new Error("traversePropertyPath - parameter is not a tensor, for path "+e);return{obj:n,objProp:r}}};function zG(e,t,n){return ru(()=>{let r=Nh(e,t.depthwise_filter,t.pointwise_filter,n,"same");return r=gu(r,t.bias),r})}function UG(e,t,n=!1){return ru(()=>{const r=gh(n?gu(ml(e,t.conv0.filters,[2,2],"same"),t.conv0.bias):zG(e,t.conv0,[2,2])),a=zG(r,t.conv1,[1,1]),s=zG(gh(gu(r,a)),t.conv2,[1,1]);return gh(gu(r,gu(a,s)))})}function WG(e,t,n=!1,r=!0){return ru(()=>{const a=gh(n?gu(ml(e,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):zG(e,t.conv0,r?[2,2]:[1,1])),s=zG(a,t.conv1,[1,1]),i=zG(gh(gu(a,s)),t.conv2,[1,1]),o=zG(gh(gu(a,gu(s,i))),t.conv3,[1,1]);return gh(gu(a,gu(s,gu(i,o))))})}function VG(e,t,n="same",r=!1){return ru(()=>{const a=gu(ml(e,t.filters,[1,1],n),t.bias);return r?gh(a):a})}function jG(e,t){Object.keys(e).forEach(n=>{t.some(e=>e.originalPath===n)||e[n].dispose()})}function GG(e,t){return(n,r,a,s)=>{const i=Qh(e(n*r*a*a),[a,a,n,r]),o=Yh(e(r));return t.push({paramPath:s+"/filters"},{paramPath:s+"/bias"}),{filters:i,bias:o}}}function HG(e,t){return(n,r,a)=>{const s=Xh(e(n*r),[n,r]),i=Yh(e(r));return t.push({paramPath:a+"/weights"},{paramPath:a+"/bias"}),{weights:s,bias:i}}}var qG=class{constructor(e,t,n){this.depthwise_filter=e,this.pointwise_filter=t,this.bias=n}};function KG(e,t){return(n,r,a)=>{const s=Qh(e(9*n),[3,3,n,1]),i=Qh(e(n*r),[1,1,n,r]),o=Yh(e(r));return t.push({paramPath:a+"/depthwise_filter"},{paramPath:a+"/pointwise_filter"},{paramPath:a+"/bias"}),new qG(s,i,o)}}function YG(e){return t=>{const n=e(t+"/depthwise_filter",4),r=e(t+"/pointwise_filter",4),a=e(t+"/bias",1);return new qG(n,r,a)}}function XG(e,t){return(n,r,a)=>{const s=e[n];if(!$j(s,r))throw new Error(`expected weightMap[${n}] to be a Tensor${r}D, instead have ${s}`);return t.push({originalPath:n,paramPath:a||n}),s}}function QG(e){let t=e;return{extractWeights:function(e){const n=t.slice(0,e);return t=t.slice(e),n},getRemainingWeights:function(){return t}}}function ZG(e,t){const n=GG(e,t),r=KG(e,t);function a(e,t,a,s=!1){return{conv0:s?n(e,t,3,a+"/conv0"):r(e,t,a+"/conv0"),conv1:r(t,t,a+"/conv1"),conv2:r(t,t,a+"/conv2")}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:function(e,t,n,s=!1){const{conv0:i,conv1:o,conv2:u}=a(e,t,n,s);return{conv0:i,conv1:o,conv2:u,conv3:r(t,t,n+"/conv3")}}}}function JG(e){return t=>({filters:e(t+"/filters",4),bias:e(t+"/bias",1)})}function eH(e,t){const n=XG(e,t),r=JG(n),a=YG(n);return{extractDenseBlock3Params:function(e,t=!1){return{conv0:t?r(e+"/conv0"):a(e+"/conv0"),conv1:a(e+"/conv1"),conv2:a(e+"/conv2")}},extractDenseBlock4Params:function(e,t=!1){return{conv0:t?r(e+"/conv0"):a(e+"/conv0"),conv1:a(e+"/conv1"),conv2:a(e+"/conv2"),conv3:a(e+"/conv3")}}}}var tH=class extends BG{constructor(){super("FaceFeatureExtractor")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("FaceFeatureExtractor - load model before inference");return ru(()=>{let n=WG(aG(Ai(e.toBatchTensor(112,!0),"float32"),[122.782,117.001,104.298]).div(255),t.dense0,!0);return n=WG(n,t.dense1),n=WG(n,t.dense2),n=WG(n,t.dense3),n=Ku(n,[7,7],[2,2],"valid"),n})}async forward(e){return this.forwardInput(await DG(e))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(e){return function(e){const t=[],{extractDenseBlock4Params:n}=eH(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2"),dense3:n("dense3")};return jG(e,t),{params:r,paramMappings:t}}(e)}extractParams(e){return function(e){const t=[],{extractWeights:n,getRemainingWeights:r}=QG(e),{extractDenseBlock4Params:a}=ZG(n,t),s=a(3,32,"dense0",!0),i=a(32,64,"dense1"),o=a(64,128,"dense2"),u=a(128,256,"dense3");if(0!==r().length)throw new Error("weights remaing after extract: "+r().length);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o,dense3:u}}}(e)}};function nH(e,t){return ru(()=>gu(Gi(e,t.weights),t.bias))}function rH(e){const t={},n={};return Object.keys(e).forEach(r=>{(r.startsWith("fc")?n:t)[r]=e[r]}),{featureExtractorMap:t,classifierMap:n}}var aH=class extends BG{constructor(e,t){super(e),this._faceFeatureExtractor=t}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){const{params:t}=this;if(!t)throw new Error(this._name+" - load model before inference");return ru(()=>{const n=e instanceof MG?this.faceFeatureExtractor.forwardInput(e):e;return nH(n.as2D(n.shape[0],-1),t.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){const{params:t,paramMappings:n}=this.extractClassifierParams(e);this._params=t,this._paramMappings=n}extractClassifierParams(e){return function(e,t,n){const r=[],{extractWeights:a,getRemainingWeights:s}=QG(e),i=HG(a,r)(t,n,"fc");if(0!==s().length)throw new Error("weights remaing after extract: "+s().length);return{paramMappings:r,params:{fc:i}}}(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(e){const{featureExtractorMap:t,classifierMap:n}=rH(e);return this.faceFeatureExtractor.loadFromWeightMap(t),function(e){const t=[],n=XG(e,t),r={fc:(a="fc",{weights:n(a+"/weights",2),bias:n(a+"/bias",1)})};var a;return jG(e,t),{params:r,paramMappings:t}}(n)}extractParams(e){const t=this.getClassifierChannelsIn(),n=this.getClassifierChannelsOut(),r=n*t+n,a=e.slice(0,e.length-r),s=e.slice(e.length-r);return this.faceFeatureExtractor.extractWeights(a),this.extractClassifierParams(s)}},sH=["neutral","happy","sad","angry","fearful","disgusted","surprised"],iH=class{constructor(e){if(this.neutral=0,this.happy=0,this.sad=0,this.angry=0,this.fearful=0,this.disgusted=0,this.surprised=0,7!==e.length)throw new Error("FaceExpressions.constructor - expected probabilities.length to be 7, have: "+e.length);sH.forEach((t,n)=>{this[t]=e[n]})}asSortedArray(){return sH.map(e=>({expression:e,probability:this[e]})).sort((e,t)=>t.probability-e.probability)}};function oH(e,t,n=.1,r){(Array.isArray(t)?t:[t]).forEach(t=>{const a=t instanceof iH?t:t.expressions instanceof iH?t.expressions:void 0;if(!a)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");const s=a.asSortedArray().filter(e=>e.probability>n),i=cG(t)?t.detection.box.bottomLeft:r||new Zj(0,0);new IG(s.map(e=>`${e.expression} (${Gj(e.probability)})`),i).draw(e)})}function uH(e){return cG(e)&&e.landmarks instanceof uG&&e.unshiftedLandmarks instanceof uG&&e.alignedRect instanceof nG}function lH(e,t){const{box:n}=e.detection,r=t.shiftBy(n.x,n.y),a=r.align(),{imageDims:s}=e.detection,i={landmarks:r,unshiftedLandmarks:t,alignedRect:new nG(e.detection.score,a.rescale(s.reverse()),s),angle:function(e){const t=(e,t,n,r)=>Math.atan2(r-t,n-e)%Math.PI,n={roll:void 0,pitch:void 0,yaw:void 0};if(!e||!e._positions||68!==e._positions.length)return n;const r=e._positions;n.roll=-t(r[36]._x,r[36]._y,r[45]._x,r[45]._y),n.pitch=t(0,Math.abs(r[0]._x-r[30]._x)/r[30]._x,Math.PI,Math.abs(r[16]._x-r[30]._x)/r[30]._x);const a=r.reduce((e,t)=>e<t._y?e:t._y,1/0),s=r.reduce((e,t)=>e>t._y?e:t._y,-1/0);return n.yaw=Math.PI*(e._imgDims._height/(s-a)/1.4-1),n}(t)};return{...e,...i}}var cH=class{constructor(e={}){const{drawLines:t=!0,drawPoints:n=!0,lineWidth:r,lineColor:a,pointSize:s,pointColor:i}=e;this.drawLines=t,this.drawPoints=n,this.lineWidth=r||1,this.pointSize=s||2,this.lineColor=a||"rgba(0, 255, 255, 1)",this.pointColor=i||"rgba(255, 0, 255, 1)"}},hH=class{constructor(e,t={}){this.faceLandmarks=e,this.options=new cH(t)}draw(e){const t=wG(e),{drawLines:n,drawPoints:r,lineWidth:a,lineColor:s,pointSize:i,pointColor:o}=this.options;if(n&&this.faceLandmarks instanceof lG&&(t.strokeStyle=s,t.lineWidth=a,Lj(t,this.faceLandmarks.getJawOutline()),Lj(t,this.faceLandmarks.getLeftEyeBrow()),Lj(t,this.faceLandmarks.getRightEyeBrow()),Lj(t,this.faceLandmarks.getNose()),Lj(t,this.faceLandmarks.getLeftEye(),!0),Lj(t,this.faceLandmarks.getRightEye(),!0),Lj(t,this.faceLandmarks.getMouth(),!0)),r){t.strokeStyle=o,t.fillStyle=o;const e=e=>{t.beginPath(),t.arc(e.x,e.y,i,0,2*Math.PI),t.fill()};this.faceLandmarks.positions.forEach(e)}}};function pH(e,t){(Array.isArray(t)?t:[t]).forEach(t=>{const n=t instanceof uG?t:uH(t)?t.landmarks:void 0;if(!n)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new hH(n).draw(e)})}function dH(e,t){const n=[],{extractWeights:r,getRemainingWeights:a}=QG(e),{extractConvParams:s,extractSeparableConvParams:i,extractReductionBlockParams:o,extractMainBlockParams:u}=function(e,t){const n=GG(e,t),r=KG(e,t);return{extractConvParams:n,extractSeparableConvParams:r,extractReductionBlockParams:function(e,t,a){return{separable_conv0:r(e,t,a+"/separable_conv0"),separable_conv1:r(t,t,a+"/separable_conv1"),expansion_conv:n(e,t,1,a+"/expansion_conv")}},extractMainBlockParams:function(e,t){return{separable_conv0:r(e,e,t+"/separable_conv0"),separable_conv1:r(e,e,t+"/separable_conv1"),separable_conv2:r(e,e,t+"/separable_conv2")}}}}(r,n),l={conv_in:s(3,32,3,"entry_flow/conv_in"),reduction_block_0:o(32,64,"entry_flow/reduction_block_0"),reduction_block_1:o(64,128,"entry_flow/reduction_block_1")},c={};Yj(t,0,1).forEach(e=>{c["main_block_"+e]=u(128,"middle_flow/main_block_"+e)});const h={reduction_block:o(128,256,"exit_flow/reduction_block"),separable_conv:i(256,512,"exit_flow/separable_conv")};if(0!==a().length)throw new Error("weights remaing after extract: "+a().length);return{paramMappings:n,params:{entry_flow:l,middle_flow:c,exit_flow:h}}}function fH(e,t){const n=[],{extractConvParams:r,extractSeparableConvParams:a,extractReductionBlockParams:s,extractMainBlockParams:i}=function(e,t){const n=XG(e,t),r=JG(n),a=YG(n);return{extractConvParams:r,extractSeparableConvParams:a,extractReductionBlockParams:function(e){return{separable_conv0:a(e+"/separable_conv0"),separable_conv1:a(e+"/separable_conv1"),expansion_conv:r(e+"/expansion_conv")}},extractMainBlockParams:function(e){return{separable_conv0:a(e+"/separable_conv0"),separable_conv1:a(e+"/separable_conv1"),separable_conv2:a(e+"/separable_conv2")}}}}(e,n),o={conv_in:r("entry_flow/conv_in"),reduction_block_0:s("entry_flow/reduction_block_0"),reduction_block_1:s("entry_flow/reduction_block_1")},u={};Yj(t,0,1).forEach(e=>{u["main_block_"+e]=i("middle_flow/main_block_"+e)});const l={reduction_block:s("exit_flow/reduction_block"),separable_conv:a("exit_flow/separable_conv")};return jG(e,n),{params:{entry_flow:o,middle_flow:u,exit_flow:l},paramMappings:n}}function mH(e,t,n){return gu(ml(e,t.filters,n,"same"),t.bias)}function gH(e,t,n=!0){let r=n?gh(e):e;return r=zG(r,t.separable_conv0,[1,1]),r=zG(gh(r),t.separable_conv1,[1,1]),r=Fc(r,[3,3],[2,2],"same"),r=gu(r,mH(e,t.expansion_conv,[2,2])),r}var yH=class extends BG{constructor(e){super("TinyXception"),this._numMainBlocks=e}forwardInput(e){const{params:t}=this;if(!t)throw new Error("TinyXception - load model before inference");return ru(()=>{const n=aG(Ai(e.toBatchTensor(112,!0),"float32"),[122.782,117.001,104.298]).div(255);let r=gh(mH(n,t.entry_flow.conv_in,[2,2]));return r=gH(r,t.entry_flow.reduction_block_0,!1),r=gH(r,t.entry_flow.reduction_block_1),Yj(this._numMainBlocks,0,1).forEach(e=>{r=function(e,t){let n=zG(gh(e),t.separable_conv0,[1,1]);return n=zG(gh(n),t.separable_conv1,[1,1]),n=zG(gh(n),t.separable_conv2,[1,1]),n=gu(n,e),n}(r,t.middle_flow["main_block_"+e])}),r=gH(r,t.exit_flow.reduction_block),r=gh(zG(r,t.exit_flow.separable_conv,[1,1])),r})}async forward(e){return this.forwardInput(await DG(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(e){return fH(e,this._numMainBlocks)}extractParams(e){return dH(e,this._numMainBlocks)}};var bH=class extends aH{postProcess(e,t,n){const r=n.map(({width:e,height:n})=>{const r=t/Math.max(n,e);return{width:e*r,height:n*r}}),a=r.length;return ru(()=>{const n=(e,t)=>Gh([Vl([68],e,"float32"),Vl([68],t,"float32")],1).as2D(1,136).as1D(),s=(e,t)=>{const{width:n,height:a}=r[e];return t(n,a)?Math.abs(n-a)/2:0};return e.mul(Vl([a,136],t,"float32")).sub(Gh(Array.from(Array(a),(e,t)=>n((e=>s(e,(e,t)=>e<t))(t),(e=>s(e,(e,t)=>t<e))(t))))).div(Gh(Array.from(Array(a),(e,t)=>n(r[t].width,r[t].height))))})}forwardInput(e){return ru(()=>{const t=this.runNet(e);return this.postProcess(t,e.inputSize,e.inputDimensions.map(([e,t])=>({height:e,width:t})))})}async forward(e){return this.forwardInput(await DG(e))}async detectLandmarks(e){const t=await DG(e),n=ru(()=>ap(this.forwardInput(t))),r=await Promise.all(n.map(async(e,n)=>{const r=Array.from(e.dataSync()),a=r.filter((e,t)=>jj(t)),s=r.filter((e,t)=>!jj(t));return new lG(Array(68).fill(0).map((e,t)=>new Zj(a[t],s[t])),{height:t.getInputHeight(n),width:t.getInputWidth(n)})}));return n.forEach(e=>e.dispose()),t.isBatchInput?r:r[0]}getClassifierChannelsOut(){return 136}};var vH=class extends BG{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("TinyFaceFeatureExtractor - load model before inference");return ru(()=>{let n=UG(aG(Ai(e.toBatchTensor(112,!0),"float32"),[122.782,117.001,104.298]).div(255),t.dense0,!0);return n=UG(n,t.dense1),n=UG(n,t.dense2),n=Ku(n,[14,14],[2,2],"valid"),n})}async forward(e){return this.forwardInput(await DG(e))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(e){return function(e){const t=[],{extractDenseBlock3Params:n}=eH(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2")};return jG(e,t),{params:r,paramMappings:t}}(e)}extractParams(e){return function(e){const t=[],{extractWeights:n,getRemainingWeights:r}=QG(e),{extractDenseBlock3Params:a}=ZG(n,t),s=a(3,32,"dense0",!0),i=a(32,64,"dense1"),o=a(64,128,"dense2");if(0!==r().length)throw new Error("weights remaing after extract: "+r().length);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o}}}(e)}};function xH(e,t,n,r,a="same"){const{filters:s,bias:i}=t.conv;let o=ml(e,s,n,a);return o=gu(o,i),o=function(e,t){return gu(vu(e,t.weights),t.biases)}(o,t.scale),r?gh(o):o}function wH(e,t){return xH(e,t,[1,1],!1)}function kH(e,t){return xH(e,t,[2,2],!0,"valid")}function SH(e,t){function n(n,r,a,s){const i=function(t,n,r){const a=e(t),s=a.length/(n*r*r);if(Vj(s))throw new Error(`depth has to be an integer: ${s}, weights.length: ${a.length}, numFilters: ${n}, filterSize: ${r}`);return ru(()=>qi(Qh(a,[n,s,r,r]),[2,3,1,0]))}(n,r,a),o=Yh(e(r));return t.push({paramPath:s+"/filters"},{paramPath:s+"/bias"}),{filters:i,bias:o}}function r(r,a,s,i){return{conv:n(r,a,s,i+"/conv"),scale:function(n,r){const a=Yh(e(n)),s=Yh(e(n));return t.push({paramPath:r+"/weights"},{paramPath:r+"/biases"}),{weights:a,biases:s}}(a,i+"/scale")}}return{extractConvLayerParams:r,extractResidualLayerParams:function(e,t,n,a,s=!1){return{conv1:r((s?.5:1)*e,t,n,a+"/conv1"),conv2:r(e,t,n,a+"/conv2")}}}}function IH(e,t){const n=XG(e,t);function r(e){return{conv:{filters:n(e+"/conv/filters",4),bias:n(e+"/conv/bias",1)},scale:function(e){return{weights:n(e+"/scale/weights",1),biases:n(e+"/scale/biases",1)}}(e)}}return{extractConvLayerParams:r,extractResidualLayerParams:function(e){return{conv1:r(e+"/conv1"),conv2:r(e+"/conv2")}}}}function AH(e,t){let n=function(e,t){return xH(e,t,[1,1],!0)}(e,t.conv1);return n=wH(n,t.conv2),n=gu(n,e),n=gh(n),n}function EH(e,t){let n=kH(e,t.conv1);n=wH(n,t.conv2);let r=Ku(e,2,2,"valid");const a=Pc(r.shape),s=r.shape[3]!==n.shape[3];if(r.shape[1]!==n.shape[1]||r.shape[2]!==n.shape[2]){const e=[...n.shape];e[1]=1;const t=Pc(e);n=Xu([n,t],1);const r=[...n.shape];r[2]=1;const a=Pc(r);n=Xu([n,a],2)}return r=s?Xu([r,a],3):r,n=gu(r,n),n=gh(n),n}function NH(e,t){function n(n,r,a,s,i){const o=Qh(e(n*r*a*a),[a,a,n,r]),u=Yh(e(r));return t.push({paramPath:s+"/filters"},{paramPath:`${s}/${i?"batch_norm_offset":"bias"}`}),{filters:o,bias:u}}function r(e,t,r,a){const{filters:s,bias:i}=n(e,t,r,a,!0);return{filters:s,batch_norm_offset:i}}function a(n,a,s){return{depthwise_conv:function(n,r){const a=Qh(e(9*n),[3,3,n,1]),s=Yh(e(n)),i=Yh(e(n)),o=Yh(e(n)),u=Yh(e(n));return t.push({paramPath:r+"/filters"},{paramPath:r+"/batch_norm_scale"},{paramPath:r+"/batch_norm_offset"},{paramPath:r+"/batch_norm_mean"},{paramPath:r+"/batch_norm_variance"}),{filters:a,batch_norm_scale:s,batch_norm_offset:i,batch_norm_mean:o,batch_norm_variance:u}}(n,s+"/depthwise_conv"),pointwise_conv:r(n,a,1,s+"/pointwise_conv")}}return{extractMobilenetV1Params:function(){return{conv_0:r(3,32,3,"mobilenetv1/conv_0"),conv_1:a(32,64,"mobilenetv1/conv_1"),conv_2:a(64,128,"mobilenetv1/conv_2"),conv_3:a(128,128,"mobilenetv1/conv_3"),conv_4:a(128,256,"mobilenetv1/conv_4"),conv_5:a(256,256,"mobilenetv1/conv_5"),conv_6:a(256,512,"mobilenetv1/conv_6"),conv_7:a(512,512,"mobilenetv1/conv_7"),conv_8:a(512,512,"mobilenetv1/conv_8"),conv_9:a(512,512,"mobilenetv1/conv_9"),conv_10:a(512,512,"mobilenetv1/conv_10"),conv_11:a(512,512,"mobilenetv1/conv_11"),conv_12:a(512,1024,"mobilenetv1/conv_12"),conv_13:a(1024,1024,"mobilenetv1/conv_13")}},extractPredictionLayerParams:function(){return{conv_0:r(1024,256,1,"prediction_layer/conv_0"),conv_1:r(256,512,3,"prediction_layer/conv_1"),conv_2:r(512,128,1,"prediction_layer/conv_2"),conv_3:r(128,256,3,"prediction_layer/conv_3"),conv_4:r(256,128,1,"prediction_layer/conv_4"),conv_5:r(128,256,3,"prediction_layer/conv_5"),conv_6:r(256,64,1,"prediction_layer/conv_6"),conv_7:r(64,128,3,"prediction_layer/conv_7"),box_predictor_0:{box_encoding_predictor:n(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),class_predictor:n(512,9,1,"prediction_layer/box_predictor_0/class_predictor")},box_predictor_1:{box_encoding_predictor:n(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),class_predictor:n(1024,18,1,"prediction_layer/box_predictor_1/class_predictor")},box_predictor_2:{box_encoding_predictor:n(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),class_predictor:n(512,18,1,"prediction_layer/box_predictor_2/class_predictor")},box_predictor_3:{box_encoding_predictor:n(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),class_predictor:n(256,18,1,"prediction_layer/box_predictor_3/class_predictor")},box_predictor_4:{box_encoding_predictor:n(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),class_predictor:n(256,18,1,"prediction_layer/box_predictor_4/class_predictor")},box_predictor_5:{box_encoding_predictor:n(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),class_predictor:n(128,18,1,"prediction_layer/box_predictor_5/class_predictor")}}}}}function CH(e){const t=[],{extractMobilenetV1Params:n,extractPredictionLayerParams:r}=function(e,t){const n=XG(e,t);function r(e,t,r){return{filters:n(`${e}/Conv2d_${t}_pointwise/weights`,4,r+"/filters"),batch_norm_offset:n(`${e}/Conv2d_${t}_pointwise/convolution_bn_offset`,1,r+"/batch_norm_offset")}}function a(e){const t="mobilenetv1/conv_"+e,a=`MobilenetV1/Conv2d_${e}_depthwise`,s=t+"/depthwise_conv",i=t+"/pointwise_conv";return{depthwise_conv:{filters:n(a+"/depthwise_weights",4,s+"/filters"),batch_norm_scale:n(a+"/BatchNorm/gamma",1,s+"/batch_norm_scale"),batch_norm_offset:n(a+"/BatchNorm/beta",1,s+"/batch_norm_offset"),batch_norm_mean:n(a+"/BatchNorm/moving_mean",1,s+"/batch_norm_mean"),batch_norm_variance:n(a+"/BatchNorm/moving_variance",1,s+"/batch_norm_variance")},pointwise_conv:r("MobilenetV1",e,i)}}function s(e,t){return{filters:n(e+"/weights",4,t+"/filters"),bias:n(e+"/biases",1,t+"/bias")}}function i(e){return{box_encoding_predictor:s(`Prediction/BoxPredictor_${e}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${e}/box_encoding_predictor`),class_predictor:s(`Prediction/BoxPredictor_${e}/ClassPredictor`,`prediction_layer/box_predictor_${e}/class_predictor`)}}return{extractMobilenetV1Params:function(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:a(1),conv_2:a(2),conv_3:a(3),conv_4:a(4),conv_5:a(5),conv_6:a(6),conv_7:a(7),conv_8:a(8),conv_9:a(9),conv_10:a(10),conv_11:a(11),conv_12:a(12),conv_13:a(13)}},extractPredictionLayerParams:function(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:i(0),box_predictor_1:i(1),box_predictor_2:i(2),box_predictor_3:i(3),box_predictor_4:i(4),box_predictor_5:i(5)}}}}(e,t),a=e["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!Uj(a))throw new Error("expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have "+a);const s={mobilenetv1:n(),prediction_layer:r(),output_layer:{extra_dim:a}};return jG(e,t),{params:s,paramMappings:t}}function TH(e,t,n){return ru(()=>{let r=ml(e,t.filters,n,"same");return r=gu(r,t.batch_norm_offset),cl(r,0,6)})}function RH(e,t){return ru(()=>{let n,r=TH(e,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((e,t)=>{const a=t+1,s=function(e){return[2,4,6,12].some(t=>t===e)?[2,2]:[1,1]}(a);r=function(e,t,n){return ru(()=>{let r=Nl(e,t.filters,n,"same");return r=nl(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,.0010000000474974513),cl(r,0,6)})}(r,e.depthwise_conv,s),r=TH(r,e.pointwise_conv,[1,1]),11===a&&(n=r)}),null===n)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:n}})}function _H(e,t,n){const r=e.arraySync(),a=Math.min(r[t][0],r[t][2]),s=Math.min(r[t][1],r[t][3]),i=Math.max(r[t][0],r[t][2]),o=Math.max(r[t][1],r[t][3]),u=Math.min(r[n][0],r[n][2]),l=Math.min(r[n][1],r[n][3]),c=Math.max(r[n][0],r[n][2]),h=Math.max(r[n][1],r[n][3]),p=(i-a)*(o-s),d=(c-u)*(h-l);if(p<=0||d<=0)return 0;const f=Math.max(a,u),m=Math.max(s,l),g=Math.min(i,c),y=Math.min(o,h),b=Math.max(g-f,0)*Math.max(y-m,0);return b/(p+d-b)}function FH(e,t){const{sizes:n,centers:r}=function(e){const t=ap(qi(e,[1,0])),n=[gc(t[2],t[0]),gc(t[3],t[1])];return{sizes:n,centers:[gu(t[0],bu(n[0],2)),gu(t[1],bu(n[1],2))]}}(e),a=ap(qi(t,[1,0])),s=bu(vu($l(bu(a[2],5)),n[0]),2),i=gu(vu(bu(a[0],10),n[0]),r[0]),o=bu(vu($l(bu(a[3],5)),n[1]),2),u=gu(vu(bu(a[1],10),n[1]),r[1]);return qi(Gh([gc(i,s),gc(u,o),gu(i,s),gu(u,o)]),[1,0])}function MH(e,t){return ru(()=>{const n=e.shape[0];return{boxPredictionEncoding:qu(VG(e,t.box_encoding_predictor),[n,-1,1,4]),classPrediction:qu(VG(e,t.class_predictor),[n,-1,3])}})}var DH=class{constructor({minConfidence:e,maxResults:t}={}){if(this._name="SsdMobilenetv1Options",this._minConfidence=e||.5,this._maxResults=t||100,"number"!=typeof this._minConfidence||this._minConfidence<=0||this._minConfidence>=1)throw new Error(this._name+" - expected minConfidence to be a number between 0 and 1");if("number"!=typeof this._maxResults)throw new Error(this._name+" - expected maxResults to be a number")}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}},OH=[new Zj(.738768,.874946),new Zj(2.42204,2.65704),new Zj(4.30971,7.04493),new Zj(10.246,4.59428),new Zj(12.6868,11.8741)],LH=[new Zj(1.603231,2.094468),new Zj(6.041143,7.080126),new Zj(2.882459,3.518061),new Zj(4.266906,5.178857),new Zj(9.041765,10.66308)],PH=[117.001,114.697,97.404],$H=e=>"number"==typeof e;function BH(e){return ru(()=>{const t=vu(e,Ah(.10000000149011612));return gu(gh(gc(e,t)),t)})}function zH(e,t){return ru(()=>{let n=Qc(e,[[0,0],[1,1],[1,1],[0,0]]);return n=ml(n,t.conv.filters,[1,1],"valid"),n=gc(n,t.bn.sub),n=vu(n,t.bn.truediv),n=gu(n,t.conv.bias),BH(n)})}function UH(e,t){return ru(()=>{let n=Qc(e,[[0,0],[1,1],[1,1],[0,0]]);return n=Nh(n,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),n=gu(n,t.bias),BH(n)})}function WH(e,t){const n=GG(e,t);const r=KG(e,t);return{extractConvParams:n,extractConvWithBatchNormParams:function(r,a,s){return{conv:n(r,a,3,s+"/conv"),bn:function(n,r){const a=Yh(e(n)),s=Yh(e(n));return t.push({paramPath:r+"/sub"},{paramPath:r+"/truediv"}),{sub:a,truediv:s}}(a,s+"/bn")}},extractSeparableConvParams:r}}function VH(e,t){const n=XG(e,t);function r(e){return{filters:n(e+"/filters",4),bias:n(e+"/bias",1)}}return{extractConvParams:r,extractConvWithBatchNormParams:function(e){return{conv:r(e+"/conv"),bn:function(e){return{sub:n(e+"/sub",1),truediv:n(e+"/truediv",1)}}(e+"/bn")}},extractSeparableConvParams:YG(n)}}var jH=class{constructor({inputSize:e,scoreThreshold:t}={}){if(this._name="TinyYolov2Options",this._inputSize=e||416,this._scoreThreshold=t||.5,"number"!=typeof this._inputSize||this._inputSize%32!=0)throw new Error(this._name+" - expected inputSize to be a number divisible by 32");if("number"!=typeof this._scoreThreshold||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(this._name+" - expected scoreThreshold to be a number between 0 and 1")}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}},GH=class extends BG{constructor(e){super("TinyYolov2"),function(e){if(!e)throw new Error("invalid config: "+e);if("boolean"!=typeof e.withSeparableConvs)throw new Error("config.withSeparableConvs has to be a boolean, have: "+e.withSeparableConvs);if(!$H(e.iouThreshold)||e.iouThreshold<0||e.iouThreshold>1)throw new Error("config.iouThreshold has to be a number between [0, 1], have: "+e.iouThreshold);if(!Array.isArray(e.classes)||!e.classes.length||!e.classes.every(e=>"string"==typeof e))throw new Error("config.classes has to be an array class names: string[], have: "+JSON.stringify(e.classes));if(!Array.isArray(e.anchors)||!e.anchors.length||!e.anchors.map(e=>e||{}).every(e=>$H(e.x)&&$H(e.y)))throw new Error("config.anchors has to be an array of { x: number, y: number }, have: "+JSON.stringify(e.anchors));if(e.meanRgb&&(!Array.isArray(e.meanRgb)||3!==e.meanRgb.length||!e.meanRgb.every($H)))throw new Error("config.meanRgb has to be an array of shape [number, number, number], have: "+JSON.stringify(e.meanRgb))}(e),this._config=e}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(e,t){let n=zH(e,t.conv0);return n=Fc(n,[2,2],[2,2],"same"),n=zH(n,t.conv1),n=Fc(n,[2,2],[2,2],"same"),n=zH(n,t.conv2),n=Fc(n,[2,2],[2,2],"same"),n=zH(n,t.conv3),n=Fc(n,[2,2],[2,2],"same"),n=zH(n,t.conv4),n=Fc(n,[2,2],[2,2],"same"),n=zH(n,t.conv5),n=Fc(n,[2,2],[1,1],"same"),n=zH(n,t.conv6),n=zH(n,t.conv7),VG(n,t.conv8,"valid",!1)}runMobilenet(e,t){let n=this.config.isFirstLayerConv2d?BH(VG(e,t.conv0,"valid",!1)):UH(e,t.conv0);return n=Fc(n,[2,2],[2,2],"same"),n=UH(n,t.conv1),n=Fc(n,[2,2],[2,2],"same"),n=UH(n,t.conv2),n=Fc(n,[2,2],[2,2],"same"),n=UH(n,t.conv3),n=Fc(n,[2,2],[2,2],"same"),n=UH(n,t.conv4),n=Fc(n,[2,2],[2,2],"same"),n=UH(n,t.conv5),n=Fc(n,[2,2],[1,1],"same"),n=t.conv6?UH(n,t.conv6):n,n=t.conv7?UH(n,t.conv7):n,VG(n,t.conv8,"valid",!1)}forwardInput(e,t){const{params:n}=this;if(!n)throw new Error("TinyYolov2 - load model before inference");return ru(()=>{let r=Ai(e.toBatchTensor(t,!1),"float32");return r=this.config.meanRgb?aG(r,this.config.meanRgb):r,r=r.div(255),this.config.withSeparableConvs?this.runMobilenet(r,n):this.runTinyYolov2(r,n)})}async forward(e,t){return this.forwardInput(await DG(e),t)}async detect(e,t={}){const{inputSize:n,scoreThreshold:r}=new jH(t),a=await DG(e),s=await this.forwardInput(a,n),i=ru(()=>ap(s)[0].expandDims()),o={width:a.getInputWidth(0),height:a.getInputHeight(0)},u=await this.extractBoxes(i,a.getReshapedInputDimensions(0),r);s.dispose(),i.dispose();const l=u.map(e=>e.box),c=u.map(e=>e.score),h=u.map(e=>e.classScore),p=u.map(e=>this.config.classes[e.label]);return function(e,t,n,r=!0){let a=t.map((e,t)=>({score:e,boxIndex:t})).sort((e,t)=>e.score-t.score).map(e=>e.boxIndex);const s=[];for(;a.length>0;){const t=a.pop();s.push(t);const i=a,o=[];for(let n=0;n<i.length;n++){const a=i[n],s=e[t],u=e[a];o.push(rG(s,u,r))}a=a.filter((e,t)=>o[t]<=n)}return s}(l.map(e=>e.rescale(n)),c,this.config.iouThreshold,!0).map(e=>new tG(c[e],h[e],p[e],l[e],o))}getDefaultModelName(){return""}extractParamsFromWeightMap(e){return function(e,t){const n=[],{extractConvParams:r,extractConvWithBatchNormParams:a,extractSeparableConvParams:s}=VH(e,n);let i;if(t.withSeparableConvs){const e=t.filterSizes&&t.filterSizes.length||9;i={conv0:t.isFirstLayerConv2d?r("conv0"):s("conv0"),conv1:s("conv1"),conv2:s("conv2"),conv3:s("conv3"),conv4:s("conv4"),conv5:s("conv5"),conv6:e>7?s("conv6"):void 0,conv7:e>8?s("conv7"):void 0,conv8:r("conv8")}}else i={conv0:a("conv0"),conv1:a("conv1"),conv2:a("conv2"),conv3:a("conv3"),conv4:a("conv4"),conv5:a("conv5"),conv6:a("conv6"),conv7:a("conv7"),conv8:r("conv8")};return jG(e,n),{params:i,paramMappings:n}}(e,this.config)}extractParams(e){const t=this.config.filterSizes||GH.DEFAULT_FILTER_SIZES,n=t?t.length:void 0;if(7!==n&&8!==n&&9!==n)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${n} filterSizes in config`);return function(e,t,n,r){const{extractWeights:a,getRemainingWeights:s}=QG(e),i=[],{extractConvParams:o,extractConvWithBatchNormParams:u,extractSeparableConvParams:l}=WH(a,i);let c;if(t.withSeparableConvs){const[e,a,s,i,u,h,p,d,f]=r;c={conv0:t.isFirstLayerConv2d?o(e,a,3,"conv0"):l(e,a,"conv0"),conv1:l(a,s,"conv1"),conv2:l(s,i,"conv2"),conv3:l(i,u,"conv3"),conv4:l(u,h,"conv4"),conv5:l(h,p,"conv5"),conv6:d?l(p,d,"conv6"):void 0,conv7:f?l(d,f,"conv7"):void 0,conv8:o(f||d||p,5*n,1,"conv8")}}else{const[e,t,a,s,i,l,h,p,d]=r;c={conv0:u(e,t,"conv0"),conv1:u(t,a,"conv1"),conv2:u(a,s,"conv2"),conv3:u(s,i,"conv3"),conv4:u(i,l,"conv4"),conv5:u(l,h,"conv5"),conv6:u(h,p,"conv6"),conv7:u(p,d,"conv7"),conv8:o(d,5*n,1,"conv8")}}if(0!==s().length)throw new Error("weights remaing after extract: "+s().length);return{params:c,paramMappings:i}}(e,this.config,this.boxEncodingSize,t)}async extractBoxes(e,t,n){const{width:r,height:a}=t,s=Math.max(r,a),i=s/r,o=s/a,u=e.shape[1],l=this.config.anchors.length,[c,h,p]=ru(()=>{const t=e.reshape([u,u,l,this.boxEncodingSize]);return[t.slice([0,0,0,0],[u,u,l,4]),t.slice([0,0,0,4],[u,u,l,1]),this.withClassScores?Lh(t.slice([0,0,0,5],[u,u,l,this.config.classes.length]),3):Ah(0)]}),d=[],f=await h.array(),m=await c.array();for(let e=0;e<u;e++)for(let t=0;t<u;t++)for(let r=0;r<l;r++){const a=sG(f[e][t][r][0]);if(!n||a>n){const n=(t+sG(m[e][t][r][0]))/u*i,s=(e+sG(m[e][t][r][1]))/u*o,l=Math.exp(m[e][t][r][2])*this.config.anchors[r].x/u*i,c=Math.exp(m[e][t][r][3])*this.config.anchors[r].y/u*o,h=n-l/2,f=s-c/2,g={row:e,col:t,anchor:r},{classScore:y,label:b}=this.withClassScores?await this.extractPredictedClass(p,g):{classScore:1,label:0};d.push({box:new eG(h,f,h+l,f+c),score:a,classScore:a*y,label:b,...g})}}return c.dispose(),h.dispose(),p.dispose(),d}async extractPredictedClass(e,t){const{row:n,col:r,anchor:a}=t,s=await e.array();return Array(this.config.classes.length).fill(0).map((e,t)=>s[n][r][a][t]).map((e,t)=>({classScore:e,label:t})).reduce((e,t)=>e.classScore>t.classScore?e:t)}},HH=GH;HH.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var qH=class extends jH{constructor(){super(...arguments),this._name="TinyFaceDetectorOptions"}},KH=class{async then(e){return e(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};async function YH(e,t,n,r,a=(({alignedRect:e})=>e)){const s=e.map(e=>uH(e)?a(e):e.detection),i=r||(t instanceof Ya?await LG(t,s):await OG(t,s)),o=await n(i);return i.forEach(e=>e instanceof Ya&&e.dispose()),o}var XH=[new Zj(1.603231,2.094468),new Zj(6.041143,7.080126),new Zj(2.882459,3.518061),new Zj(4.266906,5.178857),new Zj(9.041765,10.66308)],QH=[117.001,114.697,97.404],ZH={ssdMobilenetv1:new class extends BG{constructor(){super("SsdMobilenetv1")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("SsdMobilenetv1 - load model before inference");return ru(()=>{const n=Ai(e.toBatchTensor(512,!1),"float32"),r=RH(gc(bu(n,127.5),1),t.mobilenetv1),{boxPredictions:a,classPredictions:s}=function(e,t,n){return ru(()=>{const r=TH(e,n.conv_0,[1,1]),a=TH(r,n.conv_1,[2,2]),s=TH(a,n.conv_2,[1,1]),i=TH(s,n.conv_3,[2,2]),o=TH(i,n.conv_4,[1,1]),u=TH(o,n.conv_5,[2,2]),l=TH(u,n.conv_6,[1,1]),c=TH(l,n.conv_7,[2,2]),h=MH(t,n.box_predictor_0),p=MH(e,n.box_predictor_1),d=MH(a,n.box_predictor_2),f=MH(i,n.box_predictor_3),m=MH(u,n.box_predictor_4),g=MH(c,n.box_predictor_5);return{boxPredictions:Xu([h.boxPredictionEncoding,p.boxPredictionEncoding,d.boxPredictionEncoding,f.boxPredictionEncoding,m.boxPredictionEncoding,g.boxPredictionEncoding],1),classPredictions:Xu([h.classPrediction,p.classPrediction,d.classPrediction,f.classPrediction,m.classPrediction,g.classPrediction],1)}})}(r.out,r.conv11,t.prediction_layer);return function(e,t,n){return ru(()=>{const r=e.shape[0];let a=FH(qu(Ul(n.extra_dim,[r,1,1]),[-1,4]),qu(e,[-1,4]));a=qu(a,[r,a.shape[0]/r,4]);const s=Qu(Zu(t,[0,0,1],[-1,-1,-1]));let i=Zu(s,[0,0,0],[-1,-1,1]);i=qu(i,[r,i.shape[1]]);return{boxes:ap(a),scores:ap(i)}})}(a,s,t.output_layer)})}async forward(e){return this.forwardInput(await DG(e))}async locateFaces(e,t={}){const{maxResults:n,minConfidence:r}=new DH(t),a=await DG(e),{boxes:s,scores:i}=this.forwardInput(a),o=s[0],u=i[0];for(let e=1;e<s.length;e++)s[e].dispose(),i[e].dispose();const l=Array.from(u.dataSync()),c=function(e,t,n,r,a){const s=e.shape[0],i=Math.min(n,s),o=t.map((e,t)=>({score:e,boxIndex:t})).filter(e=>e.score>a).sort((e,t)=>t.score-e.score),u=[];return o.forEach(t=>{if(u.length>=i)return;const n=t.score;for(let n=u.length-1;n>=0;--n){const s=_H(e,t.boxIndex,u[n]);if(0!==s&&(t.score*=s<=r?1:0,t.score<=a))break}n===t.score&&u.push(t.boxIndex)}),u}(o,l,n,.5,r),h=a.getReshapedInputDimensions(0),p=a.inputSize,d=p/h.width,f=p/h.height,m=o.arraySync(),g=c.map(e=>{const[t,n]=[Math.max(0,m[e][0]),Math.min(1,m[e][2])].map(e=>e*f),[r,s]=[Math.max(0,m[e][1]),Math.min(1,m[e][3])].map(e=>e*d);return new nG(l[e],new oG(r,t,s-r,n-t),{height:a.getInputHeight(0),width:a.getInputWidth(0)})});return o.dispose(),u.dispose(),g}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(e){return CH(e)}extractParams(e){return function(e){const t=[],{extractWeights:n,getRemainingWeights:r}=QG(e),{extractMobilenetV1Params:a,extractPredictionLayerParams:s}=NH(n,t),i=a(),o=s(),u={extra_dim:to(n(20472),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),0!==r().length)throw new Error("weights remaing after extract: "+r().length);return{params:{mobilenetv1:i,prediction_layer:o,output_layer:u},paramMappings:t}}(e)}},tinyFaceDetector:new class extends HH{constructor(){super({withSeparableConvs:!0,iouThreshold:.4,classes:["face"],anchors:XH,meanRgb:QH,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]})}get anchors(){return this.config.anchors}async locateFaces(e,t){return(await this.detect(e,t)).map(e=>new nG(e.score,e.relativeBox,{width:e.imageWidth,height:e.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(e){return super.extractParamsFromWeightMap(e)}},tinyYolov2:new class extends HH{constructor(e=!0){super({withSeparableConvs:e,iouThreshold:.4,classes:["face"],...e?{anchors:LH,meanRgb:PH}:{anchors:OH,withClassScores:!0}})}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(e,t){return(await this.detect(e,t)).map(e=>new nG(e.score,e.relativeBox,{width:e.imageWidth,height:e.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?"tiny_yolov2_separable_conv_model":"tiny_yolov2_model"}extractParamsFromWeightMap(e){return super.extractParamsFromWeightMap(e)}},faceLandmark68Net:new class extends bH{constructor(e=new tH){super("FaceLandmark68Net",e)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}},faceLandmark68TinyNet:new class extends bH{constructor(e=new vH){super("FaceLandmark68TinyNet",e)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}},faceRecognitionNet:new class extends BG{constructor(){super("FaceRecognitionNet")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("FaceRecognitionNet - load model before inference");return ru(()=>{let n=kH(aG(Ai(e.toBatchTensor(150,!0),"float32"),[122.782,117.001,104.298]).div(255),t.conv32_down);n=Fc(n,3,2,"valid"),n=AH(n,t.conv32_1),n=AH(n,t.conv32_2),n=AH(n,t.conv32_3),n=EH(n,t.conv64_down),n=AH(n,t.conv64_1),n=AH(n,t.conv64_2),n=AH(n,t.conv64_3),n=EH(n,t.conv128_down),n=AH(n,t.conv128_1),n=AH(n,t.conv128_2),n=EH(n,t.conv256_down),n=AH(n,t.conv256_1),n=AH(n,t.conv256_2),n=EH(n,t.conv256_down_out);const r=n.mean([1,2]);return Gi(r,t.fc)})}async forward(e){return this.forwardInput(await DG(e))}async computeFaceDescriptor(e){var t;if(null==(t=null==e?void 0:e.shape)?void 0:t.some(e=>e<=0))return new Float32Array(128);const n=await DG(e),r=ru(()=>ap(this.forwardInput(n))),a=await Promise.all(r.map(e=>e.data()));return r.forEach(e=>e.dispose()),n.isBatchInput?a:a[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(e){return function(e){const t=[],{extractConvLayerParams:n,extractResidualLayerParams:r}=IH(e,t),a=n("conv32_down"),s=r("conv32_1"),i=r("conv32_2"),o=r("conv32_3"),u=r("conv64_down"),l=r("conv64_1"),c=r("conv64_2"),h=r("conv64_3"),p=r("conv128_down"),d=r("conv128_1"),f=r("conv128_2"),m=r("conv256_down"),g=r("conv256_1"),y=r("conv256_2"),b=r("conv256_down_out"),{fc:v}=e;if(t.push({originalPath:"fc",paramPath:"fc"}),!zj(v))throw new Error("expected weightMap[fc] to be a Tensor2D, instead have "+v);const x={conv32_down:a,conv32_1:s,conv32_2:i,conv32_3:o,conv64_down:u,conv64_1:l,conv64_2:c,conv64_3:h,conv128_down:p,conv128_1:d,conv128_2:f,conv256_down:m,conv256_1:g,conv256_2:y,conv256_down_out:b,fc:v};return jG(e,t),{params:x,paramMappings:t}}(e)}extractParams(e){return function(e){const{extractWeights:t,getRemainingWeights:n}=QG(e),r=[],{extractConvLayerParams:a,extractResidualLayerParams:s}=SH(t,r),i=a(4704,32,7,"conv32_down"),o=s(9216,32,3,"conv32_1"),u=s(9216,32,3,"conv32_2"),l=s(9216,32,3,"conv32_3"),c=s(36864,64,3,"conv64_down",!0),h=s(36864,64,3,"conv64_1"),p=s(36864,64,3,"conv64_2"),d=s(36864,64,3,"conv64_3"),f=s(147456,128,3,"conv128_down",!0),m=s(147456,128,3,"conv128_1"),g=s(147456,128,3,"conv128_2"),y=s(589824,256,3,"conv256_down",!0),b=s(589824,256,3,"conv256_1"),v=s(589824,256,3,"conv256_2"),x=s(589824,256,3,"conv256_down_out"),w=ru(()=>qi(Xh(t(32768),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),0!==n().length)throw new Error("weights remaing after extract: "+n().length);return{params:{conv32_down:i,conv32_1:o,conv32_2:u,conv32_3:l,conv64_down:c,conv64_1:h,conv64_2:p,conv64_3:d,conv128_down:f,conv128_1:m,conv128_2:g,conv256_down:y,conv256_1:b,conv256_2:v,conv256_down_out:x,fc:w},paramMappings:r}}(e)}},faceExpressionNet:new class extends aH{constructor(e=new tH){super("FaceExpressionNet",e)}forwardInput(e){return ru(()=>Lh(this.runNet(e)))}async forward(e){return this.forwardInput(await DG(e))}async predictExpressions(e){const t=await DG(e),n=await this.forwardInput(t),r=await Promise.all(ap(n).map(async e=>{const t=e.dataSync();return e.dispose(),t}));n.dispose();const a=r.map(e=>new iH(e));return t.isBatchInput?a:a[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}},ageGenderNet:new class extends BG{constructor(e=new yH(2)){super("AgeGenderNet"),this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){const{params:t}=this;if(!t)throw new Error(this._name+" - load model before inference");return ru(()=>{const n=e instanceof MG?this.faceFeatureExtractor.forwardInput(e):e,r=Ku(n,[7,7],[2,2],"valid").as2D(n.shape[0],-1);return{age:nH(r,t.fc.age).as1D(),gender:nH(r,t.fc.gender)}})}forwardInput(e){return ru(()=>{const{age:t,gender:n}=this.runNet(e);return{age:t,gender:Lh(n)}})}async forward(e){return this.forwardInput(await DG(e))}async predictAgeAndGender(e){const t=await DG(e),n=await this.forwardInput(t),r=ap(n.age),a=ap(n.gender),s=r.map((e,t)=>({ageTensor:e,genderTensor:a[t]})),i=await Promise.all(s.map(async({ageTensor:e,genderTensor:t})=>{const n=e.dataSync()[0],r=t.dataSync()[0],a=r>.5,s=a?"male":"female",i=a?r:1-r;return e.dispose(),t.dispose(),{age:n,gender:s,genderProbability:i}}));return n.age.dispose(),n.gender.dispose(),t.isBatchInput?i:i[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){const{params:t,paramMappings:n}=this.extractClassifierParams(e);this._params=t,this._paramMappings=n}extractClassifierParams(e){return function(e){const t=[],{extractWeights:n,getRemainingWeights:r}=QG(e),a=HG(n,t),s=a(512,1,"fc/age"),i=a(512,2,"fc/gender");if(0!==r().length)throw new Error("weights remaing after extract: "+r().length);return{paramMappings:t,params:{fc:{age:s,gender:i}}}}(e)}extractParamsFromWeightMap(e){const{featureExtractorMap:t,classifierMap:n}=rH(e);return this.faceFeatureExtractor.loadFromWeightMap(t),function(e){const t=[],n=XG(e,t);function r(e){return{weights:n(e+"/weights",2),bias:n(e+"/bias",1)}}const a={fc:{age:r("fc/age"),gender:r("fc/gender")}};return jG(e,t),{params:a,paramMappings:t}}(n)}extractParams(e){const t=e.slice(0,e.length-1539),n=e.slice(e.length-1539);return this.faceFeatureExtractor.extractWeights(t),this.extractClassifierParams(n)}}},JH=class extends KH{constructor(e,t,n){super(),this.parentTask=e,this.input=t,this.extractedFaces=n}},eq=class extends JH{async run(){const e=await this.parentTask,t=await YH(e,this.input,async e=>Promise.all(e.map(e=>ZH.faceExpressionNet.predictExpressions(e))),this.extractedFaces);return e.map((e,n)=>{return r=e,a=t[n],{...r,expressions:a};var r,a})}withAgeAndGender(){return new rq(this,this.input)}},tq=class extends eq{withAgeAndGender(){return new aq(this,this.input)}withFaceDescriptors(){return new iq(this,this.input)}},nq=class extends KH{constructor(e,t,n){super(),this.parentTask=e,this.input=t,this.extractedFaces=n}},rq=class extends nq{async run(){const e=await this.parentTask,t=await YH(e,this.input,async e=>Promise.all(e.map(e=>ZH.ageGenderNet.predictAgeAndGender(e))),this.extractedFaces);return e.map((e,n)=>{const{age:r,gender:a,genderProbability:s}=t[n];return function(e,t){return{...e,...{age:t}}}(function(e,t,n){return{...e,...{gender:t,genderProbability:n}}}(e,a,s),r)})}withFaceExpressions(){return new eq(this,this.input)}},aq=class extends rq{withFaceExpressions(){return new tq(this,this.input)}withFaceDescriptors(){return new iq(this,this.input)}},sq=class extends KH{constructor(e,t){super(),this.parentTask=e,this.input=t}},iq=class extends sq{async run(){const e=await this.parentTask;return(await YH(e,this.input,e=>Promise.all(e.map(e=>ZH.faceRecognitionNet.computeFaceDescriptor(e))),null,e=>e.landmarks.align(null,{useDlibAlignment:!0}))).map((t,n)=>function(e,t){return{...e,...{descriptor:t}}}(e[n],t))}withFaceExpressions(){return new tq(this,this.input)}withAgeAndGender(){return new aq(this,this.input)}},oq=class extends KH{constructor(e,t,n){super(),this.parentTask=e,this.input=t,this.useTinyLandmarkNet=n}get landmarkNet(){return this.useTinyLandmarkNet?ZH.faceLandmark68TinyNet:ZH.faceLandmark68Net}},uq=class extends oq{async run(){const e=await this.parentTask,t=e.map(e=>e.detection),n=this.input instanceof Ya?await LG(this.input,t):await OG(this.input,t),r=await Promise.all(n.map(e=>this.landmarkNet.detectLandmarks(e)));return n.forEach(e=>e instanceof Ya&&e.dispose()),e.map((e,t)=>lH(e,r[t]))}withFaceExpressions(){return new tq(this,this.input)}withAgeAndGender(){return new aq(this,this.input)}withFaceDescriptors(){return new iq(this,this.input)}},lq=class extends KH{constructor(e,t=new DH){super(),this.input=e,this.options=t}},cq=class extends lq{async run(){const{input:e,options:t}=this;let n;if(t instanceof qH)n=ZH.tinyFaceDetector.locateFaces(e,t);else if(t instanceof DH)n=ZH.ssdMobilenetv1.locateFaces(e,t);else{if(!(t instanceof jH))throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");n=ZH.tinyYolov2.locateFaces(e,t)}return n}runAndExtendWithFaceDetections(){return new Promise((e,t)=>{this.run().then(t=>e(t.map(e=>hG({},e)))).catch(e=>t(e))})}withFaceLandmarks(e=!1){return new uq(this.runAndExtendWithFaceDetections(),this.input,e)}withFaceExpressions(){return new eq(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new rq(this.runAndExtendWithFaceDetections(),this.input)}};class hq{constructor(){this.FACE_ANGLE_TOO_LARGE=!1,this.FACE_CLOSE_TO_BORDER=!1,this.FACE_NOT_FOUND=!1,this.FACE_TOO_SMALL=!1,this.PROBABILITY_TOO_SMALL=!1,this.TOO_MANY_FACES=!1}}function pq(e,t){if(!(e instanceof t))throw new TypeError("Cannot call a class as a function")}function dq(e,t){for(var n=0;n<t.length;n++){var r=t[n];r.enumerable=r.enumerable||!1,r.configurable=!0,"value"in r&&(r.writable=!0),Object.defineProperty(e,r.key,r)}}
-/**
-* KalmanFilter
-* @class
-* @author Wouter Bulten
-* @see {@link http://github.com/wouterbulten/kalmanjs}
-* @version Version: 1.0.0-beta
-* @copyright Copyright 2015-2018 Wouter Bulten
-* @license MIT License
-* @preserve
-*/
-var fq=function(){function e(){var t=arguments.length>0&&void 0!==arguments[0]?arguments[0]:{},n=t.R,r=void 0===n?1:n,a=t.Q,s=void 0===a?1:a,i=t.A,o=void 0===i?1:i,u=t.B,l=void 0===u?0:u,c=t.C,h=void 0===c?1:c;pq(this,e),this.R=r,this.Q=s,this.A=o,this.C=h,this.B=l,this.cov=NaN,this.x=NaN}var t,n,r;return t=e,(n=[{key:"filter",value:function(e){var t=arguments.length>1&&void 0!==arguments[1]?arguments[1]:0;if(isNaN(this.x))this.x=1/this.C*e,this.cov=1/this.C*this.Q*(1/this.C);else{var n=this.predict(t),r=this.uncertainty(),a=r*this.C*(1/(this.C*r*this.C+this.Q));this.x=n+a*(e-this.C*n),this.cov=r-a*this.C*r}return this.x}},{key:"predict",value:function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:0;return this.A*this.x+this.B*e}},{key:"uncertainty",value:function(){return this.A*this.cov*this.A+this.R}},{key:"lastMeasurement",value:function(){return this.x}},{key:"setMeasurementNoise",value:function(e){this.Q=e}},{key:"setProcessNoise",value:function(e){this.R=e}}])&&dq(t.prototype,n),r&&dq(t,r),e}();const mq=window.cv,gq={R:1,Q:50},yq=(e,t)=>bq[e].filter(t),bq={nosetipx:new fq(gq),nosetipy:new fq(gq),frontx:new fq(gq),fronty:new fq(gq),topx:new fq(gq),topy:new fq(gq),sidex:new fq(gq),sidey:new fq(gq)};class vq{constructor(e,t){this.selectedOpenCvMethod="0",this.noseTip=e[30],this.bottomNose=e[33],this.lefteyeleftcorner=e[36],this.lefteyerightcorner=e[39],this.righteyerightcorner=e[45],this.righteyeleftcorner=e[42],this.leftmouth=e[48],this.rightmouth=e[54],this.leftnostril=e[31],this.rightnostril=e[35],this.focal_length=t.width,this.center=[t.width/2,t.height/2]}estimateHeadPose(){const e=mq.matFromArray(10,2,mq.CV_64F,[this.noseTip._x,this.noseTip._y,this.bottomNose._x,this.bottomNose._y,this.leftnostril._x,this.leftnostril._y,this.rightnostril._x,this.rightnostril._y,this.lefteyeleftcorner._x,this.lefteyeleftcorner._y,this.lefteyerightcorner._x,this.lefteyerightcorner._y,this.righteyerightcorner._x,this.righteyerightcorner._y,this.righteyeleftcorner._x,this.righteyeleftcorner._y,this.leftmouth._x,this.leftmouth._y,this.rightmouth._x,this.rightmouth._y]),t=mq.matFromArray(10,3,mq.CV_64F,[0,0,0,0,-60,-78,-67,-58,-100,67,-58,-100,-262,168,-240,-115,170,-210,262,168,-240,115,170,-210,-148,-192,-181,148,-192,-181]),n=mq.matFromArray(3,3,mq.CV_64F,[this.focal_length,0,this.center[0],0,this.focal_length,this.center[1],0,0,1]),r=mq.Mat.zeros(4,1,mq.CV_64F),a=new mq.Mat,s=new mq.Mat,i=new mq.Mat;mq.solvePnPRansac(t,e,n,r,a,s,!1,100,8,.99,i,parseInt(this.selectedOpenCvMethod,10)),mq.solvePnPRefineVVS(t,e,n,r,a,s);const o=mq.matFromArray(3,1,mq.CV_64F,[0,0,1e3]),u=new mq.Mat,l=new mq.Mat;mq.projectPoints(o,a,s,n,r,u,l);let c=u.data64F[0],h=u.data64F[1];const p=mq.matFromArray(3,1,mq.CV_64F,[0,1e3,0]),d=new mq.Mat,f=new mq.Mat;mq.projectPoints(p,a,s,n,r,d,f);let m=d.data64F[0],g=d.data64F[1];const y=mq.matFromArray(3,1,mq.CV_64F,[1e3,0,0]),b=new mq.Mat,v=new mq.Mat;mq.projectPoints(y,a,s,n,r,b,v);let x=b.data64F[0],w=b.data64F[1];return this.drawingData={noseTip:this.noseTip,frontx:c,fronty:h,topx:m,topy:g,sidex:x,sidey:w},a.data64F.map(e=>e/Math.PI*180)}draw(e){const{noseTip:t,frontx:n,fronty:r,topx:a,topy:s,sidex:i,sidey:o}=this.drawingData;try{const u={_x:yq("nosetipx",t.x),_y:yq("nosetipy",t.y)},l=yq("frontx",n),c=yq("fronty",r),h=yq("topx",a),p=yq("topy",s),d=yq("sidex",i),f=yq("sidey",o);e.strokeStyle="white",e.fillStyle="white",e.beginPath(),e.arc(l,c,2,0,2*Math.PI),e.fill(),e.arc(u._x,u._y,2,0,2*Math.PI),e.fill(),e.lineWidth=2,e.beginPath(),e.strokeStyle="red",e.fillStyle="red",e.moveTo(l,c),e.lineTo(u._x,u._y),e.stroke(),e.beginPath(),e.strokeStyle="green",e.fillStyle="green",e.moveTo(h,p),e.lineTo(u._x,u._y),e.stroke(),e.beginPath(),e.strokeStyle="blue",e.fillStyle="blue",e.moveTo(d,f),e.lineTo(u._x,u._y),e.stroke()}catch(e){console.error("errrrr"),console.error(e)}}}const xq=()=>{let e={isMobile:!1,isAndroid:!1,isLinux:!1,isMac:!1,isWin:!1,isChrome:!1,isFirefox:!1,isSafari:!1,isIos:!1};if(e.isMobile=/Android|webOS|iPhone|iPad|iPod|BlackBerry/i.test(navigator.userAgent),e.isAndroid=/Android/i.test(navigator.userAgent),e.isIos=/iPhone|iPad|iPod/i.test(navigator.userAgent),e.isLinux=/linux/i.test(navigator.platform),e.isMac=/mac/i.test(navigator.platform),e.isWin=/win/i.test(navigator.platform),e.isChrome=/chrome/i.test(navigator.userAgent),e.isFirefox=/firefox/i.test(navigator.userAgent),e.isSafari=!e.isChrome&&/safari/i.test(navigator.userAgent),e.isMobile=/Android|webOS|iPhone|iPad|iPod|BlackBerry/i.test(navigator.userAgent),e.isAndroid=/Android/i.test(navigator.userAgent),e.isIos=/iPhone|iPad|iPod/i.test(navigator.userAgent),!e.isIos){navigator.userAgent.match(/Mac/)&&navigator.maxTouchPoints&&navigator.maxTouchPoints>2&&(e.isIos=!0,e.isMobile=!0)}return e},wq=window.cv;class kq{constructor(e,t,n){this.probabilityThreshold=50,this.counterSuccessfulResults=0,this.MAX_FPS_DETECTION=200,this.rendering={now:0,elapsed:0,then:0},this.MAX_NUMBER_FACES=1,this.MAX_ANGLE_TURN_HEAD=30,this.MIN_FACE_SIZE_FOR_MOBILE=224,this.MIN_FACE_SIZE_FOR_DESKTOP=350,this.MIN_OCCUPIED_SPACE_FOR_DESKTOP=15,this.MIN_OCCUPIED_SPACE_FOR_MOBILE=15,this.X_OFFSET_FROM_FRAME=15,this.Y_OFFSET_FROM_FRAME=5,this.DEFAULT_NUMBER_SUCCESSFUL_RESULTS_FOR_AUTO_CAPTURING=5,this.videoRatio=1,this.showDetectedFace=!1,this.headTurnCounter=0,this.counterSuccessfulHeadTurns=0,this.stream=e,this.isMobile=t,this.modelPath=n}setProbabilityThreshold(e){this.probabilityThreshold=e}static getInstance(e,t,n){return kq.instance||(kq.instance=new kq(e,t,n)),kq.instance}updateHtmlElements(e){this.videoElement=e}static async initDetector(e){await d.setBackend(this.BACKEND),await d.enableProdMode(),await d.ENV.set("DEBUG",!1),await d.ready(),await this.loadingModels(e),wq||console.error("open cv not found"),l.loadedModels({tinyFaceDetectorLoaded:ZH.tinyFaceDetector.isLoaded,faceLandmark68TinyNetLoaded:ZH.faceLandmark68TinyNet.isLoaded});const t=performance.now();await this.warmingUpModels(),l.detectorInitialized(performance.now()-t)}static async loadingModels(e){this.modelsLoaded()||(await ZH.tinyFaceDetector.load(e),await ZH.faceLandmark68TinyNet.load(e))}static modelsLoaded(){return ZH.tinyFaceDetector.isLoaded&&ZH.faceLandmark68TinyNet.isLoaded}static async warmingUpModels(){const e=new Image;e.src="data:image/jpg;base64,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",await this.detectAllFaces(e)}static async detectAllFaces(e){d.engine().startScope();let t=[];try{let n=e;if(this.isIos&&e instanceof HTMLVideoElement){const t=document.createElement("canvas"),r=480,a=640;t.width=r,t.height=a;t.getContext("2d").drawImage(e,0,0,r,a),n=t}t=await function(e,t=new DH){return new cq(e,t)}(n,this.tinyFaceDetectorOptions).withFaceLandmarks(!0)}catch(e){}finally{d.engine().endScope()}return t}async startDetector(){if(!this.videoElement)throw Error("Video element not found");if(0===this.videoElement.clientWidth)throw Error("Video element too small");this.rendering.then=Date.now(),this.counterSuccessfulResults=0,this.updateForIos(),this.videoRatio=function(e){const t=e.clientWidth/e.videoWidth,n=e.clientHeight/e.videoHeight;return Math.min(t,n)}(this.videoElement),kq.modelsLoaded()||await kq.initDetector(this.modelPath),this.requestAnimationFrame=requestAnimationFrame(()=>this.detection())}stopDetector(){this.requestAnimationFrame&&(cancelAnimationFrame(this.requestAnimationFrame),this.requestAnimationFrame=void 0,this.innerCanvas&&this.innerCanvas.remove(),this.innerCanvas=void 0)}updateForIos(){xq().isIos&&(this.DEFAULT_NUMBER_SUCCESSFUL_RESULTS_FOR_AUTO_CAPTURING=3,this.MAX_FPS_DETECTION=500,kq.isIos=!0)}async detection(){if(this.requestAnimationFrame=requestAnimationFrame(()=>this.detection()),this.rendering.now=Date.now(),this.rendering.elapsed=this.rendering.now-this.rendering.then,this.rendering.elapsed>this.MAX_FPS_DETECTION){this.rendering.then=this.rendering.now-this.rendering.elapsed%this.MAX_FPS_DETECTION;const e=await kq.detectAllFaces(this.videoElement);(this.showDetectedFace||kq.debug)&&this.drawFaces(e);const t=await this.checkDetectedFaceOnError(e);Object.values(t).every(e=>!e)?(this.counterSuccessfulResults++,this.autoCapturing()):(this.counterSuccessfulResults=0,this.stream.returnErrors(t))}}async checkDetectedFaceOnError(e){const t=function e(t,n){const{width:r,height:a}=new Pj(n.width,n.height);if(r<=0||a<=0)throw new Error("resizeResults - invalid dimensions: "+JSON.stringify({width:r,height:a}));if(Array.isArray(t))return t.map(t=>e(t,{width:r,height:a}));if(uH(t)){const e=t.detection.forSize(r,a),n=t.unshiftedLandmarks.forSize(e.box.width,e.box.height);return lH(hG(t,e),n)}return cG(t)?hG(t,t.detection.forSize(r,a)):t instanceof uG||t instanceof nG?t.forSize(r,a):t}
-/**
- * @license
- * Copyright 2017 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2018 Google LLC
- *
- * Use of this source code is governed by an MIT-style
- * license that can be found in the LICENSE file or at
- * https://opensource.org/licenses/MIT.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2018 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2018 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2019 Google LLC
- *
- * Use of this source code is governed by an MIT-style
- * license that can be found in the LICENSE file or at
- * https://opensource.org/licenses/MIT.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2019 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2019 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2020 Google Inc. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2020 Google LLC
- *
- * Use of this source code is governed by an MIT-style
- * license that can be found in the LICENSE file or at
- * https://opensource.org/licenses/MIT.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2020 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2020 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the License);
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an AS IS BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2021 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2021 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * https://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
- * @license
- * Copyright 2022 Google LLC. All Rights Reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * =============================================================================
- */
-/**
-* @license
-* Copyright 2018 Google LLC. All Rights Reserved.
-* Licensed under the Apache License, Version 2.0 (the "License");
-* you may not use this file except in compliance with the License.
-* You may obtain a copy of the License at
-*
-* http://www.apache.org/licenses/LICENSE-2.0
-*
-* Unless required by applicable law or agreed to in writing, software
-* distributed under the License is distributed on an "AS IS" BASIS,
-* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-* See the License for the specific language governing permissions and
-* limitations under the License.
-* =============================================================================
-*/
-/** @license See the LICENSE file. */(e,{width:this.videoElement.videoWidth,height:this.videoElement.videoHeight}),n=new hq,r=this.checkNumberFaces(t);return n.FACE_NOT_FOUND=!r&&0===t.length,n.TOO_MANY_FACES=!r&&t.length>this.MAX_NUMBER_FACES,r&&(wq&&(n.FACE_ANGLE_TOO_LARGE=!this.checkHeadRotation(t)),n.PROBABILITY_TOO_SMALL=!this.checkProbability(t),n.FACE_TOO_SMALL=!this.checkFaceSize(t),n.FACE_CLOSE_TO_BORDER=!this.checkFaceIndent(t)),n}autoCapturing(){kq.debug&&console.info(this.counterSuccessfulResults),this.counterSuccessfulResults>=this.DEFAULT_NUMBER_SUCCESSFUL_RESULTS_FOR_AUTO_CAPTURING&&(this.stream.autoCapturing(),this.counterSuccessfulResults=0)}checkNumberFaces(e){return e.length===this.MAX_NUMBER_FACES}checkProbability(e){return e[0].detection.score>+this.probabilityThreshold/100}checkFaceSize(e){const t=e[0].detection.box,{width:n,height:r,area:a}=t,{imageHeight:s,imageWidth:i}=e[0].detection,o=100/(s*i/a);return Math.min(n/this.videoRatio,r/this.videoRatio)>(this.isMobile?this.MIN_FACE_SIZE_FOR_MOBILE:this.MIN_FACE_SIZE_FOR_DESKTOP)&&o>(this.isMobile?this.MIN_OCCUPIED_SPACE_FOR_MOBILE:this.MIN_OCCUPIED_SPACE_FOR_DESKTOP)}checkFaceIndent(e){const t=e[0].detection.box,{imageHeight:n,imageWidth:r}=e[0].detection,{top:a,left:s,bottom:i,right:o}=t,u=s,l=r-o,c=n-i;return!(a<this.Y_OFFSET_FROM_FRAME||u<this.X_OFFSET_FROM_FRAME||l<this.X_OFFSET_FROM_FRAME||c<this.Y_OFFSET_FROM_FRAME)}checkHeadRotation(e){const t=document.createElement("canvas"),n=function(e,t,n=!1){const{width:r,height:a}=n?TG(t):t;return e.width=r,e.height=a,{width:r,height:a}}(t,this.videoElement,!0);t.remove();const r=new vq(e[0].landmarks.positions,n);let a=[0,0,0];try{a=r.estimateHeadPose()}catch(e){}return a[2]>this.MAX_ANGLE_TURN_HEAD||-1*a[2]>this.MAX_ANGLE_TURN_HEAD?this.headTurnCounter++:(this.counterSuccessfulHeadTurns++,this.counterSuccessfulHeadTurns>=5&&(this.headTurnCounter=0,this.counterSuccessfulHeadTurns=0)),this.headTurnCounter<=1}drawFaces(e){this.innerCanvas||this.createCanvasForDrawingMask(this.videoElement);const t=this.innerCanvas;if(!t)return console.error("no canvas for drawing");const n=t.getContext("2d");if(n){n.clearRect(0,0,t.width,t.height),n.font='small-caps 20px "Segoe UI"',n.fillStyle="white";for(const t of e){n.lineWidth=3,n.strokeStyle="deepskyblue",n.fillStyle="deepskyblue",n.globalAlpha=.6,n.beginPath(),n.rect(t.detection.box.x,t.detection.box.y,t.detection.box.width,t.detection.box.height),n.stroke(),n.globalAlpha=1,n.globalAlpha=.8,n.fillStyle="lightblue";const e=2;for(let r=0;r<t.landmarks.positions.length;r++)n.beginPath(),n.arc(t.landmarks.positions[r].x,t.landmarks.positions[r].y,e,0,2*Math.PI),n.fill()}}}createCanvasForDrawingMask(e){this.innerCanvas=document.createElement("canvas"),this.innerCanvas.className="canvas-on-video",this.innerCanvas.width=e.videoWidth,this.innerCanvas.height=e.videoHeight,this.videoElement.after(this.innerCanvas),this.innerCanvas.offsetWidth&&this.innerCanvas.setAttribute("style",`margin-left:${-this.innerCanvas.offsetWidth/2}px; margin-top: ${-this.innerCanvas.offsetHeight/2}px;`)}}kq.tinyFaceDetectorOptions=new qH({inputSize:192}),kq.BACKEND="webgl",kq.debug=!1;var Sq=function(e,t,n){return e(n={path:t,exports:{},require:function(e,t){return function(){throw new Error("Dynamic requires are not currently supported by @rollup/plugin-commonjs")}()}},n.exports),n.exports}((function(e,t){!function(){var n={};function r(e){return h(">"+d("B",e.length),e)}function a(e){return h(">"+d("H",e.length),e)}function s(e){return h(">"+d("L",e.length),e)}function i(e,t,n){var i,o,u,l,c="",p="";if("Byte"==t)(i=e.length)<=4?p=r(e)+d("\0",4-i):(p=h(">L",[n]),c=r(e));else if("Short"==t)(i=e.length)<=2?p=a(e)+d("\0\0",2-i):(p=h(">L",[n]),c=a(e));else if("Long"==t)(i=e.length)<=1?p=s(e):(p=h(">L",[n]),c=s(e));else if("Ascii"==t)(i=(o=e+"\0").length)>4?(p=h(">L",[n]),c=o):p=o+d("\0",4-i);else if("Rational"==t){if("number"==typeof e[0])i=1,u=e[0],l=e[1],o=h(">L",[u])+h(">L",[l]);else{i=e.length,o="";for(var f=0;f<i;f++)u=e[f][0],l=e[f][1],o+=h(">L",[u])+h(">L",[l])}p=h(">L",[n]),c=o}else if("SRational"==t){if("number"==typeof e[0])i=1,u=e[0],l=e[1],o=h(">l",[u])+h(">l",[l]);else{i=e.length,o="";for(f=0;f<i;f++)u=e[f][0],l=e[f][1],o+=h(">l",[u])+h(">l",[l])}p=h(">L",[n]),c=o}else"Undefined"==t&&((i=e.length)>4?(p=h(">L",[n]),c=e):p=e+d("\0",4-i));return[h(">L",[i]),p,c]}function o(e,t,n){var r,a=Object.keys(e).length,s=h(">H",[a]);r=["0th","1st"].indexOf(t)>-1?2+12*a+4:2+12*a;var o="",u="";for(var l in e)if("string"==typeof l&&(l=parseInt(l)),!("0th"==t&&[34665,34853].indexOf(l)>-1||"Exif"==t&&40965==l||"1st"==t&&[513,514].indexOf(l)>-1)){var c=e[l],p=h(">H",[l]),d=g[t][l].type,f=h(">H",[m[d]]);"number"==typeof c&&(c=[c]);var y=i(c,d,8+r+n+u.length);o+=p+f+y[0]+y[1],u+=y[2]}return[s+o,u]}function u(e){var t;if("ÿØ"==e.slice(0,2))t=function(e){for(var t,n=0;n<e.length;n++)if("ÿá"==(t=e[n]).slice(0,2)&&"Exif\0\0"==t.slice(4,10))return t;return null}(f(e)),this.tiftag=t?t.slice(10):null;else if(["II","MM"].indexOf(e.slice(0,2))>-1)this.tiftag=e;else{if("Exif"!=e.slice(0,4))throw new Error("Given file is neither JPEG nor TIFF.");this.tiftag=e.slice(6)}}if(n.version="1.0.4",n.remove=function(e){var t=!1;if("ÿØ"==e.slice(0,2));else{if("data:image/jpeg;base64,"!=e.slice(0,23)&&"data:image/jpg;base64,"!=e.slice(0,22))throw new Error("Given data is not jpeg.");e=c(e.split(",")[1]),t=!0}var n=f(e).filter((function(e){return!("ÿá"==e.slice(0,2)&&"Exif\0\0"==e.slice(4,10))})).join("");return t&&(n="data:image/jpeg;base64,"+l(n)),n},n.insert=function(e,t){var n=!1;if("Exif\0\0"!=e.slice(0,6))throw new Error("Given data is not exif.");if("ÿØ"==t.slice(0,2));else{if("data:image/jpeg;base64,"!=t.slice(0,23)&&"data:image/jpg;base64,"!=t.slice(0,22))throw new Error("Given data is not jpeg.");t=c(t.split(",")[1]),n=!0}var r="ÿá"+h(">H",[e.length+2])+e,a=function(e,t){var n=!1,r=[];e.forEach((function(a,s){"ÿá"==a.slice(0,2)&&"Exif\0\0"==a.slice(4,10)&&(n?r.unshift(s):(e[s]=t,n=!0))})),r.forEach((function(t){e.splice(t,1)})),!n&&t&&(e=[e[0],t].concat(e.slice(1)));return e.join("")}(f(t),r);return n&&(a="data:image/jpeg;base64,"+l(a)),a},n.load=function(e){var t;if("string"!=typeof e)throw new Error("'load' gots invalid type argument.");if("ÿØ"==e.slice(0,2))t=e;else if("data:image/jpeg;base64,"==e.slice(0,23)||"data:image/jpg;base64,"==e.slice(0,22))t=c(e.split(",")[1]);else{if("Exif"!=e.slice(0,4))throw new Error("'load' gots invalid file data.");t=e.slice(6)}var n={"0th":{},Exif:{},GPS:{},Interop:{},"1st":{},thumbnail:null},r=new u(t);if(null===r.tiftag)return n;"II"==r.tiftag.slice(0,2)?r.endian_mark="<":r.endian_mark=">";var a=p(r.endian_mark+"L",r.tiftag.slice(4,8))[0];n["0th"]=r.get_ifd(a,"0th");var s=n["0th"].first_ifd_pointer;if(delete n["0th"].first_ifd_pointer,34665 in n["0th"]&&(a=n["0th"][34665],n.Exif=r.get_ifd(a,"Exif")),34853 in n["0th"]&&(a=n["0th"][34853],n.GPS=r.get_ifd(a,"GPS")),40965 in n.Exif&&(a=n.Exif[40965],n.Interop=r.get_ifd(a,"Interop")),"\0\0\0\0"!=s&&(a=p(r.endian_mark+"L",s)[0],n["1st"]=r.get_ifd(a,"1st"),513 in n["1st"]&&514 in n["1st"])){var i=n["1st"][513]+n["1st"][514],o=r.tiftag.slice(n["1st"][513],i);n.thumbnail=o}return n},n.dump=function(e){var t,r,a,s,i,u,l=(t=e,JSON.parse(JSON.stringify(t))),c=!1,p=!1,d=!1,g=!1;r="0th"in l?l["0th"]:{},"Exif"in l&&Object.keys(l.Exif).length||"Interop"in l&&Object.keys(l.Interop).length?(r[34665]=1,c=!0,a=l.Exif,"Interop"in l&&Object.keys(l.Interop).length?(a[40965]=1,d=!0,s=l.Interop):Object.keys(a).indexOf(n.ExifIFD.InteroperabilityTag.toString())>-1&&delete a[40965]):Object.keys(r).indexOf(n.ImageIFD.ExifTag.toString())>-1&&delete r[34665],"GPS"in l&&Object.keys(l.GPS).length?(r[n.ImageIFD.GPSTag]=1,p=!0,i=l.GPS):Object.keys(r).indexOf(n.ImageIFD.GPSTag.toString())>-1&&delete r[n.ImageIFD.GPSTag],"1st"in l&&"thumbnail"in l&&null!=l.thumbnail&&(g=!0,l["1st"][513]=1,l["1st"][514]=1,u=l["1st"]);var y,b,v,x=o(r,"0th",0),w=x[0].length+12*c+12*p+4+x[1].length,k="",S=0,I="",A=0,E="",N=0,C="";(c&&(S=(y=o(a,"Exif",w))[0].length+12*d+y[1].length),p&&(A=(I=o(i,"GPS",w+S).join("")).length),d)&&(N=(E=o(s,"Interop",w+S+A).join("")).length);if(g&&(b=o(u,"1st",w+S+A+N),(v=function(e){var t=f(e);for(;"ÿà"<=t[1].slice(0,2)&&t[1].slice(0,2)<="ÿï";)t=[t[0]].concat(t.slice(2));return t.join("")}(l.thumbnail)).length>64e3))throw new Error("Given thumbnail is too large. max 64kB");var T="",R="",_="",F="\0\0\0\0";if(c){var M=h(">L",[D=8+w]);T=h(">H",[34665])+h(">H",[m.Long])+h(">L",[1])+M}if(p){M=h(">L",[D=8+w+S]);R=h(">H",[34853])+h(">H",[m.Long])+h(">L",[1])+M}if(d){M=h(">L",[D=8+w+S+A]);_=h(">H",[40965])+h(">H",[m.Long])+h(">L",[1])+M}if(g){var D;F=h(">L",[D=8+w+S+A+N]);var O="
\0\0\0\0
"+h(">L",[D+b[0].length+24+4+b[1].length]),L="\0\0\0\0
"+h(">L",[v.length]);C=b[0]+O+L+"\0\0\0\0"+b[1]+v}var P=x[0]+T+R+F+x[1];return c&&(k=y[0]+_+y[1]),"Exif\0\0MM\0*\0\0\0\b"+P+k+I+E+C},u.prototype={get_ifd:function(e,t){var n,r={},a=p(this.endian_mark+"H",this.tiftag.slice(e,e+2))[0],s=e+2;n=["0th","1st"].indexOf(t)>-1?"Image":t;for(var i=0;i<a;i++){e=s+12*i;var o=p(this.endian_mark+"H",this.tiftag.slice(e,e+2))[0],u=[p(this.endian_mark+"H",this.tiftag.slice(e+2,e+4))[0],p(this.endian_mark+"L",this.tiftag.slice(e+4,e+8))[0],this.tiftag.slice(e+8,e+12)];o in g[n]&&(r[o]=this.convert_value(u))}return"0th"==t&&(e=s+12*a,r.first_ifd_pointer=this.tiftag.slice(e,e+4)),r},convert_value:function(e){var t,n=null,r=e[0],a=e[1],s=e[2];if(1==r)a>4?(t=p(this.endian_mark+"L",s)[0],n=p(this.endian_mark+d("B",a),this.tiftag.slice(t,t+a))):n=p(this.endian_mark+d("B",a),s.slice(0,a));else if(2==r)a>4?(t=p(this.endian_mark+"L",s)[0],n=this.tiftag.slice(t,t+a-1)):n=s.slice(0,a-1);else if(3==r)a>2?(t=p(this.endian_mark+"L",s)[0],n=p(this.endian_mark+d("H",a),this.tiftag.slice(t,t+2*a))):n=p(this.endian_mark+d("H",a),s.slice(0,2*a));else if(4==r)a>1?(t=p(this.endian_mark+"L",s)[0],n=p(this.endian_mark+d("L",a),this.tiftag.slice(t,t+4*a))):n=p(this.endian_mark+d("L",a),s);else if(5==r)if(t=p(this.endian_mark+"L",s)[0],a>1){n=[];for(var i=0;i<a;i++)n.push([p(this.endian_mark+"L",this.tiftag.slice(t+8*i,t+4+8*i))[0],p(this.endian_mark+"L",this.tiftag.slice(t+4+8*i,t+8+8*i))[0]])}else n=[p(this.endian_mark+"L",this.tiftag.slice(t,t+4))[0],p(this.endian_mark+"L",this.tiftag.slice(t+4,t+8))[0]];else if(7==r)a>4?(t=p(this.endian_mark+"L",s)[0],n=this.tiftag.slice(t,t+a)):n=s.slice(0,a);else if(9==r)a>1?(t=p(this.endian_mark+"L",s)[0],n=p(this.endian_mark+d("l",a),this.tiftag.slice(t,t+4*a))):n=p(this.endian_mark+d("l",a),s);else{if(10!=r)throw new Error("Exif might be wrong. Got incorrect value type to decode. type:"+r);if(t=p(this.endian_mark+"L",s)[0],a>1){n=[];for(i=0;i<a;i++)n.push([p(this.endian_mark+"l",this.tiftag.slice(t+8*i,t+4+8*i))[0],p(this.endian_mark+"l",this.tiftag.slice(t+4+8*i,t+8+8*i))[0]])}else n=[p(this.endian_mark+"l",this.tiftag.slice(t,t+4))[0],p(this.endian_mark+"l",this.tiftag.slice(t+4,t+8))[0]]}return n instanceof Array&&1==n.length?n[0]:n}},"undefined"!=typeof window&&"function"==typeof window.btoa)var l=window.btoa;if(void 0===l)l=function(e){for(var t,n,r,a,s,i,o,u="",l=0,c="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";l<e.length;)a=(t=e.charCodeAt(l++))>>2,s=(3&t)<<4|(n=e.charCodeAt(l++))>>4,i=(15&n)<<2|(r=e.charCodeAt(l++))>>6,o=63&r,isNaN(n)?i=o=64:isNaN(r)&&(o=64),u=u+c.charAt(a)+c.charAt(s)+c.charAt(i)+c.charAt(o);return u};if("undefined"!=typeof window&&"function"==typeof window.atob)var c=window.atob;if(void 0===c)c=function(e){var t,n,r,a,s,i,o="",u=0,l="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";for(e=e.replace(/[^A-Za-z0-9\+\/\=]/g,"");u<e.length;)t=l.indexOf(e.charAt(u++))<<2|(a=l.indexOf(e.charAt(u++)))>>4,n=(15&a)<<4|(s=l.indexOf(e.charAt(u++)))>>2,r=(3&s)<<6|(i=l.indexOf(e.charAt(u++))),o+=String.fromCharCode(t),64!=s&&(o+=String.fromCharCode(n)),64!=i&&(o+=String.fromCharCode(r));return o};function h(e,t){if(!(t instanceof Array))throw new Error("'pack' error. Got invalid type argument.");if(e.length-1!=t.length)throw new Error("'pack' error. "+(e.length-1)+" marks, "+t.length+" elements.");var n;if("<"==e[0])n=!0;else{if(">"!=e[0])throw new Error("");n=!1}for(var r="",a=1,s=null,i=null,o=null;i=e[a];){if("b"==i.toLowerCase()){if(s=t[a-1],"b"==i&&s<0&&(s+=256),s>255||s<0)throw new Error("'pack' error.");o=String.fromCharCode(s)}else if("H"==i){if((s=t[a-1])>65535||s<0)throw new Error("'pack' error.");o=String.fromCharCode(Math.floor(s%65536/256))+String.fromCharCode(s%256),n&&(o=o.split("").reverse().join(""))}else{if("l"!=i.toLowerCase())throw new Error("'pack' error.");if(s=t[a-1],"l"==i&&s<0&&(s+=4294967296),s>4294967295||s<0)throw new Error("'pack' error.");o=String.fromCharCode(Math.floor(s/16777216))+String.fromCharCode(Math.floor(s%16777216/65536))+String.fromCharCode(Math.floor(s%65536/256))+String.fromCharCode(s%256),n&&(o=o.split("").reverse().join(""))}r+=o,a+=1}return r}function p(e,t){if("string"!=typeof t)throw new Error("'unpack' error. Got invalid type argument.");for(var n,r=0,a=1;a<e.length;a++)if("b"==e[a].toLowerCase())r+=1;else if("h"==e[a].toLowerCase())r+=2;else{if("l"!=e[a].toLowerCase())throw new Error("'unpack' error. Got invalid mark.");r+=4}if(r!=t.length)throw new Error("'unpack' error. Mismatch between symbol and string length. "+r+":"+t.length);if("<"==e[0])n=!0;else{if(">"!=e[0])throw new Error("'unpack' error.");n=!1}for(var s=[],i=0,o=1,u=null,l=null,c=null,h="";l=e[o];){if("b"==l.toLowerCase())c=1,u=(h=t.slice(i,i+c)).charCodeAt(0),"b"==l&&u>=128&&(u-=256);else if("H"==l)c=2,h=t.slice(i,i+c),n&&(h=h.split("").reverse().join("")),u=256*h.charCodeAt(0)+h.charCodeAt(1);else{if("l"!=l.toLowerCase())throw new Error("'unpack' error. "+l);c=4,h=t.slice(i,i+c),n&&(h=h.split("").reverse().join("")),u=16777216*h.charCodeAt(0)+65536*h.charCodeAt(1)+256*h.charCodeAt(2)+h.charCodeAt(3),"l"==l&&u>=2147483648&&(u-=4294967296)}s.push(u),i+=c,o+=1}return s}function d(e,t){for(var n="",r=0;r<t;r++)n+=e;return n}function f(e){if("ÿØ"!=e.slice(0,2))throw new Error("Given data isn't JPEG.");for(var t=2,n=["ÿØ"];;){if("ÿÚ"==e.slice(t,t+2)){n.push(e.slice(t));break}var r=t+p(">H",e.slice(t+2,t+4))[0]+2;if(n.push(e.slice(t,r)),(t=r)>=e.length)throw new Error("Wrong JPEG data.")}return n}var m={Byte:1,Ascii:2,Short:3,Long:4,Rational:5,Undefined:7,SLong:9,SRational:10},g={Image:{11:{name:"ProcessingSoftware",type:"Ascii"},254:{name:"NewSubfileType",type:"Long"},255:{name:"SubfileType",type:"Short"},256:{name:"ImageWidth",type:"Long"},257:{name:"ImageLength",type:"Long"},258:{name:"BitsPerSample",type:"Short"},259:{name:"Compression",type:"Short"},262:{name:"PhotometricInterpretation",type:"Short"},263:{name:"Threshholding",type:"Short"},264:{name:"CellWidth",type:"Short"},265:{name:"CellLength",type:"Short"},266:{name:"FillOrder",type:"Short"},269:{name:"DocumentName",type:"Ascii"},270:{name:"ImageDescription",type:"Ascii"},271:{name:"Make",type:"Ascii"},272:{name:"Model",type:"Ascii"},273:{name:"StripOffsets",type:"Long"},274:{name:"Orientation",type:"Short"},277:{name:"SamplesPerPixel",type:"Short"},278:{name:"RowsPerStrip",type:"Long"},279:{name:"StripByteCounts",type:"Long"},282:{name:"XResolution",type:"Rational"},283:{name:"YResolution",type:"Rational"},284:{name:"PlanarConfiguration",type:"Short"},290:{name:"GrayResponseUnit",type:"Short"},291:{name:"GrayResponseCurve",type:"Short"},292:{name:"T4Options",type:"Long"},293:{name:"T6Options",type:"Long"},296:{name:"ResolutionUnit",type:"Short"},301:{name:"TransferFunction",type:"Short"},305:{name:"Software",type:"Ascii"},306:{name:"DateTime",type:"Ascii"},315:{name:"Artist",type:"Ascii"},316:{name:"HostComputer",type:"Ascii"},317:{name:"Predictor",type:"Short"},318:{name:"WhitePoint",type:"Rational"},319:{name:"PrimaryChromaticities",type:"Rational"},320:{name:"ColorMap",type:"Short"},321:{name:"HalftoneHints",type:"Short"},322:{name:"TileWidth",type:"Short"},323:{name:"TileLength",type:"Short"},324:{name:"TileOffsets",type:"Short"},325:{name:"TileByteCounts",type:"Short"},330:{name:"SubIFDs",type:"Long"},332:{name:"InkSet",type:"Short"},333:{name:"InkNames",type:"Ascii"},334:{name:"NumberOfInks",type:"Short"},336:{name:"DotRange",type:"Byte"},337:{name:"TargetPrinter",type:"Ascii"},338:{name:"ExtraSamples",type:"Short"},339:{name:"SampleFormat",type:"Short"},340:{name:"SMinSampleValue",type:"Short"},341:{name:"SMaxSampleValue",type:"Short"},342:{name:"TransferRange",type:"Short"},343:{name:"ClipPath",type:"Byte"},344:{name:"XClipPathUnits",type:"Long"},345:{name:"YClipPathUnits",type:"Long"},346:{name:"Indexed",type:"Short"},347:{name:"JPEGTables",type:"Undefined"},351:{name:"OPIProxy",type:"Short"},512:{name:"JPEGProc",type:"Long"},513:{name:"JPEGInterchangeFormat",type:"Long"},514:{name:"JPEGInterchangeFormatLength",type:"Long"},515:{name:"JPEGRestartInterval",type:"Short"},517:{name:"JPEGLosslessPredictors",type:"Short"},518:{name:"JPEGPointTransforms",type:"Short"},519:{name:"JPEGQTables",type:"Long"},520:{name:"JPEGDCTables",type:"Long"},521:{name:"JPEGACTables",type:"Long"},529:{name:"YCbCrCoefficients",type:"Rational"},530:{name:"YCbCrSubSampling",type:"Short"},531:{name:"YCbCrPositioning",type:"Short"},532:{name:"ReferenceBlackWhite",type:"Rational"},700:{name:"XMLPacket",type:"Byte"},18246:{name:"Rating",type:"Short"},18249:{name:"RatingPercent",type:"Short"},32781:{name:"ImageID",type:"Ascii"},33421:{name:"CFARepeatPatternDim",type:"Short"},33422:{name:"CFAPattern",type:"Byte"},33423:{name:"BatteryLevel",type:"Rational"},33432:{name:"Copyright",type:"Ascii"},33434:{name:"ExposureTime",type:"Rational"},34377:{name:"ImageResources",type:"Byte"},34665:{name:"ExifTag",type:"Long"},34675:{name:"InterColorProfile",type:"Undefined"},34853:{name:"GPSTag",type:"Long"},34857:{name:"Interlace",type:"Short"},34858:{name:"TimeZoneOffset",type:"Long"},34859:{name:"SelfTimerMode",type:"Short"},37387:{name:"FlashEnergy",type:"Rational"},37388:{name:"SpatialFrequencyResponse",type:"Undefined"},37389:{name:"Noise",type:"Undefined"},37390:{name:"FocalPlaneXResolution",type:"Rational"},37391:{name:"FocalPlaneYResolution",type:"Rational"},37392:{name:"FocalPlaneResolutionUnit",type:"Short"},37393:{name:"ImageNumber",type:"Long"},37394:{name:"SecurityClassification",type:"Ascii"},37395:{name:"ImageHistory",type:"Ascii"},37397:{name:"ExposureIndex",type:"Rational"},37398:{name:"TIFFEPStandardID",type:"Byte"},37399:{name:"SensingMethod",type:"Short"},40091:{name:"XPTitle",type:"Byte"},40092:{name:"XPComment",type:"Byte"},40093:{name:"XPAuthor",type:"Byte"},40094:{name:"XPKeywords",type:"Byte"},40095:{name:"XPSubject",type:"Byte"},50341:{name:"PrintImageMatching",type:"Undefined"},50706:{name:"DNGVersion",type:"Byte"},50707:{name:"DNGBackwardVersion",type:"Byte"},50708:{name:"UniqueCameraModel",type:"Ascii"},50709:{name:"LocalizedCameraModel",type:"Byte"},50710:{name:"CFAPlaneColor",type:"Byte"},50711:{name:"CFALayout",type:"Short"},50712:{name:"LinearizationTable",type:"Short"},50713:{name:"BlackLevelRepeatDim",type:"Short"},50714:{name:"BlackLevel",type:"Rational"},50715:{name:"BlackLevelDeltaH",type:"SRational"},50716:{name:"BlackLevelDeltaV",type:"SRational"},50717:{name:"WhiteLevel",type:"Short"},50718:{name:"DefaultScale",type:"Rational"},50719:{name:"DefaultCropOrigin",type:"Short"},50720:{name:"DefaultCropSize",type:"Short"},50721:{name:"ColorMatrix1",type:"SRational"},50722:{name:"ColorMatrix2",type:"SRational"},50723:{name:"CameraCalibration1",type:"SRational"},50724:{name:"CameraCalibration2",type:"SRational"},50725:{name:"ReductionMatrix1",type:"SRational"},50726:{name:"ReductionMatrix2",type:"SRational"},50727:{name:"AnalogBalance",type:"Rational"},50728:{name:"AsShotNeutral",type:"Short"},50729:{name:"AsShotWhiteXY",type:"Rational"},50730:{name:"BaselineExposure",type:"SRational"},50731:{name:"BaselineNoise",type:"Rational"},50732:{name:"BaselineSharpness",type:"Rational"},50733:{name:"BayerGreenSplit",type:"Long"},50734:{name:"LinearResponseLimit",type:"Rational"},50735:{name:"CameraSerialNumber",type:"Ascii"},50736:{name:"LensInfo",type:"Rational"},50737:{name:"ChromaBlurRadius",type:"Rational"},50738:{name:"AntiAliasStrength",type:"Rational"},50739:{name:"ShadowScale",type:"SRational"},50740:{name:"DNGPrivateData",type:"Byte"},50741:{name:"MakerNoteSafety",type:"Short"},50778:{name:"CalibrationIlluminant1",type:"Short"},50779:{name:"CalibrationIlluminant2",type:"Short"},50780:{name:"BestQualityScale",type:"Rational"},50781:{name:"RawDataUniqueID",type:"Byte"},50827:{name:"OriginalRawFileName",type:"Byte"},50828:{name:"OriginalRawFileData",type:"Undefined"},50829:{name:"ActiveArea",type:"Short"},50830:{name:"MaskedAreas",type:"Short"},50831:{name:"AsShotICCProfile",type:"Undefined"},50832:{name:"AsShotPreProfileMatrix",type:"SRational"},50833:{name:"CurrentICCProfile",type:"Undefined"},50834:{name:"CurrentPreProfileMatrix",type:"SRational"},50879:{name:"ColorimetricReference",type:"Short"},50931:{name:"CameraCalibrationSignature",type:"Byte"},50932:{name:"ProfileCalibrationSignature",type:"Byte"},50934:{name:"AsShotProfileName",type:"Byte"},50935:{name:"NoiseReductionApplied",type:"Rational"},50936:{name:"ProfileName",type:"Byte"},50937:{name:"ProfileHueSatMapDims",type:"Long"},50938:{name:"ProfileHueSatMapData1",type:"Float"},50939:{name:"ProfileHueSatMapData2",type:"Float"},50940:{name:"ProfileToneCurve",type:"Float"},50941:{name:"ProfileEmbedPolicy",type:"Long"},50942:{name:"ProfileCopyright",type:"Byte"},50964:{name:"ForwardMatrix1",type:"SRational"},50965:{name:"ForwardMatrix2",type:"SRational"},50966:{name:"PreviewApplicationName",type:"Byte"},50967:{name:"PreviewApplicationVersion",type:"Byte"},50968:{name:"PreviewSettingsName",type:"Byte"},50969:{name:"PreviewSettingsDigest",type:"Byte"},50970:{name:"PreviewColorSpace",type:"Long"},50971:{name:"PreviewDateTime",type:"Ascii"},50972:{name:"RawImageDigest",type:"Undefined"},50973:{name:"OriginalRawFileDigest",type:"Undefined"},50974:{name:"SubTileBlockSize",type:"Long"},50975:{name:"RowInterleaveFactor",type:"Long"},50981:{name:"ProfileLookTableDims",type:"Long"},50982:{name:"ProfileLookTableData",type:"Float"},51008:{name:"OpcodeList1",type:"Undefined"},51009:{name:"OpcodeList2",type:"Undefined"},51022:{name:"OpcodeList3",type:"Undefined"}},Exif:{33434:{name:"ExposureTime",type:"Rational"},33437:{name:"FNumber",type:"Rational"},34850:{name:"ExposureProgram",type:"Short"},34852:{name:"SpectralSensitivity",type:"Ascii"},34855:{name:"ISOSpeedRatings",type:"Short"},34856:{name:"OECF",type:"Undefined"},34864:{name:"SensitivityType",type:"Short"},34865:{name:"StandardOutputSensitivity",type:"Long"},34866:{name:"RecommendedExposureIndex",type:"Long"},34867:{name:"ISOSpeed",type:"Long"},34868:{name:"ISOSpeedLatitudeyyy",type:"Long"},34869:{name:"ISOSpeedLatitudezzz",type:"Long"},36864:{name:"ExifVersion",type:"Undefined"},36867:{name:"DateTimeOriginal",type:"Ascii"},36868:{name:"DateTimeDigitized",type:"Ascii"},37121:{name:"ComponentsConfiguration",type:"Undefined"},37122:{name:"CompressedBitsPerPixel",type:"Rational"},37377:{name:"ShutterSpeedValue",type:"SRational"},37378:{name:"ApertureValue",type:"Rational"},37379:{name:"BrightnessValue",type:"SRational"},37380:{name:"ExposureBiasValue",type:"SRational"},37381:{name:"MaxApertureValue",type:"Rational"},37382:{name:"SubjectDistance",type:"Rational"},37383:{name:"MeteringMode",type:"Short"},37384:{name:"LightSource",type:"Short"},37385:{name:"Flash",type:"Short"},37386:{name:"FocalLength",type:"Rational"},37396:{name:"SubjectArea",type:"Short"},37500:{name:"MakerNote",type:"Undefined"},37510:{name:"UserComment",type:"Ascii"},37520:{name:"SubSecTime",type:"Ascii"},37521:{name:"SubSecTimeOriginal",type:"Ascii"},37522:{name:"SubSecTimeDigitized",type:"Ascii"},40960:{name:"FlashpixVersion",type:"Undefined"},40961:{name:"ColorSpace",type:"Short"},40962:{name:"PixelXDimension",type:"Long"},40963:{name:"PixelYDimension",type:"Long"},40964:{name:"RelatedSoundFile",type:"Ascii"},40965:{name:"InteroperabilityTag",type:"Long"},41483:{name:"FlashEnergy",type:"Rational"},41484:{name:"SpatialFrequencyResponse",type:"Undefined"},41486:{name:"FocalPlaneXResolution",type:"Rational"},41487:{name:"FocalPlaneYResolution",type:"Rational"},41488:{name:"FocalPlaneResolutionUnit",type:"Short"},41492:{name:"SubjectLocation",type:"Short"},41493:{name:"ExposureIndex",type:"Rational"},41495:{name:"SensingMethod",type:"Short"},41728:{name:"FileSource",type:"Undefined"},41729:{name:"SceneType",type:"Undefined"},41730:{name:"CFAPattern",type:"Undefined"},41985:{name:"CustomRendered",type:"Short"},41986:{name:"ExposureMode",type:"Short"},41987:{name:"WhiteBalance",type:"Short"},41988:{name:"DigitalZoomRatio",type:"Rational"},41989:{name:"FocalLengthIn35mmFilm",type:"Short"},41990:{name:"SceneCaptureType",type:"Short"},41991:{name:"GainControl",type:"Short"},41992:{name:"Contrast",type:"Short"},41993:{name:"Saturation",type:"Short"},41994:{name:"Sharpness",type:"Short"},41995:{name:"DeviceSettingDescription",type:"Undefined"},41996:{name:"SubjectDistanceRange",type:"Short"},42016:{name:"ImageUniqueID",type:"Ascii"},42032:{name:"CameraOwnerName",type:"Ascii"},42033:{name:"BodySerialNumber",type:"Ascii"},42034:{name:"LensSpecification",type:"Rational"},42035:{name:"LensMake",type:"Ascii"},42036:{name:"LensModel",type:"Ascii"},42037:{name:"LensSerialNumber",type:"Ascii"},42240:{name:"Gamma",type:"Rational"}},GPS:{0:{name:"GPSVersionID",type:"Byte"},1:{name:"GPSLatitudeRef",type:"Ascii"},2:{name:"GPSLatitude",type:"Rational"},3:{name:"GPSLongitudeRef",type:"Ascii"},4:{name:"GPSLongitude",type:"Rational"},5:{name:"GPSAltitudeRef",type:"Byte"},6:{name:"GPSAltitude",type:"Rational"},7:{name:"GPSTimeStamp",type:"Rational"},8:{name:"GPSSatellites",type:"Ascii"},9:{name:"GPSStatus",type:"Ascii"},10:{name:"GPSMeasureMode",type:"Ascii"},11:{name:"GPSDOP",type:"Rational"},12:{name:"GPSSpeedRef",type:"Ascii"},13:{name:"GPSSpeed",type:"Rational"},14:{name:"GPSTrackRef",type:"Ascii"},15:{name:"GPSTrack",type:"Rational"},16:{name:"GPSImgDirectionRef",type:"Ascii"},17:{name:"GPSImgDirection",type:"Rational"},18:{name:"GPSMapDatum",type:"Ascii"},19:{name:"GPSDestLatitudeRef",type:"Ascii"},20:{name:"GPSDestLatitude",type:"Rational"},21:{name:"GPSDestLongitudeRef",type:"Ascii"},22:{name:"GPSDestLongitude",type:"Rational"},23:{name:"GPSDestBearingRef",type:"Ascii"},24:{name:"GPSDestBearing",type:"Rational"},25:{name:"GPSDestDistanceRef",type:"Ascii"},26:{name:"GPSDestDistance",type:"Rational"},27:{name:"GPSProcessingMethod",type:"Undefined"},28:{name:"GPSAreaInformation",type:"Undefined"},29:{name:"GPSDateStamp",type:"Ascii"},30:{name:"GPSDifferential",type:"Short"},31:{name:"GPSHPositioningError",type:"Rational"}},Interop:{1:{name:"InteroperabilityIndex",type:"Ascii"}}};g["0th"]=g.Image,g["1st"]=g.Image,n.TAGS=g,n.ImageIFD={ProcessingSoftware:11,NewSubfileType:254,SubfileType:255,ImageWidth:256,ImageLength:257,BitsPerSample:258,Compression:259,PhotometricInterpretation:262,Threshholding:263,CellWidth:264,CellLength:265,FillOrder:266,DocumentName:269,ImageDescription:270,Make:271,Model:272,StripOffsets:273,Orientation:274,SamplesPerPixel:277,RowsPerStrip:278,StripByteCounts:279,XResolution:282,YResolution:283,PlanarConfiguration:284,GrayResponseUnit:290,GrayResponseCurve:291,T4Options:292,T6Options:293,ResolutionUnit:296,TransferFunction:301,Software:305,DateTime:306,Artist:315,HostComputer:316,Predictor:317,WhitePoint:318,PrimaryChromaticities:319,ColorMap:320,HalftoneHints:321,TileWidth:322,TileLength:323,TileOffsets:324,TileByteCounts:325,SubIFDs:330,InkSet:332,InkNames:333,NumberOfInks:334,DotRange:336,TargetPrinter:337,ExtraSamples:338,SampleFormat:339,SMinSampleValue:340,SMaxSampleValue:341,TransferRange:342,ClipPath:343,XClipPathUnits:344,YClipPathUnits:345,Indexed:346,JPEGTables:347,OPIProxy:351,JPEGProc:512,JPEGInterchangeFormat:513,JPEGInterchangeFormatLength:514,JPEGRestartInterval:515,JPEGLosslessPredictors:517,JPEGPointTransforms:518,JPEGQTables:519,JPEGDCTables:520,JPEGACTables:521,YCbCrCoefficients:529,YCbCrSubSampling:530,YCbCrPositioning:531,ReferenceBlackWhite:532,XMLPacket:700,Rating:18246,RatingPercent:18249,ImageID:32781,CFARepeatPatternDim:33421,CFAPattern:33422,BatteryLevel:33423,Copyright:33432,ExposureTime:33434,ImageResources:34377,ExifTag:34665,InterColorProfile:34675,GPSTag:34853,Interlace:34857,TimeZoneOffset:34858,SelfTimerMode:34859,FlashEnergy:37387,SpatialFrequencyResponse:37388,Noise:37389,FocalPlaneXResolution:37390,FocalPlaneYResolution:37391,FocalPlaneResolutionUnit:37392,ImageNumber:37393,SecurityClassification:37394,ImageHistory:37395,ExposureIndex:37397,TIFFEPStandardID:37398,SensingMethod:37399,XPTitle:40091,XPComment:40092,XPAuthor:40093,XPKeywords:40094,XPSubject:40095,PrintImageMatching:50341,DNGVersion:50706,DNGBackwardVersion:50707,UniqueCameraModel:50708,LocalizedCameraModel:50709,CFAPlaneColor:50710,CFALayout:50711,LinearizationTable:50712,BlackLevelRepeatDim:50713,BlackLevel:50714,BlackLevelDeltaH:50715,BlackLevelDeltaV:50716,WhiteLevel:50717,DefaultScale:50718,DefaultCropOrigin:50719,DefaultCropSize:50720,ColorMatrix1:50721,ColorMatrix2:50722,CameraCalibration1:50723,CameraCalibration2:50724,ReductionMatrix1:50725,ReductionMatrix2:50726,AnalogBalance:50727,AsShotNeutral:50728,AsShotWhiteXY:50729,BaselineExposure:50730,BaselineNoise:50731,BaselineSharpness:50732,BayerGreenSplit:50733,LinearResponseLimit:50734,CameraSerialNumber:50735,LensInfo:50736,ChromaBlurRadius:50737,AntiAliasStrength:50738,ShadowScale:50739,DNGPrivateData:50740,MakerNoteSafety:50741,CalibrationIlluminant1:50778,CalibrationIlluminant2:50779,BestQualityScale:50780,RawDataUniqueID:50781,OriginalRawFileName:50827,OriginalRawFileData:50828,ActiveArea:50829,MaskedAreas:50830,AsShotICCProfile:50831,AsShotPreProfileMatrix:50832,CurrentICCProfile:50833,CurrentPreProfileMatrix:50834,ColorimetricReference:50879,CameraCalibrationSignature:50931,ProfileCalibrationSignature:50932,AsShotProfileName:50934,NoiseReductionApplied:50935,ProfileName:50936,ProfileHueSatMapDims:50937,ProfileHueSatMapData1:50938,ProfileHueSatMapData2:50939,ProfileToneCurve:50940,ProfileEmbedPolicy:50941,ProfileCopyright:50942,ForwardMatrix1:50964,ForwardMatrix2:50965,PreviewApplicationName:50966,PreviewApplicationVersion:50967,PreviewSettingsName:50968,PreviewSettingsDigest:50969,PreviewColorSpace:50970,PreviewDateTime:50971,RawImageDigest:50972,OriginalRawFileDigest:50973,SubTileBlockSize:50974,RowInterleaveFactor:50975,ProfileLookTableDims:50981,ProfileLookTableData:50982,OpcodeList1:51008,OpcodeList2:51009,OpcodeList3:51022,NoiseProfile:51041},n.ExifIFD={ExposureTime:33434,FNumber:33437,ExposureProgram:34850,SpectralSensitivity:34852,ISOSpeedRatings:34855,OECF:34856,SensitivityType:34864,StandardOutputSensitivity:34865,RecommendedExposureIndex:34866,ISOSpeed:34867,ISOSpeedLatitudeyyy:34868,ISOSpeedLatitudezzz:34869,ExifVersion:36864,DateTimeOriginal:36867,DateTimeDigitized:36868,ComponentsConfiguration:37121,CompressedBitsPerPixel:37122,ShutterSpeedValue:37377,ApertureValue:37378,BrightnessValue:37379,ExposureBiasValue:37380,MaxApertureValue:37381,SubjectDistance:37382,MeteringMode:37383,LightSource:37384,Flash:37385,FocalLength:37386,SubjectArea:37396,MakerNote:37500,UserComment:37510,SubSecTime:37520,SubSecTimeOriginal:37521,SubSecTimeDigitized:37522,FlashpixVersion:40960,ColorSpace:40961,PixelXDimension:40962,PixelYDimension:40963,RelatedSoundFile:40964,InteroperabilityTag:40965,FlashEnergy:41483,SpatialFrequencyResponse:41484,FocalPlaneXResolution:41486,FocalPlaneYResolution:41487,FocalPlaneResolutionUnit:41488,SubjectLocation:41492,ExposureIndex:41493,SensingMethod:41495,FileSource:41728,SceneType:41729,CFAPattern:41730,CustomRendered:41985,ExposureMode:41986,WhiteBalance:41987,DigitalZoomRatio:41988,FocalLengthIn35mmFilm:41989,SceneCaptureType:41990,GainControl:41991,Contrast:41992,Saturation:41993,Sharpness:41994,DeviceSettingDescription:41995,SubjectDistanceRange:41996,ImageUniqueID:42016,CameraOwnerName:42032,BodySerialNumber:42033,LensSpecification:42034,LensMake:42035,LensModel:42036,LensSerialNumber:42037,Gamma:42240},n.GPSIFD={GPSVersionID:0,GPSLatitudeRef:1,GPSLatitude:2,GPSLongitudeRef:3,GPSLongitude:4,GPSAltitudeRef:5,GPSAltitude:6,GPSTimeStamp:7,GPSSatellites:8,GPSStatus:9,GPSMeasureMode:10,GPSDOP:11,GPSSpeedRef:12,GPSSpeed:13,GPSTrackRef:14,GPSTrack:15,GPSImgDirectionRef:16,GPSImgDirection:17,GPSMapDatum:18,GPSDestLatitudeRef:19,GPSDestLatitude:20,GPSDestLongitudeRef:21,GPSDestLongitude:22,GPSDestBearingRef:23,GPSDestBearing:24,GPSDestDistanceRef:25,GPSDestDistance:26,GPSProcessingMethod:27,GPSAreaInformation:28,GPSDateStamp:29,GPSDifferential:30,GPSHPositioningError:31},n.InteropIFD={InteroperabilityIndex:1},n.GPSHelper={degToDmsRational:function(e){var t=Math.abs(e),n=t%1*60,r=n%1*60;return[[Math.floor(t),1],[Math.floor(n),1],[Math.round(100*r),100]]},dmsRationalToDeg:function(e,t){var n="S"===t||"W"===t?-1:1;return(e[0][0]/e[0][1]+e[1][0]/e[1][1]/60+e[2][0]/e[2][1]/3600)*n}},e.exports&&(t=e.exports=n),t.piexif=n}()}));const Iq=async(e,t,n)=>{const r=await(e=>new Promise((t,n)=>{const r=new FileReader;r.onloadend=()=>t(r.result),r.readAsDataURL(e)}))(e),a=Sq.load(r),s=(await navigator.mediaDevices.enumerateDevices()).filter(e=>"videoinput"===e.kind).map(e=>e.label),i={width:n.width,height:n.height,usedCamera:t.label,camerasOnDevice:s},o={"0th":Object.assign(Object.assign({},a["0th"]),{[Sq.ImageIFD.Model]:i.usedCamera,[Sq.ImageIFD.ImageWidth]:i.width,[Sq.ImageIFD.ImageLength]:i.height}),Exif:Object.assign(Object.assign({},a.Exif),{[Sq.ExifIFD.UserComment]:JSON.stringify(i)}),GPS:Object.assign({},a.GPS),Interop:Object.assign({},a.Interop),"1st":Object.assign({},a["1st"]),thumbnail:a.thumbnail},u=Sq.dump(o),l=Sq.insert(u,r);return await fetch(l).then(e=>e.blob())},Aq=new Image;var Eq;Aq.src="data:image/svg+xml;base64,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",function(e){e.JPEG="image/jpeg",e.PNG="image/png"}(Eq||(Eq={}));class Nq{constructor(e,t){this.showMask=!1,this.videoSize={width:0,height:0},this.streamPaused=!1,this.pauseStream=()=>{this.streamPaused=!0,this.videoElement.pause(),this.dropMask(),this.faceDetection&&kq.getInstance().stopDetector()},this.device=e,this.detector=kq.getInstance(this,e.isMobile,t)}setShowMask(e){this.showMask=e}setFaceDetection(e){this.faceDetection=e}setProbabilityThreshold(e){kq.getInstance().setProbabilityThreshold(e)}setCallbackErrors(e){this.callbackErrors=e}setCallbackAutoCapturing(e){this.callbackAutoCapturing=e}setCallbackFaceDetectionErrors(e){this.callbackFaceDetectionErrors=e}static getInstance(e,t){return Nq.instance||(Nq.instance=new Nq(e,t)),Nq.instance}returnErrors(e){this.callbackFaceDetectionErrors(e)}autoCapturing(){this.callbackAutoCapturing()}updateHtmlElements(e,t){this.videoElement=e,this.canvasElement=t,this.detector.updateHtmlElements(e)}async startDetection(){return this.detector.startDetector()}static orientationChange(){Nq.instance&&Nq.getInstance().drawMask()}startStream(e){this.stream&&this.stream.getTracks().forEach(e=>e.stop()),this.stream=e,"srcObject"in this.videoElement?this.videoElement.srcObject=e:this.videoElement.src=window.URL.createObjectURL(e),this.videoElement.play().then(()=>{this.streamPaused=!1,l.cameraReadyEvent(),this.drawMask(),this.faceDetection&&this.startDetection()})}async initStream(e){this.startStream(e)}async resumeStream(){this.streamPaused=!1,await this.videoElement.play(),this.drawMask(),this.faceDetection&&await kq.getInstance().startDetector()}updateCanvasSize(e){this.videoSize={width:this.videoElement.videoWidth,height:this.videoElement.videoHeight},e.width=this.videoSize.width,e.height=this.videoSize.height}dropStream(){this.streamStopped()||(this.stream.getTracks().forEach(e=>e.stop()),this.videoElement.srcObject=null),this.faceDetection&&kq.getInstance().stopDetector()}streamStopped(){return!(this.stream&&this.stream.getTracks&&this.stream.getTracks().length>0)}drawMask(){this.showMask&&!this.streamPaused&&setTimeout(()=>{this.updateCanvasSize(this.canvasElement);const e=this.canvasElement,t=e.getContext("2d"),n=e.width/Aq.width,r=e.height/Aq.height,a=Math.min(n,r),s=e.width<e.height,i=!s&&this.device.isMobile?50:100,o=s?200:100,u=(e.width-Aq.width*a)/2,l=(e.height-Aq.height*a)/2;t.drawImage(Aq,0,0,Aq.width,Aq.height,u+i,l+o,Aq.width*a-2*i,Aq.height*a-2*o)},150)}dropMask(){this.canvasElement.getContext("2d").clearRect(0,0,1e4,1e4)}takePhoto(){return new Promise(async e=>{const t=document.createElement("canvas");t.width=this.videoElement.videoWidth,t.height=this.videoElement.videoHeight,e([await this.getFrame(t)])})}getFrame(e){return new Promise(t=>{e.getContext("2d").drawImage(this.videoElement,0,0,e.width,e.height),e.toBlob(e=>{if(e.type!==Eq.JPEG||this.device.isIos)t(e);else try{Iq(e,this.stream.getTracks()[0],this.videoSize).then(e=>t(e))}catch(n){t(e),this.callbackErrors(n,!1)}},Eq.PNG,1)})}}window.addEventListener("resize",Nq.orientationChange,!1),window.addEventListener("orientationchange",Nq.orientationChange,!1);let Cq=class{constructor(e){Object(u.f)(this,e),this.eventVideoStarted=Object(u.b)(this,"videoStarted",7),this.eventCloseCamera=Object(u.b)(this,"closeCamera",7),this.errorCameraEvent=Object(u.b)(this,"errorCamera",7),this.eventMakePhoto=Object(u.b)(this,"makePhoto",7),this.eventTakePhoto=Object(u.b)(this,"takePhoto",7),this.callbackErrors=(e,t)=>{t?(this.errorCameraEvent.emit(e),this.eventCloseCamera.emit()):this.errorCameraEvent.emit(e)},this.callbackAutoCapturing=()=>{this.device.isMobile?this.eventMakePhoto.emit():this.eventTakePhoto.emit()},this.callbackFaceDetectionErrors=e=>{l.detectionError(e)}}componentDidLoad(){this.startStream()}render(){const e="camera "+(this.device.isMobile?"cameraMobile":"");return Object(u.d)("div",{class:e},Object(u.d)("video",{loop:!0,autoplay:!0,playsinline:!0,muted:!0,class:"cameraVideo",ref:e=>this.cameraVideo=e}),Object(u.d)("canvas",{class:"cameraCanvas",ref:e=>this.cameraCanvas=e}))}startStream(){Nq.instance||Nq.getInstance(this.device,this.modelPath);const e=Nq.getInstance();e.updateHtmlElements(this.cameraVideo,this.cameraCanvas),e.setShowMask(this.showMask),e.setFaceDetection(this.faceDetection),e.setProbabilityThreshold(this.probabilityThreshold),e.setCallbackErrors(this.callbackErrors),e.setCallbackAutoCapturing(this.callbackAutoCapturing),e.setCallbackFaceDetectionErrors(this.callbackFaceDetectionErrors)}};Cq.style=".camera{width:100%;height:100%;color:white;display:flex;align-items:center;justify-content:center;position:relative}.cameraCanvas{transform:scale(-1, 1);-webkit-transform:scale(-1, 1);max-width:100%;max-height:100%;z-index:2}.cameraVideo{transform:scale(-1, 1);-webkit-transform:scale(-1, 1);z-index:1;position:absolute;max-width:100%;max-height:100%}.cameraMobile{position:fixed;top:0;left:0;background:black}.canvas-on-video{max-width:100%;max-height:100%;position:absolute;left:50%;top:50%;z-index:2;transform:scale(-1, 1)}";let Tq=class{constructor(e){Object(u.f)(this,e),this.updateState=()=>{this._showMask="true"===this.show_mask,this._disable_control_panel="true"===this.disable_control_panel,this._face_detection="true"===this.face_detection,this._stopAfterCapturing="true"===this.stop_after_capturing,this._probabilityThreshold=this.probability_threshold?+this.probability_threshold:50,this.url_logo=this.url_logo?this.url_logo:"data:image/png;charset=utf-8;base64,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",this._debug="true"===this.debug,kq.debug=this._debug},this.photoIsReady=e=>{this._stopAfterCapturing&&this.closeCamera(),l.captureEvent(e)},this.enableButton=()=>{this._resolutionOnPhoto=!0},this.updateState(),this._cameraStatus=0,this._resolutionOnPhoto=!1,this._device=xq(),this._mobile=this._device.isMobile,this._mobileMakePhoto=!1}videoStarted(){this._resolutionOnPhoto=!0}async retakePhoto(){this._mobileMakePhoto=!1,await Nq.getInstance().resumeStream()}takePhoto(){Nq.getInstance().takePhoto().then(e=>{this.photoIsReady(e)})}async openCamera(){this._cameraStatus=1;const e=this._device.isMac?{audio:!1,video:{facingMode:"user",width:1920,height:1080}}:{audio:!1,video:{facingMode:"user",width:{ideal:1920},height:{ideal:1080}}};setTimeout(()=>{navigator.mediaDevices.getUserMedia(e).then(e=>{Nq.getInstance().initStream(e),this.videoStarted()}).catch(e=>{this.closeCamera(),l.errorEvent(e)})},100)}async loadModels(e){await kq.initDetector(e.detail.path)}errorCamera(e){l.errorEvent(e.detail.message)}closeCamera(){Nq.instance&&(this._cameraStatus=0,l.closeEvent(),Nq.getInstance().dropStream(),this._mobileMakePhoto=!1,this._resolutionOnPhoto=!1)}makePhoto(){this._mobile?(this._mobileMakePhoto=!0,Nq.getInstance().pauseStream()):this.takePhoto()}componentWillLoad(){l.init(this.component),navigator.mediaDevices||l.errorEvent("This browser does not support webRTC")}componentDidUpdate(){this.updateState()}render(){let e;return e=this._cameraStatus?Object(u.d)("camera-comp",{class:"block",showMask:this._showMask,device:this._device,faceDetection:this._face_detection,modelPath:this.model_path,probabilityThreshold:this._probabilityThreshold}):Object(u.d)("img",{src:this.url_logo,class:"logo",style:this.logo_style?JSON.parse(this.logo_style):"",alt:"logo"}),Object(u.d)("div",{class:"cameraContainer",id:"cameraContainer",style:{backgroundColor:this.background_color}},Object(u.d)("div",{class:"wrapperCamera",style:{height:this._disable_control_panel?"100%":"65%"}},e),Object(u.d)("control-panel",{class:"block",mobile:this._mobile,disableControlPanel:this._disable_control_panel,cameraStatus:this._cameraStatus,resolutionOnPhoto:this._resolutionOnPhoto,mobileMakePhoto:this._mobileMakePhoto,faceDetection:this._face_detection}))}get component(){return Object(u.c)(this)}};Tq.style=".cameraContainer{width:100%;height:100%;z-index:10;position:relative;overflow:hidden}.logo{max-height:450px;max-width:450px}.wrapperCamera{display:flex;justify-content:center;align-items:center}.block{width:100%;height:100%}@media screen and (orientation: portrait){.history{width:100vw;font-size:20px}@media screen and (max-height: 667px){.logo{background-size:120px;max-height:120px;max-width:120px}}@media screen and (min-height: 668px) and (max-height: 812px){.logo{background-size:140px;max-height:140px;max-width:140px}}@media (max-height: 667px){.logo{margin:0 auto}}}@media screen and (max-height: 414px){.logo{background-size:140px;max-height:140px;max-width:140px}}";const Rq=(e,t)=>[...t];let _q=class{constructor(e){Object(u.f)(this,e),this.eventOpenCamera=Object(u.b)(this,"openCamera",7),this.eventCloseCamera=Object(u.b)(this,"closeCamera",7),this.eventMakePhoto=Object(u.b)(this,"makePhoto",7),this.eventRetakePhoto=Object(u.b)(this,"retakePhoto",7),this.eventTakePhoto=Object(u.b)(this,"takePhoto",7),this.makePhoto=()=>{this.eventMakePhoto.emit()},this.closeCamera=()=>{this.eventCloseCamera.emit()},this.openCamera=()=>{this.eventOpenCamera.emit()},this.takePhoto=()=>{this.eventTakePhoto.emit()},this.retakePhoto=async()=>{this.eventRetakePhoto.emit()}}render(){return Object(u.d)("div",{class:"controlPanel"+(this.mobile?" controlMobile":"")},this.mobile?Object(u.d)("div",null,this.cameraStatus?Object(u.d)("div",null,this.mobileMakePhoto?Object(u.d)("div",null,Object(u.d)("div",{onClick:this.retakePhoto,class:"retake"},"retake"),Object(u.d)("img",{onClick:this.takePhoto,alt:"use this photo",class:"shoot takePhoto",src:"data:image/svg+xml;base64,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"})):Object(u.d)("button",{onClick:this.makePhoto,disabled:!this.resolutionOnPhoto,class:"shoot"}),Object(u.d)("div",{onClick:this.closeCamera,class:"close"},"close")):Object(u.d)("div",null,this.disableControlPanel?Object(u.d)("div",null):Object(u.d)("button",{onClick:this.openCamera},"Turn on camera"))):Object(u.d)(Rq,null,this.disableControlPanel?Object(u.d)("div",null):Object(u.d)(Rq,null,this.cameraStatus?Object(u.d)(Rq,null,Object(u.d)("button",{onClick:this.makePhoto,disabled:!this.resolutionOnPhoto},"Capture"),Object(u.d)("button",{onClick:this.closeCamera,class:"cameraOff"},"Close camera")):Object(u.d)("button",{onClick:this.openCamera},"Open camera"))))}};_q.style='.controlPanel{height:35%;display:flex;justify-content:space-evenly;align-items:center;width:100%;max-width:450px;margin:auto}.buttons{width:100%;max-width:450px;display:flex;justify-content:space-evenly;align-items:center}button{width:164px;height:50px;border:none;margin-bottom:10px;-webkit-border-radius:10px;-moz-border-radius:10px;border-radius:10px;color:#fff;font-family:"Avenir", Helvetica, sans-serif;font-weight:600;font-size:16px;background-color:#4ca585;outline:none;cursor:pointer}button[disabled]{background-color:#aaa;cursor:not-allowed;pointer-events:none}.cameraOff{background-color:#667a87}.cameraOff:hover{background-color:#92a8b5}.shoot{width:50px;height:50px;border-radius:25px;border:2px solid white;outline:none;background-color:red;position:fixed;left:calc(50% - 25px);bottom:10px}.shoot[disabled]{background-color:#aaa;cursor:not-allowed;pointer-events:none}.close{position:fixed;right:6vw;top:2vh;text-shadow:0 0 4px black;color:white;font-size:20px;user-select:none}.retake{position:fixed;right:6vw;bottom:35px;text-shadow:0 0 4px black;color:white;font-size:20px;user-select:none}.takePhoto{width:46px;height:46px;bottom:20px;background:none}@media screen and (orientation:landscape) and (min-width: 360px){.shoot{bottom:calc(50% - 25px);left:auto;right:5vw}}'}.call(this,"/index.js",n(124),"/",n(33),n(238).Buffer,n(242).setImmediate)},function(e,t,n){"use strict";(function(e){
-/*!
- * The buffer module from node.js, for the browser.
- *
- * @author   Feross Aboukhadijeh <http://feross.org>
- * @license  MIT
- */
-var r=n(239),a=n(240),s=n(241);function i(){return u.TYPED_ARRAY_SUPPORT?2147483647:1073741823}function o(e,t){if(i()<t)throw new RangeError("Invalid typed array length");return u.TYPED_ARRAY_SUPPORT?(e=new Uint8Array(t)).__proto__=u.prototype:(null===e&&(e=new u(t)),e.length=t),e}function u(e,t,n){if(!(u.TYPED_ARRAY_SUPPORT||this instanceof u))return new u(e,t,n);if("number"==typeof e){if("string"==typeof t)throw new Error("If encoding is specified then the first argument must be a string");return h(this,e)}return l(this,e,t,n)}function l(e,t,n,r){if("number"==typeof t)throw new TypeError('"value" argument must not be a number');return"undefined"!=typeof ArrayBuffer&&t instanceof ArrayBuffer?function(e,t,n,r){if(t.byteLength,n<0||t.byteLength<n)throw new RangeError("'offset' is out of bounds");if(t.byteLength<n+(r||0))throw new RangeError("'length' is out of bounds");t=void 0===n&&void 0===r?new Uint8Array(t):void 0===r?new Uint8Array(t,n):new Uint8Array(t,n,r);u.TYPED_ARRAY_SUPPORT?(e=t).__proto__=u.prototype:e=p(e,t);return e}(e,t,n,r):"string"==typeof t?function(e,t,n){"string"==typeof n&&""!==n||(n="utf8");if(!u.isEncoding(n))throw new TypeError('"encoding" must be a valid string encoding');var r=0|f(t,n),a=(e=o(e,r)).write(t,n);a!==r&&(e=e.slice(0,a));return e}(e,t,n):function(e,t){if(u.isBuffer(t)){var n=0|d(t.length);return 0===(e=o(e,n)).length||t.copy(e,0,0,n),e}if(t){if("undefined"!=typeof ArrayBuffer&&t.buffer instanceof ArrayBuffer||"length"in t)return"number"!=typeof t.length||(r=t.length)!=r?o(e,0):p(e,t);if("Buffer"===t.type&&s(t.data))return p(e,t.data)}var r;throw new TypeError("First argument must be a string, Buffer, ArrayBuffer, Array, or array-like object.")}(e,t)}function c(e){if("number"!=typeof e)throw new TypeError('"size" argument must be a number');if(e<0)throw new RangeError('"size" argument must not be negative')}function h(e,t){if(c(t),e=o(e,t<0?0:0|d(t)),!u.TYPED_ARRAY_SUPPORT)for(var n=0;n<t;++n)e[n]=0;return e}function p(e,t){var n=t.length<0?0:0|d(t.length);e=o(e,n);for(var r=0;r<n;r+=1)e[r]=255&t[r];return e}function d(e){if(e>=i())throw new RangeError("Attempt to allocate Buffer larger than maximum size: 0x"+i().toString(16)+" bytes");return 0|e}function f(e,t){if(u.isBuffer(e))return e.length;if("undefined"!=typeof ArrayBuffer&&"function"==typeof ArrayBuffer.isView&&(ArrayBuffer.isView(e)||e instanceof ArrayBuffer))return e.byteLength;"string"!=typeof e&&(e=""+e);var n=e.length;if(0===n)return 0;for(var r=!1;;)switch(t){case"ascii":case"latin1":case"binary":return n;case"utf8":case"utf-8":case void 0:return z(e).length;case"ucs2":case"ucs-2":case"utf16le":case"utf-16le":return 2*n;case"hex":return n>>>1;case"base64":return U(e).length;default:if(r)return z(e).length;t=(""+t).toLowerCase(),r=!0}}function m(e,t,n){var r=!1;if((void 0===t||t<0)&&(t=0),t>this.length)return"";if((void 0===n||n>this.length)&&(n=this.length),n<=0)return"";if((n>>>=0)<=(t>>>=0))return"";for(e||(e="utf8");;)switch(e){case"hex":return T(this,t,n);case"utf8":case"utf-8":return E(this,t,n);case"ascii":return N(this,t,n);case"latin1":case"binary":return C(this,t,n);case"base64":return A(this,t,n);case"ucs2":case"ucs-2":case"utf16le":case"utf-16le":return R(this,t,n);default:if(r)throw new TypeError("Unknown encoding: "+e);e=(e+"").toLowerCase(),r=!0}}function g(e,t,n){var r=e[t];e[t]=e[n],e[n]=r}function y(e,t,n,r,a){if(0===e.length)return-1;if("string"==typeof n?(r=n,n=0):n>2147483647?n=2147483647:n<-2147483648&&(n=-2147483648),n=+n,isNaN(n)&&(n=a?0:e.length-1),n<0&&(n=e.length+n),n>=e.length){if(a)return-1;n=e.length-1}else if(n<0){if(!a)return-1;n=0}if("string"==typeof t&&(t=u.from(t,r)),u.isBuffer(t))return 0===t.length?-1:b(e,t,n,r,a);if("number"==typeof t)return t&=255,u.TYPED_ARRAY_SUPPORT&&"function"==typeof Uint8Array.prototype.indexOf?a?Uint8Array.prototype.indexOf.call(e,t,n):Uint8Array.prototype.lastIndexOf.call(e,t,n):b(e,[t],n,r,a);throw new TypeError("val must be string, number or Buffer")}function b(e,t,n,r,a){var s,i=1,o=e.length,u=t.length;if(void 0!==r&&("ucs2"===(r=String(r).toLowerCase())||"ucs-2"===r||"utf16le"===r||"utf-16le"===r)){if(e.length<2||t.length<2)return-1;i=2,o/=2,u/=2,n/=2}function l(e,t){return 1===i?e[t]:e.readUInt16BE(t*i)}if(a){var c=-1;for(s=n;s<o;s++)if(l(e,s)===l(t,-1===c?0:s-c)){if(-1===c&&(c=s),s-c+1===u)return c*i}else-1!==c&&(s-=s-c),c=-1}else for(n+u>o&&(n=o-u),s=n;s>=0;s--){for(var h=!0,p=0;p<u;p++)if(l(e,s+p)!==l(t,p)){h=!1;break}if(h)return s}return-1}function v(e,t,n,r){n=Number(n)||0;var a=e.length-n;r?(r=Number(r))>a&&(r=a):r=a;var s=t.length;if(s%2!=0)throw new TypeError("Invalid hex string");r>s/2&&(r=s/2);for(var i=0;i<r;++i){var o=parseInt(t.substr(2*i,2),16);if(isNaN(o))return i;e[n+i]=o}return i}function x(e,t,n,r){return W(z(t,e.length-n),e,n,r)}function w(e,t,n,r){return W(function(e){for(var t=[],n=0;n<e.length;++n)t.push(255&e.charCodeAt(n));return t}(t),e,n,r)}function k(e,t,n,r){return w(e,t,n,r)}function S(e,t,n,r){return W(U(t),e,n,r)}function I(e,t,n,r){return W(function(e,t){for(var n,r,a,s=[],i=0;i<e.length&&!((t-=2)<0);++i)n=e.charCodeAt(i),r=n>>8,a=n%256,s.push(a),s.push(r);return s}(t,e.length-n),e,n,r)}function A(e,t,n){return 0===t&&n===e.length?r.fromByteArray(e):r.fromByteArray(e.slice(t,n))}function E(e,t,n){n=Math.min(e.length,n);for(var r=[],a=t;a<n;){var s,i,o,u,l=e[a],c=null,h=l>239?4:l>223?3:l>191?2:1;if(a+h<=n)switch(h){case 1:l<128&&(c=l);break;case 2:128==(192&(s=e[a+1]))&&(u=(31&l)<<6|63&s)>127&&(c=u);break;case 3:s=e[a+1],i=e[a+2],128==(192&s)&&128==(192&i)&&(u=(15&l)<<12|(63&s)<<6|63&i)>2047&&(u<55296||u>57343)&&(c=u);break;case 4:s=e[a+1],i=e[a+2],o=e[a+3],128==(192&s)&&128==(192&i)&&128==(192&o)&&(u=(15&l)<<18|(63&s)<<12|(63&i)<<6|63&o)>65535&&u<1114112&&(c=u)}null===c?(c=65533,h=1):c>65535&&(c-=65536,r.push(c>>>10&1023|55296),c=56320|1023&c),r.push(c),a+=h}return function(e){var t=e.length;if(t<=4096)return String.fromCharCode.apply(String,e);var n="",r=0;for(;r<t;)n+=String.fromCharCode.apply(String,e.slice(r,r+=4096));return n}(r)}t.Buffer=u,t.SlowBuffer=function(e){+e!=e&&(e=0);return u.alloc(+e)},t.INSPECT_MAX_BYTES=50,u.TYPED_ARRAY_SUPPORT=void 0!==e.TYPED_ARRAY_SUPPORT?e.TYPED_ARRAY_SUPPORT:function(){try{var e=new Uint8Array(1);return e.__proto__={__proto__:Uint8Array.prototype,foo:function(){return 42}},42===e.foo()&&"function"==typeof e.subarray&&0===e.subarray(1,1).byteLength}catch(e){return!1}}(),t.kMaxLength=i(),u.poolSize=8192,u._augment=function(e){return e.__proto__=u.prototype,e},u.from=function(e,t,n){return l(null,e,t,n)},u.TYPED_ARRAY_SUPPORT&&(u.prototype.__proto__=Uint8Array.prototype,u.__proto__=Uint8Array,"undefined"!=typeof Symbol&&Symbol.species&&u[Symbol.species]===u&&Object.defineProperty(u,Symbol.species,{value:null,configurable:!0})),u.alloc=function(e,t,n){return function(e,t,n,r){return c(t),t<=0?o(e,t):void 0!==n?"string"==typeof r?o(e,t).fill(n,r):o(e,t).fill(n):o(e,t)}(null,e,t,n)},u.allocUnsafe=function(e){return h(null,e)},u.allocUnsafeSlow=function(e){return h(null,e)},u.isBuffer=function(e){return!(null==e||!e._isBuffer)},u.compare=function(e,t){if(!u.isBuffer(e)||!u.isBuffer(t))throw new TypeError("Arguments must be Buffers");if(e===t)return 0;for(var n=e.length,r=t.length,a=0,s=Math.min(n,r);a<s;++a)if(e[a]!==t[a]){n=e[a],r=t[a];break}return n<r?-1:r<n?1:0},u.isEncoding=function(e){switch(String(e).toLowerCase()){case"hex":case"utf8":case"utf-8":case"ascii":case"latin1":case"binary":case"base64":case"ucs2":case"ucs-2":case"utf16le":case"utf-16le":return!0;default:return!1}},u.concat=function(e,t){if(!s(e))throw new TypeError('"list" argument must be an Array of Buffers');if(0===e.length)return u.alloc(0);var n;if(void 0===t)for(t=0,n=0;n<e.length;++n)t+=e[n].length;var r=u.allocUnsafe(t),a=0;for(n=0;n<e.length;++n){var i=e[n];if(!u.isBuffer(i))throw new TypeError('"list" argument must be an Array of Buffers');i.copy(r,a),a+=i.length}return r},u.byteLength=f,u.prototype._isBuffer=!0,u.prototype.swap16=function(){var e=this.length;if(e%2!=0)throw new RangeError("Buffer size must be a multiple of 16-bits");for(var t=0;t<e;t+=2)g(this,t,t+1);return this},u.prototype.swap32=function(){var e=this.length;if(e%4!=0)throw new RangeError("Buffer size must be a multiple of 32-bits");for(var t=0;t<e;t+=4)g(this,t,t+3),g(this,t+1,t+2);return this},u.prototype.swap64=function(){var e=this.length;if(e%8!=0)throw new RangeError("Buffer size must be a multiple of 64-bits");for(var t=0;t<e;t+=8)g(this,t,t+7),g(this,t+1,t+6),g(this,t+2,t+5),g(this,t+3,t+4);return this},u.prototype.toString=function(){var e=0|this.length;return 0===e?"":0===arguments.length?E(this,0,e):m.apply(this,arguments)},u.prototype.equals=function(e){if(!u.isBuffer(e))throw new TypeError("Argument must be a Buffer");return this===e||0===u.compare(this,e)},u.prototype.inspect=function(){var e="",n=t.INSPECT_MAX_BYTES;return this.length>0&&(e=this.toString("hex",0,n).match(/.{2}/g).join(" "),this.length>n&&(e+=" ... ")),"<Buffer "+e+">"},u.prototype.compare=function(e,t,n,r,a){if(!u.isBuffer(e))throw new TypeError("Argument must be a Buffer");if(void 0===t&&(t=0),void 0===n&&(n=e?e.length:0),void 0===r&&(r=0),void 0===a&&(a=this.length),t<0||n>e.length||r<0||a>this.length)throw new RangeError("out of range index");if(r>=a&&t>=n)return 0;if(r>=a)return-1;if(t>=n)return 1;if(this===e)return 0;for(var s=(a>>>=0)-(r>>>=0),i=(n>>>=0)-(t>>>=0),o=Math.min(s,i),l=this.slice(r,a),c=e.slice(t,n),h=0;h<o;++h)if(l[h]!==c[h]){s=l[h],i=c[h];break}return s<i?-1:i<s?1:0},u.prototype.includes=function(e,t,n){return-1!==this.indexOf(e,t,n)},u.prototype.indexOf=function(e,t,n){return y(this,e,t,n,!0)},u.prototype.lastIndexOf=function(e,t,n){return y(this,e,t,n,!1)},u.prototype.write=function(e,t,n,r){if(void 0===t)r="utf8",n=this.length,t=0;else if(void 0===n&&"string"==typeof t)r=t,n=this.length,t=0;else{if(!isFinite(t))throw new Error("Buffer.write(string, encoding, offset[, length]) is no longer supported");t|=0,isFinite(n)?(n|=0,void 0===r&&(r="utf8")):(r=n,n=void 0)}var a=this.length-t;if((void 0===n||n>a)&&(n=a),e.length>0&&(n<0||t<0)||t>this.length)throw new RangeError("Attempt to write outside buffer bounds");r||(r="utf8");for(var s=!1;;)switch(r){case"hex":return v(this,e,t,n);case"utf8":case"utf-8":return x(this,e,t,n);case"ascii":return w(this,e,t,n);case"latin1":case"binary":return k(this,e,t,n);case"base64":return S(this,e,t,n);case"ucs2":case"ucs-2":case"utf16le":case"utf-16le":return I(this,e,t,n);default:if(s)throw new TypeError("Unknown encoding: "+r);r=(""+r).toLowerCase(),s=!0}},u.prototype.toJSON=function(){return{type:"Buffer",data:Array.prototype.slice.call(this._arr||this,0)}};function N(e,t,n){var r="";n=Math.min(e.length,n);for(var a=t;a<n;++a)r+=String.fromCharCode(127&e[a]);return r}function C(e,t,n){var r="";n=Math.min(e.length,n);for(var a=t;a<n;++a)r+=String.fromCharCode(e[a]);return r}function T(e,t,n){var r=e.length;(!t||t<0)&&(t=0),(!n||n<0||n>r)&&(n=r);for(var a="",s=t;s<n;++s)a+=B(e[s]);return a}function R(e,t,n){for(var r=e.slice(t,n),a="",s=0;s<r.length;s+=2)a+=String.fromCharCode(r[s]+256*r[s+1]);return a}function _(e,t,n){if(e%1!=0||e<0)throw new RangeError("offset is not uint");if(e+t>n)throw new RangeError("Trying to access beyond buffer length")}function F(e,t,n,r,a,s){if(!u.isBuffer(e))throw new TypeError('"buffer" argument must be a Buffer instance');if(t>a||t<s)throw new RangeError('"value" argument is out of bounds');if(n+r>e.length)throw new RangeError("Index out of range")}function M(e,t,n,r){t<0&&(t=65535+t+1);for(var a=0,s=Math.min(e.length-n,2);a<s;++a)e[n+a]=(t&255<<8*(r?a:1-a))>>>8*(r?a:1-a)}function D(e,t,n,r){t<0&&(t=4294967295+t+1);for(var a=0,s=Math.min(e.length-n,4);a<s;++a)e[n+a]=t>>>8*(r?a:3-a)&255}function O(e,t,n,r,a,s){if(n+r>e.length)throw new RangeError("Index out of range");if(n<0)throw new RangeError("Index out of range")}function L(e,t,n,r,s){return s||O(e,0,n,4),a.write(e,t,n,r,23,4),n+4}function P(e,t,n,r,s){return s||O(e,0,n,8),a.write(e,t,n,r,52,8),n+8}u.prototype.slice=function(e,t){var n,r=this.length;if((e=~~e)<0?(e+=r)<0&&(e=0):e>r&&(e=r),(t=void 0===t?r:~~t)<0?(t+=r)<0&&(t=0):t>r&&(t=r),t<e&&(t=e),u.TYPED_ARRAY_SUPPORT)(n=this.subarray(e,t)).__proto__=u.prototype;else{var a=t-e;n=new u(a,void 0);for(var s=0;s<a;++s)n[s]=this[s+e]}return n},u.prototype.readUIntLE=function(e,t,n){e|=0,t|=0,n||_(e,t,this.length);for(var r=this[e],a=1,s=0;++s<t&&(a*=256);)r+=this[e+s]*a;return r},u.prototype.readUIntBE=function(e,t,n){e|=0,t|=0,n||_(e,t,this.length);for(var r=this[e+--t],a=1;t>0&&(a*=256);)r+=this[e+--t]*a;return r},u.prototype.readUInt8=function(e,t){return t||_(e,1,this.length),this[e]},u.prototype.readUInt16LE=function(e,t){return t||_(e,2,this.length),this[e]|this[e+1]<<8},u.prototype.readUInt16BE=function(e,t){return t||_(e,2,this.length),this[e]<<8|this[e+1]},u.prototype.readUInt32LE=function(e,t){return t||_(e,4,this.length),(this[e]|this[e+1]<<8|this[e+2]<<16)+16777216*this[e+3]},u.prototype.readUInt32BE=function(e,t){return t||_(e,4,this.length),16777216*this[e]+(this[e+1]<<16|this[e+2]<<8|this[e+3])},u.prototype.readIntLE=function(e,t,n){e|=0,t|=0,n||_(e,t,this.length);for(var r=this[e],a=1,s=0;++s<t&&(a*=256);)r+=this[e+s]*a;return r>=(a*=128)&&(r-=Math.pow(2,8*t)),r},u.prototype.readIntBE=function(e,t,n){e|=0,t|=0,n||_(e,t,this.length);for(var r=t,a=1,s=this[e+--r];r>0&&(a*=256);)s+=this[e+--r]*a;return s>=(a*=128)&&(s-=Math.pow(2,8*t)),s},u.prototype.readInt8=function(e,t){return t||_(e,1,this.length),128&this[e]?-1*(255-this[e]+1):this[e]},u.prototype.readInt16LE=function(e,t){t||_(e,2,this.length);var n=this[e]|this[e+1]<<8;return 32768&n?4294901760|n:n},u.prototype.readInt16BE=function(e,t){t||_(e,2,this.length);var n=this[e+1]|this[e]<<8;return 32768&n?4294901760|n:n},u.prototype.readInt32LE=function(e,t){return t||_(e,4,this.length),this[e]|this[e+1]<<8|this[e+2]<<16|this[e+3]<<24},u.prototype.readInt32BE=function(e,t){return t||_(e,4,this.length),this[e]<<24|this[e+1]<<16|this[e+2]<<8|this[e+3]},u.prototype.readFloatLE=function(e,t){return t||_(e,4,this.length),a.read(this,e,!0,23,4)},u.prototype.readFloatBE=function(e,t){return t||_(e,4,this.length),a.read(this,e,!1,23,4)},u.prototype.readDoubleLE=function(e,t){return t||_(e,8,this.length),a.read(this,e,!0,52,8)},u.prototype.readDoubleBE=function(e,t){return t||_(e,8,this.length),a.read(this,e,!1,52,8)},u.prototype.writeUIntLE=function(e,t,n,r){(e=+e,t|=0,n|=0,r)||F(this,e,t,n,Math.pow(2,8*n)-1,0);var a=1,s=0;for(this[t]=255&e;++s<n&&(a*=256);)this[t+s]=e/a&255;return t+n},u.prototype.writeUIntBE=function(e,t,n,r){(e=+e,t|=0,n|=0,r)||F(this,e,t,n,Math.pow(2,8*n)-1,0);var a=n-1,s=1;for(this[t+a]=255&e;--a>=0&&(s*=256);)this[t+a]=e/s&255;return t+n},u.prototype.writeUInt8=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,1,255,0),u.TYPED_ARRAY_SUPPORT||(e=Math.floor(e)),this[t]=255&e,t+1},u.prototype.writeUInt16LE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,2,65535,0),u.TYPED_ARRAY_SUPPORT?(this[t]=255&e,this[t+1]=e>>>8):M(this,e,t,!0),t+2},u.prototype.writeUInt16BE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,2,65535,0),u.TYPED_ARRAY_SUPPORT?(this[t]=e>>>8,this[t+1]=255&e):M(this,e,t,!1),t+2},u.prototype.writeUInt32LE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,4,4294967295,0),u.TYPED_ARRAY_SUPPORT?(this[t+3]=e>>>24,this[t+2]=e>>>16,this[t+1]=e>>>8,this[t]=255&e):D(this,e,t,!0),t+4},u.prototype.writeUInt32BE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,4,4294967295,0),u.TYPED_ARRAY_SUPPORT?(this[t]=e>>>24,this[t+1]=e>>>16,this[t+2]=e>>>8,this[t+3]=255&e):D(this,e,t,!1),t+4},u.prototype.writeIntLE=function(e,t,n,r){if(e=+e,t|=0,!r){var a=Math.pow(2,8*n-1);F(this,e,t,n,a-1,-a)}var s=0,i=1,o=0;for(this[t]=255&e;++s<n&&(i*=256);)e<0&&0===o&&0!==this[t+s-1]&&(o=1),this[t+s]=(e/i>>0)-o&255;return t+n},u.prototype.writeIntBE=function(e,t,n,r){if(e=+e,t|=0,!r){var a=Math.pow(2,8*n-1);F(this,e,t,n,a-1,-a)}var s=n-1,i=1,o=0;for(this[t+s]=255&e;--s>=0&&(i*=256);)e<0&&0===o&&0!==this[t+s+1]&&(o=1),this[t+s]=(e/i>>0)-o&255;return t+n},u.prototype.writeInt8=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,1,127,-128),u.TYPED_ARRAY_SUPPORT||(e=Math.floor(e)),e<0&&(e=255+e+1),this[t]=255&e,t+1},u.prototype.writeInt16LE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,2,32767,-32768),u.TYPED_ARRAY_SUPPORT?(this[t]=255&e,this[t+1]=e>>>8):M(this,e,t,!0),t+2},u.prototype.writeInt16BE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,2,32767,-32768),u.TYPED_ARRAY_SUPPORT?(this[t]=e>>>8,this[t+1]=255&e):M(this,e,t,!1),t+2},u.prototype.writeInt32LE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,4,2147483647,-2147483648),u.TYPED_ARRAY_SUPPORT?(this[t]=255&e,this[t+1]=e>>>8,this[t+2]=e>>>16,this[t+3]=e>>>24):D(this,e,t,!0),t+4},u.prototype.writeInt32BE=function(e,t,n){return e=+e,t|=0,n||F(this,e,t,4,2147483647,-2147483648),e<0&&(e=4294967295+e+1),u.TYPED_ARRAY_SUPPORT?(this[t]=e>>>24,this[t+1]=e>>>16,this[t+2]=e>>>8,this[t+3]=255&e):D(this,e,t,!1),t+4},u.prototype.writeFloatLE=function(e,t,n){return L(this,e,t,!0,n)},u.prototype.writeFloatBE=function(e,t,n){return L(this,e,t,!1,n)},u.prototype.writeDoubleLE=function(e,t,n){return P(this,e,t,!0,n)},u.prototype.writeDoubleBE=function(e,t,n){return P(this,e,t,!1,n)},u.prototype.copy=function(e,t,n,r){if(n||(n=0),r||0===r||(r=this.length),t>=e.length&&(t=e.length),t||(t=0),r>0&&r<n&&(r=n),r===n)return 0;if(0===e.length||0===this.length)return 0;if(t<0)throw new RangeError("targetStart out of bounds");if(n<0||n>=this.length)throw new RangeError("sourceStart out of bounds");if(r<0)throw new RangeError("sourceEnd out of bounds");r>this.length&&(r=this.length),e.length-t<r-n&&(r=e.length-t+n);var a,s=r-n;if(this===e&&n<t&&t<r)for(a=s-1;a>=0;--a)e[a+t]=this[a+n];else if(s<1e3||!u.TYPED_ARRAY_SUPPORT)for(a=0;a<s;++a)e[a+t]=this[a+n];else Uint8Array.prototype.set.call(e,this.subarray(n,n+s),t);return s},u.prototype.fill=function(e,t,n,r){if("string"==typeof e){if("string"==typeof t?(r=t,t=0,n=this.length):"string"==typeof n&&(r=n,n=this.length),1===e.length){var a=e.charCodeAt(0);a<256&&(e=a)}if(void 0!==r&&"string"!=typeof r)throw new TypeError("encoding must be a string");if("string"==typeof r&&!u.isEncoding(r))throw new TypeError("Unknown encoding: "+r)}else"number"==typeof e&&(e&=255);if(t<0||this.length<t||this.length<n)throw new RangeError("Out of range index");if(n<=t)return this;var s;if(t>>>=0,n=void 0===n?this.length:n>>>0,e||(e=0),"number"==typeof e)for(s=t;s<n;++s)this[s]=e;else{var i=u.isBuffer(e)?e:z(new u(e,r).toString()),o=i.length;for(s=0;s<n-t;++s)this[s+t]=i[s%o]}return this};var $=/[^+\/0-9A-Za-z-_]/g;function B(e){return e<16?"0"+e.toString(16):e.toString(16)}function z(e,t){var n;t=t||1/0;for(var r=e.length,a=null,s=[],i=0;i<r;++i){if((n=e.charCodeAt(i))>55295&&n<57344){if(!a){if(n>56319){(t-=3)>-1&&s.push(239,191,189);continue}if(i+1===r){(t-=3)>-1&&s.push(239,191,189);continue}a=n;continue}if(n<56320){(t-=3)>-1&&s.push(239,191,189),a=n;continue}n=65536+(a-55296<<10|n-56320)}else a&&(t-=3)>-1&&s.push(239,191,189);if(a=null,n<128){if((t-=1)<0)break;s.push(n)}else if(n<2048){if((t-=2)<0)break;s.push(n>>6|192,63&n|128)}else if(n<65536){if((t-=3)<0)break;s.push(n>>12|224,n>>6&63|128,63&n|128)}else{if(!(n<1114112))throw new Error("Invalid code point");if((t-=4)<0)break;s.push(n>>18|240,n>>12&63|128,n>>6&63|128,63&n|128)}}return s}function U(e){return r.toByteArray(function(e){if((e=function(e){return e.trim?e.trim():e.replace(/^\s+|\s+$/g,"")}(e).replace($,"")).length<2)return"";for(;e.length%4!=0;)e+="=";return e}(e))}function W(e,t,n,r){for(var a=0;a<r&&!(a+n>=t.length||a>=e.length);++a)t[a+n]=e[a];return a}}).call(this,n(33))},function(e,t,n){"use strict";t.byteLength=function(e){var t=l(e),n=t[0],r=t[1];return 3*(n+r)/4-r},t.toByteArray=function(e){var t,n,r=l(e),i=r[0],o=r[1],u=new s(function(e,t,n){return 3*(t+n)/4-n}(0,i,o)),c=0,h=o>0?i-4:i;for(n=0;n<h;n+=4)t=a[e.charCodeAt(n)]<<18|a[e.charCodeAt(n+1)]<<12|a[e.charCodeAt(n+2)]<<6|a[e.charCodeAt(n+3)],u[c++]=t>>16&255,u[c++]=t>>8&255,u[c++]=255&t;2===o&&(t=a[e.charCodeAt(n)]<<2|a[e.charCodeAt(n+1)]>>4,u[c++]=255&t);1===o&&(t=a[e.charCodeAt(n)]<<10|a[e.charCodeAt(n+1)]<<4|a[e.charCodeAt(n+2)]>>2,u[c++]=t>>8&255,u[c++]=255&t);return u},t.fromByteArray=function(e){for(var t,n=e.length,a=n%3,s=[],i=0,o=n-a;i<o;i+=16383)s.push(c(e,i,i+16383>o?o:i+16383));1===a?(t=e[n-1],s.push(r[t>>2]+r[t<<4&63]+"==")):2===a&&(t=(e[n-2]<<8)+e[n-1],s.push(r[t>>10]+r[t>>4&63]+r[t<<2&63]+"="));return s.join("")};for(var r=[],a=[],s="undefined"!=typeof Uint8Array?Uint8Array:Array,i="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/",o=0,u=i.length;o<u;++o)r[o]=i[o],a[i.charCodeAt(o)]=o;function l(e){var t=e.length;if(t%4>0)throw new Error("Invalid string. Length must be a multiple of 4");var n=e.indexOf("=");return-1===n&&(n=t),[n,n===t?0:4-n%4]}function c(e,t,n){for(var a,s,i=[],o=t;o<n;o+=3)a=(e[o]<<16&16711680)+(e[o+1]<<8&65280)+(255&e[o+2]),i.push(r[(s=a)>>18&63]+r[s>>12&63]+r[s>>6&63]+r[63&s]);return i.join("")}a["-".charCodeAt(0)]=62,a["_".charCodeAt(0)]=63},function(e,t){
-/*! ieee754. BSD-3-Clause License. Feross Aboukhadijeh <https://feross.org/opensource> */
-t.read=function(e,t,n,r,a){var s,i,o=8*a-r-1,u=(1<<o)-1,l=u>>1,c=-7,h=n?a-1:0,p=n?-1:1,d=e[t+h];for(h+=p,s=d&(1<<-c)-1,d>>=-c,c+=o;c>0;s=256*s+e[t+h],h+=p,c-=8);for(i=s&(1<<-c)-1,s>>=-c,c+=r;c>0;i=256*i+e[t+h],h+=p,c-=8);if(0===s)s=1-l;else{if(s===u)return i?NaN:1/0*(d?-1:1);i+=Math.pow(2,r),s-=l}return(d?-1:1)*i*Math.pow(2,s-r)},t.write=function(e,t,n,r,a,s){var i,o,u,l=8*s-a-1,c=(1<<l)-1,h=c>>1,p=23===a?Math.pow(2,-24)-Math.pow(2,-77):0,d=r?0:s-1,f=r?1:-1,m=t<0||0===t&&1/t<0?1:0;for(t=Math.abs(t),isNaN(t)||t===1/0?(o=isNaN(t)?1:0,i=c):(i=Math.floor(Math.log(t)/Math.LN2),t*(u=Math.pow(2,-i))<1&&(i--,u*=2),(t+=i+h>=1?p/u:p*Math.pow(2,1-h))*u>=2&&(i++,u/=2),i+h>=c?(o=0,i=c):i+h>=1?(o=(t*u-1)*Math.pow(2,a),i+=h):(o=t*Math.pow(2,h-1)*Math.pow(2,a),i=0));a>=8;e[n+d]=255&o,d+=f,o/=256,a-=8);for(i=i<<a|o,l+=a;l>0;e[n+d]=255&i,d+=f,i/=256,l-=8);e[n+d-f]|=128*m}},function(e,t){var n={}.toString;e.exports=Array.isArray||function(e){return"[object Array]"==n.call(e)}},function(e,t,n){(function(e){var r=void 0!==e&&e||"undefined"!=typeof self&&self||window,a=Function.prototype.apply;function s(e,t){this._id=e,this._clearFn=t}t.setTimeout=function(){return new s(a.call(setTimeout,r,arguments),clearTimeout)},t.setInterval=function(){return new s(a.call(setInterval,r,arguments),clearInterval)},t.clearTimeout=t.clearInterval=function(e){e&&e.close()},s.prototype.unref=s.prototype.ref=function(){},s.prototype.close=function(){this._clearFn.call(r,this._id)},t.enroll=function(e,t){clearTimeout(e._idleTimeoutId),e._idleTimeout=t},t.unenroll=function(e){clearTimeout(e._idleTimeoutId),e._idleTimeout=-1},t._unrefActive=t.active=function(e){clearTimeout(e._idleTimeoutId);var t=e._idleTimeout;t>=0&&(e._idleTimeoutId=setTimeout((function(){e._onTimeout&&e._onTimeout()}),t))},n(243),t.setImmediate="undefined"!=typeof self&&self.setImmediate||void 0!==e&&e.setImmediate||this&&this.setImmediate,t.clearImmediate="undefined"!=typeof self&&self.clearImmediate||void 0!==e&&e.clearImmediate||this&&this.clearImmediate}).call(this,n(33))},function(e,t,n){(function(e,t){!function(e,n){"use strict";if(!e.setImmediate){var r,a,s,i,o,u=1,l={},c=!1,h=e.document,p=Object.getPrototypeOf&&Object.getPrototypeOf(e);p=p&&p.setTimeout?p:e,"[object process]"==={}.toString.call(e.process)?r=function(e){t.nextTick((function(){f(e)}))}:!function(){if(e.postMessage&&!e.importScripts){var t=!0,n=e.onmessage;return e.onmessage=function(){t=!1},e.postMessage("","*"),e.onmessage=n,t}}()?e.MessageChannel?((s=new MessageChannel).port1.onmessage=function(e){f(e.data)},r=function(e){s.port2.postMessage(e)}):h&&"onreadystatechange"in h.createElement("script")?(a=h.documentElement,r=function(e){var t=h.createElement("script");t.onreadystatechange=function(){f(e),t.onreadystatechange=null,a.removeChild(t),t=null},a.appendChild(t)}):r=function(e){setTimeout(f,0,e)}:(i="setImmediate$"+Math.random()+"$",o=function(t){t.source===e&&"string"==typeof t.data&&0===t.data.indexOf(i)&&f(+t.data.slice(i.length))},e.addEventListener?e.addEventListener("message",o,!1):e.attachEvent("onmessage",o),r=function(t){e.postMessage(i+t,"*")}),p.setImmediate=function(e){"function"!=typeof e&&(e=new Function(""+e));for(var t=new Array(arguments.length-1),n=0;n<t.length;n++)t[n]=arguments[n+1];var a={callback:e,args:t};return l[u]=a,r(u),u++},p.clearImmediate=d}function d(e){delete l[e]}function f(e){if(c)setTimeout(f,0,e);else{var t=l[e];if(t){c=!0;try{!function(e){var t=e.callback,n=e.args;switch(n.length){case 0:t();break;case 1:t(n[0]);break;case 2:t(n[0],n[1]);break;case 3:t(n[0],n[1],n[2]);break;default:t.apply(void 0,n)}}(t)}finally{d(e),c=!1}}}}}("undefined"==typeof self?void 0===e?this:e:self)}).call(this,n(33),n(124))},function(e,t,n){"use strict";n.r(t);n(127),n(132),n(151),n(157),n(158),n(161),n(162),n(84),n(164),n(167),n(175),n(178),n(179),n(180),n(183),n(184),n(185),n(186),n(187),n(188),n(189),n(190),n(192),n(193),n(194),n(195),n(196),n(197),n(198),n(199),n(204),n(205),n(206),n(207),n(208),n(209),n(218),n(219),n(220),n(222),n(223),n(224),n(225),n(226),n(227),n(228),n(229),n(230),n(231),n(232),n(233),n(234);var r=n(0);const a=(e,t)=>"undefined"==typeof window?Promise.resolve():Object(r.e)().then(()=>Object(r.a)([["camera-comp_3",[[1,"camera-component",{url_logo:[1],show_mask:[1],background_color:[1],disable_control_panel:[1],stop_after_capturing:[1],face_detection:[1],model_path:[1],probability_threshold:[1],debug:[1],logo_style:[1],_resolutionOnPhoto:[32],_cameraStatus:[32],_mobileMakePhoto:[32],_stopAfterCapturing:[32],_showMask:[32],_disable_control_panel:[32],_face_detection:[32],_probabilityThreshold:[32],_debug:[32],video:[32],canvas:[32]},[[0,"videoStarted","videoStarted"],[0,"retakePhoto","retakePhoto"],[0,"takePhoto","takePhoto"],[0,"openCamera","openCamera"],[0,"loadModels","loadModels"],[0,"errorCamera","errorCamera"],[0,"closeCamera","closeCamera"],[0,"makePhoto","makePhoto"]]],[0,"camera-comp",{showMask:[4,"show-mask"],modelPath:[1,"model-path"],device:[16],probabilityThreshold:[2,"probability-threshold"],faceDetection:[4,"face-detection"]}],[0,"control-panel",{mobile:[4],disableControlPanel:[4,"disable-control-panel"],cameraStatus:[2,"camera-status"],resolutionOnPhoto:[4,"resolution-on-photo"],mobileMakePhoto:[4,"mobile-make-photo"],faceDetection:[4,"face-detection"]}]]]],t));function s(e){return(s="function"==typeof Symbol&&"symbol"==typeof Symbol.iterator?function(e){return typeof e}:function(e){return e&&"function"==typeof Symbol&&e.constructor===Symbol&&e!==Symbol.prototype?"symbol":typeof e})(e)}function i(){/*! regenerator-runtime -- Copyright (c) 2014-present, Facebook, Inc. -- license (MIT): https://github.com/facebook/regenerator/blob/main/LICENSE */i=function(){return e};var e={},t=Object.prototype,n=t.hasOwnProperty,r=Object.defineProperty||function(e,t,n){e[t]=n.value},a="function"==typeof Symbol?Symbol:{},o=a.iterator||"@@iterator",u=a.asyncIterator||"@@asyncIterator",l=a.toStringTag||"@@toStringTag";function c(e,t,n){return Object.defineProperty(e,t,{value:n,enumerable:!0,configurable:!0,writable:!0}),e[t]}try{c({},"")}catch(e){c=function(e,t,n){return e[t]=n}}function h(e,t,n,a){var s=t&&t.prototype instanceof f?t:f,i=Object.create(s.prototype),o=new N(a||[]);return r(i,"_invoke",{value:S(e,n,o)}),i}function p(e,t,n){try{return{type:"normal",arg:e.call(t,n)}}catch(e){return{type:"throw",arg:e}}}e.wrap=h;var d={};function f(){}function m(){}function g(){}var y={};c(y,o,(function(){return this}));var b=Object.getPrototypeOf,v=b&&b(b(C([])));v&&v!==t&&n.call(v,o)&&(y=v);var x=g.prototype=f.prototype=Object.create(y);function w(e){["next","throw","return"].forEach((function(t){c(e,t,(function(e){return this._invoke(t,e)}))}))}function k(e,t){var a;r(this,"_invoke",{value:function(r,i){function o(){return new t((function(a,o){!function r(a,i,o,u){var l=p(e[a],e,i);if("throw"!==l.type){var c=l.arg,h=c.value;return h&&"object"==s(h)&&n.call(h,"__await")?t.resolve(h.__await).then((function(e){r("next",e,o,u)}),(function(e){r("throw",e,o,u)})):t.resolve(h).then((function(e){c.value=e,o(c)}),(function(e){return r("throw",e,o,u)}))}u(l.arg)}(r,i,a,o)}))}return a=a?a.then(o,o):o()}})}function S(e,t,n){var r="suspendedStart";return function(a,s){if("executing"===r)throw new Error("Generator is already running");if("completed"===r){if("throw"===a)throw s;return T()}for(n.method=a,n.arg=s;;){var i=n.delegate;if(i){var o=I(i,n);if(o){if(o===d)continue;return o}}if("next"===n.method)n.sent=n._sent=n.arg;else if("throw"===n.method){if("suspendedStart"===r)throw r="completed",n.arg;n.dispatchException(n.arg)}else"return"===n.method&&n.abrupt("return",n.arg);r="executing";var u=p(e,t,n);if("normal"===u.type){if(r=n.done?"completed":"suspendedYield",u.arg===d)continue;return{value:u.arg,done:n.done}}"throw"===u.type&&(r="completed",n.method="throw",n.arg=u.arg)}}}function I(e,t){var n=t.method,r=e.iterator[n];if(void 0===r)return t.delegate=null,"throw"===n&&e.iterator.return&&(t.method="return",t.arg=void 0,I(e,t),"throw"===t.method)||"return"!==n&&(t.method="throw",t.arg=new TypeError("The iterator does not provide a '"+n+"' method")),d;var a=p(r,e.iterator,t.arg);if("throw"===a.type)return t.method="throw",t.arg=a.arg,t.delegate=null,d;var s=a.arg;return s?s.done?(t[e.resultName]=s.value,t.next=e.nextLoc,"return"!==t.method&&(t.method="next",t.arg=void 0),t.delegate=null,d):s:(t.method="throw",t.arg=new TypeError("iterator result is not an object"),t.delegate=null,d)}function A(e){var t={tryLoc:e[0]};1 in e&&(t.catchLoc=e[1]),2 in e&&(t.finallyLoc=e[2],t.afterLoc=e[3]),this.tryEntries.push(t)}function E(e){var t=e.completion||{};t.type="normal",delete t.arg,e.completion=t}function N(e){this.tryEntries=[{tryLoc:"root"}],e.forEach(A,this),this.reset(!0)}function C(e){if(e){var t=e[o];if(t)return t.call(e);if("function"==typeof e.next)return e;if(!isNaN(e.length)){var r=-1,a=function t(){for(;++r<e.length;)if(n.call(e,r))return t.value=e[r],t.done=!1,t;return t.value=void 0,t.done=!0,t};return a.next=a}}return{next:T}}function T(){return{value:void 0,done:!0}}return m.prototype=g,r(x,"constructor",{value:g,configurable:!0}),r(g,"constructor",{value:m,configurable:!0}),m.displayName=c(g,l,"GeneratorFunction"),e.isGeneratorFunction=function(e){var t="function"==typeof e&&e.constructor;return!!t&&(t===m||"GeneratorFunction"===(t.displayName||t.name))},e.mark=function(e){return Object.setPrototypeOf?Object.setPrototypeOf(e,g):(e.__proto__=g,c(e,l,"GeneratorFunction")),e.prototype=Object.create(x),e},e.awrap=function(e){return{__await:e}},w(k.prototype),c(k.prototype,u,(function(){return this})),e.AsyncIterator=k,e.async=function(t,n,r,a,s){void 0===s&&(s=Promise);var i=new k(h(t,n,r,a),s);return e.isGeneratorFunction(n)?i:i.next().then((function(e){return e.done?e.value:i.next()}))},w(x),c(x,l,"Generator"),c(x,o,(function(){return this})),c(x,"toString",(function(){return"[object Generator]"})),e.keys=function(e){var t=Object(e),n=[];for(var r in t)n.push(r);return n.reverse(),function e(){for(;n.length;){var r=n.pop();if(r in t)return e.value=r,e.done=!1,e}return e.done=!0,e}},e.values=C,N.prototype={constructor:N,reset:function(e){if(this.prev=0,this.next=0,this.sent=this._sent=void 0,this.done=!1,this.delegate=null,this.method="next",this.arg=void 0,this.tryEntries.forEach(E),!e)for(var t in this)"t"===t.charAt(0)&&n.call(this,t)&&!isNaN(+t.slice(1))&&(this[t]=void 0)},stop:function(){this.done=!0;var e=this.tryEntries[0].completion;if("throw"===e.type)throw e.arg;return this.rval},dispatchException:function(e){if(this.done)throw e;var t=this;function r(n,r){return i.type="throw",i.arg=e,t.next=n,r&&(t.method="next",t.arg=void 0),!!r}for(var a=this.tryEntries.length-1;a>=0;--a){var s=this.tryEntries[a],i=s.completion;if("root"===s.tryLoc)return r("end");if(s.tryLoc<=this.prev){var o=n.call(s,"catchLoc"),u=n.call(s,"finallyLoc");if(o&&u){if(this.prev<s.catchLoc)return r(s.catchLoc,!0);if(this.prev<s.finallyLoc)return r(s.finallyLoc)}else if(o){if(this.prev<s.catchLoc)return r(s.catchLoc,!0)}else{if(!u)throw new Error("try statement without catch or finally");if(this.prev<s.finallyLoc)return r(s.finallyLoc)}}}},abrupt:function(e,t){for(var r=this.tryEntries.length-1;r>=0;--r){var a=this.tryEntries[r];if(a.tryLoc<=this.prev&&n.call(a,"finallyLoc")&&this.prev<a.finallyLoc){var s=a;break}}s&&("break"===e||"continue"===e)&&s.tryLoc<=t&&t<=s.finallyLoc&&(s=null);var i=s?s.completion:{};return i.type=e,i.arg=t,s?(this.method="next",this.next=s.finallyLoc,d):this.complete(i)},complete:function(e,t){if("throw"===e.type)throw e.arg;return"break"===e.type||"continue"===e.type?this.next=e.arg:"return"===e.type?(this.rval=this.arg=e.arg,this.method="return",this.next="end"):"normal"===e.type&&t&&(this.next=t),d},finish:function(e){for(var t=this.tryEntries.length-1;t>=0;--t){var n=this.tryEntries[t];if(n.finallyLoc===e)return this.complete(n.completion,n.afterLoc),E(n),d}},catch:function(e){for(var t=this.tryEntries.length-1;t>=0;--t){var n=this.tryEntries[t];if(n.tryLoc===e){var r=n.completion;if("throw"===r.type){var a=r.arg;E(n)}return a}}throw new Error("illegal catch attempt")},delegateYield:function(e,t,n){return this.delegate={iterator:C(e),resultName:t,nextLoc:n},"next"===this.method&&(this.arg=void 0),d}},e}function o(e,t,n,r,a,s,i){try{var o=e[s](i),u=o.value}catch(e){return void n(e)}o.done?t(u):Promise.resolve(u).then(r,a)}function u(e){return function(){var t=this,n=arguments;return new Promise((function(r,a){var s=e.apply(t,n);function i(e){o(s,r,a,i,u,"next",e)}function u(e){o(s,r,a,i,u,"throw",e)}i(void 0)}))}}!function(){if("undefined"!=typeof window&&void 0!==window.Reflect&&void 0!==window.customElements){var e=HTMLElement;window.HTMLElement=function(){return Reflect.construct(e,[],this.constructor)},HTMLElement.prototype=e.prototype,HTMLElement.prototype.constructor=HTMLElement,Object.setPrototypeOf(HTMLElement,e)}}(),function(){var e=[];if("undefined"!=typeof window){var t=window;t.customElements&&(!t.Element||t.Element.prototype.closest&&t.Element.prototype.matches&&t.Element.prototype.remove&&t.Element.prototype.getRootNode)||e.push(Promise.resolve().then(n.t.bind(null,245,7)));"function"==typeof Object.assign&&Object.entries&&Array.prototype.find&&Array.prototype.includes&&String.prototype.startsWith&&String.prototype.endsWith&&(!t.NodeList||t.NodeList.prototype.forEach)&&t.fetch&&function(){try{var e=new URL("b","http://a");return e.pathname="c%20d","http://a/c%20d"===e.href&&e.searchParams}catch(e){return!1}}()&&"undefined"!=typeof WeakMap||e.push(Promise.resolve().then(n.t.bind(null,246,7)))}return Promise.all(e)}().then((function(){a(window)})),window.AcuantPassiveLiveness=function(){var e,t=null,n=null,r=null,a=null,s=null,o=null,l=null,c=null,h=null,p=null,d={FACE_NOT_FOUND:"FACE NOT FOUND",TOO_MANY_FACES:"TOO MANY FACES",FACE_ANGLE_TOO_LARGE:"FACE ANGLE TOO LARGE",PROBABILITY_TOO_SMALL:"PROBABILITY TOO SMALL",FACE_TOO_SMALL:"FACE TOO SMALL",FACE_CLOSE_TO_BORDER:"TOO CLOSE TO THE FRAME"},f={start:function(e){var t=arguments.length>1&&void 0!==arguments[1]?arguments[1]:null;m()?(t&&(d=t),x(),I(e),w()):B(e.onCaptured)},startManualCapture:function(e){return B(e)},end:function(){return k()},getLiveness:(e=u(i().mark((function e(t,n){var r,a,s,o,u,l,c;return i().wrap((function(e){for(;;)switch(e.prev=e.next){case 0:return r=t.endpoint,a=t.token,s=t.subscriptionId,o=t.image,u={Settings:{SubscriptionId:s,AdditionalSettings:{OS:"UNKNOWN"}},Image:o},e.prev=2,e.next=5,fetch(r+"/api/v1/liveness",{method:"POST",body:JSON.stringify(u),headers:{"Content-Type":"application/json",Authorization:"Basic "+a,Accept:"application/json"}});case 5:return l=e.sent,e.next=8,l.json();case 8:c=e.sent,n(c),e.next=16;break;case 12:e.prev=12,e.t0=e.catch(2),console.log("Error while getting liveness: "+e.t0),n(null);case 16:case"end":return e.stop()}}),e,null,[[2,12]])}))),function(t,n){return e.apply(this,arguments)})},m=function(){return"mediaDevices"in navigator&&g()},g=function(){var e,t=!1;return e=navigator.userAgent||navigator.vendor||window.opera,(/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino|android|playbook|silk/i.test(e)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(e.substr(0,4)))&&(t=!0),(t||y())&&!v()},y=function(){return/iPad|iPhone|iPod/.test(navigator.platform)||b()},b=function(){return"MacIntel"===navigator.platform&&navigator.maxTouchPoints>1},v=function(){return navigator.userAgent.toLowerCase().indexOf("firefox")>-1},x=function(){if(!(n=document.getElementById("acuant-face-capture-camera"))){var e=document.getElementById("acuant-face-capture-container");(n=document.createElement("camera-component")).id="acuant-face-capture-camera",n.model_path=acuantConfig.path,n.url_logo=" ",n.logo_style='{"display": "none" }',n.show_mask=!0,n.disable_control_panel=!0,n.stop_after_capturing=!0,n.face_detection=!0,n.background_color="white",n.probability_threshold=70,e.appendChild(n)}},w=function(){var e,t,r,a;(-1==(a=navigator.userAgent.toLowerCase()).indexOf("safari")||a.indexOf("chrome")>-1)&&(e=window.navigator.standalone,t=window.navigator.userAgent.toLowerCase(),r=/mobile.*safari/.test(t),!/iphone|ipod|ipad/.test(t)||r||e)?n.parentElement.requestFullscreen().then((function(){S()})).catch((function(){console.log("Error while requesting full screen. Launching camera."),S()})):S()},k=function(){if(n){var e=new Event("closeCamera");n.dispatchEvent(e)}},S=function(){var e=new Event("openCamera");n.dispatchEvent(e)},I=function(e){D(e.onError),L(e.onOpened),P(e.onClosed),_(e.onDetection),C(e.onCaptured),E(e.onPhotoTaken),N(e.onPhotoRetake),A(e.onDetectorInitialized)},A=function(e){p=e,n.addEventListener("detectorInitialized",p)},E=function(e){r=e,n.addEventListener("makePhoto",r)},N=function(e){a=e,n.addEventListener("retakePhoto",a)},C=function(e){s=T(e),n.addEventListener("capture",s)},T=function(e){return function(t){q();var n=t.detail.photos[0];R(n,(function(t){document.fullscreenElement?document.exitFullscreen().then((function(){U(t,e)})):U(t,e)}))}},R=function(e,t){var n=new FileReader;n.onloadend=function(){return t(n.result)},n.readAsDataURL(e)},_=function(e){o=F(e),n.addEventListener("detectionError",o)},F=function(e){return function(t){var n=t.detail.error,r=M(n,!0);e(d[r])}},M=function(e,t){return Object.keys(e).find((function(n){return e[n]===t}))},D=function(e){l=O(e),n.addEventListener("error",l)},O=function(e){return function(t){q();var n={code:1,message:"Camera permission not granted"};"isTrusted"in t?document.fullscreenElement?document.exitFullscreen().then((function(){e(n)})):e(n):e({code:2,message:"Unexpected error: "+t})}},L=function(e){c=e,n.addEventListener("openCamera",c)},P=function(e){h=$(e),n.addEventListener("closeCamera",h)},$=function(e){return function(){q(),document.fullscreenElement?document.exitFullscreen().then((function(){e()})):e()}},B=function(e){t||((t=document.createElement("input")).type="file",t.capture="user",t.accept="image/*",t.onchange=function(t){return z(t,e)}),t.click()},z=function(e,t){var n=e.target,r=new FileReader;r.onload=function(e){return U(e.target.result,t)},r.readAsDataURL(n.files[0])},U=function(e,t){var n=document.createElement("img"),r=document.createElement("canvas");n.onload=function(){var e=1080,a=720,s=n.width,i=n.height,o=r.getContext("2d");(s>i?i:s)>a?s<i?a=(e=a)*(i/s):e=a*(s/i):(e=n.width,a=n.height);r.width=e,r.height=a,o.drawImage(n,0,0,e,a);var u=r.toDataURL("image/jpeg",.8);o.clearRect(0,0,e,a);var l=j(u);W(l).then(t)},n.src=e},W=function(){var e=u(i().mark((function e(t){var n,r;return i().wrap((function(e){for(;;)switch(e.prev=e.next){case 0:if(n=G(t),!AcuantJavascriptWebSdk.singleWorkerModel){e.next=11;break}return e.next=4,new Promise((function(e){AcuantJavascriptWebSdk.startImageWorker(e)}));case 4:return e.next=6,V(n);case 6:return r=e.sent,AcuantJavascriptWebSdk.endImageWorker(),e.abrupt("return",r);case 11:return e.next=13,V(n);case 13:return e.abrupt("return",e.sent);case 14:case"end":return e.stop()}}),e)})));return function(t){return e.apply(this,arguments)}}();function V(e){return new Promise((function(t){AcuantJavascriptWebSdk.sign(e,{onSuccess:function(e){return t(H(e))},onFail:t})}))}var j=function(e){return AcuantJavascriptWebSdk.addMetadata(e,{dateTimeOriginal:(new Date).toUTCString()})},G=function(e){for(var t=window.atob(e.split("base64,")[1]),n=t.length,r=new Uint8Array(new ArrayBuffer(n)),a=0;a<n;a++)r[a]=t.charCodeAt(a);return r},H=function(e){for(var t="",n=new Uint8Array(e),r=n.byteLength,a=0;a<r;a++)t+=String.fromCharCode(n[a]);return window.btoa(t)},q=function(){n.removeEventListener("makePhoto",r),n.removeEventListener("retakePhoto",a),n.removeEventListener("capture",s),n.removeEventListener("detectionError",o),n.removeEventListener("error",l),n.removeEventListener("openCamera",c),n.removeEventListener("closeCamera",h),n.removeEventListener("detectorInitialized",p),r=null,a=null,s=null,o=null,l=null,c=null,h=null,p=null,n=null};return f}()},function(e,t){var n,r,a,s,i;(function(){"use strict";var e=new Set("annotation-xml color-profile font-face font-face-src font-face-uri font-face-format font-face-name missing-glyph".split(" "));function t(t){var n=e.has(t);return t=/^[a-z][.0-9_a-z]*-[\-.0-9_a-z]*$/.test(t),!n&&t}function n(e){var t=e.isConnected;if(void 0!==t)return t;for(;e&&!(e.__CE_isImportDocument||e instanceof Document);)e=e.parentNode||(window.ShadowRoot&&e instanceof ShadowRoot?e.host:void 0);return!(!e||!(e.__CE_isImportDocument||e instanceof Document))}function r(e,t){for(;t&&t!==e&&!t.nextSibling;)t=t.parentNode;return t&&t!==e?t.nextSibling:null}function a(e,t,n){n=void 0===n?new Set:n;for(var s=e;s;){if(s.nodeType===Node.ELEMENT_NODE){var i=s;t(i);var o=i.localName;if("link"===o&&"import"===i.getAttribute("rel")){if((s=i.import)instanceof Node&&!n.has(s))for(n.add(s),s=s.firstChild;s;s=s.nextSibling)a(s,t,n);s=r(e,i);continue}if("template"===o){s=r(e,i);continue}if(i=i.__CE_shadowRoot)for(i=i.firstChild;i;i=i.nextSibling)a(i,t,n)}s=s.firstChild?s.firstChild:r(e,s)}}function s(e,t,n){e[t]=n}function i(){this.a=new Map,this.g=new Map,this.c=[],this.f=[],this.b=!1}function o(e,t){e.b&&a(t,(function(t){return u(e,t)}))}function u(e,t){if(e.b&&!t.__CE_patched){t.__CE_patched=!0;for(var n=0;n<e.c.length;n++)e.c[n](t);for(n=0;n<e.f.length;n++)e.f[n](t)}}function l(e,t){var n=[];for(a(t,(function(e){return n.push(e)})),t=0;t<n.length;t++){var r=n[t];1===r.__CE_state?e.connectedCallback(r):p(e,r)}}function c(e,t){var n=[];for(a(t,(function(e){return n.push(e)})),t=0;t<n.length;t++){var r=n[t];1===r.__CE_state&&e.disconnectedCallback(r)}}function h(e,t,n){var r=(n=void 0===n?{}:n).u||new Set,s=n.i||function(t){return p(e,t)},i=[];if(a(t,(function(t){if("link"===t.localName&&"import"===t.getAttribute("rel")){var n=t.import;n instanceof Node&&(n.__CE_isImportDocument=!0,n.__CE_hasRegistry=!0),n&&"complete"===n.readyState?n.__CE_documentLoadHandled=!0:t.addEventListener("load",(function(){var n=t.import;if(!n.__CE_documentLoadHandled){n.__CE_documentLoadHandled=!0;var a=new Set(r);a.delete(n),h(e,n,{u:a,i:s})}}))}else i.push(t)}),r),e.b)for(t=0;t<i.length;t++)u(e,i[t]);for(t=0;t<i.length;t++)s(i[t])}function p(e,t){if(void 0===t.__CE_state){var r=t.ownerDocument;if((r.defaultView||r.__CE_isImportDocument&&r.__CE_hasRegistry)&&(r=e.a.get(t.localName))){r.constructionStack.push(t);var a=r.constructorFunction;try{try{if(new a!==t)throw Error("The custom element constructor did not produce the element being upgraded.")}finally{r.constructionStack.pop()}}catch(e){throw t.__CE_state=2,e}if(t.__CE_state=1,t.__CE_definition=r,r.attributeChangedCallback)for(r=r.observedAttributes,a=0;a<r.length;a++){var s=r[a],i=t.getAttribute(s);null!==i&&e.attributeChangedCallback(t,s,null,i,null)}n(t)&&e.connectedCallback(t)}}}function d(e){var t=document;this.c=e,this.a=t,this.b=void 0,h(this.c,this.a),"loading"===this.a.readyState&&(this.b=new MutationObserver(this.f.bind(this)),this.b.observe(this.a,{childList:!0,subtree:!0}))}function f(e){e.b&&e.b.disconnect()}function m(){var e=this;this.b=this.a=void 0,this.c=new Promise((function(t){e.b=t,e.a&&t(e.a)}))}function g(e){if(e.a)throw Error("Already resolved.");e.a=void 0,e.b&&e.b(void 0)}function y(e){this.c=!1,this.a=e,this.j=new Map,this.f=function(e){return e()},this.b=!1,this.g=[],this.o=new d(e)}i.prototype.connectedCallback=function(e){var t=e.__CE_definition;t.connectedCallback&&t.connectedCallback.call(e)},i.prototype.disconnectedCallback=function(e){var t=e.__CE_definition;t.disconnectedCallback&&t.disconnectedCallback.call(e)},i.prototype.attributeChangedCallback=function(e,t,n,r,a){var s=e.__CE_definition;s.attributeChangedCallback&&-1<s.observedAttributes.indexOf(t)&&s.attributeChangedCallback.call(e,t,n,r,a)},d.prototype.f=function(e){var t=this.a.readyState;for("interactive"!==t&&"complete"!==t||f(this),t=0;t<e.length;t++)for(var n=e[t].addedNodes,r=0;r<n.length;r++)h(this.c,n[r])},y.prototype.l=function(e,n){var r=this;if(!(n instanceof Function))throw new TypeError("Custom element constructors must be functions.");if(!t(e))throw new SyntaxError("The element name '"+e+"' is not valid.");if(this.a.a.get(e))throw Error("A custom element with name '"+e+"' has already been defined.");if(this.c)throw Error("A custom element is already being defined.");this.c=!0;try{var a=function(e){var t=s[e];if(void 0!==t&&!(t instanceof Function))throw Error("The '"+e+"' callback must be a function.");return t},s=n.prototype;if(!(s instanceof Object))throw new TypeError("The custom element constructor's prototype is not an object.");var i=a("connectedCallback"),o=a("disconnectedCallback"),u=a("adoptedCallback"),l=a("attributeChangedCallback"),c=n.observedAttributes||[]}catch(e){return}finally{this.c=!1}n={localName:e,constructorFunction:n,connectedCallback:i,disconnectedCallback:o,adoptedCallback:u,attributeChangedCallback:l,observedAttributes:c,constructionStack:[]},function(e,t,n){e.a.set(t,n),e.g.set(n.constructorFunction,n)}(this.a,e,n),this.g.push(n),this.b||(this.b=!0,this.f((function(){return function(e){if(!1!==e.b){e.b=!1;for(var t=e.g,n=[],r=new Map,a=0;a<t.length;a++)r.set(t[a].localName,[]);for(h(e.a,document,{i:function(t){if(void 0===t.__CE_state){var a=t.localName,s=r.get(a);s?s.push(t):e.a.a.get(a)&&n.push(t)}}}),a=0;a<n.length;a++)p(e.a,n[a]);for(;0<t.length;){var s=t.shift();a=s.localName,s=r.get(s.localName);for(var i=0;i<s.length;i++)p(e.a,s[i]);(a=e.j.get(a))&&g(a)}}}(r)})))},y.prototype.i=function(e){h(this.a,e)},y.prototype.get=function(e){if(e=this.a.a.get(e))return e.constructorFunction},y.prototype.m=function(e){if(!t(e))return Promise.reject(new SyntaxError("'"+e+"' is not a valid custom element name."));var n=this.j.get(e);return n||(n=new m,this.j.set(e,n),this.a.a.get(e)&&!this.g.some((function(t){return t.localName===e}))&&g(n)),n.c},y.prototype.s=function(e){f(this.o);var t=this.f;this.f=function(n){return e((function(){return t(n)}))}},window.CustomElementRegistry=y,y.prototype.define=y.prototype.l,y.prototype.upgrade=y.prototype.i,y.prototype.get=y.prototype.get,y.prototype.whenDefined=y.prototype.m,y.prototype.polyfillWrapFlushCallback=y.prototype.s;var b=window.Document.prototype.createElement,v=window.Document.prototype.createElementNS,x=window.Document.prototype.importNode,w=window.Document.prototype.prepend,k=window.Document.prototype.append,S=window.DocumentFragment.prototype.prepend,I=window.DocumentFragment.prototype.append,A=window.Node.prototype.cloneNode,E=window.Node.prototype.appendChild,N=window.Node.prototype.insertBefore,C=window.Node.prototype.removeChild,T=window.Node.prototype.replaceChild,R=Object.getOwnPropertyDescriptor(window.Node.prototype,"textContent"),_=window.Element.prototype.attachShadow,F=Object.getOwnPropertyDescriptor(window.Element.prototype,"innerHTML"),M=window.Element.prototype.getAttribute,D=window.Element.prototype.setAttribute,O=window.Element.prototype.removeAttribute,L=window.Element.prototype.getAttributeNS,P=window.Element.prototype.setAttributeNS,$=window.Element.prototype.removeAttributeNS,B=window.Element.prototype.insertAdjacentElement,z=window.Element.prototype.insertAdjacentHTML,U=window.Element.prototype.prepend,W=window.Element.prototype.append,V=window.Element.prototype.before,j=window.Element.prototype.after,G=window.Element.prototype.replaceWith,H=window.Element.prototype.remove,q=window.HTMLElement,K=Object.getOwnPropertyDescriptor(window.HTMLElement.prototype,"innerHTML"),Y=window.HTMLElement.prototype.insertAdjacentElement,X=window.HTMLElement.prototype.insertAdjacentHTML,Q=new function(){};function Z(e,t,r){function a(t){return function(r){for(var a=[],s=0;s<arguments.length;++s)a[s]=arguments[s];s=[];for(var i=[],o=0;o<a.length;o++){var u=a[o];if(u instanceof Element&&n(u)&&i.push(u),u instanceof DocumentFragment)for(u=u.firstChild;u;u=u.nextSibling)s.push(u);else s.push(u)}for(t.apply(this,a),a=0;a<i.length;a++)c(e,i[a]);if(n(this))for(a=0;a<s.length;a++)(i=s[a])instanceof Element&&l(e,i)}}void 0!==r.h&&(t.prepend=a(r.h)),void 0!==r.append&&(t.append=a(r.append))}var J,ee=window.customElements;if(!ee||ee.forcePolyfill||"function"!=typeof ee.define||"function"!=typeof ee.get){var te=new i;J=te,window.HTMLElement=function(){function e(){var e=this.constructor,t=J.g.get(e);if(!t)throw Error("The custom element being constructed was not registered with `customElements`.");var n=t.constructionStack;if(0===n.length)return n=b.call(document,t.localName),Object.setPrototypeOf(n,e.prototype),n.__CE_state=1,n.__CE_definition=t,u(J,n),n;var r=n[t=n.length-1];if(r===Q)throw Error("The HTMLElement constructor was either called reentrantly for this constructor or called multiple times.");return n[t]=Q,Object.setPrototypeOf(r,e.prototype),u(J,r),r}return e.prototype=q.prototype,Object.defineProperty(e.prototype,"constructor",{writable:!0,configurable:!0,enumerable:!1,value:e}),e}(),function(){var e=te;s(Document.prototype,"createElement",(function(t){if(this.__CE_hasRegistry){var n=e.a.get(t);if(n)return new n.constructorFunction}return t=b.call(this,t),u(e,t),t})),s(Document.prototype,"importNode",(function(t,n){return t=x.call(this,t,!!n),this.__CE_hasRegistry?h(e,t):o(e,t),t})),s(Document.prototype,"createElementNS",(function(t,n){if(this.__CE_hasRegistry&&(null===t||"http://www.w3.org/1999/xhtml"===t)){var r=e.a.get(n);if(r)return new r.constructorFunction}return t=v.call(this,t,n),u(e,t),t})),Z(e,Document.prototype,{h:w,append:k})}(),Z(te,DocumentFragment.prototype,{h:S,append:I}),function(){function e(e,r){Object.defineProperty(e,"textContent",{enumerable:r.enumerable,configurable:!0,get:r.get,set:function(e){if(this.nodeType===Node.TEXT_NODE)r.set.call(this,e);else{var a=void 0;if(this.firstChild){var s=this.childNodes,i=s.length;if(0<i&&n(this)){a=Array(i);for(var o=0;o<i;o++)a[o]=s[o]}}if(r.set.call(this,e),a)for(e=0;e<a.length;e++)c(t,a[e])}}})}var t=te;s(Node.prototype,"insertBefore",(function(e,r){if(e instanceof DocumentFragment){var a=Array.prototype.slice.apply(e.childNodes);if(e=N.call(this,e,r),n(this))for(r=0;r<a.length;r++)l(t,a[r]);return e}return a=n(e),r=N.call(this,e,r),a&&c(t,e),n(this)&&l(t,e),r})),s(Node.prototype,"appendChild",(function(e){if(e instanceof DocumentFragment){var r=Array.prototype.slice.apply(e.childNodes);if(e=E.call(this,e),n(this))for(var a=0;a<r.length;a++)l(t,r[a]);return e}return r=n(e),a=E.call(this,e),r&&c(t,e),n(this)&&l(t,e),a})),s(Node.prototype,"cloneNode",(function(e){return e=A.call(this,!!e),this.ownerDocument.__CE_hasRegistry?h(t,e):o(t,e),e})),s(Node.prototype,"removeChild",(function(e){var r=n(e),a=C.call(this,e);return r&&c(t,e),a})),s(Node.prototype,"replaceChild",(function(e,r){if(e instanceof DocumentFragment){var a=Array.prototype.slice.apply(e.childNodes);if(e=T.call(this,e,r),n(this))for(c(t,r),r=0;r<a.length;r++)l(t,a[r]);return e}a=n(e);var s=T.call(this,e,r),i=n(this);return i&&c(t,r),a&&c(t,e),i&&l(t,e),s})),R&&R.get?e(Node.prototype,R):function(e,t){e.b=!0,e.c.push(t)}(t,(function(t){e(t,{enumerable:!0,configurable:!0,get:function(){for(var e=[],t=0;t<this.childNodes.length;t++){var n=this.childNodes[t];n.nodeType!==Node.COMMENT_NODE&&e.push(n.textContent)}return e.join("")},set:function(e){for(;this.firstChild;)C.call(this,this.firstChild);null!=e&&""!==e&&E.call(this,document.createTextNode(e))}})}))}(),function(){function e(e,t){Object.defineProperty(e,"innerHTML",{enumerable:t.enumerable,configurable:!0,get:t.get,set:function(e){var r=this,s=void 0;if(n(this)&&(s=[],a(this,(function(e){e!==r&&s.push(e)}))),t.set.call(this,e),s)for(var u=0;u<s.length;u++){var l=s[u];1===l.__CE_state&&i.disconnectedCallback(l)}return this.ownerDocument.__CE_hasRegistry?h(i,this):o(i,this),e}})}function t(e,t){s(e,"insertAdjacentElement",(function(e,r){var a=n(r);return e=t.call(this,e,r),a&&c(i,r),n(e)&&l(i,r),e}))}function r(e,t){function n(e,t){for(var n=[];e!==t;e=e.nextSibling)n.push(e);for(t=0;t<n.length;t++)h(i,n[t])}s(e,"insertAdjacentHTML",(function(e,r){if("beforebegin"===(e=e.toLowerCase())){var a=this.previousSibling;t.call(this,e,r),n(a||this.parentNode.firstChild,this)}else if("afterbegin"===e)a=this.firstChild,t.call(this,e,r),n(this.firstChild,a);else if("beforeend"===e)a=this.lastChild,t.call(this,e,r),n(a||this.firstChild,null);else{if("afterend"!==e)throw new SyntaxError("The value provided ("+String(e)+") is not one of 'beforebegin', 'afterbegin', 'beforeend', or 'afterend'.");a=this.nextSibling,t.call(this,e,r),n(this.nextSibling,a)}}))}var i=te;_&&s(Element.prototype,"attachShadow",(function(e){e=_.call(this,e);var t=i;if(t.b&&!e.__CE_patched){e.__CE_patched=!0;for(var n=0;n<t.c.length;n++)t.c[n](e)}return this.__CE_shadowRoot=e})),F&&F.get?e(Element.prototype,F):K&&K.get?e(HTMLElement.prototype,K):function(e,t){e.b=!0,e.f.push(t)}(i,(function(t){e(t,{enumerable:!0,configurable:!0,get:function(){return A.call(this,!0).innerHTML},set:function(e){var t="template"===this.localName,n=t?this.content:this,r=v.call(document,this.namespaceURI,this.localName);for(r.innerHTML=e;0<n.childNodes.length;)C.call(n,n.childNodes[0]);for(e=t?r.content:r;0<e.childNodes.length;)E.call(n,e.childNodes[0])}})})),s(Element.prototype,"setAttribute",(function(e,t){if(1!==this.__CE_state)return D.call(this,e,t);var n=M.call(this,e);D.call(this,e,t),t=M.call(this,e),i.attributeChangedCallback(this,e,n,t,null)})),s(Element.prototype,"setAttributeNS",(function(e,t,n){if(1!==this.__CE_state)return P.call(this,e,t,n);var r=L.call(this,e,t);P.call(this,e,t,n),n=L.call(this,e,t),i.attributeChangedCallback(this,t,r,n,e)})),s(Element.prototype,"removeAttribute",(function(e){if(1!==this.__CE_state)return O.call(this,e);var t=M.call(this,e);O.call(this,e),null!==t&&i.attributeChangedCallback(this,e,t,null,null)})),s(Element.prototype,"removeAttributeNS",(function(e,t){if(1!==this.__CE_state)return $.call(this,e,t);var n=L.call(this,e,t);$.call(this,e,t);var r=L.call(this,e,t);n!==r&&i.attributeChangedCallback(this,t,n,r,e)})),Y?t(HTMLElement.prototype,Y):B?t(Element.prototype,B):console.warn("Custom Elements: `Element#insertAdjacentElement` was not patched."),X?r(HTMLElement.prototype,X):z?r(Element.prototype,z):console.warn("Custom Elements: `Element#insertAdjacentHTML` was not patched."),Z(i,Element.prototype,{h:U,append:W}),function(e){function t(t){return function(r){for(var a=[],s=0;s<arguments.length;++s)a[s]=arguments[s];s=[];for(var i=[],o=0;o<a.length;o++){var u=a[o];if(u instanceof Element&&n(u)&&i.push(u),u instanceof DocumentFragment)for(u=u.firstChild;u;u=u.nextSibling)s.push(u);else s.push(u)}for(t.apply(this,a),a=0;a<i.length;a++)c(e,i[a]);if(n(this))for(a=0;a<s.length;a++)(i=s[a])instanceof Element&&l(e,i)}}var r=Element.prototype;void 0!==V&&(r.before=t(V)),void 0!==V&&(r.after=t(j)),void 0!==G&&s(r,"replaceWith",(function(t){for(var r=[],a=0;a<arguments.length;++a)r[a]=arguments[a];a=[];for(var s=[],i=0;i<r.length;i++){var o=r[i];if(o instanceof Element&&n(o)&&s.push(o),o instanceof DocumentFragment)for(o=o.firstChild;o;o=o.nextSibling)a.push(o);else a.push(o)}for(i=n(this),G.apply(this,r),r=0;r<s.length;r++)c(e,s[r]);if(i)for(c(e,this),r=0;r<a.length;r++)(s=a[r])instanceof Element&&l(e,s)})),void 0!==H&&s(r,"remove",(function(){var t=n(this);H.call(this),t&&c(e,this)}))}(i)}(),document.__CE_hasRegistry=!0;var ne=new y(te);Object.defineProperty(window,"customElements",{configurable:!0,enumerable:!0,value:ne})}}).call(self),"string"!=typeof document.baseURI&&Object.defineProperty(Document.prototype,"baseURI",{enumerable:!0,configurable:!0,get:function(){var e=document.querySelector("base");return e&&e.href?e.href:document.URL}}),"function"!=typeof window.CustomEvent&&(window.CustomEvent=function(e,t){t=t||{bubbles:!1,cancelable:!1,detail:void 0};var n=document.createEvent("CustomEvent");return n.initCustomEvent(e,t.bubbles,t.cancelable,t.detail),n},window.CustomEvent.prototype=window.Event.prototype),n=Event.prototype,r=document,a=window,n.composedPath||(n.composedPath=function(){if(this.path)return this.path;var e=this.target;for(this.path=[];null!==e.parentNode;)this.path.push(e),e=e.parentNode;return this.path.push(r,a),this.path}),"function"!=typeof(
-/*!
-Element.closest and Element.matches
-https://github.com/jonathantneal/closest
-Creative Commons Zero v1.0 Universal
-*/
-s=window.Element.prototype).matches&&(s.matches=s.msMatchesSelector||s.mozMatchesSelector||s.webkitMatchesSelector||function(e){e=(this.document||this.ownerDocument).querySelectorAll(e);for(var t=0;e[t]&&e[t]!==this;)++t;return!!e[t]}),"function"!=typeof s.closest&&(s.closest=function(e){for(var t=this;t&&1===t.nodeType;){if(t.matches(e))return t;t=t.parentNode}return null}),
-/*!
-Element.getRootNode()
-*/
-function(e){function t(e){return e&&e.parentNode?t(e.parentNode):e}"function"!=typeof e.getRootNode&&(e.getRootNode=function(e){return e&&e.composed?function e(n){return(n=t(n))&&11===n.nodeType?e(n.host):n}(this):t(this)})}(Element.prototype),
-/*!
-Element.isConnected()
-*/
-function(e){"isConnected"in e||Object.defineProperty(e,"isConnected",{configurable:!0,enumerable:!0,get:function(){var e=this.getRootNode({composed:!0});return e&&9===e.nodeType}})}(Element.prototype),[Element.prototype,CharacterData.prototype,DocumentType.prototype].forEach((function(e){e.hasOwnProperty("remove")||Object.defineProperty(e,"remove",{configurable:!0,enumerable:!0,writable:!0,value:function(){null!==this.parentNode&&this.parentNode.removeChild(this)}})})),"classList"in(i=Element.prototype)||Object.defineProperty(i,"classList",{get:function(){var e=this,t=(e.getAttribute("class")||"").replace(/^\s+|\s$/g,"").split(/\s+/g);function n(){t.length>0?e.setAttribute("class",t.join(" ")):e.removeAttribute("class")}return""===t[0]&&t.splice(0,1),t.toggle=function(e,r){void 0!==r?r?t.add(e):t.remove(e):-1!==t.indexOf(e)?t.splice(t.indexOf(e),1):t.push(e),n()},t.add=function(){for(var e=[].slice.call(arguments),r=0,a=e.length;r<a;r++)-1===t.indexOf(e[r])&&t.push(e[r]);n()},t.remove=function(){for(var e=[].slice.call(arguments),r=0,a=e.length;r<a;r++)-1!==t.indexOf(e[r])&&t.splice(t.indexOf(e[r]),1);n()},t.item=function(e){return t[e]},t.contains=function(e){return-1!==t.indexOf(e)},t.replace=function(e,r){-1!==t.indexOf(e)&&t.splice(t.indexOf(e),1,r),n()},t.value=e.getAttribute("class")||"",t}}),
-/*!
-DOMTokenList
-*/
-function(e){try{document.body.classList.add()}catch(r){var t=e.add,n=e.remove;e.add=function(){for(var e=0;e<arguments.length;e++)t.call(this,arguments[e])},e.remove=function(){for(var e=0;e<arguments.length;e++)n.call(this,arguments[e])}}}(DOMTokenList.prototype)},function(e,t,n){(function(e){!function(t){"use strict";!function(e){var t={};function n(r){if(t[r])return t[r].exports;var a=t[r]={i:r,l:!1,exports:{}};return e[r].call(a.exports,a,a.exports,n),a.l=!0,a.exports}n.m=e,n.c=t,n.d=function(e,t,r){n.o(e,t)||Object.defineProperty(e,t,{enumerable:!0,get:r})},n.r=function(e){"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},n.t=function(e,t){if(1&t&&(e=n(e)),8&t)return e;if(4&t&&"object"==typeof e&&e&&e.__esModule)return e;var r=Object.create(null);if(n.r(r),Object.defineProperty(r,"default",{enumerable:!0,value:e}),2&t&&"string"!=typeof e)for(var a in e)n.d(r,a,function(t){return e[t]}.bind(null,a));return r},n.n=function(e){var t=e&&e.__esModule?function(){return e.default}:function(){return e};return n.d(t,"a",t),t},n.o=function(e,t){return Object.prototype.hasOwnProperty.call(e,t)},n.p="",n(n.s=0)}([function(e,t,n){n(1),n(55),n(62),n(68),n(70),n(71),n(72),n(73),n(75),n(76),n(78),n(87),n(88),n(89),n(98),n(99),n(101),n(102),n(103),n(105),n(106),n(107),n(108),n(110),n(111),n(112),n(113),n(114),n(115),n(116),n(117),n(118),n(127),n(130),n(131),n(133),n(135),n(136),n(137),n(138),n(139),n(141),n(143),n(146),n(148),n(150),n(151),n(153),n(154),n(155),n(156),n(157),n(159),n(160),n(162),n(163),n(164),n(165),n(166),n(167),n(168),n(169),n(170),n(172),n(173),n(183),n(184),n(185),n(189),n(191),n(192),n(193),n(194),n(195),n(196),n(198),n(201),n(202),n(203),n(204),n(208),n(209),n(212),n(213),n(214),n(215),n(216),n(217),n(218),n(219),n(221),n(222),n(223),n(226),n(227),n(228),n(229),n(230),n(231),n(232),n(233),n(234),n(235),n(236),n(237),n(238),n(240),n(241),n(243),n(248),e.exports=n(246)},function(e,t,n){var r=n(2),a=n(6),s=n(45),i=n(14),o=n(46),u=n(39),l=n(47),c=n(48),h=n(52),p=n(49),d=n(53),f=p("isConcatSpreadable"),m=d>=51||!a((function(){var e=[];return e[f]=!1,e.concat()[0]!==e})),g=h("concat"),y=function(e){if(!i(e))return!1;var t=e[f];return void 0!==t?!!t:s(e)};r({target:"Array",proto:!0,forced:!m||!g},{concat:function(e){var t,n,r,a,s,i=o(this),h=c(i,0),p=0;for(t=-1,r=arguments.length;t<r;t++)if(y(s=-1===t?i:arguments[t])){if(p+(a=u(s.length))>9007199254740991)throw TypeError("Maximum allowed index exceeded");for(n=0;n<a;n++,p++)n in s&&l(h,p,s[n])}else{if(p>=9007199254740991)throw TypeError("Maximum allowed index exceeded");l(h,p++,s)}return h.length=p,h}})},function(e,t,n){var r=n(3),a=n(4).f,s=n(18),i=n(21),o=n(22),u=n(32),l=n(44);e.exports=function(e,t){var n,c,h,p,d,f=e.target,m=e.global,g=e.stat;if(n=m?r:g?r[f]||o(f,{}):(r[f]||{}).prototype)for(c in t){if(p=t[c],h=e.noTargetGet?(d=a(n,c))&&d.value:n[c],!l(m?c:f+(g?".":"#")+c,e.forced)&&void 0!==h){if(typeof p==typeof h)continue;u(p,h)}(e.sham||h&&h.sham)&&s(p,"sham",!0),i(n,c,p,e)}}},function(t,n){var r=function(e){return e&&e.Math==Math&&e};t.exports=r("object"==typeof globalThis&&globalThis)||r("object"==typeof window&&window)||r("object"==typeof self&&self)||r("object"==typeof e&&e)||Function("return this")()},function(e,t,n){var r=n(5),a=n(7),s=n(8),i=n(9),o=n(13),u=n(15),l=n(16),c=Object.getOwnPropertyDescriptor;t.f=r?c:function(e,t){if(e=i(e),t=o(t,!0),l)try{return c(e,t)}catch(e){}if(u(e,t))return s(!a.f.call(e,t),e[t])}},function(e,t,n){var r=n(6);e.exports=!r((function(){return 7!=Object.defineProperty({},1,{get:function(){return 7}})[1]}))},function(e,t){e.exports=function(e){try{return!!e()}catch(e){return!0}}},function(e,t,n){var r={}.propertyIsEnumerable,a=Object.getOwnPropertyDescriptor,s=a&&!r.call({1:2},1);t.f=s?function(e){var t=a(this,e);return!!t&&t.enumerable}:r},function(e,t){e.exports=function(e,t){return{enumerable:!(1&e),configurable:!(2&e),writable:!(4&e),value:t}}},function(e,t,n){var r=n(10),a=n(12);e.exports=function(e){return r(a(e))}},function(e,t,n){var r=n(6),a=n(11),s="".split;e.exports=r((function(){return!Object("z").propertyIsEnumerable(0)}))?function(e){return"String"==a(e)?s.call(e,""):Object(e)}:Object},function(e,t){var n={}.toString;e.exports=function(e){return n.call(e).slice(8,-1)}},function(e,t){e.exports=function(e){if(null==e)throw TypeError("Can't call method on "+e);return e}},function(e,t,n){var r=n(14);e.exports=function(e,t){if(!r(e))return e;var n,a;if(t&&"function"==typeof(n=e.toString)&&!r(a=n.call(e)))return a;if("function"==typeof(n=e.valueOf)&&!r(a=n.call(e)))return a;if(!t&&"function"==typeof(n=e.toString)&&!r(a=n.call(e)))return a;throw TypeError("Can't convert object to primitive value")}},function(e,t){e.exports=function(e){return"object"==typeof e?null!==e:"function"==typeof e}},function(e,t){var n={}.hasOwnProperty;e.exports=function(e,t){return n.call(e,t)}},function(e,t,n){var r=n(5),a=n(6),s=n(17);e.exports=!r&&!a((function(){return 7!=Object.defineProperty(s("div"),"a",{get:function(){return 7}}).a}))},function(e,t,n){var r=n(3),a=n(14),s=r.document,i=a(s)&&a(s.createElement);e.exports=function(e){return i?s.createElement(e):{}}},function(e,t,n){var r=n(5),a=n(19),s=n(8);e.exports=r?function(e,t,n){return a.f(e,t,s(1,n))}:function(e,t,n){return e[t]=n,e}},function(e,t,n){var r=n(5),a=n(16),s=n(20),i=n(13),o=Object.defineProperty;t.f=r?o:function(e,t,n){if(s(e),t=i(t,!0),s(n),a)try{return o(e,t,n)}catch(e){}if("get"in n||"set"in n)throw TypeError("Accessors not supported");return"value"in n&&(e[t]=n.value),e}},function(e,t,n){var r=n(14);e.exports=function(e){if(!r(e))throw TypeError(String(e)+" is not an object");return e}},function(e,t,n){var r=n(3),a=n(18),s=n(15),i=n(22),o=n(23),u=n(25),l=u.get,c=u.enforce,h=String(String).split("String");(e.exports=function(e,t,n,o){var u=!!o&&!!o.unsafe,l=!!o&&!!o.enumerable,p=!!o&&!!o.noTargetGet;"function"==typeof n&&("string"!=typeof t||s(n,"name")||a(n,"name",t),c(n).source=h.join("string"==typeof t?t:"")),e!==r?(u?!p&&e[t]&&(l=!0):delete e[t],l?e[t]=n:a(e,t,n)):l?e[t]=n:i(t,n)})(Function.prototype,"toString",(function(){return"function"==typeof this&&l(this).source||o(this)}))},function(e,t,n){var r=n(3),a=n(18);e.exports=function(e,t){try{a(r,e,t)}catch(n){r[e]=t}return t}},function(e,t,n){var r=n(24),a=Function.toString;"function"!=typeof r.inspectSource&&(r.inspectSource=function(e){return a.call(e)}),e.exports=r.inspectSource},function(e,t,n){var r=n(3),a=n(22),s=r["__core-js_shared__"]||a("__core-js_shared__",{});e.exports=s},function(e,t,n){var r,a,s,i=n(26),o=n(3),u=n(14),l=n(18),c=n(15),h=n(27),p=n(31),d=o.WeakMap;if(i){var f=new d,m=f.get,g=f.has,y=f.set;r=function(e,t){return y.call(f,e,t),t},a=function(e){return m.call(f,e)||{}},s=function(e){return g.call(f,e)}}else{var b=h("state");p[b]=!0,r=function(e,t){return l(e,b,t),t},a=function(e){return c(e,b)?e[b]:{}},s=function(e){return c(e,b)}}e.exports={set:r,get:a,has:s,enforce:function(e){return s(e)?a(e):r(e,{})},getterFor:function(e){return function(t){var n;if(!u(t)||(n=a(t)).type!==e)throw TypeError("Incompatible receiver, "+e+" required");return n}}}},function(e,t,n){var r=n(3),a=n(23),s=r.WeakMap;e.exports="function"==typeof s&&/native code/.test(a(s))},function(e,t,n){var r=n(28),a=n(30),s=r("keys");e.exports=function(e){return s[e]||(s[e]=a(e))}},function(e,t,n){var r=n(29),a=n(24);(e.exports=function(e,t){return a[e]||(a[e]=void 0!==t?t:{})})("versions",[]).push({version:"3.6.5",mode:r?"pure":"global",copyright:"© 2020 Denis Pushkarev (zloirock.ru)"})},function(e,t){e.exports=!1},function(e,t){var n=0,r=Math.random();e.exports=function(e){return"Symbol("+String(void 0===e?"":e)+")_"+(++n+r).toString(36)}},function(e,t){e.exports={}},function(e,t,n){var r=n(15),a=n(33),s=n(4),i=n(19);e.exports=function(e,t){for(var n=a(t),o=i.f,u=s.f,l=0;l<n.length;l++){var c=n[l];r(e,c)||o(e,c,u(t,c))}}},function(e,t,n){var r=n(34),a=n(36),s=n(43),i=n(20);e.exports=r("Reflect","ownKeys")||function(e){var t=a.f(i(e)),n=s.f;return n?t.concat(n(e)):t}},function(e,t,n){var r=n(35),a=n(3),s=function(e){return"function"==typeof e?e:void 0};e.exports=function(e,t){return arguments.length<2?s(r[e])||s(a[e]):r[e]&&r[e][t]||a[e]&&a[e][t]}},function(e,t,n){var r=n(3);e.exports=r},function(e,t,n){var r=n(37),a=n(42).concat("length","prototype");t.f=Object.getOwnPropertyNames||function(e){return r(e,a)}},function(e,t,n){var r=n(15),a=n(9),s=n(38).indexOf,i=n(31);e.exports=function(e,t){var n,o=a(e),u=0,l=[];for(n in o)!r(i,n)&&r(o,n)&&l.push(n);for(;t.length>u;)r(o,n=t[u++])&&(~s(l,n)||l.push(n));return l}},function(e,t,n){var r=n(9),a=n(39),s=n(41),i=function(e){return function(t,n,i){var o,u=r(t),l=a(u.length),c=s(i,l);if(e&&n!=n){for(;l>c;)if((o=u[c++])!=o)return!0}else for(;l>c;c++)if((e||c in u)&&u[c]===n)return e||c||0;return!e&&-1}};e.exports={includes:i(!0),indexOf:i(!1)}},function(e,t,n){var r=n(40),a=Math.min;e.exports=function(e){return e>0?a(r(e),9007199254740991):0}},function(e,t){var n=Math.ceil,r=Math.floor;e.exports=function(e){return isNaN(e=+e)?0:(e>0?r:n)(e)}},function(e,t,n){var r=n(40),a=Math.max,s=Math.min;e.exports=function(e,t){var n=r(e);return n<0?a(n+t,0):s(n,t)}},function(e,t){e.exports=["constructor","hasOwnProperty","isPrototypeOf","propertyIsEnumerable","toLocaleString","toString","valueOf"]},function(e,t){t.f=Object.getOwnPropertySymbols},function(e,t,n){var r=n(6),a=/#|\.prototype\./,s=function(e,t){var n=o[i(e)];return n==l||n!=u&&("function"==typeof t?r(t):!!t)},i=s.normalize=function(e){return String(e).replace(a,".").toLowerCase()},o=s.data={},u=s.NATIVE="N",l=s.POLYFILL="P";e.exports=s},function(e,t,n){var r=n(11);e.exports=Array.isArray||function(e){return"Array"==r(e)}},function(e,t,n){var r=n(12);e.exports=function(e){return Object(r(e))}},function(e,t,n){var r=n(13),a=n(19),s=n(8);e.exports=function(e,t,n){var i=r(t);i in e?a.f(e,i,s(0,n)):e[i]=n}},function(e,t,n){var r=n(14),a=n(45),s=n(49)("species");e.exports=function(e,t){var n;return a(e)&&("function"!=typeof(n=e.constructor)||n!==Array&&!a(n.prototype)?r(n)&&null===(n=n[s])&&(n=void 0):n=void 0),new(void 0===n?Array:n)(0===t?0:t)}},function(e,t,n){var r=n(3),a=n(28),s=n(15),i=n(30),o=n(50),u=n(51),l=a("wks"),c=r.Symbol,h=u?c:c&&c.withoutSetter||i;e.exports=function(e){return s(l,e)||(o&&s(c,e)?l[e]=c[e]:l[e]=h("Symbol."+e)),l[e]}},function(e,t,n){var r=n(6);e.exports=!!Object.getOwnPropertySymbols&&!r((function(){return!String(Symbol())}))},function(e,t,n){var r=n(50);e.exports=r&&!Symbol.sham&&"symbol"==typeof Symbol.iterator},function(e,t,n){var r=n(6),a=n(49),s=n(53),i=a("species");e.exports=function(e){return s>=51||!r((function(){var t=[];return(t.constructor={})[i]=function(){return{foo:1}},1!==t[e](Boolean).foo}))}},function(e,t,n){var r,a,s=n(3),i=n(54),o=s.process,u=o&&o.versions,l=u&&u.v8;l?a=(r=l.split("."))[0]+r[1]:i&&(!(r=i.match(/Edge\/(\d+)/))||r[1]>=74)&&(r=i.match(/Chrome\/(\d+)/))&&(a=r[1]),e.exports=a&&+a},function(e,t,n){var r=n(34);e.exports=r("navigator","userAgent")||""},function(e,t,n){var r=n(2),a=n(56),s=n(57);r({target:"Array",proto:!0},{copyWithin:a}),s("copyWithin")},function(e,t,n){var r=n(46),a=n(41),s=n(39),i=Math.min;e.exports=[].copyWithin||function(e,t){var n=r(this),o=s(n.length),u=a(e,o),l=a(t,o),c=arguments.length>2?arguments[2]:void 0,h=i((void 0===c?o:a(c,o))-l,o-u),p=1;for(l<u&&u<l+h&&(p=-1,l+=h-1,u+=h-1);h-- >0;)l in n?n[u]=n[l]:delete n[u],u+=p,l+=p;return n}},function(e,t,n){var r=n(49),a=n(58),s=n(19),i=r("unscopables"),o=Array.prototype;null==o[i]&&s.f(o,i,{configurable:!0,value:a(null)}),e.exports=function(e){o[i][e]=!0}},function(e,t,n){var r,a=n(20),s=n(59),i=n(42),o=n(31),u=n(61),l=n(17),c=n(27)("IE_PROTO"),h=function(){},p=function(e){return"<script>"+e+"<\/script>"},d=function(){try{r=document.domain&&new ActiveXObject("htmlfile")}catch(e){}var e,t;d=r?function(e){e.write(p("")),e.close();var t=e.parentWindow.Object;return e=null,t}(r):((t=l("iframe")).style.display="none",u.appendChild(t),t.src=String("javascript:"),(e=t.contentWindow.document).open(),e.write(p("document.F=Object")),e.close(),e.F);for(var n=i.length;n--;)delete d.prototype[i[n]];return d()};o[c]=!0,e.exports=Object.create||function(e,t){var n;return null!==e?(h.prototype=a(e),n=new h,h.prototype=null,n[c]=e):n=d(),void 0===t?n:s(n,t)}},function(e,t,n){var r=n(5),a=n(19),s=n(20),i=n(60);e.exports=r?Object.defineProperties:function(e,t){s(e);for(var n,r=i(t),o=r.length,u=0;o>u;)a.f(e,n=r[u++],t[n]);return e}},function(e,t,n){var r=n(37),a=n(42);e.exports=Object.keys||function(e){return r(e,a)}},function(e,t,n){var r=n(34);e.exports=r("document","documentElement")},function(e,t,n){var r=n(2),a=n(63).every,s=n(66),i=n(67),o=s("every"),u=i("every");r({target:"Array",proto:!0,forced:!o||!u},{every:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(64),a=n(10),s=n(46),i=n(39),o=n(48),u=[].push,l=function(e){var t=1==e,n=2==e,l=3==e,c=4==e,h=6==e,p=5==e||h;return function(d,f,m,g){for(var y,b,v=s(d),x=a(v),w=r(f,m,3),k=i(x.length),S=0,I=g||o,A=t?I(d,k):n?I(d,0):void 0;k>S;S++)if((p||S in x)&&(b=w(y=x[S],S,v),e))if(t)A[S]=b;else if(b)switch(e){case 3:return!0;case 5:return y;case 6:return S;case 2:u.call(A,y)}else if(c)return!1;return h?-1:l||c?c:A}};e.exports={forEach:l(0),map:l(1),filter:l(2),some:l(3),every:l(4),find:l(5),findIndex:l(6)}},function(e,t,n){var r=n(65);e.exports=function(e,t,n){if(r(e),void 0===t)return e;switch(n){case 0:return function(){return e.call(t)};case 1:return function(n){return e.call(t,n)};case 2:return function(n,r){return e.call(t,n,r)};case 3:return function(n,r,a){return e.call(t,n,r,a)}}return function(){return e.apply(t,arguments)}}},function(e,t){e.exports=function(e){if("function"!=typeof e)throw TypeError(String(e)+" is not a function");return e}},function(e,t,n){var r=n(6);e.exports=function(e,t){var n=[][e];return!!n&&r((function(){n.call(null,t||function(){throw 1},1)}))}},function(e,t,n){var r=n(5),a=n(6),s=n(15),i=Object.defineProperty,o={},u=function(e){throw e};e.exports=function(e,t){if(s(o,e))return o[e];t||(t={});var n=[][e],l=!!s(t,"ACCESSORS")&&t.ACCESSORS,c=s(t,0)?t[0]:u,h=s(t,1)?t[1]:void 0;return o[e]=!!n&&!a((function(){if(l&&!r)return!0;var e={length:-1};l?i(e,1,{enumerable:!0,get:u}):e[1]=1,n.call(e,c,h)}))}},function(e,t,n){var r=n(2),a=n(69),s=n(57);r({target:"Array",proto:!0},{fill:a}),s("fill")},function(e,t,n){var r=n(46),a=n(41),s=n(39);e.exports=function(e){for(var t=r(this),n=s(t.length),i=arguments.length,o=a(i>1?arguments[1]:void 0,n),u=i>2?arguments[2]:void 0,l=void 0===u?n:a(u,n);l>o;)t[o++]=e;return t}},function(e,t,n){var r=n(2),a=n(63).filter,s=n(52),i=n(67),o=s("filter"),u=i("filter");r({target:"Array",proto:!0,forced:!o||!u},{filter:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(2),a=n(63).find,s=n(57),i=n(67),o=!0,u=i("find");"find"in[]&&Array(1).find((function(){o=!1})),r({target:"Array",proto:!0,forced:o||!u},{find:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}}),s("find")},function(e,t,n){var r=n(2),a=n(63).findIndex,s=n(57),i=n(67),o=!0,u=i("findIndex");"findIndex"in[]&&Array(1).findIndex((function(){o=!1})),r({target:"Array",proto:!0,forced:o||!u},{findIndex:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}}),s("findIndex")},function(e,t,n){var r=n(2),a=n(74),s=n(46),i=n(39),o=n(40),u=n(48);r({target:"Array",proto:!0},{flat:function(){var e=arguments.length?arguments[0]:void 0,t=s(this),n=i(t.length),r=u(t,0);return r.length=a(r,t,t,n,0,void 0===e?1:o(e)),r}})},function(e,t,n){var r=n(45),a=n(39),s=n(64),i=function(e,t,n,o,u,l,c,h){for(var p,d=u,f=0,m=!!c&&s(c,h,3);f<o;){if(f in n){if(p=m?m(n[f],f,t):n[f],l>0&&r(p))d=i(e,t,p,a(p.length),d,l-1)-1;else{if(d>=9007199254740991)throw TypeError("Exceed the acceptable array length");e[d]=p}d++}f++}return d};e.exports=i},function(e,t,n){var r=n(2),a=n(74),s=n(46),i=n(39),o=n(65),u=n(48);r({target:"Array",proto:!0},{flatMap:function(e){var t,n=s(this),r=i(n.length);return o(e),(t=u(n,0)).length=a(t,n,n,r,0,1,e,arguments.length>1?arguments[1]:void 0),t}})},function(e,t,n){var r=n(2),a=n(77);r({target:"Array",proto:!0,forced:[].forEach!=a},{forEach:a})},function(e,t,n){var r=n(63).forEach,a=n(66),s=n(67),i=a("forEach"),o=s("forEach");e.exports=i&&o?[].forEach:function(e){return r(this,e,arguments.length>1?arguments[1]:void 0)}},function(e,t,n){var r=n(2),a=n(79);r({target:"Array",stat:!0,forced:!n(86)((function(e){Array.from(e)}))},{from:a})},function(e,t,n){var r=n(64),a=n(46),s=n(80),i=n(81),o=n(39),u=n(47),l=n(83);e.exports=function(e){var t,n,c,h,p,d,f=a(e),m="function"==typeof this?this:Array,g=arguments.length,y=g>1?arguments[1]:void 0,b=void 0!==y,v=l(f),x=0;if(b&&(y=r(y,g>2?arguments[2]:void 0,2)),null==v||m==Array&&i(v))for(n=new m(t=o(f.length));t>x;x++)d=b?y(f[x],x):f[x],u(n,x,d);else for(p=(h=v.call(f)).next,n=new m;!(c=p.call(h)).done;x++)d=b?s(h,y,[c.value,x],!0):c.value,u(n,x,d);return n.length=x,n}},function(e,t,n){var r=n(20);e.exports=function(e,t,n,a){try{return a?t(r(n)[0],n[1]):t(n)}catch(t){var s=e.return;throw void 0!==s&&r(s.call(e)),t}}},function(e,t,n){var r=n(49),a=n(82),s=r("iterator"),i=Array.prototype;e.exports=function(e){return void 0!==e&&(a.Array===e||i[s]===e)}},function(e,t){e.exports={}},function(e,t,n){var r=n(84),a=n(82),s=n(49)("iterator");e.exports=function(e){if(null!=e)return e[s]||e["@@iterator"]||a[r(e)]}},function(e,t,n){var r=n(85),a=n(11),s=n(49)("toStringTag"),i="Arguments"==a(function(){return arguments}());e.exports=r?a:function(e){var t,n,r;return void 0===e?"Undefined":null===e?"Null":"string"==typeof(n=function(e,t){try{return e[t]}catch(e){}}(t=Object(e),s))?n:i?a(t):"Object"==(r=a(t))&&"function"==typeof t.callee?"Arguments":r}},function(e,t,n){var r={};r[n(49)("toStringTag")]="z",e.exports="[object z]"===String(r)},function(e,t,n){var r=n(49)("iterator"),a=!1;try{var s=0,i={next:function(){return{done:!!s++}},return:function(){a=!0}};i[r]=function(){return this},Array.from(i,(function(){throw 2}))}catch(e){}e.exports=function(e,t){if(!t&&!a)return!1;var n=!1;try{var s={};s[r]=function(){return{next:function(){return{done:n=!0}}}},e(s)}catch(e){}return n}},function(e,t,n){var r=n(2),a=n(38).includes,s=n(57);r({target:"Array",proto:!0,forced:!n(67)("indexOf",{ACCESSORS:!0,1:0})},{includes:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}}),s("includes")},function(e,t,n){var r=n(2),a=n(38).indexOf,s=n(66),i=n(67),o=[].indexOf,u=!!o&&1/[1].indexOf(1,-0)<0,l=s("indexOf"),c=i("indexOf",{ACCESSORS:!0,1:0});r({target:"Array",proto:!0,forced:u||!l||!c},{indexOf:function(e){return u?o.apply(this,arguments)||0:a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(9),a=n(57),s=n(82),i=n(25),o=n(90),u=i.set,l=i.getterFor("Array Iterator");e.exports=o(Array,"Array",(function(e,t){u(this,{type:"Array Iterator",target:r(e),index:0,kind:t})}),(function(){var e=l(this),t=e.target,n=e.kind,r=e.index++;return!t||r>=t.length?(e.target=void 0,{value:void 0,done:!0}):"keys"==n?{value:r,done:!1}:"values"==n?{value:t[r],done:!1}:{value:[r,t[r]],done:!1}}),"values"),s.Arguments=s.Array,a("keys"),a("values"),a("entries")},function(e,t,n){var r=n(2),a=n(91),s=n(93),i=n(96),o=n(95),u=n(18),l=n(21),c=n(49),h=n(29),p=n(82),d=n(92),f=d.IteratorPrototype,m=d.BUGGY_SAFARI_ITERATORS,g=c("iterator"),y=function(){return this};e.exports=function(e,t,n,c,d,b,v){a(n,t,c);var x,w,k,S=function(e){if(e===d&&C)return C;if(!m&&e in E)return E[e];switch(e){case"keys":case"values":case"entries":return function(){return new n(this,e)}}return function(){return new n(this)}},I=t+" Iterator",A=!1,E=e.prototype,N=E[g]||E["@@iterator"]||d&&E[d],C=!m&&N||S(d),T="Array"==t&&E.entries||N;if(T&&(x=s(T.call(new e)),f!==Object.prototype&&x.next&&(h||s(x)===f||(i?i(x,f):"function"!=typeof x[g]&&u(x,g,y)),o(x,I,!0,!0),h&&(p[I]=y))),"values"==d&&N&&"values"!==N.name&&(A=!0,C=function(){return N.call(this)}),h&&!v||E[g]===C||u(E,g,C),p[t]=C,d)if(w={values:S("values"),keys:b?C:S("keys"),entries:S("entries")},v)for(k in w)(m||A||!(k in E))&&l(E,k,w[k]);else r({target:t,proto:!0,forced:m||A},w);return w}},function(e,t,n){var r=n(92).IteratorPrototype,a=n(58),s=n(8),i=n(95),o=n(82),u=function(){return this};e.exports=function(e,t,n){var l=t+" Iterator";return e.prototype=a(r,{next:s(1,n)}),i(e,l,!1,!0),o[l]=u,e}},function(e,t,n){var r,a,s,i=n(93),o=n(18),u=n(15),l=n(49),c=n(29),h=l("iterator"),p=!1;[].keys&&("next"in(s=[].keys())?(a=i(i(s)))!==Object.prototype&&(r=a):p=!0),null==r&&(r={}),c||u(r,h)||o(r,h,(function(){return this})),e.exports={IteratorPrototype:r,BUGGY_SAFARI_ITERATORS:p}},function(e,t,n){var r=n(15),a=n(46),s=n(27),i=n(94),o=s("IE_PROTO"),u=Object.prototype;e.exports=i?Object.getPrototypeOf:function(e){return e=a(e),r(e,o)?e[o]:"function"==typeof e.constructor&&e instanceof e.constructor?e.constructor.prototype:e instanceof Object?u:null}},function(e,t,n){var r=n(6);e.exports=!r((function(){function e(){}return e.prototype.constructor=null,Object.getPrototypeOf(new e)!==e.prototype}))},function(e,t,n){var r=n(19).f,a=n(15),s=n(49)("toStringTag");e.exports=function(e,t,n){e&&!a(e=n?e:e.prototype,s)&&r(e,s,{configurable:!0,value:t})}},function(e,t,n){var r=n(20),a=n(97);e.exports=Object.setPrototypeOf||("__proto__"in{}?function(){var e,t=!1,n={};try{(e=Object.getOwnPropertyDescriptor(Object.prototype,"__proto__").set).call(n,[]),t=n instanceof Array}catch(e){}return function(n,s){return r(n),a(s),t?e.call(n,s):n.__proto__=s,n}}():void 0)},function(e,t,n){var r=n(14);e.exports=function(e){if(!r(e)&&null!==e)throw TypeError("Can't set "+String(e)+" as a prototype");return e}},function(e,t,n){var r=n(2),a=n(10),s=n(9),i=n(66),o=[].join,u=a!=Object,l=i("join",",");r({target:"Array",proto:!0,forced:u||!l},{join:function(e){return o.call(s(this),void 0===e?",":e)}})},function(e,t,n){var r=n(2),a=n(100);r({target:"Array",proto:!0,forced:a!==[].lastIndexOf},{lastIndexOf:a})},function(e,t,n){var r=n(9),a=n(40),s=n(39),i=n(66),o=n(67),u=Math.min,l=[].lastIndexOf,c=!!l&&1/[1].lastIndexOf(1,-0)<0,h=i("lastIndexOf"),p=o("indexOf",{ACCESSORS:!0,1:0}),d=c||!h||!p;e.exports=d?function(e){if(c)return l.apply(this,arguments)||0;var t=r(this),n=s(t.length),i=n-1;for(arguments.length>1&&(i=u(i,a(arguments[1]))),i<0&&(i=n+i);i>=0;i--)if(i in t&&t[i]===e)return i||0;return-1}:l},function(e,t,n){var r=n(2),a=n(63).map,s=n(52),i=n(67),o=s("map"),u=i("map");r({target:"Array",proto:!0,forced:!o||!u},{map:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(2),a=n(6),s=n(47);r({target:"Array",stat:!0,forced:a((function(){function e(){}return!(Array.of.call(e)instanceof e)}))},{of:function(){for(var e=0,t=arguments.length,n=new("function"==typeof this?this:Array)(t);t>e;)s(n,e,arguments[e++]);return n.length=t,n}})},function(e,t,n){var r=n(2),a=n(104).left,s=n(66),i=n(67),o=s("reduce"),u=i("reduce",{1:0});r({target:"Array",proto:!0,forced:!o||!u},{reduce:function(e){return a(this,e,arguments.length,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(65),a=n(46),s=n(10),i=n(39),o=function(e){return function(t,n,o,u){r(n);var l=a(t),c=s(l),h=i(l.length),p=e?h-1:0,d=e?-1:1;if(o<2)for(;;){if(p in c){u=c[p],p+=d;break}if(p+=d,e?p<0:h<=p)throw TypeError("Reduce of empty array with no initial value")}for(;e?p>=0:h>p;p+=d)p in c&&(u=n(u,c[p],p,l));return u}};e.exports={left:o(!1),right:o(!0)}},function(e,t,n){var r=n(2),a=n(104).right,s=n(66),i=n(67),o=s("reduceRight"),u=i("reduce",{1:0});r({target:"Array",proto:!0,forced:!o||!u},{reduceRight:function(e){return a(this,e,arguments.length,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(2),a=n(14),s=n(45),i=n(41),o=n(39),u=n(9),l=n(47),c=n(49),h=n(52),p=n(67),d=h("slice"),f=p("slice",{ACCESSORS:!0,0:0,1:2}),m=c("species"),g=[].slice,y=Math.max;r({target:"Array",proto:!0,forced:!d||!f},{slice:function(e,t){var n,r,c,h=u(this),p=o(h.length),d=i(e,p),f=i(void 0===t?p:t,p);if(s(h)&&("function"!=typeof(n=h.constructor)||n!==Array&&!s(n.prototype)?a(n)&&null===(n=n[m])&&(n=void 0):n=void 0,n===Array||void 0===n))return g.call(h,d,f);for(r=new(void 0===n?Array:n)(y(f-d,0)),c=0;d<f;d++,c++)d in h&&l(r,c,h[d]);return r.length=c,r}})},function(e,t,n){var r=n(2),a=n(63).some,s=n(66),i=n(67),o=s("some"),u=i("some");r({target:"Array",proto:!0,forced:!o||!u},{some:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){n(109)("Array")},function(e,t,n){var r=n(34),a=n(19),s=n(49),i=n(5),o=s("species");e.exports=function(e){var t=r(e),n=a.f;i&&t&&!t[o]&&n(t,o,{configurable:!0,get:function(){return this}})}},function(e,t,n){var r=n(2),a=n(41),s=n(40),i=n(39),o=n(46),u=n(48),l=n(47),c=n(52),h=n(67),p=c("splice"),d=h("splice",{ACCESSORS:!0,0:0,1:2}),f=Math.max,m=Math.min;r({target:"Array",proto:!0,forced:!p||!d},{splice:function(e,t){var n,r,c,h,p,d,g=o(this),y=i(g.length),b=a(e,y),v=arguments.length;if(0===v?n=r=0:1===v?(n=0,r=y-b):(n=v-2,r=m(f(s(t),0),y-b)),y+n-r>9007199254740991)throw TypeError("Maximum allowed length exceeded");for(c=u(g,r),h=0;h<r;h++)(p=b+h)in g&&l(c,h,g[p]);if(c.length=r,n<r){for(h=b;h<y-r;h++)d=h+n,(p=h+r)in g?g[d]=g[p]:delete g[d];for(h=y;h>y-r+n;h--)delete g[h-1]}else if(n>r)for(h=y-r;h>b;h--)d=h+n-1,(p=h+r-1)in g?g[d]=g[p]:delete g[d];for(h=0;h<n;h++)g[h+b]=arguments[h+2];return g.length=y-r+n,c}})},function(e,t,n){n(57)("flat")},function(e,t,n){n(57)("flatMap")},function(e,t,n){var r=n(14),a=n(19),s=n(93),i=n(49)("hasInstance"),o=Function.prototype;i in o||a.f(o,i,{value:function(e){if("function"!=typeof this||!r(e))return!1;if(!r(this.prototype))return e instanceof this;for(;e=s(e);)if(this.prototype===e)return!0;return!1}})},function(e,t,n){var r=n(5),a=n(19).f,s=Function.prototype,i=s.toString,o=/^\s*function ([^ (]*)/;r&&!("name"in s)&&a(s,"name",{configurable:!0,get:function(){try{return i.call(this).match(o)[1]}catch(e){return""}}})},function(e,t,n){n(2)({global:!0},{globalThis:n(3)})},function(e,t,n){var r=n(2),a=n(34),s=n(6),i=a("JSON","stringify"),o=/[\uD800-\uDFFF]/g,u=/^[\uD800-\uDBFF]$/,l=/^[\uDC00-\uDFFF]$/,c=function(e,t,n){var r=n.charAt(t-1),a=n.charAt(t+1);return u.test(e)&&!l.test(a)||l.test(e)&&!u.test(r)?"\\u"+e.charCodeAt(0).toString(16):e},h=s((function(){return'"\\udf06\\ud834"'!==i("\udf06\ud834")||'"\\udead"'!==i("\udead")}));i&&r({target:"JSON",stat:!0,forced:h},{stringify:function(e,t,n){var r=i.apply(null,arguments);return"string"==typeof r?r.replace(o,c):r}})},function(e,t,n){var r=n(3);n(95)(r.JSON,"JSON",!0)},function(e,t,n){var r=n(119),a=n(125);e.exports=r("Map",(function(e){return function(){return e(this,arguments.length?arguments[0]:void 0)}}),a)},function(e,t,n){var r=n(2),a=n(3),s=n(44),i=n(21),o=n(120),u=n(122),l=n(123),c=n(14),h=n(6),p=n(86),d=n(95),f=n(124);e.exports=function(e,t,n){var m=-1!==e.indexOf("Map"),g=-1!==e.indexOf("Weak"),y=m?"set":"add",b=a[e],v=b&&b.prototype,x=b,w={},k=function(e){var t=v[e];i(v,e,"add"==e?function(e){return t.call(this,0===e?0:e),this}:"delete"==e?function(e){return!(g&&!c(e))&&t.call(this,0===e?0:e)}:"get"==e?function(e){return g&&!c(e)?void 0:t.call(this,0===e?0:e)}:"has"==e?function(e){return!(g&&!c(e))&&t.call(this,0===e?0:e)}:function(e,n){return t.call(this,0===e?0:e,n),this})};if(s(e,"function"!=typeof b||!(g||v.forEach&&!h((function(){(new b).entries().next()})))))x=n.getConstructor(t,e,m,y),o.REQUIRED=!0;else if(s(e,!0)){var S=new x,I=S[y](g?{}:-0,1)!=S,A=h((function(){S.has(1)})),E=p((function(e){new b(e)})),N=!g&&h((function(){for(var e=new b,t=5;t--;)e[y](t,t);return!e.has(-0)}));E||((x=t((function(t,n){l(t,x,e);var r=f(new b,t,x);return null!=n&&u(n,r[y],r,m),r}))).prototype=v,v.constructor=x),(A||N)&&(k("delete"),k("has"),m&&k("get")),(N||I)&&k(y),g&&v.clear&&delete v.clear}return w[e]=x,r({global:!0,forced:x!=b},w),d(x,e),g||n.setStrong(x,e,m),x}},function(e,t,n){var r=n(31),a=n(14),s=n(15),i=n(19).f,o=n(30),u=n(121),l=o("meta"),c=0,h=Object.isExtensible||function(){return!0},p=function(e){i(e,l,{value:{objectID:"O"+ ++c,weakData:{}}})},d=e.exports={REQUIRED:!1,fastKey:function(e,t){if(!a(e))return"symbol"==typeof e?e:("string"==typeof e?"S":"P")+e;if(!s(e,l)){if(!h(e))return"F";if(!t)return"E";p(e)}return e[l].objectID},getWeakData:function(e,t){if(!s(e,l)){if(!h(e))return!0;if(!t)return!1;p(e)}return e[l].weakData},onFreeze:function(e){return u&&d.REQUIRED&&h(e)&&!s(e,l)&&p(e),e}};r[l]=!0},function(e,t,n){var r=n(6);e.exports=!r((function(){return Object.isExtensible(Object.preventExtensions({}))}))},function(e,t,n){var r=n(20),a=n(81),s=n(39),i=n(64),o=n(83),u=n(80),l=function(e,t){this.stopped=e,this.result=t};(e.exports=function(e,t,n,c,h){var p,d,f,m,g,y,b,v=i(t,n,c?2:1);if(h)p=e;else{if("function"!=typeof(d=o(e)))throw TypeError("Target is not iterable");if(a(d)){for(f=0,m=s(e.length);m>f;f++)if((g=c?v(r(b=e[f])[0],b[1]):v(e[f]))&&g instanceof l)return g;return new l(!1)}p=d.call(e)}for(y=p.next;!(b=y.call(p)).done;)if("object"==typeof(g=u(p,v,b.value,c))&&g&&g instanceof l)return g;return new l(!1)}).stop=function(e){return new l(!0,e)}},function(e,t){e.exports=function(e,t,n){if(!(e instanceof t))throw TypeError("Incorrect "+(n?n+" ":"")+"invocation");return e}},function(e,t,n){var r=n(14),a=n(96);e.exports=function(e,t,n){var s,i;return a&&"function"==typeof(s=t.constructor)&&s!==n&&r(i=s.prototype)&&i!==n.prototype&&a(e,i),e}},function(e,t,n){var r=n(19).f,a=n(58),s=n(126),i=n(64),o=n(123),u=n(122),l=n(90),c=n(109),h=n(5),p=n(120).fastKey,d=n(25),f=d.set,m=d.getterFor;e.exports={getConstructor:function(e,t,n,l){var c=e((function(e,r){o(e,c,t),f(e,{type:t,index:a(null),first:void 0,last:void 0,size:0}),h||(e.size=0),null!=r&&u(r,e[l],e,n)})),d=m(t),g=function(e,t,n){var r,a,s=d(e),i=y(e,t);return i?i.value=n:(s.last=i={index:a=p(t,!0),key:t,value:n,previous:r=s.last,next:void 0,removed:!1},s.first||(s.first=i),r&&(r.next=i),h?s.size++:e.size++,"F"!==a&&(s.index[a]=i)),e},y=function(e,t){var n,r=d(e),a=p(t);if("F"!==a)return r.index[a];for(n=r.first;n;n=n.next)if(n.key==t)return n};return s(c.prototype,{clear:function(){for(var e=d(this),t=e.index,n=e.first;n;)n.removed=!0,n.previous&&(n.previous=n.previous.next=void 0),delete t[n.index],n=n.next;e.first=e.last=void 0,h?e.size=0:this.size=0},delete:function(e){var t=d(this),n=y(this,e);if(n){var r=n.next,a=n.previous;delete t.index[n.index],n.removed=!0,a&&(a.next=r),r&&(r.previous=a),t.first==n&&(t.first=r),t.last==n&&(t.last=a),h?t.size--:this.size--}return!!n},forEach:function(e){for(var t,n=d(this),r=i(e,arguments.length>1?arguments[1]:void 0,3);t=t?t.next:n.first;)for(r(t.value,t.key,this);t&&t.removed;)t=t.previous},has:function(e){return!!y(this,e)}}),s(c.prototype,n?{get:function(e){var t=y(this,e);return t&&t.value},set:function(e,t){return g(this,0===e?0:e,t)}}:{add:function(e){return g(this,e=0===e?0:e,e)}}),h&&r(c.prototype,"size",{get:function(){return d(this).size}}),c},setStrong:function(e,t,n){var r=t+" Iterator",a=m(t),s=m(r);l(e,t,(function(e,t){f(this,{type:r,target:e,state:a(e),kind:t,last:void 0})}),(function(){for(var e=s(this),t=e.kind,n=e.last;n&&n.removed;)n=n.previous;return e.target&&(e.last=n=n?n.next:e.state.first)?"keys"==t?{value:n.key,done:!1}:"values"==t?{value:n.value,done:!1}:{value:[n.key,n.value],done:!1}:(e.target=void 0,{value:void 0,done:!0})}),n?"entries":"values",!n,!0),c(t)}}},function(e,t,n){var r=n(21);e.exports=function(e,t,n){for(var a in t)r(e,a,t[a],n);return e}},function(e,t,n){var r=n(5),a=n(3),s=n(44),i=n(21),o=n(15),u=n(11),l=n(124),c=n(13),h=n(6),p=n(58),d=n(36).f,f=n(4).f,m=n(19).f,g=n(128).trim,y=a.Number,b=y.prototype,v="Number"==u(p(b)),x=function(e){var t,n,r,a,s,i,o,u,l=c(e,!1);if("string"==typeof l&&l.length>2)if(43===(t=(l=g(l)).charCodeAt(0))||45===t){if(88===(n=l.charCodeAt(2))||120===n)return NaN}else if(48===t){switch(l.charCodeAt(1)){case 66:case 98:r=2,a=49;break;case 79:case 111:r=8,a=55;break;default:return+l}for(i=(s=l.slice(2)).length,o=0;o<i;o++)if((u=s.charCodeAt(o))<48||u>a)return NaN;return parseInt(s,r)}return+l};if(s("Number",!y(" 0o1")||!y("0b1")||y("+0x1"))){for(var w,k=function(e){var t=arguments.length<1?0:e,n=this;return n instanceof k&&(v?h((function(){b.valueOf.call(n)})):"Number"!=u(n))?l(new y(x(t)),n,k):x(t)},S=r?d(y):"MAX_VALUE,MIN_VALUE,NaN,NEGATIVE_INFINITY,POSITIVE_INFINITY,EPSILON,isFinite,isInteger,isNaN,isSafeInteger,MAX_SAFE_INTEGER,MIN_SAFE_INTEGER,parseFloat,parseInt,isInteger".split(","),I=0;S.length>I;I++)o(y,w=S[I])&&!o(k,w)&&m(k,w,f(y,w));k.prototype=b,b.constructor=k,i(a,"Number",k)}},function(e,t,n){var r=n(12),a="["+n(129)+"]",s=RegExp("^"+a+a+"*"),i=RegExp(a+a+"*$"),o=function(e){return function(t){var n=String(r(t));return 1&e&&(n=n.replace(s,"")),2&e&&(n=n.replace(i,"")),n}};e.exports={start:o(1),end:o(2),trim:o(3)}},function(e,t){e.exports="\t\n\v\f\r                　\u2028\u2029\ufeff"},function(e,t,n){n(2)({target:"Number",stat:!0},{EPSILON:Math.pow(2,-52)})},function(e,t,n){n(2)({target:"Number",stat:!0},{isFinite:n(132)})},function(e,t,n){var r=n(3).isFinite;e.exports=Number.isFinite||function(e){return"number"==typeof e&&r(e)}},function(e,t,n){n(2)({target:"Number",stat:!0},{isInteger:n(134)})},function(e,t,n){var r=n(14),a=Math.floor;e.exports=function(e){return!r(e)&&isFinite(e)&&a(e)===e}},function(e,t,n){n(2)({target:"Number",stat:!0},{isNaN:function(e){return e!=e}})},function(e,t,n){var r=n(2),a=n(134),s=Math.abs;r({target:"Number",stat:!0},{isSafeInteger:function(e){return a(e)&&s(e)<=9007199254740991}})},function(e,t,n){n(2)({target:"Number",stat:!0},{MAX_SAFE_INTEGER:9007199254740991})},function(e,t,n){n(2)({target:"Number",stat:!0},{MIN_SAFE_INTEGER:-9007199254740991})},function(e,t,n){var r=n(2),a=n(140);r({target:"Number",stat:!0,forced:Number.parseFloat!=a},{parseFloat:a})},function(e,t,n){var r=n(3),a=n(128).trim,s=n(129),i=r.parseFloat,o=1/i(s+"-0")!=-1/0;e.exports=o?function(e){var t=a(String(e)),n=i(t);return 0===n&&"-"==t.charAt(0)?-0:n}:i},function(e,t,n){var r=n(2),a=n(142);r({target:"Number",stat:!0,forced:Number.parseInt!=a},{parseInt:a})},function(e,t,n){var r=n(3),a=n(128).trim,s=n(129),i=r.parseInt,o=/^[+-]?0[Xx]/,u=8!==i(s+"08")||22!==i(s+"0x16");e.exports=u?function(e,t){var n=a(String(e));return i(n,t>>>0||(o.test(n)?16:10))}:i},function(e,t,n){var r=n(2),a=n(40),s=n(144),i=n(145),o=n(6),u=1..toFixed,l=Math.floor,c=function(e,t,n){return 0===t?n:t%2==1?c(e,t-1,n*e):c(e*e,t/2,n)};r({target:"Number",proto:!0,forced:u&&("0.000"!==8e-5.toFixed(3)||"1"!==.9.toFixed(0)||"1.25"!==1.255.toFixed(2)||"1000000000000000128"!==(0xde0b6b3a7640080).toFixed(0))||!o((function(){u.call({})}))},{toFixed:function(e){var t,n,r,o,u=s(this),h=a(e),p=[0,0,0,0,0,0],d="",f="0",m=function(e,t){for(var n=-1,r=t;++n<6;)r+=e*p[n],p[n]=r%1e7,r=l(r/1e7)},g=function(e){for(var t=6,n=0;--t>=0;)n+=p[t],p[t]=l(n/e),n=n%e*1e7},y=function(){for(var e=6,t="";--e>=0;)if(""!==t||0===e||0!==p[e]){var n=String(p[e]);t=""===t?n:t+i.call("0",7-n.length)+n}return t};if(h<0||h>20)throw RangeError("Incorrect fraction digits");if(u!=u)return"NaN";if(u<=-1e21||u>=1e21)return String(u);if(u<0&&(d="-",u=-u),u>1e-21)if(n=(t=function(e){for(var t=0,n=e;n>=4096;)t+=12,n/=4096;for(;n>=2;)t+=1,n/=2;return t}(u*c(2,69,1))-69)<0?u*c(2,-t,1):u/c(2,t,1),n*=4503599627370496,(t=52-t)>0){for(m(0,n),r=h;r>=7;)m(1e7,0),r-=7;for(m(c(10,r,1),0),r=t-1;r>=23;)g(1<<23),r-=23;g(1<<r),m(1,1),g(2),f=y()}else m(0,n),m(1<<-t,0),f=y()+i.call("0",h);return h>0?d+((o=f.length)<=h?"0."+i.call("0",h-o)+f:f.slice(0,o-h)+"."+f.slice(o-h)):d+f}})},function(e,t,n){var r=n(11);e.exports=function(e){if("number"!=typeof e&&"Number"!=r(e))throw TypeError("Incorrect invocation");return+e}},function(e,t,n){var r=n(40),a=n(12);e.exports="".repeat||function(e){var t=String(a(this)),n="",s=r(e);if(s<0||s==1/0)throw RangeError("Wrong number of repetitions");for(;s>0;(s>>>=1)&&(t+=t))1&s&&(n+=t);return n}},function(e,t,n){var r=n(2),a=n(147);r({target:"Object",stat:!0,forced:Object.assign!==a},{assign:a})},function(e,t,n){var r=n(5),a=n(6),s=n(60),i=n(43),o=n(7),u=n(46),l=n(10),c=Object.assign,h=Object.defineProperty;e.exports=!c||a((function(){if(r&&1!==c({b:1},c(h({},"a",{enumerable:!0,get:function(){h(this,"b",{value:3,enumerable:!1})}}),{b:2})).b)return!0;var e={},t={},n=Symbol();return e[n]=7,"abcdefghijklmnopqrst".split("").forEach((function(e){t[e]=e})),7!=c({},e)[n]||"abcdefghijklmnopqrst"!=s(c({},t)).join("")}))?function(e,t){for(var n=u(e),a=arguments.length,c=1,h=i.f,p=o.f;a>c;)for(var d,f=l(arguments[c++]),m=h?s(f).concat(h(f)):s(f),g=m.length,y=0;g>y;)d=m[y++],r&&!p.call(f,d)||(n[d]=f[d]);return n}:c},function(e,t,n){var r=n(2),a=n(5),s=n(149),i=n(46),o=n(65),u=n(19);a&&r({target:"Object",proto:!0,forced:s},{__defineGetter__:function(e,t){u.f(i(this),e,{get:o(t),enumerable:!0,configurable:!0})}})},function(e,t,n){var r=n(29),a=n(3),s=n(6);e.exports=r||!s((function(){var e=Math.random();__defineSetter__.call(null,e,(function(){})),delete a[e]}))},function(e,t,n){var r=n(2),a=n(5),s=n(149),i=n(46),o=n(65),u=n(19);a&&r({target:"Object",proto:!0,forced:s},{__defineSetter__:function(e,t){u.f(i(this),e,{set:o(t),enumerable:!0,configurable:!0})}})},function(e,t,n){var r=n(2),a=n(152).entries;r({target:"Object",stat:!0},{entries:function(e){return a(e)}})},function(e,t,n){var r=n(5),a=n(60),s=n(9),i=n(7).f,o=function(e){return function(t){for(var n,o=s(t),u=a(o),l=u.length,c=0,h=[];l>c;)n=u[c++],r&&!i.call(o,n)||h.push(e?[n,o[n]]:o[n]);return h}};e.exports={entries:o(!0),values:o(!1)}},function(e,t,n){var r=n(2),a=n(121),s=n(6),i=n(14),o=n(120).onFreeze,u=Object.freeze;r({target:"Object",stat:!0,forced:s((function(){u(1)})),sham:!a},{freeze:function(e){return u&&i(e)?u(o(e)):e}})},function(e,t,n){var r=n(2),a=n(122),s=n(47);r({target:"Object",stat:!0},{fromEntries:function(e){var t={};return a(e,(function(e,n){s(t,e,n)}),void 0,!0),t}})},function(e,t,n){var r=n(2),a=n(6),s=n(9),i=n(4).f,o=n(5),u=a((function(){i(1)}));r({target:"Object",stat:!0,forced:!o||u,sham:!o},{getOwnPropertyDescriptor:function(e,t){return i(s(e),t)}})},function(e,t,n){var r=n(2),a=n(5),s=n(33),i=n(9),o=n(4),u=n(47);r({target:"Object",stat:!0,sham:!a},{getOwnPropertyDescriptors:function(e){for(var t,n,r=i(e),a=o.f,l=s(r),c={},h=0;l.length>h;)void 0!==(n=a(r,t=l[h++]))&&u(c,t,n);return c}})},function(e,t,n){var r=n(2),a=n(6),s=n(158).f;r({target:"Object",stat:!0,forced:a((function(){return!Object.getOwnPropertyNames(1)}))},{getOwnPropertyNames:s})},function(e,t,n){var r=n(9),a=n(36).f,s={}.toString,i="object"==typeof window&&window&&Object.getOwnPropertyNames?Object.getOwnPropertyNames(window):[];e.exports.f=function(e){return i&&"[object Window]"==s.call(e)?function(e){try{return a(e)}catch(e){return i.slice()}}(e):a(r(e))}},function(e,t,n){var r=n(2),a=n(6),s=n(46),i=n(93),o=n(94);r({target:"Object",stat:!0,forced:a((function(){i(1)})),sham:!o},{getPrototypeOf:function(e){return i(s(e))}})},function(e,t,n){n(2)({target:"Object",stat:!0},{is:n(161)})},function(e,t){e.exports=Object.is||function(e,t){return e===t?0!==e||1/e==1/t:e!=e&&t!=t}},function(e,t,n){var r=n(2),a=n(6),s=n(14),i=Object.isExtensible;r({target:"Object",stat:!0,forced:a((function(){i(1)}))},{isExtensible:function(e){return!!s(e)&&(!i||i(e))}})},function(e,t,n){var r=n(2),a=n(6),s=n(14),i=Object.isFrozen;r({target:"Object",stat:!0,forced:a((function(){i(1)}))},{isFrozen:function(e){return!s(e)||!!i&&i(e)}})},function(e,t,n){var r=n(2),a=n(6),s=n(14),i=Object.isSealed;r({target:"Object",stat:!0,forced:a((function(){i(1)}))},{isSealed:function(e){return!s(e)||!!i&&i(e)}})},function(e,t,n){var r=n(2),a=n(46),s=n(60);r({target:"Object",stat:!0,forced:n(6)((function(){s(1)}))},{keys:function(e){return s(a(e))}})},function(e,t,n){var r=n(2),a=n(5),s=n(149),i=n(46),o=n(13),u=n(93),l=n(4).f;a&&r({target:"Object",proto:!0,forced:s},{__lookupGetter__:function(e){var t,n=i(this),r=o(e,!0);do{if(t=l(n,r))return t.get}while(n=u(n))}})},function(e,t,n){var r=n(2),a=n(5),s=n(149),i=n(46),o=n(13),u=n(93),l=n(4).f;a&&r({target:"Object",proto:!0,forced:s},{__lookupSetter__:function(e){var t,n=i(this),r=o(e,!0);do{if(t=l(n,r))return t.set}while(n=u(n))}})},function(e,t,n){var r=n(2),a=n(14),s=n(120).onFreeze,i=n(121),o=n(6),u=Object.preventExtensions;r({target:"Object",stat:!0,forced:o((function(){u(1)})),sham:!i},{preventExtensions:function(e){return u&&a(e)?u(s(e)):e}})},function(e,t,n){var r=n(2),a=n(14),s=n(120).onFreeze,i=n(121),o=n(6),u=Object.seal;r({target:"Object",stat:!0,forced:o((function(){u(1)})),sham:!i},{seal:function(e){return u&&a(e)?u(s(e)):e}})},function(e,t,n){var r=n(85),a=n(21),s=n(171);r||a(Object.prototype,"toString",s,{unsafe:!0})},function(e,t,n){var r=n(85),a=n(84);e.exports=r?{}.toString:function(){return"[object "+a(this)+"]"}},function(e,t,n){var r=n(2),a=n(152).values;r({target:"Object",stat:!0},{values:function(e){return a(e)}})},function(e,t,n){var r,a,s,i,o=n(2),u=n(29),l=n(3),c=n(34),h=n(174),p=n(21),d=n(126),f=n(95),m=n(109),g=n(14),y=n(65),b=n(123),v=n(11),x=n(23),w=n(122),k=n(86),S=n(175),I=n(176).set,A=n(178),E=n(179),N=n(181),C=n(180),T=n(182),R=n(25),_=n(44),F=n(49),M=n(53),D=F("species"),O="Promise",L=R.get,P=R.set,$=R.getterFor(O),B=h,z=l.TypeError,U=l.document,W=l.process,V=c("fetch"),j=C.f,G=j,H="process"==v(W),q=!!(U&&U.createEvent&&l.dispatchEvent),K=_(O,(function(){if(x(B)===String(B)){if(66===M)return!0;if(!H&&"function"!=typeof PromiseRejectionEvent)return!0}if(u&&!B.prototype.finally)return!0;if(M>=51&&/native code/.test(B))return!1;var e=B.resolve(1),t=function(e){e((function(){}),(function(){}))};return(e.constructor={})[D]=t,!(e.then((function(){}))instanceof t)})),Y=K||!k((function(e){B.all(e).catch((function(){}))})),X=function(e){var t;return!(!g(e)||"function"!=typeof(t=e.then))&&t},Q=function(e,t,n){if(!t.notified){t.notified=!0;var r=t.reactions;A((function(){for(var a=t.value,s=1==t.state,i=0;r.length>i;){var o,u,l,c=r[i++],h=s?c.ok:c.fail,p=c.resolve,d=c.reject,f=c.domain;try{h?(s||(2===t.rejection&&te(e,t),t.rejection=1),!0===h?o=a:(f&&f.enter(),o=h(a),f&&(f.exit(),l=!0)),o===c.promise?d(z("Promise-chain cycle")):(u=X(o))?u.call(o,p,d):p(o)):d(a)}catch(e){f&&!l&&f.exit(),d(e)}}t.reactions=[],t.notified=!1,n&&!t.rejection&&J(e,t)}))}},Z=function(e,t,n){var r,a;q?((r=U.createEvent("Event")).promise=t,r.reason=n,r.initEvent(e,!1,!0),l.dispatchEvent(r)):r={promise:t,reason:n},(a=l["on"+e])?a(r):"unhandledrejection"===e&&N("Unhandled promise rejection",n)},J=function(e,t){I.call(l,(function(){var n,r=t.value;if(ee(t)&&(n=T((function(){H?W.emit("unhandledRejection",r,e):Z("unhandledrejection",e,r)})),t.rejection=H||ee(t)?2:1,n.error))throw n.value}))},ee=function(e){return 1!==e.rejection&&!e.parent},te=function(e,t){I.call(l,(function(){H?W.emit("rejectionHandled",e):Z("rejectionhandled",e,t.value)}))},ne=function(e,t,n,r){return function(a){e(t,n,a,r)}},re=function(e,t,n,r){t.done||(t.done=!0,r&&(t=r),t.value=n,t.state=2,Q(e,t,!0))},ae=function(e,t,n,r){if(!t.done){t.done=!0,r&&(t=r);try{if(e===n)throw z("Promise can't be resolved itself");var a=X(n);a?A((function(){var r={done:!1};try{a.call(n,ne(ae,e,r,t),ne(re,e,r,t))}catch(n){re(e,r,n,t)}})):(t.value=n,t.state=1,Q(e,t,!1))}catch(n){re(e,{done:!1},n,t)}}};K&&(B=function(e){b(this,B,O),y(e),r.call(this);var t=L(this);try{e(ne(ae,this,t),ne(re,this,t))}catch(e){re(this,t,e)}},(r=function(e){P(this,{type:O,done:!1,notified:!1,parent:!1,reactions:[],rejection:!1,state:0,value:void 0})}).prototype=d(B.prototype,{then:function(e,t){var n=$(this),r=j(S(this,B));return r.ok="function"!=typeof e||e,r.fail="function"==typeof t&&t,r.domain=H?W.domain:void 0,n.parent=!0,n.reactions.push(r),0!=n.state&&Q(this,n,!1),r.promise},catch:function(e){return this.then(void 0,e)}}),a=function(){var e=new r,t=L(e);this.promise=e,this.resolve=ne(ae,e,t),this.reject=ne(re,e,t)},C.f=j=function(e){return e===B||e===s?new a(e):G(e)},u||"function"!=typeof h||(i=h.prototype.then,p(h.prototype,"then",(function(e,t){var n=this;return new B((function(e,t){i.call(n,e,t)})).then(e,t)}),{unsafe:!0}),"function"==typeof V&&o({global:!0,enumerable:!0,forced:!0},{fetch:function(e){return E(B,V.apply(l,arguments))}}))),o({global:!0,wrap:!0,forced:K},{Promise:B}),f(B,O,!1,!0),m(O),s=c(O),o({target:O,stat:!0,forced:K},{reject:function(e){var t=j(this);return t.reject.call(void 0,e),t.promise}}),o({target:O,stat:!0,forced:u||K},{resolve:function(e){return E(u&&this===s?B:this,e)}}),o({target:O,stat:!0,forced:Y},{all:function(e){var t=this,n=j(t),r=n.resolve,a=n.reject,s=T((function(){var n=y(t.resolve),s=[],i=0,o=1;w(e,(function(e){var u=i++,l=!1;s.push(void 0),o++,n.call(t,e).then((function(e){l||(l=!0,s[u]=e,--o||r(s))}),a)})),--o||r(s)}));return s.error&&a(s.value),n.promise},race:function(e){var t=this,n=j(t),r=n.reject,a=T((function(){var a=y(t.resolve);w(e,(function(e){a.call(t,e).then(n.resolve,r)}))}));return a.error&&r(a.value),n.promise}})},function(e,t,n){var r=n(3);e.exports=r.Promise},function(e,t,n){var r=n(20),a=n(65),s=n(49)("species");e.exports=function(e,t){var n,i=r(e).constructor;return void 0===i||null==(n=r(i)[s])?t:a(n)}},function(e,t,n){var r,a,s,i=n(3),o=n(6),u=n(11),l=n(64),c=n(61),h=n(17),p=n(177),d=i.location,f=i.setImmediate,m=i.clearImmediate,g=i.process,y=i.MessageChannel,b=i.Dispatch,v=0,x={},w=function(e){if(x.hasOwnProperty(e)){var t=x[e];delete x[e],t()}},k=function(e){return function(){w(e)}},S=function(e){w(e.data)},I=function(e){i.postMessage(e+"",d.protocol+"//"+d.host)};f&&m||(f=function(e){for(var t=[],n=1;arguments.length>n;)t.push(arguments[n++]);return x[++v]=function(){("function"==typeof e?e:Function(e)).apply(void 0,t)},r(v),v},m=function(e){delete x[e]},"process"==u(g)?r=function(e){g.nextTick(k(e))}:b&&b.now?r=function(e){b.now(k(e))}:y&&!p?(s=(a=new y).port2,a.port1.onmessage=S,r=l(s.postMessage,s,1)):!i.addEventListener||"function"!=typeof postMessage||i.importScripts||o(I)||"file:"===d.protocol?r="onreadystatechange"in h("script")?function(e){c.appendChild(h("script")).onreadystatechange=function(){c.removeChild(this),w(e)}}:function(e){setTimeout(k(e),0)}:(r=I,i.addEventListener("message",S,!1))),e.exports={set:f,clear:m}},function(e,t,n){var r=n(54);e.exports=/(iphone|ipod|ipad).*applewebkit/i.test(r)},function(e,t,n){var r,a,s,i,o,u,l,c,h=n(3),p=n(4).f,d=n(11),f=n(176).set,m=n(177),g=h.MutationObserver||h.WebKitMutationObserver,y=h.process,b=h.Promise,v="process"==d(y),x=p(h,"queueMicrotask"),w=x&&x.value;w||(r=function(){var e,t;for(v&&(e=y.domain)&&e.exit();a;){t=a.fn,a=a.next;try{t()}catch(e){throw a?i():s=void 0,e}}s=void 0,e&&e.enter()},v?i=function(){y.nextTick(r)}:g&&!m?(o=!0,u=document.createTextNode(""),new g(r).observe(u,{characterData:!0}),i=function(){u.data=o=!o}):b&&b.resolve?(l=b.resolve(void 0),c=l.then,i=function(){c.call(l,r)}):i=function(){f.call(h,r)}),e.exports=w||function(e){var t={fn:e,next:void 0};s&&(s.next=t),a||(a=t,i()),s=t}},function(e,t,n){var r=n(20),a=n(14),s=n(180);e.exports=function(e,t){if(r(e),a(t)&&t.constructor===e)return t;var n=s.f(e);return(0,n.resolve)(t),n.promise}},function(e,t,n){var r=n(65),a=function(e){var t,n;this.promise=new e((function(e,r){if(void 0!==t||void 0!==n)throw TypeError("Bad Promise constructor");t=e,n=r})),this.resolve=r(t),this.reject=r(n)};e.exports.f=function(e){return new a(e)}},function(e,t,n){var r=n(3);e.exports=function(e,t){var n=r.console;n&&n.error&&(1===arguments.length?n.error(e):n.error(e,t))}},function(e,t){e.exports=function(e){try{return{error:!1,value:e()}}catch(e){return{error:!0,value:e}}}},function(e,t,n){var r=n(2),a=n(65),s=n(180),i=n(182),o=n(122);r({target:"Promise",stat:!0},{allSettled:function(e){var t=this,n=s.f(t),r=n.resolve,u=n.reject,l=i((function(){var n=a(t.resolve),s=[],i=0,u=1;o(e,(function(e){var a=i++,o=!1;s.push(void 0),u++,n.call(t,e).then((function(e){o||(o=!0,s[a]={status:"fulfilled",value:e},--u||r(s))}),(function(e){o||(o=!0,s[a]={status:"rejected",reason:e},--u||r(s))}))})),--u||r(s)}));return l.error&&u(l.value),n.promise}})},function(e,t,n){var r=n(2),a=n(29),s=n(174),i=n(6),o=n(34),u=n(175),l=n(179),c=n(21);r({target:"Promise",proto:!0,real:!0,forced:!!s&&i((function(){s.prototype.finally.call({then:function(){}},(function(){}))}))},{finally:function(e){var t=u(this,o("Promise")),n="function"==typeof e;return this.then(n?function(n){return l(t,e()).then((function(){return n}))}:e,n?function(n){return l(t,e()).then((function(){throw n}))}:e)}}),a||"function"!=typeof s||s.prototype.finally||c(s.prototype,"finally",o("Promise").prototype.finally)},function(e,t,n){var r=n(5),a=n(3),s=n(44),i=n(124),o=n(19).f,u=n(36).f,l=n(186),c=n(187),h=n(188),p=n(21),d=n(6),f=n(25).set,m=n(109),g=n(49)("match"),y=a.RegExp,b=y.prototype,v=/a/g,x=/a/g,w=new y(v)!==v,k=h.UNSUPPORTED_Y;if(r&&s("RegExp",!w||k||d((function(){return x[g]=!1,y(v)!=v||y(x)==x||"/a/i"!=y(v,"i")})))){for(var S=function(e,t){var n,r=this instanceof S,a=l(e),s=void 0===t;if(!r&&a&&e.constructor===S&&s)return e;w?a&&!s&&(e=e.source):e instanceof S&&(s&&(t=c.call(e)),e=e.source),k&&(n=!!t&&t.indexOf("y")>-1)&&(t=t.replace(/y/g,""));var o=i(w?new y(e,t):y(e,t),r?this:b,S);return k&&n&&f(o,{sticky:n}),o},I=function(e){e in S||o(S,e,{configurable:!0,get:function(){return y[e]},set:function(t){y[e]=t}})},A=u(y),E=0;A.length>E;)I(A[E++]);b.constructor=S,S.prototype=b,p(a,"RegExp",S)}m("RegExp")},function(e,t,n){var r=n(14),a=n(11),s=n(49)("match");e.exports=function(e){var t;return r(e)&&(void 0!==(t=e[s])?!!t:"RegExp"==a(e))}},function(e,t,n){var r=n(20);e.exports=function(){var e=r(this),t="";return e.global&&(t+="g"),e.ignoreCase&&(t+="i"),e.multiline&&(t+="m"),e.dotAll&&(t+="s"),e.unicode&&(t+="u"),e.sticky&&(t+="y"),t}},function(e,t,n){var r=n(6);function a(e,t){return RegExp(e,t)}t.UNSUPPORTED_Y=r((function(){var e=a("a","y");return e.lastIndex=2,null!=e.exec("abcd")})),t.BROKEN_CARET=r((function(){var e=a("^r","gy");return e.lastIndex=2,null!=e.exec("str")}))},function(e,t,n){var r=n(2),a=n(190);r({target:"RegExp",proto:!0,forced:/./.exec!==a},{exec:a})},function(e,t,n){var r,a,s=n(187),i=n(188),o=RegExp.prototype.exec,u=String.prototype.replace,l=o,c=(r=/a/,a=/b*/g,o.call(r,"a"),o.call(a,"a"),0!==r.lastIndex||0!==a.lastIndex),h=i.UNSUPPORTED_Y||i.BROKEN_CARET,p=void 0!==/()??/.exec("")[1];(c||p||h)&&(l=function(e){var t,n,r,a,i=this,l=h&&i.sticky,d=s.call(i),f=i.source,m=0,g=e;return l&&(-1===(d=d.replace("y","")).indexOf("g")&&(d+="g"),g=String(e).slice(i.lastIndex),i.lastIndex>0&&(!i.multiline||i.multiline&&"\n"!==e[i.lastIndex-1])&&(f="(?: "+f+")",g=" "+g,m++),n=new RegExp("^(?:"+f+")",d)),p&&(n=new RegExp("^"+f+"$(?!\\s)",d)),c&&(t=i.lastIndex),r=o.call(l?n:i,g),l?r?(r.input=r.input.slice(m),r[0]=r[0].slice(m),r.index=i.lastIndex,i.lastIndex+=r[0].length):i.lastIndex=0:c&&r&&(i.lastIndex=i.global?r.index+r[0].length:t),p&&r&&r.length>1&&u.call(r[0],n,(function(){for(a=1;a<arguments.length-2;a++)void 0===arguments[a]&&(r[a]=void 0)})),r}),e.exports=l},function(e,t,n){var r=n(5),a=n(19),s=n(187),i=n(188).UNSUPPORTED_Y;r&&("g"!=/./g.flags||i)&&a.f(RegExp.prototype,"flags",{configurable:!0,get:s})},function(e,t,n){var r=n(5),a=n(188).UNSUPPORTED_Y,s=n(19).f,i=n(25).get,o=RegExp.prototype;r&&a&&s(RegExp.prototype,"sticky",{configurable:!0,get:function(){if(this!==o){if(this instanceof RegExp)return!!i(this).sticky;throw TypeError("Incompatible receiver, RegExp required")}}})},function(e,t,n){n(189);var r,a,s=n(2),i=n(14),o=(r=!1,(a=/[ac]/).exec=function(){return r=!0,/./.exec.apply(this,arguments)},!0===a.test("abc")&&r),u=/./.test;s({target:"RegExp",proto:!0,forced:!o},{test:function(e){if("function"!=typeof this.exec)return u.call(this,e);var t=this.exec(e);if(null!==t&&!i(t))throw new Error("RegExp exec method returned something other than an Object or null");return!!t}})},function(e,t,n){var r=n(21),a=n(20),s=n(6),i=n(187),o=RegExp.prototype,u=o.toString,l=s((function(){return"/a/b"!=u.call({source:"a",flags:"b"})})),c="toString"!=u.name;(l||c)&&r(RegExp.prototype,"toString",(function(){var e=a(this),t=String(e.source),n=e.flags;return"/"+t+"/"+String(void 0===n&&e instanceof RegExp&&!("flags"in o)?i.call(e):n)}),{unsafe:!0})},function(e,t,n){var r=n(119),a=n(125);e.exports=r("Set",(function(e){return function(){return e(this,arguments.length?arguments[0]:void 0)}}),a)},function(e,t,n){var r=n(2),a=n(197).codeAt;r({target:"String",proto:!0},{codePointAt:function(e){return a(this,e)}})},function(e,t,n){var r=n(40),a=n(12),s=function(e){return function(t,n){var s,i,o=String(a(t)),u=r(n),l=o.length;return u<0||u>=l?e?"":void 0:(s=o.charCodeAt(u))<55296||s>56319||u+1===l||(i=o.charCodeAt(u+1))<56320||i>57343?e?o.charAt(u):s:e?o.slice(u,u+2):i-56320+(s-55296<<10)+65536}};e.exports={codeAt:s(!1),charAt:s(!0)}},function(e,t,n){var r,a=n(2),s=n(4).f,i=n(39),o=n(199),u=n(12),l=n(200),c=n(29),h="".endsWith,p=Math.min,d=l("endsWith");a({target:"String",proto:!0,forced:!(!c&&!d&&(r=s(String.prototype,"endsWith"),r&&!r.writable)||d)},{endsWith:function(e){var t=String(u(this));o(e);var n=arguments.length>1?arguments[1]:void 0,r=i(t.length),a=void 0===n?r:p(i(n),r),s=String(e);return h?h.call(t,s,a):t.slice(a-s.length,a)===s}})},function(e,t,n){var r=n(186);e.exports=function(e){if(r(e))throw TypeError("The method doesn't accept regular expressions");return e}},function(e,t,n){var r=n(49)("match");e.exports=function(e){var t=/./;try{"/./"[e](t)}catch(n){try{return t[r]=!1,"/./"[e](t)}catch(e){}}return!1}},function(e,t,n){var r=n(2),a=n(41),s=String.fromCharCode,i=String.fromCodePoint;r({target:"String",stat:!0,forced:!!i&&1!=i.length},{fromCodePoint:function(e){for(var t,n=[],r=arguments.length,i=0;r>i;){if(t=+arguments[i++],a(t,1114111)!==t)throw RangeError(t+" is not a valid code point");n.push(t<65536?s(t):s(55296+((t-=65536)>>10),t%1024+56320))}return n.join("")}})},function(e,t,n){var r=n(2),a=n(199),s=n(12);r({target:"String",proto:!0,forced:!n(200)("includes")},{includes:function(e){return!!~String(s(this)).indexOf(a(e),arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(197).charAt,a=n(25),s=n(90),i=a.set,o=a.getterFor("String Iterator");s(String,"String",(function(e){i(this,{type:"String Iterator",string:String(e),index:0})}),(function(){var e,t=o(this),n=t.string,a=t.index;return a>=n.length?{value:void 0,done:!0}:(e=r(n,a),t.index+=e.length,{value:e,done:!1})}))},function(e,t,n){var r=n(205),a=n(20),s=n(39),i=n(12),o=n(206),u=n(207);r("match",1,(function(e,t,n){return[function(t){var n=i(this),r=null==t?void 0:t[e];return void 0!==r?r.call(t,n):new RegExp(t)[e](String(n))},function(e){var r=n(t,e,this);if(r.done)return r.value;var i=a(e),l=String(this);if(!i.global)return u(i,l);var c=i.unicode;i.lastIndex=0;for(var h,p=[],d=0;null!==(h=u(i,l));){var f=String(h[0]);p[d]=f,""===f&&(i.lastIndex=o(l,s(i.lastIndex),c)),d++}return 0===d?null:p}]}))},function(e,t,n){n(189);var r=n(21),a=n(6),s=n(49),i=n(190),o=n(18),u=s("species"),l=!a((function(){var e=/./;return e.exec=function(){var e=[];return e.groups={a:"7"},e},"7"!=="".replace(e,"$<a>")})),c="$0"==="a".replace(/./,"$0"),h=s("replace"),p=!!/./[h]&&""===/./[h]("a","$0"),d=!a((function(){var e=/(?:)/,t=e.exec;e.exec=function(){return t.apply(this,arguments)};var n="ab".split(e);return 2!==n.length||"a"!==n[0]||"b"!==n[1]}));e.exports=function(e,t,n,h){var f=s(e),m=!a((function(){var t={};return t[f]=function(){return 7},7!=""[e](t)})),g=m&&!a((function(){var t=!1,n=/a/;return"split"===e&&((n={}).constructor={},n.constructor[u]=function(){return n},n.flags="",n[f]=/./[f]),n.exec=function(){return t=!0,null},n[f](""),!t}));if(!m||!g||"replace"===e&&(!l||!c||p)||"split"===e&&!d){var y=/./[f],b=n(f,""[e],(function(e,t,n,r,a){return t.exec===i?m&&!a?{done:!0,value:y.call(t,n,r)}:{done:!0,value:e.call(n,t,r)}:{done:!1}}),{REPLACE_KEEPS_$0:c,REGEXP_REPLACE_SUBSTITUTES_UNDEFINED_CAPTURE:p}),v=b[0],x=b[1];r(String.prototype,e,v),r(RegExp.prototype,f,2==t?function(e,t){return x.call(e,this,t)}:function(e){return x.call(e,this)})}h&&o(RegExp.prototype[f],"sham",!0)}},function(e,t,n){var r=n(197).charAt;e.exports=function(e,t,n){return t+(n?r(e,t).length:1)}},function(e,t,n){var r=n(11),a=n(190);e.exports=function(e,t){var n=e.exec;if("function"==typeof n){var s=n.call(e,t);if("object"!=typeof s)throw TypeError("RegExp exec method returned something other than an Object or null");return s}if("RegExp"!==r(e))throw TypeError("RegExp#exec called on incompatible receiver");return a.call(e,t)}},function(e,t,n){var r=n(2),a=n(91),s=n(12),i=n(39),o=n(65),u=n(20),l=n(11),c=n(186),h=n(187),p=n(18),d=n(6),f=n(49),m=n(175),g=n(206),y=n(25),b=n(29),v=f("matchAll"),x=y.set,w=y.getterFor("RegExp String Iterator"),k=RegExp.prototype,S=k.exec,I="".matchAll,A=!!I&&!d((function(){"a".matchAll(/./)})),E=a((function(e,t,n,r){x(this,{type:"RegExp String Iterator",regexp:e,string:t,global:n,unicode:r,done:!1})}),"RegExp String",(function(){var e=w(this);if(e.done)return{value:void 0,done:!0};var t=e.regexp,n=e.string,r=function(e,t){var n,r=e.exec;if("function"==typeof r){if("object"!=typeof(n=r.call(e,t)))throw TypeError("Incorrect exec result");return n}return S.call(e,t)}(t,n);return null===r?{value:void 0,done:e.done=!0}:e.global?(""==String(r[0])&&(t.lastIndex=g(n,i(t.lastIndex),e.unicode)),{value:r,done:!1}):(e.done=!0,{value:r,done:!1})})),N=function(e){var t,n,r,a,s,o,l=u(this),c=String(e);return t=m(l,RegExp),void 0===(n=l.flags)&&l instanceof RegExp&&!("flags"in k)&&(n=h.call(l)),r=void 0===n?"":String(n),a=new t(t===RegExp?l.source:l,r),s=!!~r.indexOf("g"),o=!!~r.indexOf("u"),a.lastIndex=i(l.lastIndex),new E(a,c,s,o)};r({target:"String",proto:!0,forced:A},{matchAll:function(e){var t,n,r,a=s(this);if(null!=e){if(c(e)&&!~String(s("flags"in k?e.flags:h.call(e))).indexOf("g"))throw TypeError("`.matchAll` does not allow non-global regexes");if(A)return I.apply(a,arguments);if(void 0===(n=e[v])&&b&&"RegExp"==l(e)&&(n=N),null!=n)return o(n).call(e,a)}else if(A)return I.apply(a,arguments);return t=String(a),r=new RegExp(e,"g"),b?N.call(r,t):r[v](t)}}),b||v in k||p(k,v,N)},function(e,t,n){var r=n(2),a=n(210).end;r({target:"String",proto:!0,forced:n(211)},{padEnd:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(39),a=n(145),s=n(12),i=Math.ceil,o=function(e){return function(t,n,o){var u,l,c=String(s(t)),h=c.length,p=void 0===o?" ":String(o),d=r(n);return d<=h||""==p?c:(u=d-h,(l=a.call(p,i(u/p.length))).length>u&&(l=l.slice(0,u)),e?c+l:l+c)}};e.exports={start:o(!1),end:o(!0)}},function(e,t,n){var r=n(54);e.exports=/Version\/10\.\d+(\.\d+)?( Mobile\/\w+)? Safari\//.test(r)},function(e,t,n){var r=n(2),a=n(210).start;r({target:"String",proto:!0,forced:n(211)},{padStart:function(e){return a(this,e,arguments.length>1?arguments[1]:void 0)}})},function(e,t,n){var r=n(2),a=n(9),s=n(39);r({target:"String",stat:!0},{raw:function(e){for(var t=a(e.raw),n=s(t.length),r=arguments.length,i=[],o=0;n>o;)i.push(String(t[o++])),o<r&&i.push(String(arguments[o]));return i.join("")}})},function(e,t,n){n(2)({target:"String",proto:!0},{repeat:n(145)})},function(e,t,n){var r=n(205),a=n(20),s=n(46),i=n(39),o=n(40),u=n(12),l=n(206),c=n(207),h=Math.max,p=Math.min,d=Math.floor,f=/\$([$&'`]|\d\d?|<[^>]*>)/g,m=/\$([$&'`]|\d\d?)/g;r("replace",2,(function(e,t,n,r){var g=r.REGEXP_REPLACE_SUBSTITUTES_UNDEFINED_CAPTURE,y=r.REPLACE_KEEPS_$0,b=g?"$":"$0";return[function(n,r){var a=u(this),s=null==n?void 0:n[e];return void 0!==s?s.call(n,a,r):t.call(String(a),n,r)},function(e,r){if(!g&&y||"string"==typeof r&&-1===r.indexOf(b)){var s=n(t,e,this,r);if(s.done)return s.value}var u=a(e),d=String(this),f="function"==typeof r;f||(r=String(r));var m=u.global;if(m){var x=u.unicode;u.lastIndex=0}for(var w=[];;){var k=c(u,d);if(null===k)break;if(w.push(k),!m)break;""===String(k[0])&&(u.lastIndex=l(d,i(u.lastIndex),x))}for(var S,I="",A=0,E=0;E<w.length;E++){k=w[E];for(var N=String(k[0]),C=h(p(o(k.index),d.length),0),T=[],R=1;R<k.length;R++)T.push(void 0===(S=k[R])?S:String(S));var _=k.groups;if(f){var F=[N].concat(T,C,d);void 0!==_&&F.push(_);var M=String(r.apply(void 0,F))}else M=v(N,d,C,T,_,r);C>=A&&(I+=d.slice(A,C)+M,A=C+N.length)}return I+d.slice(A)}];function v(e,n,r,a,i,o){var u=r+e.length,l=a.length,c=m;return void 0!==i&&(i=s(i),c=f),t.call(o,c,(function(t,s){var o;switch(s.charAt(0)){case"$":return"$";case"&":return e;case"`":return n.slice(0,r);case"'":return n.slice(u);case"<":o=i[s.slice(1,-1)];break;default:var c=+s;if(0===c)return t;if(c>l){var h=d(c/10);return 0===h?t:h<=l?void 0===a[h-1]?s.charAt(1):a[h-1]+s.charAt(1):t}o=a[c-1]}return void 0===o?"":o}))}}))},function(e,t,n){var r=n(205),a=n(20),s=n(12),i=n(161),o=n(207);r("search",1,(function(e,t,n){return[function(t){var n=s(this),r=null==t?void 0:t[e];return void 0!==r?r.call(t,n):new RegExp(t)[e](String(n))},function(e){var r=n(t,e,this);if(r.done)return r.value;var s=a(e),u=String(this),l=s.lastIndex;i(l,0)||(s.lastIndex=0);var c=o(s,u);return i(s.lastIndex,l)||(s.lastIndex=l),null===c?-1:c.index}]}))},function(e,t,n){var r=n(205),a=n(186),s=n(20),i=n(12),o=n(175),u=n(206),l=n(39),c=n(207),h=n(190),p=n(6),d=[].push,f=Math.min,m=!p((function(){return!RegExp(4294967295,"y")}));r("split",2,(function(e,t,n){var r;return r="c"=="abbc".split(/(b)*/)[1]||4!="test".split(/(?:)/,-1).length||2!="ab".split(/(?:ab)*/).length||4!=".".split(/(.?)(.?)/).length||".".split(/()()/).length>1||"".split(/.?/).length?function(e,n){var r=String(i(this)),s=void 0===n?4294967295:n>>>0;if(0===s)return[];if(void 0===e)return[r];if(!a(e))return t.call(r,e,s);for(var o,u,l,c=[],p=(e.ignoreCase?"i":"")+(e.multiline?"m":"")+(e.unicode?"u":"")+(e.sticky?"y":""),f=0,m=new RegExp(e.source,p+"g");(o=h.call(m,r))&&!((u=m.lastIndex)>f&&(c.push(r.slice(f,o.index)),o.length>1&&o.index<r.length&&d.apply(c,o.slice(1)),l=o[0].length,f=u,c.length>=s));)m.lastIndex===o.index&&m.lastIndex++;return f===r.length?!l&&m.test("")||c.push(""):c.push(r.slice(f)),c.length>s?c.slice(0,s):c}:"0".split(void 0,0).length?function(e,n){return void 0===e&&0===n?[]:t.call(this,e,n)}:t,[function(t,n){var a=i(this),s=null==t?void 0:t[e];return void 0!==s?s.call(t,a,n):r.call(String(a),t,n)},function(e,a){var i=n(r,e,this,a,r!==t);if(i.done)return i.value;var h=s(e),p=String(this),d=o(h,RegExp),g=h.unicode,y=(h.ignoreCase?"i":"")+(h.multiline?"m":"")+(h.unicode?"u":"")+(m?"y":"g"),b=new d(m?h:"^(?:"+h.source+")",y),v=void 0===a?4294967295:a>>>0;if(0===v)return[];if(0===p.length)return null===c(b,p)?[p]:[];for(var x=0,w=0,k=[];w<p.length;){b.lastIndex=m?w:0;var S,I=c(b,m?p:p.slice(w));if(null===I||(S=f(l(b.lastIndex+(m?0:w)),p.length))===x)w=u(p,w,g);else{if(k.push(p.slice(x,w)),k.length===v)return k;for(var A=1;A<=I.length-1;A++)if(k.push(I[A]),k.length===v)return k;w=x=S}}return k.push(p.slice(x)),k}]}),!m)},function(e,t,n){var r,a=n(2),s=n(4).f,i=n(39),o=n(199),u=n(12),l=n(200),c=n(29),h="".startsWith,p=Math.min,d=l("startsWith");a({target:"String",proto:!0,forced:!(!c&&!d&&(r=s(String.prototype,"startsWith"),r&&!r.writable)||d)},{startsWith:function(e){var t=String(u(this));o(e);var n=i(p(arguments.length>1?arguments[1]:void 0,t.length)),r=String(e);return h?h.call(t,r,n):t.slice(n,n+r.length)===r}})},function(e,t,n){var r=n(2),a=n(128).trim;r({target:"String",proto:!0,forced:n(220)("trim")},{trim:function(){return a(this)}})},function(e,t,n){var r=n(6),a=n(129);e.exports=function(e){return r((function(){return!!a[e]()||"​…᠎"!="​…᠎"[e]()||a[e].name!==e}))}},function(e,t,n){var r=n(2),a=n(128).end,s=n(220)("trimEnd"),i=s?function(){return a(this)}:"".trimEnd;r({target:"String",proto:!0,forced:s},{trimEnd:i,trimRight:i})},function(e,t,n){var r=n(2),a=n(128).start,s=n(220)("trimStart"),i=s?function(){return a(this)}:"".trimStart;r({target:"String",proto:!0,forced:s},{trimStart:i,trimLeft:i})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("anchor")},{anchor:function(e){return a(this,"a","name",e)}})},function(e,t,n){var r=n(12),a=/"/g;e.exports=function(e,t,n,s){var i=String(r(e)),o="<"+t;return""!==n&&(o+=" "+n+'="'+String(s).replace(a,"&quot;")+'"'),o+">"+i+"</"+t+">"}},function(e,t,n){var r=n(6);e.exports=function(e){return r((function(){var t=""[e]('"');return t!==t.toLowerCase()||t.split('"').length>3}))}},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("big")},{big:function(){return a(this,"big","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("blink")},{blink:function(){return a(this,"blink","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("bold")},{bold:function(){return a(this,"b","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("fixed")},{fixed:function(){return a(this,"tt","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("fontcolor")},{fontcolor:function(e){return a(this,"font","color",e)}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("fontsize")},{fontsize:function(e){return a(this,"font","size",e)}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("italics")},{italics:function(){return a(this,"i","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("link")},{link:function(e){return a(this,"a","href",e)}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("small")},{small:function(){return a(this,"small","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("strike")},{strike:function(){return a(this,"strike","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("sub")},{sub:function(){return a(this,"sub","","")}})},function(e,t,n){var r=n(2),a=n(224);r({target:"String",proto:!0,forced:n(225)("sup")},{sup:function(){return a(this,"sup","","")}})},function(e,t,n){var r,a=n(3),s=n(126),i=n(120),o=n(119),u=n(239),l=n(14),c=n(25).enforce,h=n(26),p=!a.ActiveXObject&&"ActiveXObject"in a,d=Object.isExtensible,f=function(e){return function(){return e(this,arguments.length?arguments[0]:void 0)}},m=e.exports=o("WeakMap",f,u);if(h&&p){r=u.getConstructor(f,"WeakMap",!0),i.REQUIRED=!0;var g=m.prototype,y=g.delete,b=g.has,v=g.get,x=g.set;s(g,{delete:function(e){if(l(e)&&!d(e)){var t=c(this);return t.frozen||(t.frozen=new r),y.call(this,e)||t.frozen.delete(e)}return y.call(this,e)},has:function(e){if(l(e)&&!d(e)){var t=c(this);return t.frozen||(t.frozen=new r),b.call(this,e)||t.frozen.has(e)}return b.call(this,e)},get:function(e){if(l(e)&&!d(e)){var t=c(this);return t.frozen||(t.frozen=new r),b.call(this,e)?v.call(this,e):t.frozen.get(e)}return v.call(this,e)},set:function(e,t){if(l(e)&&!d(e)){var n=c(this);n.frozen||(n.frozen=new r),b.call(this,e)?x.call(this,e,t):n.frozen.set(e,t)}else x.call(this,e,t);return this}})}},function(e,t,n){var r=n(126),a=n(120).getWeakData,s=n(20),i=n(14),o=n(123),u=n(122),l=n(63),c=n(15),h=n(25),p=h.set,d=h.getterFor,f=l.find,m=l.findIndex,g=0,y=function(e){return e.frozen||(e.frozen=new b)},b=function(){this.entries=[]},v=function(e,t){return f(e.entries,(function(e){return e[0]===t}))};b.prototype={get:function(e){var t=v(this,e);if(t)return t[1]},has:function(e){return!!v(this,e)},set:function(e,t){var n=v(this,e);n?n[1]=t:this.entries.push([e,t])},delete:function(e){var t=m(this.entries,(function(t){return t[0]===e}));return~t&&this.entries.splice(t,1),!!~t}},e.exports={getConstructor:function(e,t,n,l){var h=e((function(e,r){o(e,h,t),p(e,{type:t,id:g++,frozen:void 0}),null!=r&&u(r,e[l],e,n)})),f=d(t),m=function(e,t,n){var r=f(e),i=a(s(t),!0);return!0===i?y(r).set(t,n):i[r.id]=n,e};return r(h.prototype,{delete:function(e){var t=f(this);if(!i(e))return!1;var n=a(e);return!0===n?y(t).delete(e):n&&c(n,t.id)&&delete n[t.id]},has:function(e){var t=f(this);if(!i(e))return!1;var n=a(e);return!0===n?y(t).has(e):n&&c(n,t.id)}}),r(h.prototype,n?{get:function(e){var t=f(this);if(i(e)){var n=a(e);return!0===n?y(t).get(e):n?n[t.id]:void 0}},set:function(e,t){return m(this,e,t)}}:{add:function(e){return m(this,e,!0)}}),h}}},function(e,t,n){n(119)("WeakSet",(function(e){return function(){return e(this,arguments.length?arguments[0]:void 0)}}),n(239))},function(e,t,n){var r=n(3),a=n(242),s=n(77),i=n(18);for(var o in a){var u=r[o],l=u&&u.prototype;if(l&&l.forEach!==s)try{i(l,"forEach",s)}catch(e){l.forEach=s}}},function(e,t){e.exports={CSSRuleList:0,CSSStyleDeclaration:0,CSSValueList:0,ClientRectList:0,DOMRectList:0,DOMStringList:0,DOMTokenList:1,DataTransferItemList:0,FileList:0,HTMLAllCollection:0,HTMLCollection:0,HTMLFormElement:0,HTMLSelectElement:0,MediaList:0,MimeTypeArray:0,NamedNodeMap:0,NodeList:1,PaintRequestList:0,Plugin:0,PluginArray:0,SVGLengthList:0,SVGNumberList:0,SVGPathSegList:0,SVGPointList:0,SVGStringList:0,SVGTransformList:0,SourceBufferList:0,StyleSheetList:0,TextTrackCueList:0,TextTrackList:0,TouchList:0}},function(e,t,n){n(203);var r,a=n(2),s=n(5),i=n(244),o=n(3),u=n(59),l=n(21),c=n(123),h=n(15),p=n(147),d=n(79),f=n(197).codeAt,m=n(245),g=n(95),y=n(246),b=n(25),v=o.URL,x=y.URLSearchParams,w=y.getState,k=b.set,S=b.getterFor("URL"),I=Math.floor,A=Math.pow,E=/[A-Za-z]/,N=/[\d+-.A-Za-z]/,C=/\d/,T=/^(0x|0X)/,R=/^[0-7]+$/,_=/^\d+$/,F=/^[\dA-Fa-f]+$/,M=/[\u0000\u0009\u000A\u000D #%/:?@[\\]]/,D=/[\u0000\u0009\u000A\u000D #/:?@[\\]]/,O=/^[\u0000-\u001F ]+|[\u0000-\u001F ]+$/g,L=/[\u0009\u000A\u000D]/g,P=function(e,t){var n,r,a;if("["==t.charAt(0)){if("]"!=t.charAt(t.length-1))return"Invalid host";if(!(n=B(t.slice(1,-1))))return"Invalid host";e.host=n}else if(q(e)){if(t=m(t),M.test(t))return"Invalid host";if(null===(n=$(t)))return"Invalid host";e.host=n}else{if(D.test(t))return"Invalid host";for(n="",r=d(t),a=0;a<r.length;a++)n+=G(r[a],U);e.host=n}},$=function(e){var t,n,r,a,s,i,o,u=e.split(".");if(u.length&&""==u[u.length-1]&&u.pop(),(t=u.length)>4)return e;for(n=[],r=0;r<t;r++){if(""==(a=u[r]))return e;if(s=10,a.length>1&&"0"==a.charAt(0)&&(s=T.test(a)?16:8,a=a.slice(8==s?1:2)),""===a)i=0;else{if(!(10==s?_:8==s?R:F).test(a))return e;i=parseInt(a,s)}n.push(i)}for(r=0;r<t;r++)if(i=n[r],r==t-1){if(i>=A(256,5-t))return null}else if(i>255)return null;for(o=n.pop(),r=0;r<n.length;r++)o+=n[r]*A(256,3-r);return o},B=function(e){var t,n,r,a,s,i,o,u=[0,0,0,0,0,0,0,0],l=0,c=null,h=0,p=function(){return e.charAt(h)};if(":"==p()){if(":"!=e.charAt(1))return;h+=2,c=++l}for(;p();){if(8==l)return;if(":"!=p()){for(t=n=0;n<4&&F.test(p());)t=16*t+parseInt(p(),16),h++,n++;if("."==p()){if(0==n)return;if(h-=n,l>6)return;for(r=0;p();){if(a=null,r>0){if(!("."==p()&&r<4))return;h++}if(!C.test(p()))return;for(;C.test(p());){if(s=parseInt(p(),10),null===a)a=s;else{if(0==a)return;a=10*a+s}if(a>255)return;h++}u[l]=256*u[l]+a,2!=++r&&4!=r||l++}if(4!=r)return;break}if(":"==p()){if(h++,!p())return}else if(p())return;u[l++]=t}else{if(null!==c)return;h++,c=++l}}if(null!==c)for(i=l-c,l=7;0!=l&&i>0;)o=u[l],u[l--]=u[c+i-1],u[c+--i]=o;else if(8!=l)return;return u},z=function(e){var t,n,r,a;if("number"==typeof e){for(t=[],n=0;n<4;n++)t.unshift(e%256),e=I(e/256);return t.join(".")}if("object"==typeof e){for(t="",r=function(e){for(var t=null,n=1,r=null,a=0,s=0;s<8;s++)0!==e[s]?(a>n&&(t=r,n=a),r=null,a=0):(null===r&&(r=s),++a);return a>n&&(t=r,n=a),t}(e),n=0;n<8;n++)a&&0===e[n]||(a&&(a=!1),r===n?(t+=n?":":"::",a=!0):(t+=e[n].toString(16),n<7&&(t+=":")));return"["+t+"]"}return e},U={},W=p({},U,{" ":1,'"':1,"<":1,">":1,"`":1}),V=p({},W,{"#":1,"?":1,"{":1,"}":1}),j=p({},V,{"/":1,":":1,";":1,"=":1,"@":1,"[":1,"\\":1,"]":1,"^":1,"|":1}),G=function(e,t){var n=f(e,0);return n>32&&n<127&&!h(t,e)?e:encodeURIComponent(e)},H={ftp:21,file:null,http:80,https:443,ws:80,wss:443},q=function(e){return h(H,e.scheme)},K=function(e){return""!=e.username||""!=e.password},Y=function(e){return!e.host||e.cannotBeABaseURL||"file"==e.scheme},X=function(e,t){var n;return 2==e.length&&E.test(e.charAt(0))&&(":"==(n=e.charAt(1))||!t&&"|"==n)},Q=function(e){var t;return e.length>1&&X(e.slice(0,2))&&(2==e.length||"/"===(t=e.charAt(2))||"\\"===t||"?"===t||"#"===t)},Z=function(e){var t=e.path,n=t.length;!n||"file"==e.scheme&&1==n&&X(t[0],!0)||t.pop()},J=function(e){return"."===e||"%2e"===e.toLowerCase()},ee={},te={},ne={},re={},ae={},se={},ie={},oe={},ue={},le={},ce={},he={},pe={},de={},fe={},me={},ge={},ye={},be={},ve={},xe={},we=function(e,t,n,a){var s,i,o,u,l,c=n||ee,p=0,f="",m=!1,g=!1,y=!1;for(n||(e.scheme="",e.username="",e.password="",e.host=null,e.port=null,e.path=[],e.query=null,e.fragment=null,e.cannotBeABaseURL=!1,t=t.replace(O,"")),t=t.replace(L,""),s=d(t);p<=s.length;){switch(i=s[p],c){case ee:if(!i||!E.test(i)){if(n)return"Invalid scheme";c=ne;continue}f+=i.toLowerCase(),c=te;break;case te:if(i&&(N.test(i)||"+"==i||"-"==i||"."==i))f+=i.toLowerCase();else{if(":"!=i){if(n)return"Invalid scheme";f="",c=ne,p=0;continue}if(n&&(q(e)!=h(H,f)||"file"==f&&(K(e)||null!==e.port)||"file"==e.scheme&&!e.host))return;if(e.scheme=f,n)return void(q(e)&&H[e.scheme]==e.port&&(e.port=null));f="","file"==e.scheme?c=de:q(e)&&a&&a.scheme==e.scheme?c=re:q(e)?c=oe:"/"==s[p+1]?(c=ae,p++):(e.cannotBeABaseURL=!0,e.path.push(""),c=be)}break;case ne:if(!a||a.cannotBeABaseURL&&"#"!=i)return"Invalid scheme";if(a.cannotBeABaseURL&&"#"==i){e.scheme=a.scheme,e.path=a.path.slice(),e.query=a.query,e.fragment="",e.cannotBeABaseURL=!0,c=xe;break}c="file"==a.scheme?de:se;continue;case re:if("/"!=i||"/"!=s[p+1]){c=se;continue}c=ue,p++;break;case ae:if("/"==i){c=le;break}c=ye;continue;case se:if(e.scheme=a.scheme,i==r)e.username=a.username,e.password=a.password,e.host=a.host,e.port=a.port,e.path=a.path.slice(),e.query=a.query;else if("/"==i||"\\"==i&&q(e))c=ie;else if("?"==i)e.username=a.username,e.password=a.password,e.host=a.host,e.port=a.port,e.path=a.path.slice(),e.query="",c=ve;else{if("#"!=i){e.username=a.username,e.password=a.password,e.host=a.host,e.port=a.port,e.path=a.path.slice(),e.path.pop(),c=ye;continue}e.username=a.username,e.password=a.password,e.host=a.host,e.port=a.port,e.path=a.path.slice(),e.query=a.query,e.fragment="",c=xe}break;case ie:if(!q(e)||"/"!=i&&"\\"!=i){if("/"!=i){e.username=a.username,e.password=a.password,e.host=a.host,e.port=a.port,c=ye;continue}c=le}else c=ue;break;case oe:if(c=ue,"/"!=i||"/"!=f.charAt(p+1))continue;p++;break;case ue:if("/"!=i&&"\\"!=i){c=le;continue}break;case le:if("@"==i){m&&(f="%40"+f),m=!0,o=d(f);for(var b=0;b<o.length;b++){var v=o[b];if(":"!=v||y){var x=G(v,j);y?e.password+=x:e.username+=x}else y=!0}f=""}else if(i==r||"/"==i||"?"==i||"#"==i||"\\"==i&&q(e)){if(m&&""==f)return"Invalid authority";p-=d(f).length+1,f="",c=ce}else f+=i;break;case ce:case he:if(n&&"file"==e.scheme){c=me;continue}if(":"!=i||g){if(i==r||"/"==i||"?"==i||"#"==i||"\\"==i&&q(e)){if(q(e)&&""==f)return"Invalid host";if(n&&""==f&&(K(e)||null!==e.port))return;if(u=P(e,f))return u;if(f="",c=ge,n)return;continue}"["==i?g=!0:"]"==i&&(g=!1),f+=i}else{if(""==f)return"Invalid host";if(u=P(e,f))return u;if(f="",c=pe,n==he)return}break;case pe:if(!C.test(i)){if(i==r||"/"==i||"?"==i||"#"==i||"\\"==i&&q(e)||n){if(""!=f){var w=parseInt(f,10);if(w>65535)return"Invalid port";e.port=q(e)&&w===H[e.scheme]?null:w,f=""}if(n)return;c=ge;continue}return"Invalid port"}f+=i;break;case de:if(e.scheme="file","/"==i||"\\"==i)c=fe;else{if(!a||"file"!=a.scheme){c=ye;continue}if(i==r)e.host=a.host,e.path=a.path.slice(),e.query=a.query;else if("?"==i)e.host=a.host,e.path=a.path.slice(),e.query="",c=ve;else{if("#"!=i){Q(s.slice(p).join(""))||(e.host=a.host,e.path=a.path.slice(),Z(e)),c=ye;continue}e.host=a.host,e.path=a.path.slice(),e.query=a.query,e.fragment="",c=xe}}break;case fe:if("/"==i||"\\"==i){c=me;break}a&&"file"==a.scheme&&!Q(s.slice(p).join(""))&&(X(a.path[0],!0)?e.path.push(a.path[0]):e.host=a.host),c=ye;continue;case me:if(i==r||"/"==i||"\\"==i||"?"==i||"#"==i){if(!n&&X(f))c=ye;else if(""==f){if(e.host="",n)return;c=ge}else{if(u=P(e,f))return u;if("localhost"==e.host&&(e.host=""),n)return;f="",c=ge}continue}f+=i;break;case ge:if(q(e)){if(c=ye,"/"!=i&&"\\"!=i)continue}else if(n||"?"!=i)if(n||"#"!=i){if(i!=r&&(c=ye,"/"!=i))continue}else e.fragment="",c=xe;else e.query="",c=ve;break;case ye:if(i==r||"/"==i||"\\"==i&&q(e)||!n&&("?"==i||"#"==i)){if(".."===(l=(l=f).toLowerCase())||"%2e."===l||".%2e"===l||"%2e%2e"===l?(Z(e),"/"==i||"\\"==i&&q(e)||e.path.push("")):J(f)?"/"==i||"\\"==i&&q(e)||e.path.push(""):("file"==e.scheme&&!e.path.length&&X(f)&&(e.host&&(e.host=""),f=f.charAt(0)+":"),e.path.push(f)),f="","file"==e.scheme&&(i==r||"?"==i||"#"==i))for(;e.path.length>1&&""===e.path[0];)e.path.shift();"?"==i?(e.query="",c=ve):"#"==i&&(e.fragment="",c=xe)}else f+=G(i,V);break;case be:"?"==i?(e.query="",c=ve):"#"==i?(e.fragment="",c=xe):i!=r&&(e.path[0]+=G(i,U));break;case ve:n||"#"!=i?i!=r&&("'"==i&&q(e)?e.query+="%27":e.query+="#"==i?"%23":G(i,U)):(e.fragment="",c=xe);break;case xe:i!=r&&(e.fragment+=G(i,W))}p++}},ke=function(e){var t,n,r=c(this,ke,"URL"),a=arguments.length>1?arguments[1]:void 0,i=String(e),o=k(r,{type:"URL"});if(void 0!==a)if(a instanceof ke)t=S(a);else if(n=we(t={},String(a)))throw TypeError(n);if(n=we(o,i,null,t))throw TypeError(n);var u=o.searchParams=new x,l=w(u);l.updateSearchParams(o.query),l.updateURL=function(){o.query=String(u)||null},s||(r.href=Ie.call(r),r.origin=Ae.call(r),r.protocol=Ee.call(r),r.username=Ne.call(r),r.password=Ce.call(r),r.host=Te.call(r),r.hostname=Re.call(r),r.port=_e.call(r),r.pathname=Fe.call(r),r.search=Me.call(r),r.searchParams=De.call(r),r.hash=Oe.call(r))},Se=ke.prototype,Ie=function(){var e=S(this),t=e.scheme,n=e.username,r=e.password,a=e.host,s=e.port,i=e.path,o=e.query,u=e.fragment,l=t+":";return null!==a?(l+="//",K(e)&&(l+=n+(r?":"+r:"")+"@"),l+=z(a),null!==s&&(l+=":"+s)):"file"==t&&(l+="//"),l+=e.cannotBeABaseURL?i[0]:i.length?"/"+i.join("/"):"",null!==o&&(l+="?"+o),null!==u&&(l+="#"+u),l},Ae=function(){var e=S(this),t=e.scheme,n=e.port;if("blob"==t)try{return new URL(t.path[0]).origin}catch(e){return"null"}return"file"!=t&&q(e)?t+"://"+z(e.host)+(null!==n?":"+n:""):"null"},Ee=function(){return S(this).scheme+":"},Ne=function(){return S(this).username},Ce=function(){return S(this).password},Te=function(){var e=S(this),t=e.host,n=e.port;return null===t?"":null===n?z(t):z(t)+":"+n},Re=function(){var e=S(this).host;return null===e?"":z(e)},_e=function(){var e=S(this).port;return null===e?"":String(e)},Fe=function(){var e=S(this),t=e.path;return e.cannotBeABaseURL?t[0]:t.length?"/"+t.join("/"):""},Me=function(){var e=S(this).query;return e?"?"+e:""},De=function(){return S(this).searchParams},Oe=function(){var e=S(this).fragment;return e?"#"+e:""},Le=function(e,t){return{get:e,set:t,configurable:!0,enumerable:!0}};if(s&&u(Se,{href:Le(Ie,(function(e){var t=S(this),n=String(e),r=we(t,n);if(r)throw TypeError(r);w(t.searchParams).updateSearchParams(t.query)})),origin:Le(Ae),protocol:Le(Ee,(function(e){var t=S(this);we(t,String(e)+":",ee)})),username:Le(Ne,(function(e){var t=S(this),n=d(String(e));if(!Y(t)){t.username="";for(var r=0;r<n.length;r++)t.username+=G(n[r],j)}})),password:Le(Ce,(function(e){var t=S(this),n=d(String(e));if(!Y(t)){t.password="";for(var r=0;r<n.length;r++)t.password+=G(n[r],j)}})),host:Le(Te,(function(e){var t=S(this);t.cannotBeABaseURL||we(t,String(e),ce)})),hostname:Le(Re,(function(e){var t=S(this);t.cannotBeABaseURL||we(t,String(e),he)})),port:Le(_e,(function(e){var t=S(this);Y(t)||(""==(e=String(e))?t.port=null:we(t,e,pe))})),pathname:Le(Fe,(function(e){var t=S(this);t.cannotBeABaseURL||(t.path=[],we(t,e+"",ge))})),search:Le(Me,(function(e){var t=S(this);""==(e=String(e))?t.query=null:("?"==e.charAt(0)&&(e=e.slice(1)),t.query="",we(t,e,ve)),w(t.searchParams).updateSearchParams(t.query)})),searchParams:Le(De),hash:Le(Oe,(function(e){var t=S(this);""!=(e=String(e))?("#"==e.charAt(0)&&(e=e.slice(1)),t.fragment="",we(t,e,xe)):t.fragment=null}))}),l(Se,"toJSON",(function(){return Ie.call(this)}),{enumerable:!0}),l(Se,"toString",(function(){return Ie.call(this)}),{enumerable:!0}),v){var Pe=v.createObjectURL,$e=v.revokeObjectURL;Pe&&l(ke,"createObjectURL",(function(e){return Pe.apply(v,arguments)})),$e&&l(ke,"revokeObjectURL",(function(e){return $e.apply(v,arguments)}))}g(ke,"URL"),a({global:!0,forced:!i,sham:!s},{URL:ke})},function(e,t,n){var r=n(6),a=n(49),s=n(29),i=a("iterator");e.exports=!r((function(){var e=new URL("b?a=1&b=2&c=3","http://a"),t=e.searchParams,n="";return e.pathname="c%20d",t.forEach((function(e,r){t.delete("b"),n+=r+e})),s&&!e.toJSON||!t.sort||"http://a/c%20d?a=1&c=3"!==e.href||"3"!==t.get("c")||"a=1"!==String(new URLSearchParams("?a=1"))||!t[i]||"a"!==new URL("https://a@b").username||"b"!==new URLSearchParams(new URLSearchParams("a=b")).get("a")||"xn--e1aybc"!==new URL("http://тест").host||"#%D0%B1"!==new URL("http://a#б").hash||"a1c3"!==n||"x"!==new URL("http://x",void 0).host}))},function(e,t,n){var r=/[^\0-\u007E]/,a=/[.\u3002\uFF0E\uFF61]/g,s="Overflow: input needs wider integers to process",i=Math.floor,o=String.fromCharCode,u=function(e){return e+22+75*(e<26)},l=function(e,t,n){var r=0;for(e=n?i(e/700):e>>1,e+=i(e/t);e>455;r+=36)e=i(e/35);return i(r+36*e/(e+38))},c=function(e){var t,n,r=[],a=(e=function(e){for(var t=[],n=0,r=e.length;n<r;){var a=e.charCodeAt(n++);if(a>=55296&&a<=56319&&n<r){var s=e.charCodeAt(n++);56320==(64512&s)?t.push(((1023&a)<<10)+(1023&s)+65536):(t.push(a),n--)}else t.push(a)}return t}(e)).length,c=128,h=0,p=72;for(t=0;t<e.length;t++)(n=e[t])<128&&r.push(o(n));var d=r.length,f=d;for(d&&r.push("-");f<a;){var m=2147483647;for(t=0;t<e.length;t++)(n=e[t])>=c&&n<m&&(m=n);var g=f+1;if(m-c>i((2147483647-h)/g))throw RangeError(s);for(h+=(m-c)*g,c=m,t=0;t<e.length;t++){if((n=e[t])<c&&++h>2147483647)throw RangeError(s);if(n==c){for(var y=h,b=36;;b+=36){var v=b<=p?1:b>=p+26?26:b-p;if(y<v)break;var x=y-v,w=36-v;r.push(o(u(v+x%w))),y=i(x/w)}r.push(o(u(y))),p=l(h,g,f==d),h=0,++f}}++h,++c}return r.join("")};e.exports=function(e){var t,n,s=[],i=e.toLowerCase().replace(a,".").split(".");for(t=0;t<i.length;t++)n=i[t],s.push(r.test(n)?"xn--"+c(n):n);return s.join(".")}},function(e,t,n){n(89);var r=n(2),a=n(34),s=n(244),i=n(21),o=n(126),u=n(95),l=n(91),c=n(25),h=n(123),p=n(15),d=n(64),f=n(84),m=n(20),g=n(14),y=n(58),b=n(8),v=n(247),x=n(83),w=n(49),k=a("fetch"),S=a("Headers"),I=w("iterator"),A=c.set,E=c.getterFor("URLSearchParams"),N=c.getterFor("URLSearchParamsIterator"),C=/\+/g,T=Array(4),R=function(e){return T[e-1]||(T[e-1]=RegExp("((?:%[\\da-f]{2}){"+e+"})","gi"))},_=function(e){try{return decodeURIComponent(e)}catch(t){return e}},F=function(e){var t=e.replace(C," "),n=4;try{return decodeURIComponent(t)}catch(e){for(;n;)t=t.replace(R(n--),_);return t}},M=/[!'()~]|%20/g,D={"!":"%21","'":"%27","(":"%28",")":"%29","~":"%7E","%20":"+"},O=function(e){return D[e]},L=function(e){return encodeURIComponent(e).replace(M,O)},P=function(e,t){if(t)for(var n,r,a=t.split("&"),s=0;s<a.length;)(n=a[s++]).length&&(r=n.split("="),e.push({key:F(r.shift()),value:F(r.join("="))}))},$=function(e){this.entries.length=0,P(this.entries,e)},B=function(e,t){if(e<t)throw TypeError("Not enough arguments")},z=l((function(e,t){A(this,{type:"URLSearchParamsIterator",iterator:v(E(e).entries),kind:t})}),"Iterator",(function(){var e=N(this),t=e.kind,n=e.iterator.next(),r=n.value;return n.done||(n.value="keys"===t?r.key:"values"===t?r.value:[r.key,r.value]),n})),U=function(){h(this,U,"URLSearchParams");var e,t,n,r,a,s,i,o,u,l=arguments.length>0?arguments[0]:void 0,c=this,d=[];if(A(c,{type:"URLSearchParams",entries:d,updateURL:function(){},updateSearchParams:$}),void 0!==l)if(g(l))if("function"==typeof(e=x(l)))for(n=(t=e.call(l)).next;!(r=n.call(t)).done;){if((i=(s=(a=v(m(r.value))).next).call(a)).done||(o=s.call(a)).done||!s.call(a).done)throw TypeError("Expected sequence with length 2");d.push({key:i.value+"",value:o.value+""})}else for(u in l)p(l,u)&&d.push({key:u,value:l[u]+""});else P(d,"string"==typeof l?"?"===l.charAt(0)?l.slice(1):l:l+"")},W=U.prototype;o(W,{append:function(e,t){B(arguments.length,2);var n=E(this);n.entries.push({key:e+"",value:t+""}),n.updateURL()},delete:function(e){B(arguments.length,1);for(var t=E(this),n=t.entries,r=e+"",a=0;a<n.length;)n[a].key===r?n.splice(a,1):a++;t.updateURL()},get:function(e){B(arguments.length,1);for(var t=E(this).entries,n=e+"",r=0;r<t.length;r++)if(t[r].key===n)return t[r].value;return null},getAll:function(e){B(arguments.length,1);for(var t=E(this).entries,n=e+"",r=[],a=0;a<t.length;a++)t[a].key===n&&r.push(t[a].value);return r},has:function(e){B(arguments.length,1);for(var t=E(this).entries,n=e+"",r=0;r<t.length;)if(t[r++].key===n)return!0;return!1},set:function(e,t){B(arguments.length,1);for(var n,r=E(this),a=r.entries,s=!1,i=e+"",o=t+"",u=0;u<a.length;u++)(n=a[u]).key===i&&(s?a.splice(u--,1):(s=!0,n.value=o));s||a.push({key:i,value:o}),r.updateURL()},sort:function(){var e,t,n,r=E(this),a=r.entries,s=a.slice();for(a.length=0,n=0;n<s.length;n++){for(e=s[n],t=0;t<n;t++)if(a[t].key>e.key){a.splice(t,0,e);break}t===n&&a.push(e)}r.updateURL()},forEach:function(e){for(var t,n=E(this).entries,r=d(e,arguments.length>1?arguments[1]:void 0,3),a=0;a<n.length;)r((t=n[a++]).value,t.key,this)},keys:function(){return new z(this,"keys")},values:function(){return new z(this,"values")},entries:function(){return new z(this,"entries")}},{enumerable:!0}),i(W,I,W.entries),i(W,"toString",(function(){for(var e,t=E(this).entries,n=[],r=0;r<t.length;)e=t[r++],n.push(L(e.key)+"="+L(e.value));return n.join("&")}),{enumerable:!0}),u(U,"URLSearchParams"),r({global:!0,forced:!s},{URLSearchParams:U}),s||"function"!=typeof k||"function"!=typeof S||r({global:!0,enumerable:!0,forced:!0},{fetch:function(e){var t,n,r,a=[e];return arguments.length>1&&(g(t=arguments[1])&&(n=t.body,"URLSearchParams"===f(n)&&((r=t.headers?new S(t.headers):new S).has("content-type")||r.set("content-type","application/x-www-form-urlencoded;charset=UTF-8"),t=y(t,{body:b(0,String(n)),headers:b(0,r)}))),a.push(t)),k.apply(this,a)}}),e.exports={URLSearchParams:U,getState:E}},function(e,t,n){var r=n(20),a=n(83);e.exports=function(e){var t=a(e);if("function"!=typeof t)throw TypeError(String(e)+" is not iterable");return r(t.call(e))}},function(e,t,n){n(2)({target:"URL",proto:!0,enumerable:!0},{toJSON:function(){return URL.prototype.toString.call(this)}})}])}(),function(e){"use strict";var t="URLSearchParams"in self,n="Symbol"in self&&"iterator"in Symbol,r="FileReader"in self&&"Blob"in self&&function(){try{return new Blob,!0}catch(e){return!1}}(),a="FormData"in self,s="ArrayBuffer"in self;if(s)var i=["[object Int8Array]","[object Uint8Array]","[object Uint8ClampedArray]","[object Int16Array]","[object Uint16Array]","[object Int32Array]","[object Uint32Array]","[object Float32Array]","[object Float64Array]"],o=ArrayBuffer.isView||function(e){return e&&i.indexOf(Object.prototype.toString.call(e))>-1};function u(e){if("string"!=typeof e&&(e=String(e)),/[^a-z0-9\-#$%&'*+.^_`|~]/i.test(e))throw new TypeError("Invalid character in header field name");return e.toLowerCase()}function l(e){return"string"!=typeof e&&(e=String(e)),e}function c(e){var t={next:function(){var t=e.shift();return{done:void 0===t,value:t}}};return n&&(t[Symbol.iterator]=function(){return t}),t}function h(e){this.map={},e instanceof h?e.forEach((function(e,t){this.append(t,e)}),this):Array.isArray(e)?e.forEach((function(e){this.append(e[0],e[1])}),this):e&&Object.getOwnPropertyNames(e).forEach((function(t){this.append(t,e[t])}),this)}function p(e){if(e.bodyUsed)return Promise.reject(new TypeError("Already read"));e.bodyUsed=!0}function d(e){return new Promise((function(t,n){e.onload=function(){t(e.result)},e.onerror=function(){n(e.error)}}))}function f(e){var t=new FileReader,n=d(t);return t.readAsArrayBuffer(e),n}function m(e){if(e.slice)return e.slice(0);var t=new Uint8Array(e.byteLength);return t.set(new Uint8Array(e)),t.buffer}function g(){return this.bodyUsed=!1,this._initBody=function(e){var n;this._bodyInit=e,e?"string"==typeof e?this._bodyText=e:r&&Blob.prototype.isPrototypeOf(e)?this._bodyBlob=e:a&&FormData.prototype.isPrototypeOf(e)?this._bodyFormData=e:t&&URLSearchParams.prototype.isPrototypeOf(e)?this._bodyText=e.toString():s&&r&&(n=e)&&DataView.prototype.isPrototypeOf(n)?(this._bodyArrayBuffer=m(e.buffer),this._bodyInit=new Blob([this._bodyArrayBuffer])):s&&(ArrayBuffer.prototype.isPrototypeOf(e)||o(e))?this._bodyArrayBuffer=m(e):this._bodyText=e=Object.prototype.toString.call(e):this._bodyText="",this.headers.get("content-type")||("string"==typeof e?this.headers.set("content-type","text/plain;charset=UTF-8"):this._bodyBlob&&this._bodyBlob.type?this.headers.set("content-type",this._bodyBlob.type):t&&URLSearchParams.prototype.isPrototypeOf(e)&&this.headers.set("content-type","application/x-www-form-urlencoded;charset=UTF-8"))},r&&(this.blob=function(){var e=p(this);if(e)return e;if(this._bodyBlob)return Promise.resolve(this._bodyBlob);if(this._bodyArrayBuffer)return Promise.resolve(new Blob([this._bodyArrayBuffer]));if(this._bodyFormData)throw new Error("could not read FormData body as blob");return Promise.resolve(new Blob([this._bodyText]))},this.arrayBuffer=function(){return this._bodyArrayBuffer?p(this)||Promise.resolve(this._bodyArrayBuffer):this.blob().then(f)}),this.text=function(){var e,t,n,r=p(this);if(r)return r;if(this._bodyBlob)return e=this._bodyBlob,n=d(t=new FileReader),t.readAsText(e),n;if(this._bodyArrayBuffer)return Promise.resolve(function(e){for(var t=new Uint8Array(e),n=new Array(t.length),r=0;r<t.length;r++)n[r]=String.fromCharCode(t[r]);return n.join("")}(this._bodyArrayBuffer));if(this._bodyFormData)throw new Error("could not read FormData body as text");return Promise.resolve(this._bodyText)},a&&(this.formData=function(){return this.text().then(v)}),this.json=function(){return this.text().then(JSON.parse)},this}h.prototype.append=function(e,t){e=u(e),t=l(t);var n=this.map[e];this.map[e]=n?n+", "+t:t},h.prototype.delete=function(e){delete this.map[u(e)]},h.prototype.get=function(e){return e=u(e),this.has(e)?this.map[e]:null},h.prototype.has=function(e){return this.map.hasOwnProperty(u(e))},h.prototype.set=function(e,t){this.map[u(e)]=l(t)},h.prototype.forEach=function(e,t){for(var n in this.map)this.map.hasOwnProperty(n)&&e.call(t,this.map[n],n,this)},h.prototype.keys=function(){var e=[];return this.forEach((function(t,n){e.push(n)})),c(e)},h.prototype.values=function(){var e=[];return this.forEach((function(t){e.push(t)})),c(e)},h.prototype.entries=function(){var e=[];return this.forEach((function(t,n){e.push([n,t])})),c(e)},n&&(h.prototype[Symbol.iterator]=h.prototype.entries);var y=["DELETE","GET","HEAD","OPTIONS","POST","PUT"];function b(e,t){var n,r,a=(t=t||{}).body;if(e instanceof b){if(e.bodyUsed)throw new TypeError("Already read");this.url=e.url,this.credentials=e.credentials,t.headers||(this.headers=new h(e.headers)),this.method=e.method,this.mode=e.mode,this.signal=e.signal,a||null==e._bodyInit||(a=e._bodyInit,e.bodyUsed=!0)}else this.url=String(e);if(this.credentials=t.credentials||this.credentials||"same-origin",!t.headers&&this.headers||(this.headers=new h(t.headers)),this.method=(r=(n=t.method||this.method||"GET").toUpperCase(),y.indexOf(r)>-1?r:n),this.mode=t.mode||this.mode||null,this.signal=t.signal||this.signal,this.referrer=null,("GET"===this.method||"HEAD"===this.method)&&a)throw new TypeError("Body not allowed for GET or HEAD requests");this._initBody(a)}function v(e){var t=new FormData;return e.trim().split("&").forEach((function(e){if(e){var n=e.split("="),r=n.shift().replace(/\+/g," "),a=n.join("=").replace(/\+/g," ");t.append(decodeURIComponent(r),decodeURIComponent(a))}})),t}function x(e,t){t||(t={}),this.type="default",this.status=void 0===t.status?200:t.status,this.ok=this.status>=200&&this.status<300,this.statusText="statusText"in t?t.statusText:"OK",this.headers=new h(t.headers),this.url=t.url||"",this._initBody(e)}b.prototype.clone=function(){return new b(this,{body:this._bodyInit})},g.call(b.prototype),g.call(x.prototype),x.prototype.clone=function(){return new x(this._bodyInit,{status:this.status,statusText:this.statusText,headers:new h(this.headers),url:this.url})},x.error=function(){var e=new x(null,{status:0,statusText:""});return e.type="error",e};var w=[301,302,303,307,308];x.redirect=function(e,t){if(-1===w.indexOf(t))throw new RangeError("Invalid status code");return new x(null,{status:t,headers:{location:e}})},e.DOMException=self.DOMException;try{new e.DOMException}catch(t){e.DOMException=function(e,t){this.message=e,this.name=t;var n=Error(e);this.stack=n.stack},e.DOMException.prototype=Object.create(Error.prototype),e.DOMException.prototype.constructor=e.DOMException}function k(t,n){return new Promise((function(a,s){var i=new b(t,n);if(i.signal&&i.signal.aborted)return s(new e.DOMException("Aborted","AbortError"));var o=new XMLHttpRequest;function u(){o.abort()}o.onload=function(){var e,t,n={status:o.status,statusText:o.statusText,headers:(e=o.getAllResponseHeaders()||"",t=new h,e.replace(/\r?\n[\t ]+/g," ").split(/\r?\n/).forEach((function(e){var n=e.split(":"),r=n.shift().trim();if(r){var a=n.join(":").trim();t.append(r,a)}})),t)};n.url="responseURL"in o?o.responseURL:n.headers.get("X-Request-URL");var r="response"in o?o.response:o.responseText;a(new x(r,n))},o.onerror=function(){s(new TypeError("Network request failed"))},o.ontimeout=function(){s(new TypeError("Network request failed"))},o.onabort=function(){s(new e.DOMException("Aborted","AbortError"))},o.open(i.method,i.url,!0),"include"===i.credentials?o.withCredentials=!0:"omit"===i.credentials&&(o.withCredentials=!1),"responseType"in o&&r&&(o.responseType="blob"),i.headers.forEach((function(e,t){o.setRequestHeader(t,e)})),i.signal&&(i.signal.addEventListener("abort",u),o.onreadystatechange=function(){4===o.readyState&&i.signal.removeEventListener("abort",u)}),o.send(void 0===i._bodyInit?null:i._bodyInit)}))}k.polyfill=!0,self.fetch||(self.fetch=k,self.Headers=h,self.Request=b,self.Response=x),e.Headers=h,e.Request=b,e.Response=x,e.fetch=k}({})}).call(this,n(33))}]);
\ No newline at end of file
diff --git a/public/acuant/11.9.3.508.1/AcuantCamera.js b/public/acuant/11.9.3.508.1/AcuantCamera.js
new file mode 100644
index 00000000000..b5fbcd0f695
--- /dev/null
+++ b/public/acuant/11.9.3.508.1/AcuantCamera.js
@@ -0,0 +1,1678 @@
+// eslint-disable-next-line no-unused-vars, no-var
+var AcuantCameraUI = (function () {
+    'use strict';
+    let player = null;
+  
+    let uiCanvas = null;
+    let uiContext = null;
+    let acuantCamera = null;
+  
+    let svc = {
+      start: start,
+      end: end,
+    };
+  
+    let isStarted = false;
+  
+    let onDetectedResult = null;
+    // let previousState = null;
+    let counter = null;
+    let uiStateTextElement = null;
+    let errorCallback = null;
+  
+    let timeout = null;
+    let uiTimeout = null;
+  
+    const UI_STATE = {
+      CAPTURING: -1,
+      TAP_TO_CAPTURE: -2
+    };
+  
+    let userOptions = {
+      text: {
+        NONE: 'ALIGN',
+        SMALL_DOCUMENT: 'MOVE CLOSER',
+        GOOD_DOCUMENT: null,
+        BIG_DOCUMENT: 'TOO CLOSE',
+        CAPTURING: 'CAPTURING',
+        TAP_TO_CAPTURE: 'TAP TO CAPTURE'
+      }
+    };
+  
+    const TRIGGER_TIME = 3000;
+    const DETECT_TIME_THRESHOLD = 400;//slightly increased from 300, 
+    let minTime = Number.MAX_VALUE;
+  
+    function reset() {
+      onDetectedResult = null;
+      counter = null;
+      uiStateTextElement = null;
+    }
+  
+    function end() {
+      reset();
+      isStarted = false;
+      if (timeout) {
+        clearTimeout(timeout);
+        timeout = null;
+      }
+      if (uiTimeout) {
+        clearTimeout(uiTimeout);
+        uiTimeout = null;
+      }
+      
+      if(player) {
+        player.removeEventListener('play', play, 0);
+        player = null;
+      }
+  
+      if (acuantCamera) {
+        acuantCamera.removeEventListener('acuantcameracreated', onAcuantCameraCreated);
+      }
+      if (uiCanvas) {
+        uiCanvas.removeEventListener('click', onTap);
+        uiCanvas.getContext('2d').clearRect(0, 0, uiCanvas.width, uiCanvas.height);
+      }
+      AcuantCamera.end();
+    }
+  
+    function start(captureCb, options) {
+      errorCallback = captureCb.onError;
+      if (options) {
+        userOptions = options;
+        // eslint-disable-next-line no-prototype-builtins
+        if (!userOptions.text.hasOwnProperty('BIG_DOCUMENT')) {
+          userOptions.text.BIG_DOCUMENT = 'TOO CLOSE';
+        }
+      }
+      if (AcuantCamera.isCameraSupported) {
+        if (!isStarted) {
+          isStarted = true;
+          reset();
+          startCamera(captureCb);
+        } 
+      } else {
+        onError('Camera not supported.', AcuantJavascriptWebSdk.START_FAIL_CODE);
+      }
+    }
+  
+    function handleTime(start, count){
+      if (count >= 3) {
+        return true;
+      } else {
+        let elapsed = new Date().getTime() - start;
+  
+        if (elapsed < minTime) {
+          minTime = elapsed;
+        }
+        return minTime < DETECT_TIME_THRESHOLD;
+      }
+    }
+  
+    function handleLiveCapture(response, captureCb){
+      if (response) {
+        if (onDetectedResult && onDetectedResult.state === -1) {
+          return;
+        }
+  
+        if(captureCb.onFrameAvailable){
+          captureCb.onFrameAvailable(response);
+        }
+        onDetectedResult = response;
+  
+        if (onDetectedResult.state === AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT) {
+          if (timeout === null) {
+            counter = new Date().getTime();
+            triggerCountDown(captureCb);
+          }
+        } else {
+          cancelCountDown();
+          counter = null;
+        }
+      } else {
+        cancelCountDown();
+        counter = null;
+      }
+    }
+  
+    function triggerCountDown(captureCb) {
+      if (timeout === null) {
+        timeout = setTimeout(function () {
+          onDetectedResult.state = UI_STATE.CAPTURING;
+          capture(captureCb, 'AUTO');
+        }, TRIGGER_TIME);
+      }
+    }
+  
+    function cancelCountDown() {
+      if (timeout) {
+        clearTimeout(timeout);
+        timeout = null;
+      }
+    }
+  
+    function startTapToCapture(captureCb){
+      onDetectedResult = {
+        state: UI_STATE.TAP_TO_CAPTURE
+      };
+      uiCanvas.addEventListener('click', onTap, false);
+      uiCanvas.callback = captureCb;
+    }
+  
+    function onTap(event){
+      capture(event.currentTarget.callback, 'TAP');
+    }
+  
+    function onAcuantCameraCreated() {
+      player = document.getElementById('acuant-player');
+      uiCanvas = document.getElementById('acuant-ui-canvas');
+      uiContext = uiCanvas.getContext('2d');
+  
+      uiCanvas.setAttribute('role', 'img');
+      player.addEventListener('play', play, 0);
+    }
+  
+    function startCamera(captureCb) {
+      let count = 0;
+      let startTime = new Date().getTime();
+  
+      acuantCamera = document.getElementById('acuant-camera');
+      if (acuantCamera) {
+        acuantCamera.addEventListener('acuantcameracreated', onAcuantCameraCreated);
+      }
+  
+      AcuantCamera.start((response) => {
+        if (handleTime(startTime, count)) {
+          if (minTime < DETECT_TIME_THRESHOLD) {
+            handleLiveCapture(response, captureCb);
+          } else {
+            startTapToCapture(captureCb);
+          }
+        } else {
+          count += 1;
+          startTime = new Date().getTime();
+          AcuantCamera.setRepeatFrameProcessor();
+        }
+      }, captureCb, onError);
+    }
+  
+    function capture(captureCb, capType) {
+      AcuantCamera.triggerCapture((response) => {
+        end();
+        captureCb.onCaptured(response);
+  
+        AcuantCamera.evaluateImage(response.data, response.width, response.height, response.isPortraitOrientation, capType, (result) => {
+          captureCb.onCropped(result);
+        });
+      });
+    }
+  
+    function onError(error, code) {
+      end();
+      if (errorCallback) {
+        errorCallback(error, code);
+      }
+      errorCallback = null;
+    }
+  
+    //TODO We can theoretically get rid of this and only draw when a new detect result comes back
+    // in practice this slows down both detect and the ui, needs further investigation.
+    function play() {
+      (function loopUi() {
+        if (player && !player.paused && !player.ended && isStarted) {
+          handleUi();
+          
+          //Drawing at 10 fps at most (mostly done for performance on lower end devices).
+          // Dont use set interval as it seems to queue calls potentially freezing page.
+          uiTimeout = setTimeout(loopUi, 100);
+        }
+      })();
+    }
+  
+    function drawText(text, fontWeight = 0.04, color = '#ffffff', showBackRect = true) {
+      let dimension = getDimension();
+      let currentOrientation = window.orientation;
+      let measured = uiContext.measureText(text);
+  
+      let offsetY = (Math.max(dimension.width, dimension.height) * 0.01);
+      let offsetX = (Math.max(dimension.width, dimension.height) * 0.02);
+  
+      let x = (dimension.height - offsetX - measured.width) / 2;
+      let y = -((dimension.width / 2) - offsetY);
+      let rotation = 90;
+  
+      if (currentOrientation !== 0) {
+        rotation = 0;
+        x = (dimension.width - offsetY - measured.width) / 2;
+        y = (dimension.height / 2) - offsetX + (Math.max(dimension.width, dimension.height) * 0.04);
+      }
+  
+      uiContext.rotate(rotation * Math.PI / 180);
+  
+      if (showBackRect) {
+        uiContext.fillStyle = 'rgba(0, 0, 0, 0.5)';
+        uiContext.fillRect(
+          Math.round(x - offsetY),
+          Math.round(y + offsetY),
+          Math.round(measured.width + offsetX),
+          -Math.round((Math.max(dimension.width, dimension.height) * 0.05))
+        );
+      }
+  
+      uiContext.font = (Math.ceil(Math.max(dimension.width, dimension.height) * fontWeight) || 0) + 'px Sans-serif';
+      uiContext.fillStyle = color;
+      uiContext.fillText(text, x, y);
+      updateUiStateTextElement(text);
+  
+      // undo rotation to keep the ui in the same position
+      uiContext.rotate(-(rotation * Math.PI / 180));
+    }
+  
+    // For screen readers
+    const updateUiStateTextElement = (text) => {
+      if (!uiStateTextElement) {
+        uiStateTextElement = document.createElement('p');
+        uiStateTextElement.id = 'doc-state-text';
+        uiStateTextElement.style.height = '1px';
+        uiStateTextElement.style.width = '1px';
+        uiStateTextElement.style.margin = '-1px';
+        uiStateTextElement.style.overflow = 'hidden';
+        uiStateTextElement.style.position = 'absolute';
+        uiStateTextElement.style.whiteSpace = 'nowrap';
+        uiStateTextElement.setAttribute('role', 'alert');
+        uiStateTextElement.setAttribute('aria-live', 'assertive');
+        uiCanvas.parentNode.insertBefore(uiStateTextElement, uiCanvas);
+      }
+      if (uiStateTextElement.innerHTML != text) {
+        uiStateTextElement.innerHTML = text;
+      }
+    };
+  
+    function getDimension() {
+      return {
+        height: uiCanvas.height,
+        width: uiCanvas.width
+      };
+    }
+  
+    function drawCorners(point, index) {
+      let currentOrientation = window.orientation;
+      let dimension = getDimension();
+      let offsetX = dimension.width * 0.08;
+      let offsetY = dimension.height * 0.07;
+  
+      if (currentOrientation !== 0) {
+        offsetX = dimension.width * 0.07;
+        offsetY = dimension.height * 0.08;
+      }
+  
+      switch (index.toString()) {
+      case '1':
+        offsetX = -offsetX;
+        break;
+      case '2':
+        offsetX = -offsetX;
+        offsetY = -offsetY;
+        break;
+      case '3':
+        offsetY = -offsetY;
+        break;
+      default:
+        break;
+      }
+      drawCorner(point, offsetX, offsetY);
+    }
+  
+    function handleUi() {
+  
+      //TODO something like the below would be good, but was not working right now. Since we are rushed for a release can be left for later. 
+      // Essentially there is no reason to redraw the ui if nothing changed, but right now we do.
+  
+      // if (onDetectedResult) {
+      //   console.log(onDetectedResult.state, previousState);
+  
+      //   if (previousState === onDetectedResult.state && 
+      //     onDetectedResult.state != AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT && 
+      //     onDetectedResult.state != AcuantCamera.DOCUMENT_STATE.CAPTURING) {
+      //     previousState = onDetectedResult.state;
+      //     return;
+      //   }
+  
+      //   previousState = onDetectedResult.state;
+      // }
+  
+      uiContext.clearRect(0, 0, uiCanvas.width, uiCanvas.height);
+      
+      if (!onDetectedResult) {
+        drawPoints('#000000');
+        drawText(userOptions.text.NONE);
+      } else if (onDetectedResult.state === UI_STATE.CAPTURING) {
+        drawPoints('#00ff00');
+        drawOverlay('rgba(0, 255, 0, 0.2)');
+        drawText(userOptions.text.CAPTURING, 0.05, '#00ff00', false);
+      } else if (onDetectedResult.state === UI_STATE.TAP_TO_CAPTURE) {
+        drawPoints('#000000');
+        drawText(userOptions.text.TAP_TO_CAPTURE);
+      } else if (onDetectedResult.state === AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT) {
+        drawPoints('#ffff00');
+        drawOverlay('rgba(255, 255, 0, 0.2)');
+  
+        if (userOptions.text.GOOD_DOCUMENT) {
+          drawText(userOptions.text.GOOD_DOCUMENT, 0.09, '#ff0000', false);
+        } else {
+          let diff = Math.ceil((TRIGGER_TIME - (new Date().getTime() - counter)) / 1000);
+          if (diff <= 0) {
+            diff = 1;
+          }
+          drawText(diff + '...', 0.09, '#ff0000', false);
+        }
+  
+      } else if (onDetectedResult.state === AcuantCamera.DOCUMENT_STATE.SMALL_DOCUMENT) {
+        drawPoints('#ff0000');
+        drawText(userOptions.text.SMALL_DOCUMENT);
+      } else if (onDetectedResult.state === AcuantCamera.DOCUMENT_STATE.BIG_DOCUMENT) {
+        drawPoints('#ff0000');
+        drawText(userOptions.text.BIG_DOCUMENT);
+      } else {
+        drawPoints('#000000');
+        drawText(userOptions.text.NONE);
+      }
+    }
+  
+    function drawOverlay(style) {
+      if (onDetectedResult && onDetectedResult.points && onDetectedResult.points.length === 4) {
+        uiContext.beginPath();
+  
+        uiContext.moveTo(Math.round(onDetectedResult.points[0].x), Math.round(onDetectedResult.points[0].y));
+  
+        for (let i = 1; i < onDetectedResult.points.length; i++) {
+          uiContext.lineTo(Math.round(onDetectedResult.points[i].x), Math.round(onDetectedResult.points[i].y));
+        }
+        uiContext.fillStyle = style;
+        uiContext.strokeStyle = 'rgba(0, 0, 0, 0)';
+        uiContext.stroke();
+        uiContext.fill();
+      }
+    }
+  
+    function drawCorner(point, offsetX, offsetY) {
+      uiContext.beginPath();
+      const roundedX  = Math.round(point.x);
+      const roundedY = Math.round(point.y);
+      uiContext.moveTo(roundedX, roundedY);
+      uiContext.lineTo(Math.round(roundedX + offsetX), roundedY);
+      uiContext.moveTo(roundedX, roundedY);
+      uiContext.lineTo(roundedX, Math.round(roundedY + offsetY));
+      uiContext.stroke();
+    }
+  
+    function drawPoints(fillStyle) {
+      let dimension = getDimension();
+      uiContext.lineWidth = (Math.ceil(Math.max(dimension.width, dimension.height) * 0.0025) || 1);
+      uiContext.strokeStyle = fillStyle;
+      if (onDetectedResult && onDetectedResult.points && (onDetectedResult.state === -1 || onDetectedResult.state === AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT)) {
+        for (let i in onDetectedResult.points) {
+          drawCorners(onDetectedResult.points[i], i);
+        }
+      } else {
+        let center = {
+          x: dimension.width / 2,
+          y: dimension.height / 2
+        };
+  
+        let offsetX, offsetY;
+        let targetWidth, targetHeight;
+  
+        if (dimension.width > dimension.height) {
+          targetWidth = 0.85 * dimension.width;
+          targetHeight = 0.85 * dimension.width / 1.58870;
+          if (targetHeight > 0.85 * dimension.height) {
+            targetWidth = targetWidth / targetHeight * 0.85 * dimension.height;
+            targetHeight = 0.85 * dimension.height;
+          }
+        } else {
+          targetWidth = 0.85 * dimension.height / 1.58870;
+          targetHeight = 0.85 * dimension.height;
+          if (targetWidth > 0.85 * dimension.width) {
+            targetHeight = targetHeight / targetWidth * 0.85 * dimension.width;
+            targetWidth = 0.85 * dimension.width;
+          }
+        }
+  
+        offsetX = targetWidth / 2;
+        offsetY = targetHeight / 2;
+  
+        let defaultCorners = [
+          { x: center.x - offsetX, y: center.y - offsetY },
+          { x: center.x + offsetX, y: center.y - offsetY },
+          { x: center.x + offsetX, y: center.y + offsetY },
+          { x: center.x - offsetX, y: center.y + offsetY }
+        ];
+  
+        defaultCorners.forEach((point, i) => {
+          drawCorners(point, i);
+        });
+      }
+    }
+  
+    return svc;
+  
+  })();
+  
+  /* eslint no-undef: 0 */
+  // eslint-disable-next-line no-unused-vars, no-var
+  var AcuantCamera = (() => {
+    'use strict';
+    let player = null;
+    let uiCanvas = null;
+    let uiContext = null;
+    let manualCaptureInput = null;
+    let hiddenCanvas = null;
+    let hiddenContext = null;
+  
+    const SEQUENCE_BREAK_IOS_15 = 'Camera capture failed due to unexpected sequence break. This usually indicates the camera closed or froze unexpectedly. In iOS 15+ this is intermittently occurs due to a GPU Highwater failure. Swap to manual capture until the user fully reloads the browser. Attempting to continue to use live capture can lead to further Highwater errors and can cause to OS to cut off the webpage.';
+  
+    const SEQUENCE_BREAK_OTHER = 'Camera capture failed due to unexpected sequence break. This usually indicates the camera closed or froze unexpectedly. Swap to manual capture until the user fully reloads the browser.';
+  
+    const DOCUMENT_STATE = {
+      NO_DOCUMENT: 0,
+      SMALL_DOCUMENT: 1,
+      BIG_DOCUMENT: 2,
+      GOOD_DOCUMENT: 3
+    };
+  
+    const ACUANT_DOCUMENT_TYPE = {
+      NONE: 0,
+      ID: 1,
+      PASSPORT: 2
+    };
+  
+    const TARGET_LONGSIDE_SCALE = 700;
+    const TARGET_SHORTSIDE_SCALE = 500;
+    const MAX_VIDEO_WIDTH_IOS_15 = 1920;
+  
+    let onDetectCallback = null;
+    let onErrorCallback = null;
+    let onManualCaptureCallback = null;
+    let isStarted = false;
+    let isDetecting = false;
+    let detectTimeout = null;
+    let acuantCamera;
+    let modelFromHighEntropyValues;
+  
+    let svc = {
+      start: start,
+      startManualCapture: startManualCapture,
+      triggerCapture: triggerCapture,
+      end: endCamera,
+      DOCUMENT_STATE: DOCUMENT_STATE,
+      ACUANT_DOCUMENT_TYPE: ACUANT_DOCUMENT_TYPE,
+      isCameraSupported: 'mediaDevices' in navigator && mobileCheck(),
+      isIOSWebview: webViewCheck(),
+      isIOS: isIOS,
+      setRepeatFrameProcessor: processFrameForDetection,
+      evaluateImage: evaluateImage
+    };
+      
+    function webViewCheck() {
+      const standalone = window.navigator.standalone,
+        userAgent = window.navigator.userAgent.toLowerCase(),
+        safari = /safari/.test( userAgent ),
+        ios = /iphone|ipod|ipad/.test( userAgent );
+      
+      return (ios && !safari && !standalone);
+    }
+  
+    function isFireFox(){
+      return navigator.userAgent.toLowerCase().indexOf('firefox') > -1;
+    }
+  
+    function checkIOSVersion() {
+      return iOSversion()[0] >= 13;
+    }
+  
+    function mobileCheck() {
+      let check = false;
+      // eslint-disable-next-line no-useless-escape
+      (function (a) { if (/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino|android|playbook|silk/i.test(a) || /1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(a.substr(0, 4))) check = true; })(navigator.userAgent || navigator.vendor || window.opera);  
+      return (check || isIOS()) && (!isFireFox());
+    }
+  
+    function isIOS() {
+      return ((/iPad|iPhone|iPod/.test(navigator.platform) && checkIOSVersion()) || (navigator.platform === 'MacIntel' && navigator.maxTouchPoints > 1));
+    }
+  
+    function isSamsungNote10OrS10OrNewer() {
+      let deviceModel;
+      if (modelFromHighEntropyValues) {
+        deviceModel = modelFromHighEntropyValues;
+      } else {
+        deviceModel = navigator.userAgent;
+      }
+  
+      const matchedModelNumber = deviceModel.match(/SM-[N|G|S]\d{3}/);
+      if (!matchedModelNumber) {
+        return false;
+      }
+  
+      const modelNumber = parseInt(matchedModelNumber[0].match(/\d{3}/)[0], 10);
+      const smallerModelNumber = deviceModel.match(/SM-S\d{3}/) ? 900 : 970; // S10e
+      return !isNaN(modelNumber) && modelNumber >= smallerModelNumber;
+    }
+  
+    const userConfig = getUserConfig();
+  
+    function getUserConfig() {
+      let config = {
+        frameScale: 1,
+        primaryConstraints: {
+          video: {
+            facingMode: { exact: 'environment' },
+            //1.3333 aspect ratio and lower ideal height seems to result in better focus and higher dpi on ios devices.
+            // I believe this is because the true ratio of the camera is 4:3 so this ratio means people do not need to get as close
+            // to the id to get good dpi. Cameras might struggle to focus at very close angles.
+            aspectRatio: 4 / 3,
+            resizeMode: 'none',
+  
+            //This might help performance, especially on low end devices. Might be less important after apple fixes the ios 15 bug,
+            // but right now we want to do as little redraws as possible to avoid GPU load.
+            //20221110 update removing this as this did not seem to be relevant and might be negatively affecting firefox
+            //frameRate: { min: 10, ideal: 15, max: 15 }
+          }
+        },
+        fixedHeight: null,
+        fixedWidth: null,
+      };
+  
+      if (isIOS()) {
+        if (isDeviceAffectedByIOS16Issue()) {
+          //need to bump resolution since we will be capturing from further away
+          //also change aspect ratio to make it seem like the id is closer from user's pov.
+          config.primaryConstraints.video.aspectRatio = Math.max(window.innerWidth, window.innerHeight) * 1.0 / Math.min(window.innerWidth, window.innerHeight);
+          config.primaryConstraints.video.height = { min: 1440, ideal: 2880 };
+        } else if (isiOS15()) {
+          //ios 15 is unstable with large images so we keep the resolution lower
+          config.primaryConstraints.video.width = { min: MAX_VIDEO_WIDTH_IOS_15, ideal: MAX_VIDEO_WIDTH_IOS_15 };
+        } else {
+          //other ios devices default to safe size that was in use in the past
+          config.primaryConstraints.video.height = { min: 1440, ideal: 1440 };
+        }
+      } else {
+        //todo potentially can push this higher to get better dpi (especially on newer android versions)
+        config.primaryConstraints.video.height = { min: 1440, ideal: 1440 };
+      } 
+  
+      return config;
+    }
+  
+    function enableCamera(stream) {
+      startImageWorker().then(() => {
+        isStarted = true;
+        player.srcObject = stream;
+        addEvents();
+        player.play();
+      });
+    }
+  
+    function callCameraError(error, code) {
+      document.cookie = 'AcuantCameraHasFailed=' + code;
+      endCamera();
+      if (onErrorCallback && typeof (onErrorCallback) === 'function') {
+        onErrorCallback(error, code);
+      } else {
+        console.error('No error callback set. Review implementation.');
+        console.error(error, code);
+      }
+    }
+  
+  
+    function isBackCameraLabel(cameraLabel) {
+      const backCameraKeywords = [
+        'rear',
+        'back',
+        'rück',
+        'arrière',
+        'trasera',
+        'trás',
+        'traseira',
+        'posteriore',
+        '后面',
+        '後面',
+        '背面',
+        'задней',
+        'الخلفية',
+        '후',
+        'arka',
+        'achterzijde',
+        'หลัง',
+        'baksidan',
+        'bagside',
+        'sau',
+        'bak',
+        'tylny',
+        'takakamera',
+        'belakang',
+        'אחורית',
+        'πίσω',
+        'spate',
+        'hátsó',
+        'zadní',
+        'darrere',
+        'zadná',
+        'задня',
+        'stražnja',
+        'belakang',
+        'बैक'
+      ];
+      return backCameraKeywords.some((keyword) => {
+        return cameraLabel.includes(keyword);
+      });
+    }
+  
+    //sets up constraints to target the best camera. I think this is what the original implementer went off of:
+    //https://old.reddit.com/r/javascript/comments/8eg8w5/choosing_cameras_in_javascript_with_the/
+    //could be prone to error
+    function getDevice() {
+      return new Promise((resolve) => {
+        navigator.mediaDevices.enumerateDevices()
+          .then((devices) => {
+            const minSuffixDevice = {
+              suffix: undefined,
+              device: undefined,
+            };
+            const dualWideCamera = devices.find(d => d.label === 'Back Dual Wide Camera');
+            if (isiOS164Plus() && dualWideCamera) {
+              minSuffixDevice.device = dualWideCamera;
+            } else {
+              devices
+                .filter(device => device.kind === 'videoinput')
+                .forEach(device => {
+                  if (isBackCameraLabel(device.label)) {
+                    let split = device.label.split(',');
+                    let cameraSuffixNumber = parseInt(split[0][split[0].length -1]);
+  
+                    if (cameraSuffixNumber || cameraSuffixNumber === 0) {
+                      if (minSuffixDevice.suffix === undefined || minSuffixDevice.suffix > cameraSuffixNumber) {
+                        minSuffixDevice.suffix = cameraSuffixNumber;
+                        minSuffixDevice.device = device;
+                      }
+                    }
+                  }
+                });
+            }
+            resolve(minSuffixDevice.device);
+          })
+          .catch(() => {
+            resolve();
+          });
+      });        
+    }
+  
+    function startCamera(constraints, iteration = 0) {
+      const MAX_ITERATIONS = 2;
+  
+      if (iteration === 0) {
+        acuantCamera.dispatchEvent(new Event('acuantcameracreated'));
+      }
+  
+      const getDeviceWasAbleToReadLabels = Boolean(constraints.video.deviceId);
+      navigator.mediaDevices.getUserMedia(constraints)
+        .then((stream) => {
+          if (!getDeviceWasAbleToReadLabels && iteration < MAX_ITERATIONS) {
+            // Check if there is a better camera to use once we have permissions
+            getDevice().then((device) => {
+              if (device && device.deviceId !== stream.getVideoTracks()[0].getSettings().deviceId) {
+                userConfig.primaryConstraints.video.deviceId = device.deviceId;
+                stopMediaTracks(stream);
+                startCamera(userConfig.primaryConstraints, iteration++);
+              } else {
+                enableCamera(stream);
+              }
+            });
+          } else {
+            enableCamera(stream);
+          }
+        })
+        .catch((error) => {
+          callCameraError(error, AcuantJavascriptWebSdk.START_FAIL_CODE);
+        });
+    }
+  
+    //because of an ios bug to do with rotation this method will get called on rotation as the workaround
+    //that we have come up with is to close and reboot the camera on rotation. (SEE MOBILE-1250). As such
+    //the method needs to be able to handle cases where it is called more than once.
+    function start(callback, manualCallback, errorCallback) {
+  
+      if (errorCallback) {
+        onErrorCallback = errorCallback;
+      }
+  
+      if (cameraHasFailedBefore()) {
+        errorCallback('Live capture has previously failed and was called again. User was sent to manual capture.', AcuantJavascriptWebSdk.REPEAT_FAIL_CODE);
+        startManualCapture(manualCallback);
+        return;
+      }
+  
+      acuantCamera = document.getElementById('acuant-camera');
+  
+      if(!acuantCamera) {
+        callCameraError('Expected div with \'acuant-camera\' id', AcuantJavascriptWebSdk.START_FAIL_CODE);
+        return;
+      }
+  
+      acuantCamera.style.position='relative';
+      acuantCamera.innerHTML=
+              '<video id="acuant-player" autoplay playsinline style="display:block; image-orientation:none; object-fit: fill;"></video>' +
+              '<canvas id="acuant-ui-canvas" style="display:block; position:absolute; top:0; left:0;"></canvas>';
+  
+      player = document.getElementById('acuant-player');
+  
+      hiddenCanvas = document.createElement('canvas');
+      //willReadFrequently is only enabled on Android since it's not implemented on iOS
+      hiddenContext = hiddenCanvas.getContext('2d', { willReadFrequently: !isIOS() });
+  
+      uiCanvas = document.getElementById('acuant-ui-canvas');
+  
+      if (isStarted) {
+        callCameraError('already started.', AcuantJavascriptWebSdk.START_FAIL_CODE);
+      } else if (!player || !uiCanvas) {
+        callCameraError('Missing HTML elements.', AcuantJavascriptWebSdk.START_FAIL_CODE);
+      } else {
+        uiContext = uiCanvas.getContext('2d');
+        if (callback) {
+          onDetectCallback = callback;
+        }
+  
+        setZoomConstraintThenStartCamera();
+      }
+    }
+  
+    function setZoomConstraintThenStartCamera() {
+      if (navigator.userAgentData && navigator.userAgentData.getHighEntropyValues) {
+        navigator.userAgentData.getHighEntropyValues(['model']).then(dm => {
+          if (typeof(dm) === 'string') {
+            modelFromHighEntropyValues = dm;
+          } else if (typeof(dm.model) === 'string') {
+            modelFromHighEntropyValues = dm.model;
+          }
+        }).finally(() => {
+          if (isSamsungNote10OrS10OrNewer()) {
+            userConfig.primaryConstraints.video.zoom = 2.0;
+          } else  if (isiOS17()) {
+            userConfig.primaryConstraints.video.zoom = 1.6;
+          }
+          startCamera(userConfig.primaryConstraints);
+        });
+      } else {
+        if (isSamsungNote10OrS10OrNewer()) {
+          userConfig.primaryConstraints.video.zoom = 2.0;
+        } else  if (isiOS17()) {
+          userConfig.primaryConstraints.video.zoom = 1.6;
+        }
+        startCamera(userConfig.primaryConstraints);
+      }
+    }
+
+    function isMobile() {
+      return /Android|webOS|iPhone|iPad|iPod|BlackBerry|IEMobile|Opera Mini/i.test(navigator.userAgent);
+    }
+  
+    function startManualCapture(callback) {
+      onManualCaptureCallback = callback;
+      if (!manualCaptureInput && isMobile()) {
+        manualCaptureInput = document.createElement('input');
+        manualCaptureInput.type = 'file';
+        manualCaptureInput.capture = 'environment';
+        manualCaptureInput.accept = 'image/*';
+        manualCaptureInput.onclick = function(event) {
+          if (event && event.target) {
+            event.target.value = '';
+          }
+        };
+      }
+      manualCaptureInput.onchange = onManualCapture;
+      manualCaptureInput.click();
+    }
+  
+    function getOrientation(e) {
+      const view = new DataView(e.target.result);
+      if (view.getUint16(0, false) != 0xFFD8) {
+        return -2;
+      }
+      const length = view.byteLength;
+      let offset = 2;
+      while (offset < length) {
+        if (view.getUint16(offset + 2, false) <= 8) return -1;
+        const marker = view.getUint16(offset, false);
+        offset += 2;
+        if (marker == 0xFFE1) {
+          if (view.getUint32(offset += 2, false) != 0x45786966) {
+            return -1;
+          }
+  
+          const little = view.getUint16(offset += 6, false) == 0x4949;
+          offset += view.getUint32(offset + 4, little);
+          const tags = view.getUint16(offset, little);
+          offset += 2;
+          for (let i = 0; i < tags; i++) {
+            if (view.getUint16(offset + (i * 12), little) == 0x0112) {
+              return view.getUint16(offset + (i * 12) + 8, little);
+            }
+          }
+        }
+        else if ((marker & 0xFF00) != 0xFF00) {
+          break;
+        }
+        else {
+          offset += view.getUint16(offset, false);
+        }
+      }
+      return -1;
+    }
+  
+    let captureOrientation = -1;
+  
+    function onManualCapture(event) { 
+      hiddenCanvas = document.createElement('canvas');
+      hiddenContext = hiddenCanvas.getContext('2d');
+      hiddenContext.mozImageSmoothingEnabled = false;
+      hiddenContext.webkitImageSmoothingEnabled = false;
+      hiddenContext.msImageSmoothingEnabled = false;
+      hiddenContext.imageSmoothingEnabled = false;
+  
+      let file = event.target;
+      let reader = new FileReader();
+  
+      const isHeicFile = event.target.files[0] && event.target.files[0].name && event.target.files[0].name.toLowerCase().endsWith('.heic'); 
+      if (isHeicFile) {
+        reader.onload = (e) => {
+          getImageDataFromHeic(e.target.result)
+            .then(result => {
+              captureOrientation = 6;
+              onManualCaptureCallback.onCaptured(result);
+              evaluateImage(result.data, result.width, result.height, false, 'MANUAL', onManualCaptureCallback.onCropped);
+            })
+            .catch(e => onManualCaptureCallback.onError(e.error, e.code));
+        };
+      } else {
+        reader.onload = (e) => {
+          captureOrientation = getOrientation(e);
+          let image = document.createElement('img');
+          image.onload = () => {
+            const { width, height } = calculateImageSize(image.width, image.height);
+            hiddenCanvas.width = width;
+            hiddenCanvas.height = height;
+  
+            hiddenContext.drawImage(image, 0, 0, width, height);
+            const imgData = hiddenContext.getImageData(0, 0, width, height);
+  
+            hiddenContext.clearRect(0, 0, width, height);
+            image.remove();
+  
+            onManualCaptureCallback.onCaptured({
+              data: imgData,
+              width: width,
+              height: height
+            });
+  
+            evaluateImage(imgData, width, height, false, 'MANUAL', onManualCaptureCallback.onCropped);
+          };
+          image.src = 'data:image/jpeg;base64,' + arrayBufferToBase64(e.target.result);
+        };
+      }
+  
+      if (file && file.files[0]) {
+        reader.readAsArrayBuffer(file.files[0]);
+      }
+    }
+  
+    function getImageDataFromHeic(arrayBuffer) {
+      return new Promise((resolve, reject) => {
+        const magickWasm = window['magick-wasm'];
+        if (!magickWasm) {
+          reject({
+            error: 'HEIC image processing failed. Please make sure Image Magick scripts were integrated as expected.',
+            code: AcuantJavascriptWebSdk.HEIC_NOT_SUPPORTED_CODE
+          });
+          return;
+        }
+  
+        magickWasm.initializeImageMagick().then(() => {
+          magickWasm.ImageMagick.read(new Uint8Array(arrayBuffer), image => {
+            const { width, height } = calculateImageSize(image.width, image.height);
+            hiddenCanvas.width = width;
+            hiddenCanvas.height = height;
+            image.resize(width, height);
+            image.writeToCanvas(hiddenCanvas);
+            const data = hiddenContext.getImageData(0, 0, width, height);
+            resolve({ data, width, height });
+          });
+        });
+      });
+    }
+  
+    function calculateImageSize(width, height) {
+      let maxWidth = 2560,
+        maxHeight = 1920;
+  
+      if (isiOS15()) {
+        maxWidth = MAX_VIDEO_WIDTH_IOS_15;
+        maxHeight = Math.floor(MAX_VIDEO_WIDTH_IOS_15 * 3 / 4);
+      }
+  
+      const largerDimension = width > height ? width : height;
+  
+      if (largerDimension > maxWidth) {
+        if (width < height) {
+          const aspectRatio = height / width;
+          maxHeight = maxWidth;
+          maxWidth = maxHeight / aspectRatio;
+        }
+        else {
+          const aspectRatio = width / height;
+          maxHeight = maxWidth / aspectRatio;
+        }
+      } else {
+        maxWidth = width;
+        maxHeight = height;
+      }
+  
+      return { width: maxWidth, height: maxHeight };
+    }
+  
+    function stopMediaTracks(stream) {
+      stream.getTracks().forEach((track) => {
+        track.stop();
+      });
+    }
+  
+    function isChromeForIOS() {
+      const userAgent = window.navigator.userAgent;
+      const isWebkit = userAgent.indexOf('WebKit') > -1;
+      const isChrome = userAgent.indexOf('CriOS') > -1;
+      return isWebkit && isChrome && isIOS();
+    }
+  
+    function endCamera() {
+      isStarted = false;
+      isDetecting = false;
+      captureOrientation = -1;
+      if (detectTimeout) {
+        clearTimeout(detectTimeout);
+        detectTimeout = null;
+      }
+  
+      removeEvents();
+  
+      if (player) {
+        player.pause();
+        if (player.srcObject) {
+          stopMediaTracks(player.srcObject);
+        }
+        player = null;
+      }
+  
+      if (acuantCamera) {
+        acuantCamera.innerHTML = '';
+      }
+  
+      if (manualCaptureInput) {
+        manualCaptureInput.remove();
+        manualCaptureInput = null;
+      }
+    }
+  
+    function iOSversion() {
+      if (/iP(hone|od|ad)/.test(navigator.platform)) {
+        // supports iOS 2.0 and later: <http://bit.ly/TJjs1V>
+        try {
+          const v = (navigator.appVersion).match(/OS (\d+)_(\d+)_?(\d+)?/);
+          return [parseInt(v[1], 10), parseInt(v[2], 10), parseInt(v[3] || 0, 10)];
+        } catch (_) {
+          return -1;
+        }
+      }
+      return -1;
+    }
+  
+    function isiOS144Plus() {
+      let ver = iOSversion();
+      return ver && ver != -1 && ver.length >= 2 && ver[0] == 14 && ver[1] >= 4;
+    }
+  
+    function isiOS15() {
+      let ver = iOSversion();
+      return ver && ver != -1 && ver.length >= 1 && ver[0] == 15;
+    }
+  
+    function isiOS163OrLess() {
+      let ver = iOSversion();
+      return ver && ver != -1 && ver.length >= 1 && ver[0] == 16 && ver [1] < 4;
+    }
+  
+    function isiOS164Plus() {
+      let ver = iOSversion();
+      return ver && ver != -1 && ver.length >= 1 && ver[0] == 16 && ver [1] >= 4;
+    }
+  
+    function isiOS17() {
+      let ver = iOSversion();
+      return ver && ver != -1 && ver.length >= 1 && ver[0] >= 17;
+    }
+  
+    function isDeviceAffectedByIOS16Issue() {
+      let decodedCookie = decodeURIComponent(document.cookie);
+      if (decodedCookie.includes('AcuantForceRegularCapture=true')) {
+        return false;
+      }
+      if (decodedCookie.includes('AcuantForceDistantCapture=true')) {
+        return true;
+      }
+      if (isiOS163OrLess()) {
+        let dims = [screen.width, screen.height];
+        let long = Math.max(...dims);
+        let short = Math.min(...dims);
+        if (long == 852 && short == 393) { //14 pro
+          return true;
+        }
+        if (long == 932 && short == 430) { //14 pro max
+          return true;
+        }
+        if (long == 844 && short == 390) { //13 pro
+          return true;
+        }
+        if (long == 926 && short == 428) { //13 pro max
+          return true;
+        }
+      }
+      return false;
+    }
+  
+    function isiPad13Plus() {
+      return navigator.maxTouchPoints && navigator.maxTouchPoints >= 2 && /MacIntel/.test(navigator.platform);
+    }
+  
+    function onResize() {
+      uiContext.clearRect(0, 0, uiCanvas.width, uiCanvas.height);
+  
+      if (player) {
+        if (isIOS() && (isChromeForIOS() || isiOS144Plus())) {
+          endCamera();
+          start(); 
+        } else {
+          setDimens();
+        }
+      }
+    }
+  
+    function setDimens() {
+      let targetWidth;
+      let targetHeight;
+  
+      if (player.videoWidth < player.videoHeight) {
+        userConfig.canvasScale = (player.videoHeight / player.videoWidth);
+      }
+      else {
+        userConfig.canvasScale = (player.videoWidth / player.videoHeight);
+      }
+      targetWidth = window.innerWidth;
+      targetHeight = window.innerHeight;
+      if (!userConfig.fixedHeight) userConfig.fixedHeight = acuantCamera.clientHeight;
+      if (!userConfig.fixedWidth) userConfig.fixedWidth = acuantCamera.clientWidth;
+      if (userConfig.fixedHeight && userConfig.fixedWidth && userConfig.fixedWidth > userConfig.fixedHeight) {
+        userConfig.fixedHeight = acuantCamera.clientWidth;
+        userConfig.fixedWidth = acuantCamera.clientHeight;
+      }
+  
+      if (window.matchMedia('(orientation: portrait)').matches) {
+        if (userConfig.fixedWidth) {
+          if (userConfig.fixedWidth < targetWidth) {
+            targetWidth = userConfig.fixedWidth;
+          }
+          targetHeight = targetWidth * userConfig.canvasScale;
+        } else if (userConfig.fixedHeight) {
+          let width = userConfig.fixedHeight / userConfig.canvasScale;
+          if (width < targetWidth) {
+            targetHeight = userConfig.fixedHeight;
+            targetWidth = width;
+          }
+        }
+      } else {
+        if (userConfig.fixedWidth) {
+          if (userConfig.fixedWidth < targetHeight) {
+            targetHeight = userConfig.fixedWidth;
+          }
+          targetWidth = targetHeight * userConfig.canvasScale;
+        } else if (userConfig.fixedHeight) {
+          let width = userConfig.fixedHeight / userConfig.canvasScale;
+          if (width < targetHeight) {
+            targetWidth = userConfig.fixedHeight;
+            targetHeight = width;
+          }
+        }
+      }
+      player.width = uiCanvas.width = targetWidth;
+      player.height = uiCanvas.height = targetHeight;
+    }
+  
+    //We need to restart the camera when navigator is sent to background at devices where the zoom constraint is applied,
+    //otherwise, zoom value is override when resuming capture
+    function onVisibilityChange() {
+      if (manualCaptureInput || !isStarted) {
+        return;
+      }
+  
+      if (document.visibilityState === 'visible') {
+        endCamera();
+        start();
+      }
+    }
+  
+    function runDetection() {
+      if (!player || player.paused) {
+        return;
+      }
+  
+      processFrameForDetection();
+      detectTimeout = setTimeout(runDetection, 100); // detecting at most 10 times a second, short circuits out very quickly if detect is running
+    }
+  
+    function onLoadedMetaData() {
+      if (player.videoWidth + player.videoHeight < 1000) {
+        endCamera();
+        callCameraError('Camera not supported', AcuantJavascriptWebSdk.START_FAIL_CODE);
+      } else {
+        setDimens();
+        runDetection();
+      }
+    }
+  
+    // We need this to re-create the detect timeout after browser is backgrounded/resumed since the timer is throttled by the OS
+    function restartDetectionAfterBrowserResume() {
+      if (!detectTimeout) {
+        return;
+      }
+  
+      clearTimeout(detectTimeout);
+      isDetecting = false;
+      runDetection();
+    }
+  
+    function evaluateImage(imgData, width, height, isPortraitOrientation, capType, callback) {
+      if (AcuantJavascriptWebSdk.singleWorkerModel) {
+        evaluateImageHandlingWorkers(imgData, width, height, capType, isPortraitOrientation).then(callback);
+      } else {
+        evaluateImageWithoutHandlingWorkers(imgData, width, height, capType, isPortraitOrientation).then(callback);
+      }
+    }
+  
+    async function evaluateImageHandlingWorkers(imgData, width, height, capType, isPortraitOrientation) {
+      let result = { isPortraitOrientation };
+  
+      await startImageWorker();
+  
+      const cropResult = await crop(imgData, width, height);
+      if (!cropResult) {
+        return null;
+      }
+      result = { ...result, ...cropResult };
+  
+      endImageWorker();
+  
+      await startMetricsWorker();
+  
+      const moireResult = await moire(imgData, width, height);
+      const metricsResult = await metrics(result.image);
+      result = { ...result, ...moireResult, ...metricsResult };
+  
+      endMetricsWorker();
+  
+      const { imageBase64, imageBytes } = processImage(result, capType);
+      result.image.bytes = imageBytes;
+  
+      const barcodeText = await extractBarcode(imageBase64);
+      result.image.barcodeText = barcodeText;
+  
+      const imageWithExif = await addExif(result, capType, imageBase64);
+  
+      await startImageWorker();
+  
+      const signedImage = await sign(imageWithExif);
+      result.image.data = signedImage;
+  
+      endImageWorker();
+  
+      return result;
+    }
+  
+    async function evaluateImageWithoutHandlingWorkers(imgData, width, height, capType, isPortraitOrientation) {
+      let result = { isPortraitOrientation };
+  
+      const [cropResult, moireResult] = await Promise.all([crop(imgData, width, height), moire(imgData, width, height)]);
+      if (!cropResult) {
+        return null;
+      }
+      result = { ...result, ...cropResult };
+  
+      const metricsResult = await metrics(result.image);
+      result = { ...result, ...moireResult, ...metricsResult };
+  
+      const { imageBase64, imageBytes } = processImage(result, capType);
+      result.image.bytes = imageBytes;
+  
+      const barcodeText = await extractBarcode(imageBase64);
+      result.image.barcodeText = barcodeText;
+  
+      const imageWithExif = await addExif(result, capType, imageBase64);
+      const signedImage = await sign(imageWithExif);
+      result.image.data = signedImage;
+  
+      return result;
+    }
+  
+    function startImageWorker() {
+      return new Promise(resolve => {
+        AcuantJavascriptWebSdk.startImageWorker(resolve);
+      });
+    }
+  
+    function startMetricsWorker() {
+      return new Promise(resolve => {
+        AcuantJavascriptWebSdk.startMetricsWorker(resolve);
+      });
+    }
+  
+    function endImageWorker() {
+      AcuantJavascriptWebSdk.endImageWorker();
+    }
+  
+    function endMetricsWorker() {
+      AcuantJavascriptWebSdk.endMetricsWorker();
+    }
+  
+    function crop(imgData, width, height) {
+      return new Promise(resolve => {
+        AcuantJavascriptWebSdk.crop(imgData, width, height, {
+          onSuccess: ({ image, dpi, cardType }) => resolve({ image, dpi, cardType }),
+          onFail: resolve
+        });
+      });
+    }
+  
+    function moire(imgData, width, height) {
+      return new Promise(resolve => {
+        AcuantJavascriptWebSdk.moire(imgData, width, height, {
+          onSuccess: (moire, moireraw) => resolve({ moire, moireraw }),
+          onFail: () => resolve({ moire: -1, moireraw: -1 })
+        });
+      });
+    }
+  
+    function metrics(image) {
+      return new Promise(resolve => {
+        AcuantJavascriptWebSdk.metrics(image, image.width, image.height, {
+          onSuccess: (sharpness, glare) => resolve({ sharpness, glare }),
+          onFail: () => resolve({ sharpness: -1, glare: -1 }),
+        });
+      });
+    }
+  
+    function sign(imageBase64WithExif) {
+      return new Promise(resolve => {
+        const imageData = base64ToArrayBuffer(imageBase64WithExif);
+        AcuantJavascriptWebSdk.sign(imageData, {
+          onSuccess: (signedImageData) => resolve('data:image/jpeg;base64,' + arrayBufferToBase64(signedImageData)),
+          onFail: resolve
+        });
+      });
+    }
+  
+    async function extractBarcode(imageBase64) {
+      if (!document.getElementById(AcuantJavascriptWebSdk.BARCODE_READER_ID)) {
+        return null;
+      }
+  
+      try {
+        return await scanImageBarcode(imageBase64);
+      } catch {
+        return null;
+      }
+    }
+  
+    function scanImageBarcode(image) {
+      let arr = image.split(','),
+        mime = arr[0].match(/:(.*?);/)[1],
+        bstr = atob(arr[1]),
+        n = bstr.length,
+        u8arr = new Uint8Array(n);
+      while(n--){
+        u8arr[n] = bstr.charCodeAt(n);
+      }
+      const imageFile = new File([u8arr], 'imageFile', {type:mime});
+      let html5qrcode = new Html5Qrcode(AcuantJavascriptWebSdk.BARCODE_READER_ID, { formatsToSupport: [ Html5QrcodeSupportedFormats.PDF_417 ] });
+      return html5qrcode.scanFile(imageFile, false);
+    }
+  
+    function removeEvents() {
+      if (isSamsungNote10OrS10OrNewer()) {
+        document.removeEventListener('visibilitychange', onVisibilityChange);
+      }
+      window.removeEventListener('resize', onResize);
+      if (player) {
+        player.removeEventListener('play', restartDetectionAfterBrowserResume);
+        player.removeEventListener('loadedmetadata', onLoadedMetaData);
+      }
+    }
+      
+    function addEvents() {
+      if (isSamsungNote10OrS10OrNewer()) {
+        document.addEventListener('visibilitychange', onVisibilityChange);
+      }
+      window.addEventListener('resize', onResize);
+      if (player) {    
+        player.addEventListener('play', restartDetectionAfterBrowserResume);
+        player.addEventListener('loadedmetadata', onLoadedMetaData);
+      }
+    }
+  
+    function processFrameForDetection() {
+      if (!isStarted || isDetecting) {
+        return;
+      }
+      if (player.videoWidth == 0) {
+        sequenceBreak();
+        return;
+      }
+      isDetecting = true;
+  
+      let max = Math.max(player.videoWidth, player.videoHeight);
+      let min = Math.min(player.videoWidth, player.videoHeight);
+      let newCanvasHeight = 0;
+      let newCanvasWidth = 0;
+  
+      if (max > TARGET_LONGSIDE_SCALE && min > TARGET_SHORTSIDE_SCALE) {
+        if (player.videoWidth >= player.videoHeight) {
+          userConfig.frameScale = (TARGET_LONGSIDE_SCALE / player.videoWidth);
+          newCanvasWidth = TARGET_LONGSIDE_SCALE;
+          newCanvasHeight = player.videoHeight * userConfig.frameScale;
+        }
+        else {
+          userConfig.frameScale = (TARGET_LONGSIDE_SCALE / player.videoHeight);
+          newCanvasWidth = player.videoWidth * userConfig.frameScale;
+          newCanvasHeight = TARGET_LONGSIDE_SCALE;
+        }
+      }
+      else {
+        userConfig.frameScale = 1;
+        newCanvasWidth = player.videoWidth;
+        newCanvasHeight = player.videoHeight;
+      }
+  
+      if (newCanvasWidth != hiddenCanvas.width || newCanvasHeight != hiddenCanvas.height) {
+        hiddenCanvas.width = newCanvasWidth;
+        hiddenCanvas.height = newCanvasHeight;
+      }
+  
+      if (isStarted) {
+        let imgData;
+        try {
+          hiddenContext.drawImage(player, 0, 0, player.videoWidth, player.videoHeight, 0, 0, hiddenCanvas.width, hiddenCanvas.height); //this scales down to target size
+          imgData = hiddenContext.getImageData(0, 0, hiddenCanvas.width, hiddenCanvas.height);
+          hiddenContext.clearRect(0, 0, hiddenCanvas.width, hiddenCanvas.height);
+        } catch (e) {
+          sequenceBreak();
+          return;
+        }
+  
+        detect(imgData, hiddenCanvas.width, hiddenCanvas.height);
+      }
+    }
+  
+    function detect(imgData, width, height) {
+      AcuantJavascriptWebSdk.detect(imgData, width, height, {
+        onSuccess: function (response) {
+          if (!hiddenCanvas || !player || player.paused || player.ended)
+            return;
+  
+          response.points.forEach(p => {
+            if(p.x !== undefined && p.y !== undefined){
+              p.x = (p.x / userConfig.frameScale * player.width / player.videoWidth);
+              p.y = (p.y / userConfig.frameScale * player.height / player.videoHeight);
+            }
+          });
+  
+          const minRatio = Math.min(response.dimensions.width, response.dimensions.height) / Math.min(hiddenCanvas.width, hiddenCanvas.height);
+          const maxRatio = Math.max(response.dimensions.width, response.dimensions.height) / Math.max(hiddenCanvas.width, hiddenCanvas.height);
+          const isPassport = response.type == 2;
+          let minRatioUpperBoundThreshold = 0.8;
+          let maxRatioUpperBoundThreshold = 0.85;
+          let minRatioLowerBoundThreshold = 0.6;
+          let maxRatioLowerBoundThreshold = 0.65;
+          if (isPassport) {
+            minRatioUpperBoundThreshold = 0.9;
+            maxRatioUpperBoundThreshold = 0.95;
+          }
+  
+          if (isIOS()) {
+            minRatioLowerBoundThreshold = 0.65;
+            maxRatioLowerBoundThreshold = 0.7;
+            if (isDeviceAffectedByIOS16Issue()) {
+              if (isPassport) { //intentionally large to help address issues with cvml inaccuaretly detecting passports
+                minRatioUpperBoundThreshold = 0.72;
+                maxRatioUpperBoundThreshold = 0.77;
+                minRatioLowerBoundThreshold = 0.22;
+                maxRatioLowerBoundThreshold = 0.28;
+                // minRatioLowerBoundThreshold = 0.57;
+                // maxRatioLowerBoundThreshold = 0.53;
+              } else { //id
+                minRatioUpperBoundThreshold = 0.41;
+                maxRatioUpperBoundThreshold = 0.45;
+                minRatioLowerBoundThreshold = 0.22;
+                maxRatioLowerBoundThreshold = 0.28;
+              }
+            } else if (isPassport) {
+              minRatioUpperBoundThreshold = 0.95; //closest the short side can be
+              maxRatioUpperBoundThreshold = 1;    //closest the long side can be
+              minRatioLowerBoundThreshold = 0.7;  //furthest the short side can be
+              maxRatioLowerBoundThreshold = 0.75; //furthest the long side can be
+            }
+          }
+  
+          const isTooFarAway = !response.isCorrectAspectRatio || (minRatio < minRatioLowerBoundThreshold && maxRatio < maxRatioLowerBoundThreshold);
+          const isTooClose = minRatio >= minRatioUpperBoundThreshold || maxRatio >= maxRatioUpperBoundThreshold;
+  
+          if (response.type === ACUANT_DOCUMENT_TYPE.NONE) {
+            response.state = DOCUMENT_STATE.NO_DOCUMENT;
+          } else if (isTooClose) {
+            response.state = DOCUMENT_STATE.BIG_DOCUMENT;
+          } else if (isTooFarAway) {
+            response.state = DOCUMENT_STATE.SMALL_DOCUMENT;
+          }  else {
+            response.state = DOCUMENT_STATE.GOOD_DOCUMENT;
+          }
+          onDetectCallback(response);
+          isDetecting = false;
+        },
+        onFail: function () {
+          if (!hiddenCanvas || !player || player.paused || player.ended)
+            return;
+  
+          let response = {};
+          response.state = DOCUMENT_STATE.NO_DOCUMENT;
+          onDetectCallback(response);
+          isDetecting = false;
+        }
+      });
+    }
+  
+    function triggerCapture(callback) {
+      let imgData, isPortraitOrientation;
+      try {
+        if (player.videoWidth == 0) {
+          throw 'width 0';
+        }
+        hiddenCanvas.width = player.videoWidth;
+        hiddenCanvas.height = player.videoHeight;
+      
+        hiddenContext.drawImage(player, 0, 0, hiddenCanvas.width, hiddenCanvas.height);
+        imgData = hiddenContext.getImageData(0, 0, hiddenCanvas.width, hiddenCanvas.height);
+        hiddenContext.clearRect(0, 0, hiddenCanvas.width, hiddenCanvas.height);
+        isPortraitOrientation = window.matchMedia('(orientation: portrait)').matches;
+      } catch (e) {
+        sequenceBreak();
+        return;
+      }
+          
+      callback({
+        data: imgData,
+        width: hiddenCanvas.width,
+        height: hiddenCanvas.height,
+        isPortraitOrientation,
+      });
+    }
+  
+    function processImage({ image, cardType, isPortraitOrientation }, capType) {
+      if (!hiddenCanvas || !hiddenContext) {
+        hiddenCanvas = document.createElement('canvas');
+        hiddenContext = hiddenCanvas.getContext('2d');
+      }
+  
+      hiddenCanvas.width = image.width;
+      hiddenCanvas.height = image.height;
+  
+      let mImgData = hiddenContext.createImageData(image.width, image.height);
+  
+      setImageData(mImgData.data, image.data);
+      hiddenContext.putImageData(mImgData, 0, 0);
+  
+      let isManual = capType != 'AUTO' && capType != 'TAP';
+  
+      if (cardType !== 2) {
+        if ((!isManual && isPortraitOrientation) || (isManual && captureOrientation === 6)) {
+          hiddenContext.rotate(Math.PI); //180 degrees
+          //the width and height are negative as an alternative to counting out the transform needed 
+          // to get to the middle of the image. Avoids possible mistakes with off by one pixel.
+          hiddenContext.drawImage(hiddenCanvas, 0, 0, -hiddenCanvas.width, -hiddenCanvas.height); 
+        }
+      }
+  
+      const imageBytes = hiddenContext.getImageData(0, 0, hiddenCanvas.width, hiddenCanvas.height).data;
+  
+      const quality = (typeof acuantConfig !== 'undefined' && typeof acuantConfig.jpegQuality === 'number') ? acuantConfig.jpegQuality : 1.0;
+      const imageBase64 = hiddenCanvas.toDataURL('image/jpeg', quality);
+      hiddenContext.clearRect(0, 0, hiddenCanvas.width, hiddenCanvas.height);
+  
+      if (hiddenCanvas) {
+        hiddenContext = null;
+        hiddenCanvas.remove();
+        hiddenCanvas = null;
+      }
+  
+      return { imageBase64, imageBytes };
+    }
+  
+    function getCvmlVersion() {
+      return new Promise((resolve) => {
+        AcuantJavascriptWebSdk.getCvmlVersion({
+          onSuccess: (version) => {
+            resolve(version);
+          },
+          onFail: () => {
+            resolve('unknown');
+          }
+        });
+      });
+    }
+  
+    async function addExif({ dpi, cardType, sharpness, glare, moire, moireraw }, capType, base64Img) {
+      const cvmlVersion = await getCvmlVersion();
+      const imageDescription = JSON.stringify({
+        'cvml':{ 
+          'cropping':{
+            'iscropped': true,
+            'dpi': dpi,
+            'idsize': cardType === 2 ? 'ID3': 'ID1',
+            'elapsed': -1
+          },
+          'sharpness':{
+            'normalized': sharpness,
+            'elapsed': -1
+          },
+          'moire':{
+            'normalized': moire,
+            'raw': moireraw,
+            'elapsed': -1
+          },
+          'glare':{
+            'normalized': glare,
+            'elapsed': -1
+          },
+          'version': cvmlVersion
+        },
+        'device':{
+          'version': getBrowserVersion(),
+          'capturetype': capType
+        }
+      });
+      return AcuantJavascriptWebSdk.addMetadata(base64Img, { imageDescription, dateTimeOriginal: new Date().toUTCString() });
+    }
+  
+    function getBrowserVersion() {
+      const ua = navigator.userAgent;
+      let tem, M = ua.match(/(opera|chrome|safari|firefox|msie|trident(?=\/))\/?\s*(\d+)/i) || [];
+  
+      if (/trident/i.test(M[1])) {
+        tem = /\brv[ :]+(\d+)/g.exec(ua) || [];
+        return 'IE '+(tem[1] || '');
+      }
+      if (M[1] === 'Chrome') {
+        tem = ua.match(/\b(OPR|Edge)\/(\d+)/);
+        if(tem != null) return tem.slice(1).join(' ').replace('OPR', 'Opera');
+      }
+      M = M[2] ? [M[1], M[2]]: [navigator.appName, navigator.appVersion, '-?'];
+      if ((tem = ua.match(/version\/(\d+)/i)) != null) M.splice(1, 1, tem[1]);
+      return M.join(' ');
+    }
+  
+    function setImageData(imgData, src) {
+      for (let i = 0; i < imgData.length; ++i) {
+        imgData[i] = src[i];
+      }
+    }
+  
+    function cameraHasFailedBefore() {
+      let name = 'AcuantCameraHasFailed=';
+      let decodedCookie = decodeURIComponent(document.cookie);
+      return decodedCookie.includes(name);
+    }
+  
+    function sequenceBreak() {
+      if (isiOS15() || isiPad13Plus()) {
+        callCameraError(SEQUENCE_BREAK_IOS_15, AcuantJavascriptWebSdk.SEQUENCE_BREAK_CODE);
+      } else {
+        callCameraError(SEQUENCE_BREAK_OTHER, AcuantJavascriptWebSdk.SEQUENCE_BREAK_CODE);
+      }
+    }
+  
+    function base64ToArrayBuffer(base64) {
+      const binary_string = window.atob(base64.split('base64,')[1]);
+      const len = binary_string.length;
+      const bytes = new Uint8Array(new ArrayBuffer(len));
+      for (let i = 0; i < len; i++) {
+        bytes[i] = binary_string.charCodeAt(i);
+      }
+      return bytes;
+    }
+  
+    function arrayBufferToBase64(buffer) {
+      let binary = '';
+      const bytes = new Uint8Array(buffer);
+      const len = bytes.byteLength;
+      for (let i = 0; i < len; i++) {
+        binary += String.fromCharCode(bytes[i]);
+      }
+      return window.btoa(binary);
+    }
+  
+    return svc;
+  })();
\ No newline at end of file
diff --git a/public/acuant/11.9.3.508.1/AcuantCamera.min.js b/public/acuant/11.9.3.508.1/AcuantCamera.min.js
new file mode 100644
index 00000000000..54e60da5550
--- /dev/null
+++ b/public/acuant/11.9.3.508.1/AcuantCamera.min.js
@@ -0,0 +1 @@
+var AcuantCameraUI=function(){"use strict";let e=null,t=null,i=null,a=null,n={start:function(e,i){s=e.onError,i&&(m=i,m.text.hasOwnProperty("BIG_DOCUMENT")||(m.text.BIG_DOCUMENT="TOO CLOSE"));AcuantCamera.isCameraSupported?r||(r=!0,v(),function(e){let i=0,n=(new Date).getTime();a=document.getElementById("acuant-camera"),a&&a.addEventListener("acuantcameracreated",C);AcuantCamera.start((a=>{!function(e,t){if(t>=3)return!0;{let t=(new Date).getTime()-e;return t<p&&(p=t),p<f}}(n,i)?(i+=1,n=(new Date).getTime(),AcuantCamera.setRepeatFrameProcessor()):p<f?function(e,t){if(e){if(o&&-1===o.state)return;t.onFrameAvailable&&t.onFrameAvailable(e),o=e,o.state===AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT?null===l&&(c=(new Date).getTime(),function(e){null===l&&(l=setTimeout((function(){o.state=h.CAPTURING,A(e,"AUTO")}),g))}(t)):(y(),c=null)}else y(),c=null}(a,e):function(e){o={state:h.TAP_TO_CAPTURE},t.addEventListener("click",T,!1),t.callback=e}(e)}),e,E)}(e)):E("Camera not supported.",AcuantJavascriptWebSdk.START_FAIL_CODE)},end:w},r=!1,o=null,c=null,d=null,s=null,l=null,u=null;const h={CAPTURING:-1,TAP_TO_CAPTURE:-2};let m={text:{NONE:"ALIGN",SMALL_DOCUMENT:"MOVE CLOSER",GOOD_DOCUMENT:null,BIG_DOCUMENT:"TOO CLOSE",CAPTURING:"CAPTURING",TAP_TO_CAPTURE:"TAP TO CAPTURE"}};const g=3e3,f=400;let p=Number.MAX_VALUE;function v(){o=null,c=null,d=null}function w(){v(),r=!1,l&&(clearTimeout(l),l=null),u&&(clearTimeout(u),u=null),e&&(e.removeEventListener("play",b,0),e=null),a&&a.removeEventListener("acuantcameracreated",C),t&&(t.removeEventListener("click",T),t.getContext("2d").clearRect(0,0,t.width,t.height)),AcuantCamera.end()}function y(){l&&(clearTimeout(l),l=null)}function T(e){A(e.currentTarget.callback,"TAP")}function C(){e=document.getElementById("acuant-player"),t=document.getElementById("acuant-ui-canvas"),i=t.getContext("2d"),t.setAttribute("role","img"),e.addEventListener("play",b,0)}function A(e,t){AcuantCamera.triggerCapture((i=>{w(),e.onCaptured(i),AcuantCamera.evaluateImage(i.data,i.width,i.height,i.isPortraitOrientation,t,(t=>{e.onCropped(t)}))}))}function E(e,t){w(),s&&s(e,t),s=null}function b(){!function a(){e&&!e.paused&&!e.ended&&r&&(!function(){if(i.clearRect(0,0,t.width,t.height),o)if(o.state===h.CAPTURING)D("#00ff00"),M("rgba(0, 255, 0, 0.2)"),x(m.text.CAPTURING,.05,"#00ff00",!1);else if(o.state===h.TAP_TO_CAPTURE)D("#000000"),x(m.text.TAP_TO_CAPTURE);else if(o.state===AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT)if(D("#ffff00"),M("rgba(255, 255, 0, 0.2)"),m.text.GOOD_DOCUMENT)x(m.text.GOOD_DOCUMENT,.09,"#ff0000",!1);else{let e=Math.ceil((g-((new Date).getTime()-c))/1e3);e<=0&&(e=1),x(e+"...",.09,"#ff0000",!1)}else o.state===AcuantCamera.DOCUMENT_STATE.SMALL_DOCUMENT?(D("#ff0000"),x(m.text.SMALL_DOCUMENT)):o.state===AcuantCamera.DOCUMENT_STATE.BIG_DOCUMENT?(D("#ff0000"),x(m.text.BIG_DOCUMENT)):(D("#000000"),x(m.text.NONE));else D("#000000"),x(m.text.NONE)}(),u=setTimeout(a,100))}()}function x(e,t=.04,a="#ffffff",n=!0){let r=S(),o=window.orientation,c=i.measureText(e),d=.01*Math.max(r.width,r.height),s=.02*Math.max(r.width,r.height),l=(r.height-s-c.width)/2,u=-(r.width/2-d),h=90;0!==o&&(h=0,l=(r.width-d-c.width)/2,u=r.height/2-s+.04*Math.max(r.width,r.height)),i.rotate(h*Math.PI/180),n&&(i.fillStyle="rgba(0, 0, 0, 0.5)",i.fillRect(Math.round(l-d),Math.round(u+d),Math.round(c.width+s),-Math.round(.05*Math.max(r.width,r.height)))),i.font=(Math.ceil(Math.max(r.width,r.height)*t)||0)+"px Sans-serif",i.fillStyle=a,i.fillText(e,l,u),O(e),i.rotate(-h*Math.PI/180)}const O=e=>{d||(d=document.createElement("p"),d.id="doc-state-text",d.style.height="1px",d.style.width="1px",d.style.margin="-1px",d.style.overflow="hidden",d.style.position="absolute",d.style.whiteSpace="nowrap",d.setAttribute("role","alert"),d.setAttribute("aria-live","assertive"),t.parentNode.insertBefore(d,t)),d.innerHTML!=e&&(d.innerHTML=e)};function S(){return{height:t.height,width:t.width}}function k(e,t){let a=window.orientation,n=S(),r=.08*n.width,o=.07*n.height;switch(0!==a&&(r=.07*n.width,o=.08*n.height),t.toString()){case"1":r=-r;break;case"2":r=-r,o=-o;break;case"3":o=-o}!function(e,t,a){i.beginPath();const n=Math.round(e.x),r=Math.round(e.y);i.moveTo(n,r),i.lineTo(Math.round(n+t),r),i.moveTo(n,r),i.lineTo(n,Math.round(r+a)),i.stroke()}(e,r,o)}function M(e){if(o&&o.points&&4===o.points.length){i.beginPath(),i.moveTo(Math.round(o.points[0].x),Math.round(o.points[0].y));for(let e=1;e<o.points.length;e++)i.lineTo(Math.round(o.points[e].x),Math.round(o.points[e].y));i.fillStyle=e,i.strokeStyle="rgba(0, 0, 0, 0)",i.stroke(),i.fill()}}function D(e){let t=S();if(i.lineWidth=Math.ceil(.0025*Math.max(t.width,t.height))||1,i.strokeStyle=e,o&&o.points&&(-1===o.state||o.state===AcuantCamera.DOCUMENT_STATE.GOOD_DOCUMENT))for(let e in o.points)k(o.points[e],e);else{let e,i,a,n,r={x:t.width/2,y:t.height/2};t.width>t.height?(a=.85*t.width,n=.85*t.width/1.5887,n>.85*t.height&&(a=a/n*.85*t.height,n=.85*t.height)):(a=.85*t.height/1.5887,n=.85*t.height,a>.85*t.width&&(n=n/a*.85*t.width,a=.85*t.width)),e=a/2,i=n/2,[{x:r.x-e,y:r.y-i},{x:r.x+e,y:r.y-i},{x:r.x+e,y:r.y+i},{x:r.x-e,y:r.y+i}].forEach(((e,t)=>{k(e,t)}))}}return n}(),AcuantCamera=(()=>{"use strict";let e=null,t=null,i=null,a=null,n=null,r=null;const o="Camera capture failed due to unexpected sequence break. This usually indicates the camera closed or froze unexpectedly. In iOS 15+ this is intermittently occurs due to a GPU Highwater failure. Swap to manual capture until the user fully reloads the browser. Attempting to continue to use live capture can lead to further Highwater errors and can cause to OS to cut off the webpage.",c="Camera capture failed due to unexpected sequence break. This usually indicates the camera closed or froze unexpectedly. Swap to manual capture until the user fully reloads the browser.",d={NO_DOCUMENT:0,SMALL_DOCUMENT:1,BIG_DOCUMENT:2,GOOD_DOCUMENT:3},s={NONE:0,ID:1,PASSPORT:2},l=700,u=500,h=1920;let m,g,f=null,p=null,v=null,w=!1,y=!1,T=null,C={start:D,startManualCapture:I,triggerCapture:function(t){let i,a;try{if(0==e.videoWidth)throw"width 0";n.width=e.videoWidth,n.height=e.videoHeight,r.drawImage(e,0,0,n.width,n.height),i=r.getImageData(0,0,n.width,n.height),r.clearRect(0,0,n.width,n.height),a=window.matchMedia("(orientation: portrait)").matches}catch(e){return void re()}t({data:i,width:n.width,height:n.height,isPortraitOrientation:a})},end:P,DOCUMENT_STATE:d,ACUANT_DOCUMENT_TYPE:s,isCameraSupported:"mediaDevices"in navigator&&function(){let e=!1;return function(t){(/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino|android|playbook|silk/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4)))&&(e=!0)}(navigator.userAgent||navigator.vendor||window.opera),(e||E())&&!A()}(),isIOSWebview:function(){const e=window.navigator.standalone,t=window.navigator.userAgent.toLowerCase(),i=/safari/.test(t);return/iphone|ipod|ipad/.test(t)&&!i&&!e}(),isIOS:E,setRepeatFrameProcessor:te,evaluateImage:q};function A(){return navigator.userAgent.toLowerCase().indexOf("firefox")>-1}function E(){return/iPad|iPhone|iPod/.test(navigator.platform)&&R()[0]>=13||"MacIntel"===navigator.platform&&navigator.maxTouchPoints>1}function b(){let e;e=g||navigator.userAgent;const t=e.match(/SM-[N|G|S]\d{3}/);if(!t)return!1;const i=parseInt(t[0].match(/\d{3}/)[0],10),a=e.match(/SM-S\d{3}/)?900:970;return!isNaN(i)&&i>=a}const x=function(){let e={frameScale:1,primaryConstraints:{video:{facingMode:{exact:"environment"},aspectRatio:4/3,resizeMode:"none"}},fixedHeight:null,fixedWidth:null};E()?z()?(e.primaryConstraints.video.aspectRatio=1*Math.max(window.innerWidth,window.innerHeight)/Math.min(window.innerWidth,window.innerHeight),e.primaryConstraints.video.height={min:1440,ideal:2880}):L()?e.primaryConstraints.video.width={min:h,ideal:h}:e.primaryConstraints.video.height={min:1440,ideal:1440}:e.primaryConstraints.video.height={min:1440,ideal:1440};return e}();function O(t){Q().then((()=>{w=!0,e.srcObject=t,function(){b()&&document.addEventListener("visibilitychange",J);window.addEventListener("resize",B),e&&(e.addEventListener("play",V),e.addEventListener("loadedmetadata",j))}(),e.play()}))}function S(e,t){document.cookie="AcuantCameraHasFailed="+t,P(),p&&"function"==typeof p?p(e,t):(console.error("No error callback set. Review implementation."),console.error(e,t))}function k(){return new Promise((e=>{navigator.mediaDevices.enumerateDevices().then((t=>{const i={suffix:void 0,device:void 0},a=t.find((e=>"Back Dual Wide Camera"===e.label));(function(){let e=R();return e&&-1!=e&&e.length>=1&&16==e[0]&&e[1]>=4})()&&a?i.device=a:t.filter((e=>"videoinput"===e.kind)).forEach((e=>{if(t=e.label,["rear","back","rück","arrière","trasera","trás","traseira","posteriore","后面","後面","背面","задней","الخلفية","후","arka","achterzijde","หลัง","baksidan","bagside","sau","bak","tylny","takakamera","belakang","אחורית","πίσω","spate","hátsó","zadní","darrere","zadná","задня","stražnja","belakang","बैक"].some((e=>t.includes(e)))){let t=e.label.split(","),a=parseInt(t[0][t[0].length-1]);(a||0===a)&&(void 0===i.suffix||i.suffix>a)&&(i.suffix=a,i.device=e)}var t})),e(i.device)})).catch((()=>{e()}))}))}function M(e,t=0){0===t&&m.dispatchEvent(new Event("acuantcameracreated"));const i=Boolean(e.video.deviceId);navigator.mediaDevices.getUserMedia(e).then((e=>{!i&&t<2?k().then((i=>{i&&i.deviceId!==e.getVideoTracks()[0].getSettings().deviceId?(x.primaryConstraints.video.deviceId=i.deviceId,N(e),M(x.primaryConstraints,t++)):O(e)})):O(e)})).catch((e=>{S(e,AcuantJavascriptWebSdk.START_FAIL_CODE)}))}function D(a,o,c){if(c&&(p=c),function(){let e="AcuantCameraHasFailed=";return decodeURIComponent(document.cookie).includes(e)}())return c("Live capture has previously failed and was called again. User was sent to manual capture.",AcuantJavascriptWebSdk.REPEAT_FAIL_CODE),void I(o);m=document.getElementById("acuant-camera"),m?(m.style.position="relative",m.innerHTML='<video id="acuant-player" autoplay playsinline style="display:block; image-orientation:none; object-fit: fill;"></video><canvas id="acuant-ui-canvas" style="display:block; position:absolute; top:0; left:0;"></canvas>',e=document.getElementById("acuant-player"),n=document.createElement("canvas"),r=n.getContext("2d",{willReadFrequently:!E()}),t=document.getElementById("acuant-ui-canvas"),w?S("already started.",AcuantJavascriptWebSdk.START_FAIL_CODE):e&&t?(i=t.getContext("2d"),a&&(f=a),navigator.userAgentData&&navigator.userAgentData.getHighEntropyValues?navigator.userAgentData.getHighEntropyValues(["model"]).then((e=>{"string"==typeof e?g=e:"string"==typeof e.model&&(g=e.model)})).finally((()=>{b()?x.primaryConstraints.video.zoom=2:H()&&(x.primaryConstraints.video.zoom=1.6),M(x.primaryConstraints)})):(b()?x.primaryConstraints.video.zoom=2:H()&&(x.primaryConstraints.video.zoom=1.6),M(x.primaryConstraints))):S("Missing HTML elements.",AcuantJavascriptWebSdk.START_FAIL_CODE)):S("Expected div with 'acuant-camera' id",AcuantJavascriptWebSdk.START_FAIL_CODE)}function I(e){v=e,a||(a=document.createElement("input"),a.type="file",a.capture="environment",a.accept="image/*",a.onclick=function(e){e&&e.target&&(e.target.value="")}),a.onchange=_,a.click()}let U=-1;function _(e){n=document.createElement("canvas"),r=n.getContext("2d"),r.mozImageSmoothingEnabled=!1,r.webkitImageSmoothingEnabled=!1,r.msImageSmoothingEnabled=!1,r.imageSmoothingEnabled=!1;let t=e.target,i=new FileReader;const a=e.target.files[0]&&e.target.files[0].name&&e.target.files[0].name.toLowerCase().endsWith(".heic");i.onload=a?e=>{var t;(t=e.target.result,new Promise(((e,i)=>{const a=window["magick-wasm"];a?a.initializeImageMagick().then((()=>{a.ImageMagick.read(new Uint8Array(t),(t=>{const{width:i,height:a}=W(t.width,t.height);n.width=i,n.height=a,t.resize(i,a),t.writeToCanvas(n);const o=r.getImageData(0,0,i,a);e({data:o,width:i,height:a})}))})):i({error:"HEIC image processing failed. Please make sure Image Magick scripts were integrated as expected.",code:AcuantJavascriptWebSdk.HEIC_NOT_SUPPORTED_CODE})}))).then((e=>{U=6,v.onCaptured(e),q(e.data,e.width,e.height,!1,"MANUAL",v.onCropped)})).catch((e=>v.onError(e.error,e.code)))}:e=>{U=function(e){const t=new DataView(e.target.result);if(65496!=t.getUint16(0,!1))return-2;const i=t.byteLength;let a=2;for(;a<i;){if(t.getUint16(a+2,!1)<=8)return-1;const e=t.getUint16(a,!1);if(a+=2,65505==e){if(1165519206!=t.getUint32(a+=2,!1))return-1;const e=18761==t.getUint16(a+=6,!1);a+=t.getUint32(a+4,e);const i=t.getUint16(a,e);a+=2;for(let n=0;n<i;n++)if(274==t.getUint16(a+12*n,e))return t.getUint16(a+12*n+8,e)}else{if(65280!=(65280&e))break;a+=t.getUint16(a,!1)}}return-1}(e);let t=document.createElement("img");t.onload=()=>{const{width:e,height:i}=W(t.width,t.height);n.width=e,n.height=i,r.drawImage(t,0,0,e,i);const a=r.getImageData(0,0,e,i);r.clearRect(0,0,e,i),t.remove(),v.onCaptured({data:a,width:e,height:i}),q(a,e,i,!1,"MANUAL",v.onCropped)},t.src="data:image/jpeg;base64,"+oe(e.target.result)},t&&t.files[0]&&i.readAsArrayBuffer(t.files[0])}function W(e,t){let i=2560,a=1920;L()&&(i=h,a=Math.floor(3*h/4));if((e>t?e:t)>i)if(e<t){a=i,i=a/(t/e)}else{a=i/(e/t)}else i=e,a=t;return{width:i,height:a}}function N(e){e.getTracks().forEach((e=>{e.stop()}))}function P(){w=!1,y=!1,U=-1,T&&(clearTimeout(T),T=null),function(){b()&&document.removeEventListener("visibilitychange",J);window.removeEventListener("resize",B),e&&(e.removeEventListener("play",V),e.removeEventListener("loadedmetadata",j))}(),e&&(e.pause(),e.srcObject&&N(e.srcObject),e=null),m&&(m.innerHTML=""),a&&(a.remove(),a=null)}function R(){if(/iP(hone|od|ad)/.test(navigator.platform))try{const e=navigator.appVersion.match(/OS (\d+)_(\d+)_?(\d+)?/);return[parseInt(e[1],10),parseInt(e[2],10),parseInt(e[3]||0,10)]}catch(e){return-1}return-1}function L(){let e=R();return e&&-1!=e&&e.length>=1&&15==e[0]}function H(){let e=R();return e&&-1!=e&&e.length>=1&&e[0]>=17}function z(){let e=decodeURIComponent(document.cookie);if(e.includes("AcuantForceRegularCapture=true"))return!1;if(e.includes("AcuantForceDistantCapture=true"))return!0;if(function(){let e=R();return e&&-1!=e&&e.length>=1&&16==e[0]&&e[1]<4}()){let e=[screen.width,screen.height],t=Math.max(...e),i=Math.min(...e);if(852==t&&393==i)return!0;if(932==t&&430==i)return!0;if(844==t&&390==i)return!0;if(926==t&&428==i)return!0}return!1}function B(){i.clearRect(0,0,t.width,t.height),e&&(E()&&(function(){const e=window.navigator.userAgent,t=e.indexOf("WebKit")>-1,i=e.indexOf("CriOS")>-1;return t&&i&&E()}()||function(){let e=R();return e&&-1!=e&&e.length>=2&&14==e[0]&&e[1]>=4}())?(P(),D()):F())}function F(){let i,a;if(e.videoWidth<e.videoHeight?x.canvasScale=e.videoHeight/e.videoWidth:x.canvasScale=e.videoWidth/e.videoHeight,i=window.innerWidth,a=window.innerHeight,x.fixedHeight||(x.fixedHeight=m.clientHeight),x.fixedWidth||(x.fixedWidth=m.clientWidth),x.fixedHeight&&x.fixedWidth&&x.fixedWidth>x.fixedHeight&&(x.fixedHeight=m.clientWidth,x.fixedWidth=m.clientHeight),window.matchMedia("(orientation: portrait)").matches){if(x.fixedWidth)x.fixedWidth<i&&(i=x.fixedWidth),a=i*x.canvasScale;else if(x.fixedHeight){let e=x.fixedHeight/x.canvasScale;e<i&&(a=x.fixedHeight,i=e)}}else if(x.fixedWidth)x.fixedWidth<a&&(a=x.fixedWidth),i=a*x.canvasScale;else if(x.fixedHeight){let e=x.fixedHeight/x.canvasScale;e<a&&(i=x.fixedHeight,a=e)}e.width=t.width=i,e.height=t.height=a}function J(){!a&&w&&"visible"===document.visibilityState&&(P(),D())}function G(){e&&!e.paused&&(te(),T=setTimeout(G,100))}function j(){e.videoWidth+e.videoHeight<1e3?(P(),S("Camera not supported",AcuantJavascriptWebSdk.START_FAIL_CODE)):(F(),G())}function V(){T&&(clearTimeout(T),y=!1,G())}function q(e,t,i,a,n,r){AcuantJavascriptWebSdk.singleWorkerModel?async function(e,t,i,a,n){let r={isPortraitOrientation:n};await Q();const o=await X(e,t,i);if(!o)return null;r={...r,...o},K(),await new Promise((e=>{AcuantJavascriptWebSdk.startMetricsWorker(e)}));const c=await Y(e,t,i),d=await Z(r.image);r={...r,...c,...d},AcuantJavascriptWebSdk.endMetricsWorker();const{imageBase64:s,imageBytes:l}=ie(r,a);r.image.bytes=l;const u=await ee(s);r.image.barcodeText=u;const h=await ae(r,a,s);await Q();const m=await $(h);return r.image.data=m,K(),r}(e,t,i,n,a).then(r):async function(e,t,i,a,n){let r={isPortraitOrientation:n};const[o,c]=await Promise.all([X(e,t,i),Y(e,t,i)]);if(!o)return null;r={...r,...o};const d=await Z(r.image);r={...r,...c,...d};const{imageBase64:s,imageBytes:l}=ie(r,a);r.image.bytes=l;const u=await ee(s);r.image.barcodeText=u;const h=await ae(r,a,s),m=await $(h);return r.image.data=m,r}(e,t,i,n,a).then(r)}function Q(){return new Promise((e=>{AcuantJavascriptWebSdk.startImageWorker(e)}))}function K(){AcuantJavascriptWebSdk.endImageWorker()}function X(e,t,i){return new Promise((a=>{AcuantJavascriptWebSdk.crop(e,t,i,{onSuccess:({image:e,dpi:t,cardType:i})=>a({image:e,dpi:t,cardType:i}),onFail:a})}))}function Y(e,t,i){return new Promise((a=>{AcuantJavascriptWebSdk.moire(e,t,i,{onSuccess:(e,t)=>a({moire:e,moireraw:t}),onFail:()=>a({moire:-1,moireraw:-1})})}))}function Z(e){return new Promise((t=>{AcuantJavascriptWebSdk.metrics(e,e.width,e.height,{onSuccess:(e,i)=>t({sharpness:e,glare:i}),onFail:()=>t({sharpness:-1,glare:-1})})}))}function $(e){return new Promise((t=>{const i=function(e){const t=window.atob(e.split("base64,")[1]),i=t.length,a=new Uint8Array(new ArrayBuffer(i));for(let e=0;e<i;e++)a[e]=t.charCodeAt(e);return a}(e);AcuantJavascriptWebSdk.sign(i,{onSuccess:e=>t("data:image/jpeg;base64,"+oe(e)),onFail:t})}))}async function ee(e){if(!document.getElementById(AcuantJavascriptWebSdk.BARCODE_READER_ID))return null;try{return await function(e){let t=e.split(","),i=t[0].match(/:(.*?);/)[1],a=atob(t[1]),n=a.length,r=new Uint8Array(n);for(;n--;)r[n]=a.charCodeAt(n);const o=new File([r],"imageFile",{type:i});return new Html5Qrcode(AcuantJavascriptWebSdk.BARCODE_READER_ID,{formatsToSupport:[Html5QrcodeSupportedFormats.PDF_417]}).scanFile(o,!1)}(e)}catch{return null}}function te(){if(!w||y)return;if(0==e.videoWidth)return void re();y=!0;let t=Math.max(e.videoWidth,e.videoHeight),i=Math.min(e.videoWidth,e.videoHeight),a=0,o=0;if(t>l&&i>u?e.videoWidth>=e.videoHeight?(x.frameScale=l/e.videoWidth,o=l,a=e.videoHeight*x.frameScale):(x.frameScale=l/e.videoHeight,o=e.videoWidth*x.frameScale,a=l):(x.frameScale=1,o=e.videoWidth,a=e.videoHeight),o==n.width&&a==n.height||(n.width=o,n.height=a),w){let t;try{r.drawImage(e,0,0,e.videoWidth,e.videoHeight,0,0,n.width,n.height),t=r.getImageData(0,0,n.width,n.height),r.clearRect(0,0,n.width,n.height)}catch(e){return void re()}!function(t,i,a){AcuantJavascriptWebSdk.detect(t,i,a,{onSuccess:function(t){if(!n||!e||e.paused||e.ended)return;t.points.forEach((t=>{void 0!==t.x&&void 0!==t.y&&(t.x=t.x/x.frameScale*e.width/e.videoWidth,t.y=t.y/x.frameScale*e.height/e.videoHeight)}));const i=Math.min(t.dimensions.width,t.dimensions.height)/Math.min(n.width,n.height),a=Math.max(t.dimensions.width,t.dimensions.height)/Math.max(n.width,n.height),r=2==t.type;let o=.8,c=.85,l=.6,u=.65;r&&(o=.9,c=.95),E()&&(l=.65,u=.7,z()?r?(o=.72,c=.77,l=.22,u=.28):(o=.41,c=.45,l=.22,u=.28):r&&(o=.95,c=1,l=.7,u=.75));const h=!t.isCorrectAspectRatio||i<l&&a<u,m=i>=o||a>=c;t.type===s.NONE?t.state=d.NO_DOCUMENT:t.state=m?d.BIG_DOCUMENT:h?d.SMALL_DOCUMENT:d.GOOD_DOCUMENT,f(t),y=!1},onFail:function(){if(!n||!e||e.paused||e.ended)return;let t={};t.state=d.NO_DOCUMENT,f(t),y=!1}})}(t,n.width,n.height)}}function ie({image:e,cardType:t,isPortraitOrientation:i},a){n&&r||(n=document.createElement("canvas"),r=n.getContext("2d")),n.width=e.width,n.height=e.height;let o=r.createImageData(e.width,e.height);!function(e,t){for(let i=0;i<e.length;++i)e[i]=t[i]}(o.data,e.data),r.putImageData(o,0,0);let c="AUTO"!=a&&"TAP"!=a;2!==t&&(!c&&i||c&&6===U)&&(r.rotate(Math.PI),r.drawImage(n,0,0,-n.width,-n.height));const d=r.getImageData(0,0,n.width,n.height).data,s="undefined"!=typeof acuantConfig&&"number"==typeof acuantConfig.jpegQuality?acuantConfig.jpegQuality:1,l=n.toDataURL("image/jpeg",s);return r.clearRect(0,0,n.width,n.height),n&&(r=null,n.remove(),n=null),{imageBase64:l,imageBytes:d}}async function ae({dpi:e,cardType:t,sharpness:i,glare:a,moire:n,moireraw:r},o,c){const d=await new Promise((e=>{AcuantJavascriptWebSdk.getCvmlVersion({onSuccess:t=>{e(t)},onFail:()=>{e("unknown")}})})),s=JSON.stringify({cvml:{cropping:{iscropped:!0,dpi:e,idsize:2===t?"ID3":"ID1",elapsed:-1},sharpness:{normalized:i,elapsed:-1},moire:{normalized:n,raw:r,elapsed:-1},glare:{normalized:a,elapsed:-1},version:d},device:{version:ne(),capturetype:o}});return AcuantJavascriptWebSdk.addMetadata(c,{imageDescription:s,dateTimeOriginal:(new Date).toUTCString()})}function ne(){const e=navigator.userAgent;let t,i=e.match(/(opera|chrome|safari|firefox|msie|trident(?=\/))\/?\s*(\d+)/i)||[];return/trident/i.test(i[1])?(t=/\brv[ :]+(\d+)/g.exec(e)||[],"IE "+(t[1]||"")):"Chrome"===i[1]&&(t=e.match(/\b(OPR|Edge)\/(\d+)/),null!=t)?t.slice(1).join(" ").replace("OPR","Opera"):(i=i[2]?[i[1],i[2]]:[navigator.appName,navigator.appVersion,"-?"],null!=(t=e.match(/version\/(\d+)/i))&&i.splice(1,1,t[1]),i.join(" "))}function re(){L()||navigator.maxTouchPoints&&navigator.maxTouchPoints>=2&&/MacIntel/.test(navigator.platform)?S(o,AcuantJavascriptWebSdk.SEQUENCE_BREAK_CODE):S(c,AcuantJavascriptWebSdk.SEQUENCE_BREAK_CODE)}function oe(e){let t="";const i=new Uint8Array(e),a=i.byteLength;for(let e=0;e<a;e++)t+=String.fromCharCode(i[e]);return window.btoa(t)}return C})();
\ No newline at end of file
diff --git a/public/acuant/11.9.2/AcuantImageService.min.js b/public/acuant/11.9.3.508.1/AcuantImageService.min.js
similarity index 100%
rename from public/acuant/11.9.2/AcuantImageService.min.js
rename to public/acuant/11.9.3.508.1/AcuantImageService.min.js
diff --git a/public/acuant/11.9.2/AcuantImageService.wasm b/public/acuant/11.9.3.508.1/AcuantImageService.wasm
old mode 100755
new mode 100644
similarity index 100%
rename from public/acuant/11.9.2/AcuantImageService.wasm
rename to public/acuant/11.9.3.508.1/AcuantImageService.wasm
diff --git a/public/acuant/11.9.2/AcuantImageWorker.min.js b/public/acuant/11.9.3.508.1/AcuantImageWorker.min.js
similarity index 100%
rename from public/acuant/11.9.2/AcuantImageWorker.min.js
rename to public/acuant/11.9.3.508.1/AcuantImageWorker.min.js
diff --git a/public/acuant/11.9.2/AcuantInitializerService.min.js b/public/acuant/11.9.3.508.1/AcuantInitializerService.min.js
similarity index 100%
rename from public/acuant/11.9.2/AcuantInitializerService.min.js
rename to public/acuant/11.9.3.508.1/AcuantInitializerService.min.js
diff --git a/public/acuant/11.9.2/AcuantInitializerService.wasm b/public/acuant/11.9.3.508.1/AcuantInitializerService.wasm
old mode 100755
new mode 100644
similarity index 100%
rename from public/acuant/11.9.2/AcuantInitializerService.wasm
rename to public/acuant/11.9.3.508.1/AcuantInitializerService.wasm
diff --git a/public/acuant/11.9.2/AcuantInitializerWorker.min.js b/public/acuant/11.9.3.508.1/AcuantInitializerWorker.min.js
similarity index 100%
rename from public/acuant/11.9.2/AcuantInitializerWorker.min.js
rename to public/acuant/11.9.3.508.1/AcuantInitializerWorker.min.js
diff --git a/public/acuant/11.9.2/AcuantJavascriptWebSdk.min.js b/public/acuant/11.9.3.508.1/AcuantJavascriptWebSdk.min.js
similarity index 99%
rename from public/acuant/11.9.2/AcuantJavascriptWebSdk.min.js
rename to public/acuant/11.9.3.508.1/AcuantJavascriptWebSdk.min.js
index 74f4a92d413..59d36e77e39 100644
--- a/public/acuant/11.9.2/AcuantJavascriptWebSdk.min.js
+++ b/public/acuant/11.9.3.508.1/AcuantJavascriptWebSdk.min.js
@@ -1 +1 @@
-var AcuantConfig=function(){"use strict";return{acuantVersion:"11.9.2"}}();let config={};"undefined"!=typeof acuantConfig&&0!==Object.keys(acuantConfig).length&&acuantConfig.constructor===Object&&(config=acuantConfig),document.addEventListener("DOMContentLoaded",(function(){void 0===AcuantJavascriptWebSdk&&loadAcuantSdk(),document.removeEventListener("DOMContentLoaded",this)}));var AcuantJavascriptWebSdk=void 0;function loadAcuantSdk(){AcuantJavascriptWebSdk=function(e){let t={ACUANT_IMAGE_WORKER:"AcuantImageWorker",ACUANT_METRICS_WORKER:"AcuantMetricsWorker",SEQUENCE_BREAK_CODE:"sequence-break",START_FAIL_CODE:"start-fail",REPEAT_FAIL_CODE:"repeat-fail",HEIC_NOT_SUPPORTED_CODE:"heic-not-supported",BARCODE_READER_ID:"acuant-barcode-reader",singleWorkerModel:!1,startInitializer:function(n,a=0){if(!n)return void M("startInitializer did not have a user callback set");if(y)return;L=1==a,T(i,n);let r=null;e&&e.cdnPath&&e.cdnPath.initializerUrl?r=e.cdnPath.initializerUrl:e.path&&(r=e.path),y=new Worker(O(r,"AcuantInitializerWorker.min.js",a)),y.onmessage=w,function(){if(document.getElementById(t.BARCODE_READER_ID))return;const e=document.createElement("div");e.id=t.BARCODE_READER_ID,e.style.display="none",document.body.appendChild(e)}()},endInitializer:function(){y&&(y.terminate(),y.onmessage=null,g=!1,y=null)},startImageWorker:function(e){e?S?e():(T(a,e),I(L?1:0)):M("startImageWorker did not have a user callback set")},startMetricsWorker:function(e){e?R?e():(T(r,e),b(L?1:0)):M("startMetricsWorker did not have a user callback set")},endImageWorker:function(){S.terminate(),S.onmessage=null,P=!1,S=null},endMetricsWorker:function(){R.terminate(),R.onmessage=null,C=!1,R=null},start:function(e,t=!1){if(!e)return void M("start did not have a user callback set");const i=L?1:0;this.singleWorkerModel=t,T(n,e),this.singleWorkerModel?S||I(i):(S||I(i),R||b(i))},end:function(){S&&this.endImageWorker(),R&&this.endMetricsWorker()},startWorkers:function(e,t=[this.ACUANT_IMAGE_WORKER,this.ACUANT_METRICS_WORKER],i=0){e?(T(n,e),t.includes(this.ACUANT_IMAGE_WORKER)&&!S&&I(i),t.includes(this.ACUANT_METRICS_WORKER)&&!R&&b(i)):M("startWorkers did not have a user callback set")},endWorkers:function(e=[this.ACUANT_IMAGE_WORKER,this.ACUANT_METRICS_WORKER]){e.includes(this.ACUANT_IMAGE_WORKER)&&S&&(S.terminate(),S.onmessage=null,P=!1,S=null),e.includes(this.ACUANT_METRICS_WORKER)&&R&&(R.terminate(),R.onmessage=null,C=!1,R=null)},initialize:function(e,t,i,n=0){i?(T(o,i),y?B(y,"initialize",{creds:e,endpoint:t}):this.startInitializer((()=>{B(y,"initialize",{creds:e,endpoint:t})}),n)):M("initialize did not have a user callback set")},initializeWithToken:function(e,t,i,n=0){i?(T(o,i),y?B(y,"initializeWithToken",{token:e,endpoint:t}):this.startInitializer((()=>{B(y,"initializeWithToken",{token:e,endpoint:t})}),n)):M("initializeWithToken did not have a user callback set")},crop:function(e,t,i,n){h?n?P&&null!=e?(T(l,n),B(S,"crop",{imgData:e.data,width:t,height:i})):n.onFail():M("crop did not have a user callback set"):M("SDK was not initialized")},detect:function(e,t,i,n){h?n?P&&null!=e?(T(s,n),B(S,"detect",{imgData:e.data,width:t,height:i})):n.onFail():M("detect did not have a user callback set"):M("SDK was not initialized")},metrics:function(e,t,i,n){h?n?C&&null!=e?(T(m,n),B(R,"metrics",{imgData:e.data,width:t,height:i})):n.onFail():M("metrics did not have a user callback set"):M("SDK was not initialized")},moire:function(e,t,i,n){h?n?C&&null!=e?(T(c,n),B(R,"moire",{imgData:e.data,width:t,height:i})):n.onFail():M("moire did not have a user callback set"):M("SDK was not initialized")},sign:function(e,t){h?t?P&&e?(T(p,t),B(S,"sign",{imgData:e})):t.onFail():M("sign did not have a user callback set"):M("SDK was not initialized")},verify:function(e,t){h?t?P&&e?(T(u,t),B(S,"verify",{imgData:e})):t.onFail():M("verify did not have a user callback set"):M("SDK was not initialized")},getCvmlVersion:function(e){h?e?P?(T(d,e),B(S,d)):e.onFail():M("verify did not have a user callback set"):M("SDK was not initialized")},addMetadata:function(e,{make:t=navigator.platform,model:i=navigator.userAgent,software:n="Acuant JavascriptWeb SDK "+AcuantConfig.acuantVersion,imageDescription:a=null,dateTimeOriginal:r,userComment:o="=".repeat(100)}){if(!h)return void M("SDK was not initialized");let l={},s={};l[piexif.ImageIFD.Make]=t,l[piexif.ImageIFD.Model]=i,l[piexif.ImageIFD.Software]=n,a&&(l[piexif.ImageIFD.ImageDescription]=a),s[piexif.ExifIFD.DateTimeOriginal]=r,s[piexif.ExifIFD.UserComment]=o;let m={"0th":l,Exif:s},c=piexif.dump(m);return piexif.insert(c,e)},setUnexpectedErrorCallback:function(e){T(f,e)}};const i="initStart",n="workersStart",a="imageWorkerStart",r="metricsWorkerStart",o="init",l="crop",s="detect",m="metrics",c="moire",p="sign",u="verify",d="getCvmlVersion",f="unexpectedError";let h=!1,y=null,g=!1,S=null,P=!1,R=null,C=!1,A=0,k={},D={},L=!1;function I(t=0){let i=null;e&&e.cdnPath&&e.cdnPath.imageUrl?i=e.cdnPath.imageUrl:e.path&&(i=e.path),A++,S=new Worker(O(i,"AcuantImageWorker.min.js",t)),S.onmessage=E,S.onerror=function(){M("imageWorker has failed")}}function b(t=0){let i=null;e&&e.cdnPath&&e.cdnPath.metricsUrl?i=e.cdnPath.metricsUrl:e.path&&(i=e.path),A++,R=new Worker(O(i,"AcuantMetricsWorker.min.js",t)),R.onmessage=x,R.onerror=function(){M("metricsWorker has failed")}}function w(e){if(h=!1,e){let n=e.data;if(g)if(n&&"initialize"===n.func){let e=n.status,i=k[o];t.endInitializer(),i?1==e?(h=!0,i.onSuccess()):i.onFail(e,function(e){switch(e){case 401:return"Server returned a 401 (missing credentials).";case 403:return"Server returned a 403 (invalid credentials).";case 400:return"Server returned a 400.";case 2:return"Token Validation Failed (Recieved token, but token was null/corrupt).";case 3:return"Token Validation Failed (Recieved token, but token was missing part of body).";case 4:return"Token Validation Failed (Recieved token, but token body was missing fields).";case 5:return"Token Validation Failed (Recieved token, but token body failed validation).";case 6:return"At least one param was null/invalid.";case 7:return"Incorrectly formatted message to worker.";default:return"Unexpected error code."}}(e)):M("initialize did not have a user callback set")}else M("initworker sent message without correct function tagging");else{g=!0;let e=k[i];e&&e()}}else M("initworker sent message without anything in the body")}function E(e){if(e){let t=e.data;if(P)if(t&&"detect"===t.func){const e=k[s];e?t.type&&t.x1&&t.y1&&t.x2&&t.y2&&t.x3&&t.y3&&t.x4&&t.y4?function(e,t,i,n,a,r,o,l,s,m){if(m)if(-1==e)m.onFail();else{let c=function(e,t,i,n,a,r,o,l){let s={x:e,y:t},m={x:i,y:n},c={x:a,y:r},p={x:o,y:l},u=G(s,m),d=G(m,c),f=G(c,p),h=G(p,s),y=(u+f)/2,g=(d+h)/2;return y>g?{width:y,height:g}:{width:g,height:y}}(t,i,n,a,r,o,l,s),p=function(e,t){let i=!1,n=5,a=1.42,r=1.5887;if(2==t){let t=(100+n)/100*a;e>=(100-n)/100*a&&e<=t&&(i=!0)}else if(1==t){let t=(100+n)/100*r;e>=(100-n)/100*r&&e<=t&&(i=!0)}return i}(c.width/c.height,e),u=F(c.width,c.height,2==e),d=function(e){let t=[-1,-1,-1,-1];e&&4===e.length&&(v(t,e[0],e[2]),v(t,e[1],e[3]));return t}([{x:t,y:i},{x:n,y:a},{x:r,y:o},{x:l,y:s}]);m.onSuccess({type:e,dimensions:c,dpi:u,isCorrectAspectRatio:p,points:d})}}(t.type,t.x1,t.y1,t.x2,t.y2,t.x3,t.y3,t.x4,t.y4,e):e.onFail():M("detect did not have a user callback set")}else if(t&&"crop"===t.func){const e=k[l];e?t.imgData&&t.width&&t.height&&t.type?function(e,t,i,n,a){a&&(null!=e&&t>=0&&i>=0&&n>=0?(D={image:{data:e,width:t,height:i},cardType:n,dpi:F(t,i,2==n)},a.onSuccess(D)):a.onFail())}(t.imgData,t.width,t.height,t.type,e):t.error?e.onFail(t.error):e.onFail():M("crop did not have a user callback set")}else if(t&&"sign"===t.func){const e=k[p];e?t.imgData?function(e,t){t&&(e?t.onSuccess(e):t.onFail())}(t.imgData,e):t.error?e.onFail(t.error):e.onFail():M("sign did not have a user callback set")}else if(t&&"verify"===t.func){const e=k[u];e?t.result||!1===t.result?function(e,t){t&&(e||!1===e?t.onSuccess(e):t.onFail())}(t.result,e):e.onFail():M("verify did not have a user callback set")}else if(t&&t.func===d){let e=k[d];e?function(e,t){e?t.onSuccess(e):t.onFail()}(t.cvmlVersion,e):M("getCvmlVersion did not have a user callback set")}else M("imageworker sent message without correct function tagging");else P=!0,_()}else M("imageworker sent message without anything in the body")}function x(e){if(e){let t=e.data;if(C)if(t&&"metrics"===t.func){const e=k[m];e?t.sharpness&&t.glare?function(e,t,i){if(i)if(t>=0&&e>=0){let n=Math.floor(100*e),a=Math.floor(100*t);i.onSuccess(n,a)}else i.onFail()}(t.sharpness,t.glare,e):t.error?e.onFail(t.error):e.onFail():M("metrics did not have a user callback set")}else if("moire"===t.func){const e=k[c];e?t.moire&&t.moireraw?function(e,t,i){if(i)if(e>=0&&t>=0){let n=Math.floor(100*e),a=Math.floor(100*t);i.onSuccess(n,a)}else i.onFail()}(t.moire,t.moireraw,e):t.error?e.onFail(t.error):e.onFail():M("moire did not have a user callback set")}else M("metricsworker sent message without correct function tagging");else C=!0,_()}else M("metricsworker sent message without anything in the body")}function v(e,t,i){return t.x<i.x&&t.y<i.y?(e[0]=t,e[2]=i):t.x>i.x&&t.y>i.y?(e[0]=i,e[2]=t):t.x>i.x&&t.y<i.y?(e[1]=t,e[3]=i):(e[1]=i,e[3]=t),e}function G(e,t){return Math.sqrt(Math.pow(t.x-e.x,2)+Math.pow(t.y-e.y,2))}function F(e,t,i){let n=e>t?e:t,a=i?4.92:3.37;return Math.round(n/a)}function T(e,t){k[e]=t}function M(e){let t=k[f];t?e?t(e):t():console.error("Error: ",e)}function O(e,t,i){let n;return null!=e&&e.length>0&&0==i?(n="/"===e.charAt(e.length-1)?e:e+"/",n+=t):n=0!=i?e:t,n}function B(e,t,i,n=!1){let a={func:t,data:i};n&&i&&i.imgData&&i.imgData.buffer?e.postMessage(a,[a.data.imgData.buffer]):e.postMessage(a)}function _(){const e=k[n],t=k[a],i=k[r];--A,0==A&&(e?(T(n,null),e()):t?(T(a,null),t()):i&&(T(r,null),i()))}return t}(config),"function"==typeof onAcuantSdkLoaded&&onAcuantSdkLoaded()}!function(){"use strict";let e={};function t(e){return m(">"+p("B",e.length),e)}function i(e){return m(">"+p("H",e.length),e)}function n(e){return m(">"+p("L",e.length),e)}function a(e,a,r){let o,l,s,c,u="",d="";if("Byte"==a)o=e.length,o<=4?d=t(e)+p("\0",4-o):(d=m(">L",[r]),u=t(e));else if("Short"==a)o=e.length,o<=2?d=i(e)+p("\0\0",2-o):(d=m(">L",[r]),u=i(e));else if("Long"==a)o=e.length,o<=1?d=n(e):(d=m(">L",[r]),u=n(e));else if("Ascii"==a)l=e+"\0",o=l.length,o>4?(d=m(">L",[r]),u=l):d=l+p("\0",4-o);else if("Rational"==a){if("number"==typeof e[0])o=1,s=e[0],c=e[1],l=m(">L",[s])+m(">L",[c]);else{o=e.length,l="";for(var f=0;f<o;f++)s=e[f][0],c=e[f][1],l+=m(">L",[s])+m(">L",[c])}d=m(">L",[r]),u=l}else if("SRational"==a){if("number"==typeof e[0])o=1,s=e[0],c=e[1],l=m(">l",[s])+m(">l",[c]);else{o=e.length,l="";for(f=0;f<o;f++)s=e[f][0],c=e[f][1],l+=m(">l",[s])+m(">l",[c])}d=m(">L",[r]),u=l}else"Undefined"==a&&(o=e.length,o>4?(d=m(">L",[r]),u=e):d=e+p("\0",4-o));return[m(">L",[o]),d,u]}function r(e,t,i){let n,r=Object.keys(e).length,o=m(">H",[r]);n=["0th","1st"].indexOf(t)>-1?2+12*r+4:2+12*r;let l="",s="";for(var c in e){if("string"==typeof c&&(c=parseInt(c)),"0th"==t&&[34665,34853].indexOf(c)>-1)continue;if("Exif"==t&&40965==c)continue;if("1st"==t&&[513,514].indexOf(c)>-1)continue;let r=e[c],o=m(">H",[c]),p=f[t][c].type,u=m(">H",[d[p]]);"number"==typeof r&&(r=[r]);let h=a(r,p,8+n+i+s.length);l+=o+u+h[0]+h[1],s+=h[2]}return[o+l,s]}function o(e){let t,i;if("ÿØ"==e.slice(0,2))t=u(e),i=function(e){let t;for(let i=0;i<e.length;i++)if(t=e[i],"ÿá"==t.slice(0,2)&&"Exif\0\0"==t.slice(4,10))return t;return null}(t),this.tiftag=i?i.slice(10):null;else if(["II","MM"].indexOf(e.slice(0,2))>-1)this.tiftag=e;else{if("Exif"!=e.slice(0,4))throw new Error("Given file is neither JPEG nor TIFF.");this.tiftag=e.slice(6)}}if(e.version="1.0.4",e.remove=function(e){let t=!1;if("ÿØ"==e.slice(0,2));else{if("data:image/jpeg;base64,"!=e.slice(0,23)&&"data:image/jpg;base64,"!=e.slice(0,22))throw new Error("Given data is not jpeg.");e=s(e.split(",")[1]),t=!0}let i=u(e).filter((function(e){return!("ÿá"==e.slice(0,2)&&"Exif\0\0"==e.slice(4,10))})).join("");return t&&(i="data:image/jpeg;base64,"+l(i)),i},e.insert=function(e,t){let i=!1;if("Exif\0\0"!=e.slice(0,6))throw new Error("Given data is not exif.");if("ÿØ"==t.slice(0,2));else{if("data:image/jpeg;base64,"!=t.slice(0,23)&&"data:image/jpg;base64,"!=t.slice(0,22))throw new Error("Given data is not jpeg.");t=s(t.split(",")[1]),i=!0}let n="ÿá"+m(">H",[e.length+2])+e,a=function(e,t){let i=!1,n=[];e.forEach((function(a,r){"ÿá"==a.slice(0,2)&&"Exif\0\0"==a.slice(4,10)&&(i?n.unshift(r):(e[r]=t,i=!0))})),n.forEach((function(t){e.splice(t,1)})),!i&&t&&(e=[e[0],t].concat(e.slice(1)));return e.join("")}(u(t),n);return i&&(a="data:image/jpeg;base64,"+l(a)),a},e.load=function(e){let t;if("string"!=typeof e)throw new Error("'load' gots invalid type argument.");if("ÿØ"==e.slice(0,2))t=e;else if("data:image/jpeg;base64,"==e.slice(0,23)||"data:image/jpg;base64,"==e.slice(0,22))t=s(e.split(",")[1]);else{if("Exif"!=e.slice(0,4))throw new Error("'load' gots invalid file data.");t=e.slice(6)}let i={"0th":{},Exif:{},GPS:{},Interop:{},"1st":{},thumbnail:null},n=new o(t);if(null===n.tiftag)return i;"II"==n.tiftag.slice(0,2)?n.endian_mark="<":n.endian_mark=">";let a=c(n.endian_mark+"L",n.tiftag.slice(4,8))[0];i["0th"]=n.get_ifd(a,"0th");let r=i["0th"].first_ifd_pointer;if(delete i["0th"].first_ifd_pointer,34665 in i["0th"]&&(a=i["0th"][34665],i.Exif=n.get_ifd(a,"Exif")),34853 in i["0th"]&&(a=i["0th"][34853],i.GPS=n.get_ifd(a,"GPS")),40965 in i.Exif&&(a=i.Exif[40965],i.Interop=n.get_ifd(a,"Interop")),"\0\0\0\0"!=r&&(a=c(n.endian_mark+"L",r)[0],i["1st"]=n.get_ifd(a,"1st"),513 in i["1st"]&&514 in i["1st"])){let e=i["1st"][513]+i["1st"][514],t=n.tiftag.slice(i["1st"][513],e);i.thumbnail=t}return i},e.dump=function(t){let i=(n=t,JSON.parse(JSON.stringify(n)));var n;let a,o,l,s,c,p=!1,f=!1,h=!1,y=!1;a="0th"in i?i["0th"]:{},"Exif"in i&&Object.keys(i.Exif).length||"Interop"in i&&Object.keys(i.Interop).length?(a[34665]=1,p=!0,o=i.Exif,"Interop"in i&&Object.keys(i.Interop).length?(o[40965]=1,h=!0,l=i.Interop):Object.keys(o).indexOf(e.ExifIFD.InteroperabilityTag.toString())>-1&&delete o[40965]):Object.keys(a).indexOf(e.ImageIFD.ExifTag.toString())>-1&&delete a[34665],"GPS"in i&&Object.keys(i.GPS).length?(a[e.ImageIFD.GPSTag]=1,f=!0,s=i.GPS):Object.keys(a).indexOf(e.ImageIFD.GPSTag.toString())>-1&&delete a[e.ImageIFD.GPSTag],"1st"in i&&"thumbnail"in i&&null!=i.thumbnail&&(y=!0,i["1st"][513]=1,i["1st"][514]=1,c=i["1st"]);let g,S,P,R,C,A=r(a,"0th",0),k=A[0].length+12*p+12*f+4+A[1].length,D="",L=0,I="",b=0,w="",E=0,x="";(p&&(g=r(o,"Exif",k),L=g[0].length+12*h+g[1].length),f&&(S=r(s,"GPS",k+L),I=S.join(""),b=I.length),h)&&(P=r(l,"Interop",k+L+b),w=P.join(""),E=w.length);if(y&&(R=r(c,"1st",k+L+b+E),C=function(e){let t=u(e);for(;"ÿà"<=t[1].slice(0,2)&&t[1].slice(0,2)<="ÿï";)t=[t[0]].concat(t.slice(2));return t.join("")}(i.thumbnail),C.length>64e3))throw new Error("Given thumbnail is too large. max 64kB");let v="",G="",F="",T="\0\0\0\0";if(p){var M=m(">L",[O=8+k]);v=m(">H",[34665])+m(">H",[d.Long])+m(">L",[1])+M}if(f){M=m(">L",[O=8+k+L]);G=m(">H",[34853])+m(">H",[d.Long])+m(">L",[1])+M}if(h){M=m(">L",[O=8+k+L+b]);F=m(">H",[40965])+m(">H",[d.Long])+m(">L",[1])+M}if(y){var O;T=m(">L",[O=8+k+L+b+E]);let e="
\0\0\0\0
"+m(">L",[O+R[0].length+24+4+R[1].length]),t="\0\0\0\0
"+m(">L",[C.length]);x=R[0]+e+t+"\0\0\0\0"+R[1]+C}let B=A[0]+v+G+T+A[1];return p&&(D=g[0]+F+g[1]),"Exif\0\0MM\0*\0\0\0\b"+B+D+I+w+x},o.prototype={get_ifd:function(e,t){let i,n={},a=c(this.endian_mark+"H",this.tiftag.slice(e,e+2))[0],r=e+2;i=["0th","1st"].indexOf(t)>-1?"Image":t;for(let t=0;t<a;t++){e=r+12*t;let a=c(this.endian_mark+"H",this.tiftag.slice(e,e+2))[0],o=[c(this.endian_mark+"H",this.tiftag.slice(e+2,e+4))[0],c(this.endian_mark+"L",this.tiftag.slice(e+4,e+8))[0],this.tiftag.slice(e+8,e+12)];a in f[i]&&(n[a]=this.convert_value(o))}return"0th"==t&&(e=r+12*a,n.first_ifd_pointer=this.tiftag.slice(e,e+4)),n},convert_value:function(e){let t,i=null,n=e[0],a=e[1],r=e[2];if(1==n)a>4?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("B",a),this.tiftag.slice(t,t+a))):i=c(this.endian_mark+p("B",a),r.slice(0,a));else if(2==n)a>4?(t=c(this.endian_mark+"L",r)[0],i=this.tiftag.slice(t,t+a-1)):i=r.slice(0,a-1);else if(3==n)a>2?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("H",a),this.tiftag.slice(t,t+2*a))):i=c(this.endian_mark+p("H",a),r.slice(0,2*a));else if(4==n)a>1?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("L",a),this.tiftag.slice(t,t+4*a))):i=c(this.endian_mark+p("L",a),r);else if(5==n)if(t=c(this.endian_mark+"L",r)[0],a>1){i=[];for(var o=0;o<a;o++)i.push([c(this.endian_mark+"L",this.tiftag.slice(t+8*o,t+4+8*o))[0],c(this.endian_mark+"L",this.tiftag.slice(t+4+8*o,t+8+8*o))[0]])}else i=[c(this.endian_mark+"L",this.tiftag.slice(t,t+4))[0],c(this.endian_mark+"L",this.tiftag.slice(t+4,t+8))[0]];else if(7==n)a>4?(t=c(this.endian_mark+"L",r)[0],i=this.tiftag.slice(t,t+a)):i=r.slice(0,a);else if(9==n)a>1?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("l",a),this.tiftag.slice(t,t+4*a))):i=c(this.endian_mark+p("l",a),r);else{if(10!=n)throw new Error("Exif might be wrong. Got incorrect value type to decode. type:"+n);if(t=c(this.endian_mark+"L",r)[0],a>1){i=[];for(o=0;o<a;o++)i.push([c(this.endian_mark+"l",this.tiftag.slice(t+8*o,t+4+8*o))[0],c(this.endian_mark+"l",this.tiftag.slice(t+4+8*o,t+8+8*o))[0]])}else i=[c(this.endian_mark+"l",this.tiftag.slice(t,t+4))[0],c(this.endian_mark+"l",this.tiftag.slice(t+4,t+8))[0]]}return i instanceof Array&&1==i.length?i[0]:i}},"undefined"!=typeof window&&"function"==typeof window.btoa)var l=window.btoa;if(void 0===l)l=function(e){let t,i,n,a,r,o,l,s="",m=0,c="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";for(;m<e.length;)t=e.charCodeAt(m++),i=e.charCodeAt(m++),n=e.charCodeAt(m++),a=t>>2,r=(3&t)<<4|i>>4,o=(15&i)<<2|n>>6,l=63&n,isNaN(i)?o=l=64:isNaN(n)&&(l=64),s=s+c.charAt(a)+c.charAt(r)+c.charAt(o)+c.charAt(l);return s};if("undefined"!=typeof window&&"function"==typeof window.atob)var s=window.atob;if(void 0===s)s=function(e){let t,i,n,a,r,o,l,s="",m=0,c="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";for(e=e.replace(/[^A-Za-z0-9\+\/\=]/g,"");m<e.length;)a=c.indexOf(e.charAt(m++)),r=c.indexOf(e.charAt(m++)),o=c.indexOf(e.charAt(m++)),l=c.indexOf(e.charAt(m++)),t=a<<2|r>>4,i=(15&r)<<4|o>>2,n=(3&o)<<6|l,s+=String.fromCharCode(t),64!=o&&(s+=String.fromCharCode(i)),64!=l&&(s+=String.fromCharCode(n));return s};function m(e,t){if(!(t instanceof Array))throw new Error("'pack' error. Got invalid type argument.");if(e.length-1!=t.length)throw new Error("'pack' error. "+(e.length-1)+" marks, "+t.length+" elements.");let i;if("<"==e[0])i=!0;else{if(">"!=e[0])throw new Error("");i=!1}let n="",a=1,r=null,o=null,l=null;for(;o=e[a];){if("b"==o.toLowerCase()){if(r=t[a-1],"b"==o&&r<0&&(r+=256),r>255||r<0)throw new Error("'pack' error.");l=String.fromCharCode(r)}else if("H"==o){if(r=t[a-1],r>65535||r<0)throw new Error("'pack' error.");l=String.fromCharCode(Math.floor(r%65536/256))+String.fromCharCode(r%256),i&&(l=l.split("").reverse().join(""))}else{if("l"!=o.toLowerCase())throw new Error("'pack' error.");if(r=t[a-1],"l"==o&&r<0&&(r+=4294967296),r>4294967295||r<0)throw new Error("'pack' error.");l=String.fromCharCode(Math.floor(r/16777216))+String.fromCharCode(Math.floor(r%16777216/65536))+String.fromCharCode(Math.floor(r%65536/256))+String.fromCharCode(r%256),i&&(l=l.split("").reverse().join(""))}n+=l,a+=1}return n}function c(e,t){if("string"!=typeof t)throw new Error("'unpack' error. Got invalid type argument.");let i,n=0;for(let t=1;t<e.length;t++)if("b"==e[t].toLowerCase())n+=1;else if("h"==e[t].toLowerCase())n+=2;else{if("l"!=e[t].toLowerCase())throw new Error("'unpack' error. Got invalid mark.");n+=4}if(n!=t.length)throw new Error("'unpack' error. Mismatch between symbol and string length. "+n+":"+t.length);if("<"==e[0])i=!0;else{if(">"!=e[0])throw new Error("'unpack' error.");i=!1}let a=[],r=0,o=1,l=null,s=null,m=null,c="";for(;s=e[o];){if("b"==s.toLowerCase())m=1,c=t.slice(r,r+m),l=c.charCodeAt(0),"b"==s&&l>=128&&(l-=256);else if("H"==s)m=2,c=t.slice(r,r+m),i&&(c=c.split("").reverse().join("")),l=256*c.charCodeAt(0)+c.charCodeAt(1);else{if("l"!=s.toLowerCase())throw new Error("'unpack' error. "+s);m=4,c=t.slice(r,r+m),i&&(c=c.split("").reverse().join("")),l=16777216*c.charCodeAt(0)+65536*c.charCodeAt(1)+256*c.charCodeAt(2)+c.charCodeAt(3),"l"==s&&l>=2147483648&&(l-=4294967296)}a.push(l),r+=m,o+=1}return a}function p(e,t){let i="";for(let n=0;n<t;n++)i+=e;return i}function u(e){if("ÿØ"!=e.slice(0,2))throw new Error("Given data isn't JPEG.");let t=2,i=["ÿØ"];for(;;){if("ÿÚ"==e.slice(t,t+2)){i.push(e.slice(t));break}{let n=t+c(">H",e.slice(t+2,t+4))[0]+2;i.push(e.slice(t,n)),t=n}if(t>=e.length)throw new Error("Wrong JPEG data.")}return i}var d={Byte:1,Ascii:2,Short:3,Long:4,Rational:5,Undefined:7,SLong:9,SRational:10},f={Image:{11:{name:"ProcessingSoftware",type:"Ascii"},254:{name:"NewSubfileType",type:"Long"},255:{name:"SubfileType",type:"Short"},256:{name:"ImageWidth",type:"Long"},257:{name:"ImageLength",type:"Long"},258:{name:"BitsPerSample",type:"Short"},259:{name:"Compression",type:"Short"},262:{name:"PhotometricInterpretation",type:"Short"},263:{name:"Threshholding",type:"Short"},264:{name:"CellWidth",type:"Short"},265:{name:"CellLength",type:"Short"},266:{name:"FillOrder",type:"Short"},269:{name:"DocumentName",type:"Ascii"},270:{name:"ImageDescription",type:"Ascii"},271:{name:"Make",type:"Ascii"},272:{name:"Model",type:"Ascii"},273:{name:"StripOffsets",type:"Long"},274:{name:"Orientation",type:"Short"},277:{name:"SamplesPerPixel",type:"Short"},278:{name:"RowsPerStrip",type:"Long"},279:{name:"StripByteCounts",type:"Long"},282:{name:"XResolution",type:"Rational"},283:{name:"YResolution",type:"Rational"},284:{name:"PlanarConfiguration",type:"Short"},290:{name:"GrayResponseUnit",type:"Short"},291:{name:"GrayResponseCurve",type:"Short"},292:{name:"T4Options",type:"Long"},293:{name:"T6Options",type:"Long"},296:{name:"ResolutionUnit",type:"Short"},301:{name:"TransferFunction",type:"Short"},305:{name:"Software",type:"Ascii"},306:{name:"DateTime",type:"Ascii"},315:{name:"Artist",type:"Ascii"},316:{name:"HostComputer",type:"Ascii"},317:{name:"Predictor",type:"Short"},318:{name:"WhitePoint",type:"Rational"},319:{name:"PrimaryChromaticities",type:"Rational"},320:{name:"ColorMap",type:"Short"},321:{name:"HalftoneHints",type:"Short"},322:{name:"TileWidth",type:"Short"},323:{name:"TileLength",type:"Short"},324:{name:"TileOffsets",type:"Short"},325:{name:"TileByteCounts",type:"Short"},330:{name:"SubIFDs",type:"Long"},332:{name:"InkSet",type:"Short"},333:{name:"InkNames",type:"Ascii"},334:{name:"NumberOfInks",type:"Short"},336:{name:"DotRange",type:"Byte"},337:{name:"TargetPrinter",type:"Ascii"},338:{name:"ExtraSamples",type:"Short"},339:{name:"SampleFormat",type:"Short"},340:{name:"SMinSampleValue",type:"Short"},341:{name:"SMaxSampleValue",type:"Short"},342:{name:"TransferRange",type:"Short"},343:{name:"ClipPath",type:"Byte"},344:{name:"XClipPathUnits",type:"Long"},345:{name:"YClipPathUnits",type:"Long"},346:{name:"Indexed",type:"Short"},347:{name:"JPEGTables",type:"Undefined"},351:{name:"OPIProxy",type:"Short"},512:{name:"JPEGProc",type:"Long"},513:{name:"JPEGInterchangeFormat",type:"Long"},514:{name:"JPEGInterchangeFormatLength",type:"Long"},515:{name:"JPEGRestartInterval",type:"Short"},517:{name:"JPEGLosslessPredictors",type:"Short"},518:{name:"JPEGPointTransforms",type:"Short"},519:{name:"JPEGQTables",type:"Long"},520:{name:"JPEGDCTables",type:"Long"},521:{name:"JPEGACTables",type:"Long"},529:{name:"YCbCrCoefficients",type:"Rational"},530:{name:"YCbCrSubSampling",type:"Short"},531:{name:"YCbCrPositioning",type:"Short"},532:{name:"ReferenceBlackWhite",type:"Rational"},700:{name:"XMLPacket",type:"Byte"},18246:{name:"Rating",type:"Short"},18249:{name:"RatingPercent",type:"Short"},32781:{name:"ImageID",type:"Ascii"},33421:{name:"CFARepeatPatternDim",type:"Short"},33422:{name:"CFAPattern",type:"Byte"},33423:{name:"BatteryLevel",type:"Rational"},33432:{name:"Copyright",type:"Ascii"},33434:{name:"ExposureTime",type:"Rational"},34377:{name:"ImageResources",type:"Byte"},34665:{name:"ExifTag",type:"Long"},34675:{name:"InterColorProfile",type:"Undefined"},34853:{name:"GPSTag",type:"Long"},34857:{name:"Interlace",type:"Short"},34858:{name:"TimeZoneOffset",type:"Long"},34859:{name:"SelfTimerMode",type:"Short"},37387:{name:"FlashEnergy",type:"Rational"},37388:{name:"SpatialFrequencyResponse",type:"Undefined"},37389:{name:"Noise",type:"Undefined"},37390:{name:"FocalPlaneXResolution",type:"Rational"},37391:{name:"FocalPlaneYResolution",type:"Rational"},37392:{name:"FocalPlaneResolutionUnit",type:"Short"},37393:{name:"ImageNumber",type:"Long"},37394:{name:"SecurityClassification",type:"Ascii"},37395:{name:"ImageHistory",type:"Ascii"},37397:{name:"ExposureIndex",type:"Rational"},37398:{name:"TIFFEPStandardID",type:"Byte"},37399:{name:"SensingMethod",type:"Short"},40091:{name:"XPTitle",type:"Byte"},40092:{name:"XPComment",type:"Byte"},40093:{name:"XPAuthor",type:"Byte"},40094:{name:"XPKeywords",type:"Byte"},40095:{name:"XPSubject",type:"Byte"},50341:{name:"PrintImageMatching",type:"Undefined"},50706:{name:"DNGVersion",type:"Byte"},50707:{name:"DNGBackwardVersion",type:"Byte"},50708:{name:"UniqueCameraModel",type:"Ascii"},50709:{name:"LocalizedCameraModel",type:"Byte"},50710:{name:"CFAPlaneColor",type:"Byte"},50711:{name:"CFALayout",type:"Short"},50712:{name:"LinearizationTable",type:"Short"},50713:{name:"BlackLevelRepeatDim",type:"Short"},50714:{name:"BlackLevel",type:"Rational"},50715:{name:"BlackLevelDeltaH",type:"SRational"},50716:{name:"BlackLevelDeltaV",type:"SRational"},50717:{name:"WhiteLevel",type:"Short"},50718:{name:"DefaultScale",type:"Rational"},50719:{name:"DefaultCropOrigin",type:"Short"},50720:{name:"DefaultCropSize",type:"Short"},50721:{name:"ColorMatrix1",type:"SRational"},50722:{name:"ColorMatrix2",type:"SRational"},50723:{name:"CameraCalibration1",type:"SRational"},50724:{name:"CameraCalibration2",type:"SRational"},50725:{name:"ReductionMatrix1",type:"SRational"},50726:{name:"ReductionMatrix2",type:"SRational"},50727:{name:"AnalogBalance",type:"Rational"},50728:{name:"AsShotNeutral",type:"Short"},50729:{name:"AsShotWhiteXY",type:"Rational"},50730:{name:"BaselineExposure",type:"SRational"},50731:{name:"BaselineNoise",type:"Rational"},50732:{name:"BaselineSharpness",type:"Rational"},50733:{name:"BayerGreenSplit",type:"Long"},50734:{name:"LinearResponseLimit",type:"Rational"},50735:{name:"CameraSerialNumber",type:"Ascii"},50736:{name:"LensInfo",type:"Rational"},50737:{name:"ChromaBlurRadius",type:"Rational"},50738:{name:"AntiAliasStrength",type:"Rational"},50739:{name:"ShadowScale",type:"SRational"},50740:{name:"DNGPrivateData",type:"Byte"},50741:{name:"MakerNoteSafety",type:"Short"},50778:{name:"CalibrationIlluminant1",type:"Short"},50779:{name:"CalibrationIlluminant2",type:"Short"},50780:{name:"BestQualityScale",type:"Rational"},50781:{name:"RawDataUniqueID",type:"Byte"},50827:{name:"OriginalRawFileName",type:"Byte"},50828:{name:"OriginalRawFileData",type:"Undefined"},50829:{name:"ActiveArea",type:"Short"},50830:{name:"MaskedAreas",type:"Short"},50831:{name:"AsShotICCProfile",type:"Undefined"},50832:{name:"AsShotPreProfileMatrix",type:"SRational"},50833:{name:"CurrentICCProfile",type:"Undefined"},50834:{name:"CurrentPreProfileMatrix",type:"SRational"},50879:{name:"ColorimetricReference",type:"Short"},50931:{name:"CameraCalibrationSignature",type:"Byte"},50932:{name:"ProfileCalibrationSignature",type:"Byte"},50934:{name:"AsShotProfileName",type:"Byte"},50935:{name:"NoiseReductionApplied",type:"Rational"},50936:{name:"ProfileName",type:"Byte"},50937:{name:"ProfileHueSatMapDims",type:"Long"},50938:{name:"ProfileHueSatMapData1",type:"Float"},50939:{name:"ProfileHueSatMapData2",type:"Float"},50940:{name:"ProfileToneCurve",type:"Float"},50941:{name:"ProfileEmbedPolicy",type:"Long"},50942:{name:"ProfileCopyright",type:"Byte"},50964:{name:"ForwardMatrix1",type:"SRational"},50965:{name:"ForwardMatrix2",type:"SRational"},50966:{name:"PreviewApplicationName",type:"Byte"},50967:{name:"PreviewApplicationVersion",type:"Byte"},50968:{name:"PreviewSettingsName",type:"Byte"},50969:{name:"PreviewSettingsDigest",type:"Byte"},50970:{name:"PreviewColorSpace",type:"Long"},50971:{name:"PreviewDateTime",type:"Ascii"},50972:{name:"RawImageDigest",type:"Undefined"},50973:{name:"OriginalRawFileDigest",type:"Undefined"},50974:{name:"SubTileBlockSize",type:"Long"},50975:{name:"RowInterleaveFactor",type:"Long"},50981:{name:"ProfileLookTableDims",type:"Long"},50982:{name:"ProfileLookTableData",type:"Float"},51008:{name:"OpcodeList1",type:"Undefined"},51009:{name:"OpcodeList2",type:"Undefined"},51022:{name:"OpcodeList3",type:"Undefined"}},Exif:{33434:{name:"ExposureTime",type:"Rational"},33437:{name:"FNumber",type:"Rational"},34850:{name:"ExposureProgram",type:"Short"},34852:{name:"SpectralSensitivity",type:"Ascii"},34855:{name:"ISOSpeedRatings",type:"Short"},34856:{name:"OECF",type:"Undefined"},34864:{name:"SensitivityType",type:"Short"},34865:{name:"StandardOutputSensitivity",type:"Long"},34866:{name:"RecommendedExposureIndex",type:"Long"},34867:{name:"ISOSpeed",type:"Long"},34868:{name:"ISOSpeedLatitudeyyy",type:"Long"},34869:{name:"ISOSpeedLatitudezzz",type:"Long"},36864:{name:"ExifVersion",type:"Undefined"},36867:{name:"DateTimeOriginal",type:"Ascii"},36868:{name:"DateTimeDigitized",type:"Ascii"},37121:{name:"ComponentsConfiguration",type:"Undefined"},37122:{name:"CompressedBitsPerPixel",type:"Rational"},37377:{name:"ShutterSpeedValue",type:"SRational"},37378:{name:"ApertureValue",type:"Rational"},37379:{name:"BrightnessValue",type:"SRational"},37380:{name:"ExposureBiasValue",type:"SRational"},37381:{name:"MaxApertureValue",type:"Rational"},37382:{name:"SubjectDistance",type:"Rational"},37383:{name:"MeteringMode",type:"Short"},37384:{name:"LightSource",type:"Short"},37385:{name:"Flash",type:"Short"},37386:{name:"FocalLength",type:"Rational"},37396:{name:"SubjectArea",type:"Short"},37500:{name:"MakerNote",type:"Undefined"},37510:{name:"UserComment",type:"Ascii"},37520:{name:"SubSecTime",type:"Ascii"},37521:{name:"SubSecTimeOriginal",type:"Ascii"},37522:{name:"SubSecTimeDigitized",type:"Ascii"},40960:{name:"FlashpixVersion",type:"Undefined"},40961:{name:"ColorSpace",type:"Short"},40962:{name:"PixelXDimension",type:"Long"},40963:{name:"PixelYDimension",type:"Long"},40964:{name:"RelatedSoundFile",type:"Ascii"},40965:{name:"InteroperabilityTag",type:"Long"},41483:{name:"FlashEnergy",type:"Rational"},41484:{name:"SpatialFrequencyResponse",type:"Undefined"},41486:{name:"FocalPlaneXResolution",type:"Rational"},41487:{name:"FocalPlaneYResolution",type:"Rational"},41488:{name:"FocalPlaneResolutionUnit",type:"Short"},41492:{name:"SubjectLocation",type:"Short"},41493:{name:"ExposureIndex",type:"Rational"},41495:{name:"SensingMethod",type:"Short"},41728:{name:"FileSource",type:"Undefined"},41729:{name:"SceneType",type:"Undefined"},41730:{name:"CFAPattern",type:"Undefined"},41985:{name:"CustomRendered",type:"Short"},41986:{name:"ExposureMode",type:"Short"},41987:{name:"WhiteBalance",type:"Short"},41988:{name:"DigitalZoomRatio",type:"Rational"},41989:{name:"FocalLengthIn35mmFilm",type:"Short"},41990:{name:"SceneCaptureType",type:"Short"},41991:{name:"GainControl",type:"Short"},41992:{name:"Contrast",type:"Short"},41993:{name:"Saturation",type:"Short"},41994:{name:"Sharpness",type:"Short"},41995:{name:"DeviceSettingDescription",type:"Undefined"},41996:{name:"SubjectDistanceRange",type:"Short"},42016:{name:"ImageUniqueID",type:"Ascii"},42032:{name:"CameraOwnerName",type:"Ascii"},42033:{name:"BodySerialNumber",type:"Ascii"},42034:{name:"LensSpecification",type:"Rational"},42035:{name:"LensMake",type:"Ascii"},42036:{name:"LensModel",type:"Ascii"},42037:{name:"LensSerialNumber",type:"Ascii"},42240:{name:"Gamma",type:"Rational"}},GPS:{0:{name:"GPSVersionID",type:"Byte"},1:{name:"GPSLatitudeRef",type:"Ascii"},2:{name:"GPSLatitude",type:"Rational"},3:{name:"GPSLongitudeRef",type:"Ascii"},4:{name:"GPSLongitude",type:"Rational"},5:{name:"GPSAltitudeRef",type:"Byte"},6:{name:"GPSAltitude",type:"Rational"},7:{name:"GPSTimeStamp",type:"Rational"},8:{name:"GPSSatellites",type:"Ascii"},9:{name:"GPSStatus",type:"Ascii"},10:{name:"GPSMeasureMode",type:"Ascii"},11:{name:"GPSDOP",type:"Rational"},12:{name:"GPSSpeedRef",type:"Ascii"},13:{name:"GPSSpeed",type:"Rational"},14:{name:"GPSTrackRef",type:"Ascii"},15:{name:"GPSTrack",type:"Rational"},16:{name:"GPSImgDirectionRef",type:"Ascii"},17:{name:"GPSImgDirection",type:"Rational"},18:{name:"GPSMapDatum",type:"Ascii"},19:{name:"GPSDestLatitudeRef",type:"Ascii"},20:{name:"GPSDestLatitude",type:"Rational"},21:{name:"GPSDestLongitudeRef",type:"Ascii"},22:{name:"GPSDestLongitude",type:"Rational"},23:{name:"GPSDestBearingRef",type:"Ascii"},24:{name:"GPSDestBearing",type:"Rational"},25:{name:"GPSDestDistanceRef",type:"Ascii"},26:{name:"GPSDestDistance",type:"Rational"},27:{name:"GPSProcessingMethod",type:"Undefined"},28:{name:"GPSAreaInformation",type:"Undefined"},29:{name:"GPSDateStamp",type:"Ascii"},30:{name:"GPSDifferential",type:"Short"},31:{name:"GPSHPositioningError",type:"Rational"}},Interop:{1:{name:"InteroperabilityIndex",type:"Ascii"}}};f["0th"]=f.Image,f["1st"]=f.Image,e.TAGS=f,e.ImageIFD={ProcessingSoftware:11,NewSubfileType:254,SubfileType:255,ImageWidth:256,ImageLength:257,BitsPerSample:258,Compression:259,PhotometricInterpretation:262,Threshholding:263,CellWidth:264,CellLength:265,FillOrder:266,DocumentName:269,ImageDescription:270,Make:271,Model:272,StripOffsets:273,Orientation:274,SamplesPerPixel:277,RowsPerStrip:278,StripByteCounts:279,XResolution:282,YResolution:283,PlanarConfiguration:284,GrayResponseUnit:290,GrayResponseCurve:291,T4Options:292,T6Options:293,ResolutionUnit:296,TransferFunction:301,Software:305,DateTime:306,Artist:315,HostComputer:316,Predictor:317,WhitePoint:318,PrimaryChromaticities:319,ColorMap:320,HalftoneHints:321,TileWidth:322,TileLength:323,TileOffsets:324,TileByteCounts:325,SubIFDs:330,InkSet:332,InkNames:333,NumberOfInks:334,DotRange:336,TargetPrinter:337,ExtraSamples:338,SampleFormat:339,SMinSampleValue:340,SMaxSampleValue:341,TransferRange:342,ClipPath:343,XClipPathUnits:344,YClipPathUnits:345,Indexed:346,JPEGTables:347,OPIProxy:351,JPEGProc:512,JPEGInterchangeFormat:513,JPEGInterchangeFormatLength:514,JPEGRestartInterval:515,JPEGLosslessPredictors:517,JPEGPointTransforms:518,JPEGQTables:519,JPEGDCTables:520,JPEGACTables:521,YCbCrCoefficients:529,YCbCrSubSampling:530,YCbCrPositioning:531,ReferenceBlackWhite:532,XMLPacket:700,Rating:18246,RatingPercent:18249,ImageID:32781,CFARepeatPatternDim:33421,CFAPattern:33422,BatteryLevel:33423,Copyright:33432,ExposureTime:33434,ImageResources:34377,ExifTag:34665,InterColorProfile:34675,GPSTag:34853,Interlace:34857,TimeZoneOffset:34858,SelfTimerMode:34859,FlashEnergy:37387,SpatialFrequencyResponse:37388,Noise:37389,FocalPlaneXResolution:37390,FocalPlaneYResolution:37391,FocalPlaneResolutionUnit:37392,ImageNumber:37393,SecurityClassification:37394,ImageHistory:37395,ExposureIndex:37397,TIFFEPStandardID:37398,SensingMethod:37399,XPTitle:40091,XPComment:40092,XPAuthor:40093,XPKeywords:40094,XPSubject:40095,PrintImageMatching:50341,DNGVersion:50706,DNGBackwardVersion:50707,UniqueCameraModel:50708,LocalizedCameraModel:50709,CFAPlaneColor:50710,CFALayout:50711,LinearizationTable:50712,BlackLevelRepeatDim:50713,BlackLevel:50714,BlackLevelDeltaH:50715,BlackLevelDeltaV:50716,WhiteLevel:50717,DefaultScale:50718,DefaultCropOrigin:50719,DefaultCropSize:50720,ColorMatrix1:50721,ColorMatrix2:50722,CameraCalibration1:50723,CameraCalibration2:50724,ReductionMatrix1:50725,ReductionMatrix2:50726,AnalogBalance:50727,AsShotNeutral:50728,AsShotWhiteXY:50729,BaselineExposure:50730,BaselineNoise:50731,BaselineSharpness:50732,BayerGreenSplit:50733,LinearResponseLimit:50734,CameraSerialNumber:50735,LensInfo:50736,ChromaBlurRadius:50737,AntiAliasStrength:50738,ShadowScale:50739,DNGPrivateData:50740,MakerNoteSafety:50741,CalibrationIlluminant1:50778,CalibrationIlluminant2:50779,BestQualityScale:50780,RawDataUniqueID:50781,OriginalRawFileName:50827,OriginalRawFileData:50828,ActiveArea:50829,MaskedAreas:50830,AsShotICCProfile:50831,AsShotPreProfileMatrix:50832,CurrentICCProfile:50833,CurrentPreProfileMatrix:50834,ColorimetricReference:50879,CameraCalibrationSignature:50931,ProfileCalibrationSignature:50932,AsShotProfileName:50934,NoiseReductionApplied:50935,ProfileName:50936,ProfileHueSatMapDims:50937,ProfileHueSatMapData1:50938,ProfileHueSatMapData2:50939,ProfileToneCurve:50940,ProfileEmbedPolicy:50941,ProfileCopyright:50942,ForwardMatrix1:50964,ForwardMatrix2:50965,PreviewApplicationName:50966,PreviewApplicationVersion:50967,PreviewSettingsName:50968,PreviewSettingsDigest:50969,PreviewColorSpace:50970,PreviewDateTime:50971,RawImageDigest:50972,OriginalRawFileDigest:50973,SubTileBlockSize:50974,RowInterleaveFactor:50975,ProfileLookTableDims:50981,ProfileLookTableData:50982,OpcodeList1:51008,OpcodeList2:51009,OpcodeList3:51022,NoiseProfile:51041},e.ExifIFD={ExposureTime:33434,FNumber:33437,ExposureProgram:34850,SpectralSensitivity:34852,ISOSpeedRatings:34855,OECF:34856,SensitivityType:34864,StandardOutputSensitivity:34865,RecommendedExposureIndex:34866,ISOSpeed:34867,ISOSpeedLatitudeyyy:34868,ISOSpeedLatitudezzz:34869,ExifVersion:36864,DateTimeOriginal:36867,DateTimeDigitized:36868,ComponentsConfiguration:37121,CompressedBitsPerPixel:37122,ShutterSpeedValue:37377,ApertureValue:37378,BrightnessValue:37379,ExposureBiasValue:37380,MaxApertureValue:37381,SubjectDistance:37382,MeteringMode:37383,LightSource:37384,Flash:37385,FocalLength:37386,SubjectArea:37396,MakerNote:37500,UserComment:37510,SubSecTime:37520,SubSecTimeOriginal:37521,SubSecTimeDigitized:37522,FlashpixVersion:40960,ColorSpace:40961,PixelXDimension:40962,PixelYDimension:40963,RelatedSoundFile:40964,InteroperabilityTag:40965,FlashEnergy:41483,SpatialFrequencyResponse:41484,FocalPlaneXResolution:41486,FocalPlaneYResolution:41487,FocalPlaneResolutionUnit:41488,SubjectLocation:41492,ExposureIndex:41493,SensingMethod:41495,FileSource:41728,SceneType:41729,CFAPattern:41730,CustomRendered:41985,ExposureMode:41986,WhiteBalance:41987,DigitalZoomRatio:41988,FocalLengthIn35mmFilm:41989,SceneCaptureType:41990,GainControl:41991,Contrast:41992,Saturation:41993,Sharpness:41994,DeviceSettingDescription:41995,SubjectDistanceRange:41996,ImageUniqueID:42016,CameraOwnerName:42032,BodySerialNumber:42033,LensSpecification:42034,LensMake:42035,LensModel:42036,LensSerialNumber:42037,Gamma:42240},e.GPSIFD={GPSVersionID:0,GPSLatitudeRef:1,GPSLatitude:2,GPSLongitudeRef:3,GPSLongitude:4,GPSAltitudeRef:5,GPSAltitude:6,GPSTimeStamp:7,GPSSatellites:8,GPSStatus:9,GPSMeasureMode:10,GPSDOP:11,GPSSpeedRef:12,GPSSpeed:13,GPSTrackRef:14,GPSTrack:15,GPSImgDirectionRef:16,GPSImgDirection:17,GPSMapDatum:18,GPSDestLatitudeRef:19,GPSDestLatitude:20,GPSDestLongitudeRef:21,GPSDestLongitude:22,GPSDestBearingRef:23,GPSDestBearing:24,GPSDestDistanceRef:25,GPSDestDistance:26,GPSProcessingMethod:27,GPSAreaInformation:28,GPSDateStamp:29,GPSDifferential:30,GPSHPositioningError:31},e.InteropIFD={InteroperabilityIndex:1},e.GPSHelper={degToDmsRational:function(e){let t=Math.abs(e),i=t%1*60,n=i%1*60;return[[Math.floor(t),1],[Math.floor(i),1],[Math.round(100*n),100]]},dmsRationalToDeg:function(e,t){let i="S"===t||"W"===t?-1:1;return(e[0][0]/e[0][1]+e[1][0]/e[1][1]/60+e[2][0]/e[2][1]/3600)*i}},"undefined"!=typeof exports?("undefined"!=typeof module&&module.exports&&(exports=module.exports=e),exports.piexif=e):window.piexif=e}();
\ No newline at end of file
+var AcuantConfig=function(){"use strict";return{acuantVersion:"11.9.3"}}();let config={};"undefined"!=typeof acuantConfig&&0!==Object.keys(acuantConfig).length&&acuantConfig.constructor===Object&&(config=acuantConfig),document.addEventListener("DOMContentLoaded",(function(){void 0===AcuantJavascriptWebSdk&&loadAcuantSdk(),document.removeEventListener("DOMContentLoaded",this)}));var AcuantJavascriptWebSdk=void 0;function loadAcuantSdk(){AcuantJavascriptWebSdk=function(e){let t={ACUANT_IMAGE_WORKER:"AcuantImageWorker",ACUANT_METRICS_WORKER:"AcuantMetricsWorker",SEQUENCE_BREAK_CODE:"sequence-break",START_FAIL_CODE:"start-fail",REPEAT_FAIL_CODE:"repeat-fail",HEIC_NOT_SUPPORTED_CODE:"heic-not-supported",BARCODE_READER_ID:"acuant-barcode-reader",singleWorkerModel:!1,startInitializer:function(n,a=0){if(!n)return void M("startInitializer did not have a user callback set");if(y)return;L=1==a,T(i,n);let r=null;e&&e.cdnPath&&e.cdnPath.initializerUrl?r=e.cdnPath.initializerUrl:e.path&&(r=e.path),y=new Worker(O(r,"AcuantInitializerWorker.min.js",a)),y.onmessage=w,function(){if(document.getElementById(t.BARCODE_READER_ID))return;const e=document.createElement("div");e.id=t.BARCODE_READER_ID,e.style.display="none",document.body.appendChild(e)}()},endInitializer:function(){y&&(y.terminate(),y.onmessage=null,g=!1,y=null)},startImageWorker:function(e){e?S?e():(T(a,e),I(L?1:0)):M("startImageWorker did not have a user callback set")},startMetricsWorker:function(e){e?R?e():(T(r,e),b(L?1:0)):M("startMetricsWorker did not have a user callback set")},endImageWorker:function(){S.terminate(),S.onmessage=null,P=!1,S=null},endMetricsWorker:function(){R.terminate(),R.onmessage=null,C=!1,R=null},start:function(e,t=!1){if(!e)return void M("start did not have a user callback set");const i=L?1:0;this.singleWorkerModel=t,T(n,e),this.singleWorkerModel?S||I(i):(S||I(i),R||b(i))},end:function(){S&&this.endImageWorker(),R&&this.endMetricsWorker()},startWorkers:function(e,t=[this.ACUANT_IMAGE_WORKER,this.ACUANT_METRICS_WORKER],i=0){e?(T(n,e),t.includes(this.ACUANT_IMAGE_WORKER)&&!S&&I(i),t.includes(this.ACUANT_METRICS_WORKER)&&!R&&b(i)):M("startWorkers did not have a user callback set")},endWorkers:function(e=[this.ACUANT_IMAGE_WORKER,this.ACUANT_METRICS_WORKER]){e.includes(this.ACUANT_IMAGE_WORKER)&&S&&(S.terminate(),S.onmessage=null,P=!1,S=null),e.includes(this.ACUANT_METRICS_WORKER)&&R&&(R.terminate(),R.onmessage=null,C=!1,R=null)},initialize:function(e,t,i,n=0){i?(T(o,i),y?B(y,"initialize",{creds:e,endpoint:t}):this.startInitializer((()=>{B(y,"initialize",{creds:e,endpoint:t})}),n)):M("initialize did not have a user callback set")},initializeWithToken:function(e,t,i,n=0){i?(T(o,i),y?B(y,"initializeWithToken",{token:e,endpoint:t}):this.startInitializer((()=>{B(y,"initializeWithToken",{token:e,endpoint:t})}),n)):M("initializeWithToken did not have a user callback set")},crop:function(e,t,i,n){h?n?P&&null!=e?(T(l,n),B(S,"crop",{imgData:e.data,width:t,height:i})):n.onFail():M("crop did not have a user callback set"):M("SDK was not initialized")},detect:function(e,t,i,n){h?n?P&&null!=e?(T(s,n),B(S,"detect",{imgData:e.data,width:t,height:i})):n.onFail():M("detect did not have a user callback set"):M("SDK was not initialized")},metrics:function(e,t,i,n){h?n?C&&null!=e?(T(m,n),B(R,"metrics",{imgData:e.data,width:t,height:i})):n.onFail():M("metrics did not have a user callback set"):M("SDK was not initialized")},moire:function(e,t,i,n){h?n?C&&null!=e?(T(c,n),B(R,"moire",{imgData:e.data,width:t,height:i})):n.onFail():M("moire did not have a user callback set"):M("SDK was not initialized")},sign:function(e,t){h?t?P&&e?(T(p,t),B(S,"sign",{imgData:e})):t.onFail():M("sign did not have a user callback set"):M("SDK was not initialized")},verify:function(e,t){h?t?P&&e?(T(u,t),B(S,"verify",{imgData:e})):t.onFail():M("verify did not have a user callback set"):M("SDK was not initialized")},getCvmlVersion:function(e){h?e?P?(T(d,e),B(S,d)):e.onFail():M("verify did not have a user callback set"):M("SDK was not initialized")},addMetadata:function(e,{make:t=navigator.platform,model:i=navigator.userAgent,software:n="Acuant JavascriptWeb SDK "+AcuantConfig.acuantVersion,imageDescription:a=null,dateTimeOriginal:r,userComment:o="=".repeat(100)}){if(!h)return void M("SDK was not initialized");let l={},s={};l[piexif.ImageIFD.Make]=t,l[piexif.ImageIFD.Model]=i,l[piexif.ImageIFD.Software]=n,a&&(l[piexif.ImageIFD.ImageDescription]=a),s[piexif.ExifIFD.DateTimeOriginal]=r,s[piexif.ExifIFD.UserComment]=o;let m={"0th":l,Exif:s},c=piexif.dump(m);return piexif.insert(c,e)},setUnexpectedErrorCallback:function(e){T(f,e)}};const i="initStart",n="workersStart",a="imageWorkerStart",r="metricsWorkerStart",o="init",l="crop",s="detect",m="metrics",c="moire",p="sign",u="verify",d="getCvmlVersion",f="unexpectedError";let h=!1,y=null,g=!1,S=null,P=!1,R=null,C=!1,A=0,k={},D={},L=!1;function I(t=0){let i=null;e&&e.cdnPath&&e.cdnPath.imageUrl?i=e.cdnPath.imageUrl:e.path&&(i=e.path),A++,S=new Worker(O(i,"AcuantImageWorker.min.js",t)),S.onmessage=E,S.onerror=function(){M("imageWorker has failed")}}function b(t=0){let i=null;e&&e.cdnPath&&e.cdnPath.metricsUrl?i=e.cdnPath.metricsUrl:e.path&&(i=e.path),A++,R=new Worker(O(i,"AcuantMetricsWorker.min.js",t)),R.onmessage=x,R.onerror=function(){M("metricsWorker has failed")}}function w(e){if(h=!1,e){let n=e.data;if(g)if(n&&"initialize"===n.func){let e=n.status,i=k[o];t.endInitializer(),i?1==e?(h=!0,i.onSuccess()):i.onFail(e,function(e){switch(e){case 401:return"Server returned a 401 (missing credentials).";case 403:return"Server returned a 403 (invalid credentials).";case 400:return"Server returned a 400.";case 2:return"Token Validation Failed (Recieved token, but token was null/corrupt).";case 3:return"Token Validation Failed (Recieved token, but token was missing part of body).";case 4:return"Token Validation Failed (Recieved token, but token body was missing fields).";case 5:return"Token Validation Failed (Recieved token, but token body failed validation).";case 6:return"At least one param was null/invalid.";case 7:return"Incorrectly formatted message to worker.";default:return"Unexpected error code."}}(e)):M("initialize did not have a user callback set")}else M("initworker sent message without correct function tagging");else{g=!0;let e=k[i];e&&e()}}else M("initworker sent message without anything in the body")}function E(e){if(e){let t=e.data;if(P)if(t&&"detect"===t.func){const e=k[s];e?t.type&&t.x1&&t.y1&&t.x2&&t.y2&&t.x3&&t.y3&&t.x4&&t.y4?function(e,t,i,n,a,r,o,l,s,m){if(m)if(-1==e)m.onFail();else{let c=function(e,t,i,n,a,r,o,l){let s={x:e,y:t},m={x:i,y:n},c={x:a,y:r},p={x:o,y:l},u=G(s,m),d=G(m,c),f=G(c,p),h=G(p,s),y=(u+f)/2,g=(d+h)/2;return y>g?{width:y,height:g}:{width:g,height:y}}(t,i,n,a,r,o,l,s),p=function(e,t){let i=!1,n=5,a=1.42,r=1.5887;if(2==t){let t=(100+n)/100*a;e>=(100-n)/100*a&&e<=t&&(i=!0)}else if(1==t){let t=(100+n)/100*r;e>=(100-n)/100*r&&e<=t&&(i=!0)}return i}(c.width/c.height,e),u=F(c.width,c.height,2==e),d=function(e){let t=[-1,-1,-1,-1];e&&4===e.length&&(v(t,e[0],e[2]),v(t,e[1],e[3]));return t}([{x:t,y:i},{x:n,y:a},{x:r,y:o},{x:l,y:s}]);m.onSuccess({type:e,dimensions:c,dpi:u,isCorrectAspectRatio:p,points:d})}}(t.type,t.x1,t.y1,t.x2,t.y2,t.x3,t.y3,t.x4,t.y4,e):e.onFail():M("detect did not have a user callback set")}else if(t&&"crop"===t.func){const e=k[l];e?t.imgData&&t.width&&t.height&&t.type?function(e,t,i,n,a){a&&(null!=e&&t>=0&&i>=0&&n>=0?(D={image:{data:e,width:t,height:i},cardType:n,dpi:F(t,i,2==n)},a.onSuccess(D)):a.onFail())}(t.imgData,t.width,t.height,t.type,e):t.error?e.onFail(t.error):e.onFail():M("crop did not have a user callback set")}else if(t&&"sign"===t.func){const e=k[p];e?t.imgData?function(e,t){t&&(e?t.onSuccess(e):t.onFail())}(t.imgData,e):t.error?e.onFail(t.error):e.onFail():M("sign did not have a user callback set")}else if(t&&"verify"===t.func){const e=k[u];e?t.result||!1===t.result?function(e,t){t&&(e||!1===e?t.onSuccess(e):t.onFail())}(t.result,e):e.onFail():M("verify did not have a user callback set")}else if(t&&t.func===d){let e=k[d];e?function(e,t){e?t.onSuccess(e):t.onFail()}(t.cvmlVersion,e):M("getCvmlVersion did not have a user callback set")}else M("imageworker sent message without correct function tagging");else P=!0,_()}else M("imageworker sent message without anything in the body")}function x(e){if(e){let t=e.data;if(C)if(t&&"metrics"===t.func){const e=k[m];e?t.sharpness&&t.glare?function(e,t,i){if(i)if(t>=0&&e>=0){let n=Math.floor(100*e),a=Math.floor(100*t);i.onSuccess(n,a)}else i.onFail()}(t.sharpness,t.glare,e):t.error?e.onFail(t.error):e.onFail():M("metrics did not have a user callback set")}else if("moire"===t.func){const e=k[c];e?t.moire&&t.moireraw?function(e,t,i){if(i)if(e>=0&&t>=0){let n=Math.floor(100*e),a=Math.floor(100*t);i.onSuccess(n,a)}else i.onFail()}(t.moire,t.moireraw,e):t.error?e.onFail(t.error):e.onFail():M("moire did not have a user callback set")}else M("metricsworker sent message without correct function tagging");else C=!0,_()}else M("metricsworker sent message without anything in the body")}function v(e,t,i){return t.x<i.x&&t.y<i.y?(e[0]=t,e[2]=i):t.x>i.x&&t.y>i.y?(e[0]=i,e[2]=t):t.x>i.x&&t.y<i.y?(e[1]=t,e[3]=i):(e[1]=i,e[3]=t),e}function G(e,t){return Math.sqrt(Math.pow(t.x-e.x,2)+Math.pow(t.y-e.y,2))}function F(e,t,i){let n=e>t?e:t,a=i?4.92:3.37;return Math.round(n/a)}function T(e,t){k[e]=t}function M(e){let t=k[f];t?e?t(e):t():console.error("Error: ",e)}function O(e,t,i){let n;return null!=e&&e.length>0&&0==i?(n="/"===e.charAt(e.length-1)?e:e+"/",n+=t):n=0!=i?e:t,n}function B(e,t,i,n=!1){let a={func:t,data:i};n&&i&&i.imgData&&i.imgData.buffer?e.postMessage(a,[a.data.imgData.buffer]):e.postMessage(a)}function _(){const e=k[n],t=k[a],i=k[r];--A,0==A&&(e?(T(n,null),e()):t?(T(a,null),t()):i&&(T(r,null),i()))}return t}(config),"function"==typeof onAcuantSdkLoaded&&onAcuantSdkLoaded()}!function(){"use strict";let e={};function t(e){return m(">"+p("B",e.length),e)}function i(e){return m(">"+p("H",e.length),e)}function n(e){return m(">"+p("L",e.length),e)}function a(e,a,r){let o,l,s,c,u="",d="";if("Byte"==a)o=e.length,o<=4?d=t(e)+p("\0",4-o):(d=m(">L",[r]),u=t(e));else if("Short"==a)o=e.length,o<=2?d=i(e)+p("\0\0",2-o):(d=m(">L",[r]),u=i(e));else if("Long"==a)o=e.length,o<=1?d=n(e):(d=m(">L",[r]),u=n(e));else if("Ascii"==a)l=e+"\0",o=l.length,o>4?(d=m(">L",[r]),u=l):d=l+p("\0",4-o);else if("Rational"==a){if("number"==typeof e[0])o=1,s=e[0],c=e[1],l=m(">L",[s])+m(">L",[c]);else{o=e.length,l="";for(var f=0;f<o;f++)s=e[f][0],c=e[f][1],l+=m(">L",[s])+m(">L",[c])}d=m(">L",[r]),u=l}else if("SRational"==a){if("number"==typeof e[0])o=1,s=e[0],c=e[1],l=m(">l",[s])+m(">l",[c]);else{o=e.length,l="";for(f=0;f<o;f++)s=e[f][0],c=e[f][1],l+=m(">l",[s])+m(">l",[c])}d=m(">L",[r]),u=l}else"Undefined"==a&&(o=e.length,o>4?(d=m(">L",[r]),u=e):d=e+p("\0",4-o));return[m(">L",[o]),d,u]}function r(e,t,i){let n,r=Object.keys(e).length,o=m(">H",[r]);n=["0th","1st"].indexOf(t)>-1?2+12*r+4:2+12*r;let l="",s="";for(var c in e){if("string"==typeof c&&(c=parseInt(c)),"0th"==t&&[34665,34853].indexOf(c)>-1)continue;if("Exif"==t&&40965==c)continue;if("1st"==t&&[513,514].indexOf(c)>-1)continue;let r=e[c],o=m(">H",[c]),p=f[t][c].type,u=m(">H",[d[p]]);"number"==typeof r&&(r=[r]);let h=a(r,p,8+n+i+s.length);l+=o+u+h[0]+h[1],s+=h[2]}return[o+l,s]}function o(e){let t,i;if("ÿØ"==e.slice(0,2))t=u(e),i=function(e){let t;for(let i=0;i<e.length;i++)if(t=e[i],"ÿá"==t.slice(0,2)&&"Exif\0\0"==t.slice(4,10))return t;return null}(t),this.tiftag=i?i.slice(10):null;else if(["II","MM"].indexOf(e.slice(0,2))>-1)this.tiftag=e;else{if("Exif"!=e.slice(0,4))throw new Error("Given file is neither JPEG nor TIFF.");this.tiftag=e.slice(6)}}if(e.version="1.0.4",e.remove=function(e){let t=!1;if("ÿØ"==e.slice(0,2));else{if("data:image/jpeg;base64,"!=e.slice(0,23)&&"data:image/jpg;base64,"!=e.slice(0,22))throw new Error("Given data is not jpeg.");e=s(e.split(",")[1]),t=!0}let i=u(e).filter((function(e){return!("ÿá"==e.slice(0,2)&&"Exif\0\0"==e.slice(4,10))})).join("");return t&&(i="data:image/jpeg;base64,"+l(i)),i},e.insert=function(e,t){let i=!1;if("Exif\0\0"!=e.slice(0,6))throw new Error("Given data is not exif.");if("ÿØ"==t.slice(0,2));else{if("data:image/jpeg;base64,"!=t.slice(0,23)&&"data:image/jpg;base64,"!=t.slice(0,22))throw new Error("Given data is not jpeg.");t=s(t.split(",")[1]),i=!0}let n="ÿá"+m(">H",[e.length+2])+e,a=function(e,t){let i=!1,n=[];e.forEach((function(a,r){"ÿá"==a.slice(0,2)&&"Exif\0\0"==a.slice(4,10)&&(i?n.unshift(r):(e[r]=t,i=!0))})),n.forEach((function(t){e.splice(t,1)})),!i&&t&&(e=[e[0],t].concat(e.slice(1)));return e.join("")}(u(t),n);return i&&(a="data:image/jpeg;base64,"+l(a)),a},e.load=function(e){let t;if("string"!=typeof e)throw new Error("'load' gots invalid type argument.");if("ÿØ"==e.slice(0,2))t=e;else if("data:image/jpeg;base64,"==e.slice(0,23)||"data:image/jpg;base64,"==e.slice(0,22))t=s(e.split(",")[1]);else{if("Exif"!=e.slice(0,4))throw new Error("'load' gots invalid file data.");t=e.slice(6)}let i={"0th":{},Exif:{},GPS:{},Interop:{},"1st":{},thumbnail:null},n=new o(t);if(null===n.tiftag)return i;"II"==n.tiftag.slice(0,2)?n.endian_mark="<":n.endian_mark=">";let a=c(n.endian_mark+"L",n.tiftag.slice(4,8))[0];i["0th"]=n.get_ifd(a,"0th");let r=i["0th"].first_ifd_pointer;if(delete i["0th"].first_ifd_pointer,34665 in i["0th"]&&(a=i["0th"][34665],i.Exif=n.get_ifd(a,"Exif")),34853 in i["0th"]&&(a=i["0th"][34853],i.GPS=n.get_ifd(a,"GPS")),40965 in i.Exif&&(a=i.Exif[40965],i.Interop=n.get_ifd(a,"Interop")),"\0\0\0\0"!=r&&(a=c(n.endian_mark+"L",r)[0],i["1st"]=n.get_ifd(a,"1st"),513 in i["1st"]&&514 in i["1st"])){let e=i["1st"][513]+i["1st"][514],t=n.tiftag.slice(i["1st"][513],e);i.thumbnail=t}return i},e.dump=function(t){let i=(n=t,JSON.parse(JSON.stringify(n)));var n;let a,o,l,s,c,p=!1,f=!1,h=!1,y=!1;a="0th"in i?i["0th"]:{},"Exif"in i&&Object.keys(i.Exif).length||"Interop"in i&&Object.keys(i.Interop).length?(a[34665]=1,p=!0,o=i.Exif,"Interop"in i&&Object.keys(i.Interop).length?(o[40965]=1,h=!0,l=i.Interop):Object.keys(o).indexOf(e.ExifIFD.InteroperabilityTag.toString())>-1&&delete o[40965]):Object.keys(a).indexOf(e.ImageIFD.ExifTag.toString())>-1&&delete a[34665],"GPS"in i&&Object.keys(i.GPS).length?(a[e.ImageIFD.GPSTag]=1,f=!0,s=i.GPS):Object.keys(a).indexOf(e.ImageIFD.GPSTag.toString())>-1&&delete a[e.ImageIFD.GPSTag],"1st"in i&&"thumbnail"in i&&null!=i.thumbnail&&(y=!0,i["1st"][513]=1,i["1st"][514]=1,c=i["1st"]);let g,S,P,R,C,A=r(a,"0th",0),k=A[0].length+12*p+12*f+4+A[1].length,D="",L=0,I="",b=0,w="",E=0,x="";(p&&(g=r(o,"Exif",k),L=g[0].length+12*h+g[1].length),f&&(S=r(s,"GPS",k+L),I=S.join(""),b=I.length),h)&&(P=r(l,"Interop",k+L+b),w=P.join(""),E=w.length);if(y&&(R=r(c,"1st",k+L+b+E),C=function(e){let t=u(e);for(;"ÿà"<=t[1].slice(0,2)&&t[1].slice(0,2)<="ÿï";)t=[t[0]].concat(t.slice(2));return t.join("")}(i.thumbnail),C.length>64e3))throw new Error("Given thumbnail is too large. max 64kB");let v="",G="",F="",T="\0\0\0\0";if(p){var M=m(">L",[O=8+k]);v=m(">H",[34665])+m(">H",[d.Long])+m(">L",[1])+M}if(f){M=m(">L",[O=8+k+L]);G=m(">H",[34853])+m(">H",[d.Long])+m(">L",[1])+M}if(h){M=m(">L",[O=8+k+L+b]);F=m(">H",[40965])+m(">H",[d.Long])+m(">L",[1])+M}if(y){var O;T=m(">L",[O=8+k+L+b+E]);let e="
\0\0\0\0
"+m(">L",[O+R[0].length+24+4+R[1].length]),t="\0\0\0\0
"+m(">L",[C.length]);x=R[0]+e+t+"\0\0\0\0"+R[1]+C}let B=A[0]+v+G+T+A[1];return p&&(D=g[0]+F+g[1]),"Exif\0\0MM\0*\0\0\0\b"+B+D+I+w+x},o.prototype={get_ifd:function(e,t){let i,n={},a=c(this.endian_mark+"H",this.tiftag.slice(e,e+2))[0],r=e+2;i=["0th","1st"].indexOf(t)>-1?"Image":t;for(let t=0;t<a;t++){e=r+12*t;let a=c(this.endian_mark+"H",this.tiftag.slice(e,e+2))[0],o=[c(this.endian_mark+"H",this.tiftag.slice(e+2,e+4))[0],c(this.endian_mark+"L",this.tiftag.slice(e+4,e+8))[0],this.tiftag.slice(e+8,e+12)];a in f[i]&&(n[a]=this.convert_value(o))}return"0th"==t&&(e=r+12*a,n.first_ifd_pointer=this.tiftag.slice(e,e+4)),n},convert_value:function(e){let t,i=null,n=e[0],a=e[1],r=e[2];if(1==n)a>4?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("B",a),this.tiftag.slice(t,t+a))):i=c(this.endian_mark+p("B",a),r.slice(0,a));else if(2==n)a>4?(t=c(this.endian_mark+"L",r)[0],i=this.tiftag.slice(t,t+a-1)):i=r.slice(0,a-1);else if(3==n)a>2?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("H",a),this.tiftag.slice(t,t+2*a))):i=c(this.endian_mark+p("H",a),r.slice(0,2*a));else if(4==n)a>1?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("L",a),this.tiftag.slice(t,t+4*a))):i=c(this.endian_mark+p("L",a),r);else if(5==n)if(t=c(this.endian_mark+"L",r)[0],a>1){i=[];for(var o=0;o<a;o++)i.push([c(this.endian_mark+"L",this.tiftag.slice(t+8*o,t+4+8*o))[0],c(this.endian_mark+"L",this.tiftag.slice(t+4+8*o,t+8+8*o))[0]])}else i=[c(this.endian_mark+"L",this.tiftag.slice(t,t+4))[0],c(this.endian_mark+"L",this.tiftag.slice(t+4,t+8))[0]];else if(7==n)a>4?(t=c(this.endian_mark+"L",r)[0],i=this.tiftag.slice(t,t+a)):i=r.slice(0,a);else if(9==n)a>1?(t=c(this.endian_mark+"L",r)[0],i=c(this.endian_mark+p("l",a),this.tiftag.slice(t,t+4*a))):i=c(this.endian_mark+p("l",a),r);else{if(10!=n)throw new Error("Exif might be wrong. Got incorrect value type to decode. type:"+n);if(t=c(this.endian_mark+"L",r)[0],a>1){i=[];for(o=0;o<a;o++)i.push([c(this.endian_mark+"l",this.tiftag.slice(t+8*o,t+4+8*o))[0],c(this.endian_mark+"l",this.tiftag.slice(t+4+8*o,t+8+8*o))[0]])}else i=[c(this.endian_mark+"l",this.tiftag.slice(t,t+4))[0],c(this.endian_mark+"l",this.tiftag.slice(t+4,t+8))[0]]}return i instanceof Array&&1==i.length?i[0]:i}},"undefined"!=typeof window&&"function"==typeof window.btoa)var l=window.btoa;if(void 0===l)l=function(e){let t,i,n,a,r,o,l,s="",m=0,c="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";for(;m<e.length;)t=e.charCodeAt(m++),i=e.charCodeAt(m++),n=e.charCodeAt(m++),a=t>>2,r=(3&t)<<4|i>>4,o=(15&i)<<2|n>>6,l=63&n,isNaN(i)?o=l=64:isNaN(n)&&(l=64),s=s+c.charAt(a)+c.charAt(r)+c.charAt(o)+c.charAt(l);return s};if("undefined"!=typeof window&&"function"==typeof window.atob)var s=window.atob;if(void 0===s)s=function(e){let t,i,n,a,r,o,l,s="",m=0,c="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";for(e=e.replace(/[^A-Za-z0-9\+\/\=]/g,"");m<e.length;)a=c.indexOf(e.charAt(m++)),r=c.indexOf(e.charAt(m++)),o=c.indexOf(e.charAt(m++)),l=c.indexOf(e.charAt(m++)),t=a<<2|r>>4,i=(15&r)<<4|o>>2,n=(3&o)<<6|l,s+=String.fromCharCode(t),64!=o&&(s+=String.fromCharCode(i)),64!=l&&(s+=String.fromCharCode(n));return s};function m(e,t){if(!(t instanceof Array))throw new Error("'pack' error. Got invalid type argument.");if(e.length-1!=t.length)throw new Error("'pack' error. "+(e.length-1)+" marks, "+t.length+" elements.");let i;if("<"==e[0])i=!0;else{if(">"!=e[0])throw new Error("");i=!1}let n="",a=1,r=null,o=null,l=null;for(;o=e[a];){if("b"==o.toLowerCase()){if(r=t[a-1],"b"==o&&r<0&&(r+=256),r>255||r<0)throw new Error("'pack' error.");l=String.fromCharCode(r)}else if("H"==o){if(r=t[a-1],r>65535||r<0)throw new Error("'pack' error.");l=String.fromCharCode(Math.floor(r%65536/256))+String.fromCharCode(r%256),i&&(l=l.split("").reverse().join(""))}else{if("l"!=o.toLowerCase())throw new Error("'pack' error.");if(r=t[a-1],"l"==o&&r<0&&(r+=4294967296),r>4294967295||r<0)throw new Error("'pack' error.");l=String.fromCharCode(Math.floor(r/16777216))+String.fromCharCode(Math.floor(r%16777216/65536))+String.fromCharCode(Math.floor(r%65536/256))+String.fromCharCode(r%256),i&&(l=l.split("").reverse().join(""))}n+=l,a+=1}return n}function c(e,t){if("string"!=typeof t)throw new Error("'unpack' error. Got invalid type argument.");let i,n=0;for(let t=1;t<e.length;t++)if("b"==e[t].toLowerCase())n+=1;else if("h"==e[t].toLowerCase())n+=2;else{if("l"!=e[t].toLowerCase())throw new Error("'unpack' error. Got invalid mark.");n+=4}if(n!=t.length)throw new Error("'unpack' error. Mismatch between symbol and string length. "+n+":"+t.length);if("<"==e[0])i=!0;else{if(">"!=e[0])throw new Error("'unpack' error.");i=!1}let a=[],r=0,o=1,l=null,s=null,m=null,c="";for(;s=e[o];){if("b"==s.toLowerCase())m=1,c=t.slice(r,r+m),l=c.charCodeAt(0),"b"==s&&l>=128&&(l-=256);else if("H"==s)m=2,c=t.slice(r,r+m),i&&(c=c.split("").reverse().join("")),l=256*c.charCodeAt(0)+c.charCodeAt(1);else{if("l"!=s.toLowerCase())throw new Error("'unpack' error. "+s);m=4,c=t.slice(r,r+m),i&&(c=c.split("").reverse().join("")),l=16777216*c.charCodeAt(0)+65536*c.charCodeAt(1)+256*c.charCodeAt(2)+c.charCodeAt(3),"l"==s&&l>=2147483648&&(l-=4294967296)}a.push(l),r+=m,o+=1}return a}function p(e,t){let i="";for(let n=0;n<t;n++)i+=e;return i}function u(e){if("ÿØ"!=e.slice(0,2))throw new Error("Given data isn't JPEG.");let t=2,i=["ÿØ"];for(;;){if("ÿÚ"==e.slice(t,t+2)){i.push(e.slice(t));break}{let n=t+c(">H",e.slice(t+2,t+4))[0]+2;i.push(e.slice(t,n)),t=n}if(t>=e.length)throw new Error("Wrong JPEG data.")}return i}var d={Byte:1,Ascii:2,Short:3,Long:4,Rational:5,Undefined:7,SLong:9,SRational:10},f={Image:{11:{name:"ProcessingSoftware",type:"Ascii"},254:{name:"NewSubfileType",type:"Long"},255:{name:"SubfileType",type:"Short"},256:{name:"ImageWidth",type:"Long"},257:{name:"ImageLength",type:"Long"},258:{name:"BitsPerSample",type:"Short"},259:{name:"Compression",type:"Short"},262:{name:"PhotometricInterpretation",type:"Short"},263:{name:"Threshholding",type:"Short"},264:{name:"CellWidth",type:"Short"},265:{name:"CellLength",type:"Short"},266:{name:"FillOrder",type:"Short"},269:{name:"DocumentName",type:"Ascii"},270:{name:"ImageDescription",type:"Ascii"},271:{name:"Make",type:"Ascii"},272:{name:"Model",type:"Ascii"},273:{name:"StripOffsets",type:"Long"},274:{name:"Orientation",type:"Short"},277:{name:"SamplesPerPixel",type:"Short"},278:{name:"RowsPerStrip",type:"Long"},279:{name:"StripByteCounts",type:"Long"},282:{name:"XResolution",type:"Rational"},283:{name:"YResolution",type:"Rational"},284:{name:"PlanarConfiguration",type:"Short"},290:{name:"GrayResponseUnit",type:"Short"},291:{name:"GrayResponseCurve",type:"Short"},292:{name:"T4Options",type:"Long"},293:{name:"T6Options",type:"Long"},296:{name:"ResolutionUnit",type:"Short"},301:{name:"TransferFunction",type:"Short"},305:{name:"Software",type:"Ascii"},306:{name:"DateTime",type:"Ascii"},315:{name:"Artist",type:"Ascii"},316:{name:"HostComputer",type:"Ascii"},317:{name:"Predictor",type:"Short"},318:{name:"WhitePoint",type:"Rational"},319:{name:"PrimaryChromaticities",type:"Rational"},320:{name:"ColorMap",type:"Short"},321:{name:"HalftoneHints",type:"Short"},322:{name:"TileWidth",type:"Short"},323:{name:"TileLength",type:"Short"},324:{name:"TileOffsets",type:"Short"},325:{name:"TileByteCounts",type:"Short"},330:{name:"SubIFDs",type:"Long"},332:{name:"InkSet",type:"Short"},333:{name:"InkNames",type:"Ascii"},334:{name:"NumberOfInks",type:"Short"},336:{name:"DotRange",type:"Byte"},337:{name:"TargetPrinter",type:"Ascii"},338:{name:"ExtraSamples",type:"Short"},339:{name:"SampleFormat",type:"Short"},340:{name:"SMinSampleValue",type:"Short"},341:{name:"SMaxSampleValue",type:"Short"},342:{name:"TransferRange",type:"Short"},343:{name:"ClipPath",type:"Byte"},344:{name:"XClipPathUnits",type:"Long"},345:{name:"YClipPathUnits",type:"Long"},346:{name:"Indexed",type:"Short"},347:{name:"JPEGTables",type:"Undefined"},351:{name:"OPIProxy",type:"Short"},512:{name:"JPEGProc",type:"Long"},513:{name:"JPEGInterchangeFormat",type:"Long"},514:{name:"JPEGInterchangeFormatLength",type:"Long"},515:{name:"JPEGRestartInterval",type:"Short"},517:{name:"JPEGLosslessPredictors",type:"Short"},518:{name:"JPEGPointTransforms",type:"Short"},519:{name:"JPEGQTables",type:"Long"},520:{name:"JPEGDCTables",type:"Long"},521:{name:"JPEGACTables",type:"Long"},529:{name:"YCbCrCoefficients",type:"Rational"},530:{name:"YCbCrSubSampling",type:"Short"},531:{name:"YCbCrPositioning",type:"Short"},532:{name:"ReferenceBlackWhite",type:"Rational"},700:{name:"XMLPacket",type:"Byte"},18246:{name:"Rating",type:"Short"},18249:{name:"RatingPercent",type:"Short"},32781:{name:"ImageID",type:"Ascii"},33421:{name:"CFARepeatPatternDim",type:"Short"},33422:{name:"CFAPattern",type:"Byte"},33423:{name:"BatteryLevel",type:"Rational"},33432:{name:"Copyright",type:"Ascii"},33434:{name:"ExposureTime",type:"Rational"},34377:{name:"ImageResources",type:"Byte"},34665:{name:"ExifTag",type:"Long"},34675:{name:"InterColorProfile",type:"Undefined"},34853:{name:"GPSTag",type:"Long"},34857:{name:"Interlace",type:"Short"},34858:{name:"TimeZoneOffset",type:"Long"},34859:{name:"SelfTimerMode",type:"Short"},37387:{name:"FlashEnergy",type:"Rational"},37388:{name:"SpatialFrequencyResponse",type:"Undefined"},37389:{name:"Noise",type:"Undefined"},37390:{name:"FocalPlaneXResolution",type:"Rational"},37391:{name:"FocalPlaneYResolution",type:"Rational"},37392:{name:"FocalPlaneResolutionUnit",type:"Short"},37393:{name:"ImageNumber",type:"Long"},37394:{name:"SecurityClassification",type:"Ascii"},37395:{name:"ImageHistory",type:"Ascii"},37397:{name:"ExposureIndex",type:"Rational"},37398:{name:"TIFFEPStandardID",type:"Byte"},37399:{name:"SensingMethod",type:"Short"},40091:{name:"XPTitle",type:"Byte"},40092:{name:"XPComment",type:"Byte"},40093:{name:"XPAuthor",type:"Byte"},40094:{name:"XPKeywords",type:"Byte"},40095:{name:"XPSubject",type:"Byte"},50341:{name:"PrintImageMatching",type:"Undefined"},50706:{name:"DNGVersion",type:"Byte"},50707:{name:"DNGBackwardVersion",type:"Byte"},50708:{name:"UniqueCameraModel",type:"Ascii"},50709:{name:"LocalizedCameraModel",type:"Byte"},50710:{name:"CFAPlaneColor",type:"Byte"},50711:{name:"CFALayout",type:"Short"},50712:{name:"LinearizationTable",type:"Short"},50713:{name:"BlackLevelRepeatDim",type:"Short"},50714:{name:"BlackLevel",type:"Rational"},50715:{name:"BlackLevelDeltaH",type:"SRational"},50716:{name:"BlackLevelDeltaV",type:"SRational"},50717:{name:"WhiteLevel",type:"Short"},50718:{name:"DefaultScale",type:"Rational"},50719:{name:"DefaultCropOrigin",type:"Short"},50720:{name:"DefaultCropSize",type:"Short"},50721:{name:"ColorMatrix1",type:"SRational"},50722:{name:"ColorMatrix2",type:"SRational"},50723:{name:"CameraCalibration1",type:"SRational"},50724:{name:"CameraCalibration2",type:"SRational"},50725:{name:"ReductionMatrix1",type:"SRational"},50726:{name:"ReductionMatrix2",type:"SRational"},50727:{name:"AnalogBalance",type:"Rational"},50728:{name:"AsShotNeutral",type:"Short"},50729:{name:"AsShotWhiteXY",type:"Rational"},50730:{name:"BaselineExposure",type:"SRational"},50731:{name:"BaselineNoise",type:"Rational"},50732:{name:"BaselineSharpness",type:"Rational"},50733:{name:"BayerGreenSplit",type:"Long"},50734:{name:"LinearResponseLimit",type:"Rational"},50735:{name:"CameraSerialNumber",type:"Ascii"},50736:{name:"LensInfo",type:"Rational"},50737:{name:"ChromaBlurRadius",type:"Rational"},50738:{name:"AntiAliasStrength",type:"Rational"},50739:{name:"ShadowScale",type:"SRational"},50740:{name:"DNGPrivateData",type:"Byte"},50741:{name:"MakerNoteSafety",type:"Short"},50778:{name:"CalibrationIlluminant1",type:"Short"},50779:{name:"CalibrationIlluminant2",type:"Short"},50780:{name:"BestQualityScale",type:"Rational"},50781:{name:"RawDataUniqueID",type:"Byte"},50827:{name:"OriginalRawFileName",type:"Byte"},50828:{name:"OriginalRawFileData",type:"Undefined"},50829:{name:"ActiveArea",type:"Short"},50830:{name:"MaskedAreas",type:"Short"},50831:{name:"AsShotICCProfile",type:"Undefined"},50832:{name:"AsShotPreProfileMatrix",type:"SRational"},50833:{name:"CurrentICCProfile",type:"Undefined"},50834:{name:"CurrentPreProfileMatrix",type:"SRational"},50879:{name:"ColorimetricReference",type:"Short"},50931:{name:"CameraCalibrationSignature",type:"Byte"},50932:{name:"ProfileCalibrationSignature",type:"Byte"},50934:{name:"AsShotProfileName",type:"Byte"},50935:{name:"NoiseReductionApplied",type:"Rational"},50936:{name:"ProfileName",type:"Byte"},50937:{name:"ProfileHueSatMapDims",type:"Long"},50938:{name:"ProfileHueSatMapData1",type:"Float"},50939:{name:"ProfileHueSatMapData2",type:"Float"},50940:{name:"ProfileToneCurve",type:"Float"},50941:{name:"ProfileEmbedPolicy",type:"Long"},50942:{name:"ProfileCopyright",type:"Byte"},50964:{name:"ForwardMatrix1",type:"SRational"},50965:{name:"ForwardMatrix2",type:"SRational"},50966:{name:"PreviewApplicationName",type:"Byte"},50967:{name:"PreviewApplicationVersion",type:"Byte"},50968:{name:"PreviewSettingsName",type:"Byte"},50969:{name:"PreviewSettingsDigest",type:"Byte"},50970:{name:"PreviewColorSpace",type:"Long"},50971:{name:"PreviewDateTime",type:"Ascii"},50972:{name:"RawImageDigest",type:"Undefined"},50973:{name:"OriginalRawFileDigest",type:"Undefined"},50974:{name:"SubTileBlockSize",type:"Long"},50975:{name:"RowInterleaveFactor",type:"Long"},50981:{name:"ProfileLookTableDims",type:"Long"},50982:{name:"ProfileLookTableData",type:"Float"},51008:{name:"OpcodeList1",type:"Undefined"},51009:{name:"OpcodeList2",type:"Undefined"},51022:{name:"OpcodeList3",type:"Undefined"}},Exif:{33434:{name:"ExposureTime",type:"Rational"},33437:{name:"FNumber",type:"Rational"},34850:{name:"ExposureProgram",type:"Short"},34852:{name:"SpectralSensitivity",type:"Ascii"},34855:{name:"ISOSpeedRatings",type:"Short"},34856:{name:"OECF",type:"Undefined"},34864:{name:"SensitivityType",type:"Short"},34865:{name:"StandardOutputSensitivity",type:"Long"},34866:{name:"RecommendedExposureIndex",type:"Long"},34867:{name:"ISOSpeed",type:"Long"},34868:{name:"ISOSpeedLatitudeyyy",type:"Long"},34869:{name:"ISOSpeedLatitudezzz",type:"Long"},36864:{name:"ExifVersion",type:"Undefined"},36867:{name:"DateTimeOriginal",type:"Ascii"},36868:{name:"DateTimeDigitized",type:"Ascii"},37121:{name:"ComponentsConfiguration",type:"Undefined"},37122:{name:"CompressedBitsPerPixel",type:"Rational"},37377:{name:"ShutterSpeedValue",type:"SRational"},37378:{name:"ApertureValue",type:"Rational"},37379:{name:"BrightnessValue",type:"SRational"},37380:{name:"ExposureBiasValue",type:"SRational"},37381:{name:"MaxApertureValue",type:"Rational"},37382:{name:"SubjectDistance",type:"Rational"},37383:{name:"MeteringMode",type:"Short"},37384:{name:"LightSource",type:"Short"},37385:{name:"Flash",type:"Short"},37386:{name:"FocalLength",type:"Rational"},37396:{name:"SubjectArea",type:"Short"},37500:{name:"MakerNote",type:"Undefined"},37510:{name:"UserComment",type:"Ascii"},37520:{name:"SubSecTime",type:"Ascii"},37521:{name:"SubSecTimeOriginal",type:"Ascii"},37522:{name:"SubSecTimeDigitized",type:"Ascii"},40960:{name:"FlashpixVersion",type:"Undefined"},40961:{name:"ColorSpace",type:"Short"},40962:{name:"PixelXDimension",type:"Long"},40963:{name:"PixelYDimension",type:"Long"},40964:{name:"RelatedSoundFile",type:"Ascii"},40965:{name:"InteroperabilityTag",type:"Long"},41483:{name:"FlashEnergy",type:"Rational"},41484:{name:"SpatialFrequencyResponse",type:"Undefined"},41486:{name:"FocalPlaneXResolution",type:"Rational"},41487:{name:"FocalPlaneYResolution",type:"Rational"},41488:{name:"FocalPlaneResolutionUnit",type:"Short"},41492:{name:"SubjectLocation",type:"Short"},41493:{name:"ExposureIndex",type:"Rational"},41495:{name:"SensingMethod",type:"Short"},41728:{name:"FileSource",type:"Undefined"},41729:{name:"SceneType",type:"Undefined"},41730:{name:"CFAPattern",type:"Undefined"},41985:{name:"CustomRendered",type:"Short"},41986:{name:"ExposureMode",type:"Short"},41987:{name:"WhiteBalance",type:"Short"},41988:{name:"DigitalZoomRatio",type:"Rational"},41989:{name:"FocalLengthIn35mmFilm",type:"Short"},41990:{name:"SceneCaptureType",type:"Short"},41991:{name:"GainControl",type:"Short"},41992:{name:"Contrast",type:"Short"},41993:{name:"Saturation",type:"Short"},41994:{name:"Sharpness",type:"Short"},41995:{name:"DeviceSettingDescription",type:"Undefined"},41996:{name:"SubjectDistanceRange",type:"Short"},42016:{name:"ImageUniqueID",type:"Ascii"},42032:{name:"CameraOwnerName",type:"Ascii"},42033:{name:"BodySerialNumber",type:"Ascii"},42034:{name:"LensSpecification",type:"Rational"},42035:{name:"LensMake",type:"Ascii"},42036:{name:"LensModel",type:"Ascii"},42037:{name:"LensSerialNumber",type:"Ascii"},42240:{name:"Gamma",type:"Rational"}},GPS:{0:{name:"GPSVersionID",type:"Byte"},1:{name:"GPSLatitudeRef",type:"Ascii"},2:{name:"GPSLatitude",type:"Rational"},3:{name:"GPSLongitudeRef",type:"Ascii"},4:{name:"GPSLongitude",type:"Rational"},5:{name:"GPSAltitudeRef",type:"Byte"},6:{name:"GPSAltitude",type:"Rational"},7:{name:"GPSTimeStamp",type:"Rational"},8:{name:"GPSSatellites",type:"Ascii"},9:{name:"GPSStatus",type:"Ascii"},10:{name:"GPSMeasureMode",type:"Ascii"},11:{name:"GPSDOP",type:"Rational"},12:{name:"GPSSpeedRef",type:"Ascii"},13:{name:"GPSSpeed",type:"Rational"},14:{name:"GPSTrackRef",type:"Ascii"},15:{name:"GPSTrack",type:"Rational"},16:{name:"GPSImgDirectionRef",type:"Ascii"},17:{name:"GPSImgDirection",type:"Rational"},18:{name:"GPSMapDatum",type:"Ascii"},19:{name:"GPSDestLatitudeRef",type:"Ascii"},20:{name:"GPSDestLatitude",type:"Rational"},21:{name:"GPSDestLongitudeRef",type:"Ascii"},22:{name:"GPSDestLongitude",type:"Rational"},23:{name:"GPSDestBearingRef",type:"Ascii"},24:{name:"GPSDestBearing",type:"Rational"},25:{name:"GPSDestDistanceRef",type:"Ascii"},26:{name:"GPSDestDistance",type:"Rational"},27:{name:"GPSProcessingMethod",type:"Undefined"},28:{name:"GPSAreaInformation",type:"Undefined"},29:{name:"GPSDateStamp",type:"Ascii"},30:{name:"GPSDifferential",type:"Short"},31:{name:"GPSHPositioningError",type:"Rational"}},Interop:{1:{name:"InteroperabilityIndex",type:"Ascii"}}};f["0th"]=f.Image,f["1st"]=f.Image,e.TAGS=f,e.ImageIFD={ProcessingSoftware:11,NewSubfileType:254,SubfileType:255,ImageWidth:256,ImageLength:257,BitsPerSample:258,Compression:259,PhotometricInterpretation:262,Threshholding:263,CellWidth:264,CellLength:265,FillOrder:266,DocumentName:269,ImageDescription:270,Make:271,Model:272,StripOffsets:273,Orientation:274,SamplesPerPixel:277,RowsPerStrip:278,StripByteCounts:279,XResolution:282,YResolution:283,PlanarConfiguration:284,GrayResponseUnit:290,GrayResponseCurve:291,T4Options:292,T6Options:293,ResolutionUnit:296,TransferFunction:301,Software:305,DateTime:306,Artist:315,HostComputer:316,Predictor:317,WhitePoint:318,PrimaryChromaticities:319,ColorMap:320,HalftoneHints:321,TileWidth:322,TileLength:323,TileOffsets:324,TileByteCounts:325,SubIFDs:330,InkSet:332,InkNames:333,NumberOfInks:334,DotRange:336,TargetPrinter:337,ExtraSamples:338,SampleFormat:339,SMinSampleValue:340,SMaxSampleValue:341,TransferRange:342,ClipPath:343,XClipPathUnits:344,YClipPathUnits:345,Indexed:346,JPEGTables:347,OPIProxy:351,JPEGProc:512,JPEGInterchangeFormat:513,JPEGInterchangeFormatLength:514,JPEGRestartInterval:515,JPEGLosslessPredictors:517,JPEGPointTransforms:518,JPEGQTables:519,JPEGDCTables:520,JPEGACTables:521,YCbCrCoefficients:529,YCbCrSubSampling:530,YCbCrPositioning:531,ReferenceBlackWhite:532,XMLPacket:700,Rating:18246,RatingPercent:18249,ImageID:32781,CFARepeatPatternDim:33421,CFAPattern:33422,BatteryLevel:33423,Copyright:33432,ExposureTime:33434,ImageResources:34377,ExifTag:34665,InterColorProfile:34675,GPSTag:34853,Interlace:34857,TimeZoneOffset:34858,SelfTimerMode:34859,FlashEnergy:37387,SpatialFrequencyResponse:37388,Noise:37389,FocalPlaneXResolution:37390,FocalPlaneYResolution:37391,FocalPlaneResolutionUnit:37392,ImageNumber:37393,SecurityClassification:37394,ImageHistory:37395,ExposureIndex:37397,TIFFEPStandardID:37398,SensingMethod:37399,XPTitle:40091,XPComment:40092,XPAuthor:40093,XPKeywords:40094,XPSubject:40095,PrintImageMatching:50341,DNGVersion:50706,DNGBackwardVersion:50707,UniqueCameraModel:50708,LocalizedCameraModel:50709,CFAPlaneColor:50710,CFALayout:50711,LinearizationTable:50712,BlackLevelRepeatDim:50713,BlackLevel:50714,BlackLevelDeltaH:50715,BlackLevelDeltaV:50716,WhiteLevel:50717,DefaultScale:50718,DefaultCropOrigin:50719,DefaultCropSize:50720,ColorMatrix1:50721,ColorMatrix2:50722,CameraCalibration1:50723,CameraCalibration2:50724,ReductionMatrix1:50725,ReductionMatrix2:50726,AnalogBalance:50727,AsShotNeutral:50728,AsShotWhiteXY:50729,BaselineExposure:50730,BaselineNoise:50731,BaselineSharpness:50732,BayerGreenSplit:50733,LinearResponseLimit:50734,CameraSerialNumber:50735,LensInfo:50736,ChromaBlurRadius:50737,AntiAliasStrength:50738,ShadowScale:50739,DNGPrivateData:50740,MakerNoteSafety:50741,CalibrationIlluminant1:50778,CalibrationIlluminant2:50779,BestQualityScale:50780,RawDataUniqueID:50781,OriginalRawFileName:50827,OriginalRawFileData:50828,ActiveArea:50829,MaskedAreas:50830,AsShotICCProfile:50831,AsShotPreProfileMatrix:50832,CurrentICCProfile:50833,CurrentPreProfileMatrix:50834,ColorimetricReference:50879,CameraCalibrationSignature:50931,ProfileCalibrationSignature:50932,AsShotProfileName:50934,NoiseReductionApplied:50935,ProfileName:50936,ProfileHueSatMapDims:50937,ProfileHueSatMapData1:50938,ProfileHueSatMapData2:50939,ProfileToneCurve:50940,ProfileEmbedPolicy:50941,ProfileCopyright:50942,ForwardMatrix1:50964,ForwardMatrix2:50965,PreviewApplicationName:50966,PreviewApplicationVersion:50967,PreviewSettingsName:50968,PreviewSettingsDigest:50969,PreviewColorSpace:50970,PreviewDateTime:50971,RawImageDigest:50972,OriginalRawFileDigest:50973,SubTileBlockSize:50974,RowInterleaveFactor:50975,ProfileLookTableDims:50981,ProfileLookTableData:50982,OpcodeList1:51008,OpcodeList2:51009,OpcodeList3:51022,NoiseProfile:51041},e.ExifIFD={ExposureTime:33434,FNumber:33437,ExposureProgram:34850,SpectralSensitivity:34852,ISOSpeedRatings:34855,OECF:34856,SensitivityType:34864,StandardOutputSensitivity:34865,RecommendedExposureIndex:34866,ISOSpeed:34867,ISOSpeedLatitudeyyy:34868,ISOSpeedLatitudezzz:34869,ExifVersion:36864,DateTimeOriginal:36867,DateTimeDigitized:36868,ComponentsConfiguration:37121,CompressedBitsPerPixel:37122,ShutterSpeedValue:37377,ApertureValue:37378,BrightnessValue:37379,ExposureBiasValue:37380,MaxApertureValue:37381,SubjectDistance:37382,MeteringMode:37383,LightSource:37384,Flash:37385,FocalLength:37386,SubjectArea:37396,MakerNote:37500,UserComment:37510,SubSecTime:37520,SubSecTimeOriginal:37521,SubSecTimeDigitized:37522,FlashpixVersion:40960,ColorSpace:40961,PixelXDimension:40962,PixelYDimension:40963,RelatedSoundFile:40964,InteroperabilityTag:40965,FlashEnergy:41483,SpatialFrequencyResponse:41484,FocalPlaneXResolution:41486,FocalPlaneYResolution:41487,FocalPlaneResolutionUnit:41488,SubjectLocation:41492,ExposureIndex:41493,SensingMethod:41495,FileSource:41728,SceneType:41729,CFAPattern:41730,CustomRendered:41985,ExposureMode:41986,WhiteBalance:41987,DigitalZoomRatio:41988,FocalLengthIn35mmFilm:41989,SceneCaptureType:41990,GainControl:41991,Contrast:41992,Saturation:41993,Sharpness:41994,DeviceSettingDescription:41995,SubjectDistanceRange:41996,ImageUniqueID:42016,CameraOwnerName:42032,BodySerialNumber:42033,LensSpecification:42034,LensMake:42035,LensModel:42036,LensSerialNumber:42037,Gamma:42240},e.GPSIFD={GPSVersionID:0,GPSLatitudeRef:1,GPSLatitude:2,GPSLongitudeRef:3,GPSLongitude:4,GPSAltitudeRef:5,GPSAltitude:6,GPSTimeStamp:7,GPSSatellites:8,GPSStatus:9,GPSMeasureMode:10,GPSDOP:11,GPSSpeedRef:12,GPSSpeed:13,GPSTrackRef:14,GPSTrack:15,GPSImgDirectionRef:16,GPSImgDirection:17,GPSMapDatum:18,GPSDestLatitudeRef:19,GPSDestLatitude:20,GPSDestLongitudeRef:21,GPSDestLongitude:22,GPSDestBearingRef:23,GPSDestBearing:24,GPSDestDistanceRef:25,GPSDestDistance:26,GPSProcessingMethod:27,GPSAreaInformation:28,GPSDateStamp:29,GPSDifferential:30,GPSHPositioningError:31},e.InteropIFD={InteroperabilityIndex:1},e.GPSHelper={degToDmsRational:function(e){let t=Math.abs(e),i=t%1*60,n=i%1*60;return[[Math.floor(t),1],[Math.floor(i),1],[Math.round(100*n),100]]},dmsRationalToDeg:function(e,t){let i="S"===t||"W"===t?-1:1;return(e[0][0]/e[0][1]+e[1][0]/e[1][1]/60+e[2][0]/e[2][1]/3600)*i}},"undefined"!=typeof exports?("undefined"!=typeof module&&module.exports&&(exports=module.exports=e),exports.piexif=e):window.piexif=e}();
\ No newline at end of file
diff --git a/public/acuant/11.9.2/AcuantMetricsService.min.js b/public/acuant/11.9.3.508.1/AcuantMetricsService.min.js
similarity index 100%
rename from public/acuant/11.9.2/AcuantMetricsService.min.js
rename to public/acuant/11.9.3.508.1/AcuantMetricsService.min.js
diff --git a/public/acuant/11.9.2/AcuantMetricsService.wasm b/public/acuant/11.9.3.508.1/AcuantMetricsService.wasm
old mode 100755
new mode 100644
similarity index 100%
rename from public/acuant/11.9.2/AcuantMetricsService.wasm
rename to public/acuant/11.9.3.508.1/AcuantMetricsService.wasm
diff --git a/public/acuant/11.9.2/AcuantMetricsWorker.min.js b/public/acuant/11.9.3.508.1/AcuantMetricsWorker.min.js
similarity index 100%
rename from public/acuant/11.9.2/AcuantMetricsWorker.min.js
rename to public/acuant/11.9.3.508.1/AcuantMetricsWorker.min.js
diff --git a/public/acuant/11.9.3.508.1/AcuantPassiveLiveness.min.js b/public/acuant/11.9.3.508.1/AcuantPassiveLiveness.min.js
new file mode 100644
index 00000000000..7cb94fce257
--- /dev/null
+++ b/public/acuant/11.9.3.508.1/AcuantPassiveLiveness.min.js
@@ -0,0 +1,64983 @@
+! function (e) {
+    function t(t) {
+        for (var n, a, s = t[0], i = t[1], o = 0, l = []; o < s.length; o++) a = s[o], Object.prototype.hasOwnProperty.call(r, a) && r[a] && l.push(r[a][0]), r[a] = 0;
+        for (n in i) Object.prototype.hasOwnProperty.call(i, n) && (e[n] = i[n]);
+        for (u && u(t); l.length;) l.shift()()
+    }
+    var n = {},
+        r = {
+            0: 0
+        };
+
+    function a(t) {
+        if (n[t]) return n[t].exports;
+        var r = n[t] = {
+            i: t,
+            l: !1,
+            exports: {}
+        };
+        return e[t].call(r.exports, r, r.exports, a), r.l = !0, r.exports
+    }
+    a.e = function () {
+        return Promise.resolve()
+    }, a.m = e, a.c = n, a.d = function (e, t, n) {
+        a.o(e, t) || Object.defineProperty(e, t, {
+            enumerable: !0,
+            get: n
+        })
+    }, a.r = function (e) {
+        "undefined" != typeof Symbol && Symbol.toStringTag && Object.defineProperty(e, Symbol.toStringTag, {
+            value: "Module"
+        }), Object.defineProperty(e, "__esModule", {
+            value: !0
+        })
+    }, a.t = function (e, t) {
+        if (1 & t && (e = a(e)), 8 & t) return e;
+        if (4 & t && "object" == typeof e && e && e.__esModule) return e;
+        var n = Object.create(null);
+        if (a.r(n), Object.defineProperty(n, "default", {
+                enumerable: !0,
+                value: e
+            }), 2 & t && "string" != typeof e)
+            for (var r in e) a.d(n, r, function (t) {
+                return e[t]
+            }.bind(null, r));
+        return n
+    }, a.n = function (e) {
+        var t = e && e.__esModule ? function () {
+            return e.default
+        } : function () {
+            return e
+        };
+        return a.d(t, "a", t), t
+    }, a.o = function (e, t) {
+        return Object.prototype.hasOwnProperty.call(e, t)
+    }, a.p = "", a.oe = function (e) {
+        throw console.error(e), e
+    };
+    var s = window.webpackJsonp = window.webpackJsonp || [],
+        i = s.push.bind(s);
+    s.push = t, s = s.slice();
+    for (var o = 0; o < s.length; o++) t(s[o]);
+    var u = i;
+    a(a.s = 129)
+}([function (e, t, n) {
+    "use strict";
+    n.d(t, "a", (function () {
+        return Z
+    })), n.d(t, "b", (function () {
+        return P
+    })), n.d(t, "c", (function () {
+        return L
+    })), n.d(t, "d", (function () {
+        return v
+    })), n.d(t, "e", (function () {
+        return l
+    })), n.d(t, "f", (function () {
+        return te
+    }));
+    let r, a, s = !1;
+    const i = "undefined" != typeof window ? window : {},
+        o = i.document || {
+            head: {}
+        },
+        u = {
+            $flags$: 0,
+            $resourcesUrl$: "",
+            jmp: e => e(),
+            raf: e => requestAnimationFrame(e),
+            ael: (e, t, n, r) => e.addEventListener(t, n, r),
+            rel: (e, t, n, r) => e.removeEventListener(t, n, r),
+            ce: (e, t) => new CustomEvent(e, t)
+        },
+        l = e => Promise.resolve(e),
+        c = (() => {
+            try {
+                return new CSSStyleSheet, "function" == typeof (new CSSStyleSheet).replace
+            } catch (e) {}
+            return !1
+        })(),
+        h = (e, t, n, r) => {
+            n && n.map(([n, r, a]) => {
+                const s = e,
+                    i = p(t, a),
+                    o = d(n);
+                u.ael(s, r, i, o), (t.$rmListeners$ = t.$rmListeners$ || []).push(() => u.rel(s, r, i, o))
+            })
+        },
+        p = (e, t) => n => {
+            try {
+                256 & e.$flags$ ? e.$lazyInstance$[t](n) : (e.$queuedListeners$ = e.$queuedListeners$ || []).push([t, n])
+            } catch (e) {
+                ae(e)
+            }
+        },
+        d = e => 0 != (2 & e),
+        f = new WeakMap,
+        m = e => {
+            const t = e.$cmpMeta$,
+                n = e.$hostElement$,
+                r = t.$flags$,
+                a = (t.$tagName$, () => {}),
+                s = ((e, t, n, r) => {
+                    let a = g(t),
+                        s = oe.get(a);
+                    if (e = 11 === e.nodeType ? e : o, s)
+                        if ("string" == typeof s) {
+                            e = e.head || e;
+                            let t, n = f.get(e);
+                            n || f.set(e, n = new Set), n.has(a) || (t = o.createElement("style"), t.innerHTML = s, e.insertBefore(t, e.querySelector("link")), n && n.add(a))
+                        } else e.adoptedStyleSheets.includes(s) || (e.adoptedStyleSheets = [...e.adoptedStyleSheets, s]);
+                    return a
+                })(n.shadowRoot ? n.shadowRoot : n.getRootNode(), t);
+            10 & r && (n["s-sc"] = s, n.classList.add(s + "-h")), a()
+        },
+        g = (e, t) => "sc-" + e.$tagName$,
+        y = {},
+        b = e => "object" === (e = typeof e) || "function" === e,
+        v = (e, t, ...n) => {
+            let r = null,
+                a = !1,
+                s = !1,
+                i = [];
+            const o = t => {
+                for (let n = 0; n < t.length; n++) r = t[n], Array.isArray(r) ? o(r) : null != r && "boolean" != typeof r && ((a = "function" != typeof e && !b(r)) && (r = String(r)), a && s ? i[i.length - 1].$text$ += r : i.push(a ? x(null, r) : r), s = a)
+            };
+            if (o(n), t) {
+                const e = t.className || t.class;
+                e && (t.class = "object" != typeof e ? e : Object.keys(e).filter(t => e[t]).join(" "))
+            }
+            if ("function" == typeof e) return e(null === t ? {} : t, i, k);
+            const u = x(e, null);
+            return u.$attrs$ = t, i.length > 0 && (u.$children$ = i), u
+        },
+        x = (e, t) => {
+            const n = {
+                $flags$: 0,
+                $tag$: e,
+                $text$: t,
+                $elm$: null,
+                $children$: null,
+                $attrs$: null
+            };
+            return n
+        },
+        w = {},
+        k = {
+            forEach: (e, t) => e.map(S).forEach(t),
+            map: (e, t) => e.map(S).map(t).map(I)
+        },
+        S = e => ({
+            vattrs: e.$attrs$,
+            vchildren: e.$children$,
+            vkey: e.$key$,
+            vname: e.$name$,
+            vtag: e.$tag$,
+            vtext: e.$text$
+        }),
+        I = e => {
+            if ("function" == typeof e.vtag) {
+                const t = Object.assign({}, e.vattrs);
+                return e.vkey && (t.key = e.vkey), e.vname && (t.name = e.vname), v(e.vtag, t, ...e.vchildren || [])
+            }
+            const t = x(e.vtag, e.vtext);
+            return t.$attrs$ = e.vattrs, t.$children$ = e.vchildren, t.$key$ = e.vkey, t.$name$ = e.vname, t
+        },
+        A = (e, t, n, r, a, s) => {
+            if (n !== r) {
+                let o = re(e, t),
+                    l = t.toLowerCase();
+                if ("class" === t) {
+                    const t = e.classList,
+                        a = N(n),
+                        s = N(r);
+                    t.remove(...a.filter(e => e && !s.includes(e))), t.add(...s.filter(e => e && !a.includes(e)))
+                } else if ("style" === t) {
+                    for (const t in n) r && null != r[t] || (t.includes("-") ? e.style.removeProperty(t) : e.style[t] = "");
+                    for (const t in r) n && r[t] === n[t] || (t.includes("-") ? e.style.setProperty(t, r[t]) : e.style[t] = r[t])
+                } else if ("ref" === t) r && r(e);
+                else if (o || "o" !== t[0] || "n" !== t[1]) {
+                    const i = b(r);
+                    if ((o || i && null !== r) && !a) try {
+                        if (e.tagName.includes("-")) e[t] = r;
+                        else {
+                            let a = null == r ? "" : r;
+                            "list" === t ? o = !1 : null != n && e[t] == a || (e[t] = a)
+                        }
+                    } catch (e) {}
+                    null == r || !1 === r ? !1 === r && "" !== e.getAttribute(t) || e.removeAttribute(t) : (!o || 4 & s || a) && !i && (r = !0 === r ? "" : r, e.setAttribute(t, r))
+                } else t = "-" === t[2] ? t.slice(3) : re(i, l) ? l.slice(2) : l[2] + t.slice(3), n && u.rel(e, t, n, !1), r && u.ael(e, t, r, !1)
+            }
+        },
+        E = /\s/,
+        N = e => e ? e.split(E) : [],
+        C = (e, t, n, r) => {
+            const a = 11 === t.$elm$.nodeType && t.$elm$.host ? t.$elm$.host : t.$elm$,
+                s = e && e.$attrs$ || y,
+                i = t.$attrs$ || y;
+            for (r in s) r in i || A(a, r, s[r], void 0, n, t.$flags$);
+            for (r in i) A(a, r, s[r], i[r], n, t.$flags$)
+        },
+        T = (e, t, n, a) => {
+            let s, i, u = t.$children$[n],
+                l = 0;
+            if (null !== u.$text$) s = u.$elm$ = o.createTextNode(u.$text$);
+            else if (s = u.$elm$ = o.createElement(u.$tag$), C(null, u, !1), null != r && s["s-si"] !== r && s.classList.add(s["s-si"] = r), u.$children$)
+                for (l = 0; l < u.$children$.length; ++l) i = T(e, u, l), i && s.appendChild(i);
+            return s
+        },
+        R = (e, t, n, r, s, i) => {
+            let o, u = e;
+            for (u.shadowRoot && u.tagName === a && (u = u.shadowRoot); s <= i; ++s) r[s] && (o = T(null, n, s), o && (r[s].$elm$ = o, u.insertBefore(o, t)))
+        },
+        _ = (e, t, n, r, a) => {
+            for (; t <= n; ++t)(r = e[t]) && (a = r.$elm$, D(r), a.remove())
+        },
+        F = (e, t) => e.$tag$ === t.$tag$,
+        M = (e, t) => {
+            const n = t.$elm$ = e.$elm$,
+                r = e.$children$,
+                a = t.$children$,
+                s = t.$text$;
+            null === s ? (C(e, t, !1), null !== r && null !== a ? ((e, t, n, r) => {
+                let a, s = 0,
+                    i = 0,
+                    o = t.length - 1,
+                    u = t[0],
+                    l = t[o],
+                    c = r.length - 1,
+                    h = r[0],
+                    p = r[c];
+                for (; s <= o && i <= c;) null == u ? u = t[++s] : null == l ? l = t[--o] : null == h ? h = r[++i] : null == p ? p = r[--c] : F(u, h) ? (M(u, h), u = t[++s], h = r[++i]) : F(l, p) ? (M(l, p), l = t[--o], p = r[--c]) : F(u, p) ? (M(u, p), e.insertBefore(u.$elm$, l.$elm$.nextSibling), u = t[++s], p = r[--c]) : F(l, h) ? (M(l, h), e.insertBefore(l.$elm$, u.$elm$), l = t[--o], h = r[++i]) : (a = T(t && t[i], n, i), h = r[++i], a && u.$elm$.parentNode.insertBefore(a, u.$elm$));
+                s > o ? R(e, null == r[c + 1] ? null : r[c + 1].$elm$, n, r, i, c) : i > c && _(t, s, o)
+            })(n, r, t, a) : null !== a ? (null !== e.$text$ && (n.textContent = ""), R(n, null, t, a, 0, a.length - 1)) : null !== r && _(r, 0, r.length - 1)) : e.$text$ !== s && (n.data = s)
+        },
+        D = e => {
+            e.$attrs$ && e.$attrs$.ref && e.$attrs$.ref(null), e.$children$ && e.$children$.map(D)
+        },
+        O = (e, t) => {
+            const n = e.$hostElement$,
+                s = e.$vnode$ || x(null, null),
+                i = (o = t) && o.$tag$ === w ? t : v(null, null, t);
+            var o;
+            a = n.tagName, i.$tag$ = null, i.$flags$ |= 4, e.$vnode$ = i, i.$elm$ = s.$elm$ = n.shadowRoot || n, r = n["s-sc"], M(s, i)
+        },
+        L = e => ee(e).$hostElement$,
+        P = (e, t, n) => {
+            const r = L(e);
+            return {
+                emit: e => $(r, t, {
+                    bubbles: !!(4 & n),
+                    composed: !!(2 & n),
+                    cancelable: !!(1 & n),
+                    detail: e
+                })
+            }
+        },
+        $ = (e, t, n) => {
+            const r = u.ce(t, n);
+            return e.dispatchEvent(r), r
+        },
+        B = (e, t) => {
+            t && !e.$onRenderResolve$ && t["s-p"] && t["s-p"].push(new Promise(t => e.$onRenderResolve$ = t))
+        },
+        z = (e, t) => {
+            if (e.$flags$ |= 16, 4 & e.$flags$) return void(e.$flags$ |= 512);
+            B(e, e.$ancestorComponent$);
+            return fe(() => U(e, t))
+        },
+        U = (e, t) => {
+            const n = (e.$cmpMeta$.$tagName$, () => {}),
+                r = e.$lazyInstance$;
+            let a;
+            return t && (e.$flags$ |= 256, e.$queuedListeners$ && (e.$queuedListeners$.map(([e, t]) => H(r, e, t)), e.$queuedListeners$ = null), a = H(r, "componentWillLoad")), n(), q(a, () => W(e, r, t))
+        },
+        W = async (e, t, n) => {
+            const r = e.$hostElement$,
+                a = (e.$cmpMeta$.$tagName$, () => {}),
+                s = r["s-rc"];
+            n && m(e);
+            const i = (e.$cmpMeta$.$tagName$, () => {});
+            V(e, t), s && (s.map(e => e()), r["s-rc"] = void 0), i(), a(); {
+                const t = r["s-p"],
+                    n = () => j(e);
+                0 === t.length ? n() : (Promise.all(t).then(n), e.$flags$ |= 4, t.length = 0)
+            }
+        }, V = (e, t, n) => {
+            try {
+                t = t.render(), e.$flags$ &= -17, e.$flags$ |= 2, O(e, t)
+        
+                setTimeout(() => {
+                    //alert("setting focus")
+                    let shadowRoot = document.querySelector("#acuant-face-capture-camera").shadowRoot;
+                    let manCap = shadowRoot.querySelector("#manualCaptureSelfieButton");
+                    let acceptSelfie = shadowRoot.querySelector("#acceptSelfieButton");
+                    if(manCap){
+                        manCap.focus();
+                    }
+                    if(acceptSelfie){
+                        acceptSelfie.focus();
+                        shadowRoot.querySelector("#retakeButton").tabIndex = 2;
+                    }
+                }, 1500);
+            } catch (t) {
+                ae(t, e.$hostElement$)
+            }
+            return null
+        }, j = e => {
+            e.$cmpMeta$.$tagName$;
+            const t = e.$hostElement$,
+                n = () => {},
+                r = e.$lazyInstance$,
+                a = e.$ancestorComponent$;
+            64 & e.$flags$ ? (H(r, "componentDidUpdate"), n()) : (e.$flags$ |= 64, K(t), H(r, "componentDidLoad"), n(), e.$onReadyResolve$(t), a || G()), e.$onRenderResolve$ && (e.$onRenderResolve$(), e.$onRenderResolve$ = void 0), 512 & e.$flags$ && de(() => z(e, !1)), e.$flags$ &= -517
+        }, G = e => {
+            K(o.documentElement), de(() => $(i, "appload", {
+                detail: {
+                    namespace: "fas-web-ui-component-camera"
+                }
+            }))
+        }, H = (e, t, n) => {
+            if (e && e[t]) try {
+                return e[t](n)
+            } catch (e) {
+                ae(e)
+            }
+        }, q = (e, t) => e && e.then ? e.then(t) : t(), K = e => e.classList.add("hydrated"), Y = (e, t, n, r) => {
+            const a = ee(e),
+                s = a.$instanceValues$.get(t),
+                i = a.$flags$,
+                o = a.$lazyInstance$;
+            var u, l;
+            u = n, l = r.$members$[t][0], n = null == u || b(u) ? u : 4 & l ? "false" !== u && ("" === u || !!u) : 2 & l ? parseFloat(u) : 1 & l ? String(u) : u, 8 & i && void 0 !== s || n === s || (a.$instanceValues$.set(t, n), o && 2 == (18 & i) && z(a, !1))
+        }, X = (e, t, n) => {
+            if (t.$members$) {
+                const r = Object.entries(t.$members$),
+                    a = e.prototype;
+                if (r.map(([e, [r]]) => {
+                        (31 & r || 2 & n && 32 & r) && Object.defineProperty(a, e, {
+                            get() {
+                                return t = e, ee(this).$instanceValues$.get(t);
+                                var t
+                            },
+                            set(n) {
+                                Y(this, e, n, t)
+                            },
+                            configurable: !0,
+                            enumerable: !0
+                        })
+                    }), 1 & n) {
+                    const t = new Map;
+                    a.attributeChangedCallback = function (e, n, r) {
+                        u.jmp(() => {
+                            const n = t.get(e);
+                            if (this.hasOwnProperty(n)) r = this[n], delete this[n];
+                            else if (a.hasOwnProperty(n) && "number" == typeof this[n] && this[n] == r) return;
+                            this[n] = (null !== r || "boolean" != typeof this[n]) && r
+                        })
+                    }, e.observedAttributes = r.filter(([e, t]) => 15 & t[0]).map(([e, n]) => {
+                        const r = n[1] || e;
+                        return t.set(r, e), r
+                    })
+                }
+            }
+            return e
+        }, Q = async (e, t, n, r, a) => {
+            if (0 == (32 & t.$flags$)) {
+                {
+                    if (t.$flags$ |= 32, (a = ie(n)).then) {
+                        const e = () => {};
+                        a = await a, e()
+                    }
+                    a.isProxied || (X(a, n, 2), a.isProxied = !0);
+                    const e = (n.$tagName$, () => {});
+                    t.$flags$ |= 8;
+                    try {
+                        new a(t)
+                    } catch (e) {
+                        ae(e)
+                    }
+                    t.$flags$ &= -9, e()
+                }
+                if (a.style) {
+                    let e = a.style;
+                    const t = g(n);
+                    if (!oe.has(t)) {
+                        const r = (n.$tagName$, () => {});
+                        ((e, t, n) => {
+                            let r = oe.get(e);
+                            c && n ? (r = r || new CSSStyleSheet, r.replace(t)) : r = t, oe.set(e, r)
+                        })(t, e, !!(1 & n.$flags$)), r()
+                    }
+                }
+            }
+            const s = t.$ancestorComponent$,
+                i = () => z(t, !0);
+            s && s["s-rc"] ? s["s-rc"].push(i) : i()
+        }, Z = (e, t = {}) => {
+            const n = () => {},
+                r = [],
+                a = t.exclude || [],
+                s = i.customElements,
+                l = o.head,
+                c = l.querySelector("meta[charset]"),
+                p = o.createElement("style"),
+                d = [];
+            let f, m = !0;
+            Object.assign(u, t), u.$resourcesUrl$ = new URL(t.resourcesUrl || "./", o.baseURI).href, e.map(e => {
+                e[1].map(t => {
+                    const n = {
+                        $flags$: t[0],
+                        $tagName$: t[1],
+                        $members$: t[2],
+                        $listeners$: t[3]
+                    };
+                    n.$members$ = t[2], n.$listeners$ = t[3];
+                    const i = n.$tagName$,
+                        o = class extends HTMLElement {
+                            constructor(e) {
+                                super(e), ne(e = this, n), 1 & n.$flags$ && e.attachShadow({
+                                    mode: "open"
+                                })
+                            }
+                            connectedCallback() {
+                                f && (clearTimeout(f), f = null), m ? d.push(this) : u.jmp(() => (e => {
+                                    if (0 == (1 & u.$flags$)) {
+                                        const t = ee(e),
+                                            n = t.$cmpMeta$,
+                                            r = (n.$tagName$, () => {});
+                                        if (1 & t.$flags$) h(e, t, n.$listeners$);
+                                        else {
+                                            t.$flags$ |= 1; {
+                                                let n = e;
+                                                for (; n = n.parentNode || n.host;)
+                                                    if (n["s-p"]) {
+                                                        B(t, t.$ancestorComponent$ = n);
+                                                        break
+                                                    }
+                                            }
+                                            n.$members$ && Object.entries(n.$members$).map(([t, [n]]) => {
+                                                if (31 & n && e.hasOwnProperty(t)) {
+                                                    const n = e[t];
+                                                    delete e[t], e[t] = n
+                                                }
+                                            }), Q(0, t, n)
+                                        }
+                                        r()
+                                    }
+                                })(this))
+                            }
+                            disconnectedCallback() {
+                                u.jmp(() => (e => {
+                                    if (0 == (1 & u.$flags$)) {
+                                        const t = ee(e);
+                                        t.$rmListeners$ && (t.$rmListeners$.map(e => e()), t.$rmListeners$ = void 0)
+                                    }
+                                })(this))
+                            }
+                            componentOnReady() {
+                                return ee(this).$onReadyPromise$
+                            }
+                        };
+                    n.$lazyBundleId$ = e[0], a.includes(i) || s.get(i) || (r.push(i), s.define(i, X(o, n, 1)))
+                })
+            }), p.innerHTML = r + "{visibility:hidden}.hydrated{visibility:inherit}", p.setAttribute("data-styles", ""), l.insertBefore(p, c ? c.nextSibling : l.firstChild), m = !1, d.length ? d.map(e => e.connectedCallback()) : u.jmp(() => f = setTimeout(G, 30)), n()
+        }, J = new WeakMap, ee = e => J.get(e), te = (e, t) => J.set(t.$lazyInstance$ = e, t), ne = (e, t) => {
+            const n = {
+                $flags$: 0,
+                $hostElement$: e,
+                $cmpMeta$: t,
+                $instanceValues$: new Map
+            };
+            return n.$onReadyPromise$ = new Promise(e => n.$onReadyResolve$ = e), e["s-p"] = [], e["s-rc"] = [], h(e, n, t.$listeners$), J.set(e, n)
+        }, re = (e, t) => t in e, ae = (e, t) => (0, console.error)(e, t), se = new Map, ie = (e, t, r) => {
+            const a = e.$tagName$.replace(/-/g, "_"),
+                s = e.$lazyBundleId$,
+                i = se.get(s);
+            return i ? i[a] : n(239)(`./${s}.entry.js`).then(e => (se.set(s, e), e[a]), ae)
+        }, oe = new Map, ue = [], le = [], ce = (e, t) => n => {
+            e.push(n), s || (s = !0, t && 4 & u.$flags$ ? de(pe) : u.raf(pe))
+        }, he = e => {
+            for (let t = 0; t < e.length; t++) try {
+                e[t](performance.now())
+            } catch (e) {
+                ae(e)
+            }
+            e.length = 0
+        }, pe = () => {
+            he(ue), he(le), (s = ue.length > 0) && u.raf(pe)
+        }, de = e => l().then(e), fe = ce(le, !0)
+}, function (e, t, n) {
+    (function (t) {
+        var n = function (e) {
+            return e && e.Math == Math && e
+        };
+        e.exports = n("object" == typeof globalThis && globalThis) || n("object" == typeof window && window) || n("object" == typeof self && self) || n("object" == typeof t && t) || function () {
+            return this
+        }() || Function("return this")()
+    }).call(this, n(36))
+}, function (e, t) {
+    e.exports = function (e) {
+        try {
+            return !!e()
+        } catch (e) {
+            return !0
+        }
+    }
+}, function (e, t, n) {
+    var r = n(53),
+        a = Function.prototype,
+        s = a.call,
+        i = r && a.bind.bind(s, s);
+    e.exports = r ? i : function (e) {
+        return function () {
+            return s.apply(e, arguments)
+        }
+    }
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(43).f,
+        s = n(23),
+        i = n(16),
+        o = n(70),
+        u = n(96),
+        l = n(98);
+    e.exports = function (e, t) {
+        var n, c, h, p, d, f = e.target,
+            m = e.global,
+            g = e.stat;
+        if (n = m ? r : g ? r[f] || o(f, {}) : (r[f] || {}).prototype)
+            for (c in t) {
+                if (p = t[c], h = e.dontCallGetSet ? (d = a(n, c)) && d.value : n[c], !l(m ? c : f + (g ? "." : "#") + c, e.forced) && void 0 !== h) {
+                    if (typeof p == typeof h) continue;
+                    u(p, h)
+                }(e.sham || h && h.sham) && s(p, "sham", !0), i(n, c, p, e)
+            }
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r, a, s, i = n(116),
+        o = n(9),
+        u = n(1),
+        l = n(6),
+        c = n(11),
+        h = n(10),
+        p = n(24),
+        d = n(28),
+        f = n(23),
+        m = n(16),
+        g = n(33),
+        y = n(27),
+        b = n(52),
+        v = n(32),
+        x = n(7),
+        w = n(55),
+        k = n(21),
+        S = k.enforce,
+        I = k.get,
+        A = u.Int8Array,
+        E = A && A.prototype,
+        N = u.Uint8ClampedArray,
+        C = N && N.prototype,
+        T = A && b(A),
+        R = E && b(E),
+        _ = Object.prototype,
+        F = u.TypeError,
+        M = x("toStringTag"),
+        D = w("TYPED_ARRAY_TAG"),
+        O = i && !!v && "Opera" !== p(u.opera),
+        L = !1,
+        P = {
+            Int8Array: 1,
+            Uint8Array: 1,
+            Uint8ClampedArray: 1,
+            Int16Array: 2,
+            Uint16Array: 2,
+            Int32Array: 4,
+            Uint32Array: 4,
+            Float32Array: 4,
+            Float64Array: 8
+        },
+        $ = {
+            BigInt64Array: 8,
+            BigUint64Array: 8
+        },
+        B = function (e) {
+            var t = b(e);
+            if (c(t)) {
+                var n = I(t);
+                return n && h(n, "TypedArrayConstructor") ? n.TypedArrayConstructor : B(t)
+            }
+        },
+        z = function (e) {
+            if (!c(e)) return !1;
+            var t = p(e);
+            return h(P, t) || h($, t)
+        };
+    for (r in P)(s = (a = u[r]) && a.prototype) ? S(s).TypedArrayConstructor = a : O = !1;
+    for (r in $)(s = (a = u[r]) && a.prototype) && (S(s).TypedArrayConstructor = a);
+    if ((!O || !l(T) || T === Function.prototype) && (T = function () {
+            throw F("Incorrect invocation")
+        }, O))
+        for (r in P) u[r] && v(u[r], T);
+    if ((!O || !R || R === _) && (R = T.prototype, O))
+        for (r in P) u[r] && v(u[r].prototype, R);
+    if (O && b(C) !== R && v(C, R), o && !h(R, M))
+        for (r in L = !0, g(R, M, {
+                configurable: !0,
+                get: function () {
+                    return c(this) ? this[D] : void 0
+                }
+            }), P) u[r] && f(u[r], D, r);
+    e.exports = {
+        NATIVE_ARRAY_BUFFER_VIEWS: O,
+        TYPED_ARRAY_TAG: L && D,
+        aTypedArray: function (e) {
+            if (z(e)) return e;
+            throw F("Target is not a typed array")
+        },
+        aTypedArrayConstructor: function (e) {
+            if (l(e) && (!v || y(T, e))) return e;
+            throw F(d(e) + " is not a typed array constructor")
+        },
+        exportTypedArrayMethod: function (e, t, n, r) {
+            if (o) {
+                if (n)
+                    for (var a in P) {
+                        var s = u[a];
+                        if (s && h(s.prototype, e)) try {
+                            delete s.prototype[e]
+                        } catch (n) {
+                            try {
+                                s.prototype[e] = t
+                            } catch (e) {}
+                        }
+                    }
+                R[e] && !n || m(R, e, n ? t : O && E[e] || t, r)
+            }
+        },
+        exportTypedArrayStaticMethod: function (e, t, n) {
+            var r, a;
+            if (o) {
+                if (v) {
+                    if (n)
+                        for (r in P)
+                            if ((a = u[r]) && h(a, e)) try {
+                                delete a[e]
+                            } catch (e) {}
+                    if (T[e] && !n) return;
+                    try {
+                        return m(T, e, n ? t : O && T[e] || t)
+                    } catch (e) {}
+                }
+                for (r in P) !(a = u[r]) || a[e] && !n || m(a, e, t)
+            }
+        },
+        getTypedArrayConstructor: B,
+        isView: function (e) {
+            if (!c(e)) return !1;
+            var t = p(e);
+            return "DataView" === t || h(P, t) || h($, t)
+        },
+        isTypedArray: z,
+        TypedArray: T,
+        TypedArrayPrototype: R
+    }
+}, function (e, t, n) {
+    var r = n(90),
+        a = r.all;
+    e.exports = r.IS_HTMLDDA ? function (e) {
+        return "function" == typeof e || e === a
+    } : function (e) {
+        return "function" == typeof e
+    }
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(37),
+        s = n(10),
+        i = n(55),
+        o = n(38),
+        u = n(89),
+        l = r.Symbol,
+        c = a("wks"),
+        h = u ? l.for || l : l && l.withoutSetter || i;
+    e.exports = function (e) {
+        return s(c, e) || (c[e] = o && s(l, e) ? l[e] : h("Symbol." + e)), c[e]
+    }
+}, function (e, t, n) {
+    var r = n(53),
+        a = Function.prototype.call;
+    e.exports = r ? a.bind(a) : function () {
+        return a.apply(a, arguments)
+    }
+}, function (e, t, n) {
+    var r = n(2);
+    e.exports = !r((function () {
+        return 7 != Object.defineProperty({}, 1, {
+            get: function () {
+                return 7
+            }
+        })[1]
+    }))
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(15),
+        s = r({}.hasOwnProperty);
+    e.exports = Object.hasOwn || function (e, t) {
+        return s(a(e), t)
+    }
+}, function (e, t, n) {
+    var r = n(6),
+        a = n(90),
+        s = a.all;
+    e.exports = a.IS_HTMLDDA ? function (e) {
+        return "object" == typeof e ? null !== e : r(e) || e === s
+    } : function (e) {
+        return "object" == typeof e ? null !== e : r(e)
+    }
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(91),
+        s = n(92),
+        i = n(13),
+        o = n(41),
+        u = TypeError,
+        l = Object.defineProperty,
+        c = Object.getOwnPropertyDescriptor;
+    t.f = r ? s ? function (e, t, n) {
+        if (i(e), t = o(t), i(n), "function" == typeof e && "prototype" === t && "value" in n && "writable" in n && !n.writable) {
+            var r = c(e, t);
+            r && r.writable && (e[t] = n.value, n = {
+                configurable: "configurable" in n ? n.configurable : r.configurable,
+                enumerable: "enumerable" in n ? n.enumerable : r.enumerable,
+                writable: !1
+            })
+        }
+        return l(e, t, n)
+    } : l : function (e, t, n) {
+        if (i(e), t = o(t), i(n), a) try {
+            return l(e, t, n)
+        } catch (e) {}
+        if ("get" in n || "set" in n) throw u("Accessors not supported");
+        return "value" in n && (e[t] = n.value), e
+    }
+}, function (e, t, n) {
+    var r = n(11),
+        a = String,
+        s = TypeError;
+    e.exports = function (e) {
+        if (r(e)) return e;
+        throw s(a(e) + " is not an object")
+    }
+}, function (e, t, n) {
+    var r = n(40);
+    e.exports = function (e) {
+        return r(e.length)
+    }
+}, function (e, t, n) {
+    var r = n(71),
+        a = Object;
+    e.exports = function (e) {
+        return a(r(e))
+    }
+}, function (e, t, n) {
+    var r = n(6),
+        a = n(12),
+        s = n(94),
+        i = n(70);
+    e.exports = function (e, t, n, o) {
+        o || (o = {});
+        var u = o.enumerable,
+            l = void 0 !== o.name ? o.name : t;
+        if (r(n) && s(n, l, o), o.global) u ? e[t] = n : i(t, n);
+        else {
+            try {
+                o.unsafe ? e[t] && (u = !0) : delete e[t]
+            } catch (e) {}
+            u ? e[t] = n : a.f(e, t, {
+                value: n,
+                enumerable: !1,
+                configurable: !o.nonConfigurable,
+                writable: !o.nonWritable
+            })
+        }
+        return e
+    }
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(6),
+        s = function (e) {
+            return a(e) ? e : void 0
+        };
+    e.exports = function (e, t) {
+        return arguments.length < 2 ? s(r[e]) : r[e] && r[e][t]
+    }
+}, function (e, t, n) {
+    var r = n(46),
+        a = n(3),
+        s = n(74),
+        i = n(15),
+        o = n(14),
+        u = n(161),
+        l = a([].push),
+        c = function (e) {
+            var t = 1 == e,
+                n = 2 == e,
+                a = 3 == e,
+                c = 4 == e,
+                h = 6 == e,
+                p = 7 == e,
+                d = 5 == e || h;
+            return function (f, m, g, y) {
+                for (var b, v, x = i(f), w = s(x), k = r(m, g), S = o(w), I = 0, A = y || u, E = t ? A(f, S) : n || p ? A(f, 0) : void 0; S > I; I++)
+                    if ((d || I in w) && (v = k(b = w[I], I, x), e))
+                        if (t) E[I] = v;
+                        else if (v) switch (e) {
+                    case 3:
+                        return !0;
+                    case 5:
+                        return b;
+                    case 6:
+                        return I;
+                    case 2:
+                        l(E, b)
+                } else switch (e) {
+                    case 4:
+                        return !1;
+                    case 7:
+                        l(E, b)
+                }
+                return h ? -1 : a || c ? c : E
+            }
+        };
+    e.exports = {
+        forEach: c(0),
+        map: c(1),
+        filter: c(2),
+        some: c(3),
+        every: c(4),
+        find: c(5),
+        findIndex: c(6),
+        filterReject: c(7)
+    }
+}, function (e, t, n) {
+    var r = n(6),
+        a = n(28),
+        s = TypeError;
+    e.exports = function (e) {
+        if (r(e)) return e;
+        throw s(a(e) + " is not a function")
+    }
+}, function (e, t) {
+    e.exports = !1
+}, function (e, t, n) {
+    var r, a, s, i = n(132),
+        o = n(1),
+        u = n(11),
+        l = n(23),
+        c = n(10),
+        h = n(69),
+        p = n(59),
+        d = n(60),
+        f = o.TypeError,
+        m = o.WeakMap;
+    if (i || h.state) {
+        var g = h.state || (h.state = new m);
+        g.get = g.get, g.has = g.has, g.set = g.set, r = function (e, t) {
+            if (g.has(e)) throw f("Object already initialized");
+            return t.facade = e, g.set(e, t), t
+        }, a = function (e) {
+            return g.get(e) || {}
+        }, s = function (e) {
+            return g.has(e)
+        }
+    } else {
+        var y = p("state");
+        d[y] = !0, r = function (e, t) {
+            if (c(e, y)) throw f("Object already initialized");
+            return t.facade = e, l(e, y, t), t
+        }, a = function (e) {
+            return c(e, y) ? e[y] : {}
+        }, s = function (e) {
+            return c(e, y)
+        }
+    }
+    e.exports = {
+        set: r,
+        get: a,
+        has: s,
+        enforce: function (e) {
+            return s(e) ? a(e) : r(e, {})
+        },
+        getterFor: function (e) {
+            return function (t) {
+                var n;
+                if (!u(t) || (n = a(t)).type !== e) throw f("Incompatible receiver, " + e + " required");
+                return n
+            }
+        }
+    }
+}, function (e, t, n) {
+    var r = n(74),
+        a = n(71);
+    e.exports = function (e) {
+        return r(a(e))
+    }
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(12),
+        s = n(39);
+    e.exports = r ? function (e, t, n) {
+        return a.f(e, t, s(1, n))
+    } : function (e, t, n) {
+        return e[t] = n, e
+    }
+}, function (e, t, n) {
+    var r = n(68),
+        a = n(6),
+        s = n(29),
+        i = n(7)("toStringTag"),
+        o = Object,
+        u = "Arguments" == s(function () {
+            return arguments
+        }());
+    e.exports = r ? s : function (e) {
+        var t, n, r;
+        return void 0 === e ? "Undefined" : null === e ? "Null" : "string" == typeof (n = function (e, t) {
+            try {
+                return e[t]
+            } catch (e) {}
+        }(t = o(e), i)) ? n : u ? s(t) : "Object" == (r = s(t)) && a(t.callee) ? "Arguments" : r
+    }
+}, function (e, t, n) {
+    var r = n(12).f,
+        a = n(10),
+        s = n(7)("toStringTag");
+    e.exports = function (e, t, n) {
+        e && !n && (e = e.prototype), e && !a(e, s) && r(e, s, {
+            configurable: !0,
+            value: t
+        })
+    }
+}, function (e, t) {
+    e.exports = "undefined" != typeof navigator && String(navigator.userAgent) || ""
+}, function (e, t, n) {
+    var r = n(3);
+    e.exports = r({}.isPrototypeOf)
+}, function (e, t) {
+    var n = String;
+    e.exports = function (e) {
+        try {
+            return n(e)
+        } catch (e) {
+            return "Object"
+        }
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = r({}.toString),
+        s = r("".slice);
+    e.exports = function (e) {
+        return s(a(e), 8, -1)
+    }
+}, function (e, t, n) {
+    var r = n(31),
+        a = Math.max,
+        s = Math.min;
+    e.exports = function (e, t) {
+        var n = r(e);
+        return n < 0 ? a(n + t, 0) : s(n, t)
+    }
+}, function (e, t, n) {
+    var r = n(137);
+    e.exports = function (e) {
+        var t = +e;
+        return t != t || 0 === t ? 0 : r(t)
+    }
+}, function (e, t, n) {
+    var r = n(138),
+        a = n(13),
+        s = n(139);
+    e.exports = Object.setPrototypeOf || ("__proto__" in {} ? function () {
+        var e, t = !1,
+            n = {};
+        try {
+            (e = r(Object.prototype, "__proto__", "set"))(n, []), t = n instanceof Array
+        } catch (e) {}
+        return function (n, r) {
+            return a(n), s(r), t ? e(n, r) : n.__proto__ = r, n
+        }
+    }() : void 0)
+}, function (e, t, n) {
+    var r = n(94),
+        a = n(12);
+    e.exports = function (e, t, n) {
+        return n.get && r(n.get, t, {
+            getter: !0
+        }), n.set && r(n.set, t, {
+            setter: !0
+        }), a.f(e, t, n)
+    }
+}, function (e, t, n) {
+    var r = n(24),
+        a = String;
+    e.exports = function (e) {
+        if ("Symbol" === r(e)) throw TypeError("Cannot convert a Symbol value to a string");
+        return a(e)
+    }
+}, function (e, t, n) {
+    var r, a = n(13),
+        s = n(110),
+        i = n(75),
+        o = n(60),
+        u = n(101),
+        l = n(57),
+        c = n(59),
+        h = c("IE_PROTO"),
+        p = function () {},
+        d = function (e) {
+            return "<script>" + e + "<\/script>"
+        },
+        f = function (e) {
+            e.write(d("")), e.close();
+            var t = e.parentWindow.Object;
+            return e = null, t
+        },
+        m = function () {
+            try {
+                r = new ActiveXObject("htmlfile")
+            } catch (e) {}
+            var e, t;
+            m = "undefined" != typeof document ? document.domain && r ? f(r) : ((t = l("iframe")).style.display = "none", u.appendChild(t), t.src = String("javascript:"), (e = t.contentWindow.document).open(), e.write(d("document.F=Object")), e.close(), e.F) : f(r);
+            for (var n = i.length; n--;) delete m.prototype[i[n]];
+            return m()
+        };
+    o[h] = !0, e.exports = Object.create || function (e, t) {
+        var n;
+        return null !== e ? (p.prototype = a(e), n = new p, p.prototype = null, n[h] = e) : n = m(), void 0 === t ? n : s.f(n, t)
+    }
+}, function (e, t) {
+    var n;
+    n = function () {
+        return this
+    }();
+    try {
+        n = n || new Function("return this")()
+    } catch (e) {
+        "object" == typeof window && (n = window)
+    }
+    e.exports = n
+}, function (e, t, n) {
+    var r = n(20),
+        a = n(69);
+    (e.exports = function (e, t) {
+        return a[e] || (a[e] = void 0 !== t ? t : {})
+    })("versions", []).push({
+        version: "3.29.1",
+        mode: r ? "pure" : "global",
+        copyright: "© 2014-2023 Denis Pushkarev (zloirock.ru)",
+        license: "https://github.com/zloirock/core-js/blob/v3.29.1/LICENSE",
+        source: "https://github.com/zloirock/core-js"
+    })
+}, function (e, t, n) {
+    var r = n(56),
+        a = n(2);
+    e.exports = !!Object.getOwnPropertySymbols && !a((function () {
+        var e = Symbol();
+        return !String(e) || !(Object(e) instanceof Symbol) || !Symbol.sham && r && r < 41
+    }))
+}, function (e, t) {
+    e.exports = function (e, t) {
+        return {
+            enumerable: !(1 & e),
+            configurable: !(2 & e),
+            writable: !(4 & e),
+            value: t
+        }
+    }
+}, function (e, t, n) {
+    var r = n(31),
+        a = Math.min;
+    e.exports = function (e) {
+        return e > 0 ? a(r(e), 9007199254740991) : 0
+    }
+}, function (e, t, n) {
+    var r = n(93),
+        a = n(42);
+    e.exports = function (e) {
+        var t = r(e, "string");
+        return a(t) ? t : t + ""
+    }
+}, function (e, t, n) {
+    var r = n(17),
+        a = n(6),
+        s = n(27),
+        i = n(89),
+        o = Object;
+    e.exports = i ? function (e) {
+        return "symbol" == typeof e
+    } : function (e) {
+        var t = r("Symbol");
+        return a(t) && s(t.prototype, o(e))
+    }
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(8),
+        s = n(95),
+        i = n(39),
+        o = n(22),
+        u = n(41),
+        l = n(10),
+        c = n(91),
+        h = Object.getOwnPropertyDescriptor;
+    t.f = r ? h : function (e, t) {
+        if (e = o(e), t = u(t), c) try {
+            return h(e, t)
+        } catch (e) {}
+        if (l(e, t)) return i(!a(s.f, e, t), e[t])
+    }
+}, function (e, t, n) {
+    var r = n(97),
+        a = n(75).concat("length", "prototype");
+    t.f = Object.getOwnPropertyNames || function (e) {
+        return r(e, a)
+    }
+}, function (e, t, n) {
+    var r = n(53),
+        a = Function.prototype,
+        s = a.apply,
+        i = a.call;
+    e.exports = "object" == typeof Reflect && Reflect.apply || (r ? i.bind(s) : function () {
+        return i.apply(s, arguments)
+    })
+}, function (e, t, n) {
+    var r = n(63),
+        a = n(19),
+        s = n(53),
+        i = r(r.bind);
+    e.exports = function (e, t) {
+        return a(e), void 0 === t ? e : s ? i(e, t) : function () {
+            return e.apply(t, arguments)
+        }
+    }
+}, function (e, t, n) {
+    var r = n(3);
+    e.exports = r([].slice)
+}, function (e, t, n) {
+    var r = n(1);
+    e.exports = r.Promise
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(48),
+        s = n(6),
+        i = n(98),
+        o = n(73),
+        u = n(7),
+        l = n(145),
+        c = n(104),
+        h = n(20),
+        p = n(56),
+        d = a && a.prototype,
+        f = u("species"),
+        m = !1,
+        g = s(r.PromiseRejectionEvent),
+        y = i("Promise", (function () {
+            var e = o(a),
+                t = e !== String(a);
+            if (!t && 66 === p) return !0;
+            if (h && (!d.catch || !d.finally)) return !0;
+            if (!p || p < 51 || !/native code/.test(e)) {
+                var n = new a((function (e) {
+                        e(1)
+                    })),
+                    r = function (e) {
+                        e((function () {}), (function () {}))
+                    };
+                if ((n.constructor = {})[f] = r, !(m = n.then((function () {})) instanceof r)) return !0
+            }
+            return !t && (l || c) && !g
+        }));
+    e.exports = {
+        CONSTRUCTOR: y,
+        REJECTION_EVENT: g,
+        SUBCLASSING: m
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(19),
+        a = TypeError,
+        s = function (e) {
+            var t, n;
+            this.promise = new e((function (e, r) {
+                if (void 0 !== t || void 0 !== n) throw a("Bad Promise constructor");
+                t = e, n = r
+            })), this.resolve = r(t), this.reject = r(n)
+        };
+    e.exports.f = function (e) {
+        return new s(e)
+    }
+}, function (e, t) {
+    e.exports = {}
+}, function (e, t, n) {
+    var r = n(10),
+        a = n(6),
+        s = n(15),
+        i = n(59),
+        o = n(114),
+        u = i("IE_PROTO"),
+        l = Object,
+        c = l.prototype;
+    e.exports = o ? l.getPrototypeOf : function (e) {
+        var t = s(e);
+        if (r(t, u)) return t[u];
+        var n = t.constructor;
+        return a(n) && t instanceof n ? n.prototype : t instanceof l ? c : null
+    }
+}, function (e, t, n) {
+    var r = n(2);
+    e.exports = !r((function () {
+        var e = function () {}.bind();
+        return "function" != typeof e || e.hasOwnProperty("prototype")
+    }))
+}, function (e, t) {
+    e.exports = function (e) {
+        return null == e
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = 0,
+        s = Math.random(),
+        i = r(1..toString);
+    e.exports = function (e) {
+        return "Symbol(" + (void 0 === e ? "" : e) + ")_" + i(++a + s, 36)
+    }
+}, function (e, t, n) {
+    var r, a, s = n(1),
+        i = n(26),
+        o = s.process,
+        u = s.Deno,
+        l = o && o.versions || u && u.version,
+        c = l && l.v8;
+    c && (a = (r = c.split("."))[0] > 0 && r[0] < 4 ? 1 : +(r[0] + r[1])), !a && i && (!(r = i.match(/Edge\/(\d+)/)) || r[1] >= 74) && (r = i.match(/Chrome\/(\d+)/)) && (a = +r[1]), e.exports = a
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(11),
+        s = r.document,
+        i = a(s) && a(s.createElement);
+    e.exports = function (e) {
+        return i ? s.createElement(e) : {}
+    }
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(10),
+        s = Function.prototype,
+        i = r && Object.getOwnPropertyDescriptor,
+        o = a(s, "name"),
+        u = o && "something" === function () {}.name,
+        l = o && (!r || r && i(s, "name").configurable);
+    e.exports = {
+        EXISTS: o,
+        PROPER: u,
+        CONFIGURABLE: l
+    }
+}, function (e, t, n) {
+    var r = n(37),
+        a = n(55),
+        s = r("keys");
+    e.exports = function (e) {
+        return s[e] || (s[e] = a(e))
+    }
+}, function (e, t) {
+    e.exports = {}
+}, function (e, t, n) {
+    var r = n(22),
+        a = n(30),
+        s = n(14),
+        i = function (e) {
+            return function (t, n, i) {
+                var o, u = r(t),
+                    l = s(u),
+                    c = a(i, l);
+                if (e && n != n) {
+                    for (; l > c;)
+                        if ((o = u[c++]) != o) return !0
+                } else
+                    for (; l > c; c++)
+                        if ((e || c in u) && u[c] === n) return e || c || 0;
+                return !e && -1
+            }
+        };
+    e.exports = {
+        includes: i(!0),
+        indexOf: i(!1)
+    }
+}, function (e, t, n) {
+    (function (t) {
+        var r = n(29);
+        e.exports = void 0 !== t && "process" == r(t)
+    }).call(this, n(77))
+}, function (e, t, n) {
+    var r = n(29),
+        a = n(3);
+    e.exports = function (e) {
+        if ("Function" === r(e)) return a(e)
+    }
+}, function (e, t, n) {
+    var r = n(29);
+    e.exports = Array.isArray || function (e) {
+        return "Array" == r(e)
+    }
+}, function (e, t, n) {
+    var r = n(5),
+        a = n(80),
+        s = r.aTypedArrayConstructor,
+        i = r.getTypedArrayConstructor;
+    e.exports = function (e) {
+        return s(a(e, i(e)))
+    }
+}, function (e, t, n) {
+    var r = n(214),
+        a = n(10),
+        s = n(123),
+        i = n(12).f;
+    e.exports = function (e) {
+        var t = r.Symbol || (r.Symbol = {});
+        a(t, e) || i(t, e, {
+            value: s.f(e)
+        })
+    }
+}, function (e, t) {
+    (function (t) {
+        e.exports = t
+    }).call(this, {})
+}, function (e, t, n) {
+    var r = {};
+    r[n(7)("toStringTag")] = "z", e.exports = "[object z]" === String(r)
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(70),
+        s = r["__core-js_shared__"] || a("__core-js_shared__", {});
+    e.exports = s
+}, function (e, t, n) {
+    var r = n(1),
+        a = Object.defineProperty;
+    e.exports = function (e, t) {
+        try {
+            a(r, e, {
+                value: t,
+                configurable: !0,
+                writable: !0
+            })
+        } catch (n) {
+            r[e] = t
+        }
+        return t
+    }
+}, function (e, t, n) {
+    var r = n(54),
+        a = TypeError;
+    e.exports = function (e) {
+        if (r(e)) throw a("Can't call method on " + e);
+        return e
+    }
+}, function (e, t, n) {
+    var r = n(19),
+        a = n(54);
+    e.exports = function (e, t) {
+        var n = e[t];
+        return a(n) ? void 0 : r(n)
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(6),
+        s = n(69),
+        i = r(Function.toString);
+    a(s.inspectSource) || (s.inspectSource = function (e) {
+        return i(e)
+    }), e.exports = s.inspectSource
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(2),
+        s = n(29),
+        i = Object,
+        o = r("".split);
+    e.exports = a((function () {
+        return !i("z").propertyIsEnumerable(0)
+    })) ? function (e) {
+        return "String" == s(e) ? o(e, "") : i(e)
+    } : i
+}, function (e, t) {
+    e.exports = ["constructor", "hasOwnProperty", "isPrototypeOf", "propertyIsEnumerable", "toLocaleString", "toString", "valueOf"]
+}, function (e, t) {
+    t.f = Object.getOwnPropertySymbols
+}, function (e, t) {
+    var n, r, a = e.exports = {};
+
+    function s() {
+        throw new Error("setTimeout has not been defined")
+    }
+
+    function i() {
+        throw new Error("clearTimeout has not been defined")
+    }
+
+    function o(e) {
+        if (n === setTimeout) return setTimeout(e, 0);
+        if ((n === s || !n) && setTimeout) return n = setTimeout, setTimeout(e, 0);
+        try {
+            return n(e, 0)
+        } catch (t) {
+            try {
+                return n.call(null, e, 0)
+            } catch (t) {
+                return n.call(this, e, 0)
+            }
+        }
+    }! function () {
+        try {
+            n = "function" == typeof setTimeout ? setTimeout : s
+        } catch (e) {
+            n = s
+        }
+        try {
+            r = "function" == typeof clearTimeout ? clearTimeout : i
+        } catch (e) {
+            r = i
+        }
+    }();
+    var u, l = [],
+        c = !1,
+        h = -1;
+
+    function p() {
+        c && u && (c = !1, u.length ? l = u.concat(l) : h = -1, l.length && d())
+    }
+
+    function d() {
+        if (!c) {
+            var e = o(p);
+            c = !0;
+            for (var t = l.length; t;) {
+                for (u = l, l = []; ++h < t;) u && u[h].run();
+                h = -1, t = l.length
+            }
+            u = null, c = !1,
+                function (e) {
+                    if (r === clearTimeout) return clearTimeout(e);
+                    if ((r === i || !r) && clearTimeout) return r = clearTimeout, clearTimeout(e);
+                    try {
+                        r(e)
+                    } catch (t) {
+                        try {
+                            return r.call(null, e)
+                        } catch (t) {
+                            return r.call(this, e)
+                        }
+                    }
+                }(e)
+        }
+    }
+
+    function f(e, t) {
+        this.fun = e, this.array = t
+    }
+
+    function m() {}
+    a.nextTick = function (e) {
+        var t = new Array(arguments.length - 1);
+        if (arguments.length > 1)
+            for (var n = 1; n < arguments.length; n++) t[n - 1] = arguments[n];
+        l.push(new f(e, t)), 1 !== l.length || c || o(d)
+    }, f.prototype.run = function () {
+        this.fun.apply(null, this.array)
+    }, a.title = "browser", a.browser = !0, a.env = {}, a.argv = [], a.version = "", a.versions = {}, a.on = m, a.addListener = m, a.once = m, a.off = m, a.removeListener = m, a.removeAllListeners = m, a.emit = m, a.prependListener = m, a.prependOnceListener = m, a.listeners = function (e) {
+        return []
+    }, a.binding = function (e) {
+        throw new Error("process.binding is not supported")
+    }, a.cwd = function () {
+        return "/"
+    }, a.chdir = function (e) {
+        throw new Error("process.chdir is not supported")
+    }, a.umask = function () {
+        return 0
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(17),
+        a = n(33),
+        s = n(7),
+        i = n(9),
+        o = s("species");
+    e.exports = function (e) {
+        var t = r(e);
+        i && t && !t[o] && a(t, o, {
+            configurable: !0,
+            get: function () {
+                return this
+            }
+        })
+    }
+}, function (e, t, n) {
+    var r = n(27),
+        a = TypeError;
+    e.exports = function (e, t) {
+        if (r(t, e)) return e;
+        throw a("Incorrect invocation")
+    }
+}, function (e, t, n) {
+    var r = n(13),
+        a = n(99),
+        s = n(54),
+        i = n(7)("species");
+    e.exports = function (e, t) {
+        var n, o = r(e).constructor;
+        return void 0 === o || s(n = r(o)[i]) ? t : a(n)
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(2),
+        s = n(6),
+        i = n(24),
+        o = n(17),
+        u = n(73),
+        l = function () {},
+        c = [],
+        h = o("Reflect", "construct"),
+        p = /^\s*(?:class|function)\b/,
+        d = r(p.exec),
+        f = !p.exec(l),
+        m = function (e) {
+            if (!s(e)) return !1;
+            try {
+                return h(l, c, e), !0
+            } catch (e) {
+                return !1
+            }
+        },
+        g = function (e) {
+            if (!s(e)) return !1;
+            switch (i(e)) {
+                case "AsyncFunction":
+                case "GeneratorFunction":
+                case "AsyncGeneratorFunction":
+                    return !1
+            }
+            try {
+                return f || !!d(p, u(e))
+            } catch (e) {
+                return !0
+            }
+        };
+    g.sham = !0, e.exports = !h || a((function () {
+        var e;
+        return m(m.call) || !m(Object) || !m((function () {
+            e = !0
+        })) || e
+    })) ? g : m
+}, function (e, t) {
+    e.exports = function (e) {
+        try {
+            return {
+                error: !1,
+                value: e()
+            }
+        } catch (e) {
+            return {
+                error: !0,
+                value: e
+            }
+        }
+    }
+}, function (e, t, n) {
+    var r = n(24),
+        a = n(72),
+        s = n(54),
+        i = n(51),
+        o = n(7)("iterator");
+    e.exports = function (e) {
+        if (!s(e)) return a(e, o) || a(e, "@@iterator") || i[r(e)]
+    }
+}, function (e, t, n) {
+    var r = n(97),
+        a = n(75);
+    e.exports = Object.keys || function (e) {
+        return r(e, a)
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(2);
+    e.exports = function (e, t) {
+        var n = [][e];
+        return !!n && r((function () {
+            n.call(null, t || function () {
+                return 1
+            }, 1)
+        }))
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(22),
+        a = n(111),
+        s = n(51),
+        i = n(21),
+        o = n(12).f,
+        u = n(112),
+        l = n(115),
+        c = n(20),
+        h = n(9),
+        p = i.set,
+        d = i.getterFor("Array Iterator");
+    e.exports = u(Array, "Array", (function (e, t) {
+        p(this, {
+            type: "Array Iterator",
+            target: r(e),
+            index: 0,
+            kind: t
+        })
+    }), (function () {
+        var e = d(this),
+            t = e.target,
+            n = e.kind,
+            r = e.index++;
+        return !t || r >= t.length ? (e.target = void 0, l(void 0, !0)) : l("keys" == n ? r : "values" == n ? t[r] : [r, t[r]], !1)
+    }), "values");
+    var f = s.Arguments = s.Array;
+    if (a("keys"), a("values"), a("entries"), !c && h && "values" !== f.name) try {
+        o(f, "name", {
+            value: "values"
+        })
+    } catch (e) {}
+}, function (e, t, n) {
+    "use strict";
+    var r = n(1),
+        a = n(3),
+        s = n(9),
+        i = n(116),
+        o = n(58),
+        u = n(23),
+        l = n(33),
+        c = n(167),
+        h = n(2),
+        p = n(79),
+        d = n(31),
+        f = n(40),
+        m = n(117),
+        g = n(168),
+        y = n(52),
+        b = n(32),
+        v = n(44).f,
+        x = n(118),
+        w = n(88),
+        k = n(25),
+        S = n(21),
+        I = o.PROPER,
+        A = o.CONFIGURABLE,
+        E = S.getterFor("ArrayBuffer"),
+        N = S.getterFor("DataView"),
+        C = S.set,
+        T = r.ArrayBuffer,
+        R = T,
+        _ = R && R.prototype,
+        F = r.DataView,
+        M = F && F.prototype,
+        D = Object.prototype,
+        O = r.Array,
+        L = r.RangeError,
+        P = a(x),
+        $ = a([].reverse),
+        B = g.pack,
+        z = g.unpack,
+        U = function (e) {
+            return [255 & e]
+        },
+        W = function (e) {
+            return [255 & e, e >> 8 & 255]
+        },
+        V = function (e) {
+            return [255 & e, e >> 8 & 255, e >> 16 & 255, e >> 24 & 255]
+        },
+        j = function (e) {
+            return e[3] << 24 | e[2] << 16 | e[1] << 8 | e[0]
+        },
+        G = function (e) {
+            return B(e, 23, 4)
+        },
+        H = function (e) {
+            return B(e, 52, 8)
+        },
+        q = function (e, t, n) {
+            l(e.prototype, t, {
+                configurable: !0,
+                get: function () {
+                    return n(this)[t]
+                }
+            })
+        },
+        K = function (e, t, n, r) {
+            var a = m(n),
+                s = N(e);
+            if (a + t > s.byteLength) throw L("Wrong index");
+            var i = s.bytes,
+                o = a + s.byteOffset,
+                u = w(i, o, o + t);
+            return r ? u : $(u)
+        },
+        Y = function (e, t, n, r, a, s) {
+            var i = m(n),
+                o = N(e);
+            if (i + t > o.byteLength) throw L("Wrong index");
+            for (var u = o.bytes, l = i + o.byteOffset, c = r(+a), h = 0; h < t; h++) u[l + h] = c[s ? h : t - h - 1]
+        };
+    if (i) {
+        var X = I && "ArrayBuffer" !== T.name;
+        if (h((function () {
+                T(1)
+            })) && h((function () {
+                new T(-1)
+            })) && !h((function () {
+                return new T, new T(1.5), new T(NaN), 1 != T.length || X && !A
+            }))) X && A && u(T, "name", "ArrayBuffer");
+        else {
+            (R = function (e) {
+                return p(this, _), new T(m(e))
+            }).prototype = _;
+            for (var Q, Z = v(T), J = 0; Z.length > J;)(Q = Z[J++]) in R || u(R, Q, T[Q]);
+            _.constructor = R
+        }
+        b && y(M) !== D && b(M, D);
+        var ee = new F(new R(2)),
+            te = a(M.setInt8);
+        ee.setInt8(0, 2147483648), ee.setInt8(1, 2147483649), !ee.getInt8(0) && ee.getInt8(1) || c(M, {
+            setInt8: function (e, t) {
+                te(this, e, t << 24 >> 24)
+            },
+            setUint8: function (e, t) {
+                te(this, e, t << 24 >> 24)
+            }
+        }, {
+            unsafe: !0
+        })
+    } else _ = (R = function (e) {
+        p(this, _);
+        var t = m(e);
+        C(this, {
+            type: "ArrayBuffer",
+            bytes: P(O(t), 0),
+            byteLength: t
+        }), s || (this.byteLength = t, this.detached = !1)
+    }).prototype, M = (F = function (e, t, n) {
+        p(this, M), p(e, _);
+        var r = E(e),
+            a = r.byteLength,
+            i = d(t);
+        if (i < 0 || i > a) throw L("Wrong offset");
+        if (i + (n = void 0 === n ? a - i : f(n)) > a) throw L("Wrong length");
+        C(this, {
+            type: "DataView",
+            buffer: e,
+            byteLength: n,
+            byteOffset: i,
+            bytes: r.bytes
+        }), s || (this.buffer = e, this.byteLength = n, this.byteOffset = i)
+    }).prototype, s && (q(R, "byteLength", E), q(F, "buffer", N), q(F, "byteLength", N), q(F, "byteOffset", N)), c(M, {
+        getInt8: function (e) {
+            return K(this, 1, e)[0] << 24 >> 24
+        },
+        getUint8: function (e) {
+            return K(this, 1, e)[0]
+        },
+        getInt16: function (e) {
+            var t = K(this, 2, e, arguments.length > 1 ? arguments[1] : void 0);
+            return (t[1] << 8 | t[0]) << 16 >> 16
+        },
+        getUint16: function (e) {
+            var t = K(this, 2, e, arguments.length > 1 ? arguments[1] : void 0);
+            return t[1] << 8 | t[0]
+        },
+        getInt32: function (e) {
+            return j(K(this, 4, e, arguments.length > 1 ? arguments[1] : void 0))
+        },
+        getUint32: function (e) {
+            return j(K(this, 4, e, arguments.length > 1 ? arguments[1] : void 0)) >>> 0
+        },
+        getFloat32: function (e) {
+            return z(K(this, 4, e, arguments.length > 1 ? arguments[1] : void 0), 23)
+        },
+        getFloat64: function (e) {
+            return z(K(this, 8, e, arguments.length > 1 ? arguments[1] : void 0), 52)
+        },
+        setInt8: function (e, t) {
+            Y(this, 1, e, U, t)
+        },
+        setUint8: function (e, t) {
+            Y(this, 1, e, U, t)
+        },
+        setInt16: function (e, t) {
+            Y(this, 2, e, W, t, arguments.length > 2 ? arguments[2] : void 0)
+        },
+        setUint16: function (e, t) {
+            Y(this, 2, e, W, t, arguments.length > 2 ? arguments[2] : void 0)
+        },
+        setInt32: function (e, t) {
+            Y(this, 4, e, V, t, arguments.length > 2 ? arguments[2] : void 0)
+        },
+        setUint32: function (e, t) {
+            Y(this, 4, e, V, t, arguments.length > 2 ? arguments[2] : void 0)
+        },
+        setFloat32: function (e, t) {
+            Y(this, 4, e, G, t, arguments.length > 2 ? arguments[2] : void 0)
+        },
+        setFloat64: function (e, t) {
+            Y(this, 8, e, H, t, arguments.length > 2 ? arguments[2] : void 0)
+        }
+    });
+    k(R, "ArrayBuffer"), k(F, "DataView"), e.exports = {
+        ArrayBuffer: R,
+        DataView: F
+    }
+}, function (e, t, n) {
+    var r = n(30),
+        a = n(14),
+        s = n(119),
+        i = Array,
+        o = Math.max;
+    e.exports = function (e, t, n) {
+        for (var u = a(e), l = r(t, u), c = r(void 0 === n ? u : n, u), h = i(o(c - l, 0)), p = 0; l < c; l++, p++) s(h, p, e[l]);
+        return h.length = p, h
+    }
+}, function (e, t, n) {
+    var r = n(38);
+    e.exports = r && !Symbol.sham && "symbol" == typeof Symbol.iterator
+}, function (e, t) {
+    var n = "object" == typeof document && document.all,
+        r = void 0 === n && void 0 !== n;
+    e.exports = {
+        all: n,
+        IS_HTMLDDA: r
+    }
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(2),
+        s = n(57);
+    e.exports = !r && !a((function () {
+        return 7 != Object.defineProperty(s("div"), "a", {
+            get: function () {
+                return 7
+            }
+        }).a
+    }))
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(2);
+    e.exports = r && a((function () {
+        return 42 != Object.defineProperty((function () {}), "prototype", {
+            value: 42,
+            writable: !1
+        }).prototype
+    }))
+}, function (e, t, n) {
+    var r = n(8),
+        a = n(11),
+        s = n(42),
+        i = n(72),
+        o = n(131),
+        u = n(7),
+        l = TypeError,
+        c = u("toPrimitive");
+    e.exports = function (e, t) {
+        if (!a(e) || s(e)) return e;
+        var n, u = i(e, c);
+        if (u) {
+            if (void 0 === t && (t = "default"), n = r(u, e, t), !a(n) || s(n)) return n;
+            throw l("Can't convert object to primitive value")
+        }
+        return void 0 === t && (t = "number"), o(e, t)
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(2),
+        s = n(6),
+        i = n(10),
+        o = n(9),
+        u = n(58).CONFIGURABLE,
+        l = n(73),
+        c = n(21),
+        h = c.enforce,
+        p = c.get,
+        d = String,
+        f = Object.defineProperty,
+        m = r("".slice),
+        g = r("".replace),
+        y = r([].join),
+        b = o && !a((function () {
+            return 8 !== f((function () {}), "length", {
+                value: 8
+            }).length
+        })),
+        v = String(String).split("String"),
+        x = e.exports = function (e, t, n) {
+            "Symbol(" === m(d(t), 0, 7) && (t = "[" + g(d(t), /^Symbol\(([^)]*)\)/, "$1") + "]"), n && n.getter && (t = "get " + t), n && n.setter && (t = "set " + t), (!i(e, "name") || u && e.name !== t) && (o ? f(e, "name", {
+                value: t,
+                configurable: !0
+            }) : e.name = t), b && n && i(n, "arity") && e.length !== n.arity && f(e, "length", {
+                value: n.arity
+            });
+            try {
+                n && i(n, "constructor") && n.constructor ? o && f(e, "prototype", {
+                    writable: !1
+                }) : e.prototype && (e.prototype = void 0)
+            } catch (e) {}
+            var r = h(e);
+            return i(r, "source") || (r.source = y(v, "string" == typeof t ? t : "")), e
+        };
+    Function.prototype.toString = x((function () {
+        return s(this) && p(this).source || l(this)
+    }), "toString")
+}, function (e, t, n) {
+    "use strict";
+    var r = {}.propertyIsEnumerable,
+        a = Object.getOwnPropertyDescriptor,
+        s = a && !r.call({
+            1: 2
+        }, 1);
+    t.f = s ? function (e) {
+        var t = a(this, e);
+        return !!t && t.enumerable
+    } : r
+}, function (e, t, n) {
+    var r = n(10),
+        a = n(136),
+        s = n(43),
+        i = n(12);
+    e.exports = function (e, t, n) {
+        for (var o = a(t), u = i.f, l = s.f, c = 0; c < o.length; c++) {
+            var h = o[c];
+            r(e, h) || n && r(n, h) || u(e, h, l(t, h))
+        }
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(10),
+        s = n(22),
+        i = n(61).indexOf,
+        o = n(60),
+        u = r([].push);
+    e.exports = function (e, t) {
+        var n, r = s(e),
+            l = 0,
+            c = [];
+        for (n in r) !a(o, n) && a(r, n) && u(c, n);
+        for (; t.length > l;) a(r, n = t[l++]) && (~i(c, n) || u(c, n));
+        return c
+    }
+}, function (e, t, n) {
+    var r = n(2),
+        a = n(6),
+        s = /#|\.prototype\./,
+        i = function (e, t) {
+            var n = u[o(e)];
+            return n == c || n != l && (a(t) ? r(t) : !!t)
+        },
+        o = i.normalize = function (e) {
+            return String(e).replace(s, ".").toLowerCase()
+        },
+        u = i.data = {},
+        l = i.NATIVE = "N",
+        c = i.POLYFILL = "P";
+    e.exports = i
+}, function (e, t, n) {
+    var r = n(81),
+        a = n(28),
+        s = TypeError;
+    e.exports = function (e) {
+        if (r(e)) return e;
+        throw s(a(e) + " is not a constructor")
+    }
+}, function (e, t, n) {
+    var r, a, s, i, o = n(1),
+        u = n(45),
+        l = n(46),
+        c = n(6),
+        h = n(10),
+        p = n(2),
+        d = n(101),
+        f = n(47),
+        m = n(57),
+        g = n(140),
+        y = n(102),
+        b = n(62),
+        v = o.setImmediate,
+        x = o.clearImmediate,
+        w = o.process,
+        k = o.Dispatch,
+        S = o.Function,
+        I = o.MessageChannel,
+        A = o.String,
+        E = 0,
+        N = {};
+    p((function () {
+        r = o.location
+    }));
+    var C = function (e) {
+            if (h(N, e)) {
+                var t = N[e];
+                delete N[e], t()
+            }
+        },
+        T = function (e) {
+            return function () {
+                C(e)
+            }
+        },
+        R = function (e) {
+            C(e.data)
+        },
+        _ = function (e) {
+            o.postMessage(A(e), r.protocol + "//" + r.host)
+        };
+    v && x || (v = function (e) {
+        g(arguments.length, 1);
+        var t = c(e) ? e : S(e),
+            n = f(arguments, 1);
+        return N[++E] = function () {
+            u(t, void 0, n)
+        }, a(E), E
+    }, x = function (e) {
+        delete N[e]
+    }, b ? a = function (e) {
+        w.nextTick(T(e))
+    } : k && k.now ? a = function (e) {
+        k.now(T(e))
+    } : I && !y ? (i = (s = new I).port2, s.port1.onmessage = R, a = l(i.postMessage, i)) : o.addEventListener && c(o.postMessage) && !o.importScripts && r && "file:" !== r.protocol && !p(_) ? (a = _, o.addEventListener("message", R, !1)) : a = "onreadystatechange" in m("script") ? function (e) {
+        d.appendChild(m("script")).onreadystatechange = function () {
+            d.removeChild(this), C(e)
+        }
+    } : function (e) {
+        setTimeout(T(e), 0)
+    }), e.exports = {
+        set: v,
+        clear: x
+    }
+}, function (e, t, n) {
+    var r = n(17);
+    e.exports = r("document", "documentElement")
+}, function (e, t, n) {
+    var r = n(26);
+    e.exports = /(?:ipad|iphone|ipod).*applewebkit/i.test(r)
+}, function (e, t) {
+    var n = function () {
+        this.head = null, this.tail = null
+    };
+    n.prototype = {
+        add: function (e) {
+            var t = {
+                    item: e,
+                    next: null
+                },
+                n = this.tail;
+            n ? n.next = t : this.head = t, this.tail = t
+        },
+        get: function () {
+            var e = this.head;
+            if (e) return null === (this.head = e.next) && (this.tail = null), e.item
+        }
+    }, e.exports = n
+}, function (e, t) {
+    e.exports = "object" == typeof Deno && Deno && "object" == typeof Deno.version
+}, function (e, t, n) {
+    var r = n(46),
+        a = n(8),
+        s = n(13),
+        i = n(28),
+        o = n(106),
+        u = n(14),
+        l = n(27),
+        c = n(107),
+        h = n(83),
+        p = n(147),
+        d = TypeError,
+        f = function (e, t) {
+            this.stopped = e, this.result = t
+        },
+        m = f.prototype;
+    e.exports = function (e, t, n) {
+        var g, y, b, v, x, w, k, S = n && n.that,
+            I = !(!n || !n.AS_ENTRIES),
+            A = !(!n || !n.IS_RECORD),
+            E = !(!n || !n.IS_ITERATOR),
+            N = !(!n || !n.INTERRUPTED),
+            C = r(t, S),
+            T = function (e) {
+                return g && p(g, "normal", e), new f(!0, e)
+            },
+            R = function (e) {
+                return I ? (s(e), N ? C(e[0], e[1], T) : C(e[0], e[1])) : N ? C(e, T) : C(e)
+            };
+        if (A) g = e.iterator;
+        else if (E) g = e;
+        else {
+            if (!(y = h(e))) throw d(i(e) + " is not iterable");
+            if (o(y)) {
+                for (b = 0, v = u(e); v > b; b++)
+                    if ((x = R(e[b])) && l(m, x)) return x;
+                return new f(!1)
+            }
+            g = c(e, y)
+        }
+        for (w = A ? e.next : g.next; !(k = a(w, g)).done;) {
+            try {
+                x = R(k.value)
+            } catch (e) {
+                p(g, "throw", e)
+            }
+            if ("object" == typeof x && x && l(m, x)) return x
+        }
+        return new f(!1)
+    }
+}, function (e, t, n) {
+    var r = n(7),
+        a = n(51),
+        s = r("iterator"),
+        i = Array.prototype;
+    e.exports = function (e) {
+        return void 0 !== e && (a.Array === e || i[s] === e)
+    }
+}, function (e, t, n) {
+    var r = n(8),
+        a = n(19),
+        s = n(13),
+        i = n(28),
+        o = n(83),
+        u = TypeError;
+    e.exports = function (e, t) {
+        var n = arguments.length < 2 ? o(e) : t;
+        if (a(n)) return s(r(n, e));
+        throw u(i(e) + " is not iterable")
+    }
+}, function (e, t, n) {
+    var r = n(48),
+        a = n(109),
+        s = n(49).CONSTRUCTOR;
+    e.exports = s || !a((function (e) {
+        r.all(e).then(void 0, (function () {}))
+    }))
+}, function (e, t, n) {
+    var r = n(7)("iterator"),
+        a = !1;
+    try {
+        var s = 0,
+            i = {
+                next: function () {
+                    return {
+                        done: !!s++
+                    }
+                },
+                return: function () {
+                    a = !0
+                }
+            };
+        i[r] = function () {
+            return this
+        }, Array.from(i, (function () {
+            throw 2
+        }))
+    } catch (e) {}
+    e.exports = function (e, t) {
+        if (!t && !a) return !1;
+        var n = !1;
+        try {
+            var s = {};
+            s[r] = function () {
+                return {
+                    next: function () {
+                        return {
+                            done: n = !0
+                        }
+                    }
+                }
+            }, e(s)
+        } catch (e) {}
+        return n
+    }
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(92),
+        s = n(12),
+        i = n(13),
+        o = n(22),
+        u = n(84);
+    t.f = r && !a ? Object.defineProperties : function (e, t) {
+        i(e);
+        for (var n, r = o(t), a = u(t), l = a.length, c = 0; l > c;) s.f(e, n = a[c++], r[n]);
+        return e
+    }
+}, function (e, t, n) {
+    var r = n(7),
+        a = n(35),
+        s = n(12).f,
+        i = r("unscopables"),
+        o = Array.prototype;
+    null == o[i] && s(o, i, {
+        configurable: !0,
+        value: a(null)
+    }), e.exports = function (e) {
+        o[i][e] = !0
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(8),
+        s = n(20),
+        i = n(58),
+        o = n(6),
+        u = n(165),
+        l = n(52),
+        c = n(32),
+        h = n(25),
+        p = n(23),
+        d = n(16),
+        f = n(7),
+        m = n(51),
+        g = n(113),
+        y = i.PROPER,
+        b = i.CONFIGURABLE,
+        v = g.IteratorPrototype,
+        x = g.BUGGY_SAFARI_ITERATORS,
+        w = f("iterator"),
+        k = function () {
+            return this
+        };
+    e.exports = function (e, t, n, i, f, g, S) {
+        u(n, t, i);
+        var I, A, E, N = function (e) {
+                if (e === f && F) return F;
+                if (!x && e in R) return R[e];
+                switch (e) {
+                    case "keys":
+                    case "values":
+                    case "entries":
+                        return function () {
+                            return new n(this, e)
+                        }
+                }
+                return function () {
+                    return new n(this)
+                }
+            },
+            C = t + " Iterator",
+            T = !1,
+            R = e.prototype,
+            _ = R[w] || R["@@iterator"] || f && R[f],
+            F = !x && _ || N(f),
+            M = "Array" == t && R.entries || _;
+        if (M && (I = l(M.call(new e))) !== Object.prototype && I.next && (s || l(I) === v || (c ? c(I, v) : o(I[w]) || d(I, w, k)), h(I, C, !0, !0), s && (m[C] = k)), y && "values" == f && _ && "values" !== _.name && (!s && b ? p(R, "name", "values") : (T = !0, F = function () {
+                return a(_, this)
+            })), f)
+            if (A = {
+                    values: N("values"),
+                    keys: g ? F : N("keys"),
+                    entries: N("entries")
+                }, S)
+                for (E in A)(x || T || !(E in R)) && d(R, E, A[E]);
+            else r({
+                target: t,
+                proto: !0,
+                forced: x || T
+            }, A);
+        return s && !S || R[w] === F || d(R, w, F, {
+            name: f
+        }), m[t] = F, A
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r, a, s, i = n(2),
+        o = n(6),
+        u = n(11),
+        l = n(35),
+        c = n(52),
+        h = n(16),
+        p = n(7),
+        d = n(20),
+        f = p("iterator"),
+        m = !1;
+    [].keys && ("next" in (s = [].keys()) ? (a = c(c(s))) !== Object.prototype && (r = a) : m = !0), !u(r) || i((function () {
+        var e = {};
+        return r[f].call(e) !== e
+    })) ? r = {} : d && (r = l(r)), o(r[f]) || h(r, f, (function () {
+        return this
+    })), e.exports = {
+        IteratorPrototype: r,
+        BUGGY_SAFARI_ITERATORS: m
+    }
+}, function (e, t, n) {
+    var r = n(2);
+    e.exports = !r((function () {
+        function e() {}
+        return e.prototype.constructor = null, Object.getPrototypeOf(new e) !== e.prototype
+    }))
+}, function (e, t) {
+    e.exports = function (e, t) {
+        return {
+            value: e,
+            done: t
+        }
+    }
+}, function (e, t) {
+    e.exports = "undefined" != typeof ArrayBuffer && "undefined" != typeof DataView
+}, function (e, t, n) {
+    var r = n(31),
+        a = n(40),
+        s = RangeError;
+    e.exports = function (e) {
+        if (void 0 === e) return 0;
+        var t = r(e),
+            n = a(t);
+        if (t !== n) throw s("Wrong length or index");
+        return n
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(15),
+        a = n(30),
+        s = n(14);
+    e.exports = function (e) {
+        for (var t = r(this), n = s(t), i = arguments.length, o = a(i > 1 ? arguments[1] : void 0, n), u = i > 2 ? arguments[2] : void 0, l = void 0 === u ? n : a(u, n); l > o;) t[o++] = e;
+        return t
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(41),
+        a = n(12),
+        s = n(39);
+    e.exports = function (e, t, n) {
+        var i = r(t);
+        i in e ? a.f(e, i, s(0, n)) : e[i] = n
+    }
+}, function (e, t, n) {
+    var r = n(173),
+        a = RangeError;
+    e.exports = function (e, t) {
+        var n = r(e);
+        if (n % t) throw a("Wrong offset");
+        return n
+    }
+}, function (e, t, n) {
+    var r = n(93),
+        a = TypeError;
+    e.exports = function (e) {
+        var t = r(e, "number");
+        if ("number" == typeof t) throw a("Can't convert number to bigint");
+        return BigInt(t)
+    }
+}, function (e, t, n) {
+    var r = n(19),
+        a = n(15),
+        s = n(74),
+        i = n(14),
+        o = TypeError,
+        u = function (e) {
+            return function (t, n, u, l) {
+                r(n);
+                var c = a(t),
+                    h = s(c),
+                    p = i(c),
+                    d = e ? p - 1 : 0,
+                    f = e ? -1 : 1;
+                if (u < 2)
+                    for (;;) {
+                        if (d in h) {
+                            l = h[d], d += f;
+                            break
+                        }
+                        if (d += f, e ? d < 0 : p <= d) throw o("Reduce of empty array with no initial value")
+                    }
+                for (; e ? d >= 0 : p > d; d += f) d in h && (l = n(l, h[d], d, c));
+                return l
+            }
+        };
+    e.exports = {
+        left: u(!1),
+        right: u(!0)
+    }
+}, function (e, t, n) {
+    var r = n(7);
+    t.f = r
+}, function (e, t, n) {
+    var r = n(38);
+    e.exports = r && !!Symbol.for && !!Symbol.keyFor
+}, function (e, t) {
+    e.exports = {
+        CSSRuleList: 0,
+        CSSStyleDeclaration: 0,
+        CSSValueList: 0,
+        ClientRectList: 0,
+        DOMRectList: 0,
+        DOMStringList: 0,
+        DOMTokenList: 1,
+        DataTransferItemList: 0,
+        FileList: 0,
+        HTMLAllCollection: 0,
+        HTMLCollection: 0,
+        HTMLFormElement: 0,
+        HTMLSelectElement: 0,
+        MediaList: 0,
+        MimeTypeArray: 0,
+        NamedNodeMap: 0,
+        NodeList: 1,
+        PaintRequestList: 0,
+        Plugin: 0,
+        PluginArray: 0,
+        SVGLengthList: 0,
+        SVGNumberList: 0,
+        SVGPathSegList: 0,
+        SVGPointList: 0,
+        SVGStringList: 0,
+        SVGTransformList: 0,
+        SourceBufferList: 0,
+        StyleSheetList: 0,
+        TextTrackCueList: 0,
+        TextTrackList: 0,
+        TouchList: 0
+    }
+}, function (e, t, n) {
+    var r = n(57)("span").classList,
+        a = r && r.constructor && r.constructor.prototype;
+    e.exports = a === Object.prototype ? void 0 : a
+}, function (e, t, n) {
+    "use strict";
+    var r = n(18).forEach,
+        a = n(85)("forEach");
+    e.exports = a ? [].forEach : function (e) {
+        return r(this, e, arguments.length > 1 ? arguments[1] : void 0)
+    }
+}, function (e, t) {
+    function n(e) {
+        var t = new Error("Cannot find module '" + e + "'");
+        throw t.code = "MODULE_NOT_FOUND", t
+    }
+    n.keys = function () {
+        return []
+    }, n.resolve = n, e.exports = n, n.id = 128
+}, function (e, t, n) {
+    e.exports = n(247)
+}, function (e, t, n) {
+    var r = n(68),
+        a = n(16),
+        s = n(133);
+    r || a(Object.prototype, "toString", s, {
+        unsafe: !0
+    })
+}, function (e, t, n) {
+    var r = n(8),
+        a = n(6),
+        s = n(11),
+        i = TypeError;
+    e.exports = function (e, t) {
+        var n, o;
+        if ("string" === t && a(n = e.toString) && !s(o = r(n, e))) return o;
+        if (a(n = e.valueOf) && !s(o = r(n, e))) return o;
+        if ("string" !== t && a(n = e.toString) && !s(o = r(n, e))) return o;
+        throw i("Can't convert object to primitive value")
+    }
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(6),
+        s = r.WeakMap;
+    e.exports = a(s) && /native code/.test(String(s))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(68),
+        a = n(24);
+    e.exports = r ? {}.toString : function () {
+        return "[object " + a(this) + "]"
+    }
+}, function (e, t, n) {
+    n(135), n(146), n(148), n(149), n(150), n(151)
+}, function (e, t, n) {
+    "use strict";
+    var r, a, s, i = n(4),
+        o = n(20),
+        u = n(62),
+        l = n(1),
+        c = n(8),
+        h = n(16),
+        p = n(32),
+        d = n(25),
+        f = n(78),
+        m = n(19),
+        g = n(6),
+        y = n(11),
+        b = n(79),
+        v = n(80),
+        x = n(100).set,
+        w = n(141),
+        k = n(144),
+        S = n(82),
+        I = n(103),
+        A = n(21),
+        E = n(48),
+        N = n(49),
+        C = n(50),
+        T = N.CONSTRUCTOR,
+        R = N.REJECTION_EVENT,
+        _ = N.SUBCLASSING,
+        F = A.getterFor("Promise"),
+        M = A.set,
+        D = E && E.prototype,
+        O = E,
+        L = D,
+        P = l.TypeError,
+        $ = l.document,
+        B = l.process,
+        z = C.f,
+        U = z,
+        W = !!($ && $.createEvent && l.dispatchEvent),
+        V = function (e) {
+            var t;
+            return !(!y(e) || !g(t = e.then)) && t
+        },
+        j = function (e, t) {
+            var n, r, a, s = t.value,
+                i = 1 == t.state,
+                o = i ? e.ok : e.fail,
+                u = e.resolve,
+                l = e.reject,
+                h = e.domain;
+            try {
+                o ? (i || (2 === t.rejection && Y(t), t.rejection = 1), !0 === o ? n = s : (h && h.enter(), n = o(s), h && (h.exit(), a = !0)), n === e.promise ? l(P("Promise-chain cycle")) : (r = V(n)) ? c(r, n, u, l) : u(n)) : l(s)
+            } catch (e) {
+                h && !a && h.exit(), l(e)
+            }
+        },
+        G = function (e, t) {
+            e.notified || (e.notified = !0, w((function () {
+                for (var n, r = e.reactions; n = r.get();) j(n, e);
+                e.notified = !1, t && !e.rejection && q(e)
+            })))
+        },
+        H = function (e, t, n) {
+            var r, a;
+            W ? ((r = $.createEvent("Event")).promise = t, r.reason = n, r.initEvent(e, !1, !0), l.dispatchEvent(r)) : r = {
+                promise: t,
+                reason: n
+            }, !R && (a = l["on" + e]) ? a(r) : "unhandledrejection" === e && k("Unhandled promise rejection", n)
+        },
+        q = function (e) {
+            c(x, l, (function () {
+                var t, n = e.facade,
+                    r = e.value;
+                if (K(e) && (t = S((function () {
+                        u ? B.emit("unhandledRejection", r, n) : H("unhandledrejection", n, r)
+                    })), e.rejection = u || K(e) ? 2 : 1, t.error)) throw t.value
+            }))
+        },
+        K = function (e) {
+            return 1 !== e.rejection && !e.parent
+        },
+        Y = function (e) {
+            c(x, l, (function () {
+                var t = e.facade;
+                u ? B.emit("rejectionHandled", t) : H("rejectionhandled", t, e.value)
+            }))
+        },
+        X = function (e, t, n) {
+            return function (r) {
+                e(t, r, n)
+            }
+        },
+        Q = function (e, t, n) {
+            e.done || (e.done = !0, n && (e = n), e.value = t, e.state = 2, G(e, !0))
+        },
+        Z = function (e, t, n) {
+            if (!e.done) {
+                e.done = !0, n && (e = n);
+                try {
+                    if (e.facade === t) throw P("Promise can't be resolved itself");
+                    var r = V(t);
+                    r ? w((function () {
+                        var n = {
+                            done: !1
+                        };
+                        try {
+                            c(r, t, X(Z, n, e), X(Q, n, e))
+                        } catch (t) {
+                            Q(n, t, e)
+                        }
+                    })) : (e.value = t, e.state = 1, G(e, !1))
+                } catch (t) {
+                    Q({
+                        done: !1
+                    }, t, e)
+                }
+            }
+        };
+    if (T && (L = (O = function (e) {
+            b(this, L), m(e), c(r, this);
+            var t = F(this);
+            try {
+                e(X(Z, t), X(Q, t))
+            } catch (e) {
+                Q(t, e)
+            }
+        }).prototype, (r = function (e) {
+            M(this, {
+                type: "Promise",
+                done: !1,
+                notified: !1,
+                parent: !1,
+                reactions: new I,
+                rejection: !1,
+                state: 0,
+                value: void 0
+            })
+        }).prototype = h(L, "then", (function (e, t) {
+            var n = F(this),
+                r = z(v(this, O));
+            return n.parent = !0, r.ok = !g(e) || e, r.fail = g(t) && t, r.domain = u ? B.domain : void 0, 0 == n.state ? n.reactions.add(r) : w((function () {
+                j(r, n)
+            })), r.promise
+        })), a = function () {
+            var e = new r,
+                t = F(e);
+            this.promise = e, this.resolve = X(Z, t), this.reject = X(Q, t)
+        }, C.f = z = function (e) {
+            return e === O || void 0 === e ? new a(e) : U(e)
+        }, !o && g(E) && D !== Object.prototype)) {
+        s = D.then, _ || h(D, "then", (function (e, t) {
+            var n = this;
+            return new O((function (e, t) {
+                c(s, n, e, t)
+            })).then(e, t)
+        }), {
+            unsafe: !0
+        });
+        try {
+            delete D.constructor
+        } catch (e) {}
+        p && p(D, L)
+    }
+    i({
+        global: !0,
+        constructor: !0,
+        wrap: !0,
+        forced: T
+    }, {
+        Promise: O
+    }), d(O, "Promise", !1, !0), f("Promise")
+}, function (e, t, n) {
+    var r = n(17),
+        a = n(3),
+        s = n(44),
+        i = n(76),
+        o = n(13),
+        u = a([].concat);
+    e.exports = r("Reflect", "ownKeys") || function (e) {
+        var t = s.f(o(e)),
+            n = i.f;
+        return n ? u(t, n(e)) : t
+    }
+}, function (e, t) {
+    var n = Math.ceil,
+        r = Math.floor;
+    e.exports = Math.trunc || function (e) {
+        var t = +e;
+        return (t > 0 ? r : n)(t)
+    }
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(19);
+    e.exports = function (e, t, n) {
+        try {
+            return r(a(Object.getOwnPropertyDescriptor(e, t)[n]))
+        } catch (e) {}
+    }
+}, function (e, t, n) {
+    var r = n(6),
+        a = String,
+        s = TypeError;
+    e.exports = function (e) {
+        if ("object" == typeof e || r(e)) return e;
+        throw s("Can't set " + a(e) + " as a prototype")
+    }
+}, function (e, t) {
+    var n = TypeError;
+    e.exports = function (e, t) {
+        if (e < t) throw n("Not enough arguments");
+        return e
+    }
+}, function (e, t, n) {
+    var r, a, s, i, o, u = n(1),
+        l = n(46),
+        c = n(43).f,
+        h = n(100).set,
+        p = n(103),
+        d = n(102),
+        f = n(142),
+        m = n(143),
+        g = n(62),
+        y = u.MutationObserver || u.WebKitMutationObserver,
+        b = u.document,
+        v = u.process,
+        x = u.Promise,
+        w = c(u, "queueMicrotask"),
+        k = w && w.value;
+    if (!k) {
+        var S = new p,
+            I = function () {
+                var e, t;
+                for (g && (e = v.domain) && e.exit(); t = S.get();) try {
+                    t()
+                } catch (e) {
+                    throw S.head && r(), e
+                }
+                e && e.enter()
+            };
+        d || g || m || !y || !b ? !f && x && x.resolve ? ((i = x.resolve(void 0)).constructor = x, o = l(i.then, i), r = function () {
+            o(I)
+        }) : g ? r = function () {
+            v.nextTick(I)
+        } : (h = l(h, u), r = function () {
+            h(I)
+        }) : (a = !0, s = b.createTextNode(""), new y(I).observe(s, {
+            characterData: !0
+        }), r = function () {
+            s.data = a = !a
+        }), k = function (e) {
+            S.head || r(), S.add(e)
+        }
+    }
+    e.exports = k
+}, function (e, t, n) {
+    var r = n(26);
+    e.exports = /ipad|iphone|ipod/i.test(r) && "undefined" != typeof Pebble
+}, function (e, t, n) {
+    var r = n(26);
+    e.exports = /web0s(?!.*chrome)/i.test(r)
+}, function (e, t) {
+    e.exports = function (e, t) {
+        try {
+            1 == arguments.length ? console.error(e) : console.error(e, t)
+        } catch (e) {}
+    }
+}, function (e, t, n) {
+    var r = n(104),
+        a = n(62);
+    e.exports = !r && !a && "object" == typeof window && "object" == typeof document
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(8),
+        s = n(19),
+        i = n(50),
+        o = n(82),
+        u = n(105);
+    r({
+        target: "Promise",
+        stat: !0,
+        forced: n(108)
+    }, {
+        all: function (e) {
+            var t = this,
+                n = i.f(t),
+                r = n.resolve,
+                l = n.reject,
+                c = o((function () {
+                    var n = s(t.resolve),
+                        i = [],
+                        o = 0,
+                        c = 1;
+                    u(e, (function (e) {
+                        var s = o++,
+                            u = !1;
+                        c++, a(n, t, e).then((function (e) {
+                            u || (u = !0, i[s] = e, --c || r(i))
+                        }), l)
+                    })), --c || r(i)
+                }));
+            return c.error && l(c.value), n.promise
+        }
+    })
+}, function (e, t, n) {
+    var r = n(8),
+        a = n(13),
+        s = n(72);
+    e.exports = function (e, t, n) {
+        var i, o;
+        a(e);
+        try {
+            if (!(i = s(e, "return"))) {
+                if ("throw" === t) throw n;
+                return n
+            }
+            i = r(i, e)
+        } catch (e) {
+            o = !0, i = e
+        }
+        if ("throw" === t) throw n;
+        if (o) throw i;
+        return a(i), n
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(20),
+        s = n(49).CONSTRUCTOR,
+        i = n(48),
+        o = n(17),
+        u = n(6),
+        l = n(16),
+        c = i && i.prototype;
+    if (r({
+            target: "Promise",
+            proto: !0,
+            forced: s,
+            real: !0
+        }, {
+            catch: function (e) {
+                return this.then(void 0, e)
+            }
+        }), !a && u(i)) {
+        var h = o("Promise").prototype.catch;
+        c.catch !== h && l(c, "catch", h, {
+            unsafe: !0
+        })
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(8),
+        s = n(19),
+        i = n(50),
+        o = n(82),
+        u = n(105);
+    r({
+        target: "Promise",
+        stat: !0,
+        forced: n(108)
+    }, {
+        race: function (e) {
+            var t = this,
+                n = i.f(t),
+                r = n.reject,
+                l = o((function () {
+                    var i = s(t.resolve);
+                    u(e, (function (e) {
+                        a(i, t, e).then(n.resolve, r)
+                    }))
+                }));
+            return l.error && r(l.value), n.promise
+        }
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(8),
+        s = n(50);
+    r({
+        target: "Promise",
+        stat: !0,
+        forced: n(49).CONSTRUCTOR
+    }, {
+        reject: function (e) {
+            var t = s.f(this);
+            return a(t.reject, void 0, e), t.promise
+        }
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(17),
+        s = n(20),
+        i = n(48),
+        o = n(49).CONSTRUCTOR,
+        u = n(152),
+        l = a("Promise"),
+        c = s && !o;
+    r({
+        target: "Promise",
+        stat: !0,
+        forced: s || o
+    }, {
+        resolve: function (e) {
+            return u(c && this === l ? i : this, e)
+        }
+    })
+}, function (e, t, n) {
+    var r = n(13),
+        a = n(11),
+        s = n(50);
+    e.exports = function (e, t) {
+        if (r(e), a(t) && t.constructor === e) return t;
+        var n = s.f(e);
+        return (0, n.resolve)(t), n.promise
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(154);
+    r({
+        target: "RegExp",
+        proto: !0,
+        forced: /./.exec !== a
+    }, {
+        exec: a
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r, a, s = n(8),
+        i = n(3),
+        o = n(34),
+        u = n(155),
+        l = n(156),
+        c = n(37),
+        h = n(35),
+        p = n(21).get,
+        d = n(157),
+        f = n(158),
+        m = c("native-string-replace", String.prototype.replace),
+        g = RegExp.prototype.exec,
+        y = g,
+        b = i("".charAt),
+        v = i("".indexOf),
+        x = i("".replace),
+        w = i("".slice),
+        k = (a = /b*/g, s(g, r = /a/, "a"), s(g, a, "a"), 0 !== r.lastIndex || 0 !== a.lastIndex),
+        S = l.BROKEN_CARET,
+        I = void 0 !== /()??/.exec("")[1];
+    (k || I || S || d || f) && (y = function (e) {
+        var t, n, r, a, i, l, c, d = this,
+            f = p(d),
+            A = o(e),
+            E = f.raw;
+        if (E) return E.lastIndex = d.lastIndex, t = s(y, E, A), d.lastIndex = E.lastIndex, t;
+        var N = f.groups,
+            C = S && d.sticky,
+            T = s(u, d),
+            R = d.source,
+            _ = 0,
+            F = A;
+        if (C && (T = x(T, "y", ""), -1 === v(T, "g") && (T += "g"), F = w(A, d.lastIndex), d.lastIndex > 0 && (!d.multiline || d.multiline && "\n" !== b(A, d.lastIndex - 1)) && (R = "(?: " + R + ")", F = " " + F, _++), n = new RegExp("^(?:" + R + ")", T)), I && (n = new RegExp("^" + R + "$(?!\\s)", T)), k && (r = d.lastIndex), a = s(g, C ? n : d, F), C ? a ? (a.input = w(a.input, _), a[0] = w(a[0], _), a.index = d.lastIndex, d.lastIndex += a[0].length) : d.lastIndex = 0 : k && a && (d.lastIndex = d.global ? a.index + a[0].length : r), I && a && a.length > 1 && s(m, a[0], n, (function () {
+                for (i = 1; i < arguments.length - 2; i++) void 0 === arguments[i] && (a[i] = void 0)
+            })), a && N)
+            for (a.groups = l = h(null), i = 0; i < N.length; i++) l[(c = N[i])[0]] = a[c[1]];
+        return a
+    }), e.exports = y
+}, function (e, t, n) {
+    "use strict";
+    var r = n(13);
+    e.exports = function () {
+        var e = r(this),
+            t = "";
+        return e.hasIndices && (t += "d"), e.global && (t += "g"), e.ignoreCase && (t += "i"), e.multiline && (t += "m"), e.dotAll && (t += "s"), e.unicode && (t += "u"), e.unicodeSets && (t += "v"), e.sticky && (t += "y"), t
+    }
+}, function (e, t, n) {
+    var r = n(2),
+        a = n(1).RegExp,
+        s = r((function () {
+            var e = a("a", "y");
+            return e.lastIndex = 2, null != e.exec("abcd")
+        })),
+        i = s || r((function () {
+            return !a("a", "y").sticky
+        })),
+        o = s || r((function () {
+            var e = a("^r", "gy");
+            return e.lastIndex = 2, null != e.exec("str")
+        }));
+    e.exports = {
+        BROKEN_CARET: o,
+        MISSED_STICKY: i,
+        UNSUPPORTED_Y: s
+    }
+}, function (e, t, n) {
+    var r = n(2),
+        a = n(1).RegExp;
+    e.exports = r((function () {
+        var e = a(".", "s");
+        return !(e.dotAll && e.exec("\n") && "s" === e.flags)
+    }))
+}, function (e, t, n) {
+    var r = n(2),
+        a = n(1).RegExp;
+    e.exports = r((function () {
+        var e = a("(?<a>b)", "g");
+        return "b" !== e.exec("b").groups.a || "bc" !== "b".replace(e, "$<a>c")
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(63),
+        s = n(61).indexOf,
+        i = n(85),
+        o = a([].indexOf),
+        u = !!o && 1 / o([1], 1, -0) < 0;
+    r({
+        target: "Array",
+        proto: !0,
+        forced: u || !i("indexOf")
+    }, {
+        indexOf: function (e) {
+            var t = arguments.length > 1 ? arguments[1] : void 0;
+            return u ? o(this, e, t) || 0 : s(this, e, t)
+        }
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(18).find,
+        s = n(111),
+        i = !0;
+    "find" in [] && Array(1).find((function () {
+        i = !1
+    })), r({
+        target: "Array",
+        proto: !0,
+        forced: i
+    }, {
+        find: function (e) {
+            return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+        }
+    }), s("find")
+}, function (e, t, n) {
+    var r = n(162);
+    e.exports = function (e, t) {
+        return new(r(e))(0 === t ? 0 : t)
+    }
+}, function (e, t, n) {
+    var r = n(64),
+        a = n(81),
+        s = n(11),
+        i = n(7)("species"),
+        o = Array;
+    e.exports = function (e) {
+        var t;
+        return r(e) && (t = e.constructor, (a(t) && (t === o || r(t.prototype)) || s(t) && null === (t = t[i])) && (t = void 0)), void 0 === t ? o : t
+    }
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(15),
+        s = n(84);
+    r({
+        target: "Object",
+        stat: !0,
+        forced: n(2)((function () {
+            s(1)
+        }))
+    }, {
+        keys: function (e) {
+            return s(a(e))
+        }
+    })
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(16),
+        s = Date.prototype,
+        i = r(s.toString),
+        o = r(s.getTime);
+    "Invalid Date" != String(new Date(NaN)) && a(s, "toString", (function () {
+        var e = o(this);
+        return e == e ? i(this) : "Invalid Date"
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(113).IteratorPrototype,
+        a = n(35),
+        s = n(39),
+        i = n(25),
+        o = n(51),
+        u = function () {
+            return this
+        };
+    e.exports = function (e, t, n, l) {
+        var c = t + " Iterator";
+        return e.prototype = a(r, {
+            next: s(+!l, n)
+        }), i(e, c, !1, !0), o[c] = u, e
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(63),
+        s = n(2),
+        i = n(87),
+        o = n(13),
+        u = n(30),
+        l = n(40),
+        c = n(80),
+        h = i.ArrayBuffer,
+        p = i.DataView,
+        d = p.prototype,
+        f = a(h.prototype.slice),
+        m = a(d.getUint8),
+        g = a(d.setUint8);
+    r({
+        target: "ArrayBuffer",
+        proto: !0,
+        unsafe: !0,
+        forced: s((function () {
+            return !new h(2).slice(1, void 0).byteLength
+        }))
+    }, {
+        slice: function (e, t) {
+            if (f && void 0 === t) return f(o(this), e);
+            for (var n = o(this).byteLength, r = u(e, n), a = u(void 0 === t ? n : t, n), s = new(c(this, h))(l(a - r)), i = new p(this), d = new p(s), y = 0; r < a;) g(d, y++, m(i, r++));
+            return s
+        }
+    })
+}, function (e, t, n) {
+    var r = n(16);
+    e.exports = function (e, t, n) {
+        for (var a in t) r(e, a, t[a], n);
+        return e
+    }
+}, function (e, t) {
+    var n = Array,
+        r = Math.abs,
+        a = Math.pow,
+        s = Math.floor,
+        i = Math.log,
+        o = Math.LN2;
+    e.exports = {
+        pack: function (e, t, u) {
+            var l, c, h, p = n(u),
+                d = 8 * u - t - 1,
+                f = (1 << d) - 1,
+                m = f >> 1,
+                g = 23 === t ? a(2, -24) - a(2, -77) : 0,
+                y = e < 0 || 0 === e && 1 / e < 0 ? 1 : 0,
+                b = 0;
+            for ((e = r(e)) != e || e === 1 / 0 ? (c = e != e ? 1 : 0, l = f) : (l = s(i(e) / o), e * (h = a(2, -l)) < 1 && (l--, h *= 2), (e += l + m >= 1 ? g / h : g * a(2, 1 - m)) * h >= 2 && (l++, h /= 2), l + m >= f ? (c = 0, l = f) : l + m >= 1 ? (c = (e * h - 1) * a(2, t), l += m) : (c = e * a(2, m - 1) * a(2, t), l = 0)); t >= 8;) p[b++] = 255 & c, c /= 256, t -= 8;
+            for (l = l << t | c, d += t; d > 0;) p[b++] = 255 & l, l /= 256, d -= 8;
+            return p[--b] |= 128 * y, p
+        },
+        unpack: function (e, t) {
+            var n, r = e.length,
+                s = 8 * r - t - 1,
+                i = (1 << s) - 1,
+                o = i >> 1,
+                u = s - 7,
+                l = r - 1,
+                c = e[l--],
+                h = 127 & c;
+            for (c >>= 7; u > 0;) h = 256 * h + e[l--], u -= 8;
+            for (n = h & (1 << -u) - 1, h >>= -u, u += t; u > 0;) n = 256 * n + e[l--], u -= 8;
+            if (0 === h) h = 1 - o;
+            else {
+                if (h === i) return n ? NaN : c ? -1 / 0 : 1 / 0;
+                n += a(2, t), h -= o
+            }
+            return (c ? -1 : 1) * n * a(2, h - t)
+        }
+    }
+}, function (e, t, n) {
+    n(170)("Uint8", (function (e) {
+        return function (t, n, r) {
+            return e(this, t, n, r)
+        }
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(1),
+        s = n(8),
+        i = n(9),
+        o = n(171),
+        u = n(5),
+        l = n(87),
+        c = n(79),
+        h = n(39),
+        p = n(23),
+        d = n(172),
+        f = n(40),
+        m = n(117),
+        g = n(120),
+        y = n(41),
+        b = n(10),
+        v = n(24),
+        x = n(11),
+        w = n(42),
+        k = n(35),
+        S = n(27),
+        I = n(32),
+        A = n(44).f,
+        E = n(174),
+        N = n(18).forEach,
+        C = n(78),
+        T = n(33),
+        R = n(12),
+        _ = n(43),
+        F = n(21),
+        M = n(176),
+        D = F.get,
+        O = F.set,
+        L = F.enforce,
+        P = R.f,
+        $ = _.f,
+        B = Math.round,
+        z = a.RangeError,
+        U = l.ArrayBuffer,
+        W = U.prototype,
+        V = l.DataView,
+        j = u.NATIVE_ARRAY_BUFFER_VIEWS,
+        G = u.TYPED_ARRAY_TAG,
+        H = u.TypedArray,
+        q = u.TypedArrayPrototype,
+        K = u.aTypedArrayConstructor,
+        Y = u.isTypedArray,
+        X = function (e, t) {
+            K(e);
+            for (var n = 0, r = t.length, a = new e(r); r > n;) a[n] = t[n++];
+            return a
+        },
+        Q = function (e, t) {
+            T(e, t, {
+                configurable: !0,
+                get: function () {
+                    return D(this)[t]
+                }
+            })
+        },
+        Z = function (e) {
+            var t;
+            return S(W, e) || "ArrayBuffer" == (t = v(e)) || "SharedArrayBuffer" == t
+        },
+        J = function (e, t) {
+            return Y(e) && !w(t) && t in e && d(+t) && t >= 0
+        },
+        ee = function (e, t) {
+            return t = y(t), J(e, t) ? h(2, e[t]) : $(e, t)
+        },
+        te = function (e, t, n) {
+            return t = y(t), !(J(e, t) && x(n) && b(n, "value")) || b(n, "get") || b(n, "set") || n.configurable || b(n, "writable") && !n.writable || b(n, "enumerable") && !n.enumerable ? P(e, t, n) : (e[t] = n.value, e)
+        };
+    i ? (j || (_.f = ee, R.f = te, Q(q, "buffer"), Q(q, "byteOffset"), Q(q, "byteLength"), Q(q, "length")), r({
+        target: "Object",
+        stat: !0,
+        forced: !j
+    }, {
+        getOwnPropertyDescriptor: ee,
+        defineProperty: te
+    }), e.exports = function (e, t, n) {
+        var i = e.match(/\d+/)[0] / 8,
+            u = e + (n ? "Clamped" : "") + "Array",
+            l = "get" + e,
+            h = "set" + e,
+            d = a[u],
+            y = d,
+            b = y && y.prototype,
+            v = {},
+            w = function (e, t) {
+                P(e, t, {
+                    get: function () {
+                        return function (e, t) {
+                            var n = D(e);
+                            return n.view[l](t * i + n.byteOffset, !0)
+                        }(this, t)
+                    },
+                    set: function (e) {
+                        return function (e, t, r) {
+                            var a = D(e);
+                            n && (r = (r = B(r)) < 0 ? 0 : r > 255 ? 255 : 255 & r), a.view[h](t * i + a.byteOffset, r, !0)
+                        }(this, t, e)
+                    },
+                    enumerable: !0
+                })
+            };
+        j ? o && (y = t((function (e, t, n, r) {
+            return c(e, b), M(x(t) ? Z(t) ? void 0 !== r ? new d(t, g(n, i), r) : void 0 !== n ? new d(t, g(n, i)) : new d(t) : Y(t) ? X(y, t) : s(E, y, t) : new d(m(t)), e, y)
+        })), I && I(y, H), N(A(d), (function (e) {
+            e in y || p(y, e, d[e])
+        })), y.prototype = b) : (y = t((function (e, t, n, r) {
+            c(e, b);
+            var a, o, u, l = 0,
+                h = 0;
+            if (x(t)) {
+                if (!Z(t)) return Y(t) ? X(y, t) : s(E, y, t);
+                a = t, h = g(n, i);
+                var p = t.byteLength;
+                if (void 0 === r) {
+                    if (p % i) throw z("Wrong length");
+                    if ((o = p - h) < 0) throw z("Wrong length")
+                } else if ((o = f(r) * i) + h > p) throw z("Wrong length");
+                u = o / i
+            } else u = m(t), a = new U(o = u * i);
+            for (O(e, {
+                    buffer: a,
+                    byteOffset: h,
+                    byteLength: o,
+                    length: u,
+                    view: new V(a)
+                }); l < u;) w(e, l++)
+        })), I && I(y, H), b = y.prototype = k(q)), b.constructor !== y && p(b, "constructor", y), L(b).TypedArrayConstructor = y, G && p(b, G, u);
+        var S = y != d;
+        v[u] = y, r({
+            global: !0,
+            constructor: !0,
+            forced: S,
+            sham: !j
+        }, v), "BYTES_PER_ELEMENT" in y || p(y, "BYTES_PER_ELEMENT", i), "BYTES_PER_ELEMENT" in b || p(b, "BYTES_PER_ELEMENT", i), C(u)
+    }) : e.exports = function () {}
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(2),
+        s = n(109),
+        i = n(5).NATIVE_ARRAY_BUFFER_VIEWS,
+        o = r.ArrayBuffer,
+        u = r.Int8Array;
+    e.exports = !i || !a((function () {
+        u(1)
+    })) || !a((function () {
+        new u(-1)
+    })) || !s((function (e) {
+        new u, new u(null), new u(1.5), new u(e)
+    }), !0) || a((function () {
+        return 1 !== new u(new o(2), 1, void 0).length
+    }))
+}, function (e, t, n) {
+    var r = n(11),
+        a = Math.floor;
+    e.exports = Number.isInteger || function (e) {
+        return !r(e) && isFinite(e) && a(e) === e
+    }
+}, function (e, t, n) {
+    var r = n(31),
+        a = RangeError;
+    e.exports = function (e) {
+        var t = r(e);
+        if (t < 0) throw a("The argument can't be less than 0");
+        return t
+    }
+}, function (e, t, n) {
+    var r = n(46),
+        a = n(8),
+        s = n(99),
+        i = n(15),
+        o = n(14),
+        u = n(107),
+        l = n(83),
+        c = n(106),
+        h = n(175),
+        p = n(5).aTypedArrayConstructor,
+        d = n(121);
+    e.exports = function (e) {
+        var t, n, f, m, g, y, b, v, x = s(this),
+            w = i(e),
+            k = arguments.length,
+            S = k > 1 ? arguments[1] : void 0,
+            I = void 0 !== S,
+            A = l(w);
+        if (A && !c(A))
+            for (v = (b = u(w, A)).next, w = []; !(y = a(v, b)).done;) w.push(y.value);
+        for (I && k > 2 && (S = r(S, arguments[2])), n = o(w), f = new(p(x))(n), m = h(f), t = 0; n > t; t++) g = I ? S(w[t], t) : w[t], f[t] = m ? d(g) : +g;
+        return f
+    }
+}, function (e, t, n) {
+    var r = n(24);
+    e.exports = function (e) {
+        var t = r(e);
+        return "BigInt64Array" == t || "BigUint64Array" == t
+    }
+}, function (e, t, n) {
+    var r = n(6),
+        a = n(11),
+        s = n(32);
+    e.exports = function (e, t, n) {
+        var i, o;
+        return s && r(i = t.constructor) && i !== n && a(o = i.prototype) && o !== n.prototype && s(e, o), e
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(3),
+        a = n(5),
+        s = r(n(178)),
+        i = a.aTypedArray;
+    (0, a.exportTypedArrayMethod)("copyWithin", (function (e, t) {
+        return s(i(this), e, t, arguments.length > 2 ? arguments[2] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(15),
+        a = n(30),
+        s = n(14),
+        i = n(179),
+        o = Math.min;
+    e.exports = [].copyWithin || function (e, t) {
+        var n = r(this),
+            u = s(n),
+            l = a(e, u),
+            c = a(t, u),
+            h = arguments.length > 2 ? arguments[2] : void 0,
+            p = o((void 0 === h ? u : a(h, u)) - c, u - l),
+            d = 1;
+        for (c < l && l < c + p && (d = -1, c += p - 1, l += p - 1); p-- > 0;) c in n ? n[l] = n[c] : i(n, l), l += d, c += d;
+        return n
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(28),
+        a = TypeError;
+    e.exports = function (e, t) {
+        if (!delete e[t]) throw a("Cannot delete property " + r(t) + " of " + r(e))
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).every,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("every", (function (e) {
+        return a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(118),
+        s = n(121),
+        i = n(24),
+        o = n(8),
+        u = n(3),
+        l = n(2),
+        c = r.aTypedArray,
+        h = r.exportTypedArrayMethod,
+        p = u("".slice);
+    h("fill", (function (e) {
+        var t = arguments.length;
+        c(this);
+        var n = "Big" === p(i(this), 0, 3) ? s(e) : +e;
+        return o(a, this, n, t > 1 ? arguments[1] : void 0, t > 2 ? arguments[2] : void 0)
+    }), l((function () {
+        var e = 0;
+        return new Int8Array(2).fill({
+            valueOf: function () {
+                return e++
+            }
+        }), 1 !== e
+    })))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).filter,
+        s = n(183),
+        i = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("filter", (function (e) {
+        var t = a(i(this), e, arguments.length > 1 ? arguments[1] : void 0);
+        return s(this, t)
+    }))
+}, function (e, t, n) {
+    var r = n(184),
+        a = n(65);
+    e.exports = function (e, t) {
+        return r(a(e), t)
+    }
+}, function (e, t, n) {
+    var r = n(14);
+    e.exports = function (e, t) {
+        for (var n = 0, a = r(t), s = new e(a); a > n;) s[n] = t[n++];
+        return s
+    }
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).find,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("find", (function (e) {
+        return a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).findIndex,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("findIndex", (function (e) {
+        return a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).forEach,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("forEach", (function (e) {
+        a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(61).includes,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("includes", (function (e) {
+        return a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(61).indexOf,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("indexOf", (function (e) {
+        return a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(1),
+        a = n(2),
+        s = n(3),
+        i = n(5),
+        o = n(86),
+        u = n(7)("iterator"),
+        l = r.Uint8Array,
+        c = s(o.values),
+        h = s(o.keys),
+        p = s(o.entries),
+        d = i.aTypedArray,
+        f = i.exportTypedArrayMethod,
+        m = l && l.prototype,
+        g = !a((function () {
+            m[u].call([1])
+        })),
+        y = !!m && m.values && m[u] === m.values && "values" === m.values.name,
+        b = function () {
+            return c(d(this))
+        };
+    f("entries", (function () {
+        return p(d(this))
+    }), g), f("keys", (function () {
+        return h(d(this))
+    }), g), f("values", b, g || !y, {
+        name: "values"
+    }), f(u, b, g || !y, {
+        name: "values"
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(3),
+        s = r.aTypedArray,
+        i = r.exportTypedArrayMethod,
+        o = a([].join);
+    i("join", (function (e) {
+        return o(s(this), e)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(45),
+        s = n(193),
+        i = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("lastIndexOf", (function (e) {
+        var t = arguments.length;
+        return a(s, i(this), t > 1 ? [e, arguments[1]] : [e])
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(45),
+        a = n(22),
+        s = n(31),
+        i = n(14),
+        o = n(85),
+        u = Math.min,
+        l = [].lastIndexOf,
+        c = !!l && 1 / [1].lastIndexOf(1, -0) < 0,
+        h = o("lastIndexOf"),
+        p = c || !h;
+    e.exports = p ? function (e) {
+        if (c) return r(l, this, arguments) || 0;
+        var t = a(this),
+            n = i(t),
+            o = n - 1;
+        for (arguments.length > 1 && (o = u(o, s(arguments[1]))), o < 0 && (o = n + o); o >= 0; o--)
+            if (o in t && t[o] === e) return o || 0;
+        return -1
+    } : l
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).map,
+        s = n(65),
+        i = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("map", (function (e) {
+        return a(i(this), e, arguments.length > 1 ? arguments[1] : void 0, (function (e, t) {
+            return new(s(e))(t)
+        }))
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(122).left,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("reduce", (function (e) {
+        var t = arguments.length;
+        return a(s(this), e, t, t > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(122).right,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("reduceRight", (function (e) {
+        var t = arguments.length;
+        return a(s(this), e, t, t > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = r.aTypedArray,
+        s = r.exportTypedArrayMethod,
+        i = Math.floor;
+    s("reverse", (function () {
+        for (var e, t = a(this).length, n = i(t / 2), r = 0; r < n;) e = this[r], this[r++] = this[--t], this[t] = e;
+        return this
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(1),
+        a = n(8),
+        s = n(5),
+        i = n(14),
+        o = n(120),
+        u = n(15),
+        l = n(2),
+        c = r.RangeError,
+        h = r.Int8Array,
+        p = h && h.prototype,
+        d = p && p.set,
+        f = s.aTypedArray,
+        m = s.exportTypedArrayMethod,
+        g = !l((function () {
+            var e = new Uint8ClampedArray(2);
+            return a(d, e, {
+                length: 1,
+                0: 3
+            }, 1), 3 !== e[1]
+        })),
+        y = g && s.NATIVE_ARRAY_BUFFER_VIEWS && l((function () {
+            var e = new h(2);
+            return e.set(1), e.set("2", 1), 0 !== e[0] || 2 !== e[1]
+        }));
+    m("set", (function (e) {
+        f(this);
+        var t = o(arguments.length > 1 ? arguments[1] : void 0, 1),
+            n = u(e);
+        if (g) return a(d, this, n, t);
+        var r = this.length,
+            s = i(n),
+            l = 0;
+        if (s + t > r) throw c("Wrong length");
+        for (; l < s;) this[t + l] = n[l++]
+    }), !g || y)
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(65),
+        s = n(2),
+        i = n(47),
+        o = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("slice", (function (e, t) {
+        for (var n = i(o(this), e, t), r = a(this), s = 0, u = n.length, l = new r(u); u > s;) l[s] = n[s++];
+        return l
+    }), s((function () {
+        new Int8Array(1).slice()
+    })))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(18).some,
+        s = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("some", (function (e) {
+        return a(s(this), e, arguments.length > 1 ? arguments[1] : void 0)
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(1),
+        a = n(63),
+        s = n(2),
+        i = n(19),
+        o = n(202),
+        u = n(5),
+        l = n(203),
+        c = n(204),
+        h = n(56),
+        p = n(205),
+        d = u.aTypedArray,
+        f = u.exportTypedArrayMethod,
+        m = r.Uint16Array,
+        g = m && a(m.prototype.sort),
+        y = !(!g || s((function () {
+            g(new m(2), null)
+        })) && s((function () {
+            g(new m(2), {})
+        }))),
+        b = !!g && !s((function () {
+            if (h) return h < 74;
+            if (l) return l < 67;
+            if (c) return !0;
+            if (p) return p < 602;
+            var e, t, n = new m(516),
+                r = Array(516);
+            for (e = 0; e < 516; e++) t = e % 4, n[e] = 515 - e, r[e] = e - 2 * t + 3;
+            for (g(n, (function (e, t) {
+                    return (e / 4 | 0) - (t / 4 | 0)
+                })), e = 0; e < 516; e++)
+                if (n[e] !== r[e]) return !0
+        }));
+    f("sort", (function (e) {
+        return void 0 !== e && i(e), b ? g(this, e) : o(d(this), function (e) {
+            return function (t, n) {
+                return void 0 !== e ? +e(t, n) || 0 : n != n ? -1 : t != t ? 1 : 0 === t && 0 === n ? 1 / t > 0 && 1 / n < 0 ? 1 : -1 : t > n
+            }
+        }(e))
+    }), !b || y)
+}, function (e, t, n) {
+    var r = n(88),
+        a = Math.floor,
+        s = function (e, t) {
+            var n = e.length,
+                u = a(n / 2);
+            return n < 8 ? i(e, t) : o(e, s(r(e, 0, u), t), s(r(e, u), t), t)
+        },
+        i = function (e, t) {
+            for (var n, r, a = e.length, s = 1; s < a;) {
+                for (r = s, n = e[s]; r && t(e[r - 1], n) > 0;) e[r] = e[--r];
+                r !== s++ && (e[r] = n)
+            }
+            return e
+        },
+        o = function (e, t, n, r) {
+            for (var a = t.length, s = n.length, i = 0, o = 0; i < a || o < s;) e[i + o] = i < a && o < s ? r(t[i], n[o]) <= 0 ? t[i++] : n[o++] : i < a ? t[i++] : n[o++];
+            return e
+        };
+    e.exports = s
+}, function (e, t, n) {
+    var r = n(26).match(/firefox\/(\d+)/i);
+    e.exports = !!r && +r[1]
+}, function (e, t, n) {
+    var r = n(26);
+    e.exports = /MSIE|Trident/.test(r)
+}, function (e, t, n) {
+    var r = n(26).match(/AppleWebKit\/(\d+)\./);
+    e.exports = !!r && +r[1]
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5),
+        a = n(40),
+        s = n(30),
+        i = n(65),
+        o = r.aTypedArray;
+    (0, r.exportTypedArrayMethod)("subarray", (function (e, t) {
+        var n = o(this),
+            r = n.length,
+            u = s(e, r);
+        return new(i(n))(n.buffer, n.byteOffset + u * n.BYTES_PER_ELEMENT, a((void 0 === t ? r : s(t, r)) - u))
+    }))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(1),
+        a = n(45),
+        s = n(5),
+        i = n(2),
+        o = n(47),
+        u = r.Int8Array,
+        l = s.aTypedArray,
+        c = s.exportTypedArrayMethod,
+        h = [].toLocaleString,
+        p = !!u && i((function () {
+            h.call(new u(1))
+        }));
+    c("toLocaleString", (function () {
+        return a(h, p ? o(l(this)) : l(this), o(arguments))
+    }), i((function () {
+        return [1, 2].toLocaleString() != new u([1, 2]).toLocaleString()
+    })) || !i((function () {
+        u.prototype.toLocaleString.call([1, 2])
+    })))
+}, function (e, t, n) {
+    "use strict";
+    var r = n(5).exportTypedArrayMethod,
+        a = n(2),
+        s = n(1),
+        i = n(3),
+        o = s.Uint8Array,
+        u = o && o.prototype || {},
+        l = [].toString,
+        c = i([].join);
+    a((function () {
+        l.call({})
+    })) && (l = function () {
+        return c(this)
+    });
+    var h = u.toString != l;
+    r("toString", l, h)
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(1),
+        s = n(87),
+        i = n(78),
+        o = s.ArrayBuffer;
+    r({
+        global: !0,
+        constructor: !0,
+        forced: a.ArrayBuffer !== o
+    }, {
+        ArrayBuffer: o
+    }), i("ArrayBuffer")
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(9),
+        s = n(12).f;
+    r({
+        target: "Object",
+        stat: !0,
+        forced: Object.defineProperty !== s,
+        sham: !a
+    }, {
+        defineProperty: s
+    })
+}, function (e, t, n) {
+    n(212), n(216), n(217), n(218), n(220)
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(1),
+        s = n(8),
+        i = n(3),
+        o = n(20),
+        u = n(9),
+        l = n(38),
+        c = n(2),
+        h = n(10),
+        p = n(27),
+        d = n(13),
+        f = n(22),
+        m = n(41),
+        g = n(34),
+        y = n(39),
+        b = n(35),
+        v = n(84),
+        x = n(44),
+        w = n(213),
+        k = n(76),
+        S = n(43),
+        I = n(12),
+        A = n(110),
+        E = n(95),
+        N = n(16),
+        C = n(33),
+        T = n(37),
+        R = n(59),
+        _ = n(60),
+        F = n(55),
+        M = n(7),
+        D = n(123),
+        O = n(66),
+        L = n(215),
+        P = n(25),
+        $ = n(21),
+        B = n(18).forEach,
+        z = R("hidden"),
+        U = $.set,
+        W = $.getterFor("Symbol"),
+        V = Object.prototype,
+        j = a.Symbol,
+        G = j && j.prototype,
+        H = a.TypeError,
+        q = a.QObject,
+        K = S.f,
+        Y = I.f,
+        X = w.f,
+        Q = E.f,
+        Z = i([].push),
+        J = T("symbols"),
+        ee = T("op-symbols"),
+        te = T("wks"),
+        ne = !q || !q.prototype || !q.prototype.findChild,
+        re = u && c((function () {
+            return 7 != b(Y({}, "a", {
+                get: function () {
+                    return Y(this, "a", {
+                        value: 7
+                    }).a
+                }
+            })).a
+        })) ? function (e, t, n) {
+            var r = K(V, t);
+            r && delete V[t], Y(e, t, n), r && e !== V && Y(V, t, r)
+        } : Y,
+        ae = function (e, t) {
+            var n = J[e] = b(G);
+            return U(n, {
+                type: "Symbol",
+                tag: e,
+                description: t
+            }), u || (n.description = t), n
+        },
+        se = function (e, t, n) {
+            e === V && se(ee, t, n), d(e);
+            var r = m(t);
+            return d(n), h(J, r) ? (n.enumerable ? (h(e, z) && e[z][r] && (e[z][r] = !1), n = b(n, {
+                enumerable: y(0, !1)
+            })) : (h(e, z) || Y(e, z, y(1, {})), e[z][r] = !0), re(e, r, n)) : Y(e, r, n)
+        },
+        ie = function (e, t) {
+            d(e);
+            var n = f(t),
+                r = v(n).concat(ce(n));
+            return B(r, (function (t) {
+                u && !s(oe, n, t) || se(e, t, n[t])
+            })), e
+        },
+        oe = function (e) {
+            var t = m(e),
+                n = s(Q, this, t);
+            return !(this === V && h(J, t) && !h(ee, t)) && (!(n || !h(this, t) || !h(J, t) || h(this, z) && this[z][t]) || n)
+        },
+        ue = function (e, t) {
+            var n = f(e),
+                r = m(t);
+            if (n !== V || !h(J, r) || h(ee, r)) {
+                var a = K(n, r);
+                return !a || !h(J, r) || h(n, z) && n[z][r] || (a.enumerable = !0), a
+            }
+        },
+        le = function (e) {
+            var t = X(f(e)),
+                n = [];
+            return B(t, (function (e) {
+                h(J, e) || h(_, e) || Z(n, e)
+            })), n
+        },
+        ce = function (e) {
+            var t = e === V,
+                n = X(t ? ee : f(e)),
+                r = [];
+            return B(n, (function (e) {
+                !h(J, e) || t && !h(V, e) || Z(r, J[e])
+            })), r
+        };
+    l || (N(G = (j = function () {
+        if (p(G, this)) throw H("Symbol is not a constructor");
+        var e = arguments.length && void 0 !== arguments[0] ? g(arguments[0]) : void 0,
+            t = F(e),
+            n = function (e) {
+                this === V && s(n, ee, e), h(this, z) && h(this[z], t) && (this[z][t] = !1), re(this, t, y(1, e))
+            };
+        return u && ne && re(V, t, {
+            configurable: !0,
+            set: n
+        }), ae(t, e)
+    }).prototype, "toString", (function () {
+        return W(this).tag
+    })), N(j, "withoutSetter", (function (e) {
+        return ae(F(e), e)
+    })), E.f = oe, I.f = se, A.f = ie, S.f = ue, x.f = w.f = le, k.f = ce, D.f = function (e) {
+        return ae(M(e), e)
+    }, u && (C(G, "description", {
+        configurable: !0,
+        get: function () {
+            return W(this).description
+        }
+    }), o || N(V, "propertyIsEnumerable", oe, {
+        unsafe: !0
+    }))), r({
+        global: !0,
+        constructor: !0,
+        wrap: !0,
+        forced: !l,
+        sham: !l
+    }, {
+        Symbol: j
+    }), B(v(te), (function (e) {
+        O(e)
+    })), r({
+        target: "Symbol",
+        stat: !0,
+        forced: !l
+    }, {
+        useSetter: function () {
+            ne = !0
+        },
+        useSimple: function () {
+            ne = !1
+        }
+    }), r({
+        target: "Object",
+        stat: !0,
+        forced: !l,
+        sham: !u
+    }, {
+        create: function (e, t) {
+            return void 0 === t ? b(e) : ie(b(e), t)
+        },
+        defineProperty: se,
+        defineProperties: ie,
+        getOwnPropertyDescriptor: ue
+    }), r({
+        target: "Object",
+        stat: !0,
+        forced: !l
+    }, {
+        getOwnPropertyNames: le
+    }), L(), P(j, "Symbol"), _[z] = !0
+}, function (e, t, n) {
+    var r = n(29),
+        a = n(22),
+        s = n(44).f,
+        i = n(88),
+        o = "object" == typeof window && window && Object.getOwnPropertyNames ? Object.getOwnPropertyNames(window) : [];
+    e.exports.f = function (e) {
+        return o && "Window" == r(e) ? function (e) {
+            try {
+                return s(e)
+            } catch (e) {
+                return i(o)
+            }
+        }(e) : s(a(e))
+    }
+}, function (e, t, n) {
+    var r = n(1);
+    e.exports = r
+}, function (e, t, n) {
+    var r = n(8),
+        a = n(17),
+        s = n(7),
+        i = n(16);
+    e.exports = function () {
+        var e = a("Symbol"),
+            t = e && e.prototype,
+            n = t && t.valueOf,
+            o = s("toPrimitive");
+        t && !t[o] && i(t, o, (function (e) {
+            return r(n, this)
+        }), {
+            arity: 1
+        })
+    }
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(17),
+        s = n(10),
+        i = n(34),
+        o = n(37),
+        u = n(124),
+        l = o("string-to-symbol-registry"),
+        c = o("symbol-to-string-registry");
+    r({
+        target: "Symbol",
+        stat: !0,
+        forced: !u
+    }, {
+        for: function (e) {
+            var t = i(e);
+            if (s(l, t)) return l[t];
+            var n = a("Symbol")(t);
+            return l[t] = n, c[n] = t, n
+        }
+    })
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(10),
+        s = n(42),
+        i = n(28),
+        o = n(37),
+        u = n(124),
+        l = o("symbol-to-string-registry");
+    r({
+        target: "Symbol",
+        stat: !0,
+        forced: !u
+    }, {
+        keyFor: function (e) {
+            if (!s(e)) throw TypeError(i(e) + " is not a symbol");
+            if (a(l, e)) return l[e]
+        }
+    })
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(17),
+        s = n(45),
+        i = n(8),
+        o = n(3),
+        u = n(2),
+        l = n(6),
+        c = n(42),
+        h = n(47),
+        p = n(219),
+        d = n(38),
+        f = String,
+        m = a("JSON", "stringify"),
+        g = o(/./.exec),
+        y = o("".charAt),
+        b = o("".charCodeAt),
+        v = o("".replace),
+        x = o(1..toString),
+        w = /[\uD800-\uDFFF]/g,
+        k = /^[\uD800-\uDBFF]$/,
+        S = /^[\uDC00-\uDFFF]$/,
+        I = !d || u((function () {
+            var e = a("Symbol")();
+            return "[null]" != m([e]) || "{}" != m({
+                a: e
+            }) || "{}" != m(Object(e))
+        })),
+        A = u((function () {
+            return '"\\udf06\\ud834"' !== m("\udf06\ud834") || '"\\udead"' !== m("\udead")
+        })),
+        E = function (e, t) {
+            var n = h(arguments),
+                r = p(t);
+            if (l(r) || void 0 !== e && !c(e)) return n[1] = function (e, t) {
+                if (l(r) && (t = i(r, this, f(e), t)), !c(t)) return t
+            }, s(m, null, n)
+        },
+        N = function (e, t, n) {
+            var r = y(n, t - 1),
+                a = y(n, t + 1);
+            return g(k, e) && !g(S, a) || g(S, e) && !g(k, r) ? "\\u" + x(b(e, 0), 16) : e
+        };
+    m && r({
+        target: "JSON",
+        stat: !0,
+        arity: 3,
+        forced: I || A
+    }, {
+        stringify: function (e, t, n) {
+            var r = h(arguments),
+                a = s(I ? E : m, null, r);
+            return A && "string" == typeof a ? v(a, w, N) : a
+        }
+    })
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(64),
+        s = n(6),
+        i = n(29),
+        o = n(34),
+        u = r([].push);
+    e.exports = function (e) {
+        if (s(e)) return e;
+        if (a(e)) {
+            for (var t = e.length, n = [], r = 0; r < t; r++) {
+                var l = e[r];
+                "string" == typeof l ? u(n, l) : "number" != typeof l && "Number" != i(l) && "String" != i(l) || u(n, o(l))
+            }
+            var c = n.length,
+                h = !0;
+            return function (e, t) {
+                if (h) return h = !1, t;
+                if (a(this)) return t;
+                for (var r = 0; r < c; r++)
+                    if (n[r] === e) return t
+            }
+        }
+    }
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(38),
+        s = n(2),
+        i = n(76),
+        o = n(15);
+    r({
+        target: "Object",
+        stat: !0,
+        forced: !a || s((function () {
+            i.f(1)
+        }))
+    }, {
+        getOwnPropertySymbols: function (e) {
+            var t = i.f;
+            return t ? t(o(e)) : []
+        }
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(9),
+        s = n(1),
+        i = n(3),
+        o = n(10),
+        u = n(6),
+        l = n(27),
+        c = n(34),
+        h = n(33),
+        p = n(96),
+        d = s.Symbol,
+        f = d && d.prototype;
+    if (a && u(d) && (!("description" in f) || void 0 !== d().description)) {
+        var m = {},
+            g = function () {
+                var e = arguments.length < 1 || void 0 === arguments[0] ? void 0 : c(arguments[0]),
+                    t = l(f, this) ? new d(e) : void 0 === e ? d() : d(e);
+                return "" === e && (m[t] = !0), t
+            };
+        p(g, d), g.prototype = f, f.constructor = g;
+        var y = "Symbol(test)" == String(d("test")),
+            b = i(f.valueOf),
+            v = i(f.toString),
+            x = /^Symbol\((.*)\)[^)]+$/,
+            w = i("".replace),
+            k = i("".slice);
+        h(f, "description", {
+            configurable: !0,
+            get: function () {
+                var e = b(this);
+                if (o(m, e)) return "";
+                var t = v(e),
+                    n = y ? k(t, 7, -1) : w(t, x, "$1");
+                return "" === n ? void 0 : n
+            }
+        }), r({
+            global: !0,
+            constructor: !0,
+            forced: !0
+        }, {
+            Symbol: g
+        })
+    }
+}, function (e, t, n) {
+    n(66)("iterator")
+}, function (e, t, n) {
+    "use strict";
+    var r = n(224).charAt,
+        a = n(34),
+        s = n(21),
+        i = n(112),
+        o = n(115),
+        u = s.set,
+        l = s.getterFor("String Iterator");
+    i(String, "String", (function (e) {
+        u(this, {
+            type: "String Iterator",
+            string: a(e),
+            index: 0
+        })
+    }), (function () {
+        var e, t = l(this),
+            n = t.string,
+            a = t.index;
+        return a >= n.length ? o(void 0, !0) : (e = r(n, a), t.index += e.length, o(e, !1))
+    }))
+}, function (e, t, n) {
+    var r = n(3),
+        a = n(31),
+        s = n(34),
+        i = n(71),
+        o = r("".charAt),
+        u = r("".charCodeAt),
+        l = r("".slice),
+        c = function (e) {
+            return function (t, n) {
+                var r, c, h = s(i(t)),
+                    p = a(n),
+                    d = h.length;
+                return p < 0 || p >= d ? e ? "" : void 0 : (r = u(h, p)) < 55296 || r > 56319 || p + 1 === d || (c = u(h, p + 1)) < 56320 || c > 57343 ? e ? o(h, p) : r : e ? l(h, p, p + 2) : c - 56320 + (r - 55296 << 10) + 65536
+            }
+        };
+    e.exports = {
+        codeAt: c(!1),
+        charAt: c(!0)
+    }
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(125),
+        s = n(126),
+        i = n(86),
+        o = n(23),
+        u = n(7),
+        l = u("iterator"),
+        c = u("toStringTag"),
+        h = i.values,
+        p = function (e, t) {
+            if (e) {
+                if (e[l] !== h) try {
+                    o(e, l, h)
+                } catch (t) {
+                    e[l] = h
+                }
+                if (e[c] || o(e, c, t), a[t])
+                    for (var n in i)
+                        if (e[n] !== i[n]) try {
+                            o(e, n, i[n])
+                        } catch (t) {
+                            e[n] = i[n]
+                        }
+            }
+        };
+    for (var d in a) p(r[d] && r[d].prototype, d);
+    p(s, "DOMTokenList")
+}, function (e, t, n) {
+    n(66)("asyncIterator")
+}, function (e, t, n) {
+    var r = n(17),
+        a = n(66),
+        s = n(25);
+    a("toStringTag"), s(r("Symbol"), "Symbol")
+}, function (e, t, n) {
+    var r = n(1);
+    n(25)(r.JSON, "JSON", !0)
+}, function (e, t, n) {
+    n(25)(Math, "Math", !0)
+}, function (e, t, n) {
+    n(4)({
+        target: "Object",
+        stat: !0,
+        sham: !n(9)
+    }, {
+        create: n(35)
+    })
+}, function (e, t, n) {
+    var r = n(4),
+        a = n(2),
+        s = n(15),
+        i = n(52),
+        o = n(114);
+    r({
+        target: "Object",
+        stat: !0,
+        forced: a((function () {
+            i(1)
+        })),
+        sham: !o
+    }, {
+        getPrototypeOf: function (e) {
+            return i(s(e))
+        }
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(127);
+    r({
+        target: "Array",
+        proto: !0,
+        forced: [].forEach != a
+    }, {
+        forEach: a
+    })
+}, function (e, t, n) {
+    var r = n(1),
+        a = n(125),
+        s = n(126),
+        i = n(127),
+        o = n(23),
+        u = function (e) {
+            if (e && e.forEach !== i) try {
+                o(e, "forEach", i)
+            } catch (t) {
+                e.forEach = i
+            }
+        };
+    for (var l in a) a[l] && u(r[l] && r[l].prototype);
+    u(s)
+}, function (e, t, n) {
+    var r = n(9),
+        a = n(58).EXISTS,
+        s = n(3),
+        i = n(33),
+        o = Function.prototype,
+        u = s(o.toString),
+        l = /function\b(?:\s|\/\*[\S\s]*?\*\/|\/\/[^\n\r]*[\n\r]+)*([^\s(/]*)/,
+        c = s(l.exec);
+    r && !a && i(o, "name", {
+        configurable: !0,
+        get: function () {
+            try {
+                return c(l, u(this))[1]
+            } catch (e) {
+                return ""
+            }
+        }
+    })
+}, function (e, t, n) {
+    n(4)({
+        target: "Object",
+        stat: !0
+    }, {
+        setPrototypeOf: n(32)
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(3),
+        s = n(64),
+        i = a([].reverse),
+        o = [1, 2];
+    r({
+        target: "Array",
+        proto: !0,
+        forced: String(o) === String(o.reverse())
+    }, {
+        reverse: function () {
+            return s(this) && (this.length = this.length), i(this)
+        }
+    })
+}, function (e, t, n) {
+    "use strict";
+    var r = n(4),
+        a = n(64),
+        s = n(81),
+        i = n(11),
+        o = n(30),
+        u = n(14),
+        l = n(22),
+        c = n(119),
+        h = n(7),
+        p = n(238),
+        d = n(47),
+        f = p("slice"),
+        m = h("species"),
+        g = Array,
+        y = Math.max;
+    r({
+        target: "Array",
+        proto: !0,
+        forced: !f
+    }, {
+        slice: function (e, t) {
+            var n, r, h, p = l(this),
+                f = u(p),
+                b = o(e, f),
+                v = o(void 0 === t ? f : t, f);
+            if (a(p) && (n = p.constructor, (s(n) && (n === g || a(n.prototype)) || i(n) && null === (n = n[m])) && (n = void 0), n === g || void 0 === n)) return d(p, b, v);
+            for (r = new(void 0 === n ? g : n)(y(v - b, 0)), h = 0; b < v; b++, h++) b in p && c(r, h, p[b]);
+            return r.length = h, r
+        }
+    })
+}, function (e, t, n) {
+    var r = n(2),
+        a = n(7),
+        s = n(56),
+        i = a("species");
+    e.exports = function (e) {
+        return s >= 51 || !r((function () {
+            var t = [];
+            return (t.constructor = {})[i] = function () {
+                return {
+                    foo: 1
+                }
+            }, 1 !== t[e](Boolean).foo
+        }))
+    }
+}, function (e, t, n) {
+    var r = {
+        "./camera-comp_3.entry.js": [240, 0]
+    };
+
+    function a(e) {
+        if (!n.o(r, e)) return Promise.resolve().then((function () {
+            var t = new Error("Cannot find module '" + e + "'");
+            throw t.code = "MODULE_NOT_FOUND", t
+        }));
+        var t = r[e],
+            a = t[0];
+        return n.e(t[1]).then((function () {
+            return n(a)
+        }))
+    }
+    a.keys = function () {
+        return Object.keys(r)
+    }, a.id = 239, e.exports = a
+}, function (e, t, n) {
+    "use strict";
+    n.r(t),
+        function (e, r, a, s, i, o) {
+            n.d(t, "camera_comp", (function () {
+                return Cq
+            })), n.d(t, "camera_component", (function () {
+                return Tq
+            })), n.d(t, "control_panel", (function () {
+                return _q
+            }));
+            var u = n(0);
+            class l {
+                static init(e) {
+                    this.cameraModule = e
+                }
+                static captureEvent(e) {
+                    this.cameraModule.dispatchEvent(new CustomEvent("capture", {
+                        detail: {
+                            photos: e
+                        },
+                        bubbles: !0,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+                static closeEvent() {
+                    this.cameraModule.dispatchEvent(new CustomEvent("close", {
+                        bubbles: !1,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+                static cameraReadyEvent() {
+                    this.cameraModule.dispatchEvent(new CustomEvent("cameraReady", {
+                        bubbles: !1,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+                static errorEvent(e) {
+                    this.cameraModule.dispatchEvent(new CustomEvent("error", {
+                        detail: {
+                            error: e
+                        },
+                        bubbles: !0,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+                static detectionError(e) {
+                    this.cameraModule.dispatchEvent(new CustomEvent("detectionError", {
+                        detail: {
+                            error: e
+                        },
+                        bubbles: !0,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+                static loadedModels(e) {
+                    this.cameraModule.dispatchEvent(new CustomEvent("loadedModels", {
+                        detail: {
+                            loadedModels: e
+                        },
+                        bubbles: !0,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+                static detectorInitialized(e) {
+                    this.cameraModule.dispatchEvent(new CustomEvent("detectorInitialized", {
+                        detail: {
+                            initializationTime: e
+                        },
+                        bubbles: !0,
+                        cancelable: !0,
+                        composed: !0
+                    }))
+                }
+            }
+            var c = Object.defineProperty,
+                h = (e => n(128))(),
+                p = (e, t) => {
+                    for (var n in t) c(e, n, {
+                        get: t[n],
+                        enumerable: !0
+                    })
+                },
+                d = {};
+            p(d, {
+                Abs: () => et,
+                Acos: () => tt,
+                Acosh: () => nt,
+                AdadeltaOptimizer: () => Md,
+                AdagradOptimizer: () => Dd,
+                AdamOptimizer: () => Od,
+                AdamaxOptimizer: () => Ld,
+                Add: () => rt,
+                AddN: () => at,
+                All: () => st,
+                Any: () => it,
+                ArgMax: () => ot,
+                ArgMin: () => ut,
+                Asin: () => lt,
+                Asinh: () => ct,
+                Atan: () => ht,
+                Atan2: () => dt,
+                Atanh: () => pt,
+                AvgPool: () => ft,
+                AvgPool3D: () => gt,
+                AvgPool3DGrad: () => yt,
+                AvgPoolGrad: () => mt,
+                BackendWasm: () => vj,
+                BatchMatMul: () => bt,
+                BatchToSpaceND: () => vt,
+                Bincount: () => xt,
+                BroadcastArgs: () => kt,
+                BroadcastTo: () => wt,
+                Callback: () => BS,
+                CallbackList: () => Fb,
+                Cast: () => St,
+                Ceil: () => It,
+                ClipByValue: () => At,
+                Complex: () => Et,
+                ComplexAbs: () => Nt,
+                Concat: () => Ct,
+                Conv2D: () => Tt,
+                Conv2DBackpropFilter: () => Rt,
+                Conv2DBackpropInput: () => _t,
+                Conv3D: () => Ft,
+                Conv3DBackpropFilterV2: () => Mt,
+                Conv3DBackpropInputV2: () => Dt,
+                Cos: () => Ot,
+                Cosh: () => Lt,
+                CropAndResize: () => $t,
+                Cumsum: () => Pt,
+                CustomCallback: () => Ob,
+                DataStorage: () => Q,
+                DenseBincount: () => Bt,
+                DepthToSpace: () => zt,
+                DepthwiseConv2dNative: () => Ut,
+                DepthwiseConv2dNativeBackpropFilter: () => Wt,
+                DepthwiseConv2dNativeBackpropInput: () => Vt,
+                Diag: () => jt,
+                Dilation2D: () => Gt,
+                Dilation2DBackpropFilter: () => qt,
+                Dilation2DBackpropInput: () => Ht,
+                ENV: () => Qe,
+                EarlyStopping: () => WS,
+                Einsum: () => Yt,
+                Elu: () => Xt,
+                EluGrad: () => Qt,
+                Environment: () => qe,
+                Equal: () => Jt,
+                Erf: () => Zt,
+                Exp: () => en,
+                ExpandDims: () => tn,
+                Expm1: () => nn,
+                FFT: () => rn,
+                Fill: () => an,
+                FlipLeftRight: () => sn,
+                Floor: () => on,
+                FloorDiv: () => un,
+                FromPixels: () => ta,
+                FusedBatchNorm: () => ln,
+                FusedConv2D: () => aa,
+                FusedDepthwiseConv2D: () => sa,
+                GPGPUContext: () => WD,
+                GatherNd: () => hn,
+                GatherV2: () => cn,
+                GraphModel: () => IA,
+                Greater: () => pn,
+                GreaterEqual: () => dn,
+                History: () => Db,
+                IFFT: () => mn,
+                Identity: () => fn,
+                Imag: () => gn,
+                InputSpec: () => vb,
+                IsFinite: () => yn,
+                IsInf: () => bn,
+                IsNan: () => vn,
+                KernelBackend: () => Z,
+                LRN: () => Rn,
+                LRNGrad: () => _n,
+                LayerVariable: () => gb,
+                LayersModel: () => Wv,
+                LeakyRelu: () => xn,
+                Less: () => wn,
+                LessEqual: () => kn,
+                LinSpace: () => Sn,
+                Log: () => In,
+                Log1p: () => An,
+                LogSoftmax: () => Tn,
+                LogicalAnd: () => En,
+                LogicalNot: () => Nn,
+                LogicalOr: () => Cn,
+                MathBackendWebGL: () => jO,
+                Max: () => Fn,
+                MaxPool: () => Dn,
+                MaxPool3D: () => Ln,
+                MaxPool3DGrad: () => Pn,
+                MaxPoolGrad: () => On,
+                MaxPoolWithArgmax: () => $n,
+                Maximum: () => Mn,
+                Mean: () => Bn,
+                Min: () => zn,
+                Minimum: () => Un,
+                MirrorPad: () => Wn,
+                Mod: () => Vn,
+                MomentumOptimizer: () => $d,
+                Multinomial: () => jn,
+                Multiply: () => Gn,
+                Neg: () => Hn,
+                NonMaxSuppressionV3: () => Kn,
+                NonMaxSuppressionV4: () => Yn,
+                NonMaxSuppressionV5: () => Xn,
+                NotEqual: () => qn,
+                OP_SCOPE_SUFFIX: () => Fs,
+                OneHot: () => Zn,
+                OnesLike: () => Qn,
+                Optimizer: () => Fd,
+                OptimizerConstructors: () => zd,
+                Pack: () => Jn,
+                PadV2: () => er,
+                Pool: () => tr,
+                Pow: () => nr,
+                Prelu: () => rr,
+                Prod: () => ar,
+                RMSPropOptimizer: () => Bd,
+                RNN: () => Yx,
+                Range: () => sr,
+                Rank: () => Za,
+                Real: () => ir,
+                RealDiv: () => Kt,
+                Reciprocal: () => or,
+                Reduction: () => id,
+                Relu: () => ur,
+                Relu6: () => fr,
+                Reshape: () => lr,
+                ResizeBilinear: () => pr,
+                ResizeBilinearGrad: () => dr,
+                ResizeNearestNeighbor: () => cr,
+                ResizeNearestNeighborGrad: () => hr,
+                Reverse: () => mr,
+                RotateWithOffset: () => na,
+                Round: () => gr,
+                Rsqrt: () => yr,
+                SGDOptimizer: () => Pd,
+                ScatterNd: () => br,
+                Select: () => vr,
+                Selu: () => xr,
+                Sequential: () => Hv,
+                Sigmoid: () => Ar,
+                Sign: () => Ir,
+                Sin: () => kr,
+                Sinh: () => Sr,
+                Slice: () => wr,
+                Softmax: () => _r,
+                Softplus: () => Er,
+                SpaceToBatchND: () => Tr,
+                SparseFillEmptyRows: () => Fr,
+                SparseReshape: () => Mr,
+                SparseSegmentMean: () => Dr,
+                SparseSegmentSum: () => Or,
+                SparseToDense: () => Lr,
+                SplitV: () => Rr,
+                Sqrt: () => Nr,
+                Square: () => $r,
+                SquaredDifference: () => Pr,
+                Step: () => ea,
+                StridedSlice: () => Br,
+                StringNGrams: () => zr,
+                StringSplit: () => Ur,
+                StringToHashBucketFast: () => Wr,
+                Sub: () => Vr,
+                Sum: () => Cr,
+                SymbolicTensor: () => xb,
+                Tan: () => jr,
+                Tanh: () => Gr,
+                Tensor: () => Ya,
+                TensorBuffer: () => Ha,
+                Tile: () => Hr,
+                TopK: () => qr,
+                Transform: () => Kr,
+                Transpose: () => Yr,
+                Unique: () => Xr,
+                Unpack: () => Qr,
+                UnsortedSegmentSum: () => Zr,
+                Variable: () => Qa,
+                ZerosLike: () => Jr,
+                _FusedMatMul: () => ra,
+                abs: () => xu,
+                acos: () => wu,
+                acosh: () => ku,
+                add: () => gu,
+                addN: () => Su,
+                all: () => Iu,
+                any: () => Au,
+                argMax: () => Eu,
+                argMin: () => Nu,
+                asin: () => Cu,
+                asinh: () => Tu,
+                atan: () => Ru,
+                atan2: () => _u,
+                atanh: () => Fu,
+                avgPool: () => Ku,
+                avgPool3d: () => Yu,
+                backend: () => fu,
+                backend_util: () => jd,
+                basicLSTMCell: () => el,
+                batchNorm: () => nl,
+                batchNorm2d: () => rl,
+                batchNorm3d: () => al,
+                batchNorm4d: () => sl,
+                batchToSpaceND: () => tl,
+                bincount: () => il,
+                booleanMaskAsync: () => up,
+                broadcastArgs: () => ol,
+                broadcastTo: () => ul,
+                broadcast_util: () => Yi,
+                browser: () => eo,
+                buffer: () => Ii,
+                callbacks: () => KS,
+                cast: () => Ai,
+                ceil: () => ll,
+                clipByValue: () => cl,
+                clone: () => Ei,
+                complex: () => Ds,
+                concat: () => Xu,
+                concat1d: () => hl,
+                concat2d: () => pl,
+                concat3d: () => dl,
+                concat4d: () => fl,
+                constraints: () => rg,
+                conv1d: () => gl,
+                conv2d: () => ml,
+                conv2dTranspose: () => bl,
+                conv3d: () => vl,
+                conv3dTranspose: () => wl,
+                copyRegisteredKernels: () => ya,
+                cos: () => kl,
+                cosh: () => Sl,
+                cosineWindow: () => gp,
+                cumsum: () => Il,
+                customGrad: () => cc,
+                data: () => NA,
+                denseBincount: () => Al,
+                deprecationWarn: () => Zo,
+                depthToSpace: () => El,
+                depthwiseConv2d: () => Nl,
+                deregisterOp: () => ZS,
+                device_util: () => Ss,
+                diag: () => Cl,
+                dilation2d: () => Tl,
+                disableDeprecationWarnings: () => Qo,
+                dispose: () => au,
+                disposeVariables: () => Jo,
+                div: () => bu,
+                divNoNan: () => Ml,
+                dot: () => Dl,
+                dropout: () => fp,
+                einsum: () => Ol,
+                elu: () => Ll,
+                enableDebugMode: () => Xo,
+                enableProdMode: () => Yo,
+                enclosingPowerOfTwo: () => mp,
+                engine: () => eu,
+                env: () => Ye,
+                equal: () => Rl,
+                erf: () => Pl,
+                exp: () => $l,
+                expandDims: () => Bl,
+                expm1: () => zl,
+                eye: () => Wl,
+                fft: () => Ph,
+                fill: () => Vl,
+                findBackend: () => hu,
+                findBackendFactory: () => pu,
+                floor: () => jl,
+                floorDiv: () => yu,
+                forceHalfFloat: () => HO,
+                fused: () => bp,
+                gather: () => Gl,
+                gatherND: () => dp,
+                gather_util: () => oo,
+                getBackend: () => lu,
+                getGradient: () => ha,
+                getKernel: () => ca,
+                getKernelsForBackend: () => pa,
+                getThreadsCount: () => Mj,
+                gpgpu_util: () => vD,
+                grad: () => sc,
+                grads: () => ic,
+                greater: () => Hl,
+                greaterEqual: () => ql,
+                ifft: () => $h,
+                imag: () => Kl,
+                image: () => Nd,
+                inTopKAsync: () => yp,
+                initializers: () => Ug,
+                input: () => Xv,
+                io: () => Ci,
+                irfft: () => Bh,
+                isFinite: () => Yl,
+                isInf: () => Xl,
+                isNaN: () => Ql,
+                keep: () => su,
+                kernel_impls: () => jf,
+                layers: () => ib,
+                leakyRelu: () => Zl,
+                less: () => Jl,
+                lessEqual: () => ec,
+                linalg: () => Cd,
+                linspace: () => tc,
+                loadGraphModel: () => AA,
+                loadLayersModel: () => Yv,
+                localResponseNormalization: () => nc,
+                log: () => rc,
+                log1p: () => ac,
+                logSigmoid: () => fc,
+                logSoftmax: () => bc,
+                logSumExp: () => Nc,
+                logicalAnd: () => Cc,
+                logicalNot: () => Tc,
+                logicalOr: () => Rc,
+                logicalXor: () => _c,
+                losses: () => Td,
+                matMul: () => Gi,
+                math: () => ji,
+                max: () => mc,
+                maxPool: () => Fc,
+                maxPool3d: () => Mc,
+                maxPoolWithArgmax: () => Dc,
+                maximum: () => Oc,
+                mean: () => Lc,
+                memory: () => tu,
+                meshgrid: () => Bc,
+                metrics: () => bS,
+                min: () => zc,
+                minimum: () => Uc,
+                mirrorPad: () => Wc,
+                mod: () => Vc,
+                model: () => qv,
+                models: () => DS,
+                moments: () => Gc,
+                movingAverage: () => cp,
+                mul: () => vu,
+                multiRNNCell: () => Hc,
+                multinomial: () => qc,
+                neg: () => pc,
+                nextFrame: () => Vd,
+                norm: () => lp,
+                notEqual: () => Kc,
+                oneHot: () => Hi,
+                ones: () => $c,
+                onesLike: () => Yc,
+                op: () => Ms,
+                outerProduct: () => Xc,
+                pad: () => Qc,
+                pad1d: () => Zc,
+                pad2d: () => Jc,
+                pad3d: () => eh,
+                pad4d: () => th,
+                pool: () => rh,
+                pow: () => ah,
+                prelu: () => sh,
+                print: () => Ni,
+                prod: () => ih,
+                profile: () => nu,
+                rand: () => oh,
+                randomGamma: () => ch,
+                randomNormal: () => hh,
+                randomUniform: () => ph,
+                range: () => dh,
+                ready: () => uu,
+                real: () => fh,
+                reciprocal: () => mh,
+                registerBackend: () => du,
+                registerCallbackConstructor: () => Qv,
+                registerGradient: () => fa,
+                registerKernel: () => da,
+                registerOp: () => XS,
+                regularizers: () => OS,
+                relu: () => gh,
+                relu6: () => yh,
+                removeBackend: () => cu,
+                reshape: () => qu,
+                reverse: () => bh,
+                reverse1d: () => vh,
+                reverse2d: () => xh,
+                reverse3d: () => wh,
+                reverse4d: () => kh,
+                rfft: () => Uh,
+                round: () => Sh,
+                rsqrt: () => Ih,
+                scalar: () => Ah,
+                scatterND: () => hp,
+                scatter_util: () => lo,
+                selu: () => Eh,
+                separableConv2d: () => Nh,
+                sequential: () => Kv,
+                serialization: () => Fo,
+                setBackend: () => ou,
+                setPlatform: () => mu,
+                setThreadsCount: () => Fj,
+                setWasmPath: () => Cj,
+                setWasmPaths: () => Tj,
+                setWebGLContext: () => KF,
+                setdiff1dAsync: () => Ch,
+                sigmoid: () => Qu,
+                sign: () => Th,
+                signal: () => Ed,
+                sin: () => Rh,
+                sinh: () => _h,
+                slice: () => Zu,
+                slice1d: () => Fh,
+                slice2d: () => Mh,
+                slice3d: () => Dh,
+                slice4d: () => Oh,
+                slice_util: () => fo,
+                softmax: () => Lh,
+                softplus: () => dc,
+                spaceToBatchND: () => nh,
+                sparse: () => Rd,
+                sparseToDense: () => pp,
+                spectral: () => Ad,
+                split: () => zh,
+                sqrt: () => Wh,
+                square: () => jc,
+                squaredDifference: () => Vh,
+                squeeze: () => jh,
+                stack: () => Gh,
+                step: () => Hh,
+                stridedSlice: () => qh,
+                string: () => _d,
+                sub: () => gc,
+                sum: () => yc,
+                sumOutType: () => hs,
+                tan: () => Kh,
+                tanh: () => Ju,
+                tensor: () => Ls,
+                tensor1d: () => Yh,
+                tensor2d: () => Xh,
+                tensor3d: () => to,
+                tensor4d: () => Qh,
+                tensor5d: () => Zh,
+                tensor6d: () => Jh,
+                tensor_util: () => us,
+                test_util: () => Lo,
+                tidy: () => ru,
+                tile: () => Ul,
+                time: () => iu,
+                topk: () => ep,
+                train: () => Ud,
+                transpose: () => qi,
+                truncatedNormal: () => tp,
+                unique: () => np,
+                unregisterGradient: () => ga,
+                unregisterKernel: () => ma,
+                unsortedSegmentSum: () => rp,
+                unstack: () => ap,
+                upcastType: () => cs,
+                util: () => va,
+                valueAndGrad: () => oc,
+                valueAndGrads: () => uc,
+                variable: () => sp,
+                variableGrads: () => lc,
+                version: () => Oj,
+                version_converter: () => EA,
+                version_core: () => Ko,
+                version_layers: () => mv,
+                version_wasm: () => Dj,
+                version_webgl: () => GO,
+                webgl: () => qO,
+                webgl_util: () => BF,
+                where: () => _l,
+                whereAsync: () => op,
+                zeros: () => Pc,
+                zerosLike: () => Fl
+            });
+            var f = Object.create,
+                m = Object.defineProperty,
+                g = Object.getOwnPropertyDescriptor,
+                y = Object.getOwnPropertyNames,
+                b = Object.getPrototypeOf,
+                v = Object.prototype.hasOwnProperty,
+                x = (e, t) => function () {
+                    return t || (0, e[y(e)[0]])((t = {
+                        exports: {}
+                    }).exports, t), t.exports
+                },
+                w = (e, t) => {
+                    for (var n in t) m(e, n, {
+                        get: t[n],
+                        enumerable: !0
+                    })
+                },
+                k = (e, t) => {
+                    return ((e, t, n, r) => {
+                        if (t && "object" == typeof t || "function" == typeof t)
+                            for (let a of y(t)) v.call(e, a) || !n && "default" === a || m(e, a, {
+                                get: () => t[a],
+                                enumerable: !(r = g(t, a)) || r.enumerable
+                            });
+                        return e
+                    })((n = m(null != e ? f(b(e)) : {}, "default", !t && e && e.__esModule ? {
+                        get: () => e.default,
+                        enumerable: !0
+                    } : {
+                        value: e,
+                        enumerable: !0
+                    }), m(n, "__esModule", {
+                        value: !0
+                    })), e);
+                    var n
+                },
+                S = x({
+                    "node_modules/.pnpm/long@4.0.0/node_modules/long/src/long.js"(e, t) {
+                        t.exports = r;
+                        var n = null;
+                        try {
+                            n = new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0, 97, 115, 109, 1, 0, 0, 0, 1, 13, 2, 96, 0, 1, 127, 96, 4, 127, 127, 127, 127, 1, 127, 3, 7, 6, 0, 1, 1, 1, 1, 1, 6, 6, 1, 127, 1, 65, 0, 11, 7, 50, 6, 3, 109, 117, 108, 0, 1, 5, 100, 105, 118, 95, 115, 0, 2, 5, 100, 105, 118, 95, 117, 0, 3, 5, 114, 101, 109, 95, 115, 0, 4, 5, 114, 101, 109, 95, 117, 0, 5, 8, 103, 101, 116, 95, 104, 105, 103, 104, 0, 0, 10, 191, 1, 6, 4, 0, 35, 0, 11, 36, 1, 1, 126, 32, 0, 173, 32, 1, 173, 66, 32, 134, 132, 32, 2, 173, 32, 3, 173, 66, 32, 134, 132, 126, 34, 4, 66, 32, 135, 167, 36, 0, 32, 4, 167, 11, 36, 1, 1, 126, 32, 0, 173, 32, 1, 173, 66, 32, 134, 132, 32, 2, 173, 32, 3, 173, 66, 32, 134, 132, 127, 34, 4, 66, 32, 135, 167, 36, 0, 32, 4, 167, 11, 36, 1, 1, 126, 32, 0, 173, 32, 1, 173, 66, 32, 134, 132, 32, 2, 173, 32, 3, 173, 66, 32, 134, 132, 128, 34, 4, 66, 32, 135, 167, 36, 0, 32, 4, 167, 11, 36, 1, 1, 126, 32, 0, 173, 32, 1, 173, 66, 32, 134, 132, 32, 2, 173, 32, 3, 173, 66, 32, 134, 132, 129, 34, 4, 66, 32, 135, 167, 36, 0, 32, 4, 167, 11, 36, 1, 1, 126, 32, 0, 173, 32, 1, 173, 66, 32, 134, 132, 32, 2, 173, 32, 3, 173, 66, 32, 134, 132, 130, 34, 4, 66, 32, 135, 167, 36, 0, 32, 4, 167, 11])), {}).exports
+                        } catch (e) {}
+
+                        function r(e, t, n) {
+                            this.low = 0 | e, this.high = 0 | t, this.unsigned = !!n
+                        }
+
+                        function a(e) {
+                            return !0 === (e && e.__isLong__)
+                        }
+                        Object.defineProperty(r.prototype, "__isLong__", {
+                            value: !0
+                        }), r.isLong = a;
+                        var s = {},
+                            i = {};
+
+                        function o(e, t) {
+                            var n, r, a;
+                            return t ? (a = 0 <= (e >>>= 0) && e < 256) && (r = i[e]) ? r : (n = l(e, (0 | e) < 0 ? -1 : 0, !0), a && (i[e] = n), n) : (a = -128 <= (e |= 0) && e < 128) && (r = s[e]) ? r : (n = l(e, e < 0 ? -1 : 0, !1), a && (s[e] = n), n)
+                        }
+
+                        function u(e, t) {
+                            if (isNaN(e)) return t ? b : y;
+                            if (t) {
+                                if (e < 0) return b;
+                                if (e >= f) return S
+                            } else {
+                                if (e <= -m) return I;
+                                if (e + 1 >= m) return k
+                            }
+                            return e < 0 ? u(-e, t).neg() : l(e % d | 0, e / d | 0, t)
+                        }
+
+                        function l(e, t, n) {
+                            return new r(e, t, n)
+                        }
+                        r.fromInt = o, r.fromNumber = u, r.fromBits = l;
+                        var c = Math.pow;
+
+                        function h(e, t, n) {
+                            if (0 === e.length) throw Error("empty string");
+                            if ("NaN" === e || "Infinity" === e || "+Infinity" === e || "-Infinity" === e) return y;
+                            if ("number" == typeof t ? (n = t, t = !1) : t = !!t, (n = n || 10) < 2 || 36 < n) throw RangeError("radix");
+                            var r;
+                            if ((r = e.indexOf("-")) > 0) throw Error("interior hyphen");
+                            if (0 === r) return h(e.substring(1), t, n).neg();
+                            for (var a = u(c(n, 8)), s = y, i = 0; i < e.length; i += 8) {
+                                var o = Math.min(8, e.length - i),
+                                    l = parseInt(e.substring(i, i + o), n);
+                                if (o < 8) {
+                                    var p = u(c(n, o));
+                                    s = s.mul(p).add(u(l))
+                                } else s = (s = s.mul(a)).add(u(l))
+                            }
+                            return s.unsigned = t, s
+                        }
+
+                        function p(e, t) {
+                            return "number" == typeof e ? u(e, t) : "string" == typeof e ? h(e, t) : l(e.low, e.high, "boolean" == typeof t ? t : e.unsigned)
+                        }
+                        r.fromString = h, r.fromValue = p;
+                        var d = 4294967296,
+                            f = d * d,
+                            m = f / 2,
+                            g = o(1 << 24),
+                            y = o(0);
+                        r.ZERO = y;
+                        var b = o(0, !0);
+                        r.UZERO = b;
+                        var v = o(1);
+                        r.ONE = v;
+                        var x = o(1, !0);
+                        r.UONE = x;
+                        var w = o(-1);
+                        r.NEG_ONE = w;
+                        var k = l(-1, 2147483647, !1);
+                        r.MAX_VALUE = k;
+                        var S = l(-1, -1, !0);
+                        r.MAX_UNSIGNED_VALUE = S;
+                        var I = l(0, -2147483648, !1);
+                        r.MIN_VALUE = I;
+                        var A = r.prototype;
+                        A.toInt = function () {
+                            return this.unsigned ? this.low >>> 0 : this.low
+                        }, A.toNumber = function () {
+                            return this.unsigned ? (this.high >>> 0) * d + (this.low >>> 0) : this.high * d + (this.low >>> 0)
+                        }, A.toString = function (e) {
+                            if ((e = e || 10) < 2 || 36 < e) throw RangeError("radix");
+                            if (this.isZero()) return "0";
+                            if (this.isNegative()) {
+                                if (this.eq(I)) {
+                                    var t = u(e),
+                                        n = this.div(t),
+                                        r = n.mul(t).sub(this);
+                                    return n.toString(e) + r.toInt().toString(e)
+                                }
+                                return "-" + this.neg().toString(e)
+                            }
+                            for (var a = u(c(e, 6), this.unsigned), s = this, i = "";;) {
+                                var o = s.div(a),
+                                    l = (s.sub(o.mul(a)).toInt() >>> 0).toString(e);
+                                if ((s = o).isZero()) return l + i;
+                                for (; l.length < 6;) l = "0" + l;
+                                i = "" + l + i
+                            }
+                        }, A.getHighBits = function () {
+                            return this.high
+                        }, A.getHighBitsUnsigned = function () {
+                            return this.high >>> 0
+                        }, A.getLowBits = function () {
+                            return this.low
+                        }, A.getLowBitsUnsigned = function () {
+                            return this.low >>> 0
+                        }, A.getNumBitsAbs = function () {
+                            if (this.isNegative()) return this.eq(I) ? 64 : this.neg().getNumBitsAbs();
+                            for (var e = 0 != this.high ? this.high : this.low, t = 31; t > 0 && 0 == (e & 1 << t); t--);
+                            return 0 != this.high ? t + 33 : t + 1
+                        }, A.isZero = function () {
+                            return 0 === this.high && 0 === this.low
+                        }, A.eqz = A.isZero, A.isNegative = function () {
+                            return !this.unsigned && this.high < 0
+                        }, A.isPositive = function () {
+                            return this.unsigned || this.high >= 0
+                        }, A.isOdd = function () {
+                            return 1 == (1 & this.low)
+                        }, A.isEven = function () {
+                            return 0 == (1 & this.low)
+                        }, A.equals = function (e) {
+                            return a(e) || (e = p(e)), (this.unsigned === e.unsigned || this.high >>> 31 != 1 || e.high >>> 31 != 1) && (this.high === e.high && this.low === e.low)
+                        }, A.eq = A.equals, A.notEquals = function (e) {
+                            return !this.eq(e)
+                        }, A.neq = A.notEquals, A.ne = A.notEquals, A.lessThan = function (e) {
+                            return this.comp(e) < 0
+                        }, A.lt = A.lessThan, A.lessThanOrEqual = function (e) {
+                            return this.comp(e) <= 0
+                        }, A.lte = A.lessThanOrEqual, A.le = A.lessThanOrEqual, A.greaterThan = function (e) {
+                            return this.comp(e) > 0
+                        }, A.gt = A.greaterThan, A.greaterThanOrEqual = function (e) {
+                            return this.comp(e) >= 0
+                        }, A.gte = A.greaterThanOrEqual, A.ge = A.greaterThanOrEqual, A.compare = function (e) {
+                            if (a(e) || (e = p(e)), this.eq(e)) return 0;
+                            var t = this.isNegative(),
+                                n = e.isNegative();
+                            return t && !n ? -1 : !t && n ? 1 : this.unsigned ? e.high >>> 0 > this.high >>> 0 || e.high === this.high && e.low >>> 0 > this.low >>> 0 ? -1 : 1 : this.sub(e).isNegative() ? -1 : 1
+                        }, A.comp = A.compare, A.negate = function () {
+                            return !this.unsigned && this.eq(I) ? I : this.not().add(v)
+                        }, A.neg = A.negate, A.add = function (e) {
+                            a(e) || (e = p(e));
+                            var t = this.high >>> 16,
+                                n = 65535 & this.high,
+                                r = this.low >>> 16,
+                                s = 65535 & this.low,
+                                i = e.high >>> 16,
+                                o = 65535 & e.high,
+                                u = e.low >>> 16,
+                                c = 0,
+                                h = 0,
+                                d = 0,
+                                f = 0;
+                            return d += (f += s + (65535 & e.low)) >>> 16, h += (d += r + u) >>> 16, c += (h += n + o) >>> 16, c += t + i, l((d &= 65535) << 16 | (f &= 65535), (c &= 65535) << 16 | (h &= 65535), this.unsigned)
+                        }, A.subtract = function (e) {
+                            return a(e) || (e = p(e)), this.add(e.neg())
+                        }, A.sub = A.subtract, A.multiply = function (e) {
+                            if (this.isZero()) return y;
+                            if (a(e) || (e = p(e)), n) return l(n.mul(this.low, this.high, e.low, e.high), n.get_high(), this.unsigned);
+                            if (e.isZero()) return y;
+                            if (this.eq(I)) return e.isOdd() ? I : y;
+                            if (e.eq(I)) return this.isOdd() ? I : y;
+                            if (this.isNegative()) return e.isNegative() ? this.neg().mul(e.neg()) : this.neg().mul(e).neg();
+                            if (e.isNegative()) return this.mul(e.neg()).neg();
+                            if (this.lt(g) && e.lt(g)) return u(this.toNumber() * e.toNumber(), this.unsigned);
+                            var t = this.high >>> 16,
+                                r = 65535 & this.high,
+                                s = this.low >>> 16,
+                                i = 65535 & this.low,
+                                o = e.high >>> 16,
+                                c = 65535 & e.high,
+                                h = e.low >>> 16,
+                                d = 65535 & e.low,
+                                f = 0,
+                                m = 0,
+                                b = 0,
+                                v = 0;
+                            return b += (v += i * d) >>> 16, m += (b += s * d) >>> 16, b &= 65535, m += (b += i * h) >>> 16, f += (m += r * d) >>> 16, m &= 65535, f += (m += s * h) >>> 16, m &= 65535, f += (m += i * c) >>> 16, f += t * d + r * h + s * c + i * o, l((b &= 65535) << 16 | (v &= 65535), (f &= 65535) << 16 | (m &= 65535), this.unsigned)
+                        }, A.mul = A.multiply, A.divide = function (e) {
+                            if (a(e) || (e = p(e)), e.isZero()) throw Error("division by zero");
+                            var t, r, s;
+                            if (n) return this.unsigned || -2147483648 !== this.high || -1 !== e.low || -1 !== e.high ? l((this.unsigned ? n.div_u : n.div_s)(this.low, this.high, e.low, e.high), n.get_high(), this.unsigned) : this;
+                            if (this.isZero()) return this.unsigned ? b : y;
+                            if (this.unsigned) {
+                                if (e.unsigned || (e = e.toUnsigned()), e.gt(this)) return b;
+                                if (e.gt(this.shru(1))) return x;
+                                s = b
+                            } else {
+                                if (this.eq(I)) return e.eq(v) || e.eq(w) ? I : e.eq(I) ? v : (t = this.shr(1).div(e).shl(1)).eq(y) ? e.isNegative() ? v : w : (r = this.sub(e.mul(t)), s = t.add(r.div(e)));
+                                if (e.eq(I)) return this.unsigned ? b : y;
+                                if (this.isNegative()) return e.isNegative() ? this.neg().div(e.neg()) : this.neg().div(e).neg();
+                                if (e.isNegative()) return this.div(e.neg()).neg();
+                                s = y
+                            }
+                            for (r = this; r.gte(e);) {
+                                t = Math.max(1, Math.floor(r.toNumber() / e.toNumber()));
+                                for (var i = Math.ceil(Math.log(t) / Math.LN2), o = i <= 48 ? 1 : c(2, i - 48), h = u(t), d = h.mul(e); d.isNegative() || d.gt(r);) d = (h = u(t -= o, this.unsigned)).mul(e);
+                                h.isZero() && (h = v), s = s.add(h), r = r.sub(d)
+                            }
+                            return s
+                        }, A.div = A.divide, A.modulo = function (e) {
+                            return a(e) || (e = p(e)), n ? l((this.unsigned ? n.rem_u : n.rem_s)(this.low, this.high, e.low, e.high), n.get_high(), this.unsigned) : this.sub(this.div(e).mul(e))
+                        }, A.mod = A.modulo, A.rem = A.modulo, A.not = function () {
+                            return l(~this.low, ~this.high, this.unsigned)
+                        }, A.and = function (e) {
+                            return a(e) || (e = p(e)), l(this.low & e.low, this.high & e.high, this.unsigned)
+                        }, A.or = function (e) {
+                            return a(e) || (e = p(e)), l(this.low | e.low, this.high | e.high, this.unsigned)
+                        }, A.xor = function (e) {
+                            return a(e) || (e = p(e)), l(this.low ^ e.low, this.high ^ e.high, this.unsigned)
+                        }, A.shiftLeft = function (e) {
+                            return a(e) && (e = e.toInt()), 0 == (e &= 63) ? this : e < 32 ? l(this.low << e, this.high << e | this.low >>> 32 - e, this.unsigned) : l(0, this.low << e - 32, this.unsigned)
+                        }, A.shl = A.shiftLeft, A.shiftRight = function (e) {
+                            return a(e) && (e = e.toInt()), 0 == (e &= 63) ? this : e < 32 ? l(this.low >>> e | this.high << 32 - e, this.high >> e, this.unsigned) : l(this.high >> e - 32, this.high >= 0 ? 0 : -1, this.unsigned)
+                        }, A.shr = A.shiftRight, A.shiftRightUnsigned = function (e) {
+                            if (a(e) && (e = e.toInt()), 0 === (e &= 63)) return this;
+                            var t = this.high;
+                            return e < 32 ? l(this.low >>> e | t << 32 - e, t >>> e, this.unsigned) : l(32 === e ? t : t >>> e - 32, 0, this.unsigned)
+                        }, A.shru = A.shiftRightUnsigned, A.shr_u = A.shiftRightUnsigned, A.toSigned = function () {
+                            return this.unsigned ? l(this.low, this.high, !1) : this
+                        }, A.toUnsigned = function () {
+                            return this.unsigned ? this : l(this.low, this.high, !0)
+                        }, A.toBytes = function (e) {
+                            return e ? this.toBytesLE() : this.toBytesBE()
+                        }, A.toBytesLE = function () {
+                            var e = this.high,
+                                t = this.low;
+                            return [255 & t, t >>> 8 & 255, t >>> 16 & 255, t >>> 24, 255 & e, e >>> 8 & 255, e >>> 16 & 255, e >>> 24]
+                        }, A.toBytesBE = function () {
+                            var e = this.high,
+                                t = this.low;
+                            return [e >>> 24, e >>> 16 & 255, e >>> 8 & 255, 255 & e, t >>> 24, t >>> 16 & 255, t >>> 8 & 255, 255 & t]
+                        }, r.fromBytes = function (e, t, n) {
+                            return n ? r.fromBytesLE(e, t) : r.fromBytesBE(e, t)
+                        }, r.fromBytesLE = function (e, t) {
+                            return new r(e[0] | e[1] << 8 | e[2] << 16 | e[3] << 24, e[4] | e[5] << 8 | e[6] << 16 | e[7] << 24, t)
+                        }, r.fromBytesBE = function (e, t) {
+                            return new r(e[4] << 24 | e[5] << 16 | e[6] << 8 | e[7], e[0] << 24 | e[1] << 16 | e[2] << 8 | e[3], t)
+                        }
+                    }
+                }),
+                I = x({
+                    "(disabled):node_modules/.pnpm/node-fetch@2.6.6/node_modules/node-fetch/browser.js"() {}
+                }),
+                A = x({
+                    "(disabled):util"() {}
+                }),
+                E = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/alea.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t, n = this,
+                                    r = (t = 4022871197, function (e) {
+                                        e = e.toString();
+                                        for (var n = 0; n < e.length; n++) {
+                                            var r = .02519603282416938 * (t += e.charCodeAt(n));
+                                            r -= t = r >>> 0, t = (r *= t) >>> 0, t += 4294967296 * (r -= t)
+                                        }
+                                        return 2.3283064365386963e-10 * (t >>> 0)
+                                    });
+                                n.next = function () {
+                                    var e = 2091639 * n.s0 + 2.3283064365386963e-10 * n.c;
+                                    return n.s0 = n.s1, n.s1 = n.s2, n.s2 = e - (n.c = 0 | e)
+                                }, n.c = 1, n.s0 = r(" "), n.s1 = r(" "), n.s2 = r(" "), n.s0 -= r(e), n.s0 < 0 && (n.s0 += 1), n.s1 -= r(e), n.s1 < 0 && (n.s1 += 1), n.s2 -= r(e), n.s2 < 0 && (n.s2 += 1), r = null
+                            }
+
+                            function a(e, t) {
+                                return t.c = e.c, t.s0 = e.s0, t.s1 = e.s1, t.s2 = e.s2, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = n.next;
+                                return i.int32 = function () {
+                                    return 4294967296 * n.next() | 0
+                                }, i.double = function () {
+                                    return i() + 11102230246251565e-32 * (2097152 * i() | 0)
+                                }, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.alea = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                N = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xor128.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this,
+                                    n = "";
+                                t.x = 0, t.y = 0, t.z = 0, t.w = 0, t.next = function () {
+                                    var e = t.x ^ t.x << 11;
+                                    return t.x = t.y, t.y = t.z, t.z = t.w, t.w ^= t.w >>> 19 ^ e ^ e >>> 8
+                                }, e === (0 | e) ? t.x = e : n += e;
+                                for (var r = 0; r < n.length + 64; r++) t.x ^= 0 | n.charCodeAt(r), t.next()
+                            }
+
+                            function a(e, t) {
+                                return t.x = e.x, t.y = e.y, t.z = e.z, t.w = e.w, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xor128 = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                C = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xorwow.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this,
+                                    n = "";
+                                t.next = function () {
+                                    var e = t.x ^ t.x >>> 2;
+                                    return t.x = t.y, t.y = t.z, t.z = t.w, t.w = t.v, (t.d = t.d + 362437 | 0) + (t.v = t.v ^ t.v << 4 ^ e ^ e << 1) | 0
+                                }, t.x = 0, t.y = 0, t.z = 0, t.w = 0, t.v = 0, e === (0 | e) ? t.x = e : n += e;
+                                for (var r = 0; r < n.length + 64; r++) t.x ^= 0 | n.charCodeAt(r), r == n.length && (t.d = t.x << 10 ^ t.x >>> 4), t.next()
+                            }
+
+                            function a(e, t) {
+                                return t.x = e.x, t.y = e.y, t.z = e.z, t.w = e.w, t.v = e.v, t.d = e.d, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xorwow = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                T = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xorshift7.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this;
+                                t.next = function () {
+                                        var e, n, r = t.x,
+                                            a = t.i;
+                                        return e = r[a], n = (e ^= e >>> 7) ^ e << 24, n ^= (e = r[a + 1 & 7]) ^ e >>> 10, n ^= (e = r[a + 3 & 7]) ^ e >>> 3, n ^= (e = r[a + 4 & 7]) ^ e << 7, e = r[a + 7 & 7], n ^= (e ^= e << 13) ^ e << 9, r[a] = n, t.i = a + 1 & 7, n
+                                    },
+                                    function (e, t) {
+                                        var n, r = [];
+                                        if (t === (0 | t)) r[0] = t;
+                                        else
+                                            for (t = "" + t, n = 0; n < t.length; ++n) r[7 & n] = r[7 & n] << 15 ^ t.charCodeAt(n) + r[n + 1 & 7] << 13;
+                                        for (; r.length < 8;) r.push(0);
+                                        for (n = 0; n < 8 && 0 === r[n]; ++n);
+                                        for (8 == n && (r[7] = -1), e.x = r, e.i = 0, n = 256; n > 0; --n) e.next()
+                                    }(t, e)
+                            }
+
+                            function a(e, t) {
+                                return t.x = e.x.slice(), t.i = e.i, t
+                            }
+
+                            function s(e, t) {
+                                null == e && (e = +new Date);
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && (s.x && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xorshift7 = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                R = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/xor4096.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this;
+                                t.next = function () {
+                                        var e, n, r = t.w,
+                                            a = t.X,
+                                            s = t.i;
+                                        return t.w = r = r + 1640531527 | 0, n = a[s + 34 & 127], e = a[s = s + 1 & 127], n ^= n << 13, e ^= e << 17, n ^= n >>> 15, e ^= e >>> 12, n = a[s] = n ^ e, t.i = s, n + (r ^ r >>> 16) | 0
+                                    },
+                                    function (e, t) {
+                                        var n, r, a, s, i, o = [],
+                                            u = 128;
+                                        for (t === (0 | t) ? (r = t, t = null) : (t += "\0", r = 0, u = Math.max(u, t.length)), a = 0, s = -32; s < u; ++s) t && (r ^= t.charCodeAt((s + 32) % t.length)), 0 === s && (i = r), r ^= r << 10, r ^= r >>> 15, r ^= r << 4, r ^= r >>> 13, s >= 0 && (i = i + 1640531527 | 0, a = 0 == (n = o[127 & s] ^= r + i) ? a + 1 : 0);
+                                        for (a >= 128 && (o[127 & (t && t.length || 0)] = -1), a = 127, s = 512; s > 0; --s) r = o[a + 34 & 127], n = o[a = a + 1 & 127], r ^= r << 13, n ^= n << 17, r ^= r >>> 15, n ^= n >>> 12, o[a] = r ^ n;
+                                        e.w = i, e.X = o, e.i = a
+                                    }(t, e)
+                            }
+
+                            function a(e, t) {
+                                return t.i = e.i, t.w = e.w, t.X = e.X.slice(), t
+                            }
+
+                            function s(e, t) {
+                                null == e && (e = +new Date);
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && (s.X && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xor4096 = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                _ = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/lib/tychei.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this,
+                                    n = "";
+                                t.next = function () {
+                                    var e = t.b,
+                                        n = t.c,
+                                        r = t.d,
+                                        a = t.a;
+                                    return e = e << 25 ^ e >>> 7 ^ n, n = n - r | 0, r = r << 24 ^ r >>> 8 ^ a, a = a - e | 0, t.b = e = e << 20 ^ e >>> 12 ^ n, t.c = n = n - r | 0, t.d = r << 16 ^ n >>> 16 ^ a, t.a = a - e | 0
+                                }, t.a = 0, t.b = 0, t.c = -1640531527, t.d = 1367130551, e === Math.floor(e) ? (t.a = e / 4294967296 | 0, t.b = 0 | e) : n += e;
+                                for (var r = 0; r < n.length + 20; r++) t.b ^= 0 | n.charCodeAt(r), t.next()
+                            }
+
+                            function a(e, t) {
+                                return t.a = e.a, t.b = e.b, t.c = e.c, t.d = e.d, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.tychei = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                F = x({
+                    "(disabled):crypto"() {}
+                }),
+                M = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/seedrandom.js"(e, t) {
+                        ! function (e, r) {
+                            var a, s = this,
+                                i = r.pow(256, 6),
+                                o = r.pow(2, 52),
+                                u = 2 * o;
+
+                            function l(t, n, l) {
+                                var f = [],
+                                    m = p(function e(t, n) {
+                                        var r, a = [],
+                                            s = typeof t;
+                                        if (n && "object" == s)
+                                            for (r in t) try {
+                                                a.push(e(t[r], n - 1))
+                                            } catch (e) {}
+                                        return a.length ? a : "string" == s ? t : t + "\0"
+                                    }((n = 1 == n ? {
+                                        entropy: !0
+                                    } : n || {}).entropy ? [t, d(e)] : null == t ? function () {
+                                        try {
+                                            var t;
+                                            return a && (t = a.randomBytes) ? t = t(256) : (t = new Uint8Array(256), (s.crypto || s.msCrypto).getRandomValues(t)), d(t)
+                                        } catch (t) {
+                                            var n = s.navigator,
+                                                r = n && n.plugins;
+                                            return [+new Date, s, r, s.screen, d(e)]
+                                        }
+                                    }() : t, 3), f),
+                                    g = new c(f),
+                                    y = function () {
+                                        for (var e = g.g(6), t = i, n = 0; e < o;) e = 256 * (e + n), t *= 256, n = g.g(1);
+                                        for (; e >= u;) e /= 2, t /= 2, n >>>= 1;
+                                        return (e + n) / t
+                                    };
+                                return y.int32 = function () {
+                                    return 0 | g.g(4)
+                                }, y.quick = function () {
+                                    return g.g(4) / 4294967296
+                                }, y.double = y, p(d(g.S), e), (n.pass || l || function (e, t, n, a) {
+                                    return a && (a.S && h(a, g), e.state = function () {
+                                        return h(g, {})
+                                    }), n ? (r.random = e, t) : e
+                                })(y, m, "global" in n ? n.global : this == r, n.state)
+                            }
+
+                            function c(e) {
+                                var t, n = e.length,
+                                    r = this,
+                                    a = 0,
+                                    s = r.i = r.j = 0,
+                                    i = r.S = [];
+                                for (n || (e = [n++]); a < 256;) i[a] = a++;
+                                for (a = 0; a < 256; a++) i[a] = i[s = 255 & s + e[a % n] + (t = i[a])], i[s] = t;
+                                (r.g = function (e) {
+                                    for (var t, n = 0, a = r.i, s = r.j, i = r.S; e--;) t = i[a = 255 & a + 1], n = 256 * n + i[255 & (i[a] = i[s = 255 & s + t]) + (i[s] = t)];
+                                    return r.i = a, r.j = s, n
+                                })(256)
+                            }
+
+                            function h(e, t) {
+                                return t.i = e.i, t.j = e.j, t.S = e.S.slice(), t
+                            }
+
+                            function p(e, t) {
+                                for (var n, r = e + "", a = 0; a < r.length;) t[255 & a] = 255 & (n ^= 19 * t[255 & a]) + r.charCodeAt(a++);
+                                return d(t)
+                            }
+
+                            function d(e) {
+                                return String.fromCharCode.apply(0, e)
+                            }
+                            if (r.seedrandom = l, p(r.random(), e), "object" == typeof t && t.exports) {
+                                t.exports = l;
+                                try {
+                                    a = F()
+                                } catch (e) {}
+                            } else "function" == typeof define && n(67) && define((function () {
+                                return l
+                            }))
+                        }([], Math)
+                    }
+                }),
+                D = x({
+                    "node_modules/.pnpm/seedrandom@2.4.3/node_modules/seedrandom/index.js"(e, t) {
+                        var n = E(),
+                            r = N(),
+                            a = C(),
+                            s = T(),
+                            i = R(),
+                            o = _(),
+                            u = M();
+                        u.alea = n, u.xor128 = r, u.xorwow = a, u.xorshift7 = s, u.xor4096 = i, u.tychei = o, t.exports = u
+                    }
+                }),
+                O = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/alea.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t, n = this,
+                                    r = (t = 4022871197, function (e) {
+                                        e = String(e);
+                                        for (var n = 0; n < e.length; n++) {
+                                            var r = .02519603282416938 * (t += e.charCodeAt(n));
+                                            r -= t = r >>> 0, t = (r *= t) >>> 0, t += 4294967296 * (r -= t)
+                                        }
+                                        return 2.3283064365386963e-10 * (t >>> 0)
+                                    });
+                                n.next = function () {
+                                    var e = 2091639 * n.s0 + 2.3283064365386963e-10 * n.c;
+                                    return n.s0 = n.s1, n.s1 = n.s2, n.s2 = e - (n.c = 0 | e)
+                                }, n.c = 1, n.s0 = r(" "), n.s1 = r(" "), n.s2 = r(" "), n.s0 -= r(e), n.s0 < 0 && (n.s0 += 1), n.s1 -= r(e), n.s1 < 0 && (n.s1 += 1), n.s2 -= r(e), n.s2 < 0 && (n.s2 += 1), r = null
+                            }
+
+                            function a(e, t) {
+                                return t.c = e.c, t.s0 = e.s0, t.s1 = e.s1, t.s2 = e.s2, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = n.next;
+                                return i.int32 = function () {
+                                    return 4294967296 * n.next() | 0
+                                }, i.double = function () {
+                                    return i() + 11102230246251565e-32 * (2097152 * i() | 0)
+                                }, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.alea = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                L = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xor128.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this,
+                                    n = "";
+                                t.x = 0, t.y = 0, t.z = 0, t.w = 0, t.next = function () {
+                                    var e = t.x ^ t.x << 11;
+                                    return t.x = t.y, t.y = t.z, t.z = t.w, t.w ^= t.w >>> 19 ^ e ^ e >>> 8
+                                }, e === (0 | e) ? t.x = e : n += e;
+                                for (var r = 0; r < n.length + 64; r++) t.x ^= 0 | n.charCodeAt(r), t.next()
+                            }
+
+                            function a(e, t) {
+                                return t.x = e.x, t.y = e.y, t.z = e.z, t.w = e.w, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xor128 = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                P = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xorwow.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this,
+                                    n = "";
+                                t.next = function () {
+                                    var e = t.x ^ t.x >>> 2;
+                                    return t.x = t.y, t.y = t.z, t.z = t.w, t.w = t.v, (t.d = t.d + 362437 | 0) + (t.v = t.v ^ t.v << 4 ^ e ^ e << 1) | 0
+                                }, t.x = 0, t.y = 0, t.z = 0, t.w = 0, t.v = 0, e === (0 | e) ? t.x = e : n += e;
+                                for (var r = 0; r < n.length + 64; r++) t.x ^= 0 | n.charCodeAt(r), r == n.length && (t.d = t.x << 10 ^ t.x >>> 4), t.next()
+                            }
+
+                            function a(e, t) {
+                                return t.x = e.x, t.y = e.y, t.z = e.z, t.w = e.w, t.v = e.v, t.d = e.d, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xorwow = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                $ = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xorshift7.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this;
+                                t.next = function () {
+                                        var e, n, r = t.x,
+                                            a = t.i;
+                                        return e = r[a], n = (e ^= e >>> 7) ^ e << 24, n ^= (e = r[a + 1 & 7]) ^ e >>> 10, n ^= (e = r[a + 3 & 7]) ^ e >>> 3, n ^= (e = r[a + 4 & 7]) ^ e << 7, e = r[a + 7 & 7], n ^= (e ^= e << 13) ^ e << 9, r[a] = n, t.i = a + 1 & 7, n
+                                    },
+                                    function (e, t) {
+                                        var n, r = [];
+                                        if (t === (0 | t)) r[0] = t;
+                                        else
+                                            for (t = "" + t, n = 0; n < t.length; ++n) r[7 & n] = r[7 & n] << 15 ^ t.charCodeAt(n) + r[n + 1 & 7] << 13;
+                                        for (; r.length < 8;) r.push(0);
+                                        for (n = 0; n < 8 && 0 === r[n]; ++n);
+                                        for (8 == n && (r[7] = -1), e.x = r, e.i = 0, n = 256; n > 0; --n) e.next()
+                                    }(t, e)
+                            }
+
+                            function a(e, t) {
+                                return t.x = e.x.slice(), t.i = e.i, t
+                            }
+
+                            function s(e, t) {
+                                null == e && (e = +new Date);
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && (s.x && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xorshift7 = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                B = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/xor4096.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this;
+                                t.next = function () {
+                                        var e, n, r = t.w,
+                                            a = t.X,
+                                            s = t.i;
+                                        return t.w = r = r + 1640531527 | 0, n = a[s + 34 & 127], e = a[s = s + 1 & 127], n ^= n << 13, e ^= e << 17, n ^= n >>> 15, e ^= e >>> 12, n = a[s] = n ^ e, t.i = s, n + (r ^ r >>> 16) | 0
+                                    },
+                                    function (e, t) {
+                                        var n, r, a, s, i, o = [],
+                                            u = 128;
+                                        for (t === (0 | t) ? (r = t, t = null) : (t += "\0", r = 0, u = Math.max(u, t.length)), a = 0, s = -32; s < u; ++s) t && (r ^= t.charCodeAt((s + 32) % t.length)), 0 === s && (i = r), r ^= r << 10, r ^= r >>> 15, r ^= r << 4, r ^= r >>> 13, s >= 0 && (i = i + 1640531527 | 0, a = 0 == (n = o[127 & s] ^= r + i) ? a + 1 : 0);
+                                        for (a >= 128 && (o[127 & (t && t.length || 0)] = -1), a = 127, s = 512; s > 0; --s) r = o[a + 34 & 127], n = o[a = a + 1 & 127], r ^= r << 13, n ^= n << 17, r ^= r >>> 15, n ^= n >>> 12, o[a] = r ^ n;
+                                        e.w = i, e.X = o, e.i = a
+                                    }(t, e)
+                            }
+
+                            function a(e, t) {
+                                return t.i = e.i, t.w = e.w, t.X = e.X.slice(), t
+                            }
+
+                            function s(e, t) {
+                                null == e && (e = +new Date);
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && (s.X && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.xor4096 = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                z = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/lib/tychei.js"(e, t) {
+                        ! function (e, t, n) {
+                            function r(e) {
+                                var t = this,
+                                    n = "";
+                                t.next = function () {
+                                    var e = t.b,
+                                        n = t.c,
+                                        r = t.d,
+                                        a = t.a;
+                                    return e = e << 25 ^ e >>> 7 ^ n, n = n - r | 0, r = r << 24 ^ r >>> 8 ^ a, a = a - e | 0, t.b = e = e << 20 ^ e >>> 12 ^ n, t.c = n = n - r | 0, t.d = r << 16 ^ n >>> 16 ^ a, t.a = a - e | 0
+                                }, t.a = 0, t.b = 0, t.c = -1640531527, t.d = 1367130551, e === Math.floor(e) ? (t.a = e / 4294967296 | 0, t.b = 0 | e) : n += e;
+                                for (var r = 0; r < n.length + 20; r++) t.b ^= 0 | n.charCodeAt(r), t.next()
+                            }
+
+                            function a(e, t) {
+                                return t.a = e.a, t.b = e.b, t.c = e.c, t.d = e.d, t
+                            }
+
+                            function s(e, t) {
+                                var n = new r(e),
+                                    s = t && t.state,
+                                    i = function () {
+                                        return (n.next() >>> 0) / 4294967296
+                                    };
+                                return i.double = function () {
+                                    do {
+                                        var e = ((n.next() >>> 11) + (n.next() >>> 0) / 4294967296) / (1 << 21)
+                                    } while (0 === e);
+                                    return e
+                                }, i.int32 = n.next, i.quick = i, s && ("object" == typeof s && a(s, n), i.state = function () {
+                                    return a(n, {})
+                                }), i
+                            }
+                            t && t.exports ? t.exports = s : n && n.amd ? n((function () {
+                                return s
+                            })) : this.tychei = s
+                        }(0, "object" == typeof t && t, "function" == typeof define && define)
+                    }
+                }),
+                U = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/seedrandom.js"(e, t) {
+                        ! function (e, r, a) {
+                            var s, i = a.pow(256, 6),
+                                o = a.pow(2, 52),
+                                u = 2 * o;
+
+                            function l(t, n, l) {
+                                var f = [],
+                                    m = p(function e(t, n) {
+                                        var r, a = [],
+                                            s = typeof t;
+                                        if (n && "object" == s)
+                                            for (r in t) try {
+                                                a.push(e(t[r], n - 1))
+                                            } catch (e) {}
+                                        return a.length ? a : "string" == s ? t : t + "\0"
+                                    }((n = 1 == n ? {
+                                        entropy: !0
+                                    } : n || {}).entropy ? [t, d(r)] : null == t ? function () {
+                                        try {
+                                            var t;
+                                            return s && (t = s.randomBytes) ? t = t(256) : (t = new Uint8Array(256), (e.crypto || e.msCrypto).getRandomValues(t)), d(t)
+                                        } catch (t) {
+                                            var n = e.navigator,
+                                                a = n && n.plugins;
+                                            return [+new Date, e, a, e.screen, d(r)]
+                                        }
+                                    }() : t, 3), f),
+                                    g = new c(f),
+                                    y = function () {
+                                        for (var e = g.g(6), t = i, n = 0; e < o;) e = 256 * (e + n), t *= 256, n = g.g(1);
+                                        for (; e >= u;) e /= 2, t /= 2, n >>>= 1;
+                                        return (e + n) / t
+                                    };
+                                return y.int32 = function () {
+                                    return 0 | g.g(4)
+                                }, y.quick = function () {
+                                    return g.g(4) / 4294967296
+                                }, y.double = y, p(d(g.S), r), (n.pass || l || function (e, t, n, r) {
+                                    return r && (r.S && h(r, g), e.state = function () {
+                                        return h(g, {})
+                                    }), n ? (a.random = e, t) : e
+                                })(y, m, "global" in n ? n.global : this == a, n.state)
+                            }
+
+                            function c(e) {
+                                var t, n = e.length,
+                                    r = this,
+                                    a = 0,
+                                    s = r.i = r.j = 0,
+                                    i = r.S = [];
+                                for (n || (e = [n++]); a < 256;) i[a] = a++;
+                                for (a = 0; a < 256; a++) i[a] = i[s = 255 & s + e[a % n] + (t = i[a])], i[s] = t;
+                                (r.g = function (e) {
+                                    for (var t, n = 0, a = r.i, s = r.j, i = r.S; e--;) t = i[a = 255 & a + 1], n = 256 * n + i[255 & (i[a] = i[s = 255 & s + t]) + (i[s] = t)];
+                                    return r.i = a, r.j = s, n
+                                })(256)
+                            }
+
+                            function h(e, t) {
+                                return t.i = e.i, t.j = e.j, t.S = e.S.slice(), t
+                            }
+
+                            function p(e, t) {
+                                for (var n, r = e + "", a = 0; a < r.length;) t[255 & a] = 255 & (n ^= 19 * t[255 & a]) + r.charCodeAt(a++);
+                                return d(t)
+                            }
+
+                            function d(e) {
+                                return String.fromCharCode.apply(0, e)
+                            }
+                            if (p(a.random(), r), "object" == typeof t && t.exports) {
+                                t.exports = l;
+                                try {
+                                    s = F()
+                                } catch (e) {}
+                            } else "function" == typeof define && n(67) ? define((function () {
+                                return l
+                            })) : a.seedrandom = l
+                        }("undefined" != typeof self ? self : e, [], Math)
+                    }
+                }),
+                W = x({
+                    "node_modules/.pnpm/seedrandom@3.0.5/node_modules/seedrandom/index.js"(e, t) {
+                        var n = O(),
+                            r = L(),
+                            a = P(),
+                            s = $(),
+                            i = B(),
+                            o = z(),
+                            u = U();
+                        u.alea = n, u.xor128 = r, u.xorwow = a, u.xorshift7 = s, u.xor4096 = i, u.tychei = o, t.exports = u
+                    }
+                }),
+                V = x({
+                    "(disabled):node_modules/.pnpm/string_decoder@1.1.1/node_modules/string_decoder/lib/string_decoder.js"() {}
+                }),
+                j = x({
+                    "(disabled):fs"() {}
+                }),
+                G = x({
+                    "(disabled):path"() {}
+                }),
+                H = x({
+                    "(disabled):worker_threads"() {}
+                }),
+                q = x({
+                    "(disabled):perf_hooks"() {}
+                }),
+                K = x({
+                    "(disabled):os"() {}
+                }),
+                Y = x({
+                    "node_modules/.pnpm/@tensorflow+tfjs-backend-wasm@3.13.0_@tensorflow+tfjs-core@3.13.0/node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js"(t, i) {
+                        var o, u = (o = (o = "undefined" != typeof document && document.currentScript ? document.currentScript.src : void 0) || e, function (e) {
+                            function t() {
+                                return M.buffer != L && ee(M.buffer), P
+                            }
+
+                            function n() {
+                                return M.buffer != L && ee(M.buffer), $
+                            }
+
+                            function i() {
+                                return M.buffer != L && ee(M.buffer), B
+                            }
+
+                            function u() {
+                                return M.buffer != L && ee(M.buffer), z
+                            }
+
+                            function l() {
+                                return M.buffer != L && ee(M.buffer), U
+                            }
+                            var c, h, p, d = void 0 !== (e = e || {}) ? e : {};
+                            d.ready = new Promise((function (e, t) {
+                                c = e, h = t
+                            })), void 0 !== r && r.listeners && (p = {
+                                uncaughtException: r.listeners("uncaughtException"),
+                                unhandledRejection: r.listeners("unhandledRejection")
+                            });
+                            var f, m = {};
+                            for (f in d) d.hasOwnProperty(f) && (m[f] = d[f]);
+                            var g, y, b, v, x = function (e, t) {
+                                throw t
+                            };
+                            g = "object" == typeof window, y = "function" == typeof importScripts, b = "object" == typeof r && "object" == typeof r.versions && "string" == typeof r.versions.node, v = !g && !b && !y;
+                            var w = d.ENVIRONMENT_IS_PTHREAD || !1;
+                            w && (L = d.buffer);
+                            var k, S, I, A, E, N = "";
+
+                            function C(e) {
+                                return d.locateFile ? d.locateFile(e, N) : N + e
+                            }
+                            if (b) {
+                                var T;
+                                N = y ? G().dirname(N) + "/" : a + "/", k = function (e, t) {
+                                    return A || (A = j()), E || (E = G()), e = E.normalize(e), A.readFileSync(e, t ? null : "utf8")
+                                }, I = function (e) {
+                                    var t = k(e, !0);
+                                    return t.buffer || (t = new Uint8Array(t)), V(t.buffer), t
+                                }, r.argv.length > 1 && r.argv[1].replace(/\\/g, "/"), r.argv.slice(2), r.on("uncaughtException", (function (e) {
+                                    if (!(e instanceof kt)) throw e
+                                })), r.on("unhandledRejection", ce), x = function (e) {
+                                    r.exit(e)
+                                }, d.inspect = function () {
+                                    return "[Emscripten Module object]"
+                                };
+                                try {
+                                    T = H()
+                                } catch (e) {
+                                    throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'), e
+                                }
+                                s.Worker = T.Worker
+                            } else v ? ("undefined" != typeof read && (k = function (e) {
+                                return read(e)
+                            }), I = function (e) {
+                                var t;
+                                return "function" == typeof readbuffer ? new Uint8Array(readbuffer(e)) : (V("object" == typeof (t = read(e, "binary"))), t)
+                            }, "undefined" != typeof scriptArgs && scriptArgs, "function" == typeof quit && (x = function (e) {
+                                quit(e)
+                            }), "undefined" != typeof print && ("undefined" == typeof console && (console = {}), console.log = print, console.warn = console.error = "undefined" != typeof printErr ? printErr : print)) : (g || y) && (y ? N = self.location.href : "undefined" != typeof document && document.currentScript && (N = document.currentScript.src), void 0 !== o && o && (N = o), N = 0 !== N.indexOf("blob:") ? N.substr(0, N.lastIndexOf("/") + 1) : "", b ? (k = function (e, t) {
+                                return A || (A = j()), E || (E = G()), e = E.normalize(e), A.readFileSync(e, t ? null : "utf8")
+                            }, I = function (e) {
+                                var t = k(e, !0);
+                                return t.buffer || (t = new Uint8Array(t)), V(t.buffer), t
+                            }) : (k = function (e) {
+                                var t = new XMLHttpRequest;
+                                return t.open("GET", e, !1), t.send(null), t.responseText
+                            }, y && (I = function (e) {
+                                var t = new XMLHttpRequest;
+                                return t.open("GET", e, !1), t.responseType = "arraybuffer", t.send(null), new Uint8Array(t.response)
+                            }), S = function (e, t, n) {
+                                var r = new XMLHttpRequest;
+                                r.open("GET", e, !0), r.responseType = "arraybuffer", r.onload = function () {
+                                    200 == r.status || 0 == r.status && r.response ? t(r.response) : n()
+                                }, r.onerror = n, r.send(null)
+                            }));
+                            b && "undefined" == typeof performance && (s.performance = q().performance);
+                            var R, _ = d.print || console.log.bind(console),
+                                F = d.printErr || console.warn.bind(console);
+                            for (f in m) m.hasOwnProperty(f) && (d[f] = m[f]);
+                            m = null, d.quit && (x = d.quit), d.wasmBinary && (R = d.wasmBinary);
+                            var M, D, O = d.noExitRuntime || !0;
+                            "object" != typeof WebAssembly && ce("no native wasm support detected");
+                            var L, P, $, B, z, U, W = !1;
+
+                            function V(e, t) {
+                                e || ce("Assertion failed: " + t)
+                            }
+
+                            function Y(e) {
+                                var t = d["_" + e];
+                                return V(t, "Cannot call unknown function " + e + ", make sure it is exported"), t
+                            }
+
+                            function X(e, n, r, a, s) {
+                                var i = {
+                                        string: function (e) {
+                                            var t = 0;
+                                            if (null != e && 0 !== e) {
+                                                var n = 1 + (e.length << 2);
+                                                J(e, t = yt(n), n)
+                                            }
+                                            return t
+                                        },
+                                        array: function (e) {
+                                            var n, r, a = yt(e.length);
+                                            return n = e, r = a, t().set(n, r), a
+                                        }
+                                    },
+                                    o = Y(e),
+                                    u = [],
+                                    l = 0;
+                                if (a)
+                                    for (var c = 0; c < a.length; c++) {
+                                        var h = i[r[c]];
+                                        h ? (0 === l && (l = mt()), u[c] = h(a[c])) : u[c] = a[c]
+                                    }
+                                var p, d = o.apply(null, u);
+                                return p = d, d = "string" === n ? Z(p) : "boolean" === n ? Boolean(p) : p, 0 !== l && gt(l), d
+                            }
+
+                            function Q(e, t, n) {
+                                for (var r = t + n, a = ""; !(t >= r);) {
+                                    var s = e[t++];
+                                    if (!s) return a;
+                                    if (128 & s) {
+                                        var i = 63 & e[t++];
+                                        if (192 != (224 & s)) {
+                                            var o = 63 & e[t++];
+                                            if ((s = 224 == (240 & s) ? (15 & s) << 12 | i << 6 | o : (7 & s) << 18 | i << 12 | o << 6 | 63 & e[t++]) < 65536) a += String.fromCharCode(s);
+                                            else {
+                                                var u = s - 65536;
+                                                a += String.fromCharCode(55296 | u >> 10, 56320 | 1023 & u)
+                                            }
+                                        } else a += String.fromCharCode((31 & s) << 6 | i)
+                                    } else a += String.fromCharCode(s)
+                                }
+                                return a
+                            }
+
+                            function Z(e, t) {
+                                return e ? Q(n(), e, t) : ""
+                            }
+
+                            function J(e, t, r) {
+                                return function (e, t, n, r) {
+                                    if (!(r > 0)) return 0;
+                                    for (var a = n, s = n + r - 1, i = 0; i < e.length; ++i) {
+                                        var o = e.charCodeAt(i);
+                                        if (o >= 55296 && o <= 57343 && (o = 65536 + ((1023 & o) << 10) | 1023 & e.charCodeAt(++i)), o <= 127) {
+                                            if (n >= s) break;
+                                            t[n++] = o
+                                        } else if (o <= 2047) {
+                                            if (n + 1 >= s) break;
+                                            t[n++] = 192 | o >> 6, t[n++] = 128 | 63 & o
+                                        } else if (o <= 65535) {
+                                            if (n + 2 >= s) break;
+                                            t[n++] = 224 | o >> 12, t[n++] = 128 | o >> 6 & 63, t[n++] = 128 | 63 & o
+                                        } else {
+                                            if (n + 3 >= s) break;
+                                            t[n++] = 240 | o >> 18, t[n++] = 128 | o >> 12 & 63, t[n++] = 128 | o >> 6 & 63, t[n++] = 128 | 63 & o
+                                        }
+                                    }
+                                    return t[n] = 0, n - a
+                                }(e, n(), t, r)
+                            }
+
+                            function ee(e) {
+                                L = e, d.HEAP8 = P = new Int8Array(e), d.HEAP16 = new Int16Array(e), d.HEAP32 = B = new Int32Array(e), d.HEAPU8 = $ = new Uint8Array(e), d.HEAPU16 = new Uint16Array(e), d.HEAPU32 = z = new Uint32Array(e), d.HEAPF32 = new Float32Array(e), d.HEAPF64 = U = new Float64Array(e)
+                            }
+                            var te, ne = d.INITIAL_MEMORY || 16777216;
+                            if (w) M = d.wasmMemory, L = d.buffer;
+                            else if (d.wasmMemory) M = d.wasmMemory;
+                            else if (!((M = new WebAssembly.Memory({
+                                    initial: ne / 65536,
+                                    maximum: 32768,
+                                    shared: !0
+                                })).buffer instanceof SharedArrayBuffer)) throw F("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"), b && console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"), Error("bad memory");
+                            M && (L = M.buffer), ne = L.byteLength, ee(L);
+                            var re = [],
+                                ae = [],
+                                se = [],
+                                ie = [];
+
+                            function oe() {
+                                w || ye(ae)
+                            }
+                            w || ae.push({
+                                func: function () {
+                                    Qe()
+                                }
+                            });
+                            var ue = 0,
+                                le = null;
+
+                            function ce(e) {
+                                d.onAbort && d.onAbort(e), w && console.error("Pthread aborting at " + (new Error).stack), F(e += ""), W = !0, e = "abort(" + e + "). Build with -s ASSERTIONS=1 for more info.";
+                                var t = new WebAssembly.RuntimeError(e);
+                                throw h(t), t
+                            }
+
+                            function he(e, t) {
+                                return String.prototype.startsWith ? e.startsWith(t) : 0 === e.indexOf(t)
+                            }
+
+                            function pe(e) {
+                                return he(e, "data:application/octet-stream;base64,")
+                            }
+
+                            function de(e) {
+                                return he(e, "file://")
+                            }
+                            d.preloadedImages = {}, d.preloadedAudios = {};
+                            var fe = "tfjs-backend-wasm-threaded-simd.wasm";
+
+                            function me(e) {
+                                try {
+                                    if (e == fe && R) return new Uint8Array(R);
+                                    if (I) return I(e);
+                                    throw "both async and sync fetching of the wasm failed"
+                                } catch (e) {
+                                    ce(e)
+                                }
+                            }
+                            pe(fe) || (fe = C(fe));
+                            var ge = {
+                                10664: function () {
+                                    throw "Canceled!"
+                                },
+                                10682: function (e, t) {
+                                    setTimeout((function () {
+                                        ot(e, t)
+                                    }), 0)
+                                }
+                            };
+
+                            function ye(e) {
+                                for (; e.length > 0;) {
+                                    var t = e.shift();
+                                    if ("function" != typeof t) {
+                                        var n = t.func;
+                                        "number" == typeof n ? void 0 === t.arg ? te.get(n)() : te.get(n)(t.arg) : n(void 0 === t.arg ? null : t.arg)
+                                    } else t(d)
+                                }
+                            }
+                            var be = 71,
+                                ve = 10,
+                                xe = 28,
+                                we = 16;
+
+                            function ke(e, n) {
+                                if (e <= 0 || e > t().length || !0 & e || n < 0) return -28;
+                                if (0 == n) return 0;
+                                n >= 2147483647 && (n = 1 / 0);
+                                var r = Atomics.load(i(), wt >> 2),
+                                    a = 0;
+                                if (r == e && Atomics.compareExchange(i(), wt >> 2, r, 0) == r && (a = 1, --n <= 0)) return 1;
+                                var s = Atomics.notify(i(), e >> 2, n);
+                                if (s >= 0) return s + a;
+                                throw "Atomics.notify returned an unexpected value " + s
+                            }
+
+                            function Se(e) {
+                                if (w) throw "Internal Error! cleanupThread() can only ever be called from main application thread!";
+                                if (!e) throw "Internal Error! Null pthread_ptr in cleanupThread!";
+                                var t = Ae.pthreads[e];
+                                if (t) {
+                                    i()[e + 12 >> 2] = 0;
+                                    var n = t.worker;
+                                    Ae.returnWorkerToPool(n)
+                                }
+                            }
+                            d._emscripten_futex_wake = ke;
+                            var Ie, Ae = {
+                                unusedWorkers: [],
+                                runningWorkers: [],
+                                initMainThreadBlock: function () {
+                                    for (var e = 0; e < 8; ++e) Ae.allocateUnusedWorker()
+                                },
+                                initRuntime: function () {
+                                    for (var e = Ze(228), t = 0; t < 57; ++t) u()[e / 4 + t] = 0;
+                                    i()[e + 12 >> 2] = e;
+                                    var n = e + 152;
+                                    i()[n >> 2] = n;
+                                    var r = Ze(512);
+                                    for (t = 0; t < 128; ++t) u()[r / 4 + t] = 0;
+                                    Atomics.store(u(), e + 100 >> 2, r), Atomics.store(u(), e + 40 >> 2, e), ft(e, !y, 1), st(e)
+                                },
+                                initWorker: function () {},
+                                pthreads: {},
+                                threadExitHandlers: [],
+                                setThreadStatus: function () {},
+                                runExitHandlers: function () {
+                                    for (; Ae.threadExitHandlers.length > 0;) Ae.threadExitHandlers.pop()();
+                                    w && nt() && rt()
+                                },
+                                runExitHandlersAndDeinitThread: function (e, t) {
+                                    Atomics.store(u(), e + 56 >> 2, 1), Atomics.store(u(), e + 60 >> 2, 0), Ae.runExitHandlers(), Atomics.store(u(), e + 4 >> 2, t), Atomics.store(u(), e + 0 >> 2, 1), ke(e + 0, 2147483647), ft(0, 0, 0)
+                                },
+                                threadExit: function (e) {
+                                    var t = nt();
+                                    t && (Ae.runExitHandlersAndDeinitThread(t, e), w && postMessage({
+                                        cmd: "exit"
+                                    }))
+                                },
+                                threadCancel: function () {
+                                    Ae.runExitHandlersAndDeinitThread(nt(), -1), postMessage({
+                                        cmd: "cancelDone"
+                                    })
+                                },
+                                terminateAllThreads: function () {
+                                    for (var e in Ae.pthreads)(r = Ae.pthreads[e]) && r.worker && Ae.returnWorkerToPool(r.worker);
+                                    Ae.pthreads = {};
+                                    for (var t = 0; t < Ae.unusedWorkers.length; ++t)(n = Ae.unusedWorkers[t]).terminate();
+                                    for (Ae.unusedWorkers = [], t = 0; t < Ae.runningWorkers.length; ++t) {
+                                        var n, r = (n = Ae.runningWorkers[t]).pthread;
+                                        Ae.freeThreadData(r), n.terminate()
+                                    }
+                                    Ae.runningWorkers = []
+                                },
+                                freeThreadData: function (e) {
+                                    if (e) {
+                                        if (e.threadInfoStruct) {
+                                            var t = i()[e.threadInfoStruct + 100 >> 2];
+                                            i()[e.threadInfoStruct + 100 >> 2] = 0, Je(t), Je(e.threadInfoStruct)
+                                        }
+                                        e.threadInfoStruct = 0, e.allocatedOwnStack && e.stackBase && Je(e.stackBase), e.stackBase = 0, e.worker && (e.worker.pthread = null)
+                                    }
+                                },
+                                returnWorkerToPool: function (e) {
+                                    Ae.runWithoutMainThreadQueuedCalls((function () {
+                                        delete Ae.pthreads[e.pthread.threadInfoStruct], Ae.unusedWorkers.push(e), Ae.runningWorkers.splice(Ae.runningWorkers.indexOf(e), 1), Ae.freeThreadData(e.pthread), e.pthread = void 0
+                                    }))
+                                },
+                                runWithoutMainThreadQueuedCalls: function (e) {
+                                    i()[xt >> 2] = 0;
+                                    try {
+                                        e()
+                                    } finally {
+                                        i()[xt >> 2] = 1
+                                    }
+                                },
+                                receiveObjectTransfer: function (e) {},
+                                loadWasmModuleToWorker: function (e, t) {
+                                    e.onmessage = function (n) {
+                                        var r = n.data,
+                                            a = r.cmd;
+                                        if (e.pthread && (Ae.currentProxiedOperationCallerThread = e.pthread.threadInfoStruct), r.targetThread && r.targetThread != nt()) {
+                                            var s = Ae.pthreads[r.targetThread];
+                                            return s ? s.worker.postMessage(n.data, r.transferList) : console.error('Internal error! Worker sent a message "' + a + '" to target pthread ' + r.targetThread + ", but that thread no longer exists!"), void(Ae.currentProxiedOperationCallerThread = void 0)
+                                        }
+                                        if ("processQueuedMainThreadWork" === a) at();
+                                        else if ("spawnThread" === a) Ge(n.data);
+                                        else if ("cleanupThread" === a) Se(r.thread);
+                                        else if ("killThread" === a) ! function (e) {
+                                            if (w) throw "Internal Error! killThread() can only ever be called from main application thread!";
+                                            if (!e) throw "Internal Error! Null pthread_ptr in killThread!";
+                                            i()[e + 12 >> 2] = 0;
+                                            var t = Ae.pthreads[e];
+                                            t.worker.terminate(), Ae.freeThreadData(t), Ae.runningWorkers.splice(Ae.runningWorkers.indexOf(t.worker), 1), t.worker.pthread = void 0
+                                        }(r.thread);
+                                        else if ("cancelThread" === a) ! function (e) {
+                                            if (w) throw "Internal Error! cancelThread() can only ever be called from main application thread!";
+                                            if (!e) throw "Internal Error! Null pthread_ptr in cancelThread!";
+                                            Ae.pthreads[e].worker.postMessage({
+                                                cmd: "cancel"
+                                            })
+                                        }(r.thread);
+                                        else if ("loaded" === a) e.loaded = !0, t && t(e), e.runPthread && (e.runPthread(), delete e.runPthread);
+                                        else if ("print" === a) _("Thread " + r.threadId + ": " + r.text);
+                                        else if ("printErr" === a) F("Thread " + r.threadId + ": " + r.text);
+                                        else if ("alert" === a) alert("Thread " + r.threadId + ": " + r.text);
+                                        else if ("exit" === a) e.pthread && Atomics.load(u(), e.pthread.threadInfoStruct + 64 >> 2) && Ae.returnWorkerToPool(e);
+                                        else if ("exitProcess" === a) try {
+                                            ! function (e, t) {
+                                                if (!t || !O || 0 !== e) {
+                                                    if (!t && w) throw postMessage({
+                                                        cmd: "exitProcess",
+                                                        returnCode: e
+                                                    }), new kt(e);
+                                                    O || (Ae.terminateAllThreads(), d.onExit && d.onExit(e), W = !0), x(e, new kt(e))
+                                                }
+                                            }(r.returnCode)
+                                        } catch (e) {
+                                            if (e instanceof kt) return;
+                                            throw e
+                                        } else "cancelDone" === a ? Ae.returnWorkerToPool(e) : "objectTransfer" === a ? Ae.receiveObjectTransfer(n.data) : "setimmediate" === n.data.target ? e.postMessage(n.data) : F("worker sent an unknown command " + a);
+                                        Ae.currentProxiedOperationCallerThread = void 0
+                                    }, e.onerror = function (e) {
+                                        F("pthread sent an error! " + e.filename + ":" + e.lineno + ": " + e.message)
+                                    }, b && (e.on("message", (function (t) {
+                                        e.onmessage({
+                                            data: t
+                                        })
+                                    })), e.on("error", (function (t) {
+                                        e.onerror(t)
+                                    })), e.on("exit", (function (e) {}))), e.postMessage({
+                                        cmd: "load",
+                                        urlOrBlob: d.mainScriptUrlOrBlob || o,
+                                        wasmMemory: M,
+                                        wasmModule: D
+                                    })
+                                },
+                                allocateUnusedWorker: function () {
+                                    var e = C("tfjs-backend-wasm-threaded-simd.worker.js");
+                                    Ae.unusedWorkers.push(new Worker(e))
+                                },
+                                getNewWorker: function () {
+                                    return 0 == Ae.unusedWorkers.length && (Ae.allocateUnusedWorker(), Ae.loadWasmModuleToWorker(Ae.unusedWorkers[0])), Ae.unusedWorkers.length > 0 ? Ae.unusedWorkers.pop() : null
+                                },
+                                busySpinWait: function (e) {
+                                    for (var t = performance.now() + e; performance.now() < t;);
+                                }
+                            };
+
+                            function Ee(e, n, r) {
+                                if (e <= 0 || e > t().length || !0 & e) return -28;
+                                if (g) {
+                                    if (Atomics.load(i(), e >> 2) != n) return -6;
+                                    var a = performance.now(),
+                                        s = a + r;
+                                    for (Atomics.exchange(i(), wt >> 2, e);;) {
+                                        if ((a = performance.now()) > s) return Atomics.exchange(i(), wt >> 2, 0), -73;
+                                        if (0 == Atomics.exchange(i(), wt >> 2, 0)) break;
+                                        if (at(), Atomics.load(i(), e >> 2) != n) return -6;
+                                        Atomics.exchange(i(), wt >> 2, e)
+                                    }
+                                    return 0
+                                }
+                                var o = Atomics.wait(i(), e >> 2, n, r);
+                                if ("timed-out" === o) return -73;
+                                if ("not-equal" === o) return -6;
+                                if ("ok" === o) return 0;
+                                throw "Atomics.wait returned an unexpected value " + o
+                            }
+
+                            function Ne(e, t) {
+                                for (var n = arguments.length - 2, r = mt(), a = n, s = yt(8 * a), i = s >> 3, o = 0; o < n; o++) {
+                                    var u = arguments[2 + o];
+                                    l()[i + o] = u
+                                }
+                                var c = lt(e, a, s, t);
+                                return gt(r), c
+                            }
+                            d.establishStackSpace = function (e, t) {
+                                bt(e, t), gt(e)
+                            }, d.getNoExitRuntime = function () {
+                                return O
+                            }, d.invokeEntryPoint = function (e, t) {
+                                return te.get(e)(t)
+                            }, Ie = b ? function () {
+                                var e = r.hrtime();
+                                return 1e3 * e[0] + e[1] / 1e6
+                            } : w ? function () {
+                                return performance.now() - d.__performance_now_clock_drift
+                            } : "undefined" != typeof dateNow ? dateNow : function () {
+                                return performance.now()
+                            };
+                            var Ce = [],
+                                Te = [];
+
+                            function Re(e) {
+                                try {
+                                    return M.grow(e - L.byteLength + 65535 >>> 16), ee(M.buffer), 1
+                                } catch (e) {}
+                            }
+                            var _e = {
+                                inEventHandler: 0,
+                                removeAllEventListeners: function () {
+                                    for (var e = _e.eventHandlers.length - 1; e >= 0; --e) _e._removeHandler(e);
+                                    _e.eventHandlers = [], _e.deferredCalls = []
+                                },
+                                registerRemoveEventListeners: function () {
+                                    _e.removeEventListenersRegistered || (_e.removeEventListenersRegistered = !0)
+                                },
+                                deferredCalls: [],
+                                deferCall: function (e, t, n) {
+                                    function r(e, t) {
+                                        if (e.length != t.length) return !1;
+                                        for (var n in e)
+                                            if (e[n] != t[n]) return !1;
+                                        return !0
+                                    }
+                                    for (var a in _e.deferredCalls) {
+                                        var s = _e.deferredCalls[a];
+                                        if (s.targetFunction == e && r(s.argsList, n)) return
+                                    }
+                                    _e.deferredCalls.push({
+                                        targetFunction: e,
+                                        precedence: t,
+                                        argsList: n
+                                    }), _e.deferredCalls.sort((function (e, t) {
+                                        return e.precedence < t.precedence
+                                    }))
+                                },
+                                removeDeferredCalls: function (e) {
+                                    for (var t = 0; t < _e.deferredCalls.length; ++t) _e.deferredCalls[t].targetFunction == e && (_e.deferredCalls.splice(t, 1), --t)
+                                },
+                                canPerformEventHandlerRequests: function () {
+                                    return _e.inEventHandler && _e.currentEventHandler.allowsDeferredCalls
+                                },
+                                runDeferredCalls: function () {
+                                    if (_e.canPerformEventHandlerRequests())
+                                        for (var e = 0; e < _e.deferredCalls.length; ++e) {
+                                            var t = _e.deferredCalls[e];
+                                            _e.deferredCalls.splice(e, 1), --e, t.targetFunction.apply(null, t.argsList)
+                                        }
+                                },
+                                eventHandlers: [],
+                                removeAllHandlersOnTarget: function (e, t) {
+                                    for (var n = 0; n < _e.eventHandlers.length; ++n) _e.eventHandlers[n].target != e || t && t != _e.eventHandlers[n].eventTypeString || _e._removeHandler(n--)
+                                },
+                                _removeHandler: function (e) {
+                                    var t = _e.eventHandlers[e];
+                                    t.target.removeEventListener(t.eventTypeString, t.eventListenerFunc, t.useCapture), _e.eventHandlers.splice(e, 1)
+                                },
+                                registerOrRemoveHandler: function (e) {
+                                    var t = function (t) {
+                                        ++_e.inEventHandler, _e.currentEventHandler = e, _e.runDeferredCalls(), e.handlerFunc(t), _e.runDeferredCalls(), --_e.inEventHandler
+                                    };
+                                    if (e.callbackfunc) e.eventListenerFunc = t, e.target.addEventListener(e.eventTypeString, t, e.useCapture), _e.eventHandlers.push(e), _e.registerRemoveEventListeners();
+                                    else
+                                        for (var n = 0; n < _e.eventHandlers.length; ++n) _e.eventHandlers[n].target == e.target && _e.eventHandlers[n].eventTypeString == e.eventTypeString && _e._removeHandler(n--)
+                                },
+                                queueEventHandlerOnThread_iiii: function (e, t, n, r, a) {
+                                    var s = mt(),
+                                        o = yt(12);
+                                    i()[o >> 2] = n, i()[o + 4 >> 2] = r, i()[o + 8 >> 2] = a, ct(0, e, 637534208, t, r, o), gt(s)
+                                },
+                                getTargetThreadForEventCallback: function (e) {
+                                    switch (e) {
+                                        case 1:
+                                            return 0;
+                                        case 2:
+                                            return Ae.currentProxiedOperationCallerThread;
+                                        default:
+                                            return e
+                                    }
+                                },
+                                getNodeNameForTarget: function (e) {
+                                    return e ? e == window ? "#window" : e == screen ? "#screen" : e && e.nodeName ? e.nodeName : "" : ""
+                                },
+                                fullscreenEnabled: function () {
+                                    return document.fullscreenEnabled || document.webkitFullscreenEnabled
+                                }
+                            };
+
+                            function Fe(e, t, n, r) {
+                                var a, s, o, u = mt(),
+                                    l = yt(12),
+                                    c = 0;
+                                t && (s = function (e) {
+                                    for (var t = 0, n = 0; n < e.length; ++n) {
+                                        var r = e.charCodeAt(n);
+                                        r >= 55296 && r <= 57343 && (r = 65536 + ((1023 & r) << 10) | 1023 & e.charCodeAt(++n)), r <= 127 ? ++t : t += r <= 2047 ? 2 : r <= 65535 ? 3 : 4
+                                    }
+                                    return t
+                                }(a = t) + 1, o = Ze(s), J(a, o, s), c = o), i()[l >> 2] = c, i()[l + 4 >> 2] = n, i()[l + 8 >> 2] = r, ct(0, e, 657457152, 0, c, l), gt(u)
+                            }
+                            var Me = [0, "undefined" != typeof document ? document : 0, "undefined" != typeof window ? window : 0];
+
+                            function De(e) {
+                                var t;
+                                return e = (t = e) > 2 ? Z(t) : t, Me[e] || ("undefined" != typeof document ? document.querySelector(e) : void 0)
+                            }
+
+                            function Oe(e) {
+                                return De(e)
+                            }
+
+                            function Le(e, t, n) {
+                                var r = Oe(e);
+                                if (!r) return -4;
+                                if (r.canvasSharedPtr && (i()[r.canvasSharedPtr >> 2] = t, i()[r.canvasSharedPtr + 4 >> 2] = n), !r.offscreenCanvas && r.controlTransferredOffscreen) return r.canvasSharedPtr ? (function (e, t, n, r) {
+                                    Fe(e, t = t ? Z(t) : "", n, r)
+                                }(i()[r.canvasSharedPtr + 8 >> 2], e, t, n), 1) : -4;
+                                r.offscreenCanvas && (r = r.offscreenCanvas);
+                                var a = !1;
+                                if (r.GLctxObject && r.GLctxObject.GLctx) {
+                                    var s = r.GLctxObject.GLctx.getParameter(2978);
+                                    a = 0 === s[0] && 0 === s[1] && s[2] === r.width && s[3] === r.height
+                                }
+                                return r.width = t, r.height = n, a && r.GLctxObject.GLctx.viewport(0, 0, t, n), 0
+                            }
+
+                            function Pe(e, t, n) {
+                                return w ? Ne(2, 1, e, t, n) : Le(e, t, n)
+                            }
+                            var $e, Be = {
+                                    counter: 1,
+                                    buffers: [],
+                                    programs: [],
+                                    framebuffers: [],
+                                    renderbuffers: [],
+                                    textures: [],
+                                    uniforms: [],
+                                    shaders: [],
+                                    vaos: [],
+                                    contexts: {},
+                                    offscreenCanvases: {},
+                                    timerQueriesEXT: [],
+                                    programInfos: {},
+                                    stringCache: {},
+                                    unpackAlignment: 4,
+                                    recordError: function (e) {
+                                        Be.lastError || (Be.lastError = e)
+                                    },
+                                    getNewId: function (e) {
+                                        for (var t = Be.counter++, n = e.length; n < t; n++) e[n] = null;
+                                        return t
+                                    },
+                                    getSource: function (e, t, n, r) {
+                                        for (var a = "", s = 0; s < t; ++s) {
+                                            var o = r ? i()[r + 4 * s >> 2] : -1;
+                                            a += Z(i()[n + 4 * s >> 2], o < 0 ? void 0 : o)
+                                        }
+                                        return a
+                                    },
+                                    createContext: function (e, t) {
+                                        var n = e.getContext("webgl", t);
+                                        return n ? Be.registerContext(n, t) : 0
+                                    },
+                                    registerContext: function (e, t) {
+                                        var n = Ze(8);
+                                        i()[n + 4 >> 2] = nt();
+                                        var r = {
+                                            handle: n,
+                                            attributes: t,
+                                            version: t.majorVersion,
+                                            GLctx: e
+                                        };
+                                        return e.canvas && (e.canvas.GLctxObject = r), Be.contexts[n] = r, (void 0 === t.enableExtensionsByDefault || t.enableExtensionsByDefault) && Be.initExtensions(r), n
+                                    },
+                                    makeContextCurrent: function (e) {
+                                        return Be.currentContext = Be.contexts[e], d.ctx = $e = Be.currentContext && Be.currentContext.GLctx, !(e && !$e)
+                                    },
+                                    getContext: function (e) {
+                                        return Be.contexts[e]
+                                    },
+                                    deleteContext: function (e) {
+                                        Be.currentContext === Be.contexts[e] && (Be.currentContext = null), "object" == typeof _e && _e.removeAllHandlersOnTarget(Be.contexts[e].GLctx.canvas), Be.contexts[e] && Be.contexts[e].GLctx.canvas && (Be.contexts[e].GLctx.canvas.GLctxObject = void 0), Je(Be.contexts[e].handle), Be.contexts[e] = null
+                                    },
+                                    initExtensions: function (e) {
+                                        if (e || (e = Be.currentContext), !e.initExtensionsDone) {
+                                            e.initExtensionsDone = !0;
+                                            var t, n = e.GLctx;
+                                            ! function (e) {
+                                                var t = e.getExtension("ANGLE_instanced_arrays");
+                                                t && (e.vertexAttribDivisor = function (e, n) {
+                                                    t.vertexAttribDivisorANGLE(e, n)
+                                                }, e.drawArraysInstanced = function (e, n, r, a) {
+                                                    t.drawArraysInstancedANGLE(e, n, r, a)
+                                                }, e.drawElementsInstanced = function (e, n, r, a, s) {
+                                                    t.drawElementsInstancedANGLE(e, n, r, a, s)
+                                                })
+                                            }(n),
+                                            function (e) {
+                                                var t = e.getExtension("OES_vertex_array_object");
+                                                t && (e.createVertexArray = function () {
+                                                    return t.createVertexArrayOES()
+                                                }, e.deleteVertexArray = function (e) {
+                                                    t.deleteVertexArrayOES(e)
+                                                }, e.bindVertexArray = function (e) {
+                                                    t.bindVertexArrayOES(e)
+                                                }, e.isVertexArray = function (e) {
+                                                    return t.isVertexArrayOES(e)
+                                                })
+                                            }(n),
+                                            function (e) {
+                                                var t = e.getExtension("WEBGL_draw_buffers");
+                                                t && (e.drawBuffers = function (e, n) {
+                                                    t.drawBuffersWEBGL(e, n)
+                                                })
+                                            }(n), n.disjointTimerQueryExt = n.getExtension("EXT_disjoint_timer_query"), (t = n).multiDrawWebgl = t.getExtension("WEBGL_multi_draw"), (n.getSupportedExtensions() || []).forEach((function (e) {
+                                                e.indexOf("lose_context") < 0 && e.indexOf("debug") < 0 && n.getExtension(e)
+                                            }))
+                                        }
+                                    },
+                                    populateUniformTable: function (e) {
+                                        for (var t = Be.programs[e], n = Be.programInfos[e] = {
+                                                uniforms: {},
+                                                maxUniformLength: 0,
+                                                maxAttributeLength: -1,
+                                                maxUniformBlockNameLength: -1
+                                            }, r = n.uniforms, a = $e.getProgramParameter(t, 35718), s = 0; s < a; ++s) {
+                                            var i = $e.getActiveUniform(t, s),
+                                                o = i.name;
+                                            n.maxUniformLength = Math.max(n.maxUniformLength, o.length + 1), "]" == o.slice(-1) && (o = o.slice(0, o.lastIndexOf("[")));
+                                            var u = $e.getUniformLocation(t, o);
+                                            if (u) {
+                                                var l = Be.getNewId(Be.uniforms);
+                                                r[o] = [i.size, l], Be.uniforms[l] = u;
+                                                for (var c = 1; c < i.size; ++c) {
+                                                    var h = o + "[" + c + "]";
+                                                    u = $e.getUniformLocation(t, h), l = Be.getNewId(Be.uniforms), Be.uniforms[l] = u
+                                                }
+                                            }
+                                        }
+                                    }
+                                },
+                                ze = ["default", "low-power", "high-performance"],
+                                Ue = {
+                                    mappings: {},
+                                    buffers: [null, [],
+                                        []
+                                    ],
+                                    printChar: function (e, t) {
+                                        var n = Ue.buffers[e];
+                                        0 === t || 10 === t ? ((1 === e ? _ : F)(Q(n, 0)), n.length = 0) : n.push(t)
+                                    },
+                                    varargs: void 0,
+                                    get: function () {
+                                        return Ue.varargs += 4, i()[Ue.varargs - 4 >> 2]
+                                    },
+                                    getStr: function (e) {
+                                        return Z(e)
+                                    },
+                                    get64: function (e, t) {
+                                        return e
+                                    }
+                                };
+
+                            function We(e) {
+                                return w ? Ne(3, 1, e) : 0
+                            }
+
+                            function Ve(e, t, n, r, a) {
+                                if (w) return Ne(4, 1, e, t, n, r, a)
+                            }
+
+                            function je(e, t, r, a) {
+                                if (w) return Ne(5, 1, e, t, r, a);
+                                for (var s = 0, o = 0; o < r; o++) {
+                                    for (var u = i()[t + 8 * o >> 2], l = i()[t + (8 * o + 4) >> 2], c = 0; c < l; c++) Ue.printChar(e, n()[u + c]);
+                                    s += l
+                                }
+                                return i()[a >> 2] = s, 0
+                            }
+
+                            function Ge(e) {
+                                if (w) throw "Internal Error! spawnThread() can only ever be called from main application thread!";
+                                var t = Ae.getNewWorker();
+                                if (void 0 !== t.pthread) throw "Internal error!";
+                                if (!e.pthread_ptr) throw "Internal error, no pthread ptr!";
+                                Ae.runningWorkers.push(t);
+                                for (var n = Ze(512), r = 0; r < 128; ++r) i()[n + 4 * r >> 2] = 0;
+                                var a = e.stackBase + e.stackSize,
+                                    s = Ae.pthreads[e.pthread_ptr] = {
+                                        worker: t,
+                                        stackBase: e.stackBase,
+                                        stackSize: e.stackSize,
+                                        allocatedOwnStack: e.allocatedOwnStack,
+                                        threadInfoStruct: e.pthread_ptr
+                                    },
+                                    o = s.threadInfoStruct >> 2;
+                                Atomics.store(u(), o + 16, e.detached), Atomics.store(u(), o + 25, n), Atomics.store(u(), o + 10, s.threadInfoStruct), Atomics.store(u(), o + 20, e.stackSize), Atomics.store(u(), o + 19, a), Atomics.store(u(), o + 26, e.stackSize), Atomics.store(u(), o + 28, a), Atomics.store(u(), o + 29, e.detached);
+                                var l = tt() + 40;
+                                Atomics.store(u(), o + 43, l), t.pthread = s;
+                                var c = {
+                                    cmd: "run",
+                                    start_routine: e.startRoutine,
+                                    arg: e.arg,
+                                    threadInfoStruct: e.pthread_ptr,
+                                    stackBase: e.stackBase,
+                                    stackSize: e.stackSize
+                                };
+                                t.runPthread = function () {
+                                    c.time = performance.now(), t.postMessage(c, e.transferList)
+                                }, t.loaded && (t.runPthread(), delete t.runPthread)
+                            }
+
+                            function He() {
+                                if (w) {
+                                    var e = nt();
+                                    if (e && !Atomics.load(u(), e + 56 >> 2) && 2 == Atomics.load(u(), e + 0 >> 2)) throw "Canceled!"
+                                }
+                            }
+
+                            function qe() {
+                                b || y || function e(t) {
+                                    e.shown || (e.shown = {}), e.shown[t] || (e.shown[t] = 1, F(t))
+                                }("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")
+                            }
+
+                            function Ke(e) {
+                                if (w) return Ne(6, 1, e);
+                                switch (e) {
+                                    case 30:
+                                        return 16384;
+                                    case 85:
+                                        return 131072;
+                                    case 132:
+                                    case 133:
+                                    case 12:
+                                    case 137:
+                                    case 138:
+                                    case 15:
+                                    case 235:
+                                    case 16:
+                                    case 17:
+                                    case 18:
+                                    case 19:
+                                    case 20:
+                                    case 149:
+                                    case 13:
+                                    case 10:
+                                    case 236:
+                                    case 153:
+                                    case 9:
+                                    case 21:
+                                    case 22:
+                                    case 159:
+                                    case 154:
+                                    case 14:
+                                    case 77:
+                                    case 78:
+                                    case 139:
+                                    case 82:
+                                    case 68:
+                                    case 67:
+                                    case 164:
+                                    case 11:
+                                    case 29:
+                                    case 47:
+                                    case 48:
+                                    case 95:
+                                    case 52:
+                                    case 51:
+                                    case 46:
+                                        return 200809;
+                                    case 27:
+                                    case 246:
+                                    case 127:
+                                    case 128:
+                                    case 23:
+                                    case 24:
+                                    case 160:
+                                    case 161:
+                                    case 181:
+                                    case 182:
+                                    case 242:
+                                    case 183:
+                                    case 184:
+                                    case 243:
+                                    case 244:
+                                    case 245:
+                                    case 165:
+                                    case 178:
+                                    case 179:
+                                    case 49:
+                                    case 50:
+                                    case 168:
+                                    case 169:
+                                    case 175:
+                                    case 170:
+                                    case 171:
+                                    case 172:
+                                    case 97:
+                                    case 76:
+                                    case 32:
+                                    case 173:
+                                    case 35:
+                                    case 80:
+                                    case 81:
+                                    case 79:
+                                        return -1;
+                                    case 176:
+                                    case 177:
+                                    case 7:
+                                    case 155:
+                                    case 8:
+                                    case 157:
+                                    case 125:
+                                    case 126:
+                                    case 92:
+                                    case 93:
+                                    case 129:
+                                    case 130:
+                                    case 131:
+                                    case 94:
+                                    case 91:
+                                        return 1;
+                                    case 74:
+                                    case 60:
+                                    case 69:
+                                    case 70:
+                                    case 4:
+                                        return 1024;
+                                    case 31:
+                                    case 42:
+                                    case 72:
+                                        return 32;
+                                    case 87:
+                                    case 26:
+                                    case 33:
+                                        return 2147483647;
+                                    case 34:
+                                    case 1:
+                                        return 47839;
+                                    case 38:
+                                    case 36:
+                                        return 99;
+                                    case 43:
+                                    case 37:
+                                        return 2048;
+                                    case 0:
+                                        return 2097152;
+                                    case 3:
+                                        return 65536;
+                                    case 28:
+                                        return 32768;
+                                    case 44:
+                                        return 32767;
+                                    case 75:
+                                        return 16384;
+                                    case 39:
+                                        return 1e3;
+                                    case 89:
+                                        return 700;
+                                    case 71:
+                                        return 256;
+                                    case 40:
+                                        return 255;
+                                    case 2:
+                                        return 100;
+                                    case 180:
+                                        return 64;
+                                    case 25:
+                                        return 20;
+                                    case 5:
+                                        return 16;
+                                    case 6:
+                                        return 6;
+                                    case 73:
+                                        return 4;
+                                    case 84:
+                                        return "object" == typeof navigator && navigator.hardwareConcurrency || 1
+                                }
+                                var t;
+                                return t = 28, i()[et() >> 2] = t, -1
+                            }
+                            w || Ae.initMainThreadBlock();
+                            var Ye = [null, function (e, t) {
+                                    if (w) return Ne(1, 1, e, t)
+                                }, Pe, We, Ve, je, Ke],
+                                Xe = {
+                                    e: function (e, t, n, r) {
+                                        ce("Assertion failed: " + Z(e) + ", at: " + [t ? Z(t) : "unknown filename", n, r ? Z(r) : "unknown function"])
+                                    },
+                                    r: function (e, t) {
+                                        _main(e, t)
+                                    },
+                                    x: function (e, t) {
+                                        if (e == t) postMessage({
+                                            cmd: "processQueuedMainThreadWork"
+                                        });
+                                        else if (w) postMessage({
+                                            targetThread: e,
+                                            cmd: "processThreadQueue"
+                                        });
+                                        else {
+                                            var n = Ae.pthreads[e],
+                                                r = n && n.worker;
+                                            if (!r) return;
+                                            r.postMessage({
+                                                cmd: "processThreadQueue"
+                                            })
+                                        }
+                                        return 1
+                                    },
+                                    b: function () {
+                                        ce()
+                                    },
+                                    y: function (e, t, r) {
+                                        var a = function (e, t) {
+                                            var r;
+                                            for (Te.length = 0, t >>= 2; r = n()[e++];) {
+                                                var a = r < 105;
+                                                a && 1 & t && t++, Te.push(a ? l()[t++ >> 1] : i()[t]), ++t
+                                            }
+                                            return Te
+                                        }(t, r);
+                                        return ge[e].apply(null, a)
+                                    },
+                                    j: function (e, t) {},
+                                    d: Ee,
+                                    c: ke,
+                                    f: Ie,
+                                    p: function (e, t, r) {
+                                        n().copyWithin(e, t, t + r)
+                                    },
+                                    A: function () {
+                                        return b ? K().cpus().length : navigator.hardwareConcurrency
+                                    },
+                                    u: function (e, t, n) {
+                                        Ce.length = t;
+                                        for (var r = n >> 3, a = 0; a < t; a++) Ce[a] = l()[r + a];
+                                        return (e < 0 ? ge[-e - 1] : Ye[e]).apply(null, Ce)
+                                    },
+                                    q: function (e) {
+                                        var t, r, a = n().length;
+                                        if (e <= a) return !1;
+                                        if (e > 2147483648) return !1;
+                                        for (var s = 1; s <= 4; s *= 2) {
+                                            var i = a * (1 + .2 / s);
+                                            if (i = Math.min(i, e + 100663296), Re(Math.min(2147483648, ((t = Math.max(e, i)) % (r = 65536) > 0 && (t += r - t % r), t)))) return !0
+                                        }
+                                        return !1
+                                    },
+                                    v: function (e, t, n) {
+                                        return Oe(e) ? Le(e, t, n) : Pe(e, t, n)
+                                    },
+                                    i: function (e) {},
+                                    s: function (e, t) {},
+                                    w: function (e, t) {
+                                        return n = e, r = t >> 2, a = i()[r + 6], s = {
+                                            alpha: !!i()[r + 0],
+                                            depth: !!i()[r + 1],
+                                            stencil: !!i()[r + 2],
+                                            antialias: !!i()[r + 3],
+                                            premultipliedAlpha: !!i()[r + 4],
+                                            preserveDrawingBuffer: !!i()[r + 5],
+                                            powerPreference: ze[a],
+                                            failIfMajorPerformanceCaveat: !!i()[r + 7],
+                                            majorVersion: i()[r + 8],
+                                            minorVersion: i()[r + 9],
+                                            enableExtensionsByDefault: i()[r + 10],
+                                            explicitSwapControl: i()[r + 11],
+                                            proxyContextToMainThread: i()[r + 12],
+                                            renderViaOffscreenBackBuffer: i()[r + 13]
+                                        }, (o = Oe(n)) ? s.explicitSwapControl ? 0 : Be.createContext(o, s) : 0;
+                                        var n, r, a, s, o
+                                    },
+                                    l: We,
+                                    n: Ve,
+                                    g: je,
+                                    o: function () {
+                                        Ae.initRuntime()
+                                    },
+                                    a: M || d.wasmMemory,
+                                    z: function (e) {
+                                        var t = Ae.threadExitHandlers.pop();
+                                        e && t()
+                                    },
+                                    k: function (e, t) {
+                                        Ae.threadExitHandlers.push((function () {
+                                            te.get(e)(t)
+                                        }))
+                                    },
+                                    h: function (e, t, n, r) {
+                                        if ("undefined" == typeof SharedArrayBuffer) return F("Current environment does not support SharedArrayBuffer, pthreads are not available!"), 6;
+                                        if (!e) return F("pthread_create called with a null thread pointer!"), 28;
+                                        var a = [];
+                                        if (w && 0 === a.length) return ut(687865856, e, t, n, r);
+                                        var s = 0,
+                                            o = 0,
+                                            l = 0;
+                                        t && -1 != t ? (s = i()[t >> 2], s += 81920, o = i()[t + 8 >> 2], l = 0 !== i()[t + 12 >> 2]) : s = 2097152;
+                                        var c = 0 == o;
+                                        c ? o = vt(16, s) : V((o -= s) > 0);
+                                        for (var h = Ze(228), p = 0; p < 57; ++p) u()[(h >> 2) + p] = 0;
+                                        i()[e >> 2] = h, i()[h + 12 >> 2] = h;
+                                        var d = h + 152;
+                                        i()[d >> 2] = d;
+                                        var f = {
+                                            stackBase: o,
+                                            stackSize: s,
+                                            allocatedOwnStack: c,
+                                            detached: l,
+                                            startRoutine: n,
+                                            pthread_ptr: h,
+                                            arg: r,
+                                            transferList: a
+                                        };
+                                        return w ? (f.cmd = "spawnThread", postMessage(f, a)) : Ge(f), 0
+                                    },
+                                    m: function (e, t) {
+                                        return function (e, t, n) {
+                                            if (!e) return F("pthread_join attempted on a null thread pointer!"), be;
+                                            if (w && nt() == e) return F("PThread " + e + " is attempting to join to itself!"), we;
+                                            if (!w && it() == e) return F("Main thread " + e + " is attempting to join to itself!"), we;
+                                            if (i()[e + 12 >> 2] !== e) return F("pthread_join attempted on thread " + e + ", which does not point to a valid thread, or does not exist anymore!"), be;
+                                            if (Atomics.load(u(), e + 64 >> 2)) return F("Attempted to join thread " + e + ", which was already detached!"), xe;
+                                            for (n && qe();;) {
+                                                var r = Atomics.load(u(), e + 0 >> 2);
+                                                if (1 == r) {
+                                                    var a = Atomics.load(u(), e + 4 >> 2);
+                                                    return t && (i()[t >> 2] = a), Atomics.store(u(), e + 64 >> 2, 1), w ? postMessage({
+                                                        cmd: "cleanupThread",
+                                                        thread: e
+                                                    }) : Se(e), 0
+                                                }
+                                                if (!n) return ve;
+                                                He(), w || at(), Ee(e + 0, r, w ? 100 : 1)
+                                            }
+                                        }(e, t, !0)
+                                    },
+                                    t: Ke
+                                };
+                            ! function () {
+                                var e = {
+                                    a: Xe
+                                };
+
+                                function t(e, t) {
+                                    var n = e.exports;
+                                    if (d.asm = n, te = d.asm.I, D = t, !w) {
+                                        var r = Ae.unusedWorkers.length;
+                                        Ae.unusedWorkers.forEach((function (e) {
+                                            Ae.loadWasmModuleToWorker(e, (function () {
+                                                --r || function (e) {
+                                                    if (ue--, d.monitorRunDependencies && d.monitorRunDependencies(ue), 0 == ue && le) {
+                                                        var t = le;
+                                                        le = null, t()
+                                                    }
+                                                }()
+                                            }))
+                                        }))
+                                    }
+                                }
+
+                                function n(e) {
+                                    t(e.instance, e.module)
+                                }
+
+                                function r(t) {
+                                    return function () {
+                                        if (!R && (g || y)) {
+                                            if ("function" == typeof fetch && !de(fe)) return fetch(fe, {
+                                                credentials: "same-origin"
+                                            }).then((function (e) {
+                                                if (!e.ok) throw "failed to load wasm binary file at '" + fe + "'";
+                                                return e.arrayBuffer()
+                                            })).catch((function () {
+                                                return me(fe)
+                                            }));
+                                            if (S) return new Promise((function (e, t) {
+                                                S(fe, (function (t) {
+                                                    e(new Uint8Array(t))
+                                                }), t)
+                                            }))
+                                        }
+                                        return Promise.resolve().then((function () {
+                                            return me(fe)
+                                        }))
+                                    }().then((function (t) {
+                                        return WebAssembly.instantiate(t, e)
+                                    })).then(t, (function (e) {
+                                        F("failed to asynchronously prepare wasm: " + e), ce(e)
+                                    }))
+                                }
+                                if (w || (V(!w, "addRunDependency cannot be used in a pthread worker"), ue++, d.monitorRunDependencies && d.monitorRunDependencies(ue)), d.instantiateWasm) try {
+                                    return d.instantiateWasm(e, t)
+                                } catch (e) {
+                                    return F("Module.instantiateWasm callback failed with error: " + e), !1
+                                }(R || "function" != typeof WebAssembly.instantiateStreaming || pe(fe) || de(fe) || "function" != typeof fetch ? r(n) : fetch(fe, {
+                                    credentials: "same-origin"
+                                }).then((function (t) {
+                                    return WebAssembly.instantiateStreaming(t, e).then(n, (function (e) {
+                                        return F("wasm streaming compile failed: " + e), F("falling back to ArrayBuffer instantiation"), r(n)
+                                    }))
+                                }))).catch(h)
+                            }();
+                            var Qe = d.___wasm_call_ctors = function () {
+                                return (Qe = d.___wasm_call_ctors = d.asm.B).apply(null, arguments)
+                            };
+                            d._init = function () {
+                                return (d._init = d.asm.C).apply(null, arguments)
+                            }, d._init_with_threads_count = function () {
+                                return (d._init_with_threads_count = d.asm.D).apply(null, arguments)
+                            }, d._get_threads_count = function () {
+                                return (d._get_threads_count = d.asm.E).apply(null, arguments)
+                            }, d._register_tensor = function () {
+                                return (d._register_tensor = d.asm.F).apply(null, arguments)
+                            }, d._dispose_data = function () {
+                                return (d._dispose_data = d.asm.G).apply(null, arguments)
+                            }, d._dispose = function () {
+                                return (d._dispose = d.asm.H).apply(null, arguments)
+                            }, d._Abs = function () {
+                                return (d._Abs = d.asm.J).apply(null, arguments)
+                            }, d._Add = function () {
+                                return (d._Add = d.asm.K).apply(null, arguments)
+                            }, d._AddN = function () {
+                                return (d._AddN = d.asm.L).apply(null, arguments)
+                            }, d._All = function () {
+                                return (d._All = d.asm.M).apply(null, arguments)
+                            }, d._Any = function () {
+                                return (d._Any = d.asm.N).apply(null, arguments)
+                            }, d._ArgMax = function () {
+                                return (d._ArgMax = d.asm.O).apply(null, arguments)
+                            }, d._AvgPool = function () {
+                                return (d._AvgPool = d.asm.P).apply(null, arguments)
+                            }, d._BatchMatMul = function () {
+                                return (d._BatchMatMul = d.asm.Q).apply(null, arguments)
+                            }, d._Ceil = function () {
+                                return (d._Ceil = d.asm.R).apply(null, arguments)
+                            }, d._ClipByValue = function () {
+                                return (d._ClipByValue = d.asm.S).apply(null, arguments)
+                            }, d._Conv2D = function () {
+                                return (d._Conv2D = d.asm.T).apply(null, arguments)
+                            }, d._Conv2DBackpropInput = function () {
+                                return (d._Conv2DBackpropInput = d.asm.U).apply(null, arguments)
+                            }, d._Cos = function () {
+                                return (d._Cos = d.asm.V).apply(null, arguments)
+                            }, d._Cosh = function () {
+                                return (d._Cosh = d.asm.W).apply(null, arguments)
+                            }, d._CropAndResize = function () {
+                                return (d._CropAndResize = d.asm.X).apply(null, arguments)
+                            }, d._Cumsum = function () {
+                                return (d._Cumsum = d.asm.Y).apply(null, arguments)
+                            }, d._DepthToSpace = function () {
+                                return (d._DepthToSpace = d.asm.Z).apply(null, arguments)
+                            }, d._DepthwiseConv2dNative = function () {
+                                return (d._DepthwiseConv2dNative = d.asm._).apply(null, arguments)
+                            }, d._Elu = function () {
+                                return (d._Elu = d.asm.$).apply(null, arguments)
+                            }, d._Equal = function () {
+                                return (d._Equal = d.asm.aa).apply(null, arguments)
+                            }, d._Exp = function () {
+                                return (d._Exp = d.asm.ba).apply(null, arguments)
+                            }, d._FlipLeftRight = function () {
+                                return (d._FlipLeftRight = d.asm.ca).apply(null, arguments)
+                            }, d._Floor = function () {
+                                return (d._Floor = d.asm.da).apply(null, arguments)
+                            }, d._FloorDiv = function () {
+                                return (d._FloorDiv = d.asm.ea).apply(null, arguments)
+                            }, d._FusedBatchNorm = function () {
+                                return (d._FusedBatchNorm = d.asm.fa).apply(null, arguments)
+                            }, d._FusedConv2D = function () {
+                                return (d._FusedConv2D = d.asm.ga).apply(null, arguments)
+                            }, d._FusedDepthwiseConv2D = function () {
+                                return (d._FusedDepthwiseConv2D = d.asm.ha).apply(null, arguments)
+                            }, d._Gather = function () {
+                                return (d._Gather = d.asm.ia).apply(null, arguments)
+                            }, d._GatherNd = function () {
+                                return (d._GatherNd = d.asm.ja).apply(null, arguments)
+                            }, d._Greater = function () {
+                                return (d._Greater = d.asm.ka).apply(null, arguments)
+                            }, d._GreaterEqual = function () {
+                                return (d._GreaterEqual = d.asm.la).apply(null, arguments)
+                            }, d._LeakyRelu = function () {
+                                return (d._LeakyRelu = d.asm.ma).apply(null, arguments)
+                            }, d._Less = function () {
+                                return (d._Less = d.asm.na).apply(null, arguments)
+                            }, d._LessEqual = function () {
+                                return (d._LessEqual = d.asm.oa).apply(null, arguments)
+                            }, d._Log = function () {
+                                return (d._Log = d.asm.pa).apply(null, arguments)
+                            }, d._LogicalAnd = function () {
+                                return (d._LogicalAnd = d.asm.qa).apply(null, arguments)
+                            }, d._Max = function () {
+                                return (d._Max = d.asm.ra).apply(null, arguments)
+                            }, d._MaxPool = function () {
+                                return (d._MaxPool = d.asm.sa).apply(null, arguments)
+                            }, d._Maximum = function () {
+                                return (d._Maximum = d.asm.ta).apply(null, arguments)
+                            }, d._Mean = function () {
+                                return (d._Mean = d.asm.ua).apply(null, arguments)
+                            }, d._Min = function () {
+                                return (d._Min = d.asm.va).apply(null, arguments)
+                            }, d._Minimum = function () {
+                                return (d._Minimum = d.asm.wa).apply(null, arguments)
+                            }, d._MirrorPad = function () {
+                                return (d._MirrorPad = d.asm.xa).apply(null, arguments)
+                            }, d._Multiply = function () {
+                                return (d._Multiply = d.asm.ya).apply(null, arguments)
+                            }, d._Neg = function () {
+                                return (d._Neg = d.asm.za).apply(null, arguments)
+                            }, d._NonMaxSuppressionV3 = function () {
+                                return (d._NonMaxSuppressionV3 = d.asm.Aa).apply(null, arguments)
+                            }, d._NonMaxSuppressionV4 = function () {
+                                return (d._NonMaxSuppressionV4 = d.asm.Ba).apply(null, arguments)
+                            }, d._NonMaxSuppressionV5 = function () {
+                                return (d._NonMaxSuppressionV5 = d.asm.Ca).apply(null, arguments)
+                            }, d._NotEqual = function () {
+                                return (d._NotEqual = d.asm.Da).apply(null, arguments)
+                            }, d._OneHot = function () {
+                                return (d._OneHot = d.asm.Ea).apply(null, arguments)
+                            }, d._PadV2 = function () {
+                                return (d._PadV2 = d.asm.Fa).apply(null, arguments)
+                            }, d._Pow = function () {
+                                return (d._Pow = d.asm.Ga).apply(null, arguments)
+                            }, d._Prelu = function () {
+                                return (d._Prelu = d.asm.Ha).apply(null, arguments)
+                            }, d._Prod = function () {
+                                return (d._Prod = d.asm.Ia).apply(null, arguments)
+                            }, d._RealDiv = function () {
+                                return (d._RealDiv = d.asm.Ja).apply(null, arguments)
+                            }, d._Relu = function () {
+                                return (d._Relu = d.asm.Ka).apply(null, arguments)
+                            }, d._Relu6 = function () {
+                                return (d._Relu6 = d.asm.La).apply(null, arguments)
+                            }, d._ResizeBilinear = function () {
+                                return (d._ResizeBilinear = d.asm.Ma).apply(null, arguments)
+                            }, d._Reverse = function () {
+                                return (d._Reverse = d.asm.Na).apply(null, arguments)
+                            }, d._RotateWithOffset = function () {
+                                return (d._RotateWithOffset = d.asm.Oa).apply(null, arguments)
+                            }, d._Round = function () {
+                                return (d._Round = d.asm.Pa).apply(null, arguments)
+                            }, d._Rsqrt = function () {
+                                return (d._Rsqrt = d.asm.Qa).apply(null, arguments)
+                            }, d._ScatterNd = function () {
+                                return (d._ScatterNd = d.asm.Ra).apply(null, arguments)
+                            }, d._SelectV2 = function () {
+                                return (d._SelectV2 = d.asm.Sa).apply(null, arguments)
+                            }, d._Sigmoid = function () {
+                                return (d._Sigmoid = d.asm.Ta).apply(null, arguments)
+                            }, d._Sin = function () {
+                                return (d._Sin = d.asm.Ua).apply(null, arguments)
+                            }, d._Softmax = function () {
+                                return (d._Softmax = d.asm.Va).apply(null, arguments)
+                            }, d._SparseFillEmptyRows = function () {
+                                return (d._SparseFillEmptyRows = d.asm.Wa).apply(null, arguments)
+                            }, d._SparseReshape = function () {
+                                return (d._SparseReshape = d.asm.Xa).apply(null, arguments)
+                            }, d._SparseSegmentReduction = function () {
+                                return (d._SparseSegmentReduction = d.asm.Ya).apply(null, arguments)
+                            }, d._Sqrt = function () {
+                                return (d._Sqrt = d.asm.Za).apply(null, arguments)
+                            }, d._Square = function () {
+                                return (d._Square = d.asm._a).apply(null, arguments)
+                            }, d._SquaredDifference = function () {
+                                return (d._SquaredDifference = d.asm.$a).apply(null, arguments)
+                            }, d._Step = function () {
+                                return (d._Step = d.asm.ab).apply(null, arguments)
+                            }, d._StridedSlice = function () {
+                                return (d._StridedSlice = d.asm.bb).apply(null, arguments)
+                            }, d._Sub = function () {
+                                return (d._Sub = d.asm.cb).apply(null, arguments)
+                            }, d._Sum = function () {
+                                return (d._Sum = d.asm.db).apply(null, arguments)
+                            }, d._Tan = function () {
+                                return (d._Tan = d.asm.eb).apply(null, arguments)
+                            }, d._Tanh = function () {
+                                return (d._Tanh = d.asm.fb).apply(null, arguments)
+                            }, d._Tile = function () {
+                                return (d._Tile = d.asm.gb).apply(null, arguments)
+                            }, d._TopK = function () {
+                                return (d._TopK = d.asm.hb).apply(null, arguments)
+                            }, d._Transform = function () {
+                                return (d._Transform = d.asm.ib).apply(null, arguments)
+                            }, d._Transpose = function () {
+                                return (d._Transpose = d.asm.jb).apply(null, arguments)
+                            }, d.__FusedMatMul = function () {
+                                return (d.__FusedMatMul = d.asm.kb).apply(null, arguments)
+                            };
+                            var Ze = d._malloc = function () {
+                                    return (Ze = d._malloc = d.asm.lb).apply(null, arguments)
+                                },
+                                Je = d._free = function () {
+                                    return (Je = d._free = d.asm.mb).apply(null, arguments)
+                                },
+                                et = d.___errno_location = function () {
+                                    return (et = d.___errno_location = d.asm.nb).apply(null, arguments)
+                                },
+                                tt = d._emscripten_get_global_libc = function () {
+                                    return (tt = d._emscripten_get_global_libc = d.asm.ob).apply(null, arguments)
+                                },
+                                nt = d._pthread_self = function () {
+                                    return (nt = d._pthread_self = d.asm.pb).apply(null, arguments)
+                                },
+                                rt = d.___pthread_tsd_run_dtors = function () {
+                                    return (rt = d.___pthread_tsd_run_dtors = d.asm.qb).apply(null, arguments)
+                                },
+                                at = d._emscripten_main_thread_process_queued_calls = function () {
+                                    return (at = d._emscripten_main_thread_process_queued_calls = d.asm.rb).apply(null, arguments)
+                                };
+                            d._emscripten_current_thread_process_queued_calls = function () {
+                                return (d._emscripten_current_thread_process_queued_calls = d.asm.sb).apply(null, arguments)
+                            };
+                            var st = d._emscripten_register_main_browser_thread_id = function () {
+                                    return (st = d._emscripten_register_main_browser_thread_id = d.asm.tb).apply(null, arguments)
+                                },
+                                it = d._emscripten_main_browser_thread_id = function () {
+                                    return (it = d._emscripten_main_browser_thread_id = d.asm.ub).apply(null, arguments)
+                                },
+                                ot = d.__emscripten_do_dispatch_to_thread = function () {
+                                    return (ot = d.__emscripten_do_dispatch_to_thread = d.asm.vb).apply(null, arguments)
+                                },
+                                ut = d._emscripten_sync_run_in_main_thread_4 = function () {
+                                    return (ut = d._emscripten_sync_run_in_main_thread_4 = d.asm.wb).apply(null, arguments)
+                                },
+                                lt = d._emscripten_run_in_main_runtime_thread_js = function () {
+                                    return (lt = d._emscripten_run_in_main_runtime_thread_js = d.asm.xb).apply(null, arguments)
+                                },
+                                ct = d.__emscripten_call_on_thread = function () {
+                                    return (ct = d.__emscripten_call_on_thread = d.asm.yb).apply(null, arguments)
+                                };
+                            d._emscripten_tls_init = function () {
+                                return (d._emscripten_tls_init = d.asm.zb).apply(null, arguments)
+                            };
+                            var ht, pt, dt, ft = d.__emscripten_thread_init = function () {
+                                    return (ft = d.__emscripten_thread_init = d.asm.Ab).apply(null, arguments)
+                                },
+                                mt = d.stackSave = function () {
+                                    return (mt = d.stackSave = d.asm.Bb).apply(null, arguments)
+                                },
+                                gt = d.stackRestore = function () {
+                                    return (gt = d.stackRestore = d.asm.Cb).apply(null, arguments)
+                                },
+                                yt = d.stackAlloc = function () {
+                                    return (yt = d.stackAlloc = d.asm.Db).apply(null, arguments)
+                                },
+                                bt = d._emscripten_stack_set_limits = function () {
+                                    return (bt = d._emscripten_stack_set_limits = d.asm.Eb).apply(null, arguments)
+                                },
+                                vt = d._memalign = function () {
+                                    return (vt = d._memalign = d.asm.Fb).apply(null, arguments)
+                                },
+                                xt = d.__emscripten_allow_main_runtime_queued_calls = 10656,
+                                wt = d.__emscripten_main_thread_futex = 12292;
+
+                            function kt(e) {
+                                this.name = "ExitStatus", this.message = "Program terminated with exit(" + e + ")", this.status = e
+                            }
+
+                            function St(e) {
+                                if (!(ue > 0)) {
+                                    if (w) return c(d), oe(), void postMessage({
+                                        cmd: "loaded"
+                                    });
+                                    ! function () {
+                                        if (!w) {
+                                            if (d.preRun)
+                                                for ("function" == typeof d.preRun && (d.preRun = [d.preRun]); d.preRun.length;) e = d.preRun.shift(), re.unshift(e);
+                                            var e;
+                                            ye(re)
+                                        }
+                                    }(), ue > 0 || (d.setStatus ? (d.setStatus("Running..."), setTimeout((function () {
+                                        setTimeout((function () {
+                                            d.setStatus("")
+                                        }), 1), t()
+                                    }), 1)) : t())
+                                }
+
+                                function t() {
+                                    ht || (ht = !0, d.calledRun = !0, W || (oe(), w || ye(se), c(d), d.onRuntimeInitialized && d.onRuntimeInitialized(), function () {
+                                        if (!w) {
+                                            if (d.postRun)
+                                                for ("function" == typeof d.postRun && (d.postRun = [d.postRun]); d.postRun.length;) e = d.postRun.shift(), ie.unshift(e);
+                                            var e;
+                                            ye(ie)
+                                        }
+                                    }()))
+                                }
+                            }
+                            if (d.cwrap = function (e, t, n, r) {
+                                    var a = (n = n || []).every((function (e) {
+                                        return "number" === e
+                                    }));
+                                    return "string" !== t && a && !r ? Y(e) : function () {
+                                        return X(e, t, n, arguments)
+                                    }
+                                }, d.PThread = Ae, d.PThread = Ae, d.wasmMemory = M, d.ExitStatus = kt, le = function e() {
+                                    ht || St(), ht || (le = e)
+                                }, d.run = St, d.preInit)
+                                for ("function" == typeof d.preInit && (d.preInit = [d.preInit]); d.preInit.length > 0;) d.preInit.pop()();
+                            if (w && (O = !1, Ae.initWorker()), St(), p && (pt = {
+                                    uncaughtException: r.listeners("uncaughtException").filter((function (e) {
+                                        return !p.uncaughtException.indexOf(e) > -1
+                                    })),
+                                    unhandledRejection: r.listeners("unhandledRejection").filter((function (e) {
+                                        return !p.unhandledRejection.indexOf(e) > -1
+                                    }))
+                                }), "undefined" != typeof WasmBackendModule) dt = WasmBackendModule;
+                            else {
+                                if (void 0 === e) throw new Error("Could not find wasm module in post.js");
+                                dt = e
+                            }
+                            if (pt) {
+                                var It = dt._dispose;
+                                dt._dispose = function () {
+                                    It(), pt.uncaughtException.forEach((function (e) {
+                                        r.removeListener("uncaughtException", e)
+                                    })), pt.unhandledRejection.forEach((function (e) {
+                                        r.removeListener("unhandledRejection", e)
+                                    }))
+                                }
+                            }
+                            return e.ready
+                        });
+                        "object" == typeof t && "object" == typeof i ? i.exports = u : "function" == typeof define && n(67) ? define([], (function () {
+                            return u
+                        })) : "object" == typeof t && (t.WasmBackendModuleThreadedSimd = u)
+                    }
+                }),
+                X = x({
+                    "node_modules/.pnpm/@tensorflow+tfjs-backend-wasm@3.13.0_@tensorflow+tfjs-core@3.13.0/node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js"(t, s) {
+                        var i, o = (i = (i = "undefined" != typeof document && document.currentScript ? document.currentScript.src : void 0) || e, function (e) {
+                            var t, n, s, o = void 0 !== (e = e || {}) ? e : {};
+                            o.ready = new Promise((function (e, r) {
+                                t = e, n = r
+                            })), void 0 !== r && r.listeners && (s = {
+                                uncaughtException: r.listeners("uncaughtException"),
+                                unhandledRejection: r.listeners("unhandledRejection")
+                            });
+                            var u, l, c, h, p, d = {};
+                            for (u in o) o.hasOwnProperty(u) && (d[u] = o[u]);
+                            l = "object" == typeof window, c = "function" == typeof importScripts, h = "object" == typeof r && "object" == typeof r.versions && "string" == typeof r.versions.node, p = !l && !h && !c;
+                            var f, m, g, y, b, v = "";
+                            h ? (v = c ? G().dirname(v) + "/" : a + "/", f = function (e, t) {
+                                return y || (y = j()), b || (b = G()), e = b.normalize(e), y.readFileSync(e, t ? null : "utf8")
+                            }, g = function (e) {
+                                var t = f(e, !0);
+                                return t.buffer || (t = new Uint8Array(t)), A(t.buffer), t
+                            }, r.argv.length > 1 && r.argv[1].replace(/\\/g, "/"), r.argv.slice(2), r.on("uncaughtException", (function (e) {
+                                if (!(e instanceof ce)) throw e
+                            })), r.on("unhandledRejection", V), o.inspect = function () {
+                                return "[Emscripten Module object]"
+                            }) : p ? ("undefined" != typeof read && (f = function (e) {
+                                return read(e)
+                            }), g = function (e) {
+                                var t;
+                                return "function" == typeof readbuffer ? new Uint8Array(readbuffer(e)) : (A("object" == typeof (t = read(e, "binary"))), t)
+                            }, "undefined" != typeof scriptArgs && scriptArgs, "undefined" != typeof print && ("undefined" == typeof console && (console = {}), console.log = print, console.warn = console.error = "undefined" != typeof printErr ? printErr : print)) : (l || c) && (c ? v = self.location.href : "undefined" != typeof document && document.currentScript && (v = document.currentScript.src), i && (v = i), v = 0 !== v.indexOf("blob:") ? v.substr(0, v.lastIndexOf("/") + 1) : "", f = function (e) {
+                                var t = new XMLHttpRequest;
+                                return t.open("GET", e, !1), t.send(null), t.responseText
+                            }, c && (g = function (e) {
+                                var t = new XMLHttpRequest;
+                                return t.open("GET", e, !1), t.responseType = "arraybuffer", t.send(null), new Uint8Array(t.response)
+                            }), m = function (e, t, n) {
+                                var r = new XMLHttpRequest;
+                                r.open("GET", e, !0), r.responseType = "arraybuffer", r.onload = function () {
+                                    200 == r.status || 0 == r.status && r.response ? t(r.response) : n()
+                                }, r.onerror = n, r.send(null)
+                            });
+                            var x, w, k = o.print || console.log.bind(console),
+                                S = o.printErr || console.warn.bind(console);
+                            for (u in d) d.hasOwnProperty(u) && (o[u] = d[u]);
+                            d = null, o.wasmBinary && (x = o.wasmBinary), "object" != typeof WebAssembly && V("no native wasm support detected");
+                            var I = !1;
+
+                            function A(e, t) {
+                                e || V("Assertion failed: " + t)
+                            }
+
+                            function E(e) {
+                                var t = o["_" + e];
+                                return A(t, "Cannot call unknown function " + e + ", make sure it is exported"), t
+                            }
+
+                            function N(e, t, n, r, a) {
+                                var s = {
+                                        string: function (e) {
+                                            var t = 0;
+                                            if (null != e && 0 !== e) {
+                                                var n = 1 + (e.length << 2);
+                                                ! function (e, t, n) {
+                                                    ! function (e, t, n, r) {
+                                                        if (!(r > 0)) return 0;
+                                                        for (var a = n + r - 1, s = 0; s < e.length; ++s) {
+                                                            var i = e.charCodeAt(s);
+                                                            if (i >= 55296 && i <= 57343) {
+                                                                var o = e.charCodeAt(++s);
+                                                                i = 65536 + ((1023 & i) << 10) | 1023 & o
+                                                            }
+                                                            if (i <= 127) {
+                                                                if (n >= a) break;
+                                                                t[n++] = i
+                                                            } else if (i <= 2047) {
+                                                                if (n + 1 >= a) break;
+                                                                t[n++] = 192 | i >> 6, t[n++] = 128 | 63 & i
+                                                            } else if (i <= 65535) {
+                                                                if (n + 2 >= a) break;
+                                                                t[n++] = 224 | i >> 12, t[n++] = 128 | i >> 6 & 63, t[n++] = 128 | 63 & i
+                                                            } else {
+                                                                if (n + 3 >= a) break;
+                                                                t[n++] = 240 | i >> 18, t[n++] = 128 | i >> 12 & 63, t[n++] = 128 | i >> 6 & 63, t[n++] = 128 | 63 & i
+                                                            }
+                                                        }
+                                                        t[n] = 0
+                                                    }(e, R, t, n)
+                                                }(e, t = le(n), n)
+                                            }
+                                            return t
+                                        },
+                                        array: function (e) {
+                                            var t, n, r = le(e.length);
+                                            return t = e, n = r, T.set(t, n), r
+                                        }
+                                    },
+                                    i = E(e),
+                                    o = [],
+                                    u = 0;
+                                if (r)
+                                    for (var l = 0; l < r.length; l++) {
+                                        var c = s[n[l]];
+                                        c ? (0 === u && (u = oe()), o[l] = c(r[l])) : o[l] = r[l]
+                                    }
+                                var h, p = i.apply(null, o);
+                                return h = p, p = "string" === t ? O(h) : "boolean" === t ? Boolean(h) : h, 0 !== u && ue(u), p
+                            }
+                            var C, T, R, _, F, M = "undefined" != typeof TextDecoder ? new TextDecoder("utf8") : void 0;
+
+                            function D(e, t, n) {
+                                for (var r = t + n, a = t; e[a] && !(a >= r);) ++a;
+                                if (a - t > 16 && e.subarray && M) return M.decode(e.subarray(t, a));
+                                for (var s = ""; t < a;) {
+                                    var i = e[t++];
+                                    if (128 & i) {
+                                        var o = 63 & e[t++];
+                                        if (192 != (224 & i)) {
+                                            var u = 63 & e[t++];
+                                            if ((i = 224 == (240 & i) ? (15 & i) << 12 | o << 6 | u : (7 & i) << 18 | o << 12 | u << 6 | 63 & e[t++]) < 65536) s += String.fromCharCode(i);
+                                            else {
+                                                var l = i - 65536;
+                                                s += String.fromCharCode(55296 | l >> 10, 56320 | 1023 & l)
+                                            }
+                                        } else s += String.fromCharCode((31 & i) << 6 | o)
+                                    } else s += String.fromCharCode(i)
+                                }
+                                return s
+                            }
+
+                            function O(e, t) {
+                                return e ? D(R, e, t) : ""
+                            }
+
+                            function L(e) {
+                                C = e, o.HEAP8 = T = new Int8Array(e), o.HEAP16 = new Int16Array(e), o.HEAP32 = _ = new Int32Array(e), o.HEAPU8 = R = new Uint8Array(e), o.HEAPU16 = new Uint16Array(e), o.HEAPU32 = new Uint32Array(e), o.HEAPF32 = new Float32Array(e), o.HEAPF64 = new Float64Array(e)
+                            }
+                            var P = [],
+                                $ = [],
+                                B = [],
+                                z = [];
+                            $.push({
+                                func: function () {
+                                    ne()
+                                }
+                            });
+                            var U = 0,
+                                W = null;
+
+                            function V(e) {
+                                o.onAbort && o.onAbort(e), S(e += ""), I = !0, e = "abort(" + e + "). Build with -s ASSERTIONS=1 for more info.";
+                                var t = new WebAssembly.RuntimeError(e);
+                                throw n(t), t
+                            }
+
+                            function H(e, t) {
+                                return String.prototype.startsWith ? e.startsWith(t) : 0 === e.indexOf(t)
+                            }
+
+                            function q(e) {
+                                return H(e, "data:application/octet-stream;base64,")
+                            }
+
+                            function K(e) {
+                                return H(e, "file://")
+                            }
+                            o.preloadedImages = {}, o.preloadedAudios = {};
+                            var Y, X = "tfjs-backend-wasm.wasm";
+
+                            function Q(e) {
+                                try {
+                                    if (e == X && x) return new Uint8Array(x);
+                                    if (g) return g(e);
+                                    throw "both async and sync fetching of the wasm failed"
+                                } catch (e) {
+                                    V(e)
+                                }
+                            }
+
+                            function Z(e) {
+                                for (; e.length > 0;) {
+                                    var t = e.shift();
+                                    if ("function" != typeof t) {
+                                        var n = t.func;
+                                        "number" == typeof n ? void 0 === t.arg ? F.get(n)() : F.get(n)(t.arg) : n(void 0 === t.arg ? null : t.arg)
+                                    } else t(o)
+                                }
+                            }
+
+                            function J(e) {
+                                try {
+                                    return w.grow(e - C.byteLength + 65535 >>> 16), L(w.buffer), 1
+                                } catch (e) {}
+                            }
+                            q(X) || (Y = X, X = o.locateFile ? o.locateFile(Y, v) : v + Y);
+                            var ee = {
+                                    mappings: {},
+                                    buffers: [null, [],
+                                        []
+                                    ],
+                                    printChar: function (e, t) {
+                                        var n = ee.buffers[e];
+                                        0 === t || 10 === t ? ((1 === e ? k : S)(D(n, 0)), n.length = 0) : n.push(t)
+                                    },
+                                    varargs: void 0,
+                                    get: function () {
+                                        return ee.varargs += 4, _[ee.varargs - 4 >> 2]
+                                    },
+                                    getStr: function (e) {
+                                        return O(e)
+                                    },
+                                    get64: function (e, t) {
+                                        return e
+                                    }
+                                },
+                                te = {
+                                    a: function () {
+                                        V()
+                                    },
+                                    d: function (e, t, n) {
+                                        R.copyWithin(e, t, t + n)
+                                    },
+                                    e: function (e) {
+                                        var t, n, r = R.length;
+                                        if (e > 2147483648) return !1;
+                                        for (var a = 1; a <= 4; a *= 2) {
+                                            var s = r * (1 + .2 / a);
+                                            if (s = Math.min(s, e + 100663296), J(Math.min(2147483648, ((t = Math.max(e, s)) % (n = 65536) > 0 && (t += n - t % n), t)))) return !0
+                                        }
+                                        return !1
+                                    },
+                                    f: function (e) {
+                                        return 0
+                                    },
+                                    c: function (e, t, n, r, a) {},
+                                    b: function (e, t, n, r) {
+                                        for (var a = 0, s = 0; s < n; s++) {
+                                            for (var i = _[t + 8 * s >> 2], o = _[t + (8 * s + 4) >> 2], u = 0; u < o; u++) ee.printChar(e, R[i + u]);
+                                            a += o
+                                        }
+                                        return _[r >> 2] = a, 0
+                                    },
+                                    h: function () {
+                                        return 6
+                                    },
+                                    g: function () {
+                                        return 28
+                                    },
+                                    i: function (e) {
+                                        switch (e) {
+                                            case 30:
+                                                return 16384;
+                                            case 85:
+                                                return 131072;
+                                            case 132:
+                                            case 133:
+                                            case 12:
+                                            case 137:
+                                            case 138:
+                                            case 15:
+                                            case 235:
+                                            case 16:
+                                            case 17:
+                                            case 18:
+                                            case 19:
+                                            case 20:
+                                            case 149:
+                                            case 13:
+                                            case 10:
+                                            case 236:
+                                            case 153:
+                                            case 9:
+                                            case 21:
+                                            case 22:
+                                            case 159:
+                                            case 154:
+                                            case 14:
+                                            case 77:
+                                            case 78:
+                                            case 139:
+                                            case 82:
+                                            case 68:
+                                            case 67:
+                                            case 164:
+                                            case 11:
+                                            case 29:
+                                            case 47:
+                                            case 48:
+                                            case 95:
+                                            case 52:
+                                            case 51:
+                                            case 46:
+                                                return 200809;
+                                            case 27:
+                                            case 246:
+                                            case 127:
+                                            case 128:
+                                            case 23:
+                                            case 24:
+                                            case 160:
+                                            case 161:
+                                            case 181:
+                                            case 182:
+                                            case 242:
+                                            case 183:
+                                            case 184:
+                                            case 243:
+                                            case 244:
+                                            case 245:
+                                            case 165:
+                                            case 178:
+                                            case 179:
+                                            case 49:
+                                            case 50:
+                                            case 168:
+                                            case 169:
+                                            case 175:
+                                            case 170:
+                                            case 171:
+                                            case 172:
+                                            case 97:
+                                            case 76:
+                                            case 32:
+                                            case 173:
+                                            case 35:
+                                            case 80:
+                                            case 81:
+                                            case 79:
+                                                return -1;
+                                            case 176:
+                                            case 177:
+                                            case 7:
+                                            case 155:
+                                            case 8:
+                                            case 157:
+                                            case 125:
+                                            case 126:
+                                            case 92:
+                                            case 93:
+                                            case 129:
+                                            case 130:
+                                            case 131:
+                                            case 94:
+                                            case 91:
+                                                return 1;
+                                            case 74:
+                                            case 60:
+                                            case 69:
+                                            case 70:
+                                            case 4:
+                                                return 1024;
+                                            case 31:
+                                            case 42:
+                                            case 72:
+                                                return 32;
+                                            case 87:
+                                            case 26:
+                                            case 33:
+                                                return 2147483647;
+                                            case 34:
+                                            case 1:
+                                                return 47839;
+                                            case 38:
+                                            case 36:
+                                                return 99;
+                                            case 43:
+                                            case 37:
+                                                return 2048;
+                                            case 0:
+                                                return 2097152;
+                                            case 3:
+                                                return 65536;
+                                            case 28:
+                                                return 32768;
+                                            case 44:
+                                                return 32767;
+                                            case 75:
+                                                return 16384;
+                                            case 39:
+                                                return 1e3;
+                                            case 89:
+                                                return 700;
+                                            case 71:
+                                                return 256;
+                                            case 40:
+                                                return 255;
+                                            case 2:
+                                                return 100;
+                                            case 180:
+                                                return 64;
+                                            case 25:
+                                                return 20;
+                                            case 5:
+                                                return 16;
+                                            case 6:
+                                                return 6;
+                                            case 73:
+                                                return 4;
+                                            case 84:
+                                                return "object" == typeof navigator && navigator.hardwareConcurrency || 1
+                                        }
+                                        var t;
+                                        return t = 28, _[ie() >> 2] = t, -1
+                                    }
+                                };
+                            ! function () {
+                                var e = {
+                                    a: te
+                                };
+
+                                function t(e, t) {
+                                    var n = e.exports;
+                                    o.asm = n, L((w = o.asm.j).buffer), F = o.asm.r,
+                                        function (e) {
+                                            if (U--, o.monitorRunDependencies && o.monitorRunDependencies(U), 0 == U && W) {
+                                                var t = W;
+                                                W = null, t()
+                                            }
+                                        }()
+                                }
+
+                                function r(e) {
+                                    t(e.instance)
+                                }
+
+                                function a(t) {
+                                    return function () {
+                                        if (!x && (l || c)) {
+                                            if ("function" == typeof fetch && !K(X)) return fetch(X, {
+                                                credentials: "same-origin"
+                                            }).then((function (e) {
+                                                if (!e.ok) throw "failed to load wasm binary file at '" + X + "'";
+                                                return e.arrayBuffer()
+                                            })).catch((function () {
+                                                return Q(X)
+                                            }));
+                                            if (m) return new Promise((function (e, t) {
+                                                m(X, (function (t) {
+                                                    e(new Uint8Array(t))
+                                                }), t)
+                                            }))
+                                        }
+                                        return Promise.resolve().then((function () {
+                                            return Q(X)
+                                        }))
+                                    }().then((function (t) {
+                                        return WebAssembly.instantiate(t, e)
+                                    })).then(t, (function (e) {
+                                        S("failed to asynchronously prepare wasm: " + e), V(e)
+                                    }))
+                                }
+                                if (U++, o.monitorRunDependencies && o.monitorRunDependencies(U), o.instantiateWasm) try {
+                                    return o.instantiateWasm(e, t)
+                                } catch (e) {
+                                    return S("Module.instantiateWasm callback failed with error: " + e), !1
+                                }(x || "function" != typeof WebAssembly.instantiateStreaming || q(X) || K(X) || "function" != typeof fetch ? a(r) : fetch(X, {
+                                    credentials: "same-origin"
+                                }).then((function (t) {
+                                    return WebAssembly.instantiateStreaming(t, e).then(r, (function (e) {
+                                        return S("wasm streaming compile failed: " + e), S("falling back to ArrayBuffer instantiation"), a(r)
+                                    }))
+                                }))).catch(n)
+                            }();
+                            var ne = o.___wasm_call_ctors = function () {
+                                return (ne = o.___wasm_call_ctors = o.asm.k).apply(null, arguments)
+                            };
+                            o._init = function () {
+                                return (o._init = o.asm.l).apply(null, arguments)
+                            }, o._init_with_threads_count = function () {
+                                return (o._init_with_threads_count = o.asm.m).apply(null, arguments)
+                            }, o._get_threads_count = function () {
+                                return (o._get_threads_count = o.asm.n).apply(null, arguments)
+                            }, o._register_tensor = function () {
+                                return (o._register_tensor = o.asm.o).apply(null, arguments)
+                            }, o._dispose_data = function () {
+                                return (o._dispose_data = o.asm.p).apply(null, arguments)
+                            }, o._dispose = function () {
+                                return (o._dispose = o.asm.q).apply(null, arguments)
+                            }, o._Abs = function () {
+                                return (o._Abs = o.asm.s).apply(null, arguments)
+                            }, o._Add = function () {
+                                return (o._Add = o.asm.t).apply(null, arguments)
+                            }, o._AddN = function () {
+                                return (o._AddN = o.asm.u).apply(null, arguments)
+                            }, o._All = function () {
+                                return (o._All = o.asm.v).apply(null, arguments)
+                            }, o._Any = function () {
+                                return (o._Any = o.asm.w).apply(null, arguments)
+                            }, o._ArgMax = function () {
+                                return (o._ArgMax = o.asm.x).apply(null, arguments)
+                            }, o._AvgPool = function () {
+                                return (o._AvgPool = o.asm.y).apply(null, arguments)
+                            }, o._BatchMatMul = function () {
+                                return (o._BatchMatMul = o.asm.z).apply(null, arguments)
+                            }, o._Ceil = function () {
+                                return (o._Ceil = o.asm.A).apply(null, arguments)
+                            }, o._ClipByValue = function () {
+                                return (o._ClipByValue = o.asm.B).apply(null, arguments)
+                            }, o._Conv2D = function () {
+                                return (o._Conv2D = o.asm.C).apply(null, arguments)
+                            }, o._Conv2DBackpropInput = function () {
+                                return (o._Conv2DBackpropInput = o.asm.D).apply(null, arguments)
+                            }, o._Cos = function () {
+                                return (o._Cos = o.asm.E).apply(null, arguments)
+                            }, o._Cosh = function () {
+                                return (o._Cosh = o.asm.F).apply(null, arguments)
+                            }, o._CropAndResize = function () {
+                                return (o._CropAndResize = o.asm.G).apply(null, arguments)
+                            }, o._Cumsum = function () {
+                                return (o._Cumsum = o.asm.H).apply(null, arguments)
+                            }, o._DepthToSpace = function () {
+                                return (o._DepthToSpace = o.asm.I).apply(null, arguments)
+                            }, o._DepthwiseConv2dNative = function () {
+                                return (o._DepthwiseConv2dNative = o.asm.J).apply(null, arguments)
+                            }, o._Elu = function () {
+                                return (o._Elu = o.asm.K).apply(null, arguments)
+                            }, o._Equal = function () {
+                                return (o._Equal = o.asm.L).apply(null, arguments)
+                            }, o._Exp = function () {
+                                return (o._Exp = o.asm.M).apply(null, arguments)
+                            }, o._FlipLeftRight = function () {
+                                return (o._FlipLeftRight = o.asm.N).apply(null, arguments)
+                            }, o._Floor = function () {
+                                return (o._Floor = o.asm.O).apply(null, arguments)
+                            }, o._FloorDiv = function () {
+                                return (o._FloorDiv = o.asm.P).apply(null, arguments)
+                            }, o._FusedBatchNorm = function () {
+                                return (o._FusedBatchNorm = o.asm.Q).apply(null, arguments)
+                            }, o._FusedConv2D = function () {
+                                return (o._FusedConv2D = o.asm.R).apply(null, arguments)
+                            }, o._FusedDepthwiseConv2D = function () {
+                                return (o._FusedDepthwiseConv2D = o.asm.S).apply(null, arguments)
+                            }, o._Gather = function () {
+                                return (o._Gather = o.asm.T).apply(null, arguments)
+                            }, o._GatherNd = function () {
+                                return (o._GatherNd = o.asm.U).apply(null, arguments)
+                            }, o._Greater = function () {
+                                return (o._Greater = o.asm.V).apply(null, arguments)
+                            }, o._GreaterEqual = function () {
+                                return (o._GreaterEqual = o.asm.W).apply(null, arguments)
+                            }, o._LeakyRelu = function () {
+                                return (o._LeakyRelu = o.asm.X).apply(null, arguments)
+                            }, o._Less = function () {
+                                return (o._Less = o.asm.Y).apply(null, arguments)
+                            }, o._LessEqual = function () {
+                                return (o._LessEqual = o.asm.Z).apply(null, arguments)
+                            }, o._Log = function () {
+                                return (o._Log = o.asm._).apply(null, arguments)
+                            }, o._LogicalAnd = function () {
+                                return (o._LogicalAnd = o.asm.$).apply(null, arguments)
+                            }, o._Max = function () {
+                                return (o._Max = o.asm.aa).apply(null, arguments)
+                            }, o._MaxPool = function () {
+                                return (o._MaxPool = o.asm.ba).apply(null, arguments)
+                            }, o._Maximum = function () {
+                                return (o._Maximum = o.asm.ca).apply(null, arguments)
+                            }, o._Mean = function () {
+                                return (o._Mean = o.asm.da).apply(null, arguments)
+                            }, o._Min = function () {
+                                return (o._Min = o.asm.ea).apply(null, arguments)
+                            }, o._Minimum = function () {
+                                return (o._Minimum = o.asm.fa).apply(null, arguments)
+                            }, o._MirrorPad = function () {
+                                return (o._MirrorPad = o.asm.ga).apply(null, arguments)
+                            }, o._Multiply = function () {
+                                return (o._Multiply = o.asm.ha).apply(null, arguments)
+                            }, o._Neg = function () {
+                                return (o._Neg = o.asm.ia).apply(null, arguments)
+                            }, o._NonMaxSuppressionV3 = function () {
+                                return (o._NonMaxSuppressionV3 = o.asm.ja).apply(null, arguments)
+                            }, o._NonMaxSuppressionV4 = function () {
+                                return (o._NonMaxSuppressionV4 = o.asm.ka).apply(null, arguments)
+                            }, o._NonMaxSuppressionV5 = function () {
+                                return (o._NonMaxSuppressionV5 = o.asm.la).apply(null, arguments)
+                            }, o._NotEqual = function () {
+                                return (o._NotEqual = o.asm.ma).apply(null, arguments)
+                            }, o._OneHot = function () {
+                                return (o._OneHot = o.asm.na).apply(null, arguments)
+                            }, o._PadV2 = function () {
+                                return (o._PadV2 = o.asm.oa).apply(null, arguments)
+                            }, o._Pow = function () {
+                                return (o._Pow = o.asm.pa).apply(null, arguments)
+                            }, o._Prelu = function () {
+                                return (o._Prelu = o.asm.qa).apply(null, arguments)
+                            }, o._Prod = function () {
+                                return (o._Prod = o.asm.ra).apply(null, arguments)
+                            }, o._RealDiv = function () {
+                                return (o._RealDiv = o.asm.sa).apply(null, arguments)
+                            }, o._Relu = function () {
+                                return (o._Relu = o.asm.ta).apply(null, arguments)
+                            }, o._Relu6 = function () {
+                                return (o._Relu6 = o.asm.ua).apply(null, arguments)
+                            }, o._ResizeBilinear = function () {
+                                return (o._ResizeBilinear = o.asm.va).apply(null, arguments)
+                            }, o._Reverse = function () {
+                                return (o._Reverse = o.asm.wa).apply(null, arguments)
+                            }, o._RotateWithOffset = function () {
+                                return (o._RotateWithOffset = o.asm.xa).apply(null, arguments)
+                            }, o._Round = function () {
+                                return (o._Round = o.asm.ya).apply(null, arguments)
+                            }, o._Rsqrt = function () {
+                                return (o._Rsqrt = o.asm.za).apply(null, arguments)
+                            }, o._ScatterNd = function () {
+                                return (o._ScatterNd = o.asm.Aa).apply(null, arguments)
+                            }, o._SelectV2 = function () {
+                                return (o._SelectV2 = o.asm.Ba).apply(null, arguments)
+                            }, o._Sigmoid = function () {
+                                return (o._Sigmoid = o.asm.Ca).apply(null, arguments)
+                            }, o._Sin = function () {
+                                return (o._Sin = o.asm.Da).apply(null, arguments)
+                            }, o._Softmax = function () {
+                                return (o._Softmax = o.asm.Ea).apply(null, arguments)
+                            }, o._SparseFillEmptyRows = function () {
+                                return (o._SparseFillEmptyRows = o.asm.Fa).apply(null, arguments)
+                            }, o._SparseReshape = function () {
+                                return (o._SparseReshape = o.asm.Ga).apply(null, arguments)
+                            }, o._SparseSegmentReduction = function () {
+                                return (o._SparseSegmentReduction = o.asm.Ha).apply(null, arguments)
+                            }, o._Sqrt = function () {
+                                return (o._Sqrt = o.asm.Ia).apply(null, arguments)
+                            }, o._Square = function () {
+                                return (o._Square = o.asm.Ja).apply(null, arguments)
+                            }, o._SquaredDifference = function () {
+                                return (o._SquaredDifference = o.asm.Ka).apply(null, arguments)
+                            }, o._Step = function () {
+                                return (o._Step = o.asm.La).apply(null, arguments)
+                            }, o._StridedSlice = function () {
+                                return (o._StridedSlice = o.asm.Ma).apply(null, arguments)
+                            }, o._Sub = function () {
+                                return (o._Sub = o.asm.Na).apply(null, arguments)
+                            }, o._Sum = function () {
+                                return (o._Sum = o.asm.Oa).apply(null, arguments)
+                            }, o._Tan = function () {
+                                return (o._Tan = o.asm.Pa).apply(null, arguments)
+                            }, o._Tanh = function () {
+                                return (o._Tanh = o.asm.Qa).apply(null, arguments)
+                            }, o._Tile = function () {
+                                return (o._Tile = o.asm.Ra).apply(null, arguments)
+                            }, o._TopK = function () {
+                                return (o._TopK = o.asm.Sa).apply(null, arguments)
+                            }, o._Transform = function () {
+                                return (o._Transform = o.asm.Ta).apply(null, arguments)
+                            }, o._Transpose = function () {
+                                return (o._Transpose = o.asm.Ua).apply(null, arguments)
+                            }, o.__FusedMatMul = function () {
+                                return (o.__FusedMatMul = o.asm.Va).apply(null, arguments)
+                            }, o._malloc = function () {
+                                return (o._malloc = o.asm.Wa).apply(null, arguments)
+                            }, o._free = function () {
+                                return (o._free = o.asm.Xa).apply(null, arguments)
+                            };
+                            var re, ae, se, ie = o.___errno_location = function () {
+                                    return (ie = o.___errno_location = o.asm.Ya).apply(null, arguments)
+                                },
+                                oe = o.stackSave = function () {
+                                    return (oe = o.stackSave = o.asm.Za).apply(null, arguments)
+                                },
+                                ue = o.stackRestore = function () {
+                                    return (ue = o.stackRestore = o.asm._a).apply(null, arguments)
+                                },
+                                le = o.stackAlloc = function () {
+                                    return (le = o.stackAlloc = o.asm.$a).apply(null, arguments)
+                                };
+
+                            function ce(e) {
+                                this.name = "ExitStatus", this.message = "Program terminated with exit(" + e + ")", this.status = e
+                            }
+
+                            function he(e) {
+                                function n() {
+                                    re || (re = !0, o.calledRun = !0, I || (Z($), Z(B), t(o), o.onRuntimeInitialized && o.onRuntimeInitialized(), function () {
+                                        if (o.postRun)
+                                            for ("function" == typeof o.postRun && (o.postRun = [o.postRun]); o.postRun.length;) e = o.postRun.shift(), z.unshift(e);
+                                        var e;
+                                        Z(z)
+                                    }()))
+                                }
+                                U > 0 || (function () {
+                                    if (o.preRun)
+                                        for ("function" == typeof o.preRun && (o.preRun = [o.preRun]); o.preRun.length;) e = o.preRun.shift(), P.unshift(e);
+                                    var e;
+                                    Z(P)
+                                }(), U > 0 || (o.setStatus ? (o.setStatus("Running..."), setTimeout((function () {
+                                    setTimeout((function () {
+                                        o.setStatus("")
+                                    }), 1), n()
+                                }), 1)) : n()))
+                            }
+                            if (o.cwrap = function (e, t, n, r) {
+                                    var a = (n = n || []).every((function (e) {
+                                        return "number" === e
+                                    }));
+                                    return "string" !== t && a && !r ? E(e) : function () {
+                                        return N(e, t, n, arguments)
+                                    }
+                                }, W = function e() {
+                                    re || he(), re || (W = e)
+                                }, o.run = he, o.preInit)
+                                for ("function" == typeof o.preInit && (o.preInit = [o.preInit]); o.preInit.length > 0;) o.preInit.pop()();
+                            if (he(), s && (ae = {
+                                    uncaughtException: r.listeners("uncaughtException").filter((function (e) {
+                                        return !s.uncaughtException.indexOf(e) > -1
+                                    })),
+                                    unhandledRejection: r.listeners("unhandledRejection").filter((function (e) {
+                                        return !s.unhandledRejection.indexOf(e) > -1
+                                    }))
+                                }), void 0 !== e) se = e;
+                            else {
+                                if ("undefined" == typeof WasmBackendModuleThreadedSimd) throw new Error("Could not find wasm module in post.js");
+                                se = WasmBackendModuleThreadedSimd
+                            }
+                            if (ae) {
+                                var pe = se._dispose;
+                                se._dispose = function () {
+                                    pe(), ae.uncaughtException.forEach((function (e) {
+                                        r.removeListener("uncaughtException", e)
+                                    })), ae.unhandledRejection.forEach((function (e) {
+                                        r.removeListener("unhandledRejection", e)
+                                    }))
+                                }
+                            }
+                            return e.ready
+                        });
+                        "object" == typeof t && "object" == typeof s ? s.exports = o : "function" == typeof define && n(67) ? define([], (function () {
+                            return o
+                        })) : "object" == typeof t && (t.WasmBackendModule = o)
+                    }
+                }),
+                Q = class {
+                    constructor(e, t) {
+                        this.backend = e, this.dataMover = t, this.data = new WeakMap, this.dataIdsCount = 0
+                    }
+                    get(e) {
+                        return this.data.has(e) || this.dataMover.moveData(this.backend, e), this.data.get(e)
+                    }
+                    set(e, t) {
+                        this.dataIdsCount++, this.data.set(e, t)
+                    }
+                    has(e) {
+                        return this.data.has(e)
+                    }
+                    delete(e) {
+                        return this.dataIdsCount--, this.data.delete(e)
+                    }
+                    numDataIds() {
+                        return this.dataIdsCount
+                    }
+                },
+                Z = class {
+                    refCount(e) {
+                        return J("refCount")
+                    }
+                    incRef(e) {
+                        return J("incRef")
+                    }
+                    timerAvailable() {
+                        return !0
+                    }
+                    time(e) {
+                        return J("time")
+                    }
+                    read(e) {
+                        return J("read")
+                    }
+                    readSync(e) {
+                        return J("readSync")
+                    }
+                    readToGPU(e, t) {
+                        return J("readToGPU")
+                    }
+                    numDataIds() {
+                        return J("numDataIds")
+                    }
+                    disposeData(e, t) {
+                        return J("disposeData")
+                    }
+                    write(e, t, n) {
+                        return J("write")
+                    }
+                    move(e, t, n, r, a) {
+                        return J("move")
+                    }
+                    memory() {
+                        return J("memory")
+                    }
+                    floatPrecision() {
+                        return J("floatPrecision")
+                    }
+                    epsilon() {
+                        return 32 === this.floatPrecision() ? 1e-7 : 1e-4
+                    }
+                    dispose() {
+                        return J("dispose")
+                    }
+                };
+
+            function J(e) {
+                throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)
+            }
+
+            function ee(e) {
+                let t = e.length,
+                    n = 0;
+                for (; t > 0;) n = Math.random() * t | 0, t--, ae(e, t, n)
+            }
+
+            function te(e, t) {
+                if (e.length !== t.length) throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);
+                let n = e.length,
+                    r = 0;
+                for (; n > 0;) r = Math.random() * n | 0, n--, ae(e, n, r), ae(t, n, r)
+            }
+
+            function ne(e, t, n) {
+                return Math.max(e, Math.min(t, n))
+            }
+
+            function re(e) {
+                return e % 2 == 0 ? e : e + 1
+            }
+
+            function ae(e, t, n) {
+                const r = e[t];
+                e[t] = e[n], e[n] = r
+            }
+
+            function se(e) {
+                let t = 0;
+                for (let n = 0; n < e.length; n++) t += e[n];
+                return t
+            }
+
+            function ie(e, t) {
+                const n = Math.random();
+                return t * n + (1 - n) * e
+            }
+
+            function oe(e, t) {
+                let n = 0;
+                for (let r = 0; r < e.length; r++) {
+                    const a = Number(e[r]) - Number(t[r]);
+                    n += a * a
+                }
+                return n
+            }
+
+            function ue(e, t) {
+                if (!e) throw new Error("string" == typeof t ? t : t())
+            }
+
+            function le(e, t, n = "") {
+                ue(fe(e, t), () => n + ` Shapes ${e} and ${t} must match`)
+            }
+
+            function ce(e) {
+                ue(null != e, () => "The input to the tensor constructor must be a non-null value.")
+            }
+
+            function he(e, t = [], n = !1) {
+                if (null == t && (t = []), Array.isArray(e) || Te(e) && !n)
+                    for (let r = 0; r < e.length; ++r) he(e[r], t, n);
+                else t.push(e);
+                return t
+            }
+
+            function pe(e) {
+                if (0 === e.length) return 1;
+                let t = e[0];
+                for (let n = 1; n < e.length; n++) t *= e[n];
+                return t
+            }
+
+            function de(e) {
+                return 0 === e.length
+            }
+
+            function fe(e, t) {
+                if (e === t) return !0;
+                if (null == e || null == t) return !1;
+                if (e.length !== t.length) return !1;
+                for (let n = 0; n < e.length; n++)
+                    if (e[n] !== t[n]) return !1;
+                return !0
+            }
+
+            function me(e) {
+                return e % 1 == 0
+            }
+
+            function ge(e) {
+                if (null != Math.tanh) return Math.tanh(e);
+                if (e === 1 / 0) return 1;
+                if (e === -1 / 0) return -1; {
+                    const t = Math.exp(2 * e);
+                    return (t - 1) / (t + 1)
+                }
+            }
+
+            function ye(e) {
+                const t = Math.ceil(Math.sqrt(e));
+                return [t, Math.ceil(e / t)]
+            }
+
+            function be(e) {
+                const t = new Uint32Array(e);
+                for (let n = 0; n < e; ++n) t[n] = n;
+                return ee(t), t
+            }
+
+            function ve(e, t) {
+                return t <= e.length ? e : e + " ".repeat(t - e.length)
+            }
+
+            function xe(e, t = (e => 0), n) {
+                return new Promise((r, a) => {
+                    let s = 0;
+                    const i = () => {
+                        if (e()) return void r();
+                        s++;
+                        const o = t(s);
+                        null != n && s >= n ? a() : setTimeout(i, o)
+                    };
+                    i()
+                })
+            }
+
+            function we(e, t) {
+                let n = 1,
+                    r = -1;
+                for (let t = 0; t < e.length; ++t)
+                    if (e[t] >= 0) n *= e[t];
+                    else if (-1 === e[t]) {
+                    if (-1 !== r) throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${r} and dim ${t}`);
+                    r = t
+                } else if (e[t] < 0) throw Error(`Shapes can not be < 0. Found ${e[t]} at dim ${t}`);
+                if (-1 === r) {
+                    if (t > 0 && t !== n) throw Error(`Size(${t}) must match the product of shape ${e}`);
+                    return e
+                }
+                if (0 === n) throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);
+                if (t % n != 0) throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);
+                const a = e.slice();
+                return a[r] = t / n, a
+            }
+
+            function ke(e, t) {
+                const n = t.length;
+                return ue((e = null == e ? t.map((e, t) => t) : [].concat(e)).every(e => e >= -n && e < n), () => `All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`), ue(e.every(e => me(e)), () => "All values in axis param must be integers but got axis " + e), e.map(e => e < 0 ? n + e : e)
+            }
+
+            function Se(e, t) {
+                const n = [],
+                    r = [],
+                    a = null != t && Array.isArray(t) && 0 === t.length,
+                    s = null == t || a ? null : ke(t, e).sort();
+                let i = 0;
+                for (let t = 0; t < e.length; ++t) {
+                    if (null != s) {
+                        if (s[i] === t && 1 !== e[t]) throw new Error(`Can't squeeze axis ${t} since its dim '${e[t]}' is not 1`);
+                        (null == s[i] || s[i] > t) && 1 === e[t] && (n.push(e[t]), r.push(t)), s[i] <= t && i++
+                    }
+                    1 !== e[t] && (n.push(e[t]), r.push(t))
+                }
+                return {
+                    newShape: n,
+                    keptDims: r
+                }
+            }
+
+            function Ie(e, t) {
+                let n = null;
+                if (null == e || "float32" === e) n = new Float32Array(t);
+                else if ("int32" === e) n = new Int32Array(t);
+                else {
+                    if ("bool" !== e) throw new Error("Unknown data type " + e);
+                    n = new Uint8Array(t)
+                }
+                return n
+            }
+
+            function Ae(e, t) {
+                let n = null;
+                if (null == e || "float32" === e) n = new Float32Array(t);
+                else if ("int32" === e) n = new Int32Array(t);
+                else if ("bool" === e) n = new Uint8Array(t);
+                else {
+                    if ("string" !== e) throw new Error("Unknown data type " + e);
+                    n = new Array(t)
+                }
+                return n
+            }
+
+            function Ee(e, t) {
+                for (let n = 0; n < e.length; n++) {
+                    const r = e[n];
+                    if (isNaN(r) || !isFinite(r)) throw Error(`A tensor of type ${t} being uploaded contains ${r}.`)
+                }
+            }
+
+            function Ne(e) {
+                return "bool" === e || "complex64" === e || "float32" === e || "int32" === e || "string" === e
+            }
+
+            function Ce(e, t) {
+                return "complex64" !== t && (("float32" !== t || "complex64" === e) && (("int32" !== t || "float32" === e || "complex64" === e) && ("bool" !== t || "bool" !== e)))
+            }
+
+            function Te(e) {
+                return e instanceof Float32Array || e instanceof Int32Array || e instanceof Uint8Array || e instanceof Uint8ClampedArray
+            }
+
+            function Re(e) {
+                if ("float32" === e || "int32" === e) return 4;
+                if ("complex64" === e) return 8;
+                if ("bool" === e) return 1;
+                throw new Error("Unknown dtype " + e)
+            }
+
+            function _e(e) {
+                if (null == e) return 0;
+                let t = 0;
+                return e.forEach(e => t += e.length), t
+            }
+
+            function Fe(e) {
+                return "string" == typeof e || e instanceof String
+            }
+
+            function Me(e) {
+                return "boolean" == typeof e
+            }
+
+            function De(e) {
+                return "number" == typeof e
+            }
+
+            function Oe(e) {
+                return Array.isArray(e) ? Oe(e[0]) : e instanceof Float32Array ? "float32" : e instanceof Int32Array || e instanceof Uint8Array || e instanceof Uint8ClampedArray ? "int32" : De(e) ? "float32" : Fe(e) ? "string" : Me(e) ? "bool" : "float32"
+            }
+
+            function Le(e) {
+                return !!(e && e.constructor && e.call && e.apply)
+            }
+
+            function Pe(e, t) {
+                for (let n = t; n < e; ++n)
+                    if (e % n == 0) return n;
+                return e
+            }
+
+            function $e(e) {
+                const t = e.length;
+                if (t < 2) return [];
+                const n = new Array(t - 1);
+                n[t - 2] = e[t - 1];
+                for (let r = t - 3; r >= 0; --r) n[r] = n[r + 1] * e[r + 1];
+                return n
+            }
+
+            function Be(e, t, n = !1) {
+                if (0 === e.length) return t[0];
+                const r = e.reduce((e, t) => e * t) * (n ? 2 : 1);
+                if (0 === r) return [];
+                if (r !== t.length) throw new Error(`[${e}] does not match the input size ${t.length}${n?" for a complex tensor":""}.`);
+                return function e(t, n, r, a = !1) {
+                    const s = new Array;
+                    if (1 === n.length) {
+                        const e = n[0] * (a ? 2 : 1);
+                        for (let n = 0; n < e; n++) s[n] = r[t + n]
+                    } else {
+                        const i = n[0],
+                            o = n.slice(1),
+                            u = o.reduce((e, t) => e * t) * (a ? 2 : 1);
+                        for (let n = 0; n < i; n++) s[n] = e(t + n * u, o, r, a)
+                    }
+                    return s
+                }(0, e, t, n)
+            }
+
+            function ze(e, t) {
+                const n = Ue(e, t);
+                for (let e = 0; e < n.length; e++) n[e] = 1;
+                return n
+            }
+
+            function Ue(e, t) {
+                if (null == t || "float32" === t || "complex64" === t) return new Float32Array(e);
+                if ("int32" === t) return new Int32Array(e);
+                if ("bool" === t) return new Uint8Array(e);
+                throw new Error("Unknown data type " + t)
+            }
+
+            function We(e, t) {
+                const n = e.reduce((e, t) => e * t, 1);
+                if (null == t || "float32" === t) return Be(e, new Float32Array(n));
+                if ("int32" === t) return Be(e, new Int32Array(n));
+                if ("bool" === t) return Be(e, new Uint8Array(n));
+                throw new Error("Unknown data type " + t)
+            }
+
+            function Ve(e) {
+                e.forEach(t => {
+                    ue(Number.isInteger(t) && t >= 0, () => `Tensor must have a shape comprised of positive integers but got shape [${e}].`)
+                })
+            }
+
+            function je(e, t, n) {
+                if (0 === t) return 0;
+                if (1 === t) return e[0];
+                let r = e[e.length - 1];
+                for (let t = 0; t < e.length - 1; ++t) r += n[t] * e[t];
+                return r
+            }
+
+            function Ge(e, t, n) {
+                if (0 === t) return [];
+                if (1 === t) return [e];
+                const r = new Array(t);
+                for (let t = 0; t < r.length - 1; ++t) r[t] = Math.floor(e / n[t]), e -= r[t] * n[t];
+                return r[r.length - 1] = e, r
+            }
+
+            function He(e) {
+                return e && e.then && "function" == typeof e.then
+            }
+            var qe = class {
+                constructor(e) {
+                    this.global = e, this.flags = {}, this.flagRegistry = {}, this.urlFlags = {}, this.getQueryParams = Ke, this.populateURLFlags()
+                }
+                setPlatform(e, t) {
+                    null != this.platform && (Ye().getBool("IS_TEST") || Ye().getBool("PROD") || console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${t}.`)), this.platformName = e, this.platform = t
+                }
+                registerFlag(e, t, n) {
+                    if (this.flagRegistry[e] = {
+                            evaluationFn: t,
+                            setHook: n
+                        }, null != this.urlFlags[e]) {
+                        const t = this.urlFlags[e];
+                        Ye().getBool("IS_TEST") || Ye().getBool("PROD") || console.warn(`Setting feature override from URL ${e}: ${t}.`), this.set(e, t)
+                    }
+                }
+                async getAsync(e) {
+                    return e in this.flags || (this.flags[e] = await this.evaluateFlag(e)), this.flags[e]
+                }
+                get(e) {
+                    if (e in this.flags) return this.flags[e];
+                    const t = this.evaluateFlag(e);
+                    if (He(t)) throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);
+                    return this.flags[e] = t, this.flags[e]
+                }
+                getNumber(e) {
+                    return this.get(e)
+                }
+                getBool(e) {
+                    return this.get(e)
+                }
+                getFlags() {
+                    return this.flags
+                }
+                get features() {
+                    return this.flags
+                }
+                set(e, t) {
+                    if (null == this.flagRegistry[e]) throw new Error(`Cannot set flag ${e} as it has not been registered.`);
+                    this.flags[e] = t, null != this.flagRegistry[e].setHook && this.flagRegistry[e].setHook(t)
+                }
+                evaluateFlag(e) {
+                    if (null == this.flagRegistry[e]) throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);
+                    return this.flagRegistry[e].evaluationFn()
+                }
+                setFlags(e) {
+                    this.flags = Object.assign({}, e)
+                }
+                reset() {
+                    this.flags = {}, this.urlFlags = {}, this.populateURLFlags()
+                }
+                populateURLFlags() {
+                    if (void 0 === this.global || void 0 === this.global.location || void 0 === this.global.location.search) return;
+                    const e = this.getQueryParams(this.global.location.search);
+                    if ("tfjsflags" in e) {
+                        e.tfjsflags.split(",").forEach(e => {
+                            const [t, n] = e.split(":");
+                            this.urlFlags[t] = function (e, t) {
+                                if ("true" === (t = t.toLowerCase()) || "false" === t) return "true" === t;
+                                if ("" + +t === t) return +t;
+                                throw new Error(`Could not parse value flag value ${t} for flag ${e}.`)
+                            }(t, n)
+                        })
+                    }
+                }
+            };
+
+            function Ke(e) {
+                const t = {};
+                return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g, (e, ...n) => (function (e, t, n) {
+                    e[decodeURIComponent(t)] = decodeURIComponent(n || "")
+                }(t, n[0], n[1]), n.join("="))), t
+            }
+
+            function Ye() {
+                return Qe
+            }
+            var Xe, Qe = null;
+
+            function Ze() {
+                if (null == Xe) {
+                    let e;
+                    if ("undefined" != typeof window) e = window;
+                    else if (void 0 !== s) e = s;
+                    else if (void 0 !== r) e = r;
+                    else {
+                        if ("undefined" == typeof self) throw new Error("Could not find a global object");
+                        e = self
+                    }
+                    Xe = e
+                }
+                return Xe
+            }
+
+            function Je(e, t) {
+                const n = function () {
+                    const e = Ze();
+                    return null == e._tfGlobals && (e._tfGlobals = new Map), e._tfGlobals
+                }();
+                if (n.has(e)) return n.get(e); {
+                    const r = t();
+                    return n.set(e, r), n.get(e)
+                }
+            }
+            var et = "Abs",
+                tt = "Acos",
+                nt = "Acosh",
+                rt = "Add",
+                at = "AddN",
+                st = "All",
+                it = "Any",
+                ot = "ArgMax",
+                ut = "ArgMin",
+                lt = "Asin",
+                ct = "Asinh",
+                ht = "Atan",
+                pt = "Atanh",
+                dt = "Atan2",
+                ft = "AvgPool",
+                mt = "AvgPoolGrad",
+                gt = "AvgPool3D",
+                yt = "AvgPool3DGrad",
+                bt = "BatchMatMul",
+                vt = "BatchToSpaceND",
+                xt = "Bincount",
+                wt = "BroadcastTo",
+                kt = "BroadcastArgs",
+                St = "Cast",
+                It = "Ceil",
+                At = "ClipByValue",
+                Et = "Complex",
+                Nt = "ComplexAbs",
+                Ct = "Concat",
+                Tt = "Conv2D",
+                Rt = "Conv2DBackpropFilter",
+                _t = "Conv2DBackpropInput",
+                Ft = "Conv3D",
+                Mt = "Conv3DBackpropFilterV2",
+                Dt = "Conv3DBackpropInputV2",
+                Ot = "Cos",
+                Lt = "Cosh",
+                Pt = "Cumsum",
+                $t = "CropAndResize",
+                Bt = "DenseBincount",
+                zt = "DepthToSpace",
+                Ut = "DepthwiseConv2dNative",
+                Wt = "DepthwiseConv2dNativeBackpropFilter",
+                Vt = "DepthwiseConv2dNativeBackpropInput",
+                jt = "Diag",
+                Gt = "Dilation2D",
+                Ht = "Dilation2DBackpropInput",
+                qt = "Dilation2DBackpropFilter",
+                Kt = "RealDiv",
+                Yt = "Einsum",
+                Xt = "Elu",
+                Qt = "EluGrad",
+                Zt = "Erf",
+                Jt = "Equal",
+                en = "Exp",
+                tn = "ExpandDims",
+                nn = "Expm1",
+                rn = "FFT",
+                an = "Fill",
+                sn = "FlipLeftRight",
+                on = "Floor",
+                un = "FloorDiv",
+                ln = "FusedBatchNorm",
+                cn = "GatherV2",
+                hn = "GatherNd",
+                pn = "Greater",
+                dn = "GreaterEqual",
+                fn = "Identity",
+                mn = "IFFT",
+                gn = "Imag",
+                yn = "IsFinite",
+                bn = "IsInf",
+                vn = "IsNan",
+                xn = "LeakyRelu",
+                wn = "Less",
+                kn = "LessEqual",
+                Sn = "LinSpace",
+                In = "Log",
+                An = "Log1p",
+                En = "LogicalAnd",
+                Nn = "LogicalNot",
+                Cn = "LogicalOr",
+                Tn = "LogSoftmax",
+                Rn = "LRN",
+                _n = "LRNGrad",
+                Fn = "Max",
+                Mn = "Maximum",
+                Dn = "MaxPool",
+                On = "MaxPoolGrad",
+                Ln = "MaxPool3D",
+                Pn = "MaxPool3DGrad",
+                $n = "MaxPoolWithArgmax",
+                Bn = "Mean",
+                zn = "Min",
+                Un = "Minimum",
+                Wn = "MirrorPad",
+                Vn = "Mod",
+                jn = "Multinomial",
+                Gn = "Multiply",
+                Hn = "Neg",
+                qn = "NotEqual",
+                Kn = "NonMaxSuppressionV3",
+                Yn = "NonMaxSuppressionV4",
+                Xn = "NonMaxSuppressionV5",
+                Qn = "OnesLike",
+                Zn = "OneHot",
+                Jn = "Pack",
+                er = "PadV2",
+                tr = "Pool",
+                nr = "Pow",
+                rr = "Prelu",
+                ar = "Prod",
+                sr = "Range",
+                ir = "Real",
+                or = "Reciprocal",
+                ur = "Relu",
+                lr = "Reshape",
+                cr = "ResizeNearestNeighbor",
+                hr = "ResizeNearestNeighborGrad",
+                pr = "ResizeBilinear",
+                dr = "ResizeBilinearGrad",
+                fr = "Relu6",
+                mr = "Reverse",
+                gr = "Round",
+                yr = "Rsqrt",
+                br = "ScatterNd",
+                vr = "Select",
+                xr = "Selu",
+                wr = "Slice",
+                kr = "Sin",
+                Sr = "Sinh",
+                Ir = "Sign",
+                Ar = "Sigmoid",
+                Er = "Softplus",
+                Nr = "Sqrt",
+                Cr = "Sum",
+                Tr = "SpaceToBatchND",
+                Rr = "SplitV",
+                _r = "Softmax",
+                Fr = "SparseFillEmptyRows",
+                Mr = "SparseReshape",
+                Dr = "SparseSegmentMean",
+                Or = "SparseSegmentSum",
+                Lr = "SparseToDense",
+                Pr = "SquaredDifference",
+                $r = "Square",
+                Br = "StridedSlice",
+                zr = "StringNGrams",
+                Ur = "StringSplit",
+                Wr = "StringToHashBucketFast",
+                Vr = "Sub",
+                jr = "Tan",
+                Gr = "Tanh",
+                Hr = "Tile",
+                qr = "TopK",
+                Kr = "Transform",
+                Yr = "Transpose",
+                Xr = "Unique",
+                Qr = "Unpack",
+                Zr = "UnsortedSegmentSum",
+                Jr = "ZerosLike",
+                ea = "Step",
+                ta = "FromPixels",
+                na = "RotateWithOffset",
+                ra = "_FusedMatMul",
+                aa = "FusedConv2D",
+                sa = "FusedDepthwiseConv2D";
+
+            function ia(...e) {
+                Ye().getBool("IS_TEST") || Ye().getBool("PROD") || console.warn(...e)
+            }
+
+            function oa(...e) {
+                Ye().getBool("IS_TEST") || Ye().getBool("PROD") || console.log(...e)
+            }
+            var ua = Je("kernelRegistry", () => new Map),
+                la = Je("gradRegistry", () => new Map);
+
+            function ca(e, t) {
+                const n = ba(e, t);
+                return ua.get(n)
+            }
+
+            function ha(e) {
+                return la.get(e)
+            }
+
+            function pa(e) {
+                const t = ua.entries(),
+                    n = [];
+                for (;;) {
+                    const {
+                        done: r,
+                        value: a
+                    } = t.next();
+                    if (r) break;
+                    const [s, i] = a, [o] = s.split("_");
+                    o === e && n.push(i)
+                }
+                return n
+            }
+
+            function da(e) {
+                const {
+                    kernelName: t,
+                    backendName: n
+                } = e, r = ba(t, n);
+                ua.has(r) && ia(`The kernel '${t}' for backend '${n}' is already registered`), ua.set(r, e)
+            }
+
+            function fa(e) {
+                const {
+                    kernelName: t
+                } = e;
+                la.has(t) && Ye().getBool("DEBUG") && ia(`Overriding the gradient for '${t}'`), la.set(t, e)
+            }
+
+            function ma(e, t) {
+                const n = ba(e, t);
+                if (!ua.has(n)) throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);
+                ua.delete(n)
+            }
+
+            function ga(e) {
+                if (!la.has(e)) throw new Error(`The gradient '${e}' for backend is not registered`);
+                la.delete(e)
+            }
+
+            function ya(e, t) {
+                pa(e).forEach(e => {
+                    da(Object.assign({}, e, {
+                        backendName: t
+                    }))
+                })
+            }
+
+            function ba(e, t) {
+                return `${t}_${e}`
+            }
+            var va = {};
+            w(va, {
+                arraysEqual: () => fe,
+                assert: () => ue,
+                assertNonNegativeIntegerDimensions: () => Ve,
+                assertNonNull: () => ce,
+                assertShapesMatch: () => le,
+                bytesFromStringArray: () => _e,
+                bytesPerElement: () => Re,
+                checkConversionForErrors: () => Ee,
+                clamp: () => ne,
+                computeStrides: () => $e,
+                createScalarValue: () => Da,
+                createShuffledIndices: () => be,
+                decodeString: () => Ba,
+                distSquared: () => oe,
+                encodeString: () => $a,
+                fetch: () => Pa,
+                fingerPrint64: () => Ma,
+                flatten: () => he,
+                getArrayFromDType: () => Ae,
+                getTypedArrayFromDType: () => Ie,
+                hasEncodingLoss: () => Ce,
+                hexToLong: () => ka,
+                indexToLoc: () => Ge,
+                inferDtype: () => Oe,
+                inferFromImplicitShape: () => we,
+                isBoolean: () => Me,
+                isFunction: () => Le,
+                isInt: () => me,
+                isNumber: () => De,
+                isPromise: () => He,
+                isScalarShape: () => de,
+                isString: () => Fe,
+                isTypedArray: () => Te,
+                isValidDtype: () => Ne,
+                locToIndex: () => je,
+                makeOnesTypedArray: () => ze,
+                makeZerosNestedTypedArray: () => We,
+                makeZerosTypedArray: () => Ue,
+                nearestDivisor: () => Pe,
+                nearestLargerEven: () => re,
+                now: () => La,
+                parseAxisParam: () => ke,
+                randUniform: () => ie,
+                repeatedTry: () => xe,
+                rightPad: () => ve,
+                shuffle: () => ee,
+                shuffleCombo: () => te,
+                sizeFromShape: () => pe,
+                sizeToSquarishShape: () => ye,
+                squeezeShape: () => Se,
+                sum: () => se,
+                swap: () => ae,
+                tanh: () => ge,
+                toNestedArray: () => Be,
+                toTypedArray: () => Oa
+            });
+            var xa = k(S()),
+                wa = xa.default || xa;
+
+            function ka(e) {
+                return wa.fromString(e, !0, 16)
+            }
+            var Sa = ka("c3a5c85c97cb3127"),
+                Ia = ka("b492b66fbe98f273"),
+                Aa = ka("9ae16a3b2f90404f");
+
+            function Ea(e) {
+                return e.xor(e.shru(47))
+            }
+
+            function Na(e, t, n) {
+                const r = e.slice(t, t + n);
+                return wa.fromBytes(Array.from(r), !0, !0)
+            }
+
+            function Ca(e, t) {
+                return Na(e, t, 8)
+            }
+
+            function Ta(e, t) {
+                return Na(e, t, 4)
+            }
+
+            function Ra(e, t) {
+                return 0 === t ? e : e.shru(t).or(e.shl(64 - t))
+            }
+
+            function _a(e, t, n = ka("9ddfea08eb382d69")) {
+                let r = e.xor(t).mul(n);
+                r = r.xor(r.shru(47));
+                let a = t.xor(r).mul(n);
+                return a = a.xor(a.shru(47)), a = a.mul(n), a
+            }
+
+            function Fa(e, t, n, r) {
+                return function (e, t, n, r, a, s) {
+                    a = a.add(e), s = Ra(s.add(a).add(r), 21);
+                    const i = a;
+                    return a = (a = a.add(t)).add(n), s = s.add(Ra(a, 44)), [a.add(r), s.add(i)]
+                }(Ca(e, t), Ca(e, t + 8), Ca(e, t + 16), Ca(e, t + 24), n, r)
+            }
+
+            function Ma(e, t = e.length) {
+                const n = wa.fromNumber(81, !0);
+                if (t <= 32) return t <= 16 ? function (e, t = e.length) {
+                    if (t >= 8) {
+                        const n = Aa.add(2 * t),
+                            r = Ca(e, 0).add(Aa),
+                            a = Ca(e, t - 8);
+                        return _a(Ra(a, 37).mul(n).add(r), Ra(r, 25).add(a).mul(n), n)
+                    }
+                    if (t >= 4) {
+                        const n = Aa.add(2 * t);
+                        return _a(Ta(e, 0).shl(3).add(t), Ta(e, t - 4), n)
+                    }
+                    if (t > 0) {
+                        const n = e[0] + (e[t >> 1] << 8),
+                            r = t + (e[t - 1] << 2);
+                        return Ea(Aa.mul(n).xor(Sa.mul(r))).mul(Aa)
+                    }
+                    return Aa
+                }(e, t) : function (e, t = e.length) {
+                    const n = Aa.add(2 * t),
+                        r = Ca(e, 0).mul(Ia),
+                        a = Ca(e, 8),
+                        s = Ca(e, t - 8).mul(n),
+                        i = Ca(e, t - 16).mul(Aa);
+                    return _a(Ra(r.add(a), 43).add(Ra(s, 30)).add(i), r.add(Ra(a.add(Aa), 18)).add(s), n)
+                }(e, t);
+                if (t <= 64) return function (e, t = e.length) {
+                    const n = Aa.add(2 * t),
+                        r = Ca(e, 0).mul(Aa),
+                        a = Ca(e, 8),
+                        s = Ca(e, t - 8).mul(n),
+                        i = Ca(e, t - 16).mul(Aa),
+                        o = Ra(r.add(a), 43).add(Ra(s, 30)).add(i),
+                        u = _a(o, r.add(Ra(a.add(Aa), 18)).add(s), n),
+                        l = Ca(e, 16).mul(n),
+                        c = Ca(e, 24),
+                        h = o.add(Ca(e, t - 32)).mul(n),
+                        p = u.add(Ca(e, t - 24)).mul(n);
+                    return _a(Ra(l.add(c), 43).add(Ra(h, 30)).add(p), l.add(Ra(c.add(r), 18)).add(h), n)
+                }(e, t);
+                let r = n,
+                    a = n.mul(Ia).add(113),
+                    s = Ea(a.mul(Aa).add(113)).mul(Aa),
+                    i = [wa.UZERO, wa.UZERO],
+                    o = [wa.UZERO, wa.UZERO];
+                r = r.mul(Aa).add(Ca(e, 0));
+                let u = 0;
+                const l = 64 * (t - 1 >> 6),
+                    c = l + (t - 1 & 63) - 63;
+                do {
+                    r = Ra(r.add(a).add(i[0]).add(Ca(e, u + 8)), 37).mul(Ia), a = Ra(a.add(i[1]).add(Ca(e, u + 48)), 42).mul(Ia), r = r.xor(o[1]), a = a.add(i[0]).add(Ca(e, u + 40)), s = Ra(s.add(o[0]), 33).mul(Ia), i = Fa(e, u, i[1].mul(Ia), r.add(o[0])), o = Fa(e, u + 32, s.add(o[1]), a.add(Ca(e, u + 16))), [s, r] = [r, s], u += 64
+                } while (u !== l);
+                const h = Ia.add(s.and(255).shl(1));
+                return u = c, o[0] = o[0].add(t - 1 & 63), i[0] = i[0].add(o[0]), o[0] = o[0].add(i[0]), r = Ra(r.add(a).add(i[0]).add(Ca(e, u + 8)), 37).mul(h), a = Ra(a.add(i[1]).add(Ca(e, u + 48)), 42).mul(h), r = r.xor(o[1].mul(9)), a = a.add(i[0].mul(9).add(Ca(e, u + 40))), s = Ra(s.add(o[0]), 33).mul(h), i = Fa(e, u, i[1].mul(h), r.add(o[0])), o = Fa(e, u + 32, s.add(o[1]), a.add(Ca(e, u + 16))), [s, r] = [r, s], _a(_a(i[0], o[0], h).add(Ea(a).mul(Sa)).add(s), _a(i[1], o[1], h).add(r), h)
+            }
+
+            function Da(e, t) {
+                return "string" === t ? $a(e) : Oa([e], t)
+            }
+
+            function Oa(e, t) {
+                if ("string" === t) throw new Error("Cannot convert a string[] to a TypedArray");
+                if (Array.isArray(e) && (e = he(e)), Ye().getBool("DEBUG") && Ee(e, t), function (e, t) {
+                        return e instanceof Float32Array && "float32" === t || e instanceof Int32Array && "int32" === t || e instanceof Uint8Array && "bool" === t
+                    }(e, t)) return e;
+                if (null == t || "float32" === t || "complex64" === t) return new Float32Array(e);
+                if ("int32" === t) return new Int32Array(e);
+                if ("bool" === t) {
+                    const t = new Uint8Array(e.length);
+                    for (let n = 0; n < t.length; ++n) 0 !== Math.round(e[n]) && (t[n] = 1);
+                    return t
+                }
+                throw new Error("Unknown data type " + t)
+            }
+
+            function La() {
+                return Ye().platform.now()
+            }
+
+            function Pa(e, t) {
+                return Ye().platform.fetch(e, t)
+            }
+
+            function $a(e, t = "utf-8") {
+                return t = t || "utf-8", Ye().platform.encode(e, t)
+            }
+
+            function Ba(e, t = "utf-8") {
+                return t = t || "utf-8", Ye().platform.decode(e, t)
+            }
+
+            function za(e, t, n) {
+                if ("float32" !== t) return !1;
+                for (let t = 0; t < e.length; t++) {
+                    const r = e[t];
+                    if (isNaN(r) || !isFinite(r)) return console.warn(`Found ${r} in the result of '${n}'`), !0
+                }
+                return !1
+            }
+            var Ua = class {
+                logKernelProfile(e, t, n, r, a, s) {
+                    const i = "number" == typeof r ? ve(r + "ms", 9) : r.error,
+                        o = ve(e, 25),
+                        u = t.rank,
+                        l = t.size,
+                        c = ve(t.shape.toString(), 14);
+                    let h = "";
+                    for (const e in a) {
+                        const n = a[e];
+                        if (null != n) {
+                            const r = n.shape || t.shape,
+                                a = r.length;
+                            h += `${e}: ${a}D ${a>0?r:""} `
+                        }
+                    }
+                    console.log(`%c${o}\t%c${i}\t%c${u}D ${c}\t%c${l}\t%c${h}\t%c${s}`, "font-weight:bold", "color:red", "color:blue", "color: orange", "color: green", "color: steelblue")
+                }
+            };
+
+            function Wa(e, t, n, r) {
+                const a = $e(t),
+                    s = function (e, t, n, r) {
+                        const a = pe(t),
+                            s = r[r.length - 1],
+                            i = new Array(s).fill(0),
+                            o = t.length,
+                            u = "complex64" === n ? Ga(e) : e;
+                        if (o > 1)
+                            for (let e = 0; e < a / s; e++) {
+                                const t = e * s;
+                                for (let e = 0; e < s; e++) i[e] = Math.max(i[e], Va(u[t + e], 0, n).length)
+                            }
+                        return i
+                    }(e, t, n, a),
+                    i = t.length,
+                    o = function e(t, n, r, a, s, i = !0) {
+                        const o = "complex64" === r ? 2 : 1,
+                            u = n[0],
+                            l = n.length;
+                        if (0 === l) {
+                            if ("complex64" === r) {
+                                return [Va(Ga(t)[0], 0, r)]
+                            }
+                            return "bool" === r ? [ja(t[0])] : [t[0].toString()]
+                        }
+                        if (1 === l) {
+                            if (u > 20) {
+                                const e = 3 * o;
+                                let n = Array.from(t.slice(0, e)),
+                                    a = Array.from(t.slice((u - 3) * o, u * o));
+                                return "complex64" === r && (n = Ga(n), a = Ga(a)), ["[" + n.map((e, t) => Va(e, s[t], r)).join(", ") + ", ..., " + a.map((e, t) => Va(e, s[u - 3 + t], r)).join(", ") + "]"]
+                            }
+                            return ["[" + ("complex64" === r ? Ga(t) : Array.from(t)).map((e, t) => Va(e, s[t], r)).join(", ") + "]"]
+                        }
+                        const c = n.slice(1),
+                            h = a.slice(1),
+                            p = a[0] * o,
+                            d = [];
+                        if (u > 20) {
+                            for (let n = 0; n < 3; n++) {
+                                const a = n * p,
+                                    i = a + p;
+                                d.push(...e(t.slice(a, i), c, r, h, s, !1))
+                            }
+                            d.push("...");
+                            for (let n = u - 3; n < u; n++) {
+                                const a = n * p,
+                                    i = a + p;
+                                d.push(...e(t.slice(a, i), c, r, h, s, n === u - 1))
+                            }
+                        } else
+                            for (let n = 0; n < u; n++) {
+                                const a = n * p,
+                                    i = a + p;
+                                d.push(...e(t.slice(a, i), c, r, h, s, n === u - 1))
+                            }
+                        const f = 2 === l ? "," : "";
+                        d[0] = "[" + d[0] + f;
+                        for (let e = 1; e < d.length - 1; e++) d[e] = " " + d[e] + f;
+                        let m = ",\n";
+                        for (let e = 2; e < l; e++) m += "\n";
+                        return d[d.length - 1] = " " + d[d.length - 1] + "]" + (i ? "" : m), d
+                    }(e, t, n, a, s),
+                    u = ["Tensor"];
+                return r && (u.push("  dtype: " + n), u.push("  rank: " + i), u.push(`  shape: [${t}]`), u.push("  values:")), u.push(o.map(e => "    " + e).join("\n")), u.join("\n")
+            }
+
+            function Va(e, t, n) {
+                let r;
+                return r = Array.isArray(e) ? `${parseFloat(e[0].toFixed(7))} + ${parseFloat(e[1].toFixed(7))}j` : Fe(e) ? `'${e}'` : "bool" === n ? ja(e) : parseFloat(e.toFixed(7)).toString(), ve(r, t)
+            }
+
+            function ja(e) {
+                return 0 === e ? "false" : "true"
+            }
+
+            function Ga(e) {
+                const t = [];
+                for (let n = 0; n < e.length; n += 2) t.push([e[n], e[n + 1]]);
+                return t
+            }
+            var Ha = class {
+                    constructor(e, t, n) {
+                        if (this.dtype = t, this.shape = e.slice(), this.size = pe(e), null != n) {
+                            const e = n.length;
+                            ue(e === this.size, () => `Length of values '${e}' does not match the size inferred by the shape '${this.size}'.`)
+                        }
+                        if ("complex64" === t) throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");
+                        this.values = n || Ae(t, this.size), this.strides = $e(e)
+                    }
+                    set(e, ...t) {
+                        0 === t.length && (t = [0]), ue(t.length === this.rank, () => `The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);
+                        const n = this.locToIndex(t);
+                        this.values[n] = e
+                    }
+                    get(...e) {
+                        0 === e.length && (e = [0]);
+                        let t = 0;
+                        for (const n of e) {
+                            if (n < 0 || n >= this.shape[t]) {
+                                const t = `Requested out of range element at ${e}.   Buffer shape=${this.shape}`;
+                                throw new Error(t)
+                            }
+                            t++
+                        }
+                        let n = e[e.length - 1];
+                        for (let t = 0; t < e.length - 1; ++t) n += this.strides[t] * e[t];
+                        return this.values[n]
+                    }
+                    locToIndex(e) {
+                        if (0 === this.rank) return 0;
+                        if (1 === this.rank) return e[0];
+                        let t = e[e.length - 1];
+                        for (let n = 0; n < e.length - 1; ++n) t += this.strides[n] * e[n];
+                        return t
+                    }
+                    indexToLoc(e) {
+                        if (0 === this.rank) return [];
+                        if (1 === this.rank) return [e];
+                        const t = new Array(this.shape.length);
+                        for (let n = 0; n < t.length - 1; ++n) t[n] = Math.floor(e / this.strides[n]), e -= t[n] * this.strides[n];
+                        return t[t.length - 1] = e, t
+                    }
+                    get rank() {
+                        return this.shape.length
+                    }
+                    toTensor() {
+                        return qa().makeTensor(this.values, this.shape, this.dtype)
+                    }
+                },
+                qa = null,
+                Ka = null;
+            var Ya = class {
+                constructor(e, t, n, r) {
+                    this.kept = !1, this.isDisposedInternal = !1, this.shape = e.slice(), this.dtype = t || "float32", this.size = pe(e), this.strides = $e(e), this.dataId = n, this.id = r, this.rankType = this.rank < 5 ? this.rank.toString() : "higher"
+                }
+                get rank() {
+                    return this.shape.length
+                }
+                async buffer() {
+                    const e = await this.data();
+                    return Ka.buffer(this.shape, this.dtype, e)
+                }
+                bufferSync() {
+                    return Ka.buffer(this.shape, this.dtype, this.dataSync())
+                }
+                async array() {
+                    const e = await this.data();
+                    return Be(this.shape, e, "complex64" === this.dtype)
+                }
+                arraySync() {
+                    return Be(this.shape, this.dataSync(), "complex64" === this.dtype)
+                }
+                async data() {
+                    this.throwIfDisposed();
+                    const e = qa().read(this.dataId);
+                    if ("string" === this.dtype) {
+                        const t = await e;
+                        try {
+                            return t.map(e => Ba(e))
+                        } catch (e) {
+                            throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")
+                        }
+                    }
+                    return e
+                }
+                dataToGPU(e) {
+                    return this.throwIfDisposed(), qa().readToGPU(this.dataId, e)
+                }
+                dataSync() {
+                    this.throwIfDisposed();
+                    const e = qa().readSync(this.dataId);
+                    if ("string" === this.dtype) try {
+                        return e.map(e => Ba(e))
+                    } catch (e) {
+                        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")
+                    }
+                    return e
+                }
+                async bytes() {
+                    this.throwIfDisposed();
+                    const e = await qa().read(this.dataId);
+                    return "string" === this.dtype ? e : new Uint8Array(e.buffer)
+                }
+                dispose() {
+                    this.isDisposed || (qa().disposeTensor(this), this.isDisposedInternal = !0)
+                }
+                get isDisposed() {
+                    return this.isDisposedInternal
+                }
+                throwIfDisposed() {
+                    if (this.isDisposed) throw new Error("Tensor is disposed.")
+                }
+                print(e = !1) {
+                    return Ka.print(this, e)
+                }
+                clone() {
+                    return this.throwIfDisposed(), Ka.clone(this)
+                }
+                toString(e = !1) {
+                    return Wa(this.dataSync(), this.shape, this.dtype, e)
+                }
+                cast(e) {
+                    return this.throwIfDisposed(), Ka.cast(this, e)
+                }
+                variable(e = !0, t, n) {
+                    return this.throwIfDisposed(), qa().makeVariable(this, e, t, n)
+                }
+            };
+
+            function Xa() {
+                return Je("Tensor", () => Ya)
+            }
+            Object.defineProperty(Ya, Symbol.hasInstance, {
+                value: e => !!e && null != e.data && null != e.dataSync && null != e.throwIfDisposed
+            }), Xa();
+            var Qa = class extends Ya {
+                constructor(e, t, n, r) {
+                    super(e.shape, e.dtype, e.dataId, r), this.trainable = t, this.name = n
+                }
+                assign(e) {
+                    if (e.dtype !== this.dtype) throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);
+                    if (!fe(e.shape, this.shape)) throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);
+                    qa().disposeTensor(this), this.dataId = e.dataId, qa().incRef(this, null)
+                }
+                dispose() {
+                    qa().disposeVariable(this), this.isDisposedInternal = !0
+                }
+            };
+            Object.defineProperty(Qa, Symbol.hasInstance, {
+                value: e => e instanceof Ya && null != e.assign && e.assign instanceof Function
+            });
+            var Za, Ja, es, ts, ns, rs, as, ss, is, os, us = {};
+            w(us, {
+                assertTypesMatch: () => ds,
+                getTensorsInContainer: () => ms,
+                isTensorInList: () => fs,
+                makeTypesMatch: () => ps
+            }), (Ja = Za || (Za = {})).R0 = "R0", Ja.R1 = "R1", Ja.R2 = "R2", Ja.R3 = "R3", Ja.R4 = "R4", Ja.R5 = "R5", Ja.R6 = "R6", (ts = es || (es = {})).float32 = "float32", ts.int32 = "int32", ts.bool = "int32", ts.complex64 = "complex64", (rs = ns || (ns = {})).float32 = "float32", rs.int32 = "int32", rs.bool = "bool", rs.complex64 = "complex64", (ss = as || (as = {})).float32 = "float32", ss.int32 = "float32", ss.bool = "float32", ss.complex64 = "complex64", (os = is || (is = {})).float32 = "complex64", os.int32 = "complex64", os.bool = "complex64", os.complex64 = "complex64";
+            var ls = {
+                float32: as,
+                int32: es,
+                bool: ns,
+                complex64: is
+            };
+
+            function cs(e, t) {
+                if ("string" === e || "string" === t) {
+                    if ("string" === e && "string" === t) return "string";
+                    throw new Error(`Can not upcast ${e} with ${t}`)
+                }
+                return ls[e][t]
+            }
+
+            function hs(e) {
+                return cs(e, "int32")
+            }
+
+            function ps(e, t) {
+                if (e.dtype === t.dtype) return [e, t];
+                const n = cs(e.dtype, t.dtype);
+                return [e.cast(n), t.cast(n)]
+            }
+
+            function ds(e, t) {
+                ue(e.dtype === t.dtype, () => `The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)
+            }
+
+            function fs(e, t) {
+                return t.some(t => t.id === e.id)
+            }
+
+            function ms(e) {
+                const t = [];
+                return function e(t, n, r) {
+                    if (null == t) return;
+                    if (t instanceof Ya) return void n.push(t);
+                    if (a = t, !Array.isArray(a) && "object" != typeof a) return;
+                    var a;
+                    const s = t;
+                    for (const t in s) {
+                        const a = s[t];
+                        r.has(a) || (r.add(a), e(a, n, r))
+                    }
+                }(e, t, new Set), t
+            }
+
+            function gs(e) {
+                return null != e.kernelName
+            }
+            var ys = class {
+                    constructor() {
+                        this.registeredVariables = {}, this.nextTapeNodeId = 0, this.numBytes = 0, this.numTensors = 0, this.numStringTensors = 0, this.numDataBuffers = 0, this.gradientDepth = 0, this.kernelDepth = 0, this.scopeStack = [], this.numDataMovesStack = [], this.nextScopeId = 0, this.tensorInfo = new WeakMap, this.profiling = !1, this.activeProfile = {
+                            newBytes: 0,
+                            newTensors: 0,
+                            peakBytes: 0,
+                            kernels: [],
+                            result: null,
+                            get kernelNames() {
+                                return Array.from(new Set(this.kernels.map(e => e.name)))
+                            }
+                        }
+                    }
+                    dispose() {
+                        for (const e in this.registeredVariables) this.registeredVariables[e].dispose()
+                    }
+                },
+                bs = class {
+                    constructor(e) {
+                        this.ENV = e, this.registry = {}, this.registryFactory = {}, this.pendingBackendInitId = 0, this.state = new ys
+                    }
+                    async ready() {
+                        if (null != this.pendingBackendInit) return this.pendingBackendInit.then(() => {});
+                        if (null != this.backendInstance) return;
+                        const e = this.getSortedBackends();
+                        for (let t = 0; t < e.length; t++) {
+                            const n = e[t];
+                            if (await this.initializeBackend(n).success) return void await this.setBackend(n)
+                        }
+                        throw new Error("Could not initialize any backends, all backend initializations failed.")
+                    }
+                    get backend() {
+                        if (null != this.pendingBackendInit) throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);
+                        if (null == this.backendInstance) {
+                            const {
+                                name: e,
+                                asyncInit: t
+                            } = this.initializeBackendsAndReturnBest();
+                            if (t) throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);
+                            this.setBackend(e)
+                        }
+                        return this.backendInstance
+                    }
+                    backendNames() {
+                        return Object.keys(this.registryFactory)
+                    }
+                    findBackend(e) {
+                        if (!(e in this.registry)) {
+                            if (!(e in this.registryFactory)) return null; {
+                                const {
+                                    asyncInit: t
+                                } = this.initializeBackend(e);
+                                if (t) return null
+                            }
+                        }
+                        return this.registry[e]
+                    }
+                    findBackendFactory(e) {
+                        return e in this.registryFactory ? this.registryFactory[e].factory : null
+                    }
+                    registerBackend(e, t, n = 1) {
+                        return e in this.registryFactory ? (ia(e + " backend was already registered. Reusing existing backend factory."), !1) : (this.registryFactory[e] = {
+                            factory: t,
+                            priority: n
+                        }, !0)
+                    }
+                    async setBackend(e) {
+                        if (null == this.registryFactory[e]) throw new Error(`Backend name '${e}' not found in registry`);
+                        if (this.backendName = e, null == this.registry[e]) {
+                            this.backendInstance = null;
+                            const {
+                                success: t,
+                                asyncInit: n
+                            } = this.initializeBackend(e);
+                            if (!(n ? await t : t)) return !1
+                        }
+                        return this.backendInstance = this.registry[e], this.setupRegisteredKernels(), this.profiler = new class {
+                            constructor(e, t) {
+                                this.backendTimer = e, this.logger = t, null == t && (this.logger = new Ua)
+                            }
+                            profileKernel(e, t, n) {
+                                let r;
+                                const a = () => {
+                                    r = n()
+                                };
+                                let s;
+                                const i = La();
+                                if (this.backendTimer.timerAvailable()) s = this.backendTimer.time(a);
+                                else {
+                                    a();
+                                    for (const e of r) e.dataSync();
+                                    s = Promise.resolve({
+                                        kernelMs: La() - i
+                                    })
+                                }
+                                if (Ye().getBool("CHECK_COMPUTATION_FOR_ERRORS"))
+                                    for (let t = 0; t < r.length; t++) {
+                                        const n = r[t];
+                                        n.data().then(t => {
+                                            za(t, n.dtype, e)
+                                        })
+                                    }
+                                return {
+                                    kernelName: e,
+                                    outputs: r,
+                                    inputs: t,
+                                    timeMs: s.then(e => e.kernelMs),
+                                    extraInfo: s.then(e => null != e.getExtraProfileInfo ? e.getExtraProfileInfo() : "")
+                                }
+                            }
+                            logKernelProfile(e) {
+                                const {
+                                    kernelName: t,
+                                    outputs: n,
+                                    timeMs: r,
+                                    inputs: a,
+                                    extraInfo: s
+                                } = e;
+                                n.forEach(e => {
+                                    Promise.all([e.data(), r, s]).then(n => {
+                                        this.logger.logKernelProfile(t, e, n[0], n[1], a, n[2])
+                                    })
+                                })
+                            }
+                        }(this.backendInstance), !0
+                    }
+                    setupRegisteredKernels() {
+                        pa(this.backendName).forEach(e => {
+                            null != e.setupFunc && e.setupFunc(this.backendInstance)
+                        })
+                    }
+                    disposeRegisteredKernels(e) {
+                        pa(e).forEach(t => {
+                            null != t.disposeFunc && t.disposeFunc(this.registry[e])
+                        })
+                    }
+                    initializeBackend(e) {
+                        const t = this.registryFactory[e];
+                        if (null == t) throw new Error(`Cannot initialize backend ${e}, no registration found.`);
+                        try {
+                            const n = t.factory();
+                            if (!n || n instanceof Z || "function" != typeof n.then) return this.registry[e] = n, {
+                                success: !0,
+                                asyncInit: !1
+                            }; {
+                                const t = ++this.pendingBackendInitId,
+                                    r = n.then(n => !(t < this.pendingBackendInitId) && (this.registry[e] = n, this.pendingBackendInit = null, !0)).catch(n => (t < this.pendingBackendInitId || (this.pendingBackendInit = null, ia(`Initialization of backend ${e} failed`), ia(n.stack || n.message)), !1));
+                                return this.pendingBackendInit = r, {
+                                    success: r,
+                                    asyncInit: !0
+                                }
+                            }
+                        } catch (t) {
+                            return ia(`Initialization of backend ${e} failed`), ia(t.stack || t.message), {
+                                success: !1,
+                                asyncInit: !1
+                            }
+                        }
+                    }
+                    removeBackend(e) {
+                        if (!(e in this.registryFactory)) throw new Error(e + " backend not found in registry");
+                        this.backendName === e && null != this.pendingBackendInit && this.pendingBackendInitId++, e in this.registry && (this.disposeRegisteredKernels(e), this.registry[e].dispose(), delete this.registry[e]), delete this.registryFactory[e], this.backendName === e && (this.pendingBackendInit = null, this.backendName = null, this.backendInstance = null)
+                    }
+                    getSortedBackends() {
+                        if (0 === Object.keys(this.registryFactory).length) throw new Error("No backend found in registry.");
+                        return Object.keys(this.registryFactory).sort((e, t) => this.registryFactory[t].priority - this.registryFactory[e].priority)
+                    }
+                    initializeBackendsAndReturnBest() {
+                        const e = this.getSortedBackends();
+                        for (let t = 0; t < e.length; t++) {
+                            const n = e[t],
+                                {
+                                    success: r,
+                                    asyncInit: a
+                                } = this.initializeBackend(n);
+                            if (a || r) return {
+                                name: n,
+                                asyncInit: a
+                            }
+                        }
+                        throw new Error("Could not initialize any backends, all backend initializations failed.")
+                    }
+                    moveData(e, t) {
+                        const n = this.state.tensorInfo.get(t),
+                            r = n.backend,
+                            a = this.readSync(t),
+                            s = r.refCount(t);
+                        r.disposeData(t, !0), n.backend = e, e.move(t, a, n.shape, n.dtype, s), this.shouldCheckForMemLeaks() && this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1]++
+                    }
+                    tidy(e, t) {
+                        let n, r = null;
+                        if (null == t) {
+                            if ("function" != typeof e) throw new Error("Please provide a function to tidy()");
+                            t = e
+                        } else {
+                            if ("string" != typeof e && !(e instanceof String)) throw new Error("When calling with two arguments, the first argument to tidy() must be a string");
+                            if ("function" != typeof t) throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");
+                            r = e
+                        }
+                        return this.scopedRun(() => this.startScope(r), () => this.endScope(n), () => (n = t(), n instanceof Promise && console.error("Cannot return a Promise inside of tidy."), n))
+                    }
+                    scopedRun(e, t, n) {
+                        e();
+                        try {
+                            const e = n();
+                            return t(), e
+                        } catch (e) {
+                            throw t(), e
+                        }
+                    }
+                    nextTensorId() {
+                        return bs.nextTensorId++
+                    }
+                    nextVariableId() {
+                        return bs.nextVariableId++
+                    }
+                    clone(e) {
+                        const t = xs.runKernel(fn, {
+                                x: e
+                            }),
+                            n = {
+                                x: e
+                            };
+                        return this.addTapeNode(this.state.activeScope.name, n, [t], e => ({
+                            x: () => {
+                                const t = {
+                                        x: e
+                                    },
+                                    n = {
+                                        dtype: "float32"
+                                    };
+                                return xs.runKernel(St, t, n)
+                            }
+                        }), [], {}), t
+                    }
+                    runKernel(e, t, n) {
+                        if (!(null != ca(e, this.backendName))) throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);
+                        return this.runKernelFunc({
+                            kernelName: e,
+                            inputs: t,
+                            attrs: n
+                        })
+                    }
+                    shouldCheckForMemLeaks() {
+                        return this.ENV.getBool("IS_TEST")
+                    }
+                    checkKernelForMemLeak(e, t, n) {
+                        const r = this.backend.numDataIds();
+                        let a = 0;
+                        n.forEach(e => {
+                            a += "complex64" === e.dtype ? 3 : 1
+                        });
+                        const s = this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1],
+                            i = r - t - a - s;
+                        if (i > 0) throw new Error(`Backend '${this.backendName}' has an internal memory leak (${i} data ids) after running '${e}'`)
+                    }
+                    runKernelFunc(e) {
+                        let t, n = [];
+                        const r = this.isTapeOn(),
+                            a = this.state.numBytes,
+                            s = this.state.numTensors;
+                        let i, o;
+                        this.shouldCheckForMemLeaks() && this.state.numDataMovesStack.push(0);
+                        const u = gs(e) ? e.kernelName : null != this.state.activeScope ? this.state.activeScope.name : "";
+                        if (gs(e)) {
+                            const {
+                                kernelName: t,
+                                inputs: a,
+                                attrs: s
+                            } = e, u = ca(t, this.backendName);
+                            ue(null != u, () => `Cannot find registered kernel '${t}' for backend '${this.backendName}'`), i = () => {
+                                const e = this.backend.numDataIds();
+                                o = u.kernelFunc({
+                                    inputs: a,
+                                    attrs: s,
+                                    backend: this.backend
+                                });
+                                const i = Array.isArray(o) ? o : [o];
+                                this.shouldCheckForMemLeaks() && this.checkKernelForMemLeak(t, e, i);
+                                const l = i.map(e => {
+                                    if (null != e.rank) return e;
+                                    const {
+                                        dataId: t,
+                                        shape: n,
+                                        dtype: r
+                                    } = e;
+                                    return this.makeTensorFromDataId(t, n, r)
+                                });
+                                if (r) {
+                                    const e = this.getTensorsForGradient(t, a, l);
+                                    n = this.saveTensorsForBackwardMode(e)
+                                }
+                                return l
+                            }
+                        } else {
+                            const {
+                                forwardFunc: t
+                            } = e, a = e => {
+                                r && (n = e.map(e => this.keep(this.clone(e))))
+                            };
+                            i = () => {
+                                const e = this.backend.numDataIds();
+                                o = this.tidy(() => t(this.backend, a));
+                                const n = Array.isArray(o) ? o : [o];
+                                return this.shouldCheckForMemLeaks() && this.checkKernelForMemLeak(u, e, n), n
+                            }
+                        }
+                        const {
+                            inputs: l,
+                            attrs: c
+                        } = e, h = gs(e) ? null : e.backwardsFunc;
+                        let p;
+                        return this.scopedRun(() => this.state.kernelDepth++, () => this.state.kernelDepth--, () => {
+                            this.ENV.getBool("DEBUG") || this.state.profiling ? (p = this.profiler.profileKernel(u, l, () => i()), this.ENV.getBool("DEBUG") && this.profiler.logKernelProfile(p), t = p.outputs) : t = i()
+                        }), r && this.addTapeNode(u, l, t, h, n, c), this.state.profiling && this.state.activeProfile.kernels.push({
+                            name: u,
+                            bytesAdded: this.state.numBytes - a,
+                            totalBytesSnapshot: this.state.numBytes,
+                            tensorsAdded: this.state.numTensors - s,
+                            totalTensorsSnapshot: this.state.numTensors,
+                            inputShapes: Object.keys(l).map(e => null != l[e] ? l[e].shape : null),
+                            outputShapes: t.map(e => e.shape),
+                            kernelTimeMs: p.timeMs,
+                            extraInfo: p.extraInfo
+                        }), Array.isArray(o) ? t : t[0]
+                    }
+                    saveTensorsForBackwardMode(e) {
+                        return e.map(e => this.keep(this.clone(e)))
+                    }
+                    getTensorsForGradient(e, t, n) {
+                        const r = ha(e);
+                        if (null != r) {
+                            const e = r.inputsToSave || [],
+                                a = r.outputsToSave || [];
+                            let s;
+                            r.saveAllInputs ? (ue(Array.isArray(t), () => "saveAllInputs is true, expected inputs to be an array."), s = Object.keys(t).map(e => t[e])) : s = e.map(e => t[e]);
+                            const i = n.filter((e, t) => a[t]);
+                            return s.concat(i)
+                        }
+                        return []
+                    }
+                    makeTensor(e, t, n, r) {
+                        if (null == e) throw new Error("Values passed to engine.makeTensor() are null");
+                        n = n || "float32", r = r || this.backend;
+                        let a = e;
+                        "string" === n && Fe(e[0]) && (a = e.map(e => $a(e)));
+                        const s = r.write(a, t, n),
+                            i = new Ya(t, n, s, this.nextTensorId());
+                        if (this.trackTensor(i, r), "string" === n) {
+                            const e = this.state.tensorInfo.get(s),
+                                t = _e(a);
+                            this.state.numBytes += t - e.bytes, e.bytes = t
+                        }
+                        return i
+                    }
+                    makeTensorFromDataId(e, t, n, r) {
+                        const a = new Ya(t, n = n || "float32", e, this.nextTensorId());
+                        return this.trackTensor(a, r), a
+                    }
+                    makeVariable(e, t = !0, n, r) {
+                        n = n || this.nextVariableId().toString(), null != r && r !== e.dtype && (e = e.cast(r));
+                        const a = new Qa(e, t, n, this.nextTensorId());
+                        if (null != this.state.registeredVariables[a.name]) throw new Error(`Variable with name ${a.name} was already registered`);
+                        return this.state.registeredVariables[a.name] = a, this.incRef(a, this.backend), a
+                    }
+                    trackTensor(e, t) {
+                        this.state.numTensors++, "string" === e.dtype && this.state.numStringTensors++;
+                        let n = 0;
+                        "complex64" !== e.dtype && "string" !== e.dtype && (n = e.size * Re(e.dtype)), this.state.numBytes += n, this.state.tensorInfo.has(e.dataId) || (this.state.numDataBuffers++, this.state.tensorInfo.set(e.dataId, {
+                            backend: t || this.backend,
+                            dtype: e.dtype,
+                            shape: e.shape,
+                            bytes: n
+                        })), e instanceof Qa || this.track(e)
+                    }
+                    incRef(e, t) {
+                        this.trackTensor(e, t), this.backend.incRef(e.dataId)
+                    }
+                    removeDataId(e, t) {
+                        this.state.tensorInfo.has(e) && this.state.tensorInfo.get(e).backend === t && (this.state.tensorInfo.delete(e), this.state.numDataBuffers--)
+                    }
+                    disposeTensor(e) {
+                        if (!this.state.tensorInfo.has(e.dataId)) return;
+                        const t = this.state.tensorInfo.get(e.dataId);
+                        if (this.state.numTensors--, "string" === e.dtype && (this.state.numStringTensors--, this.state.numBytes -= t.bytes), "complex64" !== e.dtype && "string" !== e.dtype) {
+                            const t = e.size * Re(e.dtype);
+                            this.state.numBytes -= t
+                        }
+                        t.backend.disposeData(e.dataId) && this.removeDataId(e.dataId, t.backend)
+                    }
+                    disposeVariables() {
+                        for (const e in this.state.registeredVariables) {
+                            const t = this.state.registeredVariables[e];
+                            this.disposeVariable(t)
+                        }
+                    }
+                    disposeVariable(e) {
+                        this.disposeTensor(e), null != this.state.registeredVariables[e.name] && delete this.state.registeredVariables[e.name]
+                    }
+                    memory() {
+                        const e = this.backend.memory();
+                        return e.numTensors = this.state.numTensors, e.numDataBuffers = this.state.numDataBuffers, e.numBytes = this.state.numBytes, this.state.numStringTensors > 0 && (e.unreliable = !0, null == e.reasons && (e.reasons = []), e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")), e
+                    }
+                    async profile(e) {
+                        this.state.profiling = !0;
+                        const t = this.state.numBytes,
+                            n = this.state.numTensors;
+                        this.state.activeProfile.kernels = [], this.state.activeProfile.result = await e(), this.state.profiling = !1, this.state.activeProfile.peakBytes = Math.max(...this.state.activeProfile.kernels.map(e => e.totalBytesSnapshot)), this.state.activeProfile.newBytes = this.state.numBytes - t, this.state.activeProfile.newTensors = this.state.numTensors - n;
+                        for (const e of this.state.activeProfile.kernels) e.kernelTimeMs = await e.kernelTimeMs, e.extraInfo = await e.extraInfo;
+                        return this.state.activeProfile
+                    }
+                    isTapeOn() {
+                        return this.state.gradientDepth > 0 && 0 === this.state.kernelDepth
+                    }
+                    addTapeNode(e, t, n, r, a, s) {
+                        const i = {
+                                id: this.state.nextTapeNodeId++,
+                                kernelName: e,
+                                inputs: t,
+                                outputs: n,
+                                saved: a
+                            },
+                            o = ha(e);
+                        null != o && (r = o.gradFunc), null != r && (i.gradient = e => (e = e.map((e, t) => {
+                            if (null == e) {
+                                const e = n[t],
+                                    r = Ue(e.size, e.dtype);
+                                return this.makeTensor(r, e.shape, e.dtype)
+                            }
+                            return e
+                        }), r(e.length > 1 ? e : e[0], a, s))), this.state.activeTape.push(i)
+                    }
+                    keep(e) {
+                        return e.kept = !0, e
+                    }
+                    startTape() {
+                        0 === this.state.gradientDepth && (this.state.activeTape = []), this.state.gradientDepth++
+                    }
+                    endTape() {
+                        this.state.gradientDepth--
+                    }
+                    startScope(e) {
+                        const t = {
+                            track: [],
+                            name: "unnamed scope",
+                            id: this.state.nextScopeId++
+                        };
+                        e && (t.name = e), this.state.scopeStack.push(t), this.state.activeScope = t
+                    }
+                    endScope(e) {
+                        const t = ms(e),
+                            n = new Set(t.map(e => e.id));
+                        for (let e = 0; e < this.state.activeScope.track.length; e++) {
+                            const t = this.state.activeScope.track[e];
+                            t.kept || n.has(t.id) || t.dispose()
+                        }
+                        const r = this.state.scopeStack.pop();
+                        this.state.activeScope = 0 === this.state.scopeStack.length ? null : this.state.scopeStack[this.state.scopeStack.length - 1], t.forEach(e => {
+                            e.kept || e.scopeId !== r.id || this.track(e)
+                        })
+                    }
+                    gradients(e, t, n, r = !1) {
+                        if (ue(t.length > 0, () => "gradients() received an empty list of xs."), null != n && "float32" !== n.dtype) throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);
+                        const a = this.scopedRun(() => this.startTape(), () => this.endTape(), () => this.tidy("forward", e));
+                        ue(a instanceof Ya, () => "The result y returned by f() must be a tensor.");
+                        const s = function (e, t, n) {
+                            const r = {},
+                                a = {};
+                            for (let e = 0; e < t.length; e++) r[t[e].id] = !0;
+                            for (let n = 0; n < e.length; n++) {
+                                const s = e[n],
+                                    i = s.inputs;
+                                for (const e in i) {
+                                    const n = i[e];
+                                    let o = !1;
+                                    for (let e = 0; e < t.length; e++)
+                                        if (r[n.id]) {
+                                            s.outputs.forEach(e => r[e.id] = !0), o = !0, a[s.id] = !0;
+                                            break
+                                        } if (o) break
+                                }
+                            }
+                            const s = {};
+                            s[n.id] = !0;
+                            const i = {};
+                            for (let t = e.length - 1; t >= 0; t--) {
+                                const n = e[t],
+                                    r = n.inputs;
+                                for (let e = 0; e < n.outputs.length; e++)
+                                    if (s[n.outputs[e].id]) {
+                                        for (const e in r) s[r[e].id] = !0, i[n.id] = !0;
+                                        break
+                                    }
+                            }
+                            const o = [];
+                            for (let t = 0; t < e.length; t++) {
+                                const n = e[t];
+                                if (a[n.id] && i[n.id]) {
+                                    const e = {};
+                                    for (const t in n.inputs) {
+                                        const a = n.inputs[t];
+                                        r[a.id] && (e[t] = a)
+                                    }
+                                    const t = Object.assign({}, n);
+                                    t.inputs = e, t.outputs = n.outputs, o.push(t)
+                                }
+                            }
+                            return o
+                        }(this.state.activeTape, t, a);
+                        if (!r && 0 === s.length && t.length > 0) throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");
+                        return this.tidy("backward", () => {
+                            const e = {};
+                            e[a.id] = null == n ? function (e) {
+                                    const t = ze(pe(e), "float32");
+                                    return xs.makeTensor(t, e, "float32")
+                                }(a.shape) : n,
+                                function (e, t, n, r) {
+                                    for (let a = t.length - 1; a >= 0; a--) {
+                                        const s = t[a],
+                                            i = [];
+                                        if (s.outputs.forEach(t => {
+                                                const n = e[t.id];
+                                                null != n ? i.push(n) : i.push(null)
+                                            }), null == s.gradient) throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);
+                                        const o = s.gradient(i);
+                                        for (const t in s.inputs) {
+                                            if (!(t in o)) throw new Error(`Cannot backprop through input ${t}. Available gradients found: ${Object.keys(o)}.`);
+                                            const a = n(() => o[t]());
+                                            if ("float32" !== a.dtype) throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${t} must have 'float32' dtype, but has '${a.dtype}'`);
+                                            const i = s.inputs[t];
+                                            if (!fe(a.shape, i.shape)) throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${t}' has shape '${a.shape}', which does not match the shape of the input '${i.shape}'`);
+                                            if (null == e[i.id]) e[i.id] = a;
+                                            else {
+                                                const t = e[i.id];
+                                                e[i.id] = r(t, a), t.dispose()
+                                            }
+                                        }
+                                    }
+                                }(e, s, e => this.tidy(e), ws);
+                            const r = t.map(t => e[t.id]);
+                            return 0 === this.state.gradientDepth && (this.state.activeTape.forEach(e => {
+                                for (const t of e.saved) t.dispose()
+                            }), this.state.activeTape = null), {
+                                value: a,
+                                grads: r
+                            }
+                        })
+                    }
+                    customGrad(e) {
+                        return ue(Le(e), () => "The f passed in customGrad(f) must be a function."), (...t) => {
+                            let n;
+                            ue(t.every(e => e instanceof Ya), () => "The args passed in customGrad(f)(x1, x2,...) must all be tensors");
+                            const r = {};
+                            t.forEach((e, t) => {
+                                r[t] = e
+                            });
+                            return this.runKernelFunc({
+                                forwardFunc: (r, a) => (n = e(...t, a), ue(n.value instanceof Ya, () => "The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"), ue(Le(n.gradFunc), () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."), n.value),
+                                backwardsFunc: (e, r) => {
+                                    const a = n.gradFunc(e, r),
+                                        s = Array.isArray(a) ? a : [a];
+                                    ue(s.length === t.length, () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."), ue(s.every(e => e instanceof Ya), () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");
+                                    const i = {};
+                                    return s.forEach((e, t) => {
+                                        i[t] = () => e
+                                    }), i
+                                },
+                                inputs: r
+                            })
+                        }
+                    }
+                    readSync(e) {
+                        return this.state.tensorInfo.get(e).backend.readSync(e)
+                    }
+                    read(e) {
+                        return this.state.tensorInfo.get(e).backend.read(e)
+                    }
+                    readToGPU(e, t) {
+                        return this.state.tensorInfo.get(e).backend.readToGPU(e, t)
+                    }
+                    async time(e) {
+                        const t = La(),
+                            n = await this.backend.time(e);
+                        return n.wallMs = La() - t, n
+                    }
+                    track(e) {
+                        return null != this.state.activeScope && (e.scopeId = this.state.activeScope.id, this.state.activeScope.track.push(e)), e
+                    }
+                    get registeredVariables() {
+                        return this.state.registeredVariables
+                    }
+                    reset() {
+                        this.pendingBackendInitId++, this.state.dispose(), this.ENV.reset(), this.state = new ys;
+                        for (const e in this.registry) this.disposeRegisteredKernels(e), this.registry[e].dispose(), delete this.registry[e];
+                        this.backendName = null, this.backendInstance = null, this.pendingBackendInit = null
+                    }
+                };
+
+            function vs() {
+                const e = Ze();
+                if (null == e._tfengine) {
+                    const t = new qe(e);
+                    e._tfengine = new bs(t)
+                }
+                var t;
+                return t = e._tfengine.ENV, Qe = t, qa = () => e._tfengine, e._tfengine
+            }
+            bs.nextTensorId = 0, bs.nextVariableId = 0;
+            var xs = vs();
+
+            function ws(e, t) {
+                const n = {
+                    a: e,
+                    b: t
+                };
+                return xs.runKernel(rt, n)
+            }
+            var ks, Ss = {};
+
+            function Is(e) {
+                ks = e
+            }
+
+            function As(e) {
+                if (void 0 !== ks) return ks;
+                if (e || "undefined" != typeof navigator && null != navigator) {
+                    if (e || (e = navigator), "ReactNative" === e.product) return !0;
+                    const t = e.userAgent || e.vendor || ("undefined" != typeof window ? window.opera : "");
+                    if (!t) {
+                        const t = e;
+                        return t.userAgentData && t.userAgentData.mobile
+                    }
+                    return /(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t) || /1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0, 4))
+                }
+                return !1
+            }
+
+            function Es() {
+                return "undefined" != typeof window && null != window.document || "undefined" != typeof WorkerGlobalScope
+            }
+            w(Ss, {
+                isBrowser: () => Es,
+                isMobile: () => As,
+                mockIsMobile: () => Is
+            });
+            var Ns = Ye();
+
+            function Cs(e, t) {
+                let n = e;
+                if (Te(e)) return "string" === t ? [] : [e.length];
+                if (!Array.isArray(e)) return [];
+                const r = [];
+                for (; Array.isArray(n) || Te(n) && "string" !== t;) r.push(n.length), n = n[0];
+                return Array.isArray(e) && Ye().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY") && function e(t, n, r) {
+                    if (r = r || [], !Array.isArray(t) && !Te(t)) return void ue(0 === n.length, () => `Element arr[${r.join("][")}] is a primitive, but should be an array/TypedArray of ${n[0]} elements`);
+                    ue(n.length > 0, () => `Element arr[${r.join("][")}] should be a primitive, but is an array of ${t.length} elements`), ue(t.length === n[0], () => `Element arr[${r.join("][")}] should have ${n[0]} elements, but has ${t.length} elements`);
+                    const a = n.slice(1);
+                    for (let n = 0; n < t.length; ++n) e(t[n], a, r.concat(n))
+                }(e, r, []), r
+            }
+
+            function Ts(e, t, n, r) {
+                if ("string_or_numeric" !== e) {
+                    if (null == e) throw new Error("Expected dtype cannot be null.");
+                    if ("numeric" !== e && e !== t || "numeric" === e && "string" === t) throw new Error(`Argument '${n}' passed to '${r}' must be ${e} tensor, but got ${t} tensor`)
+                }
+            }
+
+            function Rs(e, t, n, r = "numeric") {
+                if (e instanceof Ya) return Ts(r, e.dtype, t, n), e;
+                let a = Oe(e);
+                if ("string" !== a && ["bool", "int32", "float32"].indexOf(r) >= 0 && (a = r), Ts(r, a, t, n), null == e || !Te(e) && !Array.isArray(e) && "number" != typeof e && "boolean" != typeof e && "string" != typeof e) {
+                    const r = null == e ? "null" : e.constructor.name;
+                    throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${r}'`)
+                }
+                const s = Cs(e, a);
+                Te(e) || Array.isArray(e) || (e = [e]);
+                const i = "string" !== a ? Oa(e, a) : he(e, [], !0);
+                return xs.makeTensor(i, s, a)
+            }
+
+            function _s(e, t, n, r = "numeric") {
+                if (!Array.isArray(e)) throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);
+                return e.map((e, a) => Rs(e, `${t}[${a}]`, n, r))
+            }
+            Ns.registerFlag("DEBUG", () => !1, e => {
+                e && console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")
+            }), Ns.registerFlag("IS_BROWSER", () => Es()), Ns.registerFlag("IS_NODE", () => void 0 !== r && void 0 !== r.versions && void 0 !== r.versions.node), Ns.registerFlag("IS_CHROME", () => "undefined" != typeof navigator && null != navigator && null != navigator.userAgent && /Chrome/.test(navigator.userAgent) && /Google Inc/.test(navigator.vendor)), Ns.registerFlag("PROD", () => !1), Ns.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY", () => Ns.getBool("DEBUG")), Ns.registerFlag("DEPRECATION_WARNINGS_ENABLED", () => !0), Ns.registerFlag("IS_TEST", () => !1), Ns.registerFlag("CHECK_COMPUTATION_FOR_ERRORS", () => !0), Ns.registerFlag("WRAP_TO_IMAGEBITMAP", () => !1);
+            var Fs = "__op";
+
+            function Ms(e) {
+                const t = Object.keys(e);
+                if (1 !== t.length) throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);
+                let n = t[0];
+                const r = e[n];
+                n.endsWith("_") && (n = n.substring(0, n.length - 1)), n += Fs;
+                const a = (...e) => {
+                    xs.startScope(n);
+                    try {
+                        const t = r(...e);
+                        return He(t) && console.error("Cannot return a Promise inside of tidy."), xs.endScope(t), t
+                    } catch (e) {
+                        throw xs.endScope(null), e
+                    }
+                };
+                return Object.defineProperty(a, "name", {
+                    value: n,
+                    configurable: !0
+                }), a
+            }
+            var Ds = Ms({
+                complex_: function (e, t) {
+                    const n = Rs(e, "real", "complex"),
+                        r = Rs(t, "imag", "complex");
+                    le(n.shape, r.shape, `real and imag shapes, ${n.shape} and ${r.shape}, must match in call to tf.complex().`);
+                    const a = {
+                        real: n,
+                        imag: r
+                    };
+                    return xs.runKernel(Et, a)
+                }
+            });
+
+            function Os(e, t, n, r) {
+                if (null == r && (r = Oe(e)), "complex64" === r) throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");
+                if (!Te(e) && !Array.isArray(e) && "number" != typeof e && "boolean" != typeof e && "string" != typeof e) throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");
+                if (null != t) {
+                    Ve(t);
+                    const e = pe(t),
+                        r = pe(n);
+                    ue(e === r, () => `Based on the provided shape, [${t}], the tensor should have ${e} values but has ${r}`);
+                    for (let e = 0; e < n.length; ++e) {
+                        const r = n[e],
+                            a = e !== n.length - 1 || r !== pe(t.slice(e));
+                        ue(n[e] === t[e] || !a, () => `Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)
+                    }
+                }
+                return Te(e) || Array.isArray(e) || (e = [e]), t = t || n, e = "string" !== r ? Oa(e, r) : he(e, [], !0), xs.makeTensor(e, t, r)
+            }
+
+            function Ls(e, t, n) {
+                return Os(e, t, Cs(e, n), n)
+            }
+            var Ps = {
+                float32: 4,
+                float16: 2,
+                int32: 4,
+                uint16: 2,
+                uint8: 1,
+                bool: 1,
+                complex64: 8
+            };
+            async function $s(e, t) {
+                const n = [],
+                    r = [],
+                    a = Array.isArray(e) ? e.map(e => e.name) : Object.keys(e);
+                for (let s = 0; s < a.length; ++s) {
+                    const i = a[s],
+                        o = Array.isArray(e) ? e[s].tensor : e[i];
+                    if ("float32" !== o.dtype && "int32" !== o.dtype && "bool" !== o.dtype && "string" !== o.dtype && "complex64" !== o.dtype) throw new Error(`Unsupported dtype in weight '${i}': ${o.dtype}`);
+                    const u = {
+                        name: i,
+                        shape: o.shape,
+                        dtype: o.dtype
+                    };
+                    if ("string" === o.dtype) {
+                        const e = new Promise(async e => {
+                            const t = await o.bytes(),
+                                n = t.reduce((e, t) => e + t.length, 0) + 4 * t.length,
+                                r = new Uint8Array(n);
+                            let a = 0;
+                            for (let e = 0; e < t.length; e++) {
+                                const n = t[e],
+                                    s = new Uint8Array(new Uint32Array([n.length]).buffer);
+                                r.set(s, a), a += 4, r.set(n, a), a += n.length
+                            }
+                            e(r)
+                        });
+                        r.push(e)
+                    } else r.push(o.data());
+                    null != t && (u.group = t), n.push(u)
+                }
+                return {
+                    data: zs(await Promise.all(r)),
+                    specs: n
+                }
+            }
+
+            function Bs(e, t) {
+                const n = {};
+                let r, a = 0;
+                for (const s of t) {
+                    const t = s.name,
+                        i = s.dtype,
+                        o = s.shape,
+                        u = pe(o);
+                    let l;
+                    if ("quantization" in s) {
+                        const n = s.quantization;
+                        if ("uint8" === n.dtype || "uint16" === n.dtype) {
+                            if (!("min" in n) || !("scale" in n)) throw new Error(`Weight ${s.name} with quantization ${n.dtype} doesn't have corresponding metadata min and scale.`)
+                        } else {
+                            if ("float16" !== n.dtype) throw new Error(`Weight ${s.name} has unknown quantization dtype ${n.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);
+                            if ("float32" !== i) throw new Error(`Weight ${s.name} is quantized with ${n.dtype} which only supports weights of type float32 not ${i}.`)
+                        }
+                        const o = Ps[n.dtype],
+                            c = e.slice(a, a + u * o),
+                            h = "uint8" === n.dtype ? new Uint8Array(c) : new Uint16Array(c);
+                        if ("float32" === i)
+                            if ("uint8" === n.dtype || "uint16" === n.dtype) {
+                                l = new Float32Array(h.length);
+                                for (let e = 0; e < h.length; e++) {
+                                    const t = h[e];
+                                    l[e] = t * n.scale + n.min
+                                }
+                            } else {
+                                if ("float16" !== n.dtype) throw new Error(`Unsupported quantization type ${n.dtype} for weight type float32.`);
+                                void 0 === r && (r = Ks()), l = r(h)
+                            }
+                        else {
+                            if ("int32" !== i) throw new Error(`Unsupported dtype in weight '${t}': ${i}`);
+                            if ("uint8" !== n.dtype && "uint16" !== n.dtype) throw new Error(`Unsupported quantization type ${n.dtype} for weight type int32.`);
+                            l = new Int32Array(h.length);
+                            for (let e = 0; e < h.length; e++) {
+                                const t = h[e];
+                                l[e] = Math.round(t * n.scale + n.min)
+                            }
+                        }
+                        a += u * o
+                    } else if ("string" === i) {
+                        const t = pe(s.shape);
+                        l = [];
+                        for (let n = 0; n < t; n++) {
+                            const t = new Uint32Array(e.slice(a, a + 4))[0];
+                            a += 4;
+                            const n = new Uint8Array(e.slice(a, a + t));
+                            l.push(n), a += t
+                        }
+                    } else {
+                        const r = Ps[i],
+                            s = e.slice(a, a + u * r);
+                        if ("float32" === i) l = new Float32Array(s);
+                        else if ("int32" === i) l = new Int32Array(s);
+                        else if ("bool" === i) l = new Uint8Array(s);
+                        else {
+                            if ("complex64" !== i) throw new Error(`Unsupported dtype in weight '${t}': ${i}`); {
+                                l = new Float32Array(s);
+                                const e = new Float32Array(l.length / 2),
+                                    r = new Float32Array(l.length / 2);
+                                for (let t = 0; t < e.length; t++) e[t] = l[2 * t], r[t] = l[2 * t + 1];
+                                const a = Ls(e, o, "float32"),
+                                    i = Ls(r, o, "float32");
+                                n[t] = Ds(a, i), a.dispose(), i.dispose()
+                            }
+                        }
+                        a += u * r
+                    }
+                    "complex64" !== i && (n[t] = Ls(l, o, i))
+                }
+                return n
+            }
+
+            function zs(e) {
+                if (null === e) throw new Error("Invalid input value: " + JSON.stringify(e));
+                let t = 0;
+                const n = [];
+                e.forEach(e => {
+                    if (t += e.byteLength, n.push(e.byteLength === e.buffer.byteLength ? e : new e.constructor(e)), !(e instanceof Float32Array || e instanceof Int32Array || e instanceof Uint8Array)) throw new Error("Unsupported TypedArray subtype: " + e.constructor.name)
+                });
+                const r = new Uint8Array(t);
+                let a = 0;
+                return n.forEach(e => {
+                    r.set(new Uint8Array(e.buffer), a), a += e.byteLength
+                }), r.buffer
+            }
+            var Us = void 0 !== i && ("undefined" == typeof Blob || "undefined" == typeof atob || "undefined" == typeof btoa);
+
+            function Ws(e) {
+                return Us ? i.byteLength(e) : new Blob([e]).size
+            }
+
+            function Vs(e) {
+                if (1 === e.length) return e[0];
+                let t = 0;
+                e.forEach(e => {
+                    t += e.byteLength
+                });
+                const n = new Uint8Array(t);
+                let r = 0;
+                return e.forEach(e => {
+                    n.set(new Uint8Array(e), r), r += e.byteLength
+                }), n.buffer
+            }
+
+            function js(e) {
+                for (e = e.trim(); e.endsWith("/");) e = e.slice(0, e.length - 1);
+                const t = e.split("/");
+                return t[t.length - 1]
+            }
+
+            function Gs(e, t) {
+                const n = {
+                    modelTopology: e.modelTopology,
+                    format: e.format,
+                    generatedBy: e.generatedBy,
+                    convertedBy: e.convertedBy,
+                    weightsManifest: t
+                };
+                return null != e.signature && (n.signature = e.signature), null != e.userDefinedMetadata && (n.userDefinedMetadata = e.userDefinedMetadata), null != e.modelInitializer && (n.modelInitializer = e.modelInitializer), null != e.trainingConfig && (n.trainingConfig = e.trainingConfig), n
+            }
+            async function Hs(e, t) {
+                const n = {
+                    modelTopology: e.modelTopology,
+                    format: e.format,
+                    generatedBy: e.generatedBy,
+                    convertedBy: e.convertedBy
+                };
+                if (null != e.trainingConfig && (n.trainingConfig = e.trainingConfig), null != e.weightsManifest) {
+                    const [r, a] = await t(e.weightsManifest);
+                    n.weightSpecs = r, n.weightData = a
+                }
+                return null != e.signature && (n.signature = e.signature), null != e.userDefinedMetadata && (n.userDefinedMetadata = e.userDefinedMetadata), null != e.modelInitializer && (n.modelInitializer = e.modelInitializer), n
+            }
+
+            function qs(e) {
+                if (e.modelTopology instanceof ArrayBuffer) throw new Error("Expected JSON model topology, received ArrayBuffer.");
+                return {
+                    dateSaved: new Date,
+                    modelTopologyType: "JSON",
+                    modelTopologyBytes: null == e.modelTopology ? 0 : Ws(JSON.stringify(e.modelTopology)),
+                    weightSpecsBytes: null == e.weightSpecs ? 0 : Ws(JSON.stringify(e.weightSpecs)),
+                    weightDataBytes: null == e.weightData ? 0 : e.weightData.byteLength
+                }
+            }
+
+            function Ks() {
+                const e = function () {
+                        const e = e => {
+                                let t = e << 13,
+                                    n = 0;
+                                for (; 0 == (8388608 & t);) n -= 8388608, t <<= 1;
+                                return t &= -8388609, n += 947912704, t | n
+                            },
+                            t = new Uint32Array(2048);
+                        t[0] = 0;
+                        for (let n = 1; n < 1024; n++) t[n] = e(n);
+                        for (let e = 1024; e < 2048; e++) t[e] = 939524096 + (e - 1024 << 13);
+                        return t
+                    }(),
+                    t = function () {
+                        const e = new Uint32Array(64);
+                        e[0] = 0, e[31] = 1199570944, e[32] = 2147483648, e[63] = 3347054592;
+                        for (let t = 1; t < 31; t++) e[t] = t << 23;
+                        for (let t = 33; t < 63; t++) e[t] = 2147483648 + (t - 32 << 23);
+                        return e
+                    }(),
+                    n = function () {
+                        const e = new Uint32Array(64);
+                        for (let t = 0; t < 64; t++) e[t] = 1024;
+                        return e[0] = e[32] = 0, e
+                    }();
+                return r => {
+                    const a = new ArrayBuffer(4 * r.length),
+                        s = new Uint32Array(a);
+                    for (let a = 0; a < r.length; a++) {
+                        const i = r[a],
+                            o = e[n[i >> 10] + (1023 & i)] + t[i >> 10];
+                        s[a] = o
+                    }
+                    return new Float32Array(a)
+                }
+            }
+            var Ys = class {
+                    constructor() {
+                        this.saveRouters = [], this.loadRouters = []
+                    }
+                    static getInstance() {
+                        return null == Ys.instance && (Ys.instance = new Ys), Ys.instance
+                    }
+                    static registerSaveRouter(e) {
+                        Ys.getInstance().saveRouters.push(e)
+                    }
+                    static registerLoadRouter(e) {
+                        Ys.getInstance().loadRouters.push(e)
+                    }
+                    static getSaveHandlers(e) {
+                        return Ys.getHandlers(e, "save")
+                    }
+                    static getLoadHandlers(e, t) {
+                        return Ys.getHandlers(e, "load", t)
+                    }
+                    static getHandlers(e, t, n) {
+                        const r = [];
+                        return ("load" === t ? Ys.getInstance().loadRouters : Ys.getInstance().saveRouters).forEach(t => {
+                            const a = t(e, n);
+                            null !== a && r.push(a)
+                        }), r
+                    }
+                },
+                Xs = e => Ys.registerSaveRouter(e),
+                Qs = e => Ys.registerLoadRouter(e),
+                Zs = e => Ys.getSaveHandlers(e),
+                Js = (e, t) => Ys.getLoadHandlers(e, t);
+
+            function ei() {
+                if (!Ye().getBool("IS_BROWSER")) throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");
+                const e = "undefined" == typeof window ? self : window,
+                    t = e.indexedDB || e.mozIndexedDB || e.webkitIndexedDB || e.msIndexedDB || e.shimIndexedDB;
+                if (null == t) throw new Error("The current browser does not appear to support IndexedDB.");
+                return t
+            }
+
+            function ti(e) {
+                const t = e.result;
+                t.createObjectStore("models_store", {
+                    keyPath: "modelPath"
+                }), t.createObjectStore("model_info_store", {
+                    keyPath: "modelPath"
+                })
+            }
+            var ni = class {
+                constructor(e) {
+                    if (this.indexedDB = ei(), null == e || !e) throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");
+                    this.modelPath = e
+                }
+                async save(e) {
+                    if (e.modelTopology instanceof ArrayBuffer) throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");
+                    return this.databaseAction(this.modelPath, e)
+                }
+                async load() {
+                    return this.databaseAction(this.modelPath)
+                }
+                databaseAction(e, t) {
+                    return new Promise((e, n) => {
+                        const r = this.indexedDB.open("tensorflowjs", 1);
+                        r.onupgradeneeded = () => ti(r), r.onsuccess = () => {
+                            const a = r.result;
+                            if (null == t) {
+                                const t = a.transaction("models_store", "readonly"),
+                                    r = t.objectStore("models_store").get(this.modelPath);
+                                r.onsuccess = () => {
+                                    if (null == r.result) return a.close(), n(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));
+                                    e(r.result.modelArtifacts)
+                                }, r.onerror = e => (a.close(), n(r.error)), t.oncomplete = () => a.close()
+                            } else {
+                                const r = qs(t),
+                                    s = a.transaction("model_info_store", "readwrite");
+                                let i = s.objectStore("model_info_store");
+                                const o = i.put({
+                                    modelPath: this.modelPath,
+                                    modelArtifactsInfo: r
+                                });
+                                let u;
+                                o.onsuccess = () => {
+                                    u = a.transaction("models_store", "readwrite");
+                                    const o = u.objectStore("models_store").put({
+                                        modelPath: this.modelPath,
+                                        modelArtifacts: t,
+                                        modelArtifactsInfo: r
+                                    });
+                                    o.onsuccess = () => e({
+                                        modelArtifactsInfo: r
+                                    }), o.onerror = e => {
+                                        i = s.objectStore("model_info_store");
+                                        const t = i.delete(this.modelPath);
+                                        t.onsuccess = () => (a.close(), n(o.error)), t.onerror = e => (a.close(), n(o.error))
+                                    }
+                                }, o.onerror = e => (a.close(), n(o.error)), s.oncomplete = () => {
+                                    null == u ? a.close() : u.oncomplete = () => a.close()
+                                }
+                            }
+                        }, r.onerror = e => n(r.error)
+                    })
+                }
+            };
+            ni.URL_SCHEME = "indexeddb://";
+            var ri = e => {
+                return Ye().getBool("IS_BROWSER") && !Array.isArray(e) && e.startsWith(ni.URL_SCHEME) ? (t = e.slice(ni.URL_SCHEME.length), new ni(t)) : null;
+                var t
+            };
+            Ys.registerSaveRouter(ri), Ys.registerLoadRouter(ri);
+            var ai = "tensorflowjs_models",
+                si = "info",
+                ii = "model_topology",
+                oi = "weight_specs",
+                ui = "weight_data",
+                li = "model_metadata";
+
+            function ci(e) {
+                return {
+                    info: [ai, e, si].join("/"),
+                    topology: [ai, e, ii].join("/"),
+                    weightSpecs: [ai, e, oi].join("/"),
+                    weightData: [ai, e, ui].join("/"),
+                    modelMetadata: [ai, e, li].join("/")
+                }
+            }
+
+            function hi(e) {
+                for (const t of Object.values(e)) window.localStorage.removeItem(t)
+            }
+
+            function pi(e) {
+                const t = e.split("/");
+                if (t.length < 3) throw new Error("Invalid key format: " + e);
+                return t.slice(1, t.length - 1).join("/")
+            }
+            var di = class {
+                constructor(e) {
+                    if (!Ye().getBool("IS_BROWSER") || "undefined" == typeof window || void 0 === window.localStorage) throw new Error("The current environment does not support local storage.");
+                    if (this.LS = window.localStorage, null == e || !e) throw new Error("For local storage, modelPath must not be null, undefined or empty.");
+                    this.modelPath = e, this.keys = ci(this.modelPath)
+                }
+                async save(e) {
+                    if (e.modelTopology instanceof ArrayBuffer) throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet."); {
+                        const t = JSON.stringify(e.modelTopology),
+                            n = JSON.stringify(e.weightSpecs),
+                            r = qs(e);
+                        try {
+                            this.LS.setItem(this.keys.info, JSON.stringify(r)), this.LS.setItem(this.keys.topology, t), this.LS.setItem(this.keys.weightSpecs, n), this.LS.setItem(this.keys.weightData, function (e) {
+                                if (Us) return i.from(e).toString("base64");
+                                const t = new Uint8Array(e);
+                                let n = "";
+                                for (let e = 0, r = t.length; e < r; e++) n += String.fromCharCode(t[e]);
+                                return btoa(n)
+                            }(e.weightData));
+                            const a = {
+                                format: e.format,
+                                generatedBy: e.generatedBy,
+                                convertedBy: e.convertedBy,
+                                signature: null != e.signature ? e.signature : void 0,
+                                userDefinedMetadata: null != e.userDefinedMetadata ? e.userDefinedMetadata : void 0,
+                                modelInitializer: null != e.modelInitializer ? e.modelInitializer : void 0,
+                                trainingConfig: null != e.trainingConfig ? e.trainingConfig : void 0
+                            };
+                            return this.LS.setItem(this.keys.modelMetadata, JSON.stringify(a)), {
+                                modelArtifactsInfo: r
+                            }
+                        } catch (e) {
+                            throw hi(this.keys), new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${r.modelTopologyBytes}, weightSpecsBytes=${r.weightSpecsBytes}, weightDataBytes=${r.weightDataBytes}.`)
+                        }
+                    }
+                }
+                async load() {
+                    const e = JSON.parse(this.LS.getItem(this.keys.info));
+                    if (null == e) throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);
+                    if ("JSON" !== e.modelTopologyType) throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");
+                    const t = {},
+                        n = JSON.parse(this.LS.getItem(this.keys.topology));
+                    if (null == n) throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);
+                    t.modelTopology = n;
+                    const r = JSON.parse(this.LS.getItem(this.keys.weightSpecs));
+                    if (null == r) throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);
+                    t.weightSpecs = r;
+                    const a = this.LS.getItem(this.keys.modelMetadata);
+                    if (null != a) {
+                        const e = JSON.parse(a);
+                        t.format = e.format, t.generatedBy = e.generatedBy, t.convertedBy = e.convertedBy, null != e.signature && (t.signature = e.signature), null != e.userDefinedMetadata && (t.userDefinedMetadata = e.userDefinedMetadata), null != e.modelInitializer && (t.modelInitializer = e.modelInitializer), null != e.trainingConfig && (t.trainingConfig = e.trainingConfig)
+                    }
+                    const s = this.LS.getItem(this.keys.weightData);
+                    if (null == s) throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);
+                    return t.weightData = function (e) {
+                        if (Us) {
+                            const t = i.from(e, "base64");
+                            return t.buffer.slice(t.byteOffset, t.byteOffset + t.byteLength)
+                        }
+                        const t = atob(e),
+                            n = new Uint8Array(t.length);
+                        for (let e = 0; e < t.length; ++e) n.set([t.charCodeAt(e)], e);
+                        return n.buffer
+                    }(s), t
+                }
+            };
+            di.URL_SCHEME = "localstorage://";
+            var fi = e => {
+                return Ye().getBool("IS_BROWSER") && !Array.isArray(e) && e.startsWith(di.URL_SCHEME) ? (t = e.slice(di.URL_SCHEME.length), new di(t)) : null;
+                var t
+            };
+            Ys.registerSaveRouter(fi), Ys.registerLoadRouter(fi);
+            var mi = class {
+                constructor() {
+                    this.managers = {}
+                }
+                static getInstance() {
+                    return null == mi.instance && (mi.instance = new mi), mi.instance
+                }
+                static registerManager(e, t) {
+                    ue(null != e, () => "scheme must not be undefined or null."), e.endsWith("://") && (e = e.slice(0, e.indexOf("://"))), ue(e.length > 0, () => "scheme must not be an empty string.");
+                    const n = mi.getInstance();
+                    ue(null == n.managers[e], () => `A model store manager is already registered for scheme '${e}'.`), n.managers[e] = t
+                }
+                static getManager(e) {
+                    const t = this.getInstance().managers[e];
+                    if (null == t) throw new Error(`Cannot find model manager for scheme '${e}'`);
+                    return t
+                }
+                static getSchemes() {
+                    return Object.keys(this.getInstance().managers)
+                }
+            };
+
+            function gi(e) {
+                if (-1 === e.indexOf("://")) throw new Error("The url string provided does not contain a scheme. Supported schemes are: " + mi.getSchemes().join(","));
+                return {
+                    scheme: e.split("://")[0],
+                    path: e.split("://")[1]
+                }
+            }
+            async function yi(e, t, n = !1) {
+                ue(e !== t, () => `Old path and new path are the same: '${e}'`);
+                const r = Ys.getLoadHandlers(e);
+                ue(r.length > 0, () => `Copying failed because no load handler is found for source URL ${e}.`), ue(r.length < 2, () => `Copying failed because more than one (${r.length}) load handlers for source URL ${e}.`);
+                const a = r[0],
+                    s = Ys.getSaveHandlers(t);
+                ue(s.length > 0, () => `Copying failed because no save handler is found for destination URL ${t}.`), ue(s.length < 2, () => `Copying failed because more than one (${r.length}) save handlers for destination URL ${t}.`);
+                const i = s[0],
+                    o = gi(e).scheme,
+                    u = gi(e).path,
+                    l = o === gi(e).scheme,
+                    c = await a.load();
+                n && l && await mi.getManager(o).removeModel(u);
+                const h = await i.save(c);
+                return n && !l && await mi.getManager(o).removeModel(u), h.modelArtifactsInfo
+            }
+            async function bi() {
+                const e = mi.getSchemes(),
+                    t = {};
+                for (const n of e) {
+                    const e = await mi.getManager(n).listModels();
+                    for (const r in e) {
+                        t[n + "://" + r] = e[r]
+                    }
+                }
+                return t
+            }
+            async function vi(e) {
+                const t = gi(e);
+                return mi.getManager(t.scheme).removeModel(t.path)
+            }
+            async function xi(e, t) {
+                return yi(e, t, !1)
+            }
+            async function wi(e, t) {
+                return yi(e, t, !0)
+            }
+            if (Ye().get("IS_BROWSER")) {
+                Ye().setPlatform("browser", new class {
+                    fetch(e, t) {
+                        return fetch(e, t)
+                    }
+                    now() {
+                        return performance.now()
+                    }
+                    encode(e, t) {
+                        if ("utf-8" !== t && "utf8" !== t) throw new Error("Browser's encoder only supports utf-8, but got " + t);
+                        return null == this.textEncoder && (this.textEncoder = new TextEncoder), this.textEncoder.encode(e)
+                    }
+                    decode(e, t) {
+                        return new TextDecoder(t).decode(e)
+                    }
+                });
+                try {
+                    mi.registerManager(di.URL_SCHEME, new class {
+                        constructor() {
+                            ue(Ye().getBool("IS_BROWSER"), () => "Current environment is not a web browser"), ue("undefined" == typeof window || void 0 !== window.localStorage, () => "Current browser does not appear to support localStorage"), this.LS = window.localStorage
+                        }
+                        async listModels() {
+                            const e = {},
+                                t = ai + "/",
+                                n = "/" + si;
+                            for (let r = 0; r < this.LS.length; ++r) {
+                                const a = this.LS.key(r);
+                                if (a.startsWith(t) && a.endsWith(n)) {
+                                    e[pi(a)] = JSON.parse(this.LS.getItem(a))
+                                }
+                            }
+                            return e
+                        }
+                        async removeModel(e) {
+                            var t;
+                            const n = ci(e = (t = e).startsWith(di.URL_SCHEME) ? t.slice(di.URL_SCHEME.length) : t);
+                            if (null == this.LS.getItem(n.info)) throw new Error(`Cannot find model at path '${e}'`);
+                            const r = JSON.parse(this.LS.getItem(n.info));
+                            return hi(n), r
+                        }
+                    })
+                } catch (e) {}
+                try {
+                    mi.registerManager(ni.URL_SCHEME, new class {
+                        constructor() {
+                            this.indexedDB = ei()
+                        }
+                        async listModels() {
+                            return new Promise((e, t) => {
+                                const n = this.indexedDB.open("tensorflowjs", 1);
+                                n.onupgradeneeded = () => ti(n), n.onsuccess = () => {
+                                    const r = n.result,
+                                        a = r.transaction("model_info_store", "readonly"),
+                                        s = a.objectStore("model_info_store").getAll();
+                                    s.onsuccess = () => {
+                                        const t = {};
+                                        for (const e of s.result) t[e.modelPath] = e.modelArtifactsInfo;
+                                        e(t)
+                                    }, s.onerror = e => (r.close(), t(s.error)), a.oncomplete = () => r.close()
+                                }, n.onerror = e => t(n.error)
+                            })
+                        }
+                        async removeModel(e) {
+                            var t;
+                            return e = (t = e).startsWith(ni.URL_SCHEME) ? t.slice(ni.URL_SCHEME.length) : t, new Promise((t, n) => {
+                                const r = this.indexedDB.open("tensorflowjs", 1);
+                                r.onupgradeneeded = () => ti(r), r.onsuccess = () => {
+                                    const a = r.result,
+                                        s = a.transaction("model_info_store", "readwrite"),
+                                        i = s.objectStore("model_info_store"),
+                                        o = i.get(e);
+                                    let u;
+                                    o.onsuccess = () => {
+                                        if (null == o.result) return a.close(), n(new Error(`Cannot find model with path '${e}' in IndexedDB.`)); {
+                                            const r = i.delete(e),
+                                                s = () => {
+                                                    u = a.transaction("models_store", "readwrite");
+                                                    const r = u.objectStore("models_store").delete(e);
+                                                    r.onsuccess = () => t(o.result.modelArtifactsInfo), r.onerror = e => n(o.error)
+                                                };
+                                            r.onsuccess = s, r.onerror = e => (s(), a.close(), n(o.error))
+                                        }
+                                    }, o.onerror = e => (a.close(), n(o.error)), s.oncomplete = () => {
+                                        null == u ? a.close() : u.oncomplete = () => a.close()
+                                    }
+                                }, r.onerror = e => n(r.error)
+                            })
+                        }
+                    })
+                } catch (e) {}
+            }
+            var ki, Si = () => I();
+
+            function Ii(e, t = "float32", n) {
+                return t = t || "float32", Ve(e), new Ha(e, t, n)
+            }
+            Ye().get("IS_NODE") && Ye().setPlatform("node", new class {
+                constructor() {
+                    this.util = A(), this.textEncoder = new this.util.TextEncoder
+                }
+                fetch(e, t) {
+                    return null != Ye().global.fetch ? Ye().global.fetch(e, t) : (null == ki && (ki = Si()), ki(e, t))
+                }
+                now() {
+                    const e = r.hrtime();
+                    return 1e3 * e[0] + e[1] / 1e6
+                }
+                encode(e, t) {
+                    if ("utf-8" !== t && "utf8" !== t) throw new Error("Node built-in encoder only supports utf-8, but got " + t);
+                    return this.textEncoder.encode(e)
+                }
+                decode(e, t) {
+                    return 0 === e.length ? "" : new this.util.TextDecoder(t).decode(e)
+                }
+            });
+            var Ai = Ms({
+                cast_: function (e, t) {
+                    const n = Rs(e, "x", "cast");
+                    if (!Ne(t)) throw new Error("Failed to cast to unknown dtype " + t);
+                    if ("string" === t && "string" !== n.dtype || "string" !== t && "string" === n.dtype) throw new Error("Only strings can be casted to strings");
+                    const r = {
+                            x: n
+                        },
+                        a = {
+                            dtype: t
+                        };
+                    return xs.runKernel(St, r, a)
+                }
+            });
+            var Ei = Ms({
+                clone_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "clone", "string_or_numeric")
+                    };
+                    return xs.runKernel(fn, t)
+                }
+            });
+
+            function Ni(e, t = !1) {
+                console.log(e.toString(t))
+            }
+            vs(), Ka = {
+                buffer: Ii,
+                cast: Ai,
+                clone: Ei,
+                print: Ni
+            };
+            var Ci = {};
+            w(Ci, {
+                browserFiles: () => _i,
+                browserHTTPRequest: () => zi,
+                concatenateArrayBuffers: () => Vs,
+                copyModel: () => xi,
+                decodeWeights: () => Bs,
+                encodeWeights: () => $s,
+                fromMemory: () => Wi,
+                getLoadHandlers: () => Js,
+                getModelArtifactsForJSON: () => Hs,
+                getModelArtifactsInfoForJSON: () => qs,
+                getSaveHandlers: () => Zs,
+                http: () => Bi,
+                isHTTPScheme: () => Pi,
+                listModels: () => bi,
+                loadWeights: () => Di,
+                moveModel: () => wi,
+                registerLoadRouter: () => Qs,
+                registerSaveRouter: () => Xs,
+                removeModel: () => vi,
+                weightsLoaderFactory: () => Oi,
+                withSaveHandler: () => Vi
+            });
+
+            function Ti(e) {
+                return new Promise(e => setTimeout(e)).then(e)
+            }
+            var Ri = class {
+                constructor(e) {
+                    if (!Ye().getBool("IS_BROWSER")) throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");
+                    e.startsWith(Ri.URL_SCHEME) && (e = e.slice(Ri.URL_SCHEME.length)), null != e && 0 !== e.length || (e = "model"), this.modelJsonFileName = e + ".json", this.weightDataFileName = e + ".weights.bin"
+                }
+                async save(e) {
+                    if ("undefined" == typeof document) throw new Error("Browser downloads are not supported in this environment since `document` is not present");
+                    const t = window.URL.createObjectURL(new Blob([e.weightData], {
+                        type: "application/octet-stream"
+                    }));
+                    if (e.modelTopology instanceof ArrayBuffer) throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet."); {
+                        const n = Gs(e, [{
+                                paths: ["./" + this.weightDataFileName],
+                                weights: e.weightSpecs
+                            }]),
+                            r = window.URL.createObjectURL(new Blob([JSON.stringify(n)], {
+                                type: "application/json"
+                            })),
+                            a = null == this.modelJsonAnchor ? document.createElement("a") : this.modelJsonAnchor;
+                        if (a.download = this.modelJsonFileName, a.href = r, await Ti(() => a.dispatchEvent(new MouseEvent("click"))), null != e.weightData) {
+                            const e = null == this.weightDataAnchor ? document.createElement("a") : this.weightDataAnchor;
+                            e.download = this.weightDataFileName, e.href = t, await Ti(() => e.dispatchEvent(new MouseEvent("click")))
+                        }
+                        return {
+                            modelArtifactsInfo: qs(e)
+                        }
+                    }
+                }
+            };
+            Ri.URL_SCHEME = "downloads://";
+
+            function _i(e) {
+                return new class {
+                    constructor(e) {
+                        if (null == e || e.length < 1) throw new Error("When calling browserFiles, at least 1 file is required, but received " + e);
+                        this.jsonFile = e[0], this.weightsFiles = e.slice(1)
+                    }
+                    async load() {
+                        return new Promise((e, t) => {
+                            const n = new FileReader;
+                            n.onload = n => {
+                                const r = JSON.parse(n.target.result),
+                                    a = r.modelTopology;
+                                if (null == a) return void t(new Error("modelTopology field is missing from file " + this.jsonFile.name));
+                                if (null == r.weightsManifest) return void t(new Error("weightManifest field is missing from file " + this.jsonFile.name));
+                                if (0 === this.weightsFiles.length) return void e({
+                                    modelTopology: a
+                                });
+                                const s = Hs(r, e => this.loadWeights(e));
+                                e(s)
+                            }, n.onerror = e => t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`), n.readAsText(this.jsonFile)
+                        })
+                    }
+                    loadWeights(e) {
+                        const t = [],
+                            n = [];
+                        for (const r of e) t.push(...r.weights), n.push(...r.paths);
+                        const r = this.checkManifestAndWeightFiles(e),
+                            a = n.map(e => this.loadWeightsFile(e, r[e]));
+                        return Promise.all(a).then(e => [t, Vs(e)])
+                    }
+                    loadWeightsFile(e, t) {
+                        return new Promise((n, r) => {
+                            const a = new FileReader;
+                            a.onload = e => {
+                                const t = e.target.result;
+                                n(t)
+                            }, a.onerror = t => r(`Failed to weights data from file of path '${e}'.`), a.readAsArrayBuffer(t)
+                        })
+                    }
+                    checkManifestAndWeightFiles(e) {
+                        const t = [],
+                            n = this.weightsFiles.map(e => js(e.name)),
+                            r = {};
+                        for (const a of e) a.paths.forEach(e => {
+                            const a = js(e);
+                            if (-1 !== t.indexOf(a)) throw new Error(`Duplicate file basename found in weights manifest: '${a}'`);
+                            if (t.push(a), -1 === n.indexOf(a)) throw new Error(`Weight file with basename '${a}' is not provided.`);
+                            r[e] = this.weightsFiles[n.indexOf(a)]
+                        });
+                        if (t.length !== this.weightsFiles.length) throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);
+                        return r
+                    }
+                }(e)
+            }
+
+            function Fi(e, t, n, r) {
+                var a, s, i;
+                ue(null != (a = e) && Array.isArray(a) && a.length > 0, () => "promises must be a none empty array"), i = r = null == r ? 1 : r, ue((s = n = null == n ? 0 : n) >= 0 && s <= 1, () => "Progress fraction must be in range [0, 1], but got startFraction " + s), ue(i >= 0 && i <= 1, () => "Progress fraction must be in range [0, 1], but got endFraction " + i), ue(i >= s, () => `startFraction must be no more than endFraction, but got startFraction ${s} and endFraction ${i}`);
+                let o = 0;
+                return Promise.all(e.map(a => (a.then(a => {
+                    const s = n + ++o / e.length * (r - n);
+                    return t(s), a
+                }), a)))
+            }
+            async function Mi(e, t) {
+                null == t && (t = {});
+                const n = null == t.fetchFunc ? Ye().platform.fetch : t.fetchFunc,
+                    r = e.map(e => n(e, t.requestInit, {
+                        isBinary: !0
+                    })),
+                    a = (null == t.onProgress ? await Promise.all(r) : await Fi(r, t.onProgress, 0, .5)).map(e => e.arrayBuffer());
+                return null == t.onProgress ? await Promise.all(a) : await Fi(a, t.onProgress, .5, 1)
+            }
+            async function Di(e, t = "", n, r) {
+                return Oi(e => Mi(e, {
+                    requestInit: r
+                }))(e, t, n)
+            }
+
+            function Oi(e) {
+                return async (t, n = "", r) => {
+                    const a = t.map(() => !1),
+                        s = {},
+                        i = null != r ? r.map(() => !1) : [],
+                        o = [];
+                    if (t.forEach((e, t) => {
+                            let n = 0;
+                            e.weights.forEach(e => {
+                                const u = "quantization" in e ? e.quantization.dtype : e.dtype,
+                                    l = Ps[u] * pe(e.shape),
+                                    c = () => {
+                                        a[t] = !0, null == s[t] && (s[t] = []), s[t].push({
+                                            manifestEntry: e,
+                                            groupOffset: n,
+                                            sizeBytes: l
+                                        })
+                                    };
+                                null != r ? r.forEach((t, n) => {
+                                    t === e.name && (c(), i[n] = !0)
+                                }) : c(), o.push(e.name), n += l
+                            })
+                        }), !i.every(e => e)) {
+                        const e = r.filter((e, t) => !i[t]);
+                        throw new Error(`Could not find weights in manifest with names: ${e.join(", ")}. \nManifest JSON has weights with names: ${o.join(", ")}.`)
+                    }
+                    const u = a.reduce((e, t, n) => (t && e.push(n), e), []),
+                        l = [];
+                    u.forEach(e => {
+                        t[e].paths.forEach(e => {
+                            const t = n + (n.endsWith("/") ? "" : "/") + e;
+                            l.push(t)
+                        })
+                    });
+                    const c = await e(l),
+                        h = {};
+                    let p = 0;
+                    return u.forEach(e => {
+                        const n = t[e].paths.length;
+                        let r = 0;
+                        for (let e = 0; e < n; e++) r += c[p + e].byteLength;
+                        const a = new ArrayBuffer(r),
+                            i = new Uint8Array(a);
+                        let o = 0;
+                        for (let e = 0; e < n; e++) {
+                            const t = new Uint8Array(c[p + e]);
+                            i.set(t, o), o += t.byteLength
+                        }
+                        s[e].forEach(e => {
+                            const t = Bs(a.slice(e.groupOffset, e.groupOffset + e.sizeBytes), [e.manifestEntry]);
+                            for (const e in t) h[e] = t[e]
+                        }), p += n
+                    }), h
+                }
+            }
+            Ys.registerSaveRouter(e => Ye().getBool("IS_BROWSER") && !Array.isArray(e) && e.startsWith(Ri.URL_SCHEME) ? function (e = "model") {
+                return new Ri(e)
+            }(e.slice(Ri.URL_SCHEME.length)) : null);
+            var Li = class {
+                constructor(e, t) {
+                    if (this.DEFAULT_METHOD = "POST", null == t && (t = {}), this.weightPathPrefix = t.weightPathPrefix, this.onProgress = t.onProgress, this.weightUrlConverter = t.weightUrlConverter, null != t.fetchFunc ? (ue("function" == typeof t.fetchFunc, () => "Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"), this.fetch = t.fetchFunc) : this.fetch = Ye().platform.fetch, ue(null != e && e.length > 0, () => "URL path for http must not be null, undefined or empty."), Array.isArray(e) && ue(2 === e.length, () => `URL paths for http must have a length of 2, (actual length is ${e.length}).`), this.path = e, null != t.requestInit && null != t.requestInit.body) throw new Error("requestInit is expected to have no pre-existing body, but has one.");
+                    this.requestInit = t.requestInit || {}
+                }
+                async save(e) {
+                    if (e.modelTopology instanceof ArrayBuffer) throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");
+                    const t = Object.assign({
+                        method: this.DEFAULT_METHOD
+                    }, this.requestInit);
+                    t.body = new FormData;
+                    const n = Gs(e, [{
+                        paths: ["./model.weights.bin"],
+                        weights: e.weightSpecs
+                    }]);
+                    t.body.append("model.json", new Blob([JSON.stringify(n)], {
+                        type: "application/json"
+                    }), "model.json"), null != e.weightData && t.body.append("model.weights.bin", new Blob([e.weightData], {
+                        type: "application/octet-stream"
+                    }), "model.weights.bin");
+                    const r = await this.fetch(this.path, t);
+                    if (r.ok) return {
+                        modelArtifactsInfo: qs(e),
+                        responses: [r]
+                    };
+                    throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${r.status}.`)
+                }
+                async load() {
+                    const e = await this.fetch(this.path, this.requestInit);
+                    if (!e.ok) throw new Error(`Request to ${this.path} failed with status code ${e.status}. Please verify this URL points to the model JSON of the model to load.`);
+                    let t;
+                    try {
+                        t = await e.json()
+                    } catch (e) {
+                        let t = `Failed to parse model JSON of response from ${this.path}.`;
+                        throw this.path.endsWith(".pb") ? t += " Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository." : t += " Please make sure the server is serving valid JSON for this request.", new Error(t)
+                    }
+                    const n = t.modelTopology,
+                        r = t.weightsManifest;
+                    if (null == n && null == r) throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);
+                    return Hs(t, e => this.loadWeights(e))
+                }
+                async loadWeights(e) {
+                    const t = Array.isArray(this.path) ? this.path[1] : this.path,
+                        [n, r] = function (e) {
+                            const t = e.lastIndexOf("/"),
+                                n = e.lastIndexOf("?"),
+                                r = e.substring(0, t),
+                                a = n > t ? e.substring(n) : "";
+                            return [r + "/", a]
+                        }(t),
+                        a = this.weightPathPrefix || n,
+                        s = [];
+                    for (const t of e) s.push(...t.weights);
+                    const i = [],
+                        o = [];
+                    for (const t of e)
+                        for (const e of t.paths) null != this.weightUrlConverter ? o.push(this.weightUrlConverter(e)) : i.push(a + e + r);
+                    this.weightUrlConverter && i.push(...await Promise.all(o));
+                    return [s, Vs(await Mi(i, {
+                        requestInit: this.requestInit,
+                        fetchFunc: this.fetch,
+                        onProgress: this.onProgress
+                    }))]
+                }
+            };
+
+            function Pi(e) {
+                return null != e.match(Li.URL_SCHEME_REGEX)
+            }
+            Li.URL_SCHEME_REGEX = /^https?:\/\//;
+            var $i = (e, t) => {
+                if ("undefined" == typeof fetch && (null == t || null == t.fetchFunc)) return null; {
+                    let n = !0;
+                    if (n = Array.isArray(e) ? e.every(e => Pi(e)) : Pi(e), n) return Bi(e, t)
+                }
+                return null
+            };
+
+            function Bi(e, t) {
+                return new Li(e, t)
+            }
+
+            function zi(e, t) {
+                return Bi(e, t)
+            }
+            Ys.registerSaveRouter($i), Ys.registerLoadRouter($i);
+            var Ui = class {
+                constructor(e) {
+                    this.modelArtifacts = e
+                }
+                async load() {
+                    return this.modelArtifacts
+                }
+            };
+
+            function Wi(e, t, n, r) {
+                if (1 === arguments.length) {
+                    return null != e.modelTopology || null != e.weightSpecs ? new Ui(e) : (console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."), new Ui({
+                        modelTopology: e
+                    }))
+                }
+                return console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."), new Ui({
+                    modelTopology: e,
+                    weightSpecs: t,
+                    weightData: n,
+                    trainingConfig: r
+                })
+            }
+
+            function Vi(e) {
+                return new class {
+                    constructor(e) {
+                        this.saveHandler = e
+                    }
+                    async save(e) {
+                        return this.saveHandler(e)
+                    }
+                }(e)
+            }
+            var ji = {};
+            w(ji, {
+                confusionMatrix: () => Ki
+            });
+            var Gi = Ms({
+                matMul_: function (e, t, n = !1, r = !1) {
+                    let a = Rs(e, "a", "matMul"),
+                        s = Rs(t, "b", "matMul");
+                    [a, s] = ps(a, s);
+                    const i = {
+                            a: a,
+                            b: s
+                        },
+                        o = {
+                            transposeA: n,
+                            transposeB: r
+                        };
+                    return xs.runKernel(bt, i, o)
+                }
+            });
+            var Hi = Ms({
+                oneHot_: function (e, t, n = 1, r = 0) {
+                    if (t < 2) throw new Error("Error in oneHot: depth must be >=2, but it is " + t);
+                    const a = {
+                            indices: Rs(e, "indices", "oneHot", "int32")
+                        },
+                        s = {
+                            depth: t,
+                            onValue: n,
+                            offValue: r
+                        };
+                    return xs.runKernel(Zn, a, s)
+                }
+            });
+            var qi = Ms({
+                transpose_: function (e, t) {
+                    const n = Rs(e, "x", "transpose");
+                    if (null == t && (t = n.shape.map((e, t) => t).reverse()), ue(n.rank === t.length, () => `Error in transpose: rank of input ${n.rank} must match length of perm ${t}.`), t.forEach(e => {
+                            ue(e >= 0 && e < n.rank, () => `All entries in 'perm' must be between 0 and ${n.rank-1} but got ${t}`)
+                        }), n.rank <= 1) return n.clone();
+                    const r = {
+                            x: n
+                        },
+                        a = {
+                            perm: t
+                        };
+                    return xs.runKernel(Yr, r, a)
+                }
+            });
+            var Ki = Ms({
+                    confusionMatrix_: function (e, t, n) {
+                        const r = Rs(e, "labels", "confusionMatrix"),
+                            a = Rs(t, "predictions", "confusionMatrix");
+                        ue(null == n || n > 0 && Number.isInteger(n), () => "If provided, numClasses must be a positive integer, but got " + n), ue(1 === r.rank, () => "Expected the rank of labels to be 1, but got " + r.rank), ue(1 === a.rank, () => "Expected the rank of predictions to be 1, but got " + a.rank), ue(r.shape[0] === a.shape[0], () => `Mismatch in the number of examples: ${r.shape[0]} vs. ${a.shape[0]}. Labels and predictions should have the same number of elements.`), ue(n > 0 && Number.isInteger(n), () => "numClasses is required to be a positive integer, but got " + n);
+                        const s = Hi(Ai(r, "int32"), n),
+                            i = Hi(Ai(a, "int32"), n),
+                            o = qi(s),
+                            u = Gi(o, i);
+                        return Ai(u, "int32")
+                    }
+                }),
+                Yi = {};
+
+            function Xi(e, t) {
+                const n = e.length,
+                    r = [];
+                for (let a = 0; a < n; a++) {
+                    const s = n - 1 - a,
+                        i = e[s] || 1;
+                    (t[t.length - 1 - a] || 1) > 1 && 1 === i && r.unshift(s)
+                }
+                return r
+            }
+
+            function Qi(e, t) {
+                const n = [];
+                for (let r = 0; r < t.length; r++) {
+                    const a = e[e.length - r - 1],
+                        s = t.length - r - 1,
+                        i = t[s];
+                    (null == a || 1 === a && i > 1) && n.unshift(s)
+                }
+                return n
+            }
+
+            function Zi(e, t) {
+                const n = [],
+                    r = Math.max(e.length, t.length);
+                for (let a = 0; a < r; a++) {
+                    let r = e[e.length - a - 1];
+                    null == r && (r = 1);
+                    let s = t[t.length - a - 1];
+                    if (null == s && (s = 1), 1 === r) n.unshift(s);
+                    else if (1 === s) n.unshift(r);
+                    else {
+                        if (r !== s) {
+                            throw Error(`Operands could not be broadcast together with shapes ${e} and ${t}.`)
+                        }
+                        n.unshift(r)
+                    }
+                }
+                return n
+            }
+            w(Yi, {
+                assertAndGetBroadcastShape: () => Zi,
+                getBroadcastDims: () => Xi,
+                getReductionAxes: () => Qi
+            });
+            var Ji, eo = {};
+
+            function to(e, t, n) {
+                if (ce(e), null != t && 3 !== t.length) throw new Error("tensor3d() requires shape to have three numbers");
+                const r = Cs(e, n);
+                if (3 !== r.length && 1 !== r.length) throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");
+                if (1 === r.length && null == t) throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");
+                return Os(e, t, r, n)
+            }
+
+            function no(e, t = 3) {
+                if (t > 4) throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");
+                if (null == e) throw new Error("pixels passed to tf.browser.fromPixels() can not be null");
+                let n = !1,
+                    r = !1,
+                    a = !1,
+                    s = !1,
+                    i = !1,
+                    o = !1;
+                if (e.data instanceof Uint8Array) n = !0;
+                else if ("undefined" != typeof ImageData && e instanceof ImageData) r = !0;
+                else if ("undefined" != typeof HTMLVideoElement && e instanceof HTMLVideoElement) a = !0;
+                else if ("undefined" != typeof HTMLImageElement && e instanceof HTMLImageElement) s = !0;
+                else if (null != e.getContext) i = !0;
+                else {
+                    if (!("undefined" != typeof ImageBitmap && e instanceof ImageBitmap)) throw new Error("pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was " + e.constructor.name);
+                    o = !0
+                }
+                if (a) {
+                    const t = 2;
+                    if (a && e.readyState < t) throw new Error("The video element has not loaded data yet. Please wait for `loadeddata` event on the <video> element.")
+                }
+                if (null != ca(ta, xs.backendName)) {
+                    const n = {
+                            pixels: e
+                        },
+                        r = {
+                            numChannels: t
+                        };
+                    return xs.runKernel(ta, n, r)
+                }
+                const [u, l] = a ? [e.videoWidth, e.videoHeight] : [e.width, e.height];
+                let c, h;
+                if (i) c = e.getContext("2d").getImageData(0, 0, u, l).data;
+                else if (r || n) c = e.data;
+                else if (s || a || o) {
+                    if (null == Ji)
+                        if ("undefined" == typeof document) {
+                            if ("undefined" == typeof OffscreenCanvas || "undefined" == typeof OffscreenCanvasRenderingContext2D) throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");
+                            Ji = new OffscreenCanvas(1, 1).getContext("2d")
+                        } else Ji = document.createElement("canvas").getContext("2d");
+                    Ji.canvas.width = u, Ji.canvas.height = l, Ji.drawImage(e, 0, 0, u, l), c = Ji.getImageData(0, 0, u, l).data
+                }
+                if (4 === t) h = new Int32Array(c);
+                else {
+                    const e = u * l;
+                    h = new Int32Array(e * t);
+                    for (let n = 0; n < e; n++)
+                        for (let e = 0; e < t; ++e) h[n * t + e] = c[4 * n + e]
+                }
+                return to(h, [l, u, t], "int32")
+            }
+
+            function ro(e) {
+                return "undefined" != typeof window && "undefined" != typeof ImageBitmap && window.hasOwnProperty("createImageBitmap") && !(e instanceof ImageBitmap) && function (e) {
+                    return null != e && 0 !== e.width && 0 !== e.height
+                }(e) && ! function (e) {
+                    return null != e && e.data instanceof Uint8Array
+                }(e)
+            }
+            async function ao(e, t = 3) {
+                let n = null;
+                if (Ye().getBool("WRAP_TO_IMAGEBITMAP") && ro(e)) {
+                    let t;
+                    try {
+                        t = await createImageBitmap(e, {
+                            premultiplyAlpha: "none"
+                        })
+                    } catch (e) {
+                        t = null
+                    }
+                    n = null != t && t.width === e.width && t.height === e.height ? t : e
+                } else n = e;
+                return no(n, t)
+            }
+            async function so(e, t) {
+                let n = Rs(e, "img", "toPixels");
+                if (!(e instanceof Ya)) {
+                    const e = n;
+                    n = Ai(e, "int32"), e.dispose()
+                }
+                if (2 !== n.rank && 3 !== n.rank) throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${n.rank}.`);
+                const [r, a] = n.shape.slice(0, 2), s = 2 === n.rank ? 1 : n.shape[2];
+                if (s > 4 || 2 === s) throw new Error("toPixels only supports depth of size 1, 3 or 4 but got " + s);
+                if ("float32" !== n.dtype && "int32" !== n.dtype) throw new Error(`Unsupported type for toPixels: ${n.dtype}. Please use float32 or int32 tensors.`);
+                const i = await n.data(),
+                    o = "float32" === n.dtype ? 255 : 1,
+                    u = new Uint8ClampedArray(a * r * 4);
+                for (let e = 0; e < r * a; ++e) {
+                    const t = [0, 0, 0, 255];
+                    for (let r = 0; r < s; r++) {
+                        const a = i[e * s + r];
+                        if ("float32" === n.dtype) {
+                            if (a < 0 || a > 1) throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${a}.`)
+                        } else if ("int32" === n.dtype && (a < 0 || a > 255)) throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${a}.`);
+                        1 === s ? (t[0] = a * o, t[1] = a * o, t[2] = a * o) : t[r] = a * o
+                    }
+                    const r = 4 * e;
+                    u[r + 0] = Math.round(t[0]), u[r + 1] = Math.round(t[1]), u[r + 2] = Math.round(t[2]), u[r + 3] = Math.round(t[3])
+                }
+                if (null != t) {
+                    t.width = a, t.height = r;
+                    const e = t.getContext("2d"),
+                        n = new ImageData(u, a, r);
+                    e.putImageData(n, 0, 0)
+                }
+                return n !== e && n.dispose(), u
+            }
+            w(eo, {
+                fromPixels: () => io,
+                fromPixelsAsync: () => ao,
+                toPixels: () => so
+            });
+            var io = Ms({
+                    fromPixels_: no
+                }),
+                oo = {};
+
+            function uo(e, t) {
+                const n = e.shape.length,
+                    r = t.shape.length;
+                if (n < 1) throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${n}.`);
+                if (r < 1) throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${r}.`);
+                if ("int32" !== t.dtype) throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);
+                if (t.shape[r - 1] > n) throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[r-1]} vs. ${n}`);
+                if (0 === pe(e.shape)) throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);
+                const a = t.shape,
+                    s = a[a.length - 1];
+                let i = 1;
+                for (let e = 0; e < a.length - 1; ++e) i *= a[e];
+                const o = e.shape,
+                    u = a.slice();
+                u.pop();
+                let l = 1;
+                for (let e = s; e < n; ++e) l *= o[e], u.push(o[e]);
+                const c = [...$e(e.shape).map(e => e / l), 1].slice(0, s);
+                return [u, i, l, c]
+            }
+            w(oo, {
+                prepareAndValidate: () => uo
+            });
+            var lo = {};
+
+            function co(e, t, n) {
+                const r = t.rank > 1 ? t.shape[t.rank - 1] : 1,
+                    a = t.rank > 1 ? t.rank - 1 : 1,
+                    s = `Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${r}, and batchDim: ${a}.`;
+                if (n.rank < a) throw new Error(s + ` update.rank < ${a}. `);
+                if (e.length < r + (n.rank - a)) throw new Error(s + " Output shape length < " + (r + (n.rank - a)));
+                if (n.rank !== a + e.length - r) throw new Error(s + " update.rank != " + (a + e.length - r));
+                for (let e = 0; e < a; ++e)
+                    if (n.shape[e] !== t.shape[e]) throw new Error(s + ` updates.shape[${e}] (${n.shape[e]}) != indices.shape[${e}] (${t.shape[e]}).`);
+                for (let t = 0; t < n.rank - a; ++t)
+                    if (n.shape[t + a] !== e[t + r]) throw new Error(s + ` updates.shape[${t+a}] (${n.shape[t+a]}) != shape[${t+a}] (${e[t+a]})`)
+            }
+
+            function ho(e, t, n) {
+                if (t.rank < 1) throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);
+                if (e.rank < 1) throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${e.rank}.`);
+                if ("int32" !== t.dtype) throw new Error("The dtype of 'indices' should be int32, but got dtype: " + t.dtype);
+                if (n.length < 1) throw new Error("Output rank must be greater or equal to 1, but got shape: " + n);
+                if (0 === n.length) {
+                    if (0 === t.size) throw new Error("Indices specified for empty output. indices shape: " + t.shape);
+                    if (0 === e.size) throw new Error("Updates specified for empty output. updates shape: " + e.shape)
+                }
+                co(n, t, e)
+            }
+
+            function po(e, t, n) {
+                const r = t.shape.length,
+                    a = r > 1 ? t.shape[r - 1] : 1,
+                    s = n.length;
+                let i = 1;
+                for (let e = a; e < s; ++e) i *= n[e];
+                const o = a < 1 ? 1 : a;
+                return {
+                    sliceRank: a,
+                    numUpdates: pe(t.shape) / o,
+                    sliceSize: i,
+                    strides: [...$e(n.slice(0, a)), 1],
+                    outputSize: pe(n)
+                }
+            }
+            w(lo, {
+                calculateShapes: () => po,
+                validateInput: () => ho,
+                validateUpdateShape: () => co
+            });
+            var fo = {};
+            w(fo, {
+                assertParamsValid: () => mo,
+                computeFlatOffset: () => Co,
+                computeOutShape: () => yo,
+                getNormalizedAxes: () => wo,
+                isSliceContinous: () => No,
+                maskToAxes: () => go,
+                parseSliceParams: () => To,
+                sliceInfo: () => Ro,
+                startForAxis: () => Ao,
+                startIndicesWithElidedDims: () => ko,
+                stopForAxis: () => Eo,
+                stopIndicesWithElidedDims: () => So,
+                stridesForAxis: () => Io,
+                stridesWithElidedDims: () => bo
+            });
+
+            function mo(e, t, n) {
+                const r = e.shape.length;
+                ue(r === t.length, () => `Error in slice${r}D: Length of begin ${t} must match the rank of the array (${r}).`), ue(r === n.length, () => `Error in slice${r}D: Length of size ${n} must match the rank of the array (${r}).`);
+                for (let a = 0; a < r; ++a) ue(t[a] + n[a] <= e.shape[a], () => `Error in slice${r}D: begin[${a}] + size[${a}] (${t[a]+n[a]}) would overflow input.shape[${a}] (${e.shape[a]})`)
+            }
+
+            function go(e) {
+                const t = [];
+                let n = 0;
+                for (; e > 0;) 1 & e && t.push(n), e /= 2, n++;
+                return t
+            }
+
+            function yo(e, t, n) {
+                const r = [];
+                for (let a = 0; a < e.length; a++) r[a] = Math.ceil((t[a] - e[a]) / n[a]);
+                return r
+            }
+
+            function bo(e, t, n, r) {
+                const a = [...e];
+                for (let e = a.length; e < r.length; e++) a.push(1);
+                for (let e = 0; e < n; e++) 0 === e ? a[t] = 1 : (a.splice(t, 0, 1), a.pop());
+                return a
+            }
+
+            function vo(e, t, n) {
+                return n <= e ? n : n - (t - 1)
+            }
+
+            function xo(e, t) {
+                const n = [];
+                for (let r = 0; r < e; r++) n.push(t + r);
+                return n
+            }
+
+            function wo(e, t, n, r, a, s, i, o, u) {
+                const l = e.length;
+                let c = new Array(l),
+                    h = new Array(l),
+                    p = new Array(l);
+                if (t.length && n > 0) {
+                    const u = t[0],
+                        l = n + 1;
+                    c = ko(i, u, l, r, e), h = So(o, u, l, a, e), p = bo(s, u, l, e)
+                } else
+                    for (let t = 0; t < l; t++) c[t] = Ao(i, r, s, e, t, u), h[t] = Eo(o, a, s, e, t, u), p[t] = Io(s, t, u);
+                return {
+                    begin: c,
+                    end: h,
+                    strides: p
+                }
+            }
+
+            function ko(e, t, n, r, a) {
+                const s = [...a],
+                    i = xo(n, t);
+                for (let a = 0; a < s.length; a++)
+                    if (i.indexOf(a) > -1) s[a] = 0;
+                    else {
+                        const i = vo(t, n, a);
+                        let o = r[i];
+                        e & 1 << i && (o = 0), s[a] = o
+                    } return s
+            }
+
+            function So(e, t, n, r, a) {
+                const s = [...a],
+                    i = xo(n, t);
+                for (let a = 0; a < s.length; a++)
+                    if (i.indexOf(a) > -1) s[a] = Number.MAX_SAFE_INTEGER;
+                    else {
+                        const i = vo(t, n, a);
+                        let o = r[i];
+                        e & 1 << i && (o = Number.MAX_SAFE_INTEGER), s[a] = o
+                    } for (let e = 0; e < s.length; e++) {
+                    const t = a[e];
+                    s[e] < 0 && (s[e] += t), s[e] = ne(0, s[e], a[e])
+                }
+                return s
+            }
+
+            function Io(e, t, n) {
+                let r = e[t];
+                return (n & 1 << t || null == r) && (r = 1), r
+            }
+
+            function Ao(e, t, n, r, a, s) {
+                let i = t[a];
+                const o = n[a] || 1;
+                (e & 1 << a || s & 1 << a || null == i) && (i = o > 0 ? Number.MIN_SAFE_INTEGER : Number.MAX_SAFE_INTEGER);
+                const u = r[a];
+                return i < 0 && (i += u), i = ne(0, i, u - 1), i
+            }
+
+            function Eo(e, t, n, r, a, s) {
+                let i = t[a];
+                const o = n[a] || 1;
+                (e & 1 << a || s & 1 << a || null == i) && (i = o > 0 ? Number.MAX_SAFE_INTEGER : Number.MIN_SAFE_INTEGER);
+                const u = r[a];
+                return i < 0 && (i += u), i = o > 0 ? ne(0, i, u) : ne(-1, i, u - 1), i
+            }
+
+            function No(e, t, n) {
+                let r = n.length;
+                for (let e = 0; e < n.length; e++)
+                    if (n[e] > 1) {
+                        r = e;
+                        break
+                    } for (let a = r + 1; a < n.length; a++)
+                    if (t[a] > 0 || n[a] !== e[a]) return !1;
+                return !0
+            }
+
+            function Co(e, t) {
+                let n = e.length > 0 ? e[e.length - 1] : 1;
+                for (let r = 0; r < e.length - 1; r++) n += e[r] * t[r];
+                return n
+            }
+
+            function To(e, t, n) {
+                let r;
+                const a = e.shape.length;
+                let s;
+                return r = "number" == typeof t ? [t, ...new Array(a - 1).fill(0)] : t.length < a ? t.concat(new Array(a - t.length).fill(0)) : t.slice(), r.forEach(e => {
+                    ue(-1 !== e, () => "slice() does not support negative begin indexing.")
+                }), s = null == n ? new Array(a).fill(-1) : "number" == typeof n ? [n, ...new Array(a - 1).fill(-1)] : n.length < a ? n.concat(new Array(a - n.length).fill(-1)) : n, s = s.map((t, n) => t >= 0 ? t : (ue(-1 === t, () => `Negative size values should be exactly -1 but got ${t} for the slice() size at index ${n}.`), e.shape[n] - r[n])), [r, s]
+            }
+
+            function Ro(e, t, n, r, a, s, i, o, u) {
+                let l;
+                if (null == r ? (l = new Array(t.length), l.fill(1)) : l = r, null != i && 0 != (i & i - 1)) throw new Error("Multiple ellipses in slice is not allowed.");
+                let c = !1;
+                const h = {
+                    dims: l.length,
+                    numAddAxisAfterEllipsis: 0,
+                    begin: t.slice(),
+                    end: n.slice(),
+                    strides: l.slice(),
+                    beginMask: a,
+                    endMask: s,
+                    ellipsisMask: i,
+                    newAxisMask: o,
+                    shrinkAxisMask: u
+                };
+                for (let e = 0; e < h.dims; e++) c && 0 != (1 << e & o) && h.numAddAxisAfterEllipsis++, 1 << e & i && (c = !0);
+                c || (h.ellipsisMask |= 1 << h.dims, h.dims++);
+                const p = {
+                    dims: e.length,
+                    beginMask: 0,
+                    endMask: 0,
+                    beginValid: !1,
+                    endValid: !1
+                };
+                ! function (e, t) {
+                    t.beginMask = 0, t.endMask = 0, t.shrinkAxisMask = 0;
+                    let n = 0;
+                    t.beginValid = null != e.begin, t.endValid = null != e.end, t.begin = new Array(t.dims), t.end = new Array(t.dims), t.strides = new Array(t.dims), t.finalShapeGatherIndices = [], t.finalShapeGatherIndicesSparse = [], t.inputShapeGatherIndicesSparse = new Array(t.dims);
+                    for (let r = 0; r < e.dims; r++)
+                        if (1 << r & e.ellipsisMask) {
+                            const a = Math.min(t.dims - (e.dims - r) + 1 + e.numAddAxisAfterEllipsis, t.dims);
+                            for (; n < a; n++) t.begin[n] = 0, t.end[n] = 0, t.strides[n] = 1, t.beginMask |= 1 << n, t.endMask |= 1 << n, t.finalShapeGatherIndices.push(n), t.finalShapeGatherIndicesSparse.push(-1), t.inputShapeGatherIndicesSparse[n] = r
+                        } else if (1 << r & e.newAxisMask) t.finalShapeGatherIndices.push(-2), t.finalShapeGatherIndicesSparse.push(-1);
+                    else {
+                        if (n === t.begin.length) throw Error(`Index out of range using input dim ${n}; input has only ${t.dims} dims, ${t.begin.length}.`);
+                        null != e.begin && (t.begin[n] = e.begin[r]), null != e.end && (t.end[n] = e.end[r]), t.strides[n] = e.strides[r], e.beginMask & 1 << r && (t.beginMask |= 1 << n), e.endMask & 1 << r && (t.endMask |= 1 << n), e.shrinkAxisMask & 1 << r ? (t.finalShapeGatherIndices.push(-1), t.finalShapeGatherIndicesSparse.push(-1), t.shrinkAxisMask |= 1 << n) : (t.finalShapeGatherIndices.push(n), t.finalShapeGatherIndicesSparse.push(r)), t.inputShapeGatherIndicesSparse[n] = r, n++
+                    }
+                }(h, p);
+                let d = !0,
+                    f = !0,
+                    m = !0;
+                const g = [],
+                    y = [];
+                for (let t = 0; t < e.length; ++t) {
+                    if (0 === p.strides[t]) throw Error(`strides[${t}] must be non-zero`);
+                    const n = !!(p.shrinkAxisMask & 1 << t),
+                        r = e[t];
+                    if (-1 === r) {
+                        g.push(n ? 1 : -1);
+                        continue
+                    }
+                    const a = [p.beginMask & 1 << t, p.endMask & 1 << t],
+                        s = [p.strides[t] > 0 ? 0 : -1, p.strides[t] > 0 ? r : r - 1];
+                    if (n && p.strides[t] <= 0) throw Error("only stride 1 allowed on non-range indexing.");
+                    m = m && 1 === p.strides[t];
+                    const i = !!(p.beginMask & 1 << t && p.endMask & 1 << t);
+                    if (p.beginValid && p.endValid) {
+                        if (n) {
+                            const e = p.begin[t] < 0 ? r + p.begin[t] : p.begin[t];
+                            if (p.begin[t] = e, p.end[t] = p.begin[t] + 1, e < 0 || e >= r) throw Error(`slice index ${p.begin[t]} of dimension ${t} out of bounds.`)
+                        } else p.begin[t] = _o(p.begin[t], 0, p.strides[t], r, a, s), p.end[t] = _o(p.end[t], 1, p.strides[t], r, a, s);
+                        const e = 1 === p.strides[t] && 0 === p.begin[t] && p.end[t] === r;
+                        d = d && e, f = f && (0 === t && 1 === p.strides[t] || e)
+                    } else d = d && 1 === p.strides[t] && i, f = f && (0 === t && 1 === p.strides[t] || i);
+                    let o, u = !1;
+                    if (p.beginValid && p.endValid ? (o = p.end[t] - p.begin[t], u = !0) : n ? (o = 1, u = !0) : i && r >= 0 && (o = p.strides[t] < 0 ? -r : r, u = !0), u) {
+                        let e;
+                        e = 0 === o || o < 0 != p.strides[t] < 0 ? 0 : Math.trunc(o / p.strides[t]) + (o % p.strides[t] != 0 ? 1 : 0), g.push(e)
+                    } else g.push(-1)
+                }
+                for (let e = 0; e < p.finalShapeGatherIndices.length; ++e) {
+                    const t = p.finalShapeGatherIndices[e];
+                    t >= 0 ? y.push(g[t]) : -2 === t && y.push(1)
+                }
+                return {
+                    finalShapeSparse: y.filter((e, t) => -2 !== p.finalShapeGatherIndices[t]),
+                    finalShape: y,
+                    isIdentity: d,
+                    sliceDim0: f,
+                    isSimpleSlice: m,
+                    begin: p.begin,
+                    end: p.end,
+                    strides: p.strides
+                }
+            }
+
+            function _o(e, t, n, r, a, s) {
+                if (a[t]) return n > 0 ? s[t] : s[t + 1 & 1]; {
+                    const t = e < 0 ? r + e : e;
+                    return t < s[0] ? s[0] : t > s[1] ? s[1] : t
+                }
+            }
+            var Fo = {};
+            w(Fo, {
+                Serializable: () => Mo,
+                SerializationMap: () => Do,
+                registerClass: () => Oo
+            });
+            var Mo = class {
+                    getClassName() {
+                        return this.constructor.className
+                    }
+                    static fromConfig(e, t) {
+                        return new e(t)
+                    }
+                },
+                Do = class {
+                    constructor() {
+                        this.classNameMap = {}
+                    }
+                    static getMap() {
+                        return null == Do.instance && (Do.instance = new Do), Do.instance
+                    }
+                    static register(e) {
+                        Do.getMap().classNameMap[e.className] = [e, e.fromConfig]
+                    }
+                };
+
+            function Oo(e) {
+                ue(null != e.className, () => "Class being registered does not have the static className property defined."), ue("string" == typeof e.className, () => "className is required to be a string, but got type " + typeof e.className), ue(e.className.length > 0, () => "Class being registered has an empty-string as its className, which is disallowed."), Do.register(e)
+            }
+            var Lo = {};
+            w(Lo, {
+                TEST_EPSILON_FLOAT16: () => Po,
+                encodeStrings: () => qo,
+                expectArrayBuffersEqual: () => Ho,
+                expectArraysClose: () => $o,
+                expectArraysEqual: () => Wo,
+                expectNumbersClose: () => Vo,
+                expectPromiseToFail: () => Uo,
+                expectValuesInRange: () => Go,
+                testEpsilon: () => Bo
+            });
+            var Po = .1;
+
+            function $o(e, t, n) {
+                return null == n && (n = Bo()), zo(e, t, (e, t) => jo(e, t, n))
+            }
+
+            function Bo() {
+                return 32 === xs.backend.floatPrecision() ? .001 : Po
+            }
+
+            function zo(e, t, n) {
+                let r = !0;
+                if ((Te(e) || Te(t)) && (r = !1), Te(e) && Te(t) && (r = !0), r) {
+                    const n = e.constructor.name,
+                        r = t.constructor.name;
+                    if (n !== r) throw new Error(`Arrays are of different type. Actual: ${n}. Expected: ${r}`)
+                }
+                if (Array.isArray(e) && Array.isArray(t)) {
+                    const n = Cs(e),
+                        r = Cs(t);
+                    if (!fe(n, r)) throw new Error(`Arrays have different shapes. Actual: [${n}]. Expected: [${r}]`)
+                }
+                const a = Te(e) ? e : he(e),
+                    s = Te(t) ? t : he(t);
+                if (a.length !== s.length) throw new Error(`Arrays have different lengths actual: ${a.length} vs expected: ${s.length}.\nActual:   ${a}.\nExpected: ${s}.`);
+                for (let e = 0; e < s.length; ++e) {
+                    const t = a[e],
+                        r = s[e];
+                    if (!n(t, r)) throw new Error(`Arrays differ: actual[${e}] = ${t}, expected[${e}] = ${r}.\nActual:   ${a}.\nExpected: ${s}.`)
+                }
+            }
+
+            function Uo(e, t) {
+                e().then(() => t.fail(), () => t())
+            }
+
+            function Wo(e, t) {
+                const n = "string" == typeof t || "number" == typeof t || "boolean" == typeof t ? [t] : t;
+                return Fe(e) || Fe(e[0]) || Fe(t) || Fe(t[0]) ? zo(e, n, (e, t) => e == t) : zo(e, t, (e, t) => jo(e, t, 0))
+            }
+
+            function Vo(e, t, n) {
+                if (null == n && (n = Bo()), !jo(e, t, n)) throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`)
+            }
+
+            function jo(e, t, n) {
+                return !isFinite(e) && !isFinite(t) || !(isNaN(e) || isNaN(t) || Math.abs(e - t) > n)
+            }
+
+            function Go(e, t, n) {
+                for (let r = 0; r < e.length; r++)
+                    if (e[r] < t || e[r] > n) throw new Error(`Value out of range:${e[r]} low: ${t}, high: ${n}`)
+            }
+
+            function Ho(e, t) {
+                const n = new Float32Array(e),
+                    r = new Float32Array(t);
+                if (n.length !== r.length) throw new Error(`Expected ArrayBuffer to be of length ${r.length}, but it was ${n.length}`);
+                for (let e = 0; e < r.length; e++)
+                    if (n[e] !== r[e]) throw new Error(`Expected ArrayBuffer value at ${e} to be ${r[e]} but got ${n[e]} instead`)
+            }
+
+            function qo(e) {
+                for (let t = 0; t < e.length; t++) {
+                    const n = e[t];
+                    Array.isArray(n) ? qo(n) : e[t] = $a(n)
+                }
+                return e
+            }
+            var Ko = "3.13.0";
+
+            function Yo() {
+                Ye().set("PROD", !0)
+            }
+
+            function Xo() {
+                Ye().set("DEBUG", !0)
+            }
+
+            function Qo() {
+                Ye().set("DEPRECATION_WARNINGS_ENABLED", !1), console.warn("TensorFlow.js deprecation warnings have been disabled.")
+            }
+
+            function Zo(e) {
+                Ye().getBool("DEPRECATION_WARNINGS_ENABLED") && console.warn(e + " You can disable deprecation warnings with tf.disableDeprecationWarnings().")
+            }
+
+            function Jo() {
+                xs.disposeVariables()
+            }
+
+            function eu() {
+                return xs
+            }
+
+            function tu() {
+                return xs.memory()
+            }
+
+            function nu(e) {
+                return xs.profile(e)
+            }
+
+            function ru(e, t) {
+                return xs.tidy(e, t)
+            }
+
+            function au(e) {
+                ms(e).forEach(e => e.dispose())
+            }
+
+            function su(e) {
+                return xs.keep(e)
+            }
+
+            function iu(e) {
+                return xs.time(e)
+            }
+
+            function ou(e) {
+                return xs.setBackend(e)
+            }
+
+            function uu() {
+                return xs.ready()
+            }
+
+            function lu() {
+                return xs.backendName
+            }
+
+            function cu(e) {
+                xs.removeBackend(e)
+            }
+
+            function hu(e) {
+                return xs.findBackend(e)
+            }
+
+            function pu(e) {
+                return xs.findBackendFactory(e)
+            }
+
+            function du(e, t, n = 1) {
+                return xs.registerBackend(e, t, n)
+            }
+
+            function fu() {
+                return xs.backend
+            }
+
+            function mu(e, t) {
+                Ye().setPlatform(e, t)
+            }
+            var gu = Ms({
+                add_: function (e, t) {
+                    let n = Rs(e, "a", "add"),
+                        r = Rs(t, "b", "add");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(rt, a)
+                }
+            });
+            var yu = Ms({
+                floorDiv_: function (e, t) {
+                    let n = Rs(e, "a", "floorDiv"),
+                        r = Rs(t, "b", "floorDiv");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(un, a)
+                }
+            });
+            var bu = Ms({
+                div_: function (e, t) {
+                    let n = Rs(e, "a", "div"),
+                        r = Rs(t, "b", "div");
+                    if ([n, r] = ps(n, r), "int32" === n.dtype && "int32" === r.dtype) return yu(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Kt, a, {})
+                }
+            });
+            var vu = Ms({
+                mul_: function (e, t) {
+                    let n = Rs(e, "a", "mul"),
+                        r = Rs(t, "b", "mul");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Gn, a)
+                }
+            });
+            var xu = Ms({
+                abs_: function (e) {
+                    const t = Rs(e, "x", "abs");
+                    if ("complex64" === t.dtype) {
+                        const e = {
+                            x: t
+                        };
+                        return xs.runKernel(Nt, e)
+                    } {
+                        const e = {
+                            x: t
+                        };
+                        return xs.runKernel(et, e)
+                    }
+                }
+            });
+            var wu = Ms({
+                acos_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "acos")
+                    };
+                    return xs.runKernel(tt, t)
+                }
+            });
+            var ku = Ms({
+                acosh_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "acosh")
+                    };
+                    return xs.runKernel(nt, t)
+                }
+            });
+            var Su = Ms({
+                addN_: function (e) {
+                    ue(Array.isArray(e), () => "The argument passed to tf.addN() must be a list of tensors"), ue(e.length >= 1, () => "Must pass at least one tensor to tf.addN(), but got " + e.length);
+                    const t = e.map((e, t) => Rs(e, "tensors" + t, "addN")),
+                        n = t[0];
+                    t.forEach(e => {
+                        if (e.dtype !== n.dtype) throw new Error("All tensors passed to tf.addN() must have the same dtype")
+                    }), t.forEach(e => {
+                        if (!fe(e.shape, n.shape)) throw new Error("All tensors passed to tf.addN() must have the same shape")
+                    });
+                    const r = t;
+                    return xs.runKernel(at, r)
+                }
+            });
+            var Iu = Ms({
+                all_: function (e, t = null, n = !1) {
+                    const r = {
+                            x: Rs(e, "x", "all", "bool")
+                        },
+                        a = {
+                            axis: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(st, r, a)
+                }
+            });
+            var Au = Ms({
+                any_: function (e, t = null, n = !1) {
+                    const r = {
+                            x: Rs(e, "x", "any", "bool")
+                        },
+                        a = {
+                            axis: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(it, r, a)
+                }
+            });
+            var Eu = Ms({
+                argMax_: function (e, t = 0) {
+                    const n = {
+                            x: Rs(e, "x", "argMax")
+                        },
+                        r = {
+                            axis: t
+                        };
+                    return xs.runKernel(ot, n, r)
+                }
+            });
+            var Nu = Ms({
+                argMin_: function (e, t = 0) {
+                    const n = {
+                            x: Rs(e, "x", "argMin")
+                        },
+                        r = {
+                            axis: t
+                        };
+                    return xs.runKernel(ut, n, r)
+                }
+            });
+            var Cu = Ms({
+                asin_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "asin")
+                    };
+                    return xs.runKernel(lt, t)
+                }
+            });
+            var Tu = Ms({
+                asinh_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "asinh")
+                    };
+                    return xs.runKernel(ct, t)
+                }
+            });
+            var Ru = Ms({
+                atan_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "atan")
+                    };
+                    return xs.runKernel(ht, t)
+                }
+            });
+            var _u = Ms({
+                atan2_: function (e, t) {
+                    let n = Rs(e, "a", "atan2"),
+                        r = Rs(t, "b", "atan2");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(dt, a)
+                }
+            });
+            var Fu = Ms({
+                atanh_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "atanh")
+                    };
+                    return xs.runKernel(pt, t)
+                }
+            });
+
+            function Mu(e, t, n, r, a = "NHWC", s) {
+                return Lu(e, [...t, e[3]], n, s, r, null, null, Gu(a))
+            }
+
+            function Du(e, t, n, r, a, s, i = "channelsLast") {
+                const [o, u] = Bu(t);
+                let l;
+                if ("channelsLast" === i) l = [o, u, e[3], e[3]];
+                else {
+                    if ("channelsFirst" !== i) throw new Error("Unknown dataFormat " + i);
+                    l = [o, u, e[1], e[1]]
+                }
+                return Lu(e, l, n, r, a, s, !1, i)
+            }
+
+            function Ou(e, t, n, r, a, s, i = "NDHWC") {
+                const [o, u, l] = zu(t);
+                let c, h;
+                if ("NDHWC" === i) h = "channelsLast", c = [o, u, l, e[4], e[4]];
+                else {
+                    if ("NCDHW" !== i) throw new Error("Unknown dataFormat " + i);
+                    h = "channelsFirst", c = [o, u, l, e[1], e[1]]
+                }
+                return Pu(e, c, n, r, a, !1, h, s)
+            }
+
+            function Lu(e, t, n, r, a, s, i = !1, o = "channelsLast") {
+                let [u, l, c, h] = [-1, -1, -1, -1];
+                if ("channelsLast" === o)[u, l, c, h] = e;
+                else {
+                    if ("channelsFirst" !== o) throw new Error("Unknown dataFormat " + o);
+                    [u, h, l, c] = e
+                }
+                const [p, d, , f] = t, [m, g] = Bu(n), [y, b] = Bu(r), v = Uu(p, y), x = Uu(d, b), {
+                    padInfo: w,
+                    outHeight: k,
+                    outWidth: S
+                } = function (e, t, n, r, a, s, i, o, u) {
+                    let l, c, h;
+                    if ("number" == typeof e) {
+                        l = {
+                            top: e,
+                            bottom: e,
+                            left: e,
+                            right: e,
+                            type: 0 === e ? "VALID" : "NUMBER"
+                        };
+                        const a = function (e, t, n, r, a) {
+                            null == r && (r = $u(e, t, n));
+                            const s = e[0],
+                                i = e[1],
+                                o = Wu((s - t + 2 * r) / n + 1, a),
+                                u = Wu((i - t + 2 * r) / n + 1, a);
+                            return [o, u]
+                        }([t, n], s, r, e, o);
+                        c = a[0], h = a[1]
+                    } else if ("same" === e) {
+                        c = Math.ceil(t / r), h = Math.ceil(n / a);
+                        const e = Math.max(0, (c - 1) * r + s - t),
+                            o = Math.max(0, (h - 1) * a + i - n),
+                            u = Math.floor(e / 2),
+                            p = e - u,
+                            d = Math.floor(o / 2);
+                        l = {
+                            top: u,
+                            bottom: p,
+                            left: d,
+                            right: o - d,
+                            type: "SAME"
+                        }
+                    } else if ("valid" === e) l = {
+                        top: 0,
+                        bottom: 0,
+                        left: 0,
+                        right: 0,
+                        type: "VALID"
+                    }, c = Math.ceil((t - s + 1) / r), h = Math.ceil((n - i + 1) / a);
+                    else {
+                        if ("object" != typeof e) throw Error("Unknown padding parameter: " + e); {
+                            const p = "channelsLast" === u ? e[1][0] : e[2][0],
+                                d = "channelsLast" === u ? e[1][1] : e[2][1],
+                                f = "channelsLast" === u ? e[2][0] : e[3][0],
+                                m = "channelsLast" === u ? e[2][1] : e[3][1];
+                            l = {
+                                top: p,
+                                bottom: d,
+                                left: f,
+                                right: m,
+                                type: 0 === p && 0 === d && 0 === f && 0 === m ? "VALID" : "EXPLICIT"
+                            }, c = Wu((t - s + p + d) / r + 1, o), h = Wu((n - i + f + m) / a + 1, o)
+                        }
+                    }
+                    return {
+                        padInfo: l,
+                        outHeight: c,
+                        outWidth: h
+                    }
+                }(a, l, c, m, g, v, x, s, o), I = i ? f * h : f;
+                let A;
+                return "channelsFirst" === o ? A = [u, I, k, S] : "channelsLast" === o && (A = [u, k, S, I]), {
+                    batchSize: u,
+                    dataFormat: o,
+                    inHeight: l,
+                    inWidth: c,
+                    inChannels: h,
+                    outHeight: k,
+                    outWidth: S,
+                    outChannels: I,
+                    padInfo: w,
+                    strideHeight: m,
+                    strideWidth: g,
+                    filterHeight: p,
+                    filterWidth: d,
+                    effectiveFilterHeight: v,
+                    effectiveFilterWidth: x,
+                    dilationHeight: y,
+                    dilationWidth: b,
+                    inShape: e,
+                    outShape: A,
+                    filterShape: t
+                }
+            }
+
+            function Pu(e, t, n, r, a, s = !1, i = "channelsLast", o) {
+                let [u, l, c, h, p] = [-1, -1, -1, -1, -1];
+                if ("channelsLast" === i)[u, l, c, h, p] = e;
+                else {
+                    if ("channelsFirst" !== i) throw new Error("Unknown dataFormat " + i);
+                    [u, p, l, c, h] = e
+                }
+                const [d, f, m, , g] = t, [y, b, v] = zu(n), [x, w, k] = zu(r), S = Uu(d, x), I = Uu(f, w), A = Uu(m, k), {
+                    padInfo: E,
+                    outDepth: N,
+                    outHeight: C,
+                    outWidth: T
+                } = function (e, t, n, r, a, s, i, o, u, l, c) {
+                    let h, p, d, f;
+                    if ("number" == typeof e) {
+                        h = {
+                            top: e,
+                            bottom: e,
+                            left: e,
+                            right: e,
+                            front: e,
+                            back: e,
+                            type: 0 === e ? "VALID" : "NUMBER"
+                        };
+                        const s = function (e, t, n, r, a, s) {
+                            null == a && (a = $u(e, t, r));
+                            const i = e[0],
+                                o = e[1],
+                                u = e[2],
+                                l = Wu((i - t + 2 * a) / r + 1, s),
+                                c = Wu((o - t + 2 * a) / r + 1, s),
+                                h = Wu((u - t + 2 * a) / r + 1, s);
+                            return [l, c, h, n]
+                        }([t, n, r, 1], o, 1, a, e, c);
+                        p = s[0], d = s[1], f = s[2]
+                    } else if ("same" === e) {
+                        p = Math.ceil(t / a), d = Math.ceil(n / s), f = Math.ceil(r / i);
+                        const e = (p - 1) * a + o - t,
+                            c = (d - 1) * s + u - n,
+                            m = (f - 1) * i + l - r,
+                            g = Math.floor(e / 2),
+                            y = e - g,
+                            b = Math.floor(c / 2),
+                            v = c - b,
+                            x = Math.floor(m / 2);
+                        h = {
+                            top: b,
+                            bottom: v,
+                            left: x,
+                            right: m - x,
+                            front: g,
+                            back: y,
+                            type: "SAME"
+                        }
+                    } else {
+                        if ("valid" !== e) throw Error("Unknown padding parameter: " + e);
+                        h = {
+                            top: 0,
+                            bottom: 0,
+                            left: 0,
+                            right: 0,
+                            front: 0,
+                            back: 0,
+                            type: "VALID"
+                        }, p = Math.ceil((t - o + 1) / a), d = Math.ceil((n - u + 1) / s), f = Math.ceil((r - l + 1) / i)
+                    }
+                    return {
+                        padInfo: h,
+                        outDepth: p,
+                        outHeight: d,
+                        outWidth: f
+                    }
+                }(a, l, c, h, y, b, v, S, I, A, o), R = s ? g * p : g;
+                let _;
+                return "channelsFirst" === i ? _ = [u, R, N, C, T] : "channelsLast" === i && (_ = [u, N, C, T, R]), {
+                    batchSize: u,
+                    dataFormat: i,
+                    inDepth: l,
+                    inHeight: c,
+                    inWidth: h,
+                    inChannels: p,
+                    outDepth: N,
+                    outHeight: C,
+                    outWidth: T,
+                    outChannels: R,
+                    padInfo: E,
+                    strideDepth: y,
+                    strideHeight: b,
+                    strideWidth: v,
+                    filterDepth: d,
+                    filterHeight: f,
+                    filterWidth: m,
+                    effectiveFilterDepth: S,
+                    effectiveFilterHeight: I,
+                    effectiveFilterWidth: A,
+                    dilationDepth: x,
+                    dilationHeight: w,
+                    dilationWidth: k,
+                    inShape: e,
+                    outShape: _,
+                    filterShape: t
+                }
+            }
+
+            function $u(e, t, n, r = 1) {
+                const a = Uu(t, r);
+                return Math.floor((e[0] * (n - 1) - n + a) / 2)
+            }
+
+            function Bu(e) {
+                return "number" == typeof e ? [e, e, e] : 2 === e.length ? [e[0], e[1], 1] : e
+            }
+
+            function zu(e) {
+                return "number" == typeof e ? [e, e, e] : e
+            }
+
+            function Uu(e, t) {
+                return t <= 1 ? e : e + (e - 1) * (t - 1)
+            }
+
+            function Wu(e, t) {
+                if (!t) return Math.trunc(e);
+                switch (t) {
+                    case "round":
+                        return Math.round(e);
+                    case "ceil":
+                        return Math.ceil(e);
+                    case "floor":
+                        return Math.floor(e);
+                    default:
+                        throw new Error("Unknown roundingMode " + t)
+                }
+            }
+
+            function Vu(e) {
+                const [t, n, r] = Bu(e);
+                return 1 === t && 1 === n && 1 === r
+            }
+
+            function ju(e, t) {
+                return Vu(e) || Vu(t)
+            }
+
+            function Gu(e) {
+                if ("NHWC" === e) return "channelsLast";
+                if ("NCHW" === e) return "channelsFirst";
+                throw new Error("Unknown dataFormat " + e)
+            }
+
+            function Hu(e, t, n) {
+                if (null != n) {
+                    if ("string" == typeof t) throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);
+                    if ("number" == typeof t) ue(me(t), () => `Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);
+                    else {
+                        if ("object" != typeof t) throw Error(`Error in ${e}: Unknown padding parameter: ${t}`);
+                        t.forEach(t => {
+                            t.forEach(t => {
+                                ue(me(t), () => `Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`)
+                            })
+                        })
+                    }
+                }
+            }
+            var qu = Ms({
+                reshape_: function (e, t) {
+                    const n = {
+                            x: Rs(e, "x", "reshape", "string_or_numeric")
+                        },
+                        r = {
+                            shape: t
+                        };
+                    return xs.runKernel(lr, n, r)
+                }
+            });
+            var Ku = Ms({
+                avgPool_: function (e, t, n, r, a) {
+                    const s = Rs(e, "x", "avgPool", "float32");
+                    ue(ju(n, 1), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '1'`);
+                    let i = s,
+                        o = !1;
+                    3 === s.rank && (o = !0, i = qu(s, [1, s.shape[0], s.shape[1], s.shape[2]])), ue(4 === i.rank, () => `Error in avgPool: x must be rank 4 but got rank ${i.rank}.`), Hu("avgPool", r, a);
+                    const u = {
+                            x: i
+                        },
+                        l = {
+                            filterSize: t,
+                            strides: n,
+                            pad: r,
+                            dimRoundingMode: a
+                        };
+                    let c = xs.runKernel(ft, u, l);
+                    return c = Ai(c, s.dtype), o ? qu(c, [c.shape[1], c.shape[2], c.shape[3]]) : c
+                }
+            });
+            var Yu = Ms({
+                avgPool3d_: function (e, t, n, r, a, s = "NDHWC") {
+                    const i = Rs(e, "x", "avgPool3d", "float32");
+                    let o = i,
+                        u = !1;
+                    4 === i.rank && (u = !0, o = qu(i, [1, i.shape[0], i.shape[1], i.shape[2], i.shape[3]])), ue(5 === o.rank, () => `Error in avgPool3d: x must be rank 5 but got rank ${o.rank}.`), ue("NDHWC" === s, () => "Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of " + s), Hu("avgPool3d", r, a);
+                    const l = {
+                            x: o
+                        },
+                        c = {
+                            filterSize: t,
+                            strides: n,
+                            pad: r,
+                            dimRoundingMode: a,
+                            dataFormat: s
+                        };
+                    let h = xs.runKernel(gt, l, c);
+                    return h = Ai(h, o.dtype), u ? qu(h, [h.shape[1], h.shape[2], h.shape[3], h.shape[4]]) : h
+                }
+            });
+            var Xu = Ms({
+                concat_: function (e, t = 0) {
+                    ue(e.length >= 1, () => "Pass at least one tensor to concat");
+                    const n = _s(e, "tensors", "concat", "string_or_numeric");
+                    if ("complex64" === n[0].dtype && n.forEach(e => {
+                            if ("complex64" !== e.dtype) throw new Error(`Cannot concatenate complex64 tensors with a tensor\n          with dtype ${e.dtype}. `)
+                        }), 1 === n.length) return Ei(n[0]);
+                    const r = n,
+                        a = {
+                            axis: t
+                        };
+                    return xs.runKernel(Ct, r, a)
+                }
+            });
+            var Qu = Ms({
+                sigmoid_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "sigmoid", "float32")
+                    };
+                    return xs.runKernel(Ar, t)
+                }
+            });
+            var Zu = Ms({
+                slice_: function (e, t, n) {
+                    const r = Rs(e, "x", "slice", "string_or_numeric");
+                    if (0 === r.rank) throw new Error("Slicing scalar is not possible");
+                    const a = {
+                            x: r
+                        },
+                        s = {
+                            begin: t,
+                            size: n
+                        };
+                    return xs.runKernel(wr, a, s)
+                }
+            });
+            var Ju = Ms({
+                tanh_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "tanh", "float32")
+                    };
+                    return xs.runKernel(Gr, t)
+                }
+            });
+            var el = Ms({
+                basicLSTMCell_: function (e, t, n, r, a, s) {
+                    const i = Rs(e, "forgetBias", "basicLSTMCell"),
+                        o = Rs(t, "lstmKernel", "basicLSTMCell"),
+                        u = Rs(n, "lstmBias", "basicLSTMCell"),
+                        l = Rs(r, "data", "basicLSTMCell"),
+                        c = Rs(a, "c", "basicLSTMCell"),
+                        h = Rs(s, "h", "basicLSTMCell"),
+                        p = Xu([l, h], 1),
+                        d = Gi(p, o),
+                        f = gu(d, u),
+                        m = f.shape[0],
+                        g = f.shape[1] / 4,
+                        y = [m, g],
+                        b = Zu(f, [0, 0], y),
+                        v = Zu(f, [0, g], y),
+                        x = Zu(f, [0, 2 * g], y),
+                        w = Zu(f, [0, 3 * g], y),
+                        k = gu(vu(Qu(b), Ju(v)), vu(c, Qu(gu(i, x))));
+                    return [k, vu(Ju(k), Qu(w))]
+                }
+            });
+            var tl = Ms({
+                batchToSpaceND_: function (e, t, n) {
+                    const r = Rs(e, "x", "batchToSpaceND"),
+                        a = t.reduce((e, t) => e * t);
+                    ue(r.rank >= 1 + t.length, () => `input rank is ${r.rank} but should be > than blockShape.length ${t.length}`), ue(n.length === t.length, () => `crops.length is ${n.length} but should be equal to blockShape.length  ${t.length}`), ue(r.shape[0] % a == 0, () => `input tensor batch is ${r.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${a}`);
+                    const s = {
+                            x: r
+                        },
+                        i = {
+                            blockShape: t,
+                            crops: n
+                        };
+                    return xs.runKernel(vt, s, i)
+                }
+            });
+            var nl = Ms({
+                batchNorm_: function (e, t, n, r, a, s) {
+                    null == s && (s = .001);
+                    const i = Rs(e, "x", "batchNorm"),
+                        o = Rs(t, "mean", "batchNorm"),
+                        u = Rs(n, "variance", "batchNorm");
+                    let l, c;
+                    null != a && (l = Rs(a, "scale", "batchNorm")), null != r && (c = Rs(r, "offset", "batchNorm")), ue(o.rank === u.rank, () => "Batch normalization gradient requires mean and variance to have equal ranks."), ue(null == c || o.rank === c.rank, () => "Batch normalization gradient requires mean and offset to have equal ranks."), ue(null == l || o.rank === l.rank, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+                    const h = {
+                            x: function (e) {
+                                let t;
+                                return t = 0 === e.rank || 1 === e.rank ? qu(e, [1, 1, 1, e.size]) : 2 === e.rank ? qu(e, [1, 1, e.shape[0], e.shape[1]]) : 3 === e.rank ? qu(e, [1, e.shape[0], e.shape[1], e.shape[2]]) : e, t
+                            }(i),
+                            scale: l,
+                            offset: c,
+                            mean: o,
+                            variance: u
+                        },
+                        p = {
+                            varianceEpsilon: s
+                        },
+                        d = xs.runKernel(ln, h, p);
+                    return qu(d, i.shape)
+                }
+            });
+            var rl = Ms({
+                batchNorm2d_: function (e, t, n, r, a, s) {
+                    const i = Rs(e, "x", "batchNorm"),
+                        o = Rs(t, "mean", "batchNorm"),
+                        u = Rs(n, "variance", "batchNorm");
+                    let l, c;
+                    return null != a && (l = Rs(a, "scale", "batchNorm")), null != r && (c = Rs(r, "offset", "batchNorm")), ue(2 === i.rank, () => `Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`), ue(2 === o.rank || 1 === o.rank, () => `Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${o.rank}.`), ue(2 === u.rank || 1 === u.rank, () => `Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${u.rank}.`), null != l && ue(2 === l.rank || 1 === l.rank, () => `Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${l.rank}.`), null != c && ue(2 === c.rank || 1 === c.rank, () => `Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${c.rank}.`), nl(i, o, u, c, l, s)
+                }
+            });
+            var al = Ms({
+                batchNorm3d_: function (e, t, n, r, a, s) {
+                    const i = Rs(e, "x", "batchNorm"),
+                        o = Rs(t, "mean", "batchNorm"),
+                        u = Rs(n, "variance", "batchNorm");
+                    let l, c;
+                    return null != a && (l = Rs(a, "scale", "batchNorm")), null != r && (c = Rs(r, "offset", "batchNorm")), ue(3 === i.rank, () => `Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`), ue(3 === o.rank || 1 === o.rank, () => `Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${o.rank}.`), ue(3 === u.rank || 1 === u.rank, () => `Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${u.rank}.`), null != l && ue(3 === l.rank || 1 === l.rank, () => `Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${l.rank}.`), null != c && ue(3 === c.rank || 1 === c.rank, () => `Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${c.rank}.`), nl(i, o, u, c, l, s)
+                }
+            });
+            var sl = Ms({
+                batchNorm4d_: function (e, t, n, r, a, s) {
+                    const i = Rs(e, "x", "batchNorm"),
+                        o = Rs(t, "mean", "batchNorm"),
+                        u = Rs(n, "variance", "batchNorm");
+                    let l, c;
+                    return null != a && (l = Rs(a, "scale", "batchNorm")), null != r && (c = Rs(r, "offset", "batchNorm")), ue(4 === i.rank, () => `Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`), ue(4 === o.rank || 1 === o.rank, () => `Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${o.rank}.`), ue(4 === u.rank || 1 === u.rank, () => `Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${u.rank}.`), null != l && ue(4 === l.rank || 1 === l.rank, () => `Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${l.rank}.`), null != c && ue(4 === c.rank || 1 === c.rank, () => `Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${c.rank}.`), nl(i, o, u, c, l, s)
+                }
+            });
+            var il = Ms({
+                bincount_: function (e, t, n) {
+                    const r = Rs(e, "x", "bincount"),
+                        a = Rs(t, "weights", "bincount");
+                    ue("int32" === r.dtype, () => "Error in bincount: input dtype must be int32, but got " + r.dtype), ue(n >= 0, () => `size must be non-negative, but got ${n}.`), ue(a.size === r.size || 0 === a.size, () => `Error in bincount: weights must have the same size as input or0-length, but got input shape: ${r.shape}, weights shape: ${a.shape}.`);
+                    const s = {
+                            x: r,
+                            weights: a
+                        },
+                        i = {
+                            size: n
+                        };
+                    return xs.runKernel(xt, s, i)
+                }
+            });
+            var ol = Ms({
+                broadcastArgs_: function (e, t) {
+                    const n = Rs(e, "s0", "broadcastArgs", "int32"),
+                        r = Rs(t, "s1", "broadcastArgs", "int32");
+                    if (1 !== n.rank) throw new Error("broadcastArgs(): first input must be a vector (rank=1). Has rank " + n.rank);
+                    if (1 !== r.rank) throw new Error("broadcastArgs(): second input must be a vector (rank=1). Has rank " + r.rank);
+                    const a = {
+                        s0: n,
+                        s1: r
+                    };
+                    return xs.runKernel(kt, a)
+                }
+            });
+            var ul = Ms({
+                broadcastTo_: function (e, t) {
+                    let n = Rs(e, "broadcastTo", "x");
+                    const r = n.shape;
+                    if (t.some(e => !(e > 0) || e % 1 != 0)) throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);
+                    if (t.length < n.rank) throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);
+                    if (t.length > n.rank) {
+                        const e = n.shape.slice();
+                        for (; e.length < t.length;) e.unshift(1);
+                        n = qu(n, e)
+                    }
+                    const a = n.shape,
+                        s = Array.from(t);
+                    for (let e = t.length - 1; e >= 0; e--)
+                        if (a[e] === t[e]) s[e] = 1;
+                        else if (1 !== n.shape[e]) throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${t}].`);
+                    if (0 === s.map((e, t) => e > 1 ? t : -1).filter(e => e >= 0).length) return Ei(n);
+                    const i = {
+                            x: n
+                        },
+                        o = {
+                            reps: s
+                        };
+                    return xs.runKernel(Hr, i, o)
+                }
+            });
+            var ll = Ms({
+                ceil_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "ceil", "float32")
+                    };
+                    return xs.runKernel(It, t)
+                }
+            });
+            var cl = Ms({
+                clipByValue_: function (e, t, n) {
+                    const r = Rs(e, "x", "clipByValue");
+                    ue(t <= n, () => `Error in clip: min (${t}) must be less than or equal to max (${n}).`);
+                    const a = {
+                            x: r
+                        },
+                        s = {
+                            clipValueMin: t,
+                            clipValueMax: n
+                        };
+                    return xs.runKernel(At, a, s)
+                }
+            });
+            var hl = Ms({
+                concat1d_: function (e) {
+                    return Xu(e, 0)
+                }
+            });
+            var pl = Ms({
+                concat2d_: function (e, t) {
+                    return Xu(e, t)
+                }
+            });
+            var dl = Ms({
+                concat3d_: function (e, t) {
+                    return Xu(e, t)
+                }
+            });
+            var fl = Ms({
+                concat4d_: function (e, t) {
+                    return Xu(e, t)
+                }
+            });
+            var ml = Ms({
+                conv2d_: function (e, t, n, r, a = "NHWC", s = [1, 1], i) {
+                    const o = Rs(e, "x", "conv2d", "float32"),
+                        u = Rs(t, "filter", "conv2d", "float32");
+                    let l = o,
+                        c = !1;
+                    3 === o.rank && (c = !0, l = qu(o, [1, o.shape[0], o.shape[1], o.shape[2]])), ue(4 === l.rank, () => `Error in conv2d: input must be rank 4, but got rank ${l.rank}.`), ue(4 === u.rank, () => `Error in conv2d: filter must be rank 4, but got rank ${u.rank}.`), Hu("conv2d", r, i);
+                    const h = "NHWC" === a ? l.shape[3] : l.shape[1];
+                    ue(h === u.shape[2], () => `Error in conv2d: depth of input (${h}) must match input depth for filter ${u.shape[2]}.`), ue(ju(n, s), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);
+                    const p = {
+                            x: l,
+                            filter: u
+                        },
+                        d = {
+                            strides: n,
+                            pad: r,
+                            dataFormat: a,
+                            dilations: s,
+                            dimRoundingMode: i
+                        },
+                        f = xs.runKernel(Tt, p, d);
+                    return c ? qu(f, [f.shape[1], f.shape[2], f.shape[3]]) : f
+                }
+            });
+            var gl = Ms({
+                conv1d_: function (e, t, n, r, a = "NWC", s = 1, i) {
+                    const o = Rs(e, "x", "conv1d"),
+                        u = Rs(t, "filter", "conv1d");
+                    let l = o,
+                        c = !1;
+                    2 === o.rank && (c = !0, l = qu(o, [1, o.shape[0], o.shape[1]])), ue(3 === l.rank, () => `Error in conv1d: input must be rank 3, but got rank ${l.rank}.`), ue(3 === u.rank, () => `Error in conv1d: filter must be rank 3, but got rank ${u.rank}.`), Hu("conv1d", r, i), ue(l.shape[2] === u.shape[1], () => `Error in conv1d: depth of input (${l.shape[2]}) must match input depth for filter ${u.shape[1]}.`), ue(ju(n, s), () => `Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${s}'`), ue("NWC" === a, () => `Error in conv1d: got dataFormat of ${a} but only NWC is currently supported.`);
+                    const h = qu(u, [1, u.shape[0], u.shape[1], u.shape[2]]),
+                        p = qu(l, [l.shape[0], 1, l.shape[1], l.shape[2]]),
+                        d = ml(p, h, [1, n], r, "NHWC", [1, s], i);
+                    return qu(d, c ? [d.shape[2], d.shape[3]] : [d.shape[0], d.shape[2], d.shape[3]])
+                }
+            });
+            var yl = Ms({
+                conv2DBackpropInput_: function (e, t, n, r, a, s = "NHWC", i) {
+                    ue(e.length === t.rank, () => `Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);
+                    let o = e,
+                        u = t,
+                        l = !1;
+                    3 === t.rank && (l = !0, u = qu(t, [1, t.shape[0], t.shape[1], t.shape[2]]), o = [1, e[0], e[1], e[2]]), ue(4 === o.length, () => `Error in conv2dDerInput: inShape must be length 4, but got length ${o.length}.`), ue(4 === u.rank, () => "Error in conv2dDerInput: dy must be rank 4, but got rank " + u.rank), ue(4 === n.rank, () => "Error in conv2dDerInput: filter must be rank 4, but got rank " + n.rank);
+                    const c = "NHWC" === s ? o[3] : o[1],
+                        h = "NHWC" === s ? u.shape[3] : u.shape[1];
+                    ue(c === n.shape[2], () => `Error in conv2dDerInput: depth of input (${c}) must match input depth for filter ${n.shape[2]}.`), ue(h === n.shape[3], () => `Error in conv2dDerInput: depth of output (${h}) must match output depth for filter ${n.shape[3]}.`), Hu("conv2dDerInput", a, i);
+                    const p = {
+                            dy: u,
+                            filter: n
+                        },
+                        d = {
+                            strides: r,
+                            pad: a,
+                            dataFormat: s,
+                            dimRoundingMode: i,
+                            inputShape: o
+                        },
+                        f = xs.runKernel(_t, p, d);
+                    return l ? qu(f, [f.shape[1], f.shape[2], f.shape[3]]) : f
+                }
+            });
+            var bl = Ms({
+                conv2dTranspose_: function (e, t, n, r, a, s) {
+                    const i = Rs(e, "x", "conv2dTranspose"),
+                        o = Rs(t, "filter", "conv2dTranspose");
+                    return yl(n, i, o, r, a, "NHWC", s)
+                }
+            });
+            var vl = Ms({
+                conv3d_: function (e, t, n, r, a = "NDHWC", s = [1, 1, 1]) {
+                    const i = Rs(e, "x", "conv3d"),
+                        o = Rs(t, "filter", "conv3d");
+                    let u = i,
+                        l = !1;
+                    4 === i.rank && (l = !0, u = qu(i, [1, i.shape[0], i.shape[1], i.shape[2], i.shape[3]])), ue(5 === u.rank, () => `Error in conv3d: input must be rank 5, but got rank ${u.rank}.`), ue(5 === o.rank, () => `Error in conv3d: filter must be rank 5, but got rank ${o.rank}.`), ue(u.shape[4] === o.shape[3], () => `Error in conv3d: depth of input (${u.shape[4]}) must match input depth for filter ${o.shape[3]}.`), ue(ju(n, s), () => `Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`), ue("NDHWC" === a, () => `Error in conv3d: got dataFormat of ${a} but only NDHWC is currently supported.`);
+                    const c = {
+                            x: u,
+                            filter: o
+                        },
+                        h = {
+                            strides: n,
+                            pad: r,
+                            dataFormat: a,
+                            dilations: s
+                        },
+                        p = xs.runKernel(Ft, c, h);
+                    return l ? qu(p, [p.shape[1], p.shape[2], p.shape[3], p.shape[4]]) : p
+                }
+            });
+            var xl = Ms({
+                conv3DBackpropInput_: function (e, t, n, r, a) {
+                    ue(e.length === t.rank, () => `Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);
+                    let s = e,
+                        i = t,
+                        o = !1;
+                    4 === t.rank && (o = !0, i = qu(t, [1, t.shape[0], t.shape[1], t.shape[2], t.shape[3]]), s = [1, e[0], e[1], e[2], e[3]]);
+                    const u = s[4],
+                        l = i.shape[4];
+                    ue(5 === s.length, () => `Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`), ue(5 === i.rank, () => "Error in conv3dDerInput: dy must be rank 5, but got rank " + i.rank), ue(5 === n.rank, () => "Error in conv3dDerInput: filter must be rank 5, but got rank " + n.rank), ue(u === n.shape[3], () => `Error in conv3dDerInput: depth of input (${u}) must match input depth for filter ${n.shape[3]}.`), ue(l === n.shape[4], () => `Error in conv3dDerInput: depth of output (${l}) must match output depth for filter ${n.shape[4]}.`);
+                    const c = {
+                            dy: i,
+                            filter: n
+                        },
+                        h = {
+                            pad: a,
+                            strides: r,
+                            inputShape: s
+                        },
+                        p = xs.runKernel(Dt, c, h);
+                    return o ? qu(p, [p.shape[1], p.shape[2], p.shape[3], p.shape[4]]) : p
+                }
+            });
+            var wl = Ms({
+                conv3dTranspose_: function (e, t, n, r, a) {
+                    const s = Rs(e, "x", "conv3dTranspose"),
+                        i = Rs(t, "filter", "conv3dTranspose");
+                    return xl(n, s, i, r, a)
+                }
+            });
+            var kl = Ms({
+                cos_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "cos", "float32")
+                    };
+                    return xs.runKernel(Ot, t)
+                }
+            });
+            var Sl = Ms({
+                cosh_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "cosh", "float32")
+                    };
+                    return xs.runKernel(Lt, t)
+                }
+            });
+            var Il = Ms({
+                cumsum_: function (e, t = 0, n = !1, r = !1) {
+                    const a = {
+                            x: Rs(e, "x", "cumsum")
+                        },
+                        s = {
+                            axis: t,
+                            exclusive: n,
+                            reverse: r
+                        };
+                    return xs.runKernel(Pt, a, s)
+                }
+            });
+            var Al = Ms({
+                denseBincount_: function (e, t, n, r = !1) {
+                    const a = Rs(e, "x", "denseBincount"),
+                        s = Rs(t, "weights", "denseBincount");
+                    ue("int32" === a.dtype, () => "Error in denseBincount: input dtype must be int32, but got " + a.dtype), ue(a.rank <= 2, () => `Error in denseBincount: input must be at most rank 2, but got rank ${a.rank}.`), ue(n >= 0, () => `size must be non-negative, but got ${n}.`), ue(s.size === a.size || 0 === s.size, () => `Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${a.shape}, weights shape: ${s.shape}.`);
+                    const i = {
+                            x: a,
+                            weights: s
+                        },
+                        o = {
+                            size: n,
+                            binaryOutput: r
+                        };
+                    return xs.runKernel(Bt, i, o)
+                }
+            });
+            var El = Ms({
+                depthToSpace_: function (e, t, n = "NHWC") {
+                    const r = Rs(e, "x", "depthToSpace", "float32"),
+                        a = "NHWC" === n ? r.shape[1] : r.shape[2],
+                        s = "NHWC" === n ? r.shape[2] : r.shape[3],
+                        i = "NHWC" === n ? r.shape[3] : r.shape[1];
+                    ue(t > 1, () => "blockSize should be > 1 for depthToSpace, but was: " + t), ue(a * t >= 0, () => `Negative dimension size caused by overflow when multiplying\n    ${a} and ${t}  for depthToSpace with input shape\n    ${r.shape}`), ue(s * t >= 0, () => `Negative dimension size caused by overflow when multiplying\n    ${s} and ${t} for depthToSpace with input shape\n        ${r.shape}`), ue(i % (t * t) == 0, () => `Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${r.shape}`);
+                    const o = {
+                            x: r
+                        },
+                        u = {
+                            blockSize: t,
+                            dataFormat: n
+                        };
+                    return xs.runKernel(zt, o, u)
+                }
+            });
+            var Nl = Ms({
+                depthwiseConv2d_: function (e, t, n, r, a = "NHWC", s = [1, 1], i) {
+                    const o = Rs(e, "x", "depthwiseConv2d", "float32"),
+                        u = Rs(t, "filter", "depthwiseConv2d", "float32");
+                    let l = o,
+                        c = !1;
+                    3 === o.rank && (c = !0, l = qu(o, [1, o.shape[0], o.shape[1], o.shape[2]])), ue(4 === l.rank, () => `Error in depthwiseConv2d: input must be rank 4, but got rank ${l.rank}.`), ue(4 === u.rank, () => `Error in depthwiseConv2d: filter must be rank 4, but got rank ${u.rank}.`), ue(l.shape[3] === u.shape[2], () => `Error in depthwiseConv2d: number of input channels (${l.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`), Hu("depthwiseConv2d", r, i);
+                    const h = {
+                            x: l,
+                            filter: u
+                        },
+                        p = {
+                            strides: n,
+                            pad: r,
+                            dataFormat: a,
+                            dilations: s,
+                            dimRoundingMode: i
+                        },
+                        d = xs.runKernel(Ut, h, p);
+                    return c ? qu(d, [d.shape[1], d.shape[2], d.shape[3]]) : d
+                }
+            });
+            var Cl = Ms({
+                diag_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "diag")
+                    };
+                    return xs.runKernel(jt, t)
+                }
+            });
+            var Tl = Ms({
+                dilation2d_: function (e, t, n, r, a = [1, 1], s = "NHWC") {
+                    const i = Rs(e, "x", "dilation2d"),
+                        o = Rs(t, "filter", "dilation2d");
+                    ue(3 === i.rank || 4 === i.rank, () => `Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`), ue(3 === o.rank, () => `Error in dilation2d: filter must be rank 3, but got rank ${o.rank}.`), ue("NHWC" === s, () => "Error in dilation2d: Only NHWC is currently supported, but got dataFormat of " + s);
+                    let u = i,
+                        l = !1;
+                    3 === i.rank && (u = qu(i, [1, i.shape[0], i.shape[1], i.shape[2]]), l = !0);
+                    const c = {
+                            x: u,
+                            filter: o
+                        },
+                        h = {
+                            strides: n,
+                            pad: r,
+                            dilations: a
+                        },
+                        p = xs.runKernel(Gt, c, h);
+                    return l ? qu(p, [p.shape[1], p.shape[2], p.shape[3]]) : p
+                }
+            });
+            var Rl = Ms({
+                equal_: function (e, t) {
+                    let n = Rs(e, "a", "equal", "string_or_numeric"),
+                        r = Rs(t, "b", "equal", "string_or_numeric");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Jt, a)
+                }
+            });
+            var _l = Ms({
+                where_: function (e, t, n) {
+                    const r = Rs(t, "a", "where"),
+                        a = Rs(n, "b", "where"),
+                        s = Rs(e, "condition", "where", "bool"),
+                        i = Zi(Zi(s.shape, r.shape), a.shape),
+                        o = {
+                            condition: ul(s, i),
+                            t: ul(r, i),
+                            e: ul(a, i)
+                        };
+                    return xs.runKernel(vr, o)
+                }
+            });
+            var Fl = Ms({
+                zerosLike_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "zerosLike")
+                    };
+                    return xs.runKernel(Jr, t)
+                }
+            });
+            var Ml = Ms({
+                divNoNan_: function (e, t) {
+                    let n = Rs(e, "a", "div"),
+                        r = Rs(t, "b", "div");
+                    [n, r] = ps(n, r);
+                    const a = bu(n, r),
+                        s = Fl(a),
+                        i = Rl(r, s);
+                    return _l(i, s, a)
+                }
+            });
+            var Dl = Ms({
+                dot_: function (e, t) {
+                    const n = Rs(e, "t1", "dot"),
+                        r = Rs(t, "t2", "dot");
+                    ue(!(1 !== n.rank && 2 !== n.rank || 1 !== r.rank && 2 !== r.rank), () => `Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${r.rank}.`);
+                    const a = 1 === n.rank ? n.size : n.shape[1],
+                        s = 1 === r.rank ? r.size : r.shape[0];
+                    if (ue(a === s, () => `Error in dot: inner dimensions of inputs must match, but got ${a} and ${s}.`), 1 === n.rank && 1 === r.rank) {
+                        const e = qu(n, [1, -1]),
+                            t = qu(r, [-1, 1]),
+                            a = Gi(e, t);
+                        return qu(a, [])
+                    }
+                    if (1 === n.rank && 2 === r.rank) {
+                        const e = qu(n, [1, -1]),
+                            t = qu(r, [r.shape[0], r.shape[1]]),
+                            a = Gi(e, t);
+                        return qu(a, [a.size])
+                    }
+                    if (2 === n.rank && 1 === r.rank) {
+                        const e = qu(r, [-1, 1]),
+                            t = Gi(n, e);
+                        return qu(t, [t.size])
+                    } {
+                        const e = qu(r, [r.shape[0], r.shape[1]]);
+                        return Gi(n, e)
+                    }
+                }
+            });
+            var Ol = Ms({
+                einsum_: function (e, ...t) {
+                    const n = t.map((e, t) => Rs(e, "tensors" + t, "einsum")),
+                        r = {
+                            equation: e
+                        };
+                    return xs.runKernel(Yt, n, r)
+                }
+            });
+            var Ll = Ms({
+                elu_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "elu", "float32")
+                    };
+                    return xs.runKernel(Xt, t)
+                }
+            });
+            var Pl = Ms({
+                erf_: function (e) {
+                    let t = Rs(e, "x", "erf");
+                    ue("int32" === t.dtype || "float32" === t.dtype, () => "Input dtype must be `int32` or `float32`."), "int32" === t.dtype && (t = Ai(t, "float32"));
+                    const n = {
+                        x: t
+                    };
+                    return xs.runKernel(Zt, n)
+                }
+            });
+            var $l = Ms({
+                exp_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "exp")
+                    };
+                    return xs.runKernel(en, t)
+                }
+            });
+            var Bl = Ms({
+                expandDims_: function (e, t = 0) {
+                    const n = Rs(e, "x", "expandDims", "string_or_numeric");
+                    ue(t <= n.rank, () => "Axis must be <= rank of the tensor");
+                    const r = {
+                            input: n
+                        },
+                        a = {
+                            dim: t
+                        };
+                    return xs.runKernel(tn, r, a)
+                }
+            });
+            var zl = Ms({
+                expm1_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "expm1")
+                    };
+                    return xs.runKernel(nn, t)
+                }
+            });
+            var Ul = Ms({
+                tile_: function (e, t) {
+                    const n = Rs(e, "x", "tile", "string_or_numeric");
+                    ue(n.rank === t.length, () => `Error in transpose: rank of input ${n.rank} must match length of reps ${t}.`);
+                    const r = {
+                            x: n
+                        },
+                        a = {
+                            reps: t
+                        };
+                    return xs.runKernel(Hr, r, a)
+                }
+            });
+            var Wl = Ms({
+                eye_: function (e, t, n, r = "float32") {
+                    null == t && (t = e);
+                    const a = Ii([e, t], r),
+                        s = e <= t ? e : t;
+                    for (let e = 0; e < s; ++e) a.set(1, e, e);
+                    const i = qu(a.toTensor(), [e, t]);
+                    if (null == n) return i;
+                    if (1 === n.length) return Ul(Bl(i, 0), [n[0], 1, 1]);
+                    if (2 === n.length) return Ul(Bl(Bl(i, 0), 0), [n[0], n[1], 1, 1]);
+                    if (3 === n.length) return Ul(Bl(Bl(Bl(i, 0), 0), 0), [n[0], n[1], n[2], 1, 1]);
+                    throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${n.length}D.`)
+                }
+            });
+
+            function Vl(e, t, n) {
+                const r = {
+                    shape: e,
+                    value: t,
+                    dtype: n
+                };
+                return xs.runKernel(an, {}, r)
+            }
+            var jl = Ms({
+                floor_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "floor", "float32")
+                    };
+                    return xs.runKernel(on, t)
+                }
+            });
+            var Gl = Ms({
+                gather_: function (e, t, n = 0, r = 0) {
+                    const a = {
+                            x: Rs(e, "x", "gather"),
+                            indices: Rs(t, "indices", "gather", "int32")
+                        },
+                        s = {
+                            axis: n,
+                            batchDims: r
+                        };
+                    return xs.runKernel(cn, a, s)
+                }
+            });
+            var Hl = Ms({
+                greater_: function (e, t) {
+                    let n = Rs(e, "a", "greater", "string_or_numeric"),
+                        r = Rs(t, "b", "greater", "string_or_numeric");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(pn, a)
+                }
+            });
+            var ql = Ms({
+                greaterEqual_: function (e, t) {
+                    let n = Rs(e, "a", "greaterEqual", "string_or_numeric"),
+                        r = Rs(t, "b", "greaterEqual", "string_or_numeric");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(dn, a)
+                }
+            });
+            var Kl = Ms({
+                imag_: function (e) {
+                    const t = {
+                        input: Rs(e, "input", "imag")
+                    };
+                    return xs.runKernel(gn, t)
+                }
+            });
+            var Yl = Ms({
+                isFinite_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "isFinite")
+                    };
+                    return xs.runKernel(yn, t)
+                }
+            });
+            var Xl = Ms({
+                isInf_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "isInf")
+                    };
+                    return xs.runKernel(bn, t)
+                }
+            });
+            var Ql = Ms({
+                isNaN_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "isNaN")
+                    };
+                    return xs.runKernel(vn, t)
+                }
+            });
+            var Zl = Ms({
+                leakyRelu_: function (e, t = .2) {
+                    const n = {
+                            x: Rs(e, "x", "leakyRelu")
+                        },
+                        r = {
+                            alpha: t
+                        };
+                    return xs.runKernel(xn, n, r)
+                }
+            });
+            var Jl = Ms({
+                less_: function (e, t) {
+                    let n = Rs(e, "a", "less", "string_or_numeric"),
+                        r = Rs(t, "b", "less", "string_or_numeric");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(wn, a)
+                }
+            });
+            var ec = Ms({
+                lessEqual_: function (e, t) {
+                    let n = Rs(e, "a", "lessEqual", "string_or_numeric"),
+                        r = Rs(t, "b", "lessEqual", "string_or_numeric");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(kn, a)
+                }
+            });
+
+            function tc(e, t, n) {
+                if (n <= 0) throw new Error("The number of values should be positive.");
+                const r = {
+                    start: e,
+                    stop: t,
+                    num: n
+                };
+                return xs.runKernel(Sn, {}, r)
+            }
+            var nc = Ms({
+                localResponseNormalization_: function (e, t = 5, n = 1, r = 1, a = .5) {
+                    const s = Rs(e, "x", "localResponseNormalization");
+                    ue(4 === s.rank || 3 === s.rank, () => `Error in localResponseNormalization: x must be rank 3 or 4 but got\n               rank ${s.rank}.`), ue(me(t), () => `Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);
+                    let i = s,
+                        o = !1;
+                    3 === s.rank && (o = !0, i = qu(s, [1, s.shape[0], s.shape[1], s.shape[2]]));
+                    const u = {
+                            x: i
+                        },
+                        l = {
+                            depthRadius: t,
+                            bias: n,
+                            alpha: r,
+                            beta: a
+                        },
+                        c = xs.runKernel(Rn, u, l);
+                    return o ? qu(c, [c.shape[1], c.shape[2], c.shape[3]]) : c
+                }
+            });
+            var rc = Ms({
+                log_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "log", "float32")
+                    };
+                    return xs.runKernel(In, t)
+                }
+            });
+            var ac = Ms({
+                log1p_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "log1p")
+                    };
+                    return xs.runKernel(An, t)
+                }
+            });
+
+            function sc(e) {
+                return ue(Le(e), () => "The f passed in grad(f) must be a function"), (t, n) => {
+                    const r = Rs(t, "x", "tf.grad", "string_or_numeric"),
+                        a = null != n ? Rs(n, "dy", "tf.grad") : null;
+                    return xs.tidy(() => {
+                        const {
+                            value: t,
+                            grads: n
+                        } = xs.gradients(() => e(r), [r], a);
+                        return null != a && le(t.shape, a.shape, "The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"), hc(n), n[0]
+                    })
+                }
+            }
+
+            function ic(e) {
+                return ue(Le(e), () => "The f passed in grads(f) must be a function"), (t, n) => {
+                    ue(Array.isArray(t), () => "The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");
+                    const r = _s(t, "args", "tf.grads", "string_or_numeric"),
+                        a = null != n ? Rs(n, "dy", "tf.grads") : null;
+                    return xs.tidy(() => {
+                        const {
+                            value: t,
+                            grads: n
+                        } = xs.gradients(() => e(...r), r, a);
+                        return null != a && le(t.shape, a.shape, "The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"), hc(n), n
+                    })
+                }
+            }
+
+            function oc(e) {
+                return ue(Le(e), () => "The f passed in valueAndGrad(f) must be a function"), (t, n) => {
+                    ue(t instanceof Ya, () => "The x passed in valueAndGrad(f)(x) must be a tensor"), ue(null == n || n instanceof Ya, () => "The dy passed in valueAndGrad(f)(x, dy) must be a tensor");
+                    const {
+                        grads: r,
+                        value: a
+                    } = xs.gradients(() => e(t), [t], n);
+                    return hc(r), {
+                        grad: r[0],
+                        value: a
+                    }
+                }
+            }
+
+            function uc(e) {
+                return ue(Le(e), () => "The f passed in valueAndGrads(f) must be a function"), (t, n) => {
+                    ue(Array.isArray(t) && t.every(e => e instanceof Ya), () => "The args passed in valueAndGrads(f)(args) must be array of tensors"), ue(null == n || n instanceof Ya, () => "The dy passed in valueAndGrads(f)(args, dy) must be a tensor");
+                    const r = xs.gradients(() => e(...t), t, n);
+                    return null != n && le(r.value.shape, n.shape, "The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"), hc(r.grads), r
+                }
+            }
+
+            function lc(e, t) {
+                ue(Le(e), () => "The f passed in variableGrads(f) must be a function"), ue(null == t || Array.isArray(t) && t.every(e => e instanceof Qa), () => "The varList passed in variableGrads(f, varList) must be an array of variables");
+                const n = null != t;
+                if (!n) {
+                    t = [];
+                    for (const e in xs.registeredVariables) t.push(xs.registeredVariables[e])
+                }
+                const r = n ? t.filter(e => !e.trainable) : null,
+                    a = t.length;
+                ue((t = t.filter(e => e.trainable)).length > 0, () => `variableGrads() expects at least one of the input variables to be trainable, but none of the ${a} variables is trainable.`);
+                const {
+                    value: s,
+                    grads: i
+                } = xs.gradients(e, t, null, !0);
+                ue(i.some(e => null != e), () => "Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."), ue(0 === s.rank, () => `The f passed in variableGrads(f) must return a scalar, but it returned a rank-${s.rank} tensor`);
+                const o = {};
+                return t.forEach((e, t) => {
+                    null != i[t] && (o[e.name] = i[t])
+                }), null != r && r.forEach(e => o[e.name] = null), {
+                    value: s,
+                    grads: o
+                }
+            }
+
+            function cc(e) {
+                return xs.customGrad(e)
+            }
+
+            function hc(e) {
+                if (e.filter(e => null == e).length > 0) throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that\n    the f you passed encloses all operations that lead from x to y.")
+            }
+            var pc = Ms({
+                neg_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "neg")
+                    };
+                    return xs.runKernel(Hn, t)
+                }
+            });
+            var dc = Ms({
+                softplus_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "softplus")
+                    };
+                    return xs.runKernel(Er, t)
+                }
+            });
+            var fc = Ms({
+                logSigmoid_: function (e) {
+                    const t = Rs(e, "x", "logSigmoid");
+                    return cc(e => ({
+                        value: pc(dc(pc(e))),
+                        gradFunc: t => vu(t, Qu(pc(e)))
+                    }))(t)
+                }
+            });
+            var mc = Ms({
+                max_: function (e, t = null, n = !1) {
+                    const r = {
+                            x: Rs(e, "x", "max")
+                        },
+                        a = {
+                            reductionIndices: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(Fn, r, a)
+                }
+            });
+            var gc = Ms({
+                sub_: function (e, t) {
+                    let n = Rs(e, "a", "sub"),
+                        r = Rs(t, "b", "sub");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Vr, a)
+                }
+            });
+            var yc = Ms({
+                sum_: function (e, t = null, n = !1) {
+                    let r = Rs(e, "x", "sum");
+                    "bool" === r.dtype && (r = Ai(r, "int32"));
+                    const a = {
+                            x: r
+                        },
+                        s = {
+                            axis: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(Cr, a, s)
+                }
+            });
+            var bc = Ms({
+                logSoftmax_: function (e, t = -1) {
+                    const n = Rs(e, "logits", "logSoftmax");
+                    if (-1 === t && (t = n.rank - 1), t !== n.rank - 1) throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and axis was ${t}`);
+                    return cc((e, n) => {
+                        const r = mc(e, t, !0),
+                            a = gc(e, r),
+                            s = gc(Ai(a, "float32"), rc(yc($l(a), t, !0)));
+                        n([s]);
+                        return {
+                            value: s,
+                            gradFunc: (e, n) => {
+                                const [r] = n, a = $l(r);
+                                return gc(e, vu(yc(e, t, !0), a))
+                            }
+                        }
+                    })(n)
+                }
+            });
+
+            function vc(e, t) {
+                for (let n = 0; n < e.length; ++n)
+                    if (e[e.length - n - 1] !== t - 1 - n) return !1;
+                return !0
+            }
+
+            function xc(e, t, n) {
+                const r = e.length + t.length,
+                    a = [];
+                let s = 0,
+                    i = 0;
+                for (let o = 0; o < r; o++) - 1 === n.indexOf(o) ? a.push(e[s++]) : a.push(t[i++]);
+                return a
+            }
+
+            function wc(e, t) {
+                const n = [],
+                    r = e.length;
+                for (let a = 0; a < r; a++) - 1 === t.indexOf(a) && n.push(e[a]);
+                return [n, t.map(t => e[t])]
+            }
+
+            function kc(e, t) {
+                return xc(e, t.map(e => 1), t)
+            }
+
+            function Sc(e, t, n) {
+                ue(vc(t, n), () => `${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)
+            }
+
+            function Ic(e, t) {
+                if (vc(e, t)) return null;
+                const n = [];
+                for (let r = 0; r < t; ++r) - 1 === e.indexOf(r) && n.push(r);
+                return e.forEach(e => n.push(e)), n
+            }
+
+            function Ac(e) {
+                return e.map((e, t) => [t, e]).sort((e, t) => e[1] - t[1]).map(e => e[0])
+            }
+
+            function Ec(e, t) {
+                const n = [];
+                for (let r = t - e; r < t; ++r) n.push(r);
+                return n
+            }
+            var Nc = Ms({
+                logSumExp_: function (e, t = null, n = !1) {
+                    const r = Rs(e, "x", "logSumExp"),
+                        a = ke(t, r.shape),
+                        s = mc(r, a, !0),
+                        i = gc(r, s),
+                        o = $l(i),
+                        u = yc(o, a),
+                        l = rc(u),
+                        c = gu(qu(s, l.shape), l);
+                    if (n) {
+                        const e = kc(c.shape, a);
+                        return qu(c, e)
+                    }
+                    return c
+                }
+            });
+            var Cc = Ms({
+                logicalAnd_: function (e, t) {
+                    const n = Rs(e, "a", "logicalAnd", "bool"),
+                        r = Rs(t, "b", "logicalAnd", "bool");
+                    Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(En, a)
+                }
+            });
+            var Tc = Ms({
+                logicalNot_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "logicalNot", "bool")
+                    };
+                    return xs.runKernel(Nn, t)
+                }
+            });
+            var Rc = Ms({
+                logicalOr_: function (e, t) {
+                    const n = Rs(e, "a", "logicalOr", "bool"),
+                        r = Rs(t, "b", "logicalOr", "bool");
+                    Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Cn, a)
+                }
+            });
+            var _c = Ms({
+                logicalXor_: function (e, t) {
+                    const n = Rs(e, "a", "logicalXor", "bool"),
+                        r = Rs(t, "b", "logicalXor", "bool");
+                    return Zi(n.shape, r.shape), Cc(Rc(e, t), Tc(Cc(e, t)))
+                }
+            });
+            var Fc = Ms({
+                maxPool_: function (e, t, n, r, a) {
+                    const s = Rs(e, "x", "maxPool");
+                    let i = s,
+                        o = !1;
+                    3 === s.rank && (o = !0, i = qu(s, [1, s.shape[0], s.shape[1], s.shape[2]])), ue(4 === i.rank, () => `Error in maxPool: input must be rank 4 but got rank ${i.rank}.`), ue(ju(n, 1), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '1'`), Hu("maxPool", r, a);
+                    const u = {
+                            x: i
+                        },
+                        l = {
+                            filterSize: t,
+                            strides: n,
+                            pad: r,
+                            dimRoundingMode: a
+                        },
+                        c = xs.runKernel(Dn, u, l);
+                    return o ? qu(c, [c.shape[1], c.shape[2], c.shape[3]]) : c
+                }
+            });
+            var Mc = Ms({
+                maxPool3d_: function (e, t = [1, 1, 1], n, r, a, s = "NDHWC") {
+                    const i = Rs(e, "x", "maxPool3d");
+                    let o = i,
+                        u = !1;
+                    4 === i.rank && (u = !0, o = qu(i, [1, i.shape[0], i.shape[1], i.shape[2], i.shape[3]])), ue(5 === o.rank, () => `Error in maxPool3d: x must be rank 5 but got rank ${o.rank}.`), ue("NDHWC" === s, () => "Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of " + s), Hu("maxPool3d", r, a);
+                    const l = {
+                            x: o
+                        },
+                        c = {
+                            filterSize: t,
+                            strides: n,
+                            pad: r,
+                            dimRoundingMode: a,
+                            dataFormat: s
+                        },
+                        h = xs.runKernel(Ln, l, c);
+                    return u ? qu(h, [h.shape[1], h.shape[2], h.shape[3], h.shape[4]]) : h
+                }
+            });
+            var Dc = Ms({
+                maxPoolWithArgmax_: function (e, t, n, r, a = !1) {
+                    const s = {
+                            x: Rs(e, "x", "maxPoolWithArgmax")
+                        },
+                        i = {
+                            filterSize: t,
+                            strides: n,
+                            pad: r,
+                            includeBatchInIndex: a
+                        },
+                        o = xs.runKernel($n, s, i);
+                    return {
+                        result: o[0],
+                        indexes: o[1]
+                    }
+                }
+            });
+            var Oc = Ms({
+                maximum_: function (e, t) {
+                    let n = Rs(e, "a", "maximum"),
+                        r = Rs(t, "b", "maximum");
+                    [n, r] = ps(n, r), "bool" === n.dtype && (n = Ai(n, "int32"), r = Ai(r, "int32")), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Mn, a)
+                }
+            });
+            var Lc = Ms({
+                mean_: function (e, t = null, n = !1) {
+                    const r = {
+                            x: Rs(e, "x", "mean")
+                        },
+                        a = {
+                            axis: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(Bn, r, a)
+                }
+            });
+
+            function Pc(e, t = "float32") {
+                if ("complex64" === t) {
+                    const t = Pc(e, "float32"),
+                        n = Pc(e, "float32");
+                    return Ds(t, n)
+                }
+                const n = Ue(pe(e), t);
+                return xs.makeTensor(n, e, t)
+            }
+
+            function $c(e, t = "float32") {
+                if ("complex64" === t) {
+                    const t = $c(e, "float32"),
+                        n = Pc(e, "float32");
+                    return Ds(t, n)
+                }
+                const n = ze(pe(e), t);
+                return xs.makeTensor(n, e, t)
+            }
+
+            function Bc(e, t, {
+                indexing: n = "xy"
+            } = {}) {
+                if ("xy" !== n && "ij" !== n) throw new TypeError(n + " is not a valid third argument to meshgrid");
+                if (void 0 === e) return [];
+                let r = Rs(e, "x", "meshgrid", e instanceof Ya ? e.dtype : "float32");
+                if (void 0 === t) return [r];
+                let a = Rs(t, "y", "meshgrid", t instanceof Ya ? t.dtype : "float32");
+                const s = pe(r.shape),
+                    i = pe(a.shape);
+                return "xy" === n ? (r = qu(r, [1, -1]), a = qu(a, [-1, 1]), [Gi($c([i, 1], r.dtype), r), Gi(a, $c([1, s], a.dtype))]) : (r = qu(r, [-1, 1]), a = qu(a, [1, -1]), [Gi(r, $c([1, i], r.dtype)), Gi($c([s, 1], a.dtype), a)])
+            }
+            var zc = Ms({
+                min_: function (e, t = null, n = !1) {
+                    const r = {
+                            x: Rs(e, "x", "min")
+                        },
+                        a = {
+                            axis: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(zn, r, a)
+                }
+            });
+            var Uc = Ms({
+                minimum_: function (e, t) {
+                    let n = Rs(e, "a", "minimum"),
+                        r = Rs(t, "b", "minimum");
+                    [n, r] = ps(n, r), "bool" === n.dtype && (n = Ai(n, "int32"), r = Ai(r, "int32")), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Un, a)
+                }
+            });
+            var Wc = Ms({
+                mirrorPad_: function (e, t, n) {
+                    ue("reflect" === n || "symmetric" === n, () => `Invalid mode. Mode must be either reflect or symmetric. Got ${n}.`);
+                    const r = Rs(e, "x", "mirrorPad");
+                    if (0 === r.rank) throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");
+                    ue(t.length === r.rank, () => `Padding doesn't match input. Must be ${r.rank}. Got ${t.length}.`);
+                    const a = "reflect" === n ? 1 : 0;
+                    for (let e = 0; e < r.rank; e++) ue(2 === t[e].length, () => "Invalid number of paddings. Must be length of 2 each."), ue(t[e][0] >= 0 && t[e][0] <= r.shape[e] - a && t[e][1] >= 0 && t[e][1] <= r.shape[e] - a, () => `Padding in dimension ${e} cannot be greater than or equal to ${r.shape[e]-a} or less than 0 for input of shape ${r.shape}`);
+                    const s = {
+                            paddings: t,
+                            mode: n
+                        },
+                        i = {
+                            x: r
+                        };
+                    return xs.runKernel(Wn, i, s)
+                }
+            });
+            var Vc = Ms({
+                mod_: function (e, t) {
+                    let n = Rs(e, "a", "mod"),
+                        r = Rs(t, "b", "mod");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Vn, a)
+                }
+            });
+            var jc = Ms({
+                square_: function (e) {
+                    const t = Rs(e, "x", "square");
+                    return xs.runKernel("Square", {
+                        x: t
+                    }, {})
+                }
+            });
+            var Gc = Ms({
+                moments_: function (e, t = null, n = !1) {
+                    const r = ke(t, (e = Rs(e, "x", "moments")).shape),
+                        a = Lc(e, r, n);
+                    let s = a.shape;
+                    n || (s = kc(a.shape, r));
+                    const i = jc(gc(Ai(e, "float32"), qu(a, s)));
+                    return {
+                        mean: a,
+                        variance: Lc(i, r, n)
+                    }
+                }
+            });
+            var Hc = Ms({
+                multiRNNCell_: function (e, t, n, r) {
+                    const a = Rs(t, "data", "multiRNNCell"),
+                        s = _s(n, "c", "multiRNNCell"),
+                        i = _s(r, "h", "multiRNNCell");
+                    let o = a;
+                    const u = [];
+                    for (let t = 0; t < e.length; t++) {
+                        const n = e[t](o, s[t], i[t]);
+                        u.push(n[0]), u.push(n[1]), o = n[1]
+                    }
+                    const l = [],
+                        c = [];
+                    for (let e = 0; e < u.length; e += 2) l.push(u[e]), c.push(u[e + 1]);
+                    return [l, c]
+                }
+            });
+            var qc = Ms({
+                multinomial_: function (e, t, n, r = !1) {
+                    const a = Rs(e, "logits", "multinomial"),
+                        s = a.size,
+                        i = a.rank;
+                    if (s < 2) throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);
+                    if (i > 2) throw new Error("Rank of probabilities must be 1 or 2, but is " + i);
+                    n = n || Math.random();
+                    const o = {
+                            logits: 1 === i ? qu(a, [1, -1]) : a
+                        },
+                        u = {
+                            numSamples: t,
+                            seed: n,
+                            normalized: r
+                        },
+                        l = xs.runKernel(jn, o, u);
+                    return 1 === i ? qu(l, [l.size]) : l
+                }
+            });
+            var Kc = Ms({
+                notEqual_: function (e, t) {
+                    let n = Rs(e, "a", "notEqual", "string_or_numeric"),
+                        r = Rs(t, "b", "notEqual", "string_or_numeric");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(qn, a)
+                }
+            });
+            var Yc = Ms({
+                onesLike_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "onesLike")
+                    };
+                    return xs.runKernel(Qn, t)
+                }
+            });
+            var Xc = Ms({
+                outerProduct_: function (e, t) {
+                    const n = Rs(e, "v1", "outerProduct"),
+                        r = Rs(t, "v2", "outerProduct");
+                    ue(1 === n.rank && 1 === r.rank, () => `Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${r.rank}.`);
+                    const a = qu(n, [-1, 1]),
+                        s = qu(r, [1, -1]);
+                    return Gi(a, s)
+                }
+            });
+            var Qc = Ms({
+                pad_: function (e, t, n = 0) {
+                    const r = Rs(e, "x", "pad");
+                    if (0 === r.rank) throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");
+                    const a = {
+                            paddings: t,
+                            constantValue: n
+                        },
+                        s = {
+                            x: r
+                        };
+                    return xs.runKernel(er, s, a)
+                }
+            });
+            var Zc = Ms({
+                pad1d_: function (e, t, n = 0) {
+                    return ue(2 === t.length, () => "Invalid number of paddings. Must be length of 2."), Qc(e, [t], n)
+                }
+            });
+            var Jc = Ms({
+                pad2d_: function (e, t, n = 0) {
+                    return ue(2 === t.length && 2 === t[0].length && 2 === t[1].length, () => "Invalid number of paddings. Must be length of 2 each."), Qc(e, t, n)
+                }
+            });
+            var eh = Ms({
+                pad3d_: function (e, t, n = 0) {
+                    return ue(3 === t.length && 2 === t[0].length && 2 === t[1].length && 2 === t[2].length, () => "Invalid number of paddings. Must be length of 2 each."), Qc(e, t, n)
+                }
+            });
+            var th = Ms({
+                pad4d_: function (e, t, n = 0) {
+                    return ue(4 === t.length && 2 === t[0].length && 2 === t[1].length && 2 === t[2].length && 2 === t[3].length, () => "Invalid number of paddings. Must be length of 2 each."), Qc(e, t, n)
+                }
+            });
+            var nh = Ms({
+                spaceToBatchND_: function (e, t, n) {
+                    const r = Rs(e, "x", "spaceToBatchND");
+                    ue(r.rank >= 1 + t.length, () => `input rank ${r.rank} should be > than [blockShape] ${t.length}`), ue(n.length === t.length, () => `paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`), ue(r.shape.reduce((e, r, a) => a > 0 && a <= t.length ? e && (r + n[a - 1][0] + n[a - 1][1]) % t[a - 1] == 0 : e, !0), () => `input spatial dimensions ${r.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);
+                    const a = {
+                            x: r
+                        },
+                        s = {
+                            blockShape: t,
+                            paddings: n
+                        };
+                    return xs.runKernel(Tr, a, s)
+                }
+            });
+            var rh = Ms({
+                pool_: function (e, t, n, r, a, s, i) {
+                    null == a && (a = [1, 1]), null == s && (s = 1), 0 === r && (r = "valid");
+                    const o = Rs(e, "x", "maxPool");
+                    let u = o,
+                        l = !1;
+                    3 === o.rank && (l = !0, u = qu(o, [1, o.shape[0], o.shape[1], o.shape[2]])), ue(ju(s, a), () => `Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${a}'`);
+                    const c = Du(u.shape, t, s, a, r),
+                        h = [c.dilationHeight, c.dilationWidth];
+                    let p;
+                    p = "same" === r ? function (e, t) {
+                        const n = e.map((e, n) => e + (e - 1) * (t[n] - 1)).map(e => e - 1),
+                            r = n.map(e => Math.floor(e / 2)),
+                            a = n.map((e, t) => e - r[t]);
+                        return n.map((e, t) => [r[t], a[t]])
+                    }([c.filterHeight, c.filterWidth], h) : [
+                        [0, 0],
+                        [0, 0]
+                    ];
+                    const d = 1 === h[0] && 1 === h[1],
+                        [f, m] = function (e, t, n) {
+                            const r = n.map(e => e[0]),
+                                a = n.map(e => e[1]),
+                                s = e.concat(r, a),
+                                i = t.map((e, t) => (e - s[t] % e) % e),
+                                o = a.map((e, t) => e + i[t]),
+                                u = t.map((e, t) => [r[t], o[t]]),
+                                l = t.map((e, t) => [0, i[t]]);
+                            return [u, l]
+                        }([c.inHeight, c.inWidth], h, p),
+                        g = d ? r : "valid",
+                        y = d ? u : nh(u, h, f),
+                        b = ("avg" === n ? () => Ku(y, t, s, g, i) : () => Fc(y, t, s, g, i))(),
+                        v = d ? b : tl(b, h, m);
+                    return l ? qu(v, [v.shape[1], v.shape[2], v.shape[3]]) : v
+                }
+            });
+            var ah = Ms({
+                pow_: function (e, t) {
+                    let n = Rs(e, "base", "pow"),
+                        r = Rs(t, "exp", "pow");
+                    [n, r] = ps(n, r);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(nr, a)
+                }
+            });
+            var sh = Ms({
+                prelu_: function (e, t) {
+                    const n = {
+                        x: Rs(e, "x", "prelu"),
+                        alpha: Rs(t, "alpha", "prelu")
+                    };
+                    return xs.runKernel(rr, n)
+                }
+            });
+            var ih = Ms({
+                prod_: function (e, t = null, n = !1) {
+                    let r = Rs(e, "x", "prod");
+                    "bool" === r.dtype && (r = Ai(r, "int32"));
+                    const a = {
+                            x: r
+                        },
+                        s = {
+                            axis: t,
+                            keepDims: n
+                        };
+                    return xs.runKernel(ar, a, s)
+                }
+            });
+            var oh = Ms({
+                    rand_: function (e, t, n) {
+                        const r = pe(e);
+                        let a = null;
+                        if (null == n || "float32" === n) a = new Float32Array(r);
+                        else if ("int32" === n) a = new Int32Array(r);
+                        else {
+                            if ("bool" !== n) throw new Error("Unknown data type " + n);
+                            a = new Uint8Array(r)
+                        }
+                        for (let e = 0; e < r; e++) a[e] = t();
+                        return xs.makeTensor(a, e, n)
+                    }
+                }),
+                uh = k(D()),
+                lh = class {
+                    constructor(e, t, n, r, a) {
+                        this.mean = e, this.stdDev = t, this.dtype = n, this.nextVal = NaN, this.truncated = r, this.truncated && (this.upper = this.mean + 2 * this.stdDev, this.lower = this.mean - 2 * this.stdDev);
+                        const s = a || Math.random();
+                        this.random = uh.alea(s.toString())
+                    }
+                    nextValue() {
+                        if (!isNaN(this.nextVal)) {
+                            const e = this.nextVal;
+                            return this.nextVal = NaN, e
+                        }
+                        let e, t, n = !1;
+                        for (; !n;) {
+                            let r, a, s;
+                            do {
+                                r = 2 * this.random() - 1, a = 2 * this.random() - 1, s = r * r + a * a
+                            } while (s >= 1 || 0 === s);
+                            const i = Math.sqrt(-2 * Math.log(s) / s);
+                            e = this.mean + this.stdDev * r * i, t = this.mean + this.stdDev * a * i, this.truncated && !this.isValidTruncated(e) || (n = !0)
+                        }
+                        return this.truncated && !this.isValidTruncated(t) || (this.nextVal = this.convertValue(t)), this.convertValue(e)
+                    }
+                    convertValue(e) {
+                        return null == this.dtype || "float32" === this.dtype ? e : Math.round(e)
+                    }
+                    isValidTruncated(e) {
+                        return e <= this.upper && e >= this.lower
+                    }
+                };
+            var ch = Ms({
+                randomGamma_: function (e, t, n = 1, r = "float32", a) {
+                    if (null == n && (n = 1), null == r && (r = "float32"), "float32" !== r && "int32" !== r) throw new Error("Unsupported data type " + r);
+                    const s = new class {
+                            constructor(e, t, n, r) {
+                                this.alpha = e, this.beta = 1 / t, this.dtype = n;
+                                const a = r || Math.random();
+                                this.randu = uh.alea(a.toString()), this.randn = new lh(0, 1, n, !1, this.randu()), this.d = e < 1 ? e + 2 / 3 : e - 1 / 3, this.c = 1 / Math.sqrt(9 * this.d)
+                            }
+                            nextValue() {
+                                let e, t, n, r, a, s;
+                                for (;;) {
+                                    do {
+                                        r = this.randn.nextValue(), s = 1 + this.c * r
+                                    } while (s <= 0);
+                                    if (s *= s * s, e = r * r, t = 1 - .331 * e * e, n = .5 * e + this.d * (1 - s + Math.log(s)), a = this.randu(), a < t || Math.log(a) < n) break
+                                }
+                                return s = 1 / this.beta * this.d * s, this.alpha < 1 && (s *= Math.pow(this.randu(), 1 / this.alpha)), this.convertValue(s)
+                            }
+                            convertValue(e) {
+                                return "float32" === this.dtype ? e : Math.round(e)
+                            }
+                        }(t, n, r, a),
+                        i = Ii(e, r);
+                    for (let e = 0; e < i.values.length; e++) i.values[e] = s.nextValue();
+                    return i.toTensor()
+                }
+            });
+            var hh = Ms({
+                randomNormal_: function (e, t = 0, n = 1, r, a) {
+                    if (null != r && "bool" === r) throw new Error("Unsupported data type " + r);
+                    const s = new lh(t, n, r, !1, a),
+                        i = Ii(e, r);
+                    for (let e = 0; e < i.values.length; e++) i.values[e] = s.nextValue();
+                    return i.toTensor()
+                }
+            });
+            var ph = Ms({
+                randomUniform_: function (e, t = 0, n = 1, r = "float32", a) {
+                    const s = Ii(e, r),
+                        i = new class {
+                            constructor(e = 0, t = 1, n, r) {
+                                if (this.canReturnFloat = () => null == this.dtype || "float32" === this.dtype, this.min = e, this.range = t - e, this.dtype = n, null == r && (r = Math.random()), "number" == typeof r && (r = r.toString()), !this.canReturnFloat() && this.range <= 1) throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);
+                                this.random = uh.alea(r)
+                            }
+                            convertValue(e) {
+                                return this.canReturnFloat() ? e : Math.round(e)
+                            }
+                            nextValue() {
+                                return this.convertValue(this.min + this.range * this.random())
+                            }
+                        }(t, n, null, a);
+                    for (let e = 0; e < s.values.length; e++) s.values[e] = i.nextValue();
+                    return s.toTensor()
+                }
+            });
+
+            function dh(e, t, n = 1, r = "float32") {
+                if (0 === n) throw new Error("Cannot have a step of zero");
+                const a = {
+                    start: e,
+                    stop: t,
+                    step: n,
+                    dtype: r
+                };
+                return xs.runKernel(sr, {}, a)
+            }
+            var fh = Ms({
+                real_: function (e) {
+                    const t = {
+                        input: Rs(e, "input", "real")
+                    };
+                    return xs.runKernel(ir, t)
+                }
+            });
+            var mh = Ms({
+                reciprocal_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "reciprocal")
+                    };
+                    return xs.runKernel(or, t)
+                }
+            });
+            var gh = Ms({
+                relu_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "relu")
+                    };
+                    return xs.runKernel(ur, t)
+                }
+            });
+            var yh = Ms({
+                relu6_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "relu6")
+                    };
+                    return xs.runKernel(fr, t)
+                }
+            });
+            var bh = Ms({
+                reverse_: function (e, t) {
+                    const n = {
+                            x: Rs(e, "x", "reverse")
+                        },
+                        r = {
+                            dims: t
+                        };
+                    return xs.runKernel(mr, n, r)
+                }
+            });
+            var vh = Ms({
+                reverse1d_: function (e) {
+                    const t = Rs(e, "x", "reverse");
+                    return ue(1 === t.rank, () => `Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`), bh(t, 0)
+                }
+            });
+            var xh = Ms({
+                reverse2d_: function (e, t) {
+                    const n = Rs(e, "x", "reverse");
+                    return ue(2 === n.rank, () => `Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`), bh(n, t)
+                }
+            });
+            var wh = Ms({
+                reverse3d_: function (e, t) {
+                    const n = Rs(e, "x", "reverse");
+                    return ue(3 === n.rank, () => `Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`), bh(n, t)
+                }
+            });
+            var kh = Ms({
+                reverse4d_: function (e, t) {
+                    const n = Rs(e, "x", "reverse");
+                    return ue(4 === n.rank, () => `Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`), bh(n, t)
+                }
+            });
+            var Sh = Ms({
+                round_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "round")
+                    };
+                    return xs.runKernel(gr, t)
+                }
+            });
+            var Ih = Ms({
+                rsqrt_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "rsqrt", "float32")
+                    };
+                    return xs.runKernel(yr, t)
+                }
+            });
+
+            function Ah(e, t) {
+                if ((Te(e) && "string" !== t || Array.isArray(e)) && "complex64" !== t) throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");
+                if ("string" === t && Te(e) && !(e instanceof Uint8Array)) throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");
+                return Os(e, [], [], t)
+            }
+            var Eh = Ms({
+                selu_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "selu")
+                    };
+                    return xs.runKernel(xr, t)
+                }
+            });
+            var Nh = Ms({
+                separableConv2d_: function (e, t, n, r, a, s = [1, 1], i = "NHWC") {
+                    const o = Rs(e, "x", "separableConv2d"),
+                        u = Rs(t, "depthwiseFilter", "separableConv2d"),
+                        l = Rs(n, "pointwiseFilter", "separableConv2d");
+                    let c = o,
+                        h = !1;
+                    if (3 === o.rank && (h = !0, c = qu(o, [1, o.shape[0], o.shape[1], o.shape[2]])), "NCHW" === i) throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");
+                    ue(4 === c.rank, () => `Error in separableConv2d: input must be rank 4, but got rank ${c.rank}.`), ue(4 === u.rank, () => `Error in separableConv2d: depthwise filter must be rank 4, but got rank ${u.rank}.`), ue(4 === l.rank, () => `Error in separableConv2d: pointwise filter must be rank 4, but got rank ${u.rank}.`), ue(1 === l.shape[0], () => `Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${l.shape[0]}.`), ue(1 === l.shape[1], () => `Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${l.shape[1]}.`);
+                    const p = u.shape[2],
+                        d = u.shape[3];
+                    ue(l.shape[2] === p * d, () => `Error in separableConv2d: the third dimension of pointwise filter must be ${p*d}, but got ${l.shape[2]}.`);
+                    const f = Nl(c, u, r, a, i, s),
+                        m = ml(f, l, 1, "valid", i);
+                    return h ? qu(m, [m.shape[1], m.shape[2], m.shape[3]]) : m
+                }
+            });
+            var Ch = async function (e, t) {
+                const n = Rs(e, "x", "setdiff1d"),
+                    r = Rs(t, "y", "setdiff1d");
+                ue(n.dtype === r.dtype, () => `x and y should have the same dtype, but got x (${n.dtype}) and y (${r.dtype}).`), ue(1 === n.rank, () => `x should be 1D tensor, but got x (${n.shape}).`), ue(1 === r.rank, () => `y should be 1D tensor, but got y (${r.shape}).`);
+                const a = await n.data(),
+                    s = await r.data(),
+                    i = new Set(s);
+                let o = 0;
+                for (let e = 0; e < a.length; e++) i.has(a[e]) || o++;
+                const u = new Ha([o], n.dtype),
+                    l = new Ha([o], "int32");
+                for (let e = 0, t = 0; e < a.length; e++) i.has(a[e]) || (u.values[t] = a[e], l.values[t] = e, t++);
+                return [u.toTensor(), l.toTensor()]
+            };
+            var Th = Ms({
+                sign_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "sign")
+                    };
+                    return xs.runKernel(Ir, t)
+                }
+            });
+            var Rh = Ms({
+                sin_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "sin", "float32")
+                    };
+                    return xs.runKernel(kr, t)
+                }
+            });
+            var _h = Ms({
+                sinh_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "sinh")
+                    };
+                    return xs.runKernel(Sr, t)
+                }
+            });
+            var Fh = Ms({
+                slice1d_: function (e, t, n) {
+                    const r = Rs(e, "x", "slice1d");
+                    return ue(1 === r.rank, () => `slice1d expects a rank-1 tensor, but got a rank-${r.rank} tensor`), Zu(r, [t], [n])
+                }
+            });
+            var Mh = Ms({
+                slice2d_: function (e, t, n) {
+                    const r = Rs(e, "x", "slice2d");
+                    return ue(2 === r.rank, () => `slice2d expects a rank-2 tensor, but got a rank-${r.rank} tensor`), Zu(r, t, n)
+                }
+            });
+            var Dh = Ms({
+                slice3d_: function (e, t, n) {
+                    const r = Rs(e, "x", "slice3d");
+                    return ue(3 === r.rank, () => `slice3d expects a rank-3 tensor, but got a rank-${r.rank} tensor`), Zu(r, t, n)
+                }
+            });
+            var Oh = Ms({
+                slice4d_: function (e, t, n) {
+                    const r = Rs(e, "x", "slice4d");
+                    return ue(4 === r.rank, () => `slice4d expects a rank-4 tensor, but got a rank-${r.rank} tensor`), Zu(r, t, n)
+                }
+            });
+            var Lh = Ms({
+                softmax_: function (e, t = -1) {
+                    const n = Rs(e, "logits", "softmax", "float32");
+                    if (-1 === t && (t = n.rank - 1), t !== n.rank - 1) throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and dim was ${t}`);
+                    const r = {
+                            logits: n
+                        },
+                        a = {
+                            dim: t
+                        };
+                    return xs.runKernel(_r, r, a)
+                }
+            });
+            var Ph = Ms({
+                fft_: function (e) {
+                    ue("complex64" === e.dtype, () => `The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);
+                    const t = {
+                        input: e
+                    };
+                    return xs.runKernel(rn, t)
+                }
+            });
+            var $h = Ms({
+                ifft_: function (e) {
+                    ue("complex64" === e.dtype, () => `The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);
+                    const t = {
+                        input: e
+                    };
+                    return xs.runKernel(mn, t)
+                }
+            });
+            var Bh = Ms({
+                irfft_: function (e) {
+                    const t = e.shape[e.shape.length - 1],
+                        n = e.size / t;
+                    let r;
+                    if (t <= 2) {
+                        const a = qu(e, [n, t]);
+                        r = $h(a)
+                    } else {
+                        const a = [n, 2 * (t - 1)],
+                            s = qu(fh(e), [n, t]),
+                            i = qu(Kl(e), [n, t]),
+                            o = bh(Zu(s, [0, 1], [n, t - 2]), 1),
+                            u = vu(bh(Zu(i, [0, 1], [n, t - 2]), 1), Ah(-1)),
+                            l = Xu([s, o], 1),
+                            c = Xu([i, u], 1),
+                            h = qu(Ds(l, c), [a[0], a[1]]);
+                        r = $h(h)
+                    }
+                    if (r = fh(r), 3 === e.rank && 0 !== e.shape[0]) {
+                        const t = r,
+                            n = e.shape[0];
+                        r = qu(r, [n, r.shape[0] / n, r.shape[1]]), t.dispose()
+                    }
+                    return r
+                }
+            });
+            var zh = Ms({
+                split_: function (e, t, n = 0) {
+                    const r = {
+                            x: Rs(e, "x", "split")
+                        },
+                        a = {
+                            numOrSizeSplits: t,
+                            axis: n
+                        };
+                    return xs.runKernel(Rr, r, a)
+                }
+            });
+            var Uh = Ms({
+                rfft_: function (e, t) {
+                    ue("float32" === e.dtype, () => "The dtype for rfft() must be real value but got " + e.dtype);
+                    let n = e.shape[e.shape.length - 1];
+                    const r = e.size / n;
+                    let a;
+                    if (null != t && t < n) {
+                        const r = e.shape.map(e => 0),
+                            s = e.shape.map(e => e);
+                        s[e.shape.length - 1] = t, a = Zu(e, r, s), n = t
+                    } else if (null != t && t > n) {
+                        const r = e.shape.map(e => e);
+                        r[e.shape.length - 1] = t - n, a = Xu([e, Pc(r)], e.shape.length - 1), n = t
+                    } else a = e;
+                    const s = Fl(a),
+                        i = qu(Ds(a, s), [r, n]),
+                        o = Ph(i),
+                        u = Math.floor(n / 2) + 1,
+                        l = fh(o),
+                        c = Kl(o),
+                        h = zh(l, [u, n - u], l.shape.length - 1),
+                        p = zh(c, [u, n - u], c.shape.length - 1),
+                        d = a.shape.slice();
+                    return d[a.shape.length - 1] = u, qu(Ds(h[0], p[0]), d)
+                }
+            });
+            var Wh = Ms({
+                sqrt_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "sqrt", "float32")
+                    };
+                    return xs.runKernel(Nr, t)
+                }
+            });
+            var Vh = Ms({
+                squaredDifference_: function (e, t) {
+                    let n = Rs(e, "a", "squaredDifference"),
+                        r = Rs(t, "b", "squaredDifference");
+                    [n, r] = ps(n, r), Zi(n.shape, r.shape);
+                    const a = {
+                        a: n,
+                        b: r
+                    };
+                    return xs.runKernel(Pr, a, {})
+                }
+            });
+            var jh = Ms({
+                squeeze_: function (e, t) {
+                    const n = Rs(e, "x", "squeeze");
+                    return qu(n, Se(n.shape, t).newShape)
+                }
+            });
+            var Gh = Ms({
+                stack_: function (e, t = 0) {
+                    const n = _s(e, "tensors", "stack", "string_or_numeric");
+                    ue(n.length >= 1, () => "Pass at least one tensor to tf.stack"), n.length > 0 && ue(t <= n[0].rank, () => "Axis must be <= rank of the tensor");
+                    const r = n,
+                        a = {
+                            axis: t
+                        };
+                    return xs.runKernel(Jn, r, a)
+                }
+            });
+            var Hh = Ms({
+                step_: function (e, t = 0) {
+                    const n = {
+                            x: Rs(e, "x", "step")
+                        },
+                        r = {
+                            alpha: t
+                        };
+                    return xs.runKernel(ea, n, r)
+                }
+            });
+            var qh = Ms({
+                stridedSlice_: function (e, t, n, r, a = 0, s = 0, i = 0, o = 0, u = 0) {
+                    const l = {
+                            x: Rs(e, "x", "stridedSlice", "string_or_numeric")
+                        },
+                        c = {
+                            begin: t,
+                            end: n,
+                            strides: r,
+                            beginMask: a,
+                            endMask: s,
+                            ellipsisMask: i,
+                            newAxisMask: o,
+                            shrinkAxisMask: u
+                        };
+                    return xs.runKernel(Br, l, c)
+                }
+            });
+            var Kh = Ms({
+                tan_: function (e) {
+                    const t = {
+                        x: Rs(e, "x", "tan", "float32")
+                    };
+                    return xs.runKernel(jr, t)
+                }
+            });
+
+            function Yh(e, t) {
+                ce(e);
+                const n = Cs(e, t);
+                if (1 !== n.length) throw new Error("tensor1d() requires values to be a flat/TypedArray");
+                return Os(e, null, n, t)
+            }
+
+            function Xh(e, t, n) {
+                if (ce(e), null != t && 2 !== t.length) throw new Error("tensor2d() requires shape to have two numbers");
+                const r = Cs(e, n);
+                if (2 !== r.length && 1 !== r.length) throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");
+                if (1 === r.length && null == t) throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");
+                return Os(e, t, r, n)
+            }
+
+            function Qh(e, t, n) {
+                if (ce(e), null != t && 4 !== t.length) throw new Error("tensor4d() requires shape to have four numbers");
+                const r = Cs(e, n);
+                if (4 !== r.length && 1 !== r.length) throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");
+                if (1 === r.length && null == t) throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");
+                return Os(e, t, r, n)
+            }
+
+            function Zh(e, t, n) {
+                if (ce(e), null != t && 5 !== t.length) throw new Error("tensor5d() requires shape to have five numbers");
+                const r = Cs(e, n);
+                if (5 !== r.length && 1 !== r.length) throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");
+                if (1 === r.length && null == t) throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");
+                return Os(e, t, r, n)
+            }
+
+            function Jh(e, t, n) {
+                if (ce(e), null != t && 6 !== t.length) throw new Error("tensor6d() requires shape to have six numbers");
+                const r = Cs(e, n);
+                if (6 !== r.length && 1 !== r.length) throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");
+                if (1 === r.length && null == t) throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");
+                return Os(e, t = t || r, r, n)
+            }
+            var ep = Ms({
+                topk_: function (e, t = 1, n = !0) {
+                    const r = Rs(e, "x", "topk");
+                    if (0 === r.rank) throw new Error("topk() expects the input to be of rank 1 or higher");
+                    const a = r.shape[r.shape.length - 1];
+                    if (t < 0) throw new Error("'k' passed to topk() must be >= 0 but got " + t);
+                    if (t > a) throw new Error(`'k' passed to topk() must be <= the last dimension (${a}) but got ${t}`);
+                    const s = {
+                            x: r
+                        },
+                        i = {
+                            k: t,
+                            sorted: n
+                        },
+                        [o, u] = xs.runKernel(qr, s, i);
+                    return {
+                        values: o,
+                        indices: u
+                    }
+                }
+            });
+            var tp = Ms({
+                truncatedNormal_: function (e, t = 0, n = 1, r, a) {
+                    if (null != r && "bool" === r) throw new Error("Unsupported data type $ { dtype }");
+                    const s = new lh(t, n, r, !0, a),
+                        i = Ii(e, r);
+                    for (let e = 0; e < i.values.length; e++) i.values[e] = s.nextValue();
+                    return i.toTensor()
+                }
+            });
+            var np = Ms({
+                unique_: function (e, t = 0) {
+                    const n = Rs(e, "x", "unique", "string_or_numeric");
+                    ue(n.rank > 0, () => "The input tensor must be at least 1D");
+                    const r = {
+                            x: n
+                        },
+                        a = {
+                            axis: t
+                        },
+                        [s, i] = xs.runKernel(Xr, r, a);
+                    return {
+                        values: s,
+                        indices: i
+                    }
+                }
+            });
+            var rp = Ms({
+                unsortedSegmentSum_: function (e, t, n) {
+                    const r = Rs(e, "x", "unsortedSegmentSum"),
+                        a = Rs(t, "segmentIds", "unsortedSegmentSum", "int32");
+                    ue(me(n), () => "numSegments must be of dtype int");
+                    const s = {
+                            x: r,
+                            segmentIds: a
+                        },
+                        i = {
+                            numSegments: n
+                        };
+                    return xs.runKernel(Zr, s, i)
+                }
+            });
+            var ap = Ms({
+                unstack_: function (e, t = 0) {
+                    const n = Rs(e, "x", "unstack", "string_or_numeric");
+                    ue(t >= -n.shape.length && t < n.shape.length, () => `Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`);
+                    const r = {
+                            value: n
+                        },
+                        a = {
+                            axis: t
+                        };
+                    return xs.runKernel(Qr, r, a)
+                }
+            });
+
+            function sp(e, t = !0, n, r) {
+                return xs.makeVariable(e, t, n, r)
+            }
+
+            function ip(e, t) {
+                const n = [];
+                for (let e = 0; e < t.length; e++) t[e] && n.push(e);
+                const r = Ii(e, "int32"),
+                    a = Ii([n.length, e.length], "int32");
+                for (let t = 0; t < n.length; t++) {
+                    const s = r.indexToLoc(n[t]),
+                        i = t * e.length;
+                    a.values.set(s, i)
+                }
+                return a.toTensor()
+            }
+            var op = async function (e) {
+                const t = Rs(e, "condition", "whereAsync", "bool"),
+                    n = await t.data(),
+                    r = ip(t.shape, n);
+                return e !== t && t.dispose(), r
+            };
+            var up = async function (e, t, n) {
+                const r = Rs(e, "tensor", "boolMask"),
+                    a = Rs(t, "mask", "boolMask", "bool"),
+                    s = null == n ? 0 : n,
+                    i = a.rank,
+                    o = r.shape;
+                ue(i > 0, () => "mask cannot be scalar"), le(o.slice(s, s + i), a.shape, "mask's shape must match the first K dimensions of tensor's shape,");
+                let u = 1;
+                for (let e = s; e < s + i; e++) u *= o[e];
+                const l = o.slice(0, s).concat([u], o.slice(s + i)),
+                    c = qu(r, l),
+                    h = qu(a, [-1]),
+                    p = await op(h),
+                    d = jh(p, [1]),
+                    f = Gl(c, d, s);
+                return e !== r && r.dispose(), t !== a && a.dispose(), d.dispose(), c.dispose(), h.dispose(), p.dispose(), f
+            };
+            var lp = Ms({
+                norm_: function (e, t = "euclidean", n = null, r = !1) {
+                    const a = function e(t, n, r = null) {
+                        if (0 === t.rank) return xu(t);
+                        if (1 !== t.rank && null === r) return e(qu(t, [-1]), n, r);
+                        if (1 === t.rank || "number" == typeof r || Array.isArray(r) && 1 === r.length) {
+                            if (1 === n) return yc(xu(t), r);
+                            if (n === 1 / 0) return mc(xu(t), r);
+                            if (n === -1 / 0) return zc(xu(t), r);
+                            if ("euclidean" === n || 2 === n) return Wh(yc(ah(xu(t), Ah(2, "int32")), r));
+                            throw new Error("Error in norm: invalid ord value: " + n)
+                        }
+                        if (Array.isArray(r) && 2 === r.length) {
+                            if (1 === n) return mc(yc(xu(t), r[0]), r[1] - 1);
+                            if (n === 1 / 0) return mc(yc(xu(t), r[1]), r[0]);
+                            if (n === -1 / 0) return zc(yc(xu(t), r[1]), r[0]);
+                            if ("fro" === n || "euclidean" === n) return Wh(yc(jc(t), r));
+                            throw new Error("Error in norm: invalid ord value: " + n)
+                        }
+                        throw new Error("Error in norm: invalid axis: " + r)
+                    }(e = Rs(e, "x", "norm"), t, n);
+                    let s = a.shape;
+                    if (r) {
+                        const t = ke(n, e.shape);
+                        s = kc(a.shape, t)
+                    }
+                    return qu(a, s)
+                }
+            });
+            var cp = Ms({
+                movingAverage_: function (e, t, n, r, a = !0) {
+                    const s = Rs(e, "v", "movingAverage"),
+                        i = Rs(t, "x", "movingAverage"),
+                        o = Rs(n, "decay", "movingAverage");
+                    ds(s, i), ue(fe(s.shape, i.shape), () => "Shape mismatch in v and x");
+                    const u = Ah(1),
+                        l = gc(u, o);
+                    let c = vu(gc(i, s), l);
+                    if (a) {
+                        ue(null != r, () => "When using zeroDebias: true, step is required.");
+                        const e = Rs(r, "step", "movingAverage");
+                        c = bu(c, gc(u, ah(o, e)))
+                    }
+                    return gu(s, c)
+                }
+            });
+            var hp = Ms({
+                scatterND_: function (e, t, n) {
+                    const r = Rs(e, "indices", "scatterND", "int32"),
+                        a = Rs(t, "updates", "scatterND");
+                    ho(a, r, n);
+                    const s = {
+                            indices: r,
+                            updates: a
+                        },
+                        i = {
+                            shape: n
+                        };
+                    return xs.runKernel(br, s, i)
+                }
+            });
+            var pp = Ms({
+                sparseToDense_: function (e, t, n, r = 0) {
+                    const a = Rs(e, "sparseIndices", "sparseToDense", "int32"),
+                        s = Rs(t, "sparseValues", "sparseToDense"),
+                        i = Rs(r, "defaultValue", "sparseToDense", s.dtype);
+                    ! function (e, t, n, r) {
+                        if ("int32" !== e.dtype) throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);
+                        if (e.rank > 2) throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);
+                        const a = e.rank > 0 ? e.shape[0] : 1,
+                            s = e.rank > 1 ? e.shape[1] : 1;
+                        if (n.length !== s) throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${s}.`);
+                        const i = t.size;
+                        if (0 !== t.rank && (1 !== t.rank || i !== a)) throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${a}]`);
+                        if (t.dtype !== r.dtype) throw new Error("sparseValues.dtype must match defaultValues.dtype")
+                    }(a, s, n, i);
+                    const o = {
+                            sparseIndices: a,
+                            sparseValues: s,
+                            defaultValue: i
+                        },
+                        u = {
+                            outputShape: n
+                        };
+                    return xs.runKernel(Lr, o, u)
+                }
+            });
+            var dp = Ms({
+                gatherND_: function (e, t) {
+                    const n = Rs(t, "indices", "gatherND", "int32"),
+                        r = {
+                            params: Rs(e, "x", "gatherND", "string_or_numeric"),
+                            indices: n
+                        };
+                    return xs.runKernel(hn, r)
+                }
+            });
+            var fp = Ms({
+                dropout_: function (e, t, n, r) {
+                    const a = Rs(e, "x", "dropout");
+                    if (ue("float32" === a.dtype, () => `x has to be a floating point tensor since it's going to be scaled, but got a ${a.dtype} tensor instead.`), ue(t >= 0 && t < 1, () => `rate must be a float in the range [0, 1), but got ${t}.`), 0 === t) return e instanceof Ya ? a.clone() : a;
+                    const s = function (e, t) {
+                            if (null == t) return e.shape.slice();
+                            if (fe(e.shape, t)) return t;
+                            if (e.shape.length === t.length) {
+                                const n = [];
+                                for (let r = 0; r < e.shape.length; r++) null == t[r] && null != e.shape[r] ? n.push(e.shape[r]) : n.push(t[r]);
+                                return n
+                            }
+                            return t
+                        }(a, n),
+                        i = 1 - t,
+                        o = bu(jl(gu(ph(s, 0, 1, "float32", r), i)), i);
+                    return vu(a, o)
+                }
+            });
+
+            function mp(e) {
+                return Math.floor(Math.pow(2, Math.ceil(Math.log(e) / Math.log(2))))
+            }
+
+            function gp(e, t, n) {
+                const r = 1 - e % 2,
+                    a = new Float32Array(e);
+                for (let s = 0; s < e; ++s) {
+                    const i = 2 * Math.PI * s / (e + r - 1);
+                    a[s] = t - n * Math.cos(i)
+                }
+                return Yh(a, "float32")
+            }
+            var yp = async function (e, t, n = 1) {
+                const r = Rs(e, "predictions", "inTopK"),
+                    a = Rs(t, "targets", "inTopK");
+                ue(r.rank > 1, () => "inTopK() expects the predictions to be of rank 2 or higher, but got " + r.rank), ue(r.rank - 1 === a.rank, () => `predictions rank should be 1 larger than targets rank, but got predictions rank ${r.rank} and targets rank ${a.rank}`), le(r.shape.slice(0, r.shape.length - 1), a.shape, "predictions's shape should be align with the targets' shape, except the last dimension.");
+                const s = r.shape[r.shape.length - 1];
+                ue(n > 0 && n <= s, () => `'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${n}`);
+                const i = await r.data(),
+                    o = await a.data(),
+                    [u, l] = [i.length / s, s],
+                    c = Ie("bool", u);
+                for (let e = 0; e < u; e++) {
+                    const t = e * l,
+                        r = i.subarray(t, t + l),
+                        a = [];
+                    for (let e = 0; e < r.length; e++) a.push({
+                        value: r[e],
+                        index: e
+                    });
+                    a.sort((e, t) => t.value - e.value), c[e] = 0;
+                    for (let t = 0; t < n; t++)
+                        if (a[t].index === o[e]) {
+                            c[e] = 1;
+                            break
+                        }
+                }
+                return e !== r && r.dispose(), t !== a && a.dispose(), Ls(c, a.shape, "bool")
+            }, bp = {};
+            w(bp, {
+                conv2d: () => Ip,
+                depthwiseConv2d: () => Np,
+                matMul: () => Cp
+            });
+            var vp = Ms({
+                conv2DBackpropFilter_: function (e, t, n, r, a, s = "NHWC", i) {
+                    let o = e;
+                    3 === e.rank && (o = qu(e, [1, e.shape[0], e.shape[1], e.shape[2]]));
+                    let u = t;
+                    3 === u.rank && (u = qu(t, [1, t.shape[0], t.shape[1], t.shape[2]])), ue(4 === o.rank, () => `Error in conv2dDerFilter: input must be rank 4, but got shape ${o.shape}.`), ue(4 === u.rank, () => `Error in conv2dDerFilter: dy must be rank 4, but got shape ${u.shape}.`), ue(4 === n.length, () => `Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);
+                    const l = "NHWC" === s ? o.shape[3] : o.shape[1],
+                        c = "NHWC" === s ? u.shape[3] : u.shape[1];
+                    ue(l === n[2], () => `Error in conv2dDerFilter: depth of input ${l}) must match input depth in filter (${n[2]}.`), ue(c === n[3], () => `Error in conv2dDerFilter: depth of dy (${c}) must match output depth for filter (${n[3]}).`), Hu("conv2dDerFilter", a, i);
+                    const h = {
+                            x: o,
+                            dy: u
+                        },
+                        p = {
+                            strides: r,
+                            pad: a,
+                            dataFormat: s,
+                            dimRoundingMode: i,
+                            filterShape: n
+                        };
+                    return xs.runKernel(Rt, h, p)
+                }
+            });
+
+            function xp(e, t, n) {
+                if (null == n || "linear" === n) return e;
+                if ("relu" === n) return vu(e, Hh(t));
+                throw new Error(`Cannot compute gradient for fused activation ${n}.`)
+            }
+
+            function wp(e, t) {
+                let n = t;
+                const r = Qi(e.shape, t.shape);
+                return r.length > 0 && (n = yc(n, r)), qu(n, e.shape)
+            }
+
+            function kp(e, t, n, r) {
+                if ("linear" === t) return e;
+                if ("relu" === t) return gh(e);
+                if ("elu" === t) return Ll(e);
+                if ("relu6" === t) return yh(e);
+                if ("prelu" === t) return sh(e, n);
+                if ("leakyrelu" === t) return Zl(e, r);
+                if ("sigmoid" === t) return Qu(e);
+                throw new Error(`Unknown fused activation ${t}.`)
+            }
+            var Sp = (e, t) => !(e > 0) || "linear" === t;
+            var Ip = Ms({
+                fusedConv2d_: function ({
+                    x: e,
+                    filter: t,
+                    strides: n,
+                    pad: r,
+                    dataFormat: a = "NHWC",
+                    dilations: s = [1, 1],
+                    dimRoundingMode: i,
+                    bias: o,
+                    activation: u = "linear",
+                    preluActivationWeights: l,
+                    leakyreluAlpha: c
+                }) {
+                    if (u = u || "linear", !1 === Sp(xs.state.gradientDepth, u)) {
+                        let h = ml(e, t, n, r, a, s, i);
+                        return null != o && (h = gu(h, o)), kp(h, u, l, c)
+                    }
+                    const h = Rs(e, "x", "conv2d", "float32"),
+                        p = Rs(t, "filter", "conv2d", "float32");
+                    let d = h,
+                        f = !1;
+                    3 === h.rank && (f = !0, d = qu(h, [1, h.shape[0], h.shape[1], h.shape[2]])), ue(4 === d.rank, () => `Error in fused conv2d: input must be rank 4, but got rank ${d.rank}.`), ue(4 === p.rank, () => `Error in fused conv2d: filter must be rank 4, but got rank ${p.rank}.`), Hu("fused conv2d", r, i), ue(d.shape[3] === p.shape[2], () => `Error in conv2d: depth of input (${d.shape[3]}) must match input depth for filter ${p.shape[2]}.`), ue(ju(n, s), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`), ue("NHWC" === a, () => `Error in conv2d: got dataFormat of ${a} but only NHWC is currently supported.`);
+                    const m = Lu(d.shape, p.shape, n, s, r, i);
+                    let g, y;
+                    null != o && (g = Rs(o, "bias", "fused conv2d"), [g] = ps(g, h), Zi(m.outShape, g.shape)), null != l && (y = Rs(l, "prelu weights", "fused conv2d"));
+                    const b = (e, t) => {
+                            const [a, i, o, l] = t, c = xp(e, o, u);
+                            ue(Vu(s), () => `Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);
+                            const h = [yl(i.shape, c, a, n, r), vp(i, c, a.shape, n, r)];
+                            if (null != l) {
+                                const e = wp(l, c);
+                                h.push(e)
+                            }
+                            return h
+                        },
+                        v = {
+                            x: d,
+                            filter: p,
+                            bias: g,
+                            preluActivationWeights: y
+                        },
+                        x = {
+                            strides: n,
+                            pad: r,
+                            dataFormat: a,
+                            dilations: s,
+                            dimRoundingMode: i,
+                            activation: u,
+                            leakyreluAlpha: c
+                        };
+                    if (null == o) {
+                        return cc((e, t, n) => {
+                            let r = xs.runKernel(aa, v, x);
+                            return n([t, e, r]), f && (r = qu(r, [r.shape[1], r.shape[2], r.shape[3]])), {
+                                value: r,
+                                gradFunc: b
+                            }
+                        })(d, p)
+                    }
+                    return cc((e, t, n, r) => {
+                        let a = xs.runKernel(aa, v, x);
+                        return r([t, e, a, n]), f && (a = qu(a, [a.shape[1], a.shape[2], a.shape[3]])), {
+                            value: a,
+                            gradFunc: b
+                        }
+                    })(d, p, g)
+                }
+            });
+            var Ap = Ms({
+                depthwiseConv2dNativeBackpropFilter_: function (e, t, n, r, a, s = [1, 1], i) {
+                    let o = e;
+                    3 === e.rank && (o = qu(e, [1, e.shape[0], e.shape[1], e.shape[2]]));
+                    let u = t;
+                    3 === u.rank && (u = qu(t, [1, t.shape[0], t.shape[1], t.shape[2]]));
+                    const l = {
+                            x: o,
+                            dy: u
+                        },
+                        c = {
+                            strides: r,
+                            pad: a,
+                            dimRoundingMode: i,
+                            dilations: s,
+                            filterShape: n
+                        };
+                    return xs.runKernel(Wt, l, c)
+                }
+            });
+            var Ep = Ms({
+                depthwiseConv2dNativeBackpropInput_: function (e, t, n, r, a, s = [1, 1], i) {
+                    let o = t,
+                        u = !1;
+                    3 === t.rank && (u = !0, o = qu(t, [1, t.shape[0], t.shape[1], t.shape[2]]));
+                    const l = {
+                            dy: o,
+                            filter: n
+                        },
+                        c = {
+                            strides: r,
+                            pad: a,
+                            dimRoundingMode: i,
+                            dilations: s,
+                            inputShape: e
+                        },
+                        h = xs.runKernel(Vt, l, c);
+                    return u ? qu(h, [h.shape[1], h.shape[2], h.shape[3]]) : h
+                }
+            });
+            var Np = Ms({
+                fusedDepthwiseConv2d_: function ({
+                    x: e,
+                    filter: t,
+                    strides: n,
+                    pad: r,
+                    dataFormat: a = "NHWC",
+                    dilations: s = [1, 1],
+                    dimRoundingMode: i,
+                    bias: o,
+                    activation: u = "linear",
+                    preluActivationWeights: l,
+                    leakyreluAlpha: c
+                }) {
+                    if (!1 === Sp(xs.state.gradientDepth, u)) {
+                        let h = Nl(e, t, n, r, a, s, i);
+                        return null != o && (h = gu(h, o)), kp(h, u, l, c)
+                    }
+                    const h = Rs(e, "x", "depthwiseConv2d", "float32"),
+                        p = Rs(t, "filter", "depthwiseConv2d", "float32");
+                    let d = h,
+                        f = !1;
+                    3 === h.rank && (f = !0, d = qu(h, [1, h.shape[0], h.shape[1], h.shape[2]])), ue(4 === d.rank, () => `Error in fused depthwiseConv2d: input must be rank 4, but got rank ${d.rank}.`), ue(4 === p.rank, () => `Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${p.rank}.`), ue(d.shape[3] === p.shape[2], () => `Error in fused depthwiseConv2d: number of input channels (${d.shape[3]}) must match the inChannels dimension in filter ${p.shape[2]}.`), null == s && (s = [1, 1]), ue(ju(n, s), () => `Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`), Hu("fused depthwiseConv2d", r, i);
+                    const m = Lu(d.shape, p.shape, n, s, r, i, !0);
+                    let g, y;
+                    null != o && (g = Rs(o, "bias", "fused conv2d"), [g] = ps(g, h), Zi(m.outShape, g.shape)), null != l && (y = Rs(l, "prelu weights", "fused depthwiseConv2d"));
+                    const b = (e, t) => {
+                            ue(Vu(s), () => `Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);
+                            const [a, o, l, c] = t, h = xp(e, l, u), p = Ep(o.shape, h, a, n, r, s, i), d = Ap(o, h, a.shape, n, r, s, i);
+                            if (null != c) {
+                                return [p, d, wp(g, h)]
+                            }
+                            return [p, d]
+                        },
+                        v = {
+                            x: d,
+                            filter: p,
+                            bias: g,
+                            preluActivationWeights: y
+                        },
+                        x = {
+                            strides: n,
+                            pad: r,
+                            dataFormat: a,
+                            dilations: s,
+                            dimRoundingMode: i,
+                            activation: u,
+                            leakyreluAlpha: c
+                        };
+                    if (null == o) {
+                        return cc((e, t, n) => {
+                            let r = xs.runKernel(sa, v, x);
+                            return n([t, e, r]), f && (r = qu(r, [r.shape[1], r.shape[2], r.shape[3]])), {
+                                value: r,
+                                gradFunc: b
+                            }
+                        })(d, p)
+                    }
+                    return cc((e, t, n, r) => {
+                        let a = xs.runKernel(sa, v, x);
+                        return r([t, e, a, n]), f && (a = qu(a, [a.shape[1], a.shape[2], a.shape[3]])), {
+                            value: a,
+                            gradFunc: b
+                        }
+                    })(d, p, g)
+                }
+            });
+            var Cp = Ms({
+                fusedMatMul_: function ({
+                    a: e,
+                    b: t,
+                    transposeA: n = !1,
+                    transposeB: r = !1,
+                    bias: a,
+                    activation: s = "linear",
+                    preluActivationWeights: i,
+                    leakyreluAlpha: o
+                }) {
+                    if (!1 === Sp(xs.state.gradientDepth, s)) {
+                        let u = Gi(e, t, n, r);
+                        return null != a && (u = gu(u, a)), kp(u, s, i, o)
+                    }
+                    let u = Rs(e, "a", "fused matMul"),
+                        l = Rs(t, "b", "fused matMul");
+                    [u, l] = ps(u, l);
+                    const c = n ? u.shape[u.rank - 2] : u.shape[u.rank - 1],
+                        h = r ? l.shape[l.rank - 1] : l.shape[l.rank - 2],
+                        p = n ? u.shape[u.rank - 1] : u.shape[u.rank - 2],
+                        d = r ? l.shape[l.rank - 2] : l.shape[l.rank - 1],
+                        f = u.shape.slice(0, -2),
+                        m = l.shape.slice(0, -2),
+                        g = pe(f),
+                        y = pe(m);
+                    ue(c === h, () => `Error in fused matMul: inner shapes (${c}) and (${h}) of Tensors with shapes ${u.shape} and ${l.shape} and transposeA=${n} and transposeB=${r} must match.`);
+                    const b = Zi(u.shape.slice(0, -2), l.shape.slice(0, -2)).concat([p, d]),
+                        v = qu(u, n ? [g, c, p] : [g, p, c]),
+                        x = qu(l, r ? [y, d, h] : [y, h, d]);
+                    let w, k;
+                    null != a && (w = Rs(a, "bias", "fused matMul"), [w] = ps(w, u), Zi(b, w.shape)), null != i && (k = Rs(i, "prelu weights", "fused matMul"));
+                    const S = (e, t) => {
+                            const [i, o, u, l] = t, c = xp(qu(e, u.shape), u, s);
+                            let h, p;
+                            if (n || r ? !n && r ? (h = Gi(c, o, !1, !1), p = Gi(c, i, !0, !1)) : n && !r ? (h = Gi(o, c, !1, !0), p = Gi(i, c, !1, !1)) : (h = Gi(o, c, !0, !0), p = Gi(c, i, !0, !0)) : (h = Gi(c, o, !1, !0), p = Gi(i, c, !0, !1)), null != a) {
+                                return [h, p, wp(l, c)]
+                            }
+                            return [h, p]
+                        },
+                        I = {
+                            a: v,
+                            b: x,
+                            bias: w,
+                            preluActivationWeights: k
+                        },
+                        A = {
+                            transposeA: n,
+                            transposeB: r,
+                            activation: s,
+                            leakyreluAlpha: o
+                        };
+                    if (null == a) {
+                        return cc((e, t, n) => {
+                            const r = xs.runKernel(ra, I, A);
+                            return n([e, t, r]), {
+                                value: qu(r, b),
+                                gradFunc: S
+                            }
+                        })(v, x)
+                    }
+                    return cc((e, t, n, r) => {
+                        const a = xs.runKernel(ra, I, A);
+                        return r([e, t, a, n]), {
+                            value: qu(a, b),
+                            gradFunc: S
+                        }
+                    })(v, x, w)
+                }
+            });
+            var Tp = Ms({
+                hammingWindow_: function (e) {
+                    return gp(e, .54, .46)
+                }
+            });
+            var Rp = Ms({
+                hannWindow_: function (e) {
+                    return gp(e, .5, .5)
+                }
+            });
+            var _p = Ms({
+                frame_: function (e, t, n, r = !1, a = 0) {
+                    let s = 0;
+                    const i = [];
+                    for (; s + t <= e.size;) i.push(Zu(e, s, t)), s += n;
+                    if (r)
+                        for (; s < e.size;) {
+                            const r = s + t - e.size,
+                                o = Xu([Zu(e, s, t - r), Vl([r], a)]);
+                            i.push(o), s += n
+                        }
+                    return 0 === i.length ? Xh([], [0, t]) : qu(Xu(i), [i.length, t])
+                }
+            });
+            var Fp = Ms({
+                stft_: function (e, t, n, r, a = Rp) {
+                    null == r && (r = mp(t));
+                    const s = _p(e, t, n),
+                        i = vu(s, a(t));
+                    return Uh(i, r)
+                }
+            });
+            var Mp = Ms({
+                cropAndResize_: function (e, t, n, r, a = "bilinear", s = 0) {
+                    const i = Rs(e, "image", "cropAndResize"),
+                        o = Rs(t, "boxes", "cropAndResize", "float32"),
+                        u = Rs(n, "boxInd", "cropAndResize", "int32"),
+                        l = o.shape[0];
+                    ue(4 === i.rank, () => `Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`), ue(2 === o.rank && 4 === o.shape[1], () => `Error in cropAndResize: boxes must be have size [${l},4] but had shape ${o.shape}.`), ue(1 === u.rank && u.shape[0] === l, () => `Error in cropAndResize: boxInd must be have size [${l}] but had shape ${o.shape}.`), ue(2 === r.length, () => `Error in cropAndResize: cropSize must be of length 2, but got length ${r.length}.`), ue(r[0] >= 1 && r[1] >= 1, () => "cropSize must be atleast [1,1], but was " + r), ue("bilinear" === a || "nearest" === a, () => "method must be bilinear or nearest, but was " + a);
+                    const c = {
+                            image: i,
+                            boxes: o,
+                            boxInd: u
+                        },
+                        h = {
+                            method: a,
+                            extrapolationValue: s,
+                            cropSize: r
+                        };
+                    return xs.runKernel($t, c, h)
+                }
+            });
+            var Dp = Ms({
+                flipLeftRight_: function (e) {
+                    const t = Rs(e, "image", "flipLeftRight", "float32");
+                    ue(4 === t.rank, () => `Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);
+                    const n = {
+                        image: t
+                    };
+                    return xs.runKernel(sn, n, {})
+                }
+            });
+            var Op = Ms({
+                grayscaleToRGB_: function (e) {
+                    const t = Rs(e, "image", "grayscaleToRGB"),
+                        n = t.rank - 1,
+                        r = t.shape[n];
+                    ue(t.rank >= 2, () => `Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`), ue(1 === r, () => `Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${r}.`);
+                    const a = new Array(t.rank);
+                    return a.fill(1, 0, n), a[n] = 3, Ul(t, a)
+                }
+            });
+            var Lp = Ms({
+                rotateWithOffset_: function (e, t, n = 0, r = .5) {
+                    const a = Rs(e, "image", "rotateWithOffset", "float32");
+                    ue(4 === a.rank, () => `Error in rotateWithOffset: image must be rank 4,but got rank ${a.rank}.`);
+                    const s = {
+                            image: a
+                        },
+                        i = {
+                            radians: t,
+                            fillValue: n,
+                            center: r
+                        };
+                    return xs.runKernel(na, s, i)
+                }
+            });
+
+            function Pp(e, t, n, r, a, s) {
+                null == r && (r = .5), null == a && (a = Number.NEGATIVE_INFINITY), null == s && (s = 0);
+                const i = e.shape[0];
+                return n = Math.min(n, i), ue(0 <= r && r <= 1, () => `iouThreshold must be in [0, 1], but was '${r}'`), ue(2 === e.rank, () => `boxes must be a 2D tensor, but was of rank '${e.rank}'`), ue(4 === e.shape[1], () => "boxes must have 4 columns, but 2nd dimension was " + e.shape[1]), ue(1 === t.rank, () => "scores must be a 1D tensor"), ue(t.shape[0] === i, () => `scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`), ue(0 <= s && s <= 1, () => `softNmsSigma must be in [0, 1], but was '${s}'`), {
+                    maxOutputSize: n,
+                    iouThreshold: r,
+                    scoreThreshold: a,
+                    softNmsSigma: s
+                }
+            }
+            var $p = Ms({
+                nonMaxSuppression_: function (e, t, n, r = .5, a = Number.NEGATIVE_INFINITY) {
+                    const s = Rs(e, "boxes", "nonMaxSuppression", "float32"),
+                        i = Rs(t, "scores", "nonMaxSuppression", "float32"),
+                        o = Pp(s, i, n, r, a),
+                        u = {
+                            maxOutputSize: n = o.maxOutputSize,
+                            iouThreshold: r = o.iouThreshold,
+                            scoreThreshold: a = o.scoreThreshold
+                        };
+                    return xs.runKernel(Kn, {
+                        boxes: s,
+                        scores: i
+                    }, u)
+                }
+            });
+
+            function Bp(e, t, n) {
+                const r = function (e, t, n) {
+                        return function (e, t, n) {
+                            let r = 0,
+                                a = e.length,
+                                s = 0,
+                                i = !1;
+                            for (; r < a;) {
+                                s = r + (a - r >>> 1);
+                                const o = n(t, e[s]);
+                                o > 0 ? r = s + 1 : (a = s, i = !o)
+                            }
+                            return i ? r : -r - 1
+                        }(e, t, n || zp)
+                    }(e, t, n),
+                    a = r < 0 ? -(r + 1) : r;
+                e.splice(a, 0, t)
+            }
+
+            function zp(e, t) {
+                return e > t ? 1 : e < t ? -1 : 0
+            }
+
+            function Up(e, t, n, r, a) {
+                return jp(e, t, n, r, a, 0)
+            }
+
+            function Wp(e, t, n, r, a, s) {
+                return jp(e, t, n, r, a, 0, !1, s, !0)
+            }
+
+            function Vp(e, t, n, r, a, s) {
+                return jp(e, t, n, r, a, s, !0)
+            }
+
+            function jp(e, t, n, r, a, s, i = !1, o = !1, u = !1) {
+                const l = [];
+                for (let e = 0; e < t.length; e++) t[e] > a && l.push({
+                    score: t[e],
+                    boxIndex: e,
+                    suppressBeginIndex: 0
+                });
+                l.sort(qp);
+                const c = s > 0 ? -.5 / s : 0,
+                    h = [],
+                    p = [];
+                for (; h.length < n && l.length > 0;) {
+                    const t = l.pop(),
+                        {
+                            score: n,
+                            boxIndex: s,
+                            suppressBeginIndex: i
+                        } = t;
+                    if (n < a) break;
+                    let o = !1;
+                    for (let n = h.length - 1; n >= i; --n) {
+                        const i = Gp(e, s, h[n]);
+                        if (i >= r) {
+                            o = !0;
+                            break
+                        }
+                        if (t.score = t.score * Hp(r, c, i), t.score <= a) break
+                    }
+                    t.suppressBeginIndex = h.length, o || (t.score === n ? (h.push(s), p.push(t.score)) : t.score > a && Bp(l, t, qp))
+                }
+                const d = h.length,
+                    f = n - d;
+                o && f > 0 && (h.push(...new Array(f).fill(0)), p.push(...new Array(f).fill(0)));
+                const m = {
+                    selectedIndices: h
+                };
+                return i && (m.selectedScores = p), u && (m.validOutputs = d), m
+            }
+
+            function Gp(e, t, n) {
+                const r = e.subarray(4 * t, 4 * t + 4),
+                    a = e.subarray(4 * n, 4 * n + 4),
+                    s = Math.min(r[0], r[2]),
+                    i = Math.min(r[1], r[3]),
+                    o = Math.max(r[0], r[2]),
+                    u = Math.max(r[1], r[3]),
+                    l = Math.min(a[0], a[2]),
+                    c = Math.min(a[1], a[3]),
+                    h = Math.max(a[0], a[2]),
+                    p = Math.max(a[1], a[3]),
+                    d = (o - s) * (u - i),
+                    f = (h - l) * (p - c);
+                if (d <= 0 || f <= 0) return 0;
+                const m = Math.max(s, l),
+                    g = Math.max(i, c),
+                    y = Math.min(o, h),
+                    b = Math.min(u, p),
+                    v = Math.max(y - m, 0) * Math.max(b - g, 0);
+                return v / (d + f - v)
+            }
+
+            function Hp(e, t, n) {
+                const r = Math.exp(t * n * n);
+                return n <= e ? r : 0
+            }
+
+            function qp(e, t) {
+                return e.score - t.score || e.score === t.score && t.boxIndex - e.boxIndex
+            }
+            var Kp = async function (e, t, n, r = .5, a = Number.NEGATIVE_INFINITY) {
+                const s = Rs(e, "boxes", "nonMaxSuppressionAsync"),
+                    i = Rs(t, "scores", "nonMaxSuppressionAsync"),
+                    o = Pp(s, i, n, r, a);
+                n = o.maxOutputSize, r = o.iouThreshold, a = o.scoreThreshold;
+                const u = await Promise.all([s.data(), i.data()]),
+                    l = u[0],
+                    c = u[1],
+                    {
+                        selectedIndices: h
+                    } = Up(l, c, n, r, a);
+                return s !== e && s.dispose(), i !== t && i.dispose(), Yh(h, "int32")
+            };
+            var Yp = Ms({
+                nonMaxSuppressionWithScore_: function (e, t, n, r = .5, a = Number.NEGATIVE_INFINITY, s = 0) {
+                    const i = Rs(e, "boxes", "nonMaxSuppression"),
+                        o = Rs(t, "scores", "nonMaxSuppression"),
+                        u = Pp(i, o, n, r, a, s),
+                        l = {
+                            boxes: i,
+                            scores: o
+                        },
+                        c = {
+                            maxOutputSize: n = u.maxOutputSize,
+                            iouThreshold: r = u.iouThreshold,
+                            scoreThreshold: a = u.scoreThreshold,
+                            softNmsSigma: s = u.softNmsSigma
+                        },
+                        h = xs.runKernel(Xn, l, c);
+                    return {
+                        selectedIndices: h[0],
+                        selectedScores: h[1]
+                    }
+                }
+            });
+            var Xp = async function (e, t, n, r = .5, a = Number.NEGATIVE_INFINITY, s = 0) {
+                const i = Rs(e, "boxes", "nonMaxSuppressionAsync"),
+                    o = Rs(t, "scores", "nonMaxSuppressionAsync"),
+                    u = Pp(i, o, n, r, a, s);
+                n = u.maxOutputSize, r = u.iouThreshold, a = u.scoreThreshold, s = u.softNmsSigma;
+                const l = await Promise.all([i.data(), o.data()]),
+                    c = l[0],
+                    h = l[1],
+                    {
+                        selectedIndices: p,
+                        selectedScores: d
+                    } = Vp(c, h, n, r, a, s);
+                return i !== e && i.dispose(), o !== t && o.dispose(), {
+                    selectedIndices: Yh(p, "int32"),
+                    selectedScores: Yh(d)
+                }
+            };
+            var Qp = Ms({
+                nonMaxSuppressionPadded_: function (e, t, n, r = .5, a = Number.NEGATIVE_INFINITY, s = !1) {
+                    const i = Rs(e, "boxes", "nonMaxSuppression"),
+                        o = Rs(t, "scores", "nonMaxSuppression"),
+                        u = Pp(i, o, n, r, a, null),
+                        l = {
+                            boxes: i,
+                            scores: o
+                        },
+                        c = {
+                            maxOutputSize: u.maxOutputSize,
+                            iouThreshold: u.iouThreshold,
+                            scoreThreshold: u.scoreThreshold,
+                            padToMaxOutputSize: s
+                        },
+                        h = xs.runKernel(Yn, l, c);
+                    return {
+                        selectedIndices: h[0],
+                        validOutputs: h[1]
+                    }
+                }
+            });
+            var Zp = async function (e, t, n, r = .5, a = Number.NEGATIVE_INFINITY, s = !1) {
+                const i = Rs(e, "boxes", "nonMaxSuppressionAsync"),
+                    o = Rs(t, "scores", "nonMaxSuppressionAsync"),
+                    u = Pp(i, o, n, r, a, null),
+                    l = u.maxOutputSize,
+                    c = u.iouThreshold,
+                    h = u.scoreThreshold,
+                    [p, d] = await Promise.all([i.data(), o.data()]),
+                    {
+                        selectedIndices: f,
+                        validOutputs: m
+                    } = Wp(p, d, l, c, h, s);
+                return i !== e && i.dispose(), o !== t && o.dispose(), {
+                    selectedIndices: Yh(f, "int32"),
+                    validOutputs: Ah(m, "int32")
+                }
+            };
+            var Jp = Ms({
+                resizeBilinear_: function (e, t, n = !1, r = !1) {
+                    const a = Rs(e, "images", "resizeBilinear");
+                    ue(3 === a.rank || 4 === a.rank, () => `Error in resizeBilinear: x must be rank 3 or 4, but got rank ${a.rank}.`), ue(2 === t.length, () => `Error in resizeBilinear: new shape must 2D, but got shape ${t}.`), ue(!1 === r || !1 === n, () => "Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");
+                    let s = a,
+                        i = !1;
+                    3 === a.rank && (i = !0, s = qu(a, [1, a.shape[0], a.shape[1], a.shape[2]]));
+                    const o = {
+                            images: s
+                        },
+                        u = {
+                            alignCorners: n,
+                            halfPixelCenters: r,
+                            size: t
+                        },
+                        l = xs.runKernel(pr, o, u);
+                    return i ? qu(l, [l.shape[1], l.shape[2], l.shape[3]]) : l
+                }
+            });
+            var ed = Ms({
+                resizeNearestNeighbor_: function (e, t, n = !1, r = !1) {
+                    const a = Rs(e, "images", "resizeNearestNeighbor");
+                    ue(3 === a.rank || 4 === a.rank, () => `Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${a.rank}.`), ue(2 === t.length, () => `Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`), ue("float32" === a.dtype || "int32" === a.dtype, () => "`images` must have `int32` or `float32` as dtype"), ue(!1 === r || !1 === n, () => "Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");
+                    let s = a,
+                        i = !1;
+                    3 === a.rank && (i = !0, s = qu(a, [1, a.shape[0], a.shape[1], a.shape[2]]));
+                    const o = {
+                            images: s
+                        },
+                        u = {
+                            alignCorners: n,
+                            halfPixelCenters: r,
+                            size: t
+                        },
+                        l = xs.runKernel(cr, o, u);
+                    return i ? qu(l, [l.shape[1], l.shape[2], l.shape[3]]) : l
+                }
+            });
+            var td = Ms({
+                threshold_: function (e, t = "binary", n = !1, r = .5) {
+                    const a = Rs(e, "image", "threshold"),
+                        s = a.shape[0] * a.shape[1];
+                    let i, o, u, l, c = vu(Yh([r]), 255);
+                    if (ue(3 === a.rank, () => `Error in threshold: image must be rank 3,but got rank ${a.rank}.`), ue(3 === a.shape[2] || 1 === a.shape[2], () => `Error in threshold: image color channel must be equal to 3 or 1but got ${a.shape[2]}.`), ue("int32" === a.dtype || "float32" === a.dtype, () => `Error in dtype: image dtype must be int32 or float32,but got dtype ${a.dtype}.`), ue("otsu" === t || "binary" === t, () => "Method must be binary or otsu, but was " + t), 3 === a.shape[2]) {
+                        [i, o, u] = zh(a, [1, 1, 1], -1);
+                        const e = vu(i, .2989),
+                            t = vu(o, .587),
+                            n = vu(u, .114);
+                        l = gu(gu(e, t), n)
+                    } else l = e;
+                    if ("otsu" === t) {
+                        c = function (e, t) {
+                            let n, r, a, s, i, o, u = Yh([-1]),
+                                l = Yh([0]),
+                                c = Yh([0]);
+                            for (let h = 0; h < e.size - 1; h++) {
+                                n = Zu(e, 0, h + 1), r = Zu(e, h + 1), i = bu(yc(n), t), o = bu(yc(r), t);
+                                const p = yc(vu(n, dh(0, n.size)));
+                                a = bu(p, yc(n));
+                                const d = Vl(r.shape, n.size),
+                                    f = gu(dh(0, r.size), d),
+                                    m = vu(r, f);
+                                s = bu(yc(m), yc(r));
+                                const g = gc(a, s),
+                                    y = gc(a, s),
+                                    b = vu(i, o);
+                                c = vu(vu(b, g), y);
+                                const v = Hl(c, l);
+                                l = _l(v, c, l), u = _l(v, Yh([h]), u)
+                            }
+                            return u
+                        }(il(Ai(Sh(l), "int32"), Ls([]), 256), s)
+                    }
+                    const h = n ? ec(l, c) : Hl(l, c);
+                    return Ai(vu(h, 255), "int32")
+                }
+            });
+            var nd = Ms({
+                transform_: function (e, t, n = "nearest", r = "constant", a = 0, s) {
+                    const i = Rs(e, "image", "transform", "float32"),
+                        o = Rs(t, "transforms", "transform", "float32");
+                    ue(4 === i.rank, () => `Error in transform: image must be rank 4,but got rank ${i.rank}.`), ue(2 === o.rank && (o.shape[0] === i.shape[0] || 1 === o.shape[0]) && 8 === o.shape[1], () => "Error in transform: Input transform should be batch x 8 or 1 x 8"), ue(null == s || 2 === s.length, () => `Error in transform: outputShape must be [height, width] or null, but got ${s}.`);
+                    const u = {
+                            image: i,
+                            transforms: o
+                        },
+                        l = {
+                            interpolation: n,
+                            fillMode: r,
+                            fillValue: a,
+                            outputShape: s
+                        };
+                    return xs.runKernel(Kr, u, l)
+                }
+            });
+            var rd = Ms({
+                bandPart_: function (e, t, n) {
+                    ue(t % 1 == 0, () => `bandPart(): numLower must be an integer, got ${t}.`), ue(n % 1 == 0, () => `bandPart(): numUpper must be an integer, got ${n}.`);
+                    const r = Rs(e, "a", "bandPart");
+                    ue(r.rank >= 2, () => `bandPart(): Rank must be at least 2, got ${r.rank}.`);
+                    const a = r.shape,
+                        [s, i] = r.shape.slice(-2);
+                    if (!(t <= s)) throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);
+                    if (!(n <= i)) throw new Error(`bandPart(): numUpper (${n}) must not be greater than the number of columns (${i}).`);
+                    t < 0 && (t = s), n < 0 && (n = i);
+                    const o = qu(dh(0, s, 1, "int32"), [-1, 1]),
+                        u = dh(0, i, 1, "int32"),
+                        l = gc(o, u),
+                        c = Cc(ec(l, Ah(+t, "int32")), ql(l, Ah(-n, "int32"))),
+                        h = Pc([s, i], r.dtype);
+                    return qu(Gh(ap(qu(r, [-1, s, i])).map(e => _l(c, e, h))), a)
+                }
+            });
+            var ad = Ms({
+                gramSchmidt_: function (e) {
+                    let t;
+                    if (Array.isArray(e)) {
+                        t = !1, ue(null != e && e.length > 0, () => "Gram-Schmidt process: input must not be null, undefined, or empty");
+                        const n = e[0].shape[0];
+                        for (let t = 1; t < e.length; ++t) ue(e[t].shape[0] === n, () => `Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[t].shape[0]} vs. ${n})`)
+                    } else t = !0, e = zh(e, e.shape[0], 0).map(e => jh(e, [0]));
+                    ue(e.length <= e[0].shape[0], () => `Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);
+                    const n = [],
+                        r = e;
+                    for (let t = 0; t < e.length; ++t) n.push(xs.tidy(() => {
+                        let e = r[t];
+                        if (t > 0)
+                            for (let r = 0; r < t; ++r) {
+                                const t = vu(yc(vu(n[r], e)), n[r]);
+                                e = gc(e, t)
+                            }
+                        return bu(e, lp(e, "euclidean"))
+                    }));
+                    return t ? Gh(n, 0) : n
+                }
+            });
+
+            function sd(e, t = !1) {
+                return xs.tidy(() => {
+                    ue(2 === e.shape.length, () => `qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);
+                    const n = e.shape[0],
+                        r = e.shape[1];
+                    let a = Wl(n),
+                        s = Ei(e);
+                    const i = Xh([
+                        [1]
+                    ], [1, 1]);
+                    let o = Ei(i);
+                    const u = n >= r ? r : n;
+                    for (let e = 0; e < u; ++e) {
+                        const t = s,
+                            u = o,
+                            l = a;
+                        [o, s, a] = xs.tidy(() => {
+                            const t = Zu(s, [e, e], [n - e, 1]),
+                                u = lp(t),
+                                l = Zu(s, [e, e], [1, 1]),
+                                c = _l(Hl(l, 0), Xh([
+                                    [-1]
+                                ]), Xh([
+                                    [1]
+                                ])),
+                                h = gc(l, vu(c, u)),
+                                p = bu(t, h);
+                            o = 1 === p.shape[0] ? Ei(i) : Xu([i, Zu(p, [1, 0], [p.shape[0] - 1, p.shape[1]])], 0);
+                            const d = pc(bu(Gi(c, h), u)),
+                                f = Zu(s, [e, 0], [n - e, r]),
+                                m = vu(d, o),
+                                g = qi(o);
+                            if (0 === e) s = gc(f, Gi(m, Gi(g, f)));
+                            else {
+                                const t = gc(f, Gi(m, Gi(g, f)));
+                                s = Xu([Zu(s, [0, 0], [e, r]), t], 0)
+                            }
+                            const y = qi(m),
+                                b = Zu(a, [0, e], [n, a.shape[1] - e]);
+                            if (0 === e) a = gc(b, Gi(Gi(b, o), y));
+                            else {
+                                const t = gc(b, Gi(Gi(b, o), y));
+                                a = Xu([Zu(a, [0, 0], [n, e]), t], 1)
+                            }
+                            return [o, s, a]
+                        }), au([t, u, l])
+                    }
+                    return !t && n > r && (a = Zu(a, [0, 0], [n, r]), s = Zu(s, [0, 0], [r, r])), [a, s]
+                })
+            }
+            var id, od, ud = Ms({
+                qr_: function (e, t = !1) {
+                    if (ue(e.rank >= 2, () => "qr() requires input tensor to have a rank >= 2, but got rank " + e.rank), 2 === e.rank) return sd(e, t); {
+                        const n = e.shape.slice(0, e.shape.length - 2).reduce((e, t) => e * t),
+                            r = ap(qu(e, [n, e.shape[e.shape.length - 2], e.shape[e.shape.length - 1]]), 0),
+                            a = [],
+                            s = [];
+                        r.forEach(e => {
+                            const [n, r] = sd(e, t);
+                            a.push(n), s.push(r)
+                        });
+                        return [qu(Gh(a, 0), e.shape), qu(Gh(s, 0), e.shape)]
+                    }
+                }
+            });
+            (od = id || (id = {}))[od.NONE = 0] = "NONE", od[od.MEAN = 1] = "MEAN", od[od.SUM = 2] = "SUM", od[od.SUM_BY_NONZERO_WEIGHTS = 3] = "SUM_BY_NONZERO_WEIGHTS";
+            var ld = Ms({
+                computeWeightedLoss_: function (e, t, n = id.SUM_BY_NONZERO_WEIGHTS) {
+                    const r = Rs(e, "losses", "computeWeightedLoss");
+                    let a = null;
+                    null != t && (a = Rs(t, "weights", "computeWeightedLoss"));
+                    const s = null == a ? r : vu(r, a);
+                    if (n === id.NONE) return s;
+                    if (n === id.SUM) return yc(s);
+                    if (n === id.MEAN) {
+                        if (null == a) return Lc(s); {
+                            const e = r.size / a.size,
+                                t = bu(yc(s), yc(a));
+                            return e > 1 ? bu(t, Ah(e)) : t
+                        }
+                    }
+                    if (n === id.SUM_BY_NONZERO_WEIGHTS) {
+                        if (null == a) return bu(yc(s), Ah(r.size)); {
+                            const e = vu(a, $c(r.shape)),
+                                t = Ai(yc(Kc(e, Ah(0))), "float32");
+                            return bu(yc(s), t)
+                        }
+                    }
+                    throw Error("Unknown reduction: " + n)
+                }
+            });
+            var cd = Ms({
+                absoluteDifference_: function (e, t, n, r = id.SUM_BY_NONZERO_WEIGHTS) {
+                    const a = Rs(e, "labels", "absoluteDifference"),
+                        s = Rs(t, "predictions", "absoluteDifference");
+                    let i = null;
+                    null != n && (i = Rs(n, "weights", "absoluteDifference")), le(a.shape, s.shape, "Error in absoluteDifference: ");
+                    const o = xu(gc(a, s));
+                    return ld(o, i, r)
+                }
+            });
+            var hd = Ms({
+                cosineDistance_: function (e, t, n, r, a = id.SUM_BY_NONZERO_WEIGHTS) {
+                    const s = Rs(e, "labels", "cosineDistance"),
+                        i = Rs(t, "predictions", "cosineDistance");
+                    let o = null;
+                    null != r && (o = Rs(r, "weights", "cosineDistance")), le(s.shape, i.shape, "Error in cosineDistance: ");
+                    const u = Ah(1),
+                        l = gc(u, yc(vu(s, i), n, !0));
+                    return ld(l, o, a)
+                }
+            });
+            var pd = Ms({
+                hingeLoss_: function (e, t, n, r = id.SUM_BY_NONZERO_WEIGHTS) {
+                    let a = Rs(e, "labels", "hingeLoss");
+                    const s = Rs(t, "predictions", "hingeLoss");
+                    let i = null;
+                    null != n && (i = Rs(n, "weights", "hingeLoss")), le(a.shape, s.shape, "Error in hingeLoss: ");
+                    const o = Ah(1);
+                    a = gc(vu(Ah(2), a), o);
+                    const u = gh(gc(o, vu(a, s)));
+                    return ld(u, i, r)
+                }
+            });
+            var dd = Ms({
+                huberLoss_: function (e, t, n, r = 1, a = id.SUM_BY_NONZERO_WEIGHTS) {
+                    const s = Rs(e, "labels", "huberLoss"),
+                        i = Rs(t, "predictions", "huberLoss");
+                    let o = null;
+                    null != n && (o = Rs(n, "weights", "huberLoss")), le(s.shape, i.shape, "Error in huberLoss: ");
+                    const u = Ah(r),
+                        l = xu(gc(i, s)),
+                        c = Uc(l, u),
+                        h = gc(l, c),
+                        p = gu(vu(Ah(.5), jc(c)), vu(u, h));
+                    return ld(p, o, a)
+                }
+            });
+            var fd = Ms({
+                logLoss_: function (e, t, n, r = 1e-7, a = id.SUM_BY_NONZERO_WEIGHTS) {
+                    const s = Rs(e, "labels", "logLoss"),
+                        i = Rs(t, "predictions", "logLoss");
+                    let o = null;
+                    null != n && (o = Rs(n, "weights", "logLoss")), le(s.shape, i.shape, "Error in logLoss: ");
+                    const u = Ah(1),
+                        l = Ah(r),
+                        c = pc(vu(s, rc(gu(i, l)))),
+                        h = vu(gc(u, s), rc(gu(gc(u, i), l))),
+                        p = gc(c, h);
+                    return ld(p, o, a)
+                }
+            });
+            var md = Ms({
+                meanSquaredError_: function (e, t, n, r = id.SUM_BY_NONZERO_WEIGHTS) {
+                    const a = Rs(e, "labels", "meanSquaredError"),
+                        s = Rs(t, "predictions", "meanSquaredError");
+                    let i = null;
+                    null != n && (i = Rs(n, "weights", "meanSquaredError")), le(a.shape, s.shape, "Error in meanSquaredError: ");
+                    const o = Vh(a, s);
+                    return ld(o, i, r)
+                }
+            });
+            var gd = Ms({
+                sigmoidCrossEntropy_: function (e, t, n, r = 0, a = id.SUM_BY_NONZERO_WEIGHTS) {
+                    let s = Rs(e, "multiClassLabels", "sigmoidCrossEntropy");
+                    const i = Rs(t, "logits", "sigmoidCrossEntropy");
+                    let o = null;
+                    if (null != n && (o = Rs(n, "weights", "sigmoidCrossEntropy")), le(s.shape, i.shape, "Error in sigmoidCrossEntropy: "), r > 0) {
+                        const e = Ah(r),
+                            t = Ah(1),
+                            n = Ah(.5);
+                        s = gu(vu(s, gc(t, e)), vu(n, e))
+                    }
+                    const u = function (e, t) {
+                        const n = Rs(e, "labels", "sigmoidCrossEntropyWithLogits"),
+                            r = Rs(t, "logits", "sigmoidCrossEntropyWithLogits");
+                        le(n.shape, r.shape, "Error in sigmoidCrossEntropyWithLogits: ");
+                        const a = gh(r),
+                            s = vu(r, n),
+                            i = ac($l(pc(xu(r))));
+                        return gu(gc(a, s), i)
+                    }(s, i);
+                    return ld(u, o, a)
+                }
+            });
+            var yd = Ms({
+                softmaxCrossEntropy_: function (e, t, n, r = 0, a = id.SUM_BY_NONZERO_WEIGHTS) {
+                    let s = Rs(e, "onehotLabels", "softmaxCrossEntropy");
+                    const i = Rs(t, "logits", "softmaxCrossEntropy");
+                    let o = null;
+                    if (null != n && (o = Rs(n, "weights", "softmaxCrossEntropy")), le(s.shape, i.shape, "Error in softmaxCrossEntropy: "), r > 0) {
+                        const e = Ah(r),
+                            t = Ah(1),
+                            n = Ah(s.shape[1]);
+                        s = gu(vu(s, gc(t, e)), bu(e, n))
+                    }
+                    const u = function (e, t, n = -1) {
+                        if (-1 === n && (n = t.rank - 1), n !== t.rank - 1) throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${n}`);
+                        return cc((e, t, r) => {
+                            const a = Nc(t, [n], !0),
+                                s = gc(Ai(t, "float32"), a);
+                            r([e, s]);
+                            const i = pc(vu(s, e));
+                            return {
+                                value: yc(i, [n]),
+                                gradFunc: (e, t) => {
+                                    const [r, a] = t, s = kc(e.shape, [n]);
+                                    return [vu(qu(e, s), gc(Ai(r, "float32"), $l(a))), vu(qu(e, s), gc($l(a), Ai(r, "float32")))]
+                                }
+                            }
+                        })(e, t)
+                    }(s, i);
+                    return ld(u, o, a)
+                }
+            });
+            var bd = Ms({
+                sparseFillEmptyRows_: function (e, t, n, r) {
+                    const a = Rs(e, "indices", "sparseFillEmptyRows", "int32"),
+                        s = Rs(t, "values", "sparseFillEmptyRows"),
+                        i = Rs(n, "denseShape", "sparseFillEmptyRows", "int32"),
+                        o = Rs(r, "defaultValue", "sparseFillEmptyRows", s.dtype);
+                    if (2 !== a.rank) throw new Error("Indices should be Tensor2D but received shape\n        " + a.shape);
+                    if (1 !== s.rank) throw new Error("Values should be Tensor1D but received shape " + s.shape);
+                    if (1 !== i.rank) throw new Error("Dense shape should be Tensor1D but received shape " + i.shape);
+                    if (0 !== o.rank) throw new Error("Default value should be a scalar but received shape " + o.shape);
+                    const u = {
+                            indices: a,
+                            values: s,
+                            denseShape: i,
+                            defaultValue: o
+                        },
+                        l = xs.runKernel(Fr, u);
+                    return {
+                        outputIndices: l[0],
+                        outputValues: l[1],
+                        emptyRowIndicator: l[2],
+                        reverseIndexMap: l[3]
+                    }
+                }
+            });
+            var vd = Ms({
+                sparseReshape_: function (e, t, n) {
+                    const r = Rs(e, "inputIndices", "sparseReshape", "int32"),
+                        a = Rs(t, "inputShape", "sparseReshape", "int32"),
+                        s = Rs(n, "newShape", "sparseReshape", "int32");
+                    if (2 !== r.rank) throw new Error("Input indices should be Tensor2D but received shape\n        " + r.shape);
+                    if (1 !== a.rank) throw new Error("Input shape should be Tensor1D but received shape " + a.shape);
+                    if (1 !== s.rank) throw new Error("New shape should be Tensor1D but received shape " + s.shape);
+                    const i = {
+                            inputIndices: r,
+                            inputShape: a,
+                            newShape: s
+                        },
+                        o = xs.runKernel(Mr, i);
+                    return {
+                        outputIndices: o[0],
+                        outputShape: o[1]
+                    }
+                }
+            });
+            var xd = Ms({
+                sparseSegmentMean_: function (e, t, n) {
+                    const r = Rs(e, "data", "sparseSegmentMean"),
+                        a = Rs(t, "indices", "sparseSegmentMean", "int32"),
+                        s = Rs(n, "segmentIds", "sparseSegmentMean", "int32");
+                    if (r.rank < 1) throw new Error("Data should be at least 1 dimensional but received scalar");
+                    if (1 !== a.rank) throw new Error("Indices should be Tensor1D but received shape\n          " + a.shape);
+                    if (1 !== s.rank) throw new Error("Segment ids should be Tensor1D but received shape\n          " + s.shape);
+                    const i = {
+                        data: r,
+                        indices: a,
+                        segmentIds: s
+                    };
+                    return xs.runKernel(Dr, i)
+                }
+            });
+            var wd = Ms({
+                sparseSegmentSum_: function (e, t, n) {
+                    const r = Rs(e, "data", "sparseSegmentSum"),
+                        a = Rs(t, "indices", "sparseSegmentSum", "int32"),
+                        s = Rs(n, "segmentIds", "sparseSegmentSum", "int32");
+                    if (r.rank < 1) throw new Error("Data should be at least 1 dimensional but received scalar");
+                    if (1 !== a.rank) throw new Error("Indices should be Tensor1D but received shape\n         " + a.shape);
+                    if (1 !== s.rank) throw new Error("Segment ids should be Tensor1D but received shape\n         " + s.shape);
+                    const i = {
+                        data: r,
+                        indices: a,
+                        segmentIds: s
+                    };
+                    return xs.runKernel(Or, i)
+                }
+            });
+            var kd = Ms({
+                stringNGrams_: function (e, t, n, r, a, s, i, o) {
+                    const u = Rs(e, "data", "stringNGrams", "string");
+                    if ("string" !== u.dtype) throw new Error("Data must be of datatype string");
+                    if (1 !== u.shape.length) throw new Error("Data must be a vector, saw: " + u.shape);
+                    const l = Rs(t, "dataSplits", "stringNGrams");
+                    if ("int32" !== l.dtype) throw new Error("Data splits must be of datatype int32");
+                    const c = {
+                            separator: n,
+                            nGramWidths: r,
+                            leftPad: a,
+                            rightPad: s,
+                            padWidth: i,
+                            preserveShortSequences: o
+                        },
+                        h = {
+                            data: u,
+                            dataSplits: l
+                        },
+                        p = xs.runKernel(zr, h, c);
+                    return {
+                        nGrams: p[0],
+                        nGramsSplits: p[1]
+                    }
+                }
+            });
+            var Sd = Ms({
+                stringSplit_: function (e, t, n = !0) {
+                    const r = Rs(e, "input", "stringSplit", "string"),
+                        a = Rs(t, "delimiter", "stringSplit", "string");
+                    if (1 !== r.rank) throw new Error("Input should be Tensor1D but received shape " + r.shape);
+                    if (0 !== a.rank) throw new Error("Delimiter should be a scalar but received shape " + a.shape);
+                    const s = {
+                            skipEmpty: n
+                        },
+                        i = {
+                            input: r,
+                            delimiter: a
+                        },
+                        o = xs.runKernel(Ur, i, s);
+                    return {
+                        indices: o[0],
+                        values: o[1],
+                        shape: o[2]
+                    }
+                }
+            });
+            var Id = Ms({
+                    stringToHashBucketFast_: function (e, t) {
+                        const n = Rs(e, "input", "stringToHashBucketFast", "string"),
+                            r = {
+                                numBuckets: t
+                            };
+                        if (t <= 0) throw new Error("Number of buckets must be at least 1");
+                        const a = {
+                            input: n
+                        };
+                        return xs.runKernel(Wr, a, r)
+                    }
+                }),
+                Ad = {
+                    fft: Ph,
+                    ifft: $h,
+                    rfft: Uh,
+                    irfft: Bh
+                },
+                Ed = {
+                    hammingWindow: Tp,
+                    hannWindow: Rp,
+                    frame: _p,
+                    stft: Fp
+                },
+                Nd = {
+                    flipLeftRight: Dp,
+                    grayscaleToRGB: Op,
+                    resizeNearestNeighbor: ed,
+                    resizeBilinear: Jp,
+                    rotateWithOffset: Lp,
+                    cropAndResize: Mp,
+                    nonMaxSuppression: $p,
+                    nonMaxSuppressionAsync: Kp,
+                    nonMaxSuppressionWithScore: Yp,
+                    nonMaxSuppressionWithScoreAsync: Xp,
+                    nonMaxSuppressionPadded: Qp,
+                    nonMaxSuppressionPaddedAsync: Zp,
+                    threshold: td,
+                    transform: nd
+                },
+                Cd = {
+                    bandPart: rd,
+                    gramSchmidt: ad,
+                    qr: ud
+                },
+                Td = {
+                    absoluteDifference: cd,
+                    computeWeightedLoss: ld,
+                    cosineDistance: hd,
+                    hingeLoss: pd,
+                    huberLoss: dd,
+                    logLoss: fd,
+                    meanSquaredError: md,
+                    sigmoidCrossEntropy: gd,
+                    softmaxCrossEntropy: yd
+                },
+                Rd = {
+                    sparseFillEmptyRows: bd,
+                    sparseReshape: vd,
+                    sparseSegmentMean: xd,
+                    sparseSegmentSum: wd
+                },
+                _d = {
+                    stringNGrams: kd,
+                    stringSplit: Sd,
+                    stringToHashBucketFast: Id
+                },
+                Fd = class extends Mo {
+                    minimize(e, t = !1, n) {
+                        const {
+                            value: r,
+                            grads: a
+                        } = this.computeGradients(e, n);
+                        if (null != n) {
+                            const e = n.map(e => ({
+                                name: e.name,
+                                tensor: a[e.name]
+                            }));
+                            this.applyGradients(e)
+                        } else this.applyGradients(a);
+                        return au(a), t ? r : (r.dispose(), null)
+                    }
+                    get iterations() {
+                        return null == this.iterations_ && (this.iterations_ = 0), this.iterations_
+                    }
+                    incrementIterations() {
+                        this.iterations_ = this.iterations + 1
+                    }
+                    computeGradients(e, t) {
+                        return lc(e, t)
+                    }
+                    dispose() {
+                        null != this.iterations_ && au(this.iterations_)
+                    }
+                    async saveIterations() {
+                        return null == this.iterations_ && (this.iterations_ = 0), {
+                            name: "iter",
+                            tensor: Ah(this.iterations_, "int32")
+                        }
+                    }
+                    async getWeights() {
+                        throw new Error("getWeights() is not implemented for this optimizer yet.")
+                    }
+                    async setWeights(e) {
+                        throw new Error("setWeights() is not implemented for this optimizer class " + this.getClassName())
+                    }
+                    async extractIterations(e) {
+                        return this.iterations_ = (await e[0].tensor.data())[0], e.slice(1)
+                    }
+                };
+            Object.defineProperty(Fd, Symbol.hasInstance, {
+                value: e => null != e.minimize && null != e.computeGradients && null != e.applyGradients
+            });
+            var Md = class extends Fd {
+                constructor(e, t, n = null) {
+                    super(), this.learningRate = e, this.rho = t, this.epsilon = n, this.accumulatedGrads = [], this.accumulatedUpdates = [], null == n && (this.epsilon = xs.backend.epsilon())
+                }
+                applyGradients(e) {
+                    (Array.isArray(e) ? e.map(e => e.name) : Object.keys(e)).forEach((t, n) => {
+                        const r = xs.registeredVariables[t];
+                        null == this.accumulatedGrads[n] && (this.accumulatedGrads[n] = {
+                            originalName: t + "/accum_grad",
+                            variable: ru(() => Fl(r).variable(!1))
+                        }), null == this.accumulatedUpdates[n] && (this.accumulatedUpdates[n] = {
+                            originalName: t + "/accum_var",
+                            variable: ru(() => Fl(r).variable(!1))
+                        });
+                        const a = Array.isArray(e) ? e[n].tensor : e[t];
+                        if (null == a) return;
+                        const s = this.accumulatedGrads[n].variable,
+                            i = this.accumulatedUpdates[n].variable;
+                        ru(() => {
+                            const e = gu(vu(s, this.rho), vu(jc(a), 1 - this.rho)),
+                                t = vu(bu(Wh(gu(i, this.epsilon)), Wh(gu(s, this.epsilon))), a),
+                                n = gu(vu(i, this.rho), vu(jc(t), 1 - this.rho));
+                            s.assign(e), i.assign(n);
+                            const o = gu(vu(t, -this.learningRate), r);
+                            r.assign(o)
+                        })
+                    }), this.incrementIterations()
+                }
+                dispose() {
+                    null != this.accumulatedUpdates && (au(this.accumulatedGrads.map(e => e.variable)), au(this.accumulatedUpdates.map(e => e.variable)))
+                }
+                async getWeights() {
+                    const e = [...this.accumulatedGrads, ...this.accumulatedUpdates];
+                    return [await this.saveIterations()].concat(e.map(e => ({
+                        name: e.originalName,
+                        tensor: e.variable
+                    })))
+                }
+                async setWeights(e) {
+                    const t = (e = await this.extractIterations(e)).length / 2;
+                    this.accumulatedGrads = e.slice(0, t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    })), this.accumulatedUpdates = e.slice(t, 2 * t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    }))
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate,
+                        rho: this.rho,
+                        epsilon: this.epsilon
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate, t.rho, t.epsilon)
+                }
+            };
+            Md.className = "Adadelta", Oo(Md);
+            var Dd = class extends Fd {
+                constructor(e, t = .1) {
+                    super(), this.learningRate = e, this.initialAccumulatorValue = t, this.accumulatedGrads = []
+                }
+                applyGradients(e) {
+                    (Array.isArray(e) ? e.map(e => e.name) : Object.keys(e)).forEach((t, n) => {
+                        const r = xs.registeredVariables[t];
+                        if (null == this.accumulatedGrads[n]) {
+                            const e = !1;
+                            this.accumulatedGrads[n] = {
+                                originalName: t + "/accumulator",
+                                variable: ru(() => Vl(r.shape, this.initialAccumulatorValue).variable(e))
+                            }
+                        }
+                        const a = Array.isArray(e) ? e[n].tensor : e[t];
+                        if (null == a) return;
+                        const s = this.accumulatedGrads[n].variable;
+                        ru(() => {
+                            const e = gu(s, jc(a));
+                            s.assign(e);
+                            const t = gu(vu(bu(a, Wh(gu(e, xs.backend.epsilon()))), -this.learningRate), r);
+                            r.assign(t)
+                        })
+                    }), this.incrementIterations()
+                }
+                dispose() {
+                    null != this.accumulatedGrads && au(this.accumulatedGrads.map(e => e.variable))
+                }
+                async getWeights() {
+                    return [await this.saveIterations()].concat(this.accumulatedGrads.map(e => ({
+                        name: e.originalName,
+                        tensor: e.variable
+                    })))
+                }
+                async setWeights(e) {
+                    e = await this.extractIterations(e);
+                    this.accumulatedGrads = e.map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    }))
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate,
+                        initialAccumulatorValue: this.initialAccumulatorValue
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate, t.initialAccumulatorValue)
+                }
+            };
+            Dd.className = "Adagrad", Oo(Dd);
+            var Od = class extends Fd {
+                constructor(e, t, n, r = null) {
+                    super(), this.learningRate = e, this.beta1 = t, this.beta2 = n, this.epsilon = r, this.accumulatedFirstMoment = [], this.accumulatedSecondMoment = [], ru(() => {
+                        this.accBeta1 = Ah(t).variable(), this.accBeta2 = Ah(n).variable()
+                    }), null == r && (this.epsilon = xs.backend.epsilon())
+                }
+                applyGradients(e) {
+                    const t = Array.isArray(e) ? e.map(e => e.name) : Object.keys(e);
+                    ru(() => {
+                        const n = gc(1, this.accBeta1),
+                            r = gc(1, this.accBeta2);
+                        t.forEach((t, a) => {
+                            const s = xs.registeredVariables[t];
+                            null == this.accumulatedFirstMoment[a] && (this.accumulatedFirstMoment[a] = {
+                                originalName: t + "/m",
+                                variable: ru(() => Fl(s).variable(!1))
+                            }), null == this.accumulatedSecondMoment[a] && (this.accumulatedSecondMoment[a] = {
+                                originalName: t + "/v",
+                                variable: ru(() => Fl(s).variable(!1))
+                            });
+                            const i = Array.isArray(e) ? e[a].tensor : e[t];
+                            if (null == i) return;
+                            const o = this.accumulatedFirstMoment[a].variable,
+                                u = this.accumulatedSecondMoment[a].variable,
+                                l = gu(vu(o, this.beta1), vu(i, 1 - this.beta1)),
+                                c = gu(vu(u, this.beta2), vu(jc(i), 1 - this.beta2)),
+                                h = bu(l, n),
+                                p = bu(c, r);
+                            o.assign(l), u.assign(c);
+                            const d = gu(vu(bu(h, gu(Wh(p), this.epsilon)), -this.learningRate), s);
+                            s.assign(d)
+                        }), this.accBeta1.assign(vu(this.accBeta1, this.beta1)), this.accBeta2.assign(vu(this.accBeta2, this.beta2))
+                    }), this.incrementIterations()
+                }
+                dispose() {
+                    this.accBeta1.dispose(), this.accBeta2.dispose(), null != this.accumulatedFirstMoment && au(this.accumulatedFirstMoment.map(e => e.variable)), null != this.accumulatedSecondMoment && au(this.accumulatedSecondMoment.map(e => e.variable))
+                }
+                async getWeights() {
+                    const e = [...this.accumulatedFirstMoment, ...this.accumulatedSecondMoment];
+                    return [await this.saveIterations()].concat(e.map(e => ({
+                        name: e.originalName,
+                        tensor: e.variable
+                    })))
+                }
+                async setWeights(e) {
+                    e = await this.extractIterations(e), ru(() => {
+                        this.accBeta1.assign(ah(this.beta1, this.iterations_ + 1)), this.accBeta2.assign(ah(this.beta2, this.iterations_ + 1))
+                    });
+                    const t = e.length / 2;
+                    this.accumulatedFirstMoment = e.slice(0, t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    })), this.accumulatedSecondMoment = e.slice(t, 2 * t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    }))
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate,
+                        beta1: this.beta1,
+                        beta2: this.beta2,
+                        epsilon: this.epsilon
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate, t.beta1, t.beta2, t.epsilon)
+                }
+            };
+            Od.className = "Adam", Oo(Od);
+            var Ld = class extends Fd {
+                constructor(e, t, n, r = null, a = 0) {
+                    super(), this.learningRate = e, this.beta1 = t, this.beta2 = n, this.epsilon = r, this.decay = a, this.accumulatedFirstMoment = [], this.accumulatedWeightedInfNorm = [], ru(() => {
+                        this.iteration = Ah(0).variable(), this.accBeta1 = Ah(t).variable()
+                    }), null == r && (this.epsilon = xs.backend.epsilon())
+                }
+                applyGradients(e) {
+                    const t = Array.isArray(e) ? e.map(e => e.name) : Object.keys(e);
+                    ru(() => {
+                        const n = gc(1, this.accBeta1),
+                            r = bu(-this.learningRate, gu(vu(this.iteration, this.decay), 1));
+                        t.forEach((t, a) => {
+                            const s = xs.registeredVariables[t];
+                            null == this.accumulatedFirstMoment[a] && (this.accumulatedFirstMoment[a] = {
+                                originalName: t + "/m",
+                                variable: Fl(s).variable(!1)
+                            }), null == this.accumulatedWeightedInfNorm[a] && (this.accumulatedWeightedInfNorm[a] = {
+                                originalName: t + "/v",
+                                variable: Fl(s).variable(!1)
+                            });
+                            const i = Array.isArray(e) ? e[a].tensor : e[t];
+                            if (null == i) return;
+                            const o = this.accumulatedFirstMoment[a].variable,
+                                u = this.accumulatedWeightedInfNorm[a].variable,
+                                l = gu(vu(o, this.beta1), vu(i, 1 - this.beta1)),
+                                c = vu(u, this.beta2),
+                                h = xu(i),
+                                p = Oc(c, h);
+                            o.assign(l), u.assign(p);
+                            const d = gu(vu(bu(r, n), bu(l, gu(p, this.epsilon))), s);
+                            s.assign(d)
+                        }), this.iteration.assign(gu(this.iteration, 1)), this.accBeta1.assign(vu(this.accBeta1, this.beta1))
+                    }), this.incrementIterations()
+                }
+                dispose() {
+                    this.accBeta1.dispose(), this.iteration.dispose(), null != this.accumulatedFirstMoment && au(this.accumulatedFirstMoment.map(e => e.variable)), null != this.accumulatedWeightedInfNorm && au(this.accumulatedWeightedInfNorm.map(e => e.variable))
+                }
+                async getWeights() {
+                    throw new Error("getWeights() is not implemented for Adamax yet.")
+                }
+                async setWeights(e) {
+                    throw new Error("setWeights() is not implemented for Adamax yet.")
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate,
+                        beta1: this.beta1,
+                        beta2: this.beta2,
+                        epsilon: this.epsilon,
+                        decay: this.decay
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate, t.beta1, t.beta2, t.epsilon, t.decay)
+                }
+            };
+            Ld.className = "Adamax", Oo(Ld);
+            var Pd = class extends Fd {
+                constructor(e) {
+                    super(), this.learningRate = e, this.setLearningRate(e)
+                }
+                applyGradients(e) {
+                    (Array.isArray(e) ? e.map(e => e.name) : Object.keys(e)).forEach((t, n) => {
+                        const r = Array.isArray(e) ? e[n].tensor : e[t];
+                        if (null == r) return;
+                        const a = xs.registeredVariables[t];
+                        ru(() => {
+                            const e = gu(vu(this.c, r), a);
+                            a.assign(e)
+                        })
+                    }), this.incrementIterations()
+                }
+                setLearningRate(e) {
+                    this.learningRate = e, null != this.c && this.c.dispose(), this.c = su(Ah(-e))
+                }
+                dispose() {
+                    this.c.dispose()
+                }
+                async getWeights() {
+                    return [await this.saveIterations()]
+                }
+                async setWeights(e) {
+                    if (0 !== (e = await this.extractIterations(e)).length) throw new Error("SGD optimizer does not have settable weights.")
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate)
+                }
+            };
+            Pd.className = "SGD", Oo(Pd);
+            var $d = class extends Pd {
+                constructor(e, t, n = !1) {
+                    super(e), this.learningRate = e, this.momentum = t, this.useNesterov = n, this.accumulations = [], this.m = Ah(this.momentum)
+                }
+                applyGradients(e) {
+                    (Array.isArray(e) ? e.map(e => e.name) : Object.keys(e)).forEach((t, n) => {
+                        const r = xs.registeredVariables[t];
+                        if (null == this.accumulations[n]) {
+                            const e = !1;
+                            this.accumulations[n] = {
+                                originalName: t + "/momentum",
+                                variable: ru(() => Fl(r).variable(e))
+                            }
+                        }
+                        const a = this.accumulations[n].variable,
+                            s = Array.isArray(e) ? e[n].tensor : e[t];
+                        null != s && ru(() => {
+                            let e;
+                            const t = gu(vu(this.m, a), s);
+                            e = this.useNesterov ? gu(vu(this.c, gu(s, vu(t, this.m))), r) : gu(vu(this.c, t), r), a.assign(t), r.assign(e)
+                        })
+                    }), this.incrementIterations()
+                }
+                dispose() {
+                    this.m.dispose(), null != this.accumulations && au(this.accumulations.map(e => e.variable))
+                }
+                setMomentum(e) {
+                    this.momentum = e
+                }
+                async getWeights() {
+                    return [await this.saveIterations()].concat(this.accumulations.map(e => ({
+                        name: e.originalName,
+                        tensor: e.variable
+                    })))
+                }
+                async setWeights(e) {
+                    e = await this.extractIterations(e);
+                    this.accumulations = e.map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    }))
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate,
+                        momentum: this.momentum,
+                        useNesterov: this.useNesterov
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate, t.momentum, t.useNesterov)
+                }
+            };
+            $d.className = "Momentum", Oo($d);
+            var Bd = class extends Fd {
+                constructor(e, t = .9, n = 0, r = null, a = !1) {
+                    if (super(), this.learningRate = e, this.decay = t, this.momentum = n, this.epsilon = r, this.accumulatedMeanSquares = [], this.accumulatedMoments = [], this.accumulatedMeanGrads = [], this.centered = a, null == r && (this.epsilon = xs.backend.epsilon()), null == e) throw new Error("learningRate for RMSPropOptimizer must be defined.")
+                }
+                applyGradients(e) {
+                    (Array.isArray(e) ? e.map(e => e.name) : Object.keys(e)).forEach((t, n) => {
+                        const r = xs.registeredVariables[t];
+                        null == this.accumulatedMeanSquares[n] && (this.accumulatedMeanSquares[n] = {
+                            originalName: t + "/rms",
+                            variable: ru(() => Fl(r).variable(!1))
+                        }), null == this.accumulatedMoments[n] && (this.accumulatedMoments[n] = {
+                            originalName: t + "/momentum",
+                            variable: ru(() => Fl(r).variable(!1))
+                        }), null == this.accumulatedMeanGrads[n] && this.centered && (this.accumulatedMeanGrads[n] = {
+                            originalName: t + "/mg",
+                            variable: ru(() => Fl(r).variable(!1))
+                        });
+                        const a = Array.isArray(e) ? e[n].tensor : e[t];
+                        if (null == a) return;
+                        const s = this.accumulatedMeanSquares[n].variable,
+                            i = this.accumulatedMoments[n].variable;
+                        ru(() => {
+                            const e = gu(vu(s, this.decay), vu(jc(a), 1 - this.decay));
+                            if (this.centered) {
+                                const t = this.accumulatedMeanGrads[n].variable,
+                                    o = gu(vu(t, this.decay), vu(a, 1 - this.decay)),
+                                    u = bu(vu(a, this.learningRate), Wh(gc(e, gu(jc(o), this.epsilon)))),
+                                    l = gu(vu(i, this.momentum), u);
+                                s.assign(e), t.assign(o), i.assign(l);
+                                const c = gc(r, l);
+                                r.assign(c)
+                            } else {
+                                const e = gu(vu(s, this.decay), vu(jc(a), 1 - this.decay)),
+                                    t = gu(vu(i, this.momentum), bu(vu(a, this.learningRate), Wh(gu(e, this.epsilon))));
+                                s.assign(e), i.assign(t);
+                                const n = gc(r, t);
+                                r.assign(n)
+                            }
+                        })
+                    }), this.incrementIterations()
+                }
+                dispose() {
+                    null != this.accumulatedMeanSquares && au(this.accumulatedMeanSquares.map(e => e.variable)), null != this.accumulatedMeanGrads && this.centered && au(this.accumulatedMeanGrads.map(e => e.variable)), null != this.accumulatedMoments && au(this.accumulatedMoments.map(e => e.variable))
+                }
+                async getWeights() {
+                    const e = [...this.accumulatedMeanSquares, ...this.accumulatedMoments];
+                    return this.centered && e.push(...this.accumulatedMeanGrads), [await this.saveIterations()].concat(e.map(e => ({
+                        name: e.originalName,
+                        tensor: e.variable
+                    })))
+                }
+                async setWeights(e) {
+                    e = await this.extractIterations(e);
+                    const t = this.centered ? e.length / 3 : e.length / 2;
+                    this.accumulatedMeanSquares = e.slice(0, t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    })), this.accumulatedMoments = e.slice(t, 2 * t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    })), this.centered && (this.accumulatedMeanGrads = e.slice(2 * t, 3 * t).map(e => ({
+                        originalName: e.name,
+                        variable: e.tensor.variable(!1)
+                    })))
+                }
+                getConfig() {
+                    return {
+                        learningRate: this.learningRate,
+                        decay: this.decay,
+                        momentum: this.momentum,
+                        epsilon: this.epsilon,
+                        centered: this.centered
+                    }
+                }
+                static fromConfig(e, t) {
+                    return new e(t.learningRate, t.decay, t.momentum, t.epsilon, t.centered)
+                }
+            };
+            Bd.className = "RMSProp", Oo(Bd);
+            var zd = class {
+                    static sgd(e) {
+                        return new Pd(e)
+                    }
+                    static momentum(e, t, n = !1) {
+                        return new $d(e, t, n)
+                    }
+                    static rmsprop(e, t = .9, n = 0, r = null, a = !1) {
+                        return new Bd(e, t, n, r, a)
+                    }
+                    static adam(e = .001, t = .9, n = .999, r = null) {
+                        return new Od(e, t, n, r)
+                    }
+                    static adadelta(e = .001, t = .95, n = null) {
+                        return new Md(e, t, n)
+                    }
+                    static adamax(e = .002, t = .9, n = .999, r = null, a = 0) {
+                        return new Ld(e, t, n, r, a)
+                    }
+                    static adagrad(e, t = .1) {
+                        return new Dd(e, t)
+                    }
+                },
+                Ud = {
+                    sgd: zd.sgd,
+                    momentum: zd.momentum,
+                    adadelta: zd.adadelta,
+                    adagrad: zd.adagrad,
+                    rmsprop: zd.rmsprop,
+                    adamax: zd.adamax,
+                    adam: zd.adam
+                },
+                Wd = "undefined" != typeof requestAnimationFrame ? requestAnimationFrame : void 0 !== o ? o : e => e();
+
+            function Vd() {
+                return new Promise(e => Wd(() => e()))
+            }
+            var jd = {};
+
+            function Gd(e, t) {
+                const n = e[0].length;
+                e.forEach((e, t) => {
+                    ue(e.length === n, () => `Error in concat${n}D: rank of tensors[${t}] must be the same as the rank of the rest (${n})`)
+                }), ue(t >= 0 && t < n, () => `Error in concat${n}D: axis must be between 0 and ${n-1}.`);
+                const r = e[0];
+                e.forEach((e, a) => {
+                    for (let s = 0; s < n; s++) ue(s === t || e[s] === r[s], () => `Error in concat${n}D: Shape of tensors[${a}] (${e}) does not match the shape of the rest (${r}) along the non-concatenated axis ${a}.`)
+                })
+            }
+
+            function Hd(e, t) {
+                const n = e[0].slice();
+                for (let r = 1; r < e.length; r++) n[t] += e[r][t];
+                return n
+            }
+            w(jd, {
+                ERF_A1: () => af,
+                ERF_A2: () => sf,
+                ERF_A3: () => of ,
+                ERF_A4: () => uf,
+                ERF_A5: () => lf,
+                ERF_P: () => rf,
+                PARALLELIZE_THRESHOLD: () => qd,
+                SELU_SCALE: () => nf,
+                SELU_SCALEALPHA: () => tf,
+                applyActivation: () => kp,
+                assertAndGetBroadcastShape: () => Zi,
+                assertAxesAreInnerMostDims: () => Sc,
+                assertParamsConsistent: () => Gd,
+                assignToTypedArray: () => mf,
+                axesAreInnerMostDims: () => vc,
+                calculateShapes: () => po,
+                checkEinsumDimSizes: () => wf,
+                checkPadOnDimRoundingMode: () => Hu,
+                combineLocations: () => xc,
+                complexWithEvenIndex: () => pf,
+                complexWithOddIndex: () => df,
+                computeConv2DInfo: () => Lu,
+                computeConv3DInfo: () => Pu,
+                computeDefaultPad: () => $u,
+                computeDilation2DInfo: () => Mu,
+                computeOptimalWindowSize: () => Kd,
+                computeOutAndReduceShapes: () => wc,
+                computeOutShape: () => Hd,
+                computePool2DInfo: () => Du,
+                computePool3DInfo: () => Ou,
+                convertConv2DDataFormat: () => Gu,
+                decodeEinsumEquation: () => vf,
+                eitherStridesOrDilationsAreOne: () => ju,
+                expandShapeToKeepDim: () => kc,
+                exponent: () => yf,
+                exponents: () => gf,
+                fromStringArrayToUint8: () => Vf,
+                fromUint8ToStringArray: () => Wf,
+                getAxesPermutation: () => Ic,
+                getBroadcastDims: () => Xi,
+                getComplexWithIndex: () => ff,
+                getEinsumComputePath: () => kf,
+                getEinsumPermutation: () => xf,
+                getFusedBiasGradient: () => wp,
+                getFusedDyActivation: () => xp,
+                getImageCenter: () => Yd,
+                getInnerMostAxes: () => Ec,
+                getPermuted: () => Qd,
+                getReductionAxes: () => Qi,
+                getReshaped: () => Xd,
+                getReshapedPermuted: () => Zd,
+                getSliceBeginCoords: () => Jd,
+                getSliceSize: () => ef,
+                getSparseFillEmptyRowsIndicesDenseShapeMismatch: () => Ef,
+                getSparseFillEmptyRowsNegativeIndexErrorMessage: () => Nf,
+                getSparseFillEmptyRowsOutOfRangeIndexErrorMessage: () => Cf,
+                getSparseReshapeEmptyTensorZeroOutputDimErrorMessage: () => _f,
+                getSparseReshapeInputOutputMismatchErrorMessage: () => Mf,
+                getSparseReshapeInputOutputMultipleErrorMessage: () => Ff,
+                getSparseReshapeMultipleNegativeOneOutputDimErrorMessage: () => Tf,
+                getSparseReshapeNegativeOutputDimErrorMessage: () => Rf,
+                getSparseSegmentReductionIndicesOutOfRangeErrorMessage: () => Pf,
+                getSparseSegmentReductionNegativeSegmentIdsErrorMessage: () => Df,
+                getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage: () => Of,
+                getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage: () => Lf,
+                getUndoAxesPermutation: () => Ac,
+                isIdentityPermutation: () => Sf,
+                log: () => oa,
+                mergeRealAndImagArrays: () => cf,
+                prepareAndValidate: () => uo,
+                prepareSplitSize: () => Af,
+                segment_util: () => $f,
+                shouldFuse: () => Sp,
+                slice_util: () => fo,
+                splitRealAndImagArrays: () => hf,
+                tupleValuesAreOne: () => Vu,
+                upcastType: () => cs,
+                validateInput: () => ho,
+                validateUpdateShape: () => co,
+                warn: () => ia
+            });
+            var qd = 30;
+
+            function Kd(e) {
+                return e <= qd ? e : Pe(e, Math.floor(Math.sqrt(e)))
+            }
+
+            function Yd(e, t, n) {
+                return [n * ("number" == typeof e ? e : e[0]), t * ("number" == typeof e ? e : e[1])]
+            }
+
+            function Xd(e, t, n, r = !0) {
+                let a = [];
+                if (r) a = a.concat(t.slice(0)), a.push(e[0] / n), a = a.concat(e.slice(1));
+                else {
+                    a = a.concat(e[0]);
+                    const n = t.length;
+                    for (let r = 0; r < n; ++r) a = a.concat([e[r + 1] / t[r], t[r]]);
+                    a = a.concat(e.slice(n + 1))
+                }
+                return a
+            }
+
+            function Qd(e, t, n = !0) {
+                const r = [];
+                if (n) {
+                    r.push(t);
+                    for (let n = t + 1; n < e; ++n) n <= 2 * t ? (r.push(n), r.push(n - (t + 1))) : r.push(n)
+                } else {
+                    const n = [],
+                        a = [];
+                    for (let r = 1; r < e; ++r) r >= 2 * t + 1 || r % 2 == 1 ? a.push(r) : n.push(r);
+                    r.push(...n), r.push(0), r.push(...a)
+                }
+                return r
+            }
+
+            function Zd(e, t, n, r = !0) {
+                const a = [];
+                r ? a.push(e[0] / n) : a.push(e[0] * n);
+                for (let n = 1; n < e.length; ++n) n <= t.length ? r ? a.push(t[n - 1] * e[n]) : a.push(e[n] / t[n - 1]) : a.push(e[n]);
+                return a
+            }
+
+            function Jd(e, t) {
+                const n = [0];
+                for (let r = 0; r < t; ++r) n.push(e[r][0]);
+                return n
+            }
+
+            function ef(e, t, n) {
+                const r = e.slice(0, 1);
+                for (let a = 0; a < n; ++a) r.push(e[a + 1] - t[a][0] - t[a][1]);
+                return r
+            }
+            var tf = 1.7580993408473768,
+                nf = 1.0507009873554805,
+                rf = .3275911,
+                af = .254829592,
+                sf = -.284496736,
+                of = 1.421413741,
+                uf = -1.453152027,
+                lf = 1.061405429;
+
+            function cf(e, t) {
+                if (e.length !== t.length) throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);
+                const n = new Float32Array(2 * e.length);
+                for (let r = 0; r < n.length; r += 2) n[r] = e[r / 2], n[r + 1] = t[r / 2];
+                return n
+            }
+
+            function hf(e) {
+                const t = new Float32Array(e.length / 2),
+                    n = new Float32Array(e.length / 2);
+                for (let r = 0; r < e.length; r += 2) t[r / 2] = e[r], n[r / 2] = e[r + 1];
+                return {
+                    real: t,
+                    imag: n
+                }
+            }
+
+            function pf(e) {
+                const t = Math.ceil(e.length / 4),
+                    n = new Float32Array(t),
+                    r = new Float32Array(t);
+                for (let t = 0; t < e.length; t += 4) n[Math.floor(t / 4)] = e[t], r[Math.floor(t / 4)] = e[t + 1];
+                return {
+                    real: n,
+                    imag: r
+                }
+            }
+
+            function df(e) {
+                const t = Math.floor(e.length / 4),
+                    n = new Float32Array(t),
+                    r = new Float32Array(t);
+                for (let t = 2; t < e.length; t += 4) n[Math.floor(t / 4)] = e[t], r[Math.floor(t / 4)] = e[t + 1];
+                return {
+                    real: n,
+                    imag: r
+                }
+            }
+
+            function ff(e, t) {
+                return {
+                    real: e[2 * t],
+                    imag: e[2 * t + 1]
+                }
+            }
+
+            function mf(e, t, n, r) {
+                e[2 * r] = t, e[2 * r + 1] = n
+            }
+
+            function gf(e, t) {
+                const n = new Float32Array(e / 2),
+                    r = new Float32Array(e / 2);
+                for (let a = 0; a < Math.ceil(e / 2); a++) {
+                    const s = (t ? 2 : -2) * Math.PI * (a / e);
+                    n[a] = Math.cos(s), r[a] = Math.sin(s)
+                }
+                return {
+                    real: n,
+                    imag: r
+                }
+            }
+
+            function yf(e, t, n) {
+                const r = (n ? 2 : -2) * Math.PI * (e / t);
+                return {
+                    real: Math.cos(r),
+                    imag: Math.sin(r)
+                }
+            }
+            var bf = /->/g;
+
+            function vf(e, t) {
+                const n = ((e = e.replace(/\s/g, "")).length - e.replace(bf, "").length) / "->".length;
+                if (n < 1) throw new Error("Equations without an arrow are not supported.");
+                if (n > 1) throw new Error('Equation must contain exactly one arrow ("->").');
+                const [r, a] = e.split("->");
+                ue(-1 === r.indexOf("..."), () => 'The ellipsis notation ("...") is not supported yet.');
+                const s = r.split(","),
+                    i = s.length;
+                if (t !== i) throw new Error(`Expected ${i} input tensors, received ${t}`);
+                if (i > 2) throw new Error("Support for more than 2 input tensors is not implemented yet.");
+                const o = [];
+                for (let e = 0; e < a.length; ++e) {
+                    const t = a[e];
+                    if (!s.some(e => -1 !== e.indexOf(t))) throw new Error(`Output subscripts contain the label ${t} not present in the input subscripts.`); - 1 === o.indexOf(t) && o.push(t)
+                }
+                for (let e = 0; e < r.length; ++e) {
+                    const t = r[e]; - 1 === o.indexOf(t) && "," !== t && o.push(t)
+                }
+                const u = new Array(s.length);
+                for (let e = 0; e < i; ++e) {
+                    if (new Set(s[e].split("")).size !== s[e].length) throw new Error(`Found duplicate axes in input component ${s[e]}. Support for duplicate axes in input is not implemented yet.`);
+                    u[e] = [];
+                    for (let t = 0; t < s[e].length; ++t) u[e].push(o.indexOf(s[e][t]))
+                }
+                const l = o.length,
+                    c = [];
+                for (let e = a.length; e < l; ++e) c.push(e);
+                return {
+                    allDims: o,
+                    summedDims: c,
+                    idDims: u
+                }
+            }
+
+            function xf(e, t) {
+                let n = new Array(e);
+                n.fill(-1);
+                for (let e = 0; e < t.length; ++e) n[t[e]] = e;
+                const r = [];
+                for (let t = 0; t < e; ++t) - 1 === n[t] && r.push(t);
+                return n = n.filter(e => -1 !== e), {
+                    permutationIndices: n,
+                    expandDims: r
+                }
+            }
+
+            function wf(e, t, n) {
+                const r = new Array(e);
+                for (let e = 0; e < n.length; ++e) {
+                    const a = n[e].shape;
+                    for (let n = 0; n < t[e].length; ++n) void 0 === r[t[e][n]] ? r[t[e][n]] = a[n] : ue(r[t[e][n]] === a[n], () => `Expected dimension ${r[t[e][n]]} at axis ${n} of input shaped ${JSON.stringify(a)}, but got dimension ${a[n]}`)
+                }
+            }
+
+            function kf(e, t) {
+                const n = e,
+                    r = [];
+                let a = 0;
+                0 === e.length && n.push(-1), a = e.length + 1;
+                for (let e = 0; e < a; ++e) r.push([]);
+                const s = [];
+                for (let e = 0; e < n.length; ++e) {
+                    const a = If(t, n[e]);
+                    for (const t of a) - 1 === s.indexOf(t) && (r[e].push(t), s.push(t))
+                }
+                return {
+                    path: n,
+                    steps: r
+                }
+            }
+
+            function Sf(e) {
+                return e.every((e, t) => e === t)
+            }
+
+            function If(e, t) {
+                const n = [];
+                for (let r = 0; r < e.length; ++r) 0 !== e[r].length && -1 === e[r].indexOf(t) && -1 !== t || n.push(r);
+                return n
+            }
+
+            function Af(e, t, n = 0) {
+                let r = [];
+                if ("number" == typeof t) ue(e.shape[n] % t == 0, () => "Number of splits must evenly divide the axis."), r = new Array(t).fill(e.shape[n] / t);
+                else {
+                    ue(t.reduce((e, t) => (-1 === t && (e += 1), e), 0) <= 1, () => "There should be only one negative value in split array.");
+                    const a = t.indexOf(-1);
+                    if (-1 !== a) {
+                        const r = t.reduce((e, t) => t > 0 ? e + t : e);
+                        t[a] = e.shape[n] - r
+                    }
+                    ue(e.shape[n] === t.reduce((e, t) => e + t), () => "The sum of sizes must match the size of the axis dimension."), r = t
+                }
+                return r
+            }
+
+            function Ef(e) {
+                return "Received SparseTensor with denseShape[0] = 0 but\n  indices.shape[0] = " + e
+            }
+
+            function Nf(e, t) {
+                return `indices(${e}, 0) is invalid: ${t} < 0`
+            }
+
+            function Cf(e, t, n) {
+                return `indices(${e}, 0) is invalid: ${t} >= ${n}`
+            }
+
+            function Tf(e, t) {
+                return `only one output dimension may be -1, not both ${e} and ${t}`
+            }
+
+            function Rf(e, t) {
+                return `size ${e} must be non-negative, not ${t}`
+            }
+
+            function _f() {
+                return "reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"
+            }
+
+            function Ff(e, t) {
+                return `Input to reshape is a SparseTensor with ${pe(e)}\n  dense values, but the requested shape requires a multiple of ${pe(t)}. inputShape=${e} outputShape= ${t}`
+            }
+
+            function Mf(e, t) {
+                return `Input to reshape is a tensor with ${pe(e)} dense values, but the requested shape has ${pe(t)}. inputShape=${e} outputShape=${t}`
+            }
+
+            function Df() {
+                return "segment ids must be >= 0"
+            }
+
+            function Of() {
+                return "segment ids are not increasing"
+            }
+
+            function Lf(e, t) {
+                return `Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`
+            }
+
+            function Pf(e, t, n) {
+                return `Bad: indices[${e}] == ${t} out of range [0, ${n})`
+            }
+            var $f = {};
+
+            function Bf(e, t) {
+                let n, r = !1;
+                for (e <= qd ? (n = e, r = !0) : n = Pe(e, Math.floor(Math.sqrt(e))); !r;) n > t || n === e ? r = !0 : n = Pe(e, n + 1);
+                return n
+            }
+
+            function zf(e, t, n) {
+                const r = [],
+                    a = e.length;
+                for (let s = 0; s < a; s++) s !== t ? r.push(e[s]) : r.push(n);
+                return r
+            }
+
+            function Uf(e, t, n, r) {
+                const a = t.shape.length,
+                    s = e.shape.length;
+                if (0 !== r && (r < -a || r > a)) throw new Error(`Expect batchDims in the range of [-${a}, ${a}], but got ${r}`);
+                if (r < 0 && (r += a), r > s) throw new Error(`batchDims (${r}) must be less than rank(x) (\n    ${s}).`);
+                if (n < r) throw new Error(`batchDims (${r}) must be less than or equal to axis (${n}).`);
+                for (let n = 0; n < r; ++n)
+                    if (e.shape[n] !== t.shape[n]) throw new Error(`x.shape[${n}]: ${e.shape[n]} should be equal to indices.shape[${n}]: ${t.shape[n]}.`);
+                const i = e.shape[n],
+                    o = [];
+                let u = 1,
+                    l = 1,
+                    c = 1;
+                for (let t = 0; t < r; ++t) o.push(e.shape[t]), u *= e.shape[t];
+                for (let t = r; t < n; t++) o.push(e.shape[t]), l *= e.shape[t];
+                for (let e = r; e < a; e++) o.push(t.shape[e]);
+                for (let t = n + 1; t < s; t++) o.push(e.shape[t]), c *= e.shape[t];
+                return {
+                    batchSize: u,
+                    sliceSize: c,
+                    outerSize: l,
+                    dimSize: i,
+                    outputShape: o
+                }
+            }
+
+            function Wf(e) {
+                try {
+                    return e.map(e => Ba(e))
+                } catch (e) {
+                    throw new Error("Failed to decode encoded string bytes into utf-8, error: " + e)
+                }
+            }
+
+            function Vf(e) {
+                return e.map(e => $a(e))
+            }
+            w($f, {
+                collectGatherOpShapeInfo: () => Uf,
+                computeOutShape: () => zf,
+                segOpComputeOptimalWindowSize: () => Bf
+            });
+            var jf = {};
+            w(jf, {
+                nonMaxSuppressionV3Impl: () => Up,
+                nonMaxSuppressionV4Impl: () => Wp,
+                nonMaxSuppressionV5Impl: () => Vp,
+                whereImpl: () => ip
+            });
+            var Gf = {
+                    kernelName: et,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => vu(e, Hh(Ai(n, "float32"), -1))
+                        }
+                    }
+                },
+                Hf = {
+                    kernelName: tt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => {
+                                const t = jc(Ai(n, "float32")),
+                                    r = Wh(gc(Ah(1), t));
+                                return pc(bu(e, r))
+                            }
+                        }
+                    }
+                },
+                qf = {
+                    kernelName: nt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => {
+                                const t = Wh(gc(jc(Ai(n, "float32")), 1));
+                                return bu(e, t)
+                            }
+                        }
+                    }
+                },
+                Kf = {
+                    kernelName: rt,
+                    inputsToSave: ["a", "b"],
+                    gradFunc: (e, t) => {
+                        const [n, r] = t, a = Zi(n.shape, r.shape);
+                        return {
+                            a: () => {
+                                let t = e;
+                                const r = Qi(n.shape, a);
+                                return r.length > 0 && (t = yc(t, r)), qu(t, n.shape)
+                            },
+                            b: () => {
+                                let t = e;
+                                const n = Qi(r.shape, a);
+                                return n.length > 0 && (t = yc(t, n)), qu(t, r.shape)
+                            }
+                        }
+                    }
+                },
+                Yf = {
+                    kernelName: at,
+                    saveAllInputs: !0,
+                    gradFunc: (e, t) => {
+                        const n = {};
+                        return t.forEach((t, r) => {
+                            n[r] = () => e.clone()
+                        }), n
+                    }
+                },
+                Xf = {
+                    kernelName: ot,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => Fl(n)
+                        }
+                    }
+                },
+                Qf = {
+                    kernelName: ut,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => Fl(n)
+                        }
+                    }
+                },
+                Zf = {
+                    kernelName: lt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => bu(e, Wh(gc(Ah(1), jc(Ai(n, "float32")))))
+                        }
+                    }
+                },
+                Jf = {
+                    kernelName: ct,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => {
+                                const t = Wh(gu(Ah(1), jc(Ai(n, "float32"))));
+                                return bu(e, t)
+                            }
+                        }
+                    }
+                },
+                em = {
+                    kernelName: dt,
+                    inputsToSave: ["a", "b"],
+                    gradFunc: (e, t) => {
+                        const [n, r] = t, a = Zi(n.shape, r.shape);
+                        return {
+                            a: () => {
+                                const t = gu(jc(n), jc(r));
+                                let s = vu(e, bu(r, t));
+                                const i = Qi(n.shape, a);
+                                return i.length > 0 && (s = yc(s, i)), qu(s, n.shape)
+                            },
+                            b: () => {
+                                const t = gu(jc(n), jc(r));
+                                let s = pc(vu(e, bu(n, t)));
+                                const i = Qi(r.shape, a);
+                                return i.length > 0 && (s = yc(s, i)), qu(s, r.shape)
+                            }
+                        }
+                    }
+                },
+                tm = {
+                    kernelName: ht,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => bu(e, gu(jc(Ai(n, "float32")), 1))
+                        }
+                    }
+                },
+                nm = {
+                    kernelName: pt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => bu(e, gc(Ah(1), jc(Ai(n, "float32"))))
+                        }
+                    }
+                };
+            var rm = Ms({
+                    avgPool3dGrad_: function (e, t, n, r, a, s) {
+                        const i = Rs(e, "dy", "avgPool3dGrad"),
+                            o = Rs(t, "input", "avgPool3dGrad");
+                        let u = i,
+                            l = o,
+                            c = !1;
+                        4 === o.rank && (c = !0, u = qu(i, [1, i.shape[0], i.shape[1], i.shape[2], i.shape[3]]), l = qu(o, [1, o.shape[0], o.shape[1], o.shape[2], o.shape[3]])), ue(5 === u.rank, () => `Error in avgPool3dGrad: dy must be rank 5 but got rank ${u.rank}.`), ue(5 === l.rank, () => `Error in avgPool3dGrad: input must be rank 5 but got rank ${l.rank}.`), Hu("avgPool3dGrad", a, s);
+                        const h = {
+                                dy: u,
+                                input: l
+                            },
+                            p = {
+                                filterSize: n,
+                                strides: r,
+                                pad: a,
+                                dimRoundingMode: s
+                            },
+                            d = xs.runKernel(yt, h, p);
+                        return c ? qu(d, [d.shape[1], d.shape[2], d.shape[3], d.shape[4]]) : d
+                    }
+                }),
+                am = {
+                    kernelName: gt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t, n) => {
+                        const [r] = t, {
+                            filterSize: a,
+                            strides: s,
+                            pad: i,
+                            dimRoundingMode: o
+                        } = n;
+                        return {
+                            x: () => rm(e, r, a, s, i, o)
+                        }
+                    }
+                };
+            var sm = Ms({
+                    avgPoolGrad_: function (e, t, n, r, a) {
+                        const s = Rs(e, "dy", "avgPoolGrad"),
+                            i = Rs(t, "input", "avgPoolGrad");
+                        ue(i.rank === s.rank, () => `Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);
+                        let o = i,
+                            u = s,
+                            l = !1;
+                        3 === i.rank && (l = !0, o = qu(i, [1, i.shape[0], i.shape[1], i.shape[2]]), u = qu(s, [1, s.shape[0], s.shape[1], s.shape[2]])), ue(4 === u.rank, () => `Error in avgPoolGrad: dy must be rank 4 but got rank ${u.rank}.`), ue(4 === o.rank, () => `Error in avgPoolGrad: input must be rank 4 but got rank ${o.rank}.`);
+                        const c = {
+                                dy: u,
+                                input: o
+                            },
+                            h = {
+                                filterSize: n,
+                                strides: r,
+                                pad: a
+                            },
+                            p = xs.runKernel(mt, c, h);
+                        return l ? qu(p, [p.shape[1], p.shape[2], p.shape[3]]) : p
+                    }
+                }),
+                im = {
+                    kernelName: ft,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t, n) => {
+                        const [r] = t, {
+                            filterSize: a,
+                            strides: s,
+                            pad: i
+                        } = n;
+                        return {
+                            x: () => sm(e, r, a, s, i)
+                        }
+                    }
+                },
+                om = {
+                    kernelName: bt,
+                    inputsToSave: ["a", "b"],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, {
+                            transposeA: s,
+                            transposeB: i
+                        } = n;
+                        return s || i ? !s && i ? {
+                            a: () => Gi(e, a, !1, !1),
+                            b: () => Gi(e, r, !0, !1)
+                        } : s && !i ? {
+                            a: () => Gi(a, e, !1, !0),
+                            b: () => Gi(r, e, !1, !1)
+                        } : {
+                            a: () => Gi(a, e, !0, !0),
+                            b: () => Gi(e, r, !0, !0)
+                        } : {
+                            a: () => Gi(e, a, !1, !0),
+                            b: () => Gi(r, e, !0, !1)
+                        }
+                    }
+                },
+                um = {
+                    kernelName: vt,
+                    gradFunc: (e, t, n) => {
+                        const {
+                            blockShape: r,
+                            crops: a
+                        } = n;
+                        return {
+                            x: () => nh(e, r, a)
+                        }
+                    }
+                },
+                lm = {
+                    kernelName: wt,
+                    gradFunc: (e, t, n) => {
+                        const r = n,
+                            a = r.inputShape,
+                            s = r.shape,
+                            i = Array.from(s);
+                        for (let e = a.length - 1; e >= 0; e--)
+                            if (a[e] === s[e]) i[e] = 1;
+                            else if (1 !== a[e]) throw new Error(`broadcastTo(): [${a}] cannot be broadcast to [${s}].`);
+                        const o = [];
+                        for (let e = 0; e < i.length; e++) i[e] > 1 && o.push(e);
+                        return {
+                            x: () => yc(e, o, !0)
+                        }
+                    }
+                },
+                cm = {
+                    kernelName: St,
+                    gradFunc: e => ({
+                        x: () => e.clone()
+                    })
+                },
+                hm = {
+                    kernelName: It,
+                    gradFunc: e => ({
+                        x: () => Fl(e)
+                    })
+                },
+                pm = {
+                    kernelName: At,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t, n) => {
+                        const [r] = t, {
+                            clipValueMin: a,
+                            clipValueMax: s
+                        } = n;
+                        return {
+                            x: () => _l(Cc(ql(r, a), ec(r, s)), e, Fl(e))
+                        }
+                    }
+                },
+                dm = {
+                    kernelName: Nt,
+                    inputsToSave: ["x"],
+                    gradFunc: Gf.gradFunc
+                },
+                fm = {
+                    kernelName: Ct,
+                    saveAllInputs: !0,
+                    gradFunc: (e, t, n) => {
+                        const r = t.map(e => e.shape),
+                            {
+                                axis: a
+                            } = n,
+                            s = ke(a, t[0].shape)[0],
+                            i = r.map(e => e[s]);
+                        return zh(e, i, s).map(e => () => e)
+                    }
+                },
+                mm = {
+                    kernelName: Tt,
+                    inputsToSave: ["x", "filter"],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, {
+                            dilations: s,
+                            strides: i,
+                            pad: o,
+                            dataFormat: u
+                        } = n;
+                        return ue(Vu(s), () => `Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`), {
+                            x: () => yl(r.shape, e, a, i, o, u),
+                            filter: () => vp(r, e, a.shape, i, o, u)
+                        }
+                    }
+                },
+                gm = {
+                    kernelName: _t,
+                    inputsToSave: ["dy", "filter"],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, {
+                            strides: s,
+                            pad: i,
+                            dataFormat: o,
+                            dimRoundingMode: u
+                        } = n;
+                        return {
+                            dy: () => ml(e, a, s, i, o, 1, u),
+                            filter: () => vp(e, r, a.shape, s, i, o, u)
+                        }
+                    }
+                };
+            var ym = Ms({
+                    conv3DBackpropFilter_: function (e, t, n, r, a) {
+                        let s = e;
+                        4 === e.rank && (s = qu(e, [1, e.shape[0], e.shape[1], e.shape[2], e.shape[3]]));
+                        let i = t;
+                        4 === i.rank && (i = qu(t, [1, t.shape[0], t.shape[1], t.shape[2], t.shape[3]])), ue(5 === s.rank, () => `Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`), ue(5 === i.rank, () => `Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`), ue(5 === n.length, () => `Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`), ue(s.shape[4] === n[3], () => `Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${n[3]}.`), ue(i.shape[4] === n[4], () => `Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${n[4]}).`);
+                        const o = {
+                                x: s,
+                                dy: i
+                            },
+                            u = {
+                                strides: r,
+                                pad: a,
+                                filterShape: n
+                            };
+                        return xs.runKernel(Mt, o, u)
+                    }
+                }),
+                bm = {
+                    kernelName: Ft,
+                    inputsToSave: ["x", "filter"],
+                    gradFunc: (e, t, n) => {
+                        const {
+                            dilations: r,
+                            strides: a,
+                            pad: s
+                        } = n;
+                        ue(Vu(r), () => `Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${r}'`);
+                        const [i, o] = t;
+                        return {
+                            x: () => xl(i.shape, e, o, a, s),
+                            filter: () => ym(i, e, o.shape, a, s)
+                        }
+                    }
+                },
+                vm = {
+                    kernelName: Ot,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => vu(pc(Rh(Ai(n, "float32"))), e)
+                        }
+                    }
+                },
+                xm = {
+                    kernelName: Lt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => vu(_h(Ai(n, "float32")), e)
+                        }
+                    }
+                },
+                wm = {
+                    kernelName: Pt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t, n) => {
+                        const [r] = t, {
+                            axis: a,
+                            exclusive: s,
+                            reverse: i
+                        } = n;
+                        return {
+                            x: () => {
+                                const t = Ic([a], r.rank);
+                                let n = Il(e, a, s, !i);
+                                return null != t && (n = qi(n, t)), n
+                            }
+                        }
+                    }
+                },
+                km = {
+                    kernelName: Ut,
+                    inputsToSave: ["x", "filter"],
+                    gradFunc: (e, t, n) => {
+                        const {
+                            dilations: r,
+                            strides: a,
+                            pad: s,
+                            dimRoundingMode: i
+                        } = n, o = null == r ? [1, 1] : r;
+                        ue(Vu(o), () => `Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);
+                        const [u, l] = t;
+                        return ue(4 === u.rank, () => `Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${u.rank}.`), ue(4 === l.rank, () => `Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${l.rank}.`), ue(u.shape[3] === l.shape[2], () => `Error in gradient of depthwiseConv2d: number of input channels (${u.shape[3]}) must match the inChannels dimension in filter ${l.shape[2]}.`), ue(ju(a, o), () => `Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${a} and dilations '${o}'.`), Hu("depthwiseConv2d", s, i), {
+                            x: () => Ep(u.shape, e, l, a, s, o, i),
+                            filter: () => Ap(u, e, l.shape, a, s, o, i)
+                        }
+                    }
+                },
+                Sm = {
+                    kernelName: Gt,
+                    inputsToSave: ["x", "filter"],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, s = {
+                            x: r,
+                            filter: a,
+                            dy: e
+                        }, i = {
+                            x: r,
+                            filter: a,
+                            dy: e
+                        };
+                        return {
+                            x: () => xs.runKernel(Ht, s, n),
+                            filter: () => xs.runKernel(qt, i, n)
+                        }
+                    }
+                },
+                Im = {
+                    kernelName: Xt,
+                    outputsToSave: [!0],
+                    gradFunc: (e, t) => {
+                        const [n] = t, r = {
+                            dy: e,
+                            y: n
+                        };
+                        return {
+                            x: () => xs.runKernel(Qt, r)
+                        }
+                    }
+                },
+                Am = {
+                    kernelName: Zt,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t, r = vu($l(pc(jc(n))), 2 / Math.sqrt(Math.PI));
+                        return {
+                            x: () => vu(e, r)
+                        }
+                    }
+                },
+                Em = {
+                    kernelName: en,
+                    outputsToSave: [!0],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => vu(e, n)
+                        }
+                    }
+                },
+                Nm = {
+                    kernelName: tn,
+                    inputsToSave: ["input"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            input: () => qu(e, n.shape)
+                        }
+                    }
+                },
+                Cm = {
+                    kernelName: nn,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => vu(e, $l(n))
+                        }
+                    }
+                },
+                Tm = {
+                    kernelName: on,
+                    gradFunc: e => ({
+                        x: () => Fl(e)
+                    })
+                },
+                Rm = {
+                    kernelName: un,
+                    inputsToSave: ["a", "b"],
+                    gradFunc: (e, t) => {
+                        const [n, r] = t, a = Zi(n.shape, r.shape);
+                        return {
+                            a: () => {
+                                const t = bu(e, Ai(r, "float32")),
+                                    s = Qi(n.shape, a);
+                                return s.length > 0 ? qu(yc(t, s), n.shape) : t
+                            },
+                            b: () => {
+                                let t = vu(e, Ai(n, "float32"));
+                                const s = Qi(r.shape, a);
+                                s.length > 0 && (t = qu(yc(t, s), r.shape));
+                                const i = jc(r);
+                                return pc(bu(t, Ai(i, "float32")))
+                            }
+                        }
+                    }
+                },
+                _m = {
+                    kernelName: ln,
+                    inputsToSave: ["x", "mean", "variance", "scale"],
+                    gradFunc: (e, t, n) => {
+                        const {
+                            varianceEpsilon: r
+                        } = n, [a, s, i, o] = t, u = null == o ? Ah(1) : o, l = Qi(s.shape, a.shape), c = [];
+                        if (1 === s.rank) {
+                            for (let e = 0; e < a.shape.length - 1; ++e) c.push(a.shape[e]);
+                            c.push(1)
+                        }
+                        const h = gc(a, s),
+                            p = vu(e, u),
+                            d = Ih(gu(i, Ah(r))),
+                            f = vu(vu(vu(d, d), d), Ah(-.5));
+                        return {
+                            x: () => 1 === s.rank ? qu(vu(vu(e, Ul(qu(d, [1, 1, 1, s.shape[0]]), c)), u), a.shape) : qu(vu(vu(e, d), u), a.shape),
+                            mean: () => {
+                                let e = vu(vu(d, Ah(-1)), p);
+                                return 1 === s.rank && (e = yc(e, l)), qu(e, s.shape)
+                            },
+                            variance: () => {
+                                let e = vu(vu(f, h), p);
+                                return 1 === s.rank && (e = yc(e, l)), qu(e, s.shape)
+                            },
+                            scale: () => {
+                                const t = vu(h, d);
+                                let n = vu(e, t);
+                                return 1 === s.rank && (n = yc(n, l)), qu(n, s.shape)
+                            },
+                            offset: () => {
+                                let t = e;
+                                return 1 === s.rank && (t = yc(t, l)), qu(t, s.shape)
+                            }
+                        }
+                    }
+                },
+                Fm = {
+                    kernelName: cn,
+                    inputsToSave: ["x", "indices"],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, {
+                            axis: s
+                        } = n, i = ke(s, r.shape)[0];
+                        return {
+                            x: () => {
+                                const t = r.shape,
+                                    n = a.size,
+                                    o = t.slice(0, i),
+                                    u = o.length,
+                                    l = t.slice(s, t.length).slice(1),
+                                    c = l.length,
+                                    h = Mm(0, u),
+                                    p = Mm(u + 1, u + 1 + c),
+                                    d = Dm([o, [n], l]),
+                                    f = qu(e, d),
+                                    m = qu(a, [n]),
+                                    g = Dm([
+                                        [u], h, p
+                                    ]),
+                                    y = qi(f, g);
+                                let b = rp(y, m, r.shape[i]);
+                                const v = Ac(g);
+                                return b = qi(b, v), b
+                            },
+                            indices: () => a
+                        }
+                    }
+                };
+
+            function Mm(e, t) {
+                const n = [];
+                for (let r = e; r < t; ++r) n.push(r);
+                return n
+            }
+
+            function Dm(e) {
+                const t = [];
+                for (let n = 0; n < e.length; ++n)
+                    for (let r = 0; r < e[n].length; ++r) t.push(e[n][r]);
+                return t
+            }
+            var Om = {
+                    kernelName: dn,
+                    inputsToSave: ["a", "b"],
+                    gradFunc: (e, t) => {
+                        const [n, r] = t;
+                        return {
+                            a: () => Fl(n),
+                            b: () => Fl(r)
+                        }
+                    }
+                },
+                Lm = {
+                    kernelName: fn,
+                    gradFunc: e => ({
+                        x: () => Ai(e, "float32")
+                    })
+                },
+                Pm = {
+                    kernelName: yn,
+                    gradFunc: e => ({
+                        x: () => Fl(e)
+                    })
+                },
+                $m = {
+                    kernelName: bn,
+                    gradFunc: e => ({
+                        x: () => Fl(e)
+                    })
+                },
+                Bm = {
+                    kernelName: vn,
+                    gradFunc: e => ({
+                        x: () => Fl(e)
+                    })
+                },
+                zm = {
+                    kernelName: xn,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t, n) => {
+                        const [r] = t, {
+                            alpha: a
+                        } = n, s = Hl(r, 0);
+                        return {
+                            x: () => _l(s, e, vu(e, a))
+                        }
+                    }
+                },
+                Um = {
+                    kernelName: An,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => bu(e, gu(n, 1))
+                        }
+                    }
+                },
+                Wm = {
+                    kernelName: In,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t) => {
+                        const [n] = t;
+                        return {
+                            x: () => bu(e, Ai(n, "float32"))
+                        }
+                    }
+                },
+                Vm = {
+                    kernelName: Tn,
+                    inputsToSave: [],
+                    outputsToSave: [!0],
+                    gradFunc: (e, t, n) => {
+                        const [r] = t, {
+                            axis: a
+                        } = n;
+                        return {
+                            logits: () => {
+                                const t = $l(r);
+                                return gc(e, vu(yc(e, a, !0), t))
+                            }
+                        }
+                    }
+                };
+            var jm = Ms({
+                    localResponseNormalizationBackprop_: function (e, t, n, r = 5, a = 1, s = 1, i = .5) {
+                        const o = {
+                                x: e,
+                                y: t,
+                                dy: n
+                            },
+                            u = {
+                                depthRadius: r,
+                                bias: a,
+                                alpha: s,
+                                beta: i
+                            };
+                        return xs.runKernel(_n, o, u)
+                    }
+                }),
+                Gm = {
+                    kernelName: Rn,
+                    inputsToSave: ["x"],
+                    outputsToSave: [!0],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, {
+                            depthRadius: s,
+                            bias: i,
+                            alpha: o,
+                            beta: u
+                        } = n;
+                        return {
+                            x: () => jm(r, a, e, s, i, o, u)
+                        }
+                    }
+                };
+
+            function Hm(e, t, n, r) {
+                return t.rank < n.rank && (t = qu(t, kc(t.shape, r))), e.rank < n.rank && (e = qu(e, kc(e.shape, r))), {
+                    x: () => vu(e, Ai(Rl(n, t), e.dtype))
+                }
+            }
+            var qm = {
+                    kernelName: Fn,
+                    inputsToSave: ["x"],
+                    outputsToSave: [!0],
+                    gradFunc: (e, t, n) => {
+                        const r = n,
+                            {
+                                reductionIndices: a
+                            } = r,
+                            s = t[0],
+                            i = Hm(e, t[1], s, ke(a, s.shape));
+                        return {
+                            x: () => i.x()
+                        }
+                    }
+                },
+                Km = {
+                    kernelName: Mn,
+                    inputsToSave: ["a", "b"],
+                    gradFunc: (e, t) => {
+                        const [n, r] = t;
+                        return {
+                            a: () => vu(e, Ai(ql(n, r), "float32")),
+                            b: () => vu(e, Ai(Jl(n, r), "float32"))
+                        }
+                    }
+                };
+            var Ym = Ms({
+                    maxPool3dGrad_: function (e, t, n, r, a, s, i) {
+                        const o = Rs(e, "dy", "maxPool3dGrad"),
+                            u = Rs(t, "input", "maxPool3dGrad"),
+                            l = Rs(n, "output", "maxPool3dGrad");
+                        let c = o,
+                            h = u,
+                            p = l,
+                            d = !1;
+                        4 === u.rank && (d = !0, c = qu(o, [1, o.shape[0], o.shape[1], o.shape[2], o.shape[3]]), h = qu(u, [1, u.shape[0], u.shape[1], u.shape[2], u.shape[3]]), p = qu(l, [1, l.shape[0], l.shape[1], l.shape[2], l.shape[3]])), ue(5 === c.rank, () => `Error in maxPool3dGrad: dy must be rank 5 but got rank ${c.rank}.`), ue(5 === h.rank, () => `Error in maxPool3dGrad: input must be rank 5 but got rank ${h.rank}.`), ue(5 === p.rank, () => `Error in maxPool3dGrad: output must be rank 5 but got rank ${p.rank}.`), Hu("maxPool3dGrad", s, i);
+                        const f = {
+                                dy: c,
+                                input: h,
+                                output: p
+                            },
+                            m = {
+                                filterSize: r,
+                                strides: a,
+                                pad: s,
+                                dimRoundingMode: i
+                            },
+                            g = xs.runKernel(Pn, f, m);
+                        return d ? qu(g, [g.shape[1], g.shape[2], g.shape[3], g.shape[4]]) : g
+                    }
+                }),
+                Xm = {
+                    kernelName: Ln,
+                    inputsToSave: ["x"],
+                    outputsToSave: [!0],
+                    gradFunc: (e, t, n) => {
+                        const [r, a] = t, {
+                            filterSize: s,
+                            strides: i,
+                            pad: o,
+                            dimRoundingMode: u
+                        } = n;
+                        return {
+                            x: () => Ym(e, r, a, s, i, o, u)
+                        }
+                    }
+                };
+            var Qm = Ms({
+                    maxPoolGrad_: function (e, t, n, r, a, s, i) {
+                        const o = Rs(e, "dy", "maxPoolGrad"),
+                            u = Rs(t, "input", "maxPoolGrad"),
+                            l = Rs(n, "output", "maxPoolGrad");
+                        ue(u.rank === o.rank, () => `Rank of input (${u.rank}) does not match rank of dy (${o.rank})`), ue(4 === o.rank, () => `Error in maxPoolGrad: dy must be rank 4 but got rank ${o.rank}.`), ue(4 === u.rank, () => `Error in maxPoolGrad: input must be rank 4 but got rank ${u.rank}.`), Hu("maxPoolGrad", s, i);
+                        const c = {
+                                dy: o,
+                                input: u,
+                                output: l
+                            },
+                            h = {
+                                filterSize: r,
+                                strides: a,
+                                pad: s,
+                                dimRoundingMode: i
+                            };
+                        return xs.runKernel(On, c, h)
+                    }
+                }),
+                Zm = {
+                    kernelName: er,
+                    inputsToSave: ["x"],
+                    gradFunc: (e, t, n) => {
+                        const r = t[0],
+                            {
+                                paddings: a
+                            } = n,
+                            s = a.map(e => e[0]);
+                        return {
+                            x: () => Zu(e, s, r.shape)
+                        }
+                    }
+                },
+                Jm = {
+                    kernelName: Tr,
+                    gradFunc: (e, t, n) => {
+                        const {
+                            blockShape: r,
+                            paddings: a
+                        } = n;
+                        return {
+                            x: () => tl(e, r, a)
+                        }
+                    }
+                },
+                eg = {
+                    kernelName: Rr,
+                    gradFunc: (e, t, n) => {
+                        const {
+                            axis: r
+                        } = n;
+                        return {
+                            x: () => Xu(e, r)
+                        }
+                    }
+                };
+            var tg = [Gf, Hf, qf, Kf, Yf, Xf, Qf, Zf, Jf, em, tm, nm, am, im, om, um, lm, cm, hm, pm, dm, fm, gm, mm, bm, vm, xm, wm, km, Sm, {
+                kernelName: Kt,
+                inputsToSave: ["a", "b"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t, a = Zi(n.shape, r.shape);
+                    return {
+                        a: () => {
+                            const t = bu(e, Ai(r, "float32")),
+                                s = Qi(n.shape, a);
+                            return s.length > 0 ? qu(yc(t, s), n.shape) : t
+                        },
+                        b: () => {
+                            let t = vu(e, Ai(n, "float32"));
+                            const s = Qi(r.shape, a);
+                            s.length > 0 && (t = qu(yc(t, s), r.shape));
+                            const i = jc(r);
+                            return pc(bu(t, Ai(i, "float32")))
+                        }
+                    }
+                }
+            }, Im, Am, Em, Nm, Cm, Rm, Tm, _m, Fm, Om, Lm, Pm, $m, Bm, zm, Um, Wm, Vm, Gm, qm, qm, Km, Xm, {
+                kernelName: Dn,
+                inputsToSave: ["x"],
+                outputsToSave: [!0],
+                gradFunc: (e, t, n) => {
+                    const [r, a] = t, {
+                        filterSize: s,
+                        strides: i,
+                        pad: o
+                    } = n;
+                    return {
+                        x: () => Qm(e, r, a, s, i, o)
+                    }
+                }
+            }, {
+                kernelName: Bn,
+                inputsToSave: ["x"],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, {
+                        axis: a
+                    } = n, s = ke(a, r.shape), i = pe(wc(r.shape, s)[1]);
+                    return {
+                        x: () => {
+                            const t = r.shape.slice();
+                            s.forEach(e => {
+                                t[e] = 1
+                            });
+                            const n = qu(e, t);
+                            return bu(vu(n, $c(r.shape, "float32")), i)
+                        }
+                    }
+                }
+            }, {
+                kernelName: zn,
+                inputsToSave: ["x"],
+                outputsToSave: [!0],
+                gradFunc: (e, t, n) => {
+                    const r = n,
+                        {
+                            axis: a
+                        } = r,
+                        [s, i] = t,
+                        o = Hm(e, i, s, ke(a, s.shape));
+                    return {
+                        x: () => o.x()
+                    }
+                }
+            }, {
+                kernelName: Un,
+                inputsToSave: ["a", "b"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t;
+                    return {
+                        a: () => vu(e, Ai(ec(n, r), "float32")),
+                        b: () => vu(e, Ai(Hl(n, r), "float32"))
+                    }
+                }
+            }, {
+                kernelName: Wn,
+                inputsToSave: ["x"],
+                gradFunc: (e, t, n) => {
+                    const r = t[0],
+                        {
+                            paddings: a
+                        } = n,
+                        s = a.map(e => e[0]);
+                    return {
+                        x: () => Zu(e, s, r.shape)
+                    }
+                }
+            }, {
+                kernelName: Vn,
+                inputsToSave: ["a", "b"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t, a = Zi(n.shape, r.shape);
+                    return {
+                        a: () => {
+                            const t = Qi(n.shape, a);
+                            return t.length > 0 ? qu(yc(e, t), n.shape) : e
+                        },
+                        b: () => {
+                            const t = vu(e, pc(jl(bu(n, r)))),
+                                s = Qi(r.shape, a);
+                            return s.length > 0 ? qu(yc(t, s), r.shape) : t
+                        }
+                    }
+                }
+            }, {
+                kernelName: Gn,
+                inputsToSave: ["a", "b"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t, a = Zi(n.shape, r.shape);
+                    return {
+                        a: () => {
+                            const t = vu(e, Ai(r, "float32")),
+                                s = Qi(n.shape, a);
+                            return s.length > 0 ? qu(yc(t, s), n.shape) : t
+                        },
+                        b: () => {
+                            const t = vu(e, Ai(n, "float32")),
+                                s = Qi(r.shape, a);
+                            return s.length > 0 ? qu(yc(t, s), r.shape) : t
+                        }
+                    }
+                }
+            }, {
+                kernelName: Hn,
+                gradFunc: e => ({
+                    x: () => pc(e)
+                })
+            }, {
+                kernelName: Zn,
+                inputsToSave: ["indices"],
+                gradFunc: (e, t) => {
+                    const n = t[0];
+                    return {
+                        indices: () => Pc(n.shape, "float32")
+                    }
+                }
+            }, {
+                kernelName: Qn,
+                gradFunc: e => ({
+                    x: () => Fl(e)
+                })
+            }, {
+                kernelName: Jn,
+                saveAllInputs: !0,
+                gradFunc: (e, t, n) => {
+                    const {
+                        axis: r
+                    } = n;
+                    return ap(e, r).map(e => () => e)
+                }
+            }, Zm, Zm, {
+                kernelName: nr,
+                inputsToSave: ["a", "b"],
+                outputsToSave: [!0],
+                gradFunc: (e, t) => {
+                    const [n, r, a] = t, s = n, i = r, o = Zi(s.shape, i.shape);
+                    return {
+                        a: () => {
+                            const t = Ai(i, "float32");
+                            let n = vu(e, vu(t, ah(s, gc(t, Ah(1)))));
+                            const r = Qi(s.shape, o);
+                            return r.length > 0 && (n = yc(n, r)), qu(n, s.shape)
+                        },
+                        b: () => {
+                            const t = Hl(s, 0),
+                                n = _l(t, rc(s), Fl(s));
+                            let r = vu(e, vu(a, n));
+                            const u = Qi(i.shape, o);
+                            return u.length > 0 && (r = yc(r, u)), qu(r, i.shape)
+                        }
+                    }
+                }
+            }, {
+                kernelName: rr,
+                inputsToSave: ["x", "alpha"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t, a = Hl(n, 0);
+                    return {
+                        x: () => _l(a, e, vu(e, r)),
+                        alpha: () => {
+                            let t = _l(a, Fl(e), vu(e, n));
+                            const s = Qi(r.shape, e.shape);
+                            return s.length > 0 && (t = yc(t, s)), qu(t, r.shape)
+                        }
+                    }
+                }
+            }, {
+                kernelName: or,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => bu(e, pc(jc(n)))
+                    }
+                }
+            }, {
+                kernelName: fr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t, r = vu(ec(n, 6), Hh(n));
+                    return {
+                        x: () => vu(e, Ai(r, "float32"))
+                    }
+                }
+            }, {
+                kernelName: ur,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(e, Ai(Hh(n), "float32"))
+                    }
+                }
+            }, {
+                kernelName: lr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => qu(e, n.shape)
+                    }
+                }
+            }, {
+                kernelName: pr,
+                inputsToSave: ["images"],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, a = {
+                        dy: e,
+                        images: r
+                    };
+                    return {
+                        images: () => xs.runKernel(dr, a, n)
+                    }
+                }
+            }, {
+                kernelName: cr,
+                inputsToSave: ["images"],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, a = {
+                        dy: e,
+                        images: r
+                    };
+                    return {
+                        images: () => xs.runKernel(hr, a, n)
+                    }
+                }
+            }, {
+                kernelName: mr,
+                gradFunc: (e, t, n) => {
+                    const {
+                        dims: r
+                    } = n, a = ke(r, e.shape);
+                    return {
+                        x: () => bh(e, a)
+                    }
+                }
+            }, {
+                kernelName: gr,
+                gradFunc: e => ({
+                    x: () => Fl(e)
+                })
+            }, {
+                kernelName: yr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => pc(bu(e, vu(ah(n, 1.5), 2)))
+                    }
+                }
+            }, {
+                kernelName: vr,
+                inputsToSave: ["condition"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        condition: () => Ai(Fl(n), "float32"),
+                        t: () => vu(e, Ai(n, e.dtype)),
+                        e: () => vu(e, Ai(Tc(n), e.dtype))
+                    }
+                }
+            }, {
+                kernelName: xr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => {
+                            const t = Hl(n, Ah(0)),
+                                r = Ah(tf),
+                                a = Ah(nf),
+                                s = vu(e, a),
+                                i = vu(vu(e, r), $l(Ai(n, "float32")));
+                            return _l(t, s, i)
+                        }
+                    }
+                }
+            }, {
+                kernelName: Ar,
+                outputsToSave: [!0],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(e, vu(n, gc(Ah(1), n)))
+                    }
+                }
+            }, {
+                kernelName: Ir,
+                gradFunc: e => ({
+                    x: () => Fl(e)
+                })
+            }, {
+                kernelName: kr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(kl(Ai(n, "float32")), e)
+                    }
+                }
+            }, {
+                kernelName: Sr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(Sl(Ai(n, "float32")), e)
+                    }
+                }
+            }, {
+                kernelName: wr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, {
+                        begin: a,
+                        size: s
+                    } = n, i = r.shape, [o, u] = To(r, a, s), l = [];
+                    for (let t = 0; t < e.rank; t++) l.push([o[t], i[t] - o[t] - u[t]]);
+                    return {
+                        x: () => Qc(e, l)
+                    }
+                }
+            }, {
+                kernelName: _r,
+                outputsToSave: [!0],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, {
+                        dim: a
+                    } = n, s = vu(e, r);
+                    return {
+                        logits: () => gc(s, vu(yc(s, [a], !0), r))
+                    }
+                }
+            }, {
+                kernelName: Er,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(e, Qu(n))
+                    }
+                }
+            }, Jm, Jm, eg, eg, {
+                kernelName: Nr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => bu(e, vu(Wh(Ai(n, "float32")), 2))
+                    }
+                }
+            }, {
+                kernelName: Pr,
+                inputsToSave: ["a", "b"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t, a = Ah(2);
+                    return {
+                        a: () => vu(e, vu(a, gc(n, r))),
+                        b: () => vu(e, vu(a, gc(r, n)))
+                    }
+                }
+            }, {
+                kernelName: $r,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(e, vu(Ai(n, "float32"), 2))
+                    }
+                }
+            }, {
+                kernelName: ea,
+                gradFunc: e => ({
+                    x: () => Fl(e)
+                })
+            }, {
+                kernelName: Vr,
+                inputsToSave: ["a", "b"],
+                gradFunc: (e, t) => {
+                    const [n, r] = t, a = Zi(n.shape, r.shape);
+                    return {
+                        a: () => {
+                            let t = e;
+                            const r = Qi(n.shape, a);
+                            return r.length > 0 && (t = yc(t, r)), qu(t, n.shape)
+                        },
+                        b: () => {
+                            let t = e;
+                            const n = Qi(r.shape, a);
+                            return n.length > 0 && (t = yc(t, n)), qu(pc(t), r.shape)
+                        }
+                    }
+                }
+            }, {
+                kernelName: Cr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, a = r.shape.slice(), {
+                        axis: s
+                    } = n;
+                    ke(s, r.shape).forEach(e => {
+                        a[e] = 1
+                    });
+                    const i = qu(e, a),
+                        o = vu(i, $c(r.shape, "float32"));
+                    return {
+                        x: () => o
+                    }
+                }
+            }, {
+                kernelName: jr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => bu(e, jc(kl(n)))
+                    }
+                }
+            }, {
+                kernelName: Gr,
+                outputsToSave: [!0],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => vu(gc(Ah(1), jc(n)), e)
+                    }
+                }
+            }, {
+                kernelName: Hr,
+                inputsToSave: ["x"],
+                gradFunc: (e, t, n) => {
+                    const [r] = t, {
+                        reps: a
+                    } = n;
+                    return {
+                        x: () => {
+                            let t = Fl(r);
+                            if (1 === r.rank)
+                                for (let n = 0; n < a[0]; ++n) t = gu(t, Zu(e, [n * r.shape[0]], [r.shape[0]]));
+                            else if (2 === r.rank)
+                                for (let n = 0; n < a[0]; ++n)
+                                    for (let s = 0; s < a[1]; ++s) t = gu(t, Zu(e, [n * r.shape[0], s * r.shape[1]], [r.shape[0], r.shape[1]]));
+                            else if (3 === r.rank)
+                                for (let n = 0; n < a[0]; ++n)
+                                    for (let s = 0; s < a[1]; ++s)
+                                        for (let i = 0; i < a[2]; ++i) t = gu(t, Zu(e, [n * r.shape[0], s * r.shape[1], i * r.shape[2]], [r.shape[0], r.shape[1], r.shape[2]]));
+                            else {
+                                if (4 !== r.rank) throw new Error(`Gradient for tile operation is not implemented for rank-${r.rank} tensors yet.`);
+                                for (let n = 0; n < a[0]; ++n)
+                                    for (let s = 0; s < a[1]; ++s)
+                                        for (let i = 0; i < a[2]; ++i)
+                                            for (let o = 0; o < a[3]; ++o) t = gu(t, Zu(e, [n * r.shape[0], s * r.shape[1], i * r.shape[2], o * r.shape[3]], [r.shape[0], r.shape[1], r.shape[2], r.shape[3]]))
+                            }
+                            return t
+                        }
+                    }
+                }
+            }, {
+                kernelName: Yr,
+                gradFunc: (e, t, n) => {
+                    const r = n,
+                        {
+                            perm: a
+                        } = r,
+                        s = Ac(a);
+                    return {
+                        x: () => qi(e, s)
+                    }
+                }
+            }, {
+                kernelName: Qr,
+                gradFunc: (e, t, n) => {
+                    const r = n,
+                        {
+                            axis: a
+                        } = r;
+                    return {
+                        value: () => Gh(e, a)
+                    }
+                }
+            }, {
+                kernelName: Zr,
+                inputsToSave: ["segmentIds"],
+                gradFunc: (e, t) => {
+                    const [n] = t;
+                    return {
+                        x: () => function (e, t) {
+                            const n = Oc(t, Fl(t)),
+                                r = Gl(e, n);
+                            let a = ql(t, Ah(0, "int32"));
+                            const s = r.rank - a.rank;
+                            for (let e = 0; e < s; ++e) a = Bl(a, e + 1);
+                            a = Cc(a, $c(r.shape, "bool"));
+                            const i = Fl(r);
+                            return _l(a, r, i)
+                        }(e, n)
+                    }
+                }
+            }, {
+                kernelName: Jr,
+                gradFunc: e => ({
+                    x: () => Fl(e)
+                })
+            }];
+            for (const e of tg) fa(e);
+            Xa().prototype.abs = function () {
+                return this.throwIfDisposed(), xu(this)
+            }, Xa().prototype.acos = function () {
+                return this.throwIfDisposed(), wu(this)
+            }, Xa().prototype.acosh = function () {
+                return this.throwIfDisposed(), ku(this)
+            }, Xa().prototype.add = function (e) {
+                return this.throwIfDisposed(), gu(this, e)
+            }, Xa().prototype.all = function (e, t) {
+                return this.throwIfDisposed(), Iu(this, e, t)
+            }, Xa().prototype.any = function (e, t) {
+                return this.throwIfDisposed(), Au(this, e, t)
+            }, Xa().prototype.argMax = function (e) {
+                return this.throwIfDisposed(), Eu(this, e)
+            }, Xa().prototype.argMin = function (e) {
+                return this.throwIfDisposed(), Nu(this, e)
+            }, Xa().prototype.asScalar = function () {
+                return this.throwIfDisposed(), ue(1 === this.size, () => "The array must have only 1 element."), qu(this, [])
+            }, Xa().prototype.asType = function (e) {
+                return this.throwIfDisposed(), Ai(this, e)
+            }, Xa().prototype.as1D = function () {
+                return this.throwIfDisposed(), qu(this, [this.size])
+            }, Xa().prototype.as2D = function (e, t) {
+                return this.throwIfDisposed(), qu(this, [e, t])
+            }, Xa().prototype.as3D = function (e, t, n) {
+                return this.throwIfDisposed(), qu(this, [e, t, n])
+            }, Xa().prototype.as4D = function (e, t, n, r) {
+                return this.throwIfDisposed(), qu(this, [e, t, n, r])
+            }, Xa().prototype.as5D = function (e, t, n, r, a) {
+                return this.throwIfDisposed(), qu(this, [e, t, n, r, a])
+            }, Xa().prototype.asin = function () {
+                return this.throwIfDisposed(), Cu(this)
+            }, Xa().prototype.asinh = function () {
+                return this.throwIfDisposed(), Tu(this)
+            }, Xa().prototype.atan = function () {
+                return this.throwIfDisposed(), Ru(this)
+            }, Xa().prototype.atan2 = function (e) {
+                return this.throwIfDisposed(), _u(this, e)
+            }, Xa().prototype.atanh = function () {
+                return this.throwIfDisposed(), Fu(this)
+            }, Xa().prototype.avgPool = function (e, t, n, r) {
+                return this.throwIfDisposed(), Ku(this, e, t, n, r)
+            }, Xa().prototype.batchToSpaceND = function (e, t) {
+                return this.throwIfDisposed(), tl(this, e, t)
+            }, Xa().prototype.batchNorm = function (e, t, n, r, a) {
+                return this.throwIfDisposed(), nl(this, e, t, n, r, a)
+            }, Xa().prototype.broadcastTo = function (e) {
+                return this.throwIfDisposed(), ul(this, e)
+            }, Xa().prototype.cast = function (e) {
+                return this.throwIfDisposed(), Ai(this, e)
+            }, Xa().prototype.ceil = function () {
+                return this.throwIfDisposed(), ll(this)
+            }, Xa().prototype.clipByValue = function (e, t) {
+                return this.throwIfDisposed(), cl(this, e, t)
+            }, Xa().prototype.concat = function (e, t) {
+                return this.throwIfDisposed(), e instanceof Ya && (e = [e]), Xu([this, ...e], t)
+            }, Xa().prototype.conv1d = function (e, t, n, r, a, s) {
+                return this.throwIfDisposed(), gl(this, e, t, n, r, a, s)
+            }, Xa().prototype.conv2dTranspose = function (e, t, n, r, a) {
+                return this.throwIfDisposed(), bl(this, e, t, n, r, a)
+            }, Xa().prototype.conv2d = function (e, t, n, r, a, s) {
+                return this.throwIfDisposed(), ml(this, e, t, n, r, a, s)
+            }, Xa().prototype.cos = function () {
+                return this.throwIfDisposed(), kl(this)
+            }, Xa().prototype.cosh = function () {
+                return this.throwIfDisposed(), Sl(this)
+            }, Xa().prototype.cumsum = function (e, t, n) {
+                return this.throwIfDisposed(), Il(this, e, t, n)
+            }, Xa().prototype.depthToSpace = function (e, t) {
+                return this.throwIfDisposed(), El(this, e, t)
+            }, Xa().prototype.depthwiseConv2d = function (e, t, n, r, a, s) {
+                return this.throwIfDisposed(), Nl(this, e, t, n, r, a, s)
+            }, Xa().prototype.dilation2d = function (e, t, n, r, a) {
+                return this.throwIfDisposed(), Tl(this, e, t, n, r, a)
+            }, Xa().prototype.divNoNan = function (e) {
+                return this.throwIfDisposed(), Ml(this, e)
+            }, Xa().prototype.div = function (e) {
+                return this.throwIfDisposed(), bu(this, e)
+            }, Xa().prototype.dot = function (e) {
+                return this.throwIfDisposed(), Dl(this, e)
+            }, Xa().prototype.elu = function () {
+                return this.throwIfDisposed(), Ll(this)
+            }, Xa().prototype.equal = function (e) {
+                return this.throwIfDisposed(), Rl(this, e)
+            }, Xa().prototype.erf = function () {
+                return this.throwIfDisposed(), Pl(this)
+            }, Xa().prototype.exp = function () {
+                return this.throwIfDisposed(), $l(this)
+            }, Xa().prototype.expandDims = function (e) {
+                return this.throwIfDisposed(), Bl(this, e)
+            }, Xa().prototype.expm1 = function () {
+                return this.throwIfDisposed(), zl(this)
+            }, Xa().prototype.fft = function () {
+                return this.throwIfDisposed(), Ph(this)
+            }, Xa().prototype.flatten = function () {
+                return this.throwIfDisposed(), qu(this, [this.size])
+            }, Xa().prototype.floor = function () {
+                return this.throwIfDisposed(), jl(this)
+            }, Xa().prototype.floorDiv = function (e) {
+                return this.throwIfDisposed(), yu(this, e)
+            }, Xa().prototype.gather = function (e, t) {
+                return this.throwIfDisposed(), Gl(this, e, t)
+            }, Xa().prototype.greaterEqual = function (e) {
+                return this.throwIfDisposed(), ql(this, e)
+            }, Xa().prototype.greater = function (e) {
+                return this.throwIfDisposed(), Hl(this, e)
+            }, Xa().prototype.ifft = function () {
+                return this.throwIfDisposed(), $h(this)
+            }, Xa().prototype.irfft = function () {
+                return this.throwIfDisposed(), Bh(this)
+            }, Xa().prototype.isFinite = function () {
+                return this.throwIfDisposed(), Yl(this)
+            }, Xa().prototype.isInf = function () {
+                return this.throwIfDisposed(), Xl(this)
+            }, Xa().prototype.isNaN = function () {
+                return this.throwIfDisposed(), Ql(this)
+            }, Xa().prototype.leakyRelu = function (e) {
+                return this.throwIfDisposed(), Zl(this, e)
+            }, Xa().prototype.lessEqual = function (e) {
+                return this.throwIfDisposed(), ec(this, e)
+            }, Xa().prototype.less = function (e) {
+                return this.throwIfDisposed(), Jl(this, e)
+            }, Xa().prototype.localResponseNormalization = function (e, t, n, r) {
+                return this.throwIfDisposed(), nc(this, e, t, n, r)
+            }, Xa().prototype.logSigmoid = function () {
+                return this.throwIfDisposed(), fc(this)
+            }, Xa().prototype.logSoftmax = function (e) {
+                return this.throwIfDisposed(), bc(this, e)
+            }, Xa().prototype.logSumExp = function (e, t) {
+                return this.throwIfDisposed(), Nc(this, e, t)
+            }, Xa().prototype.log = function () {
+                return this.throwIfDisposed(), rc(this)
+            }, Xa().prototype.log1p = function () {
+                return this.throwIfDisposed(), ac(this)
+            }, Xa().prototype.logicalAnd = function (e) {
+                return this.throwIfDisposed(), Cc(this, e)
+            }, Xa().prototype.logicalNot = function () {
+                return this.throwIfDisposed(), Tc(this)
+            }, Xa().prototype.logicalOr = function (e) {
+                return this.throwIfDisposed(), Rc(this, e)
+            }, Xa().prototype.logicalXor = function (e) {
+                return this.throwIfDisposed(), _c(this, e)
+            }, Xa().prototype.matMul = function (e, t, n) {
+                return this.throwIfDisposed(), Gi(this, e, t, n)
+            }, Xa().prototype.maxPool = function (e, t, n, r) {
+                return this.throwIfDisposed(), Fc(this, e, t, n, r)
+            }, Xa().prototype.max = function (e, t) {
+                return this.throwIfDisposed(), mc(this, e, t)
+            }, Xa().prototype.maximum = function (e) {
+                return this.throwIfDisposed(), Oc(this, e)
+            }, Xa().prototype.mean = function (e, t) {
+                return this.throwIfDisposed(), Lc(this, e, t)
+            }, Xa().prototype.min = function (e, t) {
+                return this.throwIfDisposed(), zc(this, e, t)
+            }, Xa().prototype.minimum = function (e) {
+                return this.throwIfDisposed(), Uc(this, e)
+            }, Xa().prototype.mirrorPad = function (e, t) {
+                return this.throwIfDisposed(), Wc(this, e, t)
+            }, Xa().prototype.mod = function (e) {
+                return this.throwIfDisposed(), Vc(this, e)
+            }, Xa().prototype.mul = function (e) {
+                return this.throwIfDisposed(), vu(this, e)
+            }, Xa().prototype.neg = function () {
+                return this.throwIfDisposed(), pc(this)
+            }, Xa().prototype.norm = function (e, t, n) {
+                return this.throwIfDisposed(), lp(this, e, t, n)
+            }, Xa().prototype.notEqual = function (e) {
+                return this.throwIfDisposed(), Kc(this, e)
+            }, Xa().prototype.oneHot = function (e, t = 1, n = 0) {
+                return this.throwIfDisposed(), Hi(this, e, t, n)
+            }, Xa().prototype.onesLike = function () {
+                return this.throwIfDisposed(), Yc(this)
+            }, Xa().prototype.pad = function (e, t) {
+                return this.throwIfDisposed(), Qc(this, e, t)
+            }, Xa().prototype.pool = function (e, t, n, r, a, s) {
+                return this.throwIfDisposed(), rh(this, e, t, n, r, a, s)
+            }, Xa().prototype.pow = function (e) {
+                return this.throwIfDisposed(), ah(this, e)
+            }, Xa().prototype.prelu = function (e) {
+                return this.throwIfDisposed(), sh(this, e)
+            }, Xa().prototype.prod = function (e, t) {
+                return this.throwIfDisposed(), ih(this, e, t)
+            }, Xa().prototype.reciprocal = function () {
+                return this.throwIfDisposed(), mh(this)
+            }, Xa().prototype.relu = function () {
+                return this.throwIfDisposed(), gh(this)
+            }, Xa().prototype.relu6 = function () {
+                return this.throwIfDisposed(), yh(this)
+            }, Xa().prototype.reshapeAs = function (e) {
+                return this.throwIfDisposed(), qu(this, e.shape)
+            }, Xa().prototype.reshape = function (e) {
+                return this.throwIfDisposed(), qu(this, e)
+            }, Xa().prototype.resizeBilinear = function (e, t, n) {
+                return this.throwIfDisposed(), Jp(this, e, t, n)
+            }, Xa().prototype.resizeNearestNeighbor = function (e, t, n) {
+                return this.throwIfDisposed(), ed(this, e, t, n)
+            }, Xa().prototype.reverse = function (e) {
+                return this.throwIfDisposed(), bh(this, e)
+            }, Xa().prototype.rfft = function () {
+                return this.throwIfDisposed(), Uh(this)
+            }, Xa().prototype.round = function () {
+                return this.throwIfDisposed(), Sh(this)
+            }, Xa().prototype.rsqrt = function () {
+                return this.throwIfDisposed(), Ih(this)
+            }, Xa().prototype.selu = function () {
+                return this.throwIfDisposed(), Eh(this)
+            }, Xa().prototype.separableConv2d = function (e, t, n, r, a, s) {
+                return this.throwIfDisposed(), Nh(this, e, t, n, r, a, s)
+            }, Xa().prototype.sigmoid = function () {
+                return this.throwIfDisposed(), Qu(this)
+            }, Xa().prototype.sign = function () {
+                return this.throwIfDisposed(), Th(this)
+            }, Xa().prototype.sin = function () {
+                return this.throwIfDisposed(), Rh(this)
+            }, Xa().prototype.sinh = function () {
+                return this.throwIfDisposed(), _h(this)
+            }, Xa().prototype.slice = function (e, t) {
+                return this.throwIfDisposed(), Zu(this, e, t)
+            }, Xa().prototype.softmax = function (e) {
+                return this.throwIfDisposed(), Lh(this, e)
+            }, Xa().prototype.softplus = function () {
+                return this.throwIfDisposed(), dc(this)
+            }, Xa().prototype.spaceToBatchND = function (e, t) {
+                return this.throwIfDisposed(), nh(this, e, t)
+            }, Xa().prototype.split = function (e, t) {
+                return this.throwIfDisposed(), zh(this, e, t)
+            }, Xa().prototype.sqrt = function () {
+                return this.throwIfDisposed(), Wh(this)
+            }, Xa().prototype.square = function () {
+                return this.throwIfDisposed(), jc(this)
+            }, Xa().prototype.squaredDifference = function (e) {
+                return this.throwIfDisposed(), Vh(this, e)
+            }, Xa().prototype.squeeze = function (e) {
+                return this.throwIfDisposed(), jh(this, e)
+            }, Xa().prototype.stack = function (e, t) {
+                this.throwIfDisposed();
+                const n = e instanceof Ya ? [this, e] : [this, ...e];
+                return Gh(n, t)
+            }, Xa().prototype.step = function (e) {
+                return this.throwIfDisposed(), Hh(this, e)
+            }, Xa().prototype.stridedSlice = function (e, t, n, r, a, s, i, o) {
+                return this.throwIfDisposed(), qh(this, e, t, n, r, a, s, i, o)
+            }, Xa().prototype.sub = function (e) {
+                return this.throwIfDisposed(), gc(this, e)
+            }, Xa().prototype.sum = function (e, t) {
+                return this.throwIfDisposed(), yc(this, e, t)
+            }, Xa().prototype.tan = function () {
+                return this.throwIfDisposed(), Kh(this)
+            }, Xa().prototype.tanh = function () {
+                return this.throwIfDisposed(), Ju(this)
+            }, Xa().prototype.tile = function (e) {
+                return this.throwIfDisposed(), Ul(this, e)
+            }, Xa().prototype.toBool = function () {
+                return this.throwIfDisposed(), Ai(this, "bool")
+            }, Xa().prototype.toFloat = function () {
+                return this.throwIfDisposed(), Ai(this, "float32")
+            }, Xa().prototype.toInt = function () {
+                return this.throwIfDisposed(), Ai(this, "int32")
+            }, Xa().prototype.topk = function (e, t) {
+                return this.throwIfDisposed(), ep(this, e, t)
+            }, Xa().prototype.transpose = function (e) {
+                return this.throwIfDisposed(), qi(this, e)
+            }, Xa().prototype.unique = function (e) {
+                return this.throwIfDisposed(), np(this, e)
+            }, Xa().prototype.unsortedSegmentSum = function (e, t) {
+                return this.throwIfDisposed(), rp(this, e, t)
+            }, Xa().prototype.unstack = function (e) {
+                return this.throwIfDisposed(), ap(this, e)
+            }, Xa().prototype.where = function (e, t) {
+                return this.throwIfDisposed(), _l(e, this, t)
+            }, Xa().prototype.zerosLike = function () {
+                return this.throwIfDisposed(), Fl(this)
+            };
+            var ng, rg = {};
+
+            function ag() {
+                return null == ng && (ng = fu().epsilon()), ng
+            }
+            w(rg, {
+                maxNorm: () => Pg,
+                minMaxNorm: () => zg,
+                nonNeg: () => Bg,
+                unitNorm: () => $g
+            });
+            var sg = class extends Error {
+                    constructor(e) {
+                        super(e), Object.setPrototypeOf(this, sg.prototype)
+                    }
+                },
+                ig = class extends Error {
+                    constructor(e) {
+                        super(e), Object.setPrototypeOf(this, ig.prototype)
+                    }
+                },
+                og = class extends Error {
+                    constructor(e) {
+                        super(e), Object.setPrototypeOf(this, og.prototype)
+                    }
+                },
+                ug = class extends Error {
+                    constructor(e) {
+                        super(e), Object.setPrototypeOf(this, ug.prototype)
+                    }
+                },
+                lg = class extends Error {
+                    constructor(e) {
+                        super(e), Object.setPrototypeOf(this, lg.prototype)
+                    }
+                };
+
+            function cg(e, t) {
+                if (Array.isArray(e)) {
+                    let n = [];
+                    for (let r = 0; r < t; r++) n = n.concat(e);
+                    return n
+                } {
+                    const n = new Array(t);
+                    return n.fill(e), n
+                }
+            }
+
+            function hg(e, t) {
+                if (!e) throw new lg(t)
+            }
+
+            function pg(e, t) {
+                let n = 0;
+                for (const r of e) r === t && n++;
+                return n
+            }
+
+            function dg(e) {
+                return 1 === e.length ? e[0] : e
+            }
+
+            function fg(e) {
+                return Array.isArray(e) ? e : [e]
+            }
+
+            function mg(e) {
+                const t = e.replace(/(.)([A-Z][a-z0-9]+)/g, "$1_$2").replace(/([a-z])([A-Z])/g, "$1_$2").toLowerCase();
+                return "_" !== t[0] ? t : "private" + t
+            }
+
+            function gg(e) {
+                return e.length <= 1 || -1 === e.indexOf("_") ? e : e.replace(/[_]+(\w|$)/g, (e, t) => t.toUpperCase())
+            }
+            var yg = {};
+
+            function bg(e) {
+                if (null == e) return null;
+                const t = {};
+                return t.className = e.getClassName(), t.config = e.getConfig(), t
+            }
+
+            function vg(e, t = {}, n = {}, r = "object", a = !1) {
+                if ("string" == typeof e) {
+                    const a = e;
+                    let s;
+                    if (a in n) s = n[a];
+                    else if (a in yg) s = yg[a];
+                    else if (s = t[a], null == s) throw new og(`Unknown ${r}: ${e}. This may be due to one of the following reasons:\n1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);
+                    return s
+                } {
+                    const s = e;
+                    if (null == s.className || null == s.config) throw new og(`${r}: Improper config format: ${JSON.stringify(s)}.\n'className' and 'config' must set.`);
+                    const i = s.className;
+                    let o, u;
+                    if (i in n ? [o, u] = n[i] : i in yg ? [o, u] = yg.className : i in t && ([o, u] = t[i]), null == o) throw new og(`Unknown ${r}: ${i}. This may be due to one of the following reasons:\n1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);
+                    if (null != u) {
+                        const e = {};
+                        for (const t of Object.keys(yg)) e[t] = yg[t];
+                        for (const t of Object.keys(n)) e[t] = n[t];
+                        s.config.customObjects = e;
+                        const t = Object.assign({}, yg);
+                        for (const e of Object.keys(n)) yg[e] = n[e];
+                        ! function e(t) {
+                            if (null != t && "object" == typeof t)
+                                if (Array.isArray(t)) t.forEach(t => e(t));
+                                else {
+                                    const n = Object.keys(t);
+                                    for (const r of n) {
+                                        const n = t[r];
+                                        null != n && "object" == typeof n && (Array.isArray(n) || "ndarray" !== n.type || "number" != typeof n.value ? e(n) : t[r] = n.value)
+                                    }
+                                }
+                        }(s.config);
+                        const r = u(o, s.config, n, a);
+                        return yg = Object.assign({}, t), r
+                    } {
+                        const e = Object.assign({}, yg);
+                        for (const e of Object.keys(n)) yg[e] = n[e];
+                        const t = new o(s.config);
+                        return yg = Object.assign({}, e), t
+                    }
+                }
+            }
+
+            function xg(e, t) {
+                return -1 * function (e, t) {
+                    return e < t ? -1 : e > t ? 1 : 0
+                }(e, t)
+            }
+
+            function wg(e) {
+                if (null == e) return e;
+                const t = [];
+                for (const n of e) - 1 === t.indexOf(n) && t.push(n);
+                return t
+            }
+
+            function kg(e) {
+                if (null == e) throw new og("Invalid value in obj: " + JSON.stringify(e));
+                for (const t in e)
+                    if (e.hasOwnProperty(t)) return !1;
+                return !0
+            }
+
+            function Sg(e, t, n) {
+                if (null != n && e.indexOf(n) < 0) throw new og(`${n} is not a valid ${t}.  Valid values are ${e} or null/undefined.`)
+            }
+
+            function Ig(e, t, n = 0, r = 1 / 0) {
+                return hg(n >= 0), hg(r >= n), Array.isArray(e) && e.length >= n && e.length <= r && e.every(e => typeof e === t)
+            }
+
+            function Ag(e, t) {
+                Array.isArray(e) ? (va.assert(e.length > 0, () => t + " is unexpectedly an empty array."), e.forEach((e, n) => Ag(e, `element ${n+1} of ${t}`))) : va.assert(Number.isInteger(e) && e > 0, () => `Expected ${t} to be a positive integer, but got ${function e(t){return null===t?"null":Array.isArray(t)?"["+t.map(t=>e(t)).join(",")+"]":"string"==typeof t?`"${t}"`:""+t}(e)}.`)
+            }
+
+            function Eg(e) {
+                return "relu" === e ? "relu" : "linear" === e ? "linear" : "elu" === e ? "elu" : null
+            }
+
+            function Ng(e, t) {
+                return ru(() => Wh(yc(vu(e, e), t, !0)))
+            }
+            var Cg = class extends Fo.Serializable {
+                    getConfig() {
+                        return {}
+                    }
+                },
+                Tg = class extends Cg {
+                    constructor(e) {
+                        super(), this.defaultMaxValue = 2, this.defaultAxis = 0, this.maxValue = null != e.maxValue ? e.maxValue : this.defaultMaxValue, this.axis = null != e.axis ? e.axis : this.defaultAxis
+                    }
+                    apply(e) {
+                        return ru(() => {
+                            const t = Ng(e, this.axis),
+                                n = cl(t, 0, this.maxValue);
+                            return vu(e, bu(n, gu(ag(), t)))
+                        })
+                    }
+                    getConfig() {
+                        return {
+                            maxValue: this.maxValue,
+                            axis: this.axis
+                        }
+                    }
+                };
+            Tg.className = "MaxNorm", Fo.registerClass(Tg);
+            var Rg = class extends Cg {
+                constructor(e) {
+                    super(), this.defaultAxis = 0, this.axis = null != e.axis ? e.axis : this.defaultAxis
+                }
+                apply(e) {
+                    return ru(() => bu(e, gu(ag(), Ng(e, this.axis))))
+                }
+                getConfig() {
+                    return {
+                        axis: this.axis
+                    }
+                }
+            };
+            Rg.className = "UnitNorm", Fo.registerClass(Rg);
+            var _g = class extends Cg {
+                apply(e) {
+                    return gh(e)
+                }
+            };
+            _g.className = "NonNeg", Fo.registerClass(_g);
+            var Fg = class extends Cg {
+                constructor(e) {
+                    super(), this.defaultMinValue = 0, this.defaultMaxValue = 1, this.defaultRate = 1, this.defaultAxis = 0, this.minValue = null != e.minValue ? e.minValue : this.defaultMinValue, this.maxValue = null != e.maxValue ? e.maxValue : this.defaultMaxValue, this.rate = null != e.rate ? e.rate : this.defaultRate, this.axis = null != e.axis ? e.axis : this.defaultAxis
+                }
+                apply(e) {
+                    return ru(() => {
+                        const t = Ng(e, this.axis),
+                            n = gu(vu(this.rate, cl(t, this.minValue, this.maxValue)), vu(1 - this.rate, t));
+                        return vu(e, bu(n, gu(ag(), t)))
+                    })
+                }
+                getConfig() {
+                    return {
+                        minValue: this.minValue,
+                        maxValue: this.maxValue,
+                        rate: this.rate,
+                        axis: this.axis
+                    }
+                }
+            };
+            Fg.className = "MinMaxNorm", Fo.registerClass(Fg);
+            var Mg = {
+                maxNorm: "MaxNorm",
+                minMaxNorm: "MinMaxNorm",
+                nonNeg: "NonNeg",
+                unitNorm: "UnitNorm"
+            };
+
+            function Dg(e) {
+                return bg(e)
+            }
+
+            function Og(e, t = {}) {
+                return vg(e, Fo.SerializationMap.getMap().classNameMap, t, "constraint")
+            }
+
+            function Lg(e) {
+                if (null == e) return null;
+                if ("string" == typeof e) {
+                    return Og({
+                        className: e in Mg ? Mg[e] : e,
+                        config: {}
+                    })
+                }
+                return e instanceof Cg ? e : Og(e)
+            }
+
+            function Pg(e) {
+                return new Tg(e)
+            }
+
+            function $g(e) {
+                return new Rg(e)
+            }
+
+            function Bg() {
+                return new _g
+            }
+
+            function zg(e) {
+                return new Fg(e)
+            }
+            var Ug = {};
+            w(Ug, {
+                constant: () => qy,
+                glorotNormal: () => eb,
+                glorotUniform: () => Jy,
+                heNormal: () => tb,
+                heUniform: () => nb,
+                identity: () => Qy,
+                leCunNormal: () => rb,
+                leCunUniform: () => ab,
+                ones: () => Hy,
+                orthogonal: () => sb,
+                randomNormal: () => Yy,
+                randomUniform: () => Ky,
+                truncatedNormal: () => Xy,
+                varianceScaling: () => Zy,
+                zeros: () => Gy
+            });
+            var Wg = ["channelsFirst", "channelsLast"],
+                Vg = ["nearest", "bilinear"],
+                jg = ["valid", "same", "causal"],
+                Gg = ["max", "avg"],
+                Hg = ["sum", "mul", "concat", "ave"],
+                qg = new Map;
+
+            function Kg(e) {
+                Sg(Wg, "DataFormat", e)
+            }
+
+            function Yg(e) {
+                Sg(jg, "PaddingMode", e)
+            }
+
+            function Xg(e) {
+                Sg(Gg, "PoolMode", e)
+            }
+            var Qg = [];
+
+            function Zg(e, t) {
+                Qg.push(e);
+                try {
+                    const e = t();
+                    return Qg.pop(), e
+                } catch (e) {
+                    throw Qg.pop(), e
+                }
+            }
+
+            function Jg(e) {
+                if (!ny(e)) throw new Error("Not a valid tensor name: '" + e + "'");
+                return (0 === Qg.length ? "" : Qg.join("/") + "/") + e
+            }
+
+            function ey(e) {
+                if (!ny(e)) throw new Error("Not a valid tensor name: '" + e + "'");
+                qg.has(e) || qg.set(e, 0);
+                const t = qg.get(e);
+                if (qg.set(e, qg.get(e) + 1), t > 0) {
+                    const n = `${e}_${t}`;
+                    return qg.set(n, 1), n
+                }
+                return e
+            }
+            var ty = new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);
+
+            function ny(e) {
+                return !!e.match(ty)
+            }
+
+            function ry(e, t, n) {
+                null == t && (t = 0), null == n && (n = e.length);
+                let r = 1;
+                for (let a = t; a < n; ++a) r *= e[a];
+                return r
+            }
+
+            function ay(e) {
+                if (0 === e.length) return Number.NaN;
+                let t = Number.POSITIVE_INFINITY;
+                for (let n = 0; n < e.length; n++) {
+                    const r = e[n];
+                    r < t && (t = r)
+                }
+                return t
+            }
+
+            function sy(e) {
+                if (0 === e.length) return Number.NaN;
+                let t = Number.NEGATIVE_INFINITY;
+                for (let n = 0; n < e.length; n++) {
+                    const r = e[n];
+                    r > t && (t = r)
+                }
+                return t
+            }
+
+            function iy(e, t) {
+                if (t < e) throw new og(`end (${t}) < begin (${e}) is forbidden.`);
+                const n = [];
+                for (let r = e; r < t; ++r) n.push(r);
+                return n
+            }
+
+            function oy(e, t) {
+                return Ai(e, t)
+            }
+
+            function uy(e, t = -1) {
+                const n = e.shape.slice();
+                return t < 0 && (t = n.length + t + 1), n.splice(t, 0, 1), qu(e, n)
+            }
+
+            function ly(e, t, n) {
+                return ru(() => {
+                    switch (e.rank) {
+                        case 1:
+                            return Fh(e, t, n);
+                        case 2:
+                            return Mh(e, [t, 0], [n, e.shape[1]]);
+                        case 3:
+                            return Dh(e, [t, 0, 0], [n, e.shape[1], e.shape[2]]);
+                        case 4:
+                            return Oh(e, [t, 0, 0, 0], [n, e.shape[1], e.shape[2], e.shape[3]]);
+                        case 5:
+                            return Zu(e, [t, 0, 0, 0, 0], [n, e.shape[1], e.shape[2], e.shape[3], e.shape[4]]);
+                        case 6:
+                            return Zu(e, [t, 0, 0, 0, 0, 0], [n, e.shape[1], e.shape[2], e.shape[3], e.shape[4], e.shape[5]]);
+                        default:
+                            throw new og("sliceAlongFirstAxis() received an unsupported tensor rank: " + e.rank)
+                    }
+                })
+            }
+
+            function cy(e, t, n) {
+                return ru(() => {
+                    switch (e.rank) {
+                        case 1:
+                            return Fh(e, t, n);
+                        case 2:
+                            return Mh(e, [0, t], [e.shape[0], n]);
+                        case 3:
+                            return Dh(e, [0, 0, t], [e.shape[0], e.shape[1], n]);
+                        case 4:
+                            return Oh(e, [0, 0, 0, t], [e.shape[0], e.shape[1], e.shape[2], n]);
+                        default:
+                            throw new og("sliceAlongLastAxis() received an unsupported tensor rank: " + e.rank)
+                    }
+                })
+            }
+
+            function hy(e, t, n, r) {
+                return ru(() => {
+                    switch (e.rank) {
+                        case 1:
+                            return Fh(e, t, n);
+                        case 2:
+                            switch (r) {
+                                case 1:
+                                    return ly(e, t, n);
+                                case 2:
+                                    return cy(e, t, n);
+                                default:
+                                    throw new og("The axis is not within the rank of the tensor " + r)
+                            }
+                            case 3:
+                                switch (r) {
+                                    case 1:
+                                        return ly(e, t, n);
+                                    case 2:
+                                        return Dh(e, [0, t, 0], [e.shape[0], n, e.shape[2]]);
+                                    case 3:
+                                        return cy(e, t, n);
+                                    default:
+                                        throw new og("The axis is not within the rank of the tensor " + r)
+                                }
+                                case 4:
+                                    switch (r) {
+                                        case 1:
+                                            return ly(e, t, n);
+                                        case 2:
+                                            return Oh(e, [0, t, 0, 0], [e.shape[0], n, e.shape[2], e.shape[3]]);
+                                        case 3:
+                                            return Oh(e, [0, 0, t, 0], [e.shape[0], e.shape[1], n, e.shape[3]]);
+                                        case 4:
+                                            return cy(e, t, n);
+                                        default:
+                                            throw new og("The axis is not within the rank of the tensor " + r)
+                                    }
+                                    default:
+                                        throw new og("sliceAlongLastAxis() received an unsupported tensor rank: " + e.rank)
+                    }
+                })
+            }
+
+            function py(e, t = -1) {
+                let n;
+                return t < 0 && (n = e[0].rank, t = 0 !== n ? n : 0), t === e[0].rank && (t = -1), Xu(e, t)
+            }
+
+            function dy(e, t) {
+                switch (e.rank) {
+                    case 1:
+                        return hl([e, t]);
+                    case 2:
+                        return pl([e, t], 0);
+                    case 3:
+                        return dl([e, t], 0);
+                    case 4:
+                        return fl([e, t], 0);
+                    default:
+                        throw new og("concatAlongFirstAxis() received an unsupported tensor rank: " + e.rank)
+                }
+            }
+
+            function fy(e, t) {
+                if (Array.isArray(t) || (t = [t]), e.rank !== t.length) throw new og(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);
+                return Ul(e, t)
+            }
+
+            function my(e, t = 0, n = 1, r, a) {
+                return hh(e, t, n, r, a)
+            }
+
+            function gy(e, t, n, r) {
+                if (e.rank < 2 || t.rank < 2) throw new ug(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);
+                if (t.rank >= 3) {
+                    if (e.shape.slice(-1)[0] !== t.shape.slice(-2)[0]) throw new ug(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and  y shape = ${t.shape}`)
+                }
+                if (2 === e.rank && 2 === t.rank) {
+                    const a = !1,
+                        s = !1;
+                    return bp.matMul({
+                        a: e,
+                        b: t,
+                        transposeA: a,
+                        transposeB: s,
+                        bias: r ? vy(e.rank, r, "channelsLast") : null,
+                        activation: n
+                    })
+                } {
+                    const a = e.shape.slice(),
+                        s = a.pop();
+                    e = qu(e, [-1, s]);
+                    const i = t.shape.slice(),
+                        o = i.pop(),
+                        u = i.pop(),
+                        l = [...i, o],
+                        c = Array.from({
+                            length: t.rank
+                        }, (e, n) => 0 === n ? t.rank - 2 : n <= t.rank - 2 ? n - 1 : n);
+                    t = qu(qi(t, c), [u, -1]);
+                    const h = [...a, ...l],
+                        p = !1,
+                        d = !1;
+                    return qu(bp.matMul({
+                        a: e,
+                        b: t,
+                        transposeA: p,
+                        transposeB: d,
+                        bias: r ? vy(e.rank, r, "channelsLast") : null,
+                        activation: n
+                    }), h)
+                }
+            }
+
+            function yy(e, t, n) {
+                return ru(() => (t = Array.isArray(t) ? Yh(t, "int32") : Ai(t, "int32"), Gl(e, t, n)))
+            }
+
+            function by(e) {
+                return vu(e, e)
+            }
+
+            function vy(e, t, n) {
+                const r = t.shape;
+                if (1 !== t.rank && t.rank !== e) throw new og(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);
+                if (5 === e) {
+                    if ("channelsFirst" === n) return 1 === r.length ? qu(t, [1, r[0], 1, 1, 1]) : qu(t, [1, r[3], r[0], r[1], r[2]]);
+                    if ("channelsLast" === n) return 1 === r.length ? qu(t, [1, 1, 1, 1, r[0]]) : qu(t, [1].concat(r))
+                } else if (4 === e) {
+                    if ("channelsFirst" === n) return 1 === r.length ? qu(t, [1, r[0], 1, 1]) : qu(t, [1, r[2], r[0], r[1]]);
+                    if ("channelsLast" === n) return 1 === r.length ? qu(t, [1, 1, 1, r[0]]) : qu(t, [1].concat(r))
+                } else if (3 === e) {
+                    if ("channelsFirst" === n) return 1 === r.length ? qu(t, [1, r[0], 1]) : qu(t, [1, r[1], r[0]]);
+                    if ("channelsLast" === n) return 1 === r.length ? qu(t, [1, 1, r[0]]) : qu(t, [1].concat(r))
+                } else if (e < 3) return t;
+                throw new og("Unsupported input rank by biasAdd: " + t.rank)
+            }
+
+            function xy(e, t, n) {
+                return ru(() => (null == n && (n = "channelsLast"), Kg(n), gu(e, vy(e.rank, t, n))))
+            }
+
+            function wy(e, t, n, r) {
+                return ru(() => fp(e, t, n, r))
+            }
+
+            function ky(e, t, n = !1) {
+                return n ? e() : t()
+            }
+            var Sy = ["fanIn", "fanOut", "fanAvg"],
+                Iy = ["normal", "uniform", "truncatedNormal"];
+            var Ay = class extends Fo.Serializable {
+                    fromConfigUsesCustomObjects() {
+                        return !1
+                    }
+                    getConfig() {
+                        return {}
+                    }
+                },
+                Ey = class extends Ay {
+                    apply(e, t) {
+                        return Pc(e, t)
+                    }
+                };
+            Ey.className = "Zeros", Fo.registerClass(Ey);
+            var Ny = class extends Ay {
+                apply(e, t) {
+                    return $c(e, t)
+                }
+            };
+            Ny.className = "Ones", Fo.registerClass(Ny);
+            var Cy = class extends Ay {
+                constructor(e) {
+                    if (super(), "object" != typeof e) throw new og("Expected argument of type ConstantConfig but got " + e);
+                    if (void 0 === e.value) throw new og("config must have value set but got " + e);
+                    this.value = e.value
+                }
+                apply(e, t) {
+                    return ru(() => vu(Ah(this.value), $c(e, t)))
+                }
+                getConfig() {
+                    return {
+                        value: this.value
+                    }
+                }
+            };
+            Cy.className = "Constant", Fo.registerClass(Cy);
+            var Ty = class extends Ay {
+                constructor(e) {
+                    super(), this.DEFAULT_MINVAL = -.05, this.DEFAULT_MAXVAL = .05, this.minval = e.minval || this.DEFAULT_MINVAL, this.maxval = e.maxval || this.DEFAULT_MAXVAL, this.seed = e.seed
+                }
+                apply(e, t) {
+                    return ph(e, this.minval, this.maxval, t)
+                }
+                getConfig() {
+                    return {
+                        minval: this.minval,
+                        maxval: this.maxval,
+                        seed: this.seed
+                    }
+                }
+            };
+            Ty.className = "RandomUniform", Fo.registerClass(Ty);
+            var Ry = class extends Ay {
+                constructor(e) {
+                    super(), this.DEFAULT_MEAN = 0, this.DEFAULT_STDDEV = .05, this.mean = e.mean || this.DEFAULT_MEAN, this.stddev = e.stddev || this.DEFAULT_STDDEV, this.seed = e.seed
+                }
+                apply(e, t) {
+                    if ("float32" !== (t = t || "float32") && "int32" !== t) throw new ug(`randomNormal does not support dType ${t}.`);
+                    return my(e, this.mean, this.stddev, t, this.seed)
+                }
+                getConfig() {
+                    return {
+                        mean: this.mean,
+                        stddev: this.stddev,
+                        seed: this.seed
+                    }
+                }
+            };
+            Ry.className = "RandomNormal", Fo.registerClass(Ry);
+            var _y = class extends Ay {
+                constructor(e) {
+                    super(), this.DEFAULT_MEAN = 0, this.DEFAULT_STDDEV = .05, this.mean = e.mean || this.DEFAULT_MEAN, this.stddev = e.stddev || this.DEFAULT_STDDEV, this.seed = e.seed
+                }
+                apply(e, t) {
+                    if ("float32" !== (t = t || "float32") && "int32" !== t) throw new ug(`truncatedNormal does not support dType ${t}.`);
+                    return tp(e, this.mean, this.stddev, t, this.seed)
+                }
+                getConfig() {
+                    return {
+                        mean: this.mean,
+                        stddev: this.stddev,
+                        seed: this.seed
+                    }
+                }
+            };
+            _y.className = "TruncatedNormal", Fo.registerClass(_y);
+            var Fy = class extends Ay {
+                constructor(e) {
+                    super(), this.gain = null != e.gain ? e.gain : 1
+                }
+                apply(e, t) {
+                    return ru(() => {
+                        if (2 !== e.length || e[0] !== e[1]) throw new og("Identity matrix initializer can only be used for 2D square matrices.");
+                        return vu(this.gain, Wl(e[0]))
+                    })
+                }
+                getConfig() {
+                    return {
+                        gain: this.gain
+                    }
+                }
+            };
+            Fy.className = "Identity", Fo.registerClass(Fy);
+            var My = class extends Ay {
+                constructor(e) {
+                    if (super(), e.scale < 0) throw new og("scale must be a positive float. Got: " + e.scale);
+                    var t;
+                    this.scale = null == e.scale ? 1 : e.scale, this.mode = null == e.mode ? "fanIn" : e.mode, t = this.mode, Sg(Sy, "FanMode", t), this.distribution = null == e.distribution ? "normal" : e.distribution,
+                        function (e) {
+                            Sg(Iy, "Distribution", e)
+                        }(this.distribution), this.seed = e.seed
+                }
+                apply(e, t) {
+                    const n = function (e, t = "channelsLast") {
+                            let n, r;
+                            if (Kg(t), 2 === e.length) n = e[0], r = e[1];
+                            else if (-1 !== [3, 4, 5].indexOf(e.length)) {
+                                if ("channelsFirst" === t) {
+                                    const t = ry(e, 2);
+                                    n = e[1] * t, r = e[0] * t
+                                } else if ("channelsLast" === t) {
+                                    const t = ry(e, 0, e.length - 2);
+                                    n = e[e.length - 2] * t, r = e[e.length - 1] * t
+                                }
+                            } else {
+                                const t = ry(e);
+                                n = Math.sqrt(t), r = Math.sqrt(t)
+                            }
+                            return [n, r]
+                        }(e),
+                        r = n[0],
+                        a = n[1];
+                    let s = this.scale;
+                    if ("fanIn" === this.mode ? s /= Math.max(1, r) : "fanOut" === this.mode ? s /= Math.max(1, a) : s /= Math.max(1, (r + a) / 2), "normal" === this.distribution) {
+                        const n = Math.sqrt(s);
+                        if ("float32" !== (t = t || "float32") && "int32" !== t) throw new ug(`${this.getClassName()} does not support dType ${t}.`);
+                        return tp(e, 0, n, t, this.seed)
+                    } {
+                        const n = Math.sqrt(3 * s);
+                        return ph(e, -n, n, t)
+                    }
+                }
+                getConfig() {
+                    return {
+                        scale: this.scale,
+                        mode: this.mode,
+                        distribution: this.distribution,
+                        seed: this.seed
+                    }
+                }
+            };
+            My.className = "VarianceScaling", Fo.registerClass(My);
+            var Dy = class extends My {
+                constructor(e) {
+                    super({
+                        scale: 1,
+                        mode: "fanAvg",
+                        distribution: "uniform",
+                        seed: null == e ? null : e.seed
+                    })
+                }
+                getClassName() {
+                    return My.className
+                }
+            };
+            Dy.className = "GlorotUniform", Fo.registerClass(Dy);
+            var Oy = class extends My {
+                constructor(e) {
+                    super({
+                        scale: 1,
+                        mode: "fanAvg",
+                        distribution: "normal",
+                        seed: null == e ? null : e.seed
+                    })
+                }
+                getClassName() {
+                    return My.className
+                }
+            };
+            Oy.className = "GlorotNormal", Fo.registerClass(Oy);
+            var Ly = class extends My {
+                constructor(e) {
+                    super({
+                        scale: 2,
+                        mode: "fanIn",
+                        distribution: "normal",
+                        seed: null == e ? null : e.seed
+                    })
+                }
+                getClassName() {
+                    return My.className
+                }
+            };
+            Ly.className = "HeNormal", Fo.registerClass(Ly);
+            var Py = class extends My {
+                constructor(e) {
+                    super({
+                        scale: 2,
+                        mode: "fanIn",
+                        distribution: "uniform",
+                        seed: null == e ? null : e.seed
+                    })
+                }
+                getClassName() {
+                    return My.className
+                }
+            };
+            Py.className = "HeUniform", Fo.registerClass(Py);
+            var $y = class extends My {
+                constructor(e) {
+                    super({
+                        scale: 1,
+                        mode: "fanIn",
+                        distribution: "normal",
+                        seed: null == e ? null : e.seed
+                    })
+                }
+                getClassName() {
+                    return My.className
+                }
+            };
+            $y.className = "LeCunNormal", Fo.registerClass($y);
+            var By = class extends My {
+                constructor(e) {
+                    super({
+                        scale: 1,
+                        mode: "fanIn",
+                        distribution: "uniform",
+                        seed: null == e ? null : e.seed
+                    })
+                }
+                getClassName() {
+                    return My.className
+                }
+            };
+            By.className = "LeCunNormal", Fo.registerClass(By);
+            var zy = class extends Ay {
+                constructor(e) {
+                    if (super(), this.DEFAULT_GAIN = 1, this.gain = null == e.gain ? this.DEFAULT_GAIN : e.gain, this.seed = e.seed, null != this.seed) throw new ug("Random seed is not implemented for Orthogonal Initializer yet.")
+                }
+                apply(e, t) {
+                    return ru(() => {
+                        if (e.length < 2) throw new ug("Shape must be at least 2D.");
+                        e[0] * e[1] > 2e3 && console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${e[0]*e[1]}) elements: Slowness may result.`);
+                        const t = my(e[0] > e[1] ? [e[1], e[0]] : e, 0, 1, "float32");
+                        let n = Cd.gramSchmidt(t);
+                        return e[0] > e[1] && (n = qi(n)), vu(this.gain, n)
+                    })
+                }
+                getConfig() {
+                    return {
+                        gain: this.gain,
+                        seed: this.seed
+                    }
+                }
+            };
+            zy.className = "Orthogonal", Fo.registerClass(zy);
+            var Uy = {
+                constant: "Constant",
+                glorotNormal: "GlorotNormal",
+                glorotUniform: "GlorotUniform",
+                heNormal: "HeNormal",
+                heUniform: "HeUniform",
+                identity: "Identity",
+                leCunNormal: "LeCunNormal",
+                leCunUniform: "LeCunUniform",
+                ones: "Ones",
+                orthogonal: "Orthogonal",
+                randomNormal: "RandomNormal",
+                randomUniform: "RandomUniform",
+                truncatedNormal: "TruncatedNormal",
+                varianceScaling: "VarianceScaling",
+                zeros: "Zeros"
+            };
+
+            function Wy(e, t = {}) {
+                return vg(e, Fo.SerializationMap.getMap().classNameMap, t, "initializer")
+            }
+
+            function Vy(e) {
+                return bg(e)
+            }
+
+            function jy(e) {
+                if ("string" == typeof e) {
+                    const t = e in Uy ? Uy[e] : e;
+                    if ("GlorotNormal" === t) return new Oy;
+                    if ("GlorotUniform" === t) return new Dy;
+                    if ("HeNormal" === t) return new Ly;
+                    if ("HeUniform" === t) return new Py;
+                    if ("LeCunNormal" === t) return new $y;
+                    if ("LeCunUniform" === t) return new By; {
+                        const e = {};
+                        return e.className = t, e.config = {}, Wy(e)
+                    }
+                }
+                return e instanceof Ay ? e : Wy(e)
+            }
+
+            function Gy() {
+                return new Ey
+            }
+
+            function Hy() {
+                return new Ny
+            }
+
+            function qy(e) {
+                return new Cy(e)
+            }
+
+            function Ky(e) {
+                return new Ty(e)
+            }
+
+            function Yy(e) {
+                return new Ry(e)
+            }
+
+            function Xy(e) {
+                return new _y(e)
+            }
+
+            function Qy(e) {
+                return new Fy(e)
+            }
+
+            function Zy(e) {
+                return new My(e)
+            }
+
+            function Jy(e) {
+                return new Dy(e)
+            }
+
+            function eb(e) {
+                return new Oy(e)
+            }
+
+            function tb(e) {
+                return new Ly(e)
+            }
+
+            function nb(e) {
+                return new Py(e)
+            }
+
+            function rb(e) {
+                return new $y(e)
+            }
+
+            function ab(e) {
+                return new By(e)
+            }
+
+            function sb(e) {
+                return new zy(e)
+            }
+            var ib = {};
+            w(ib, {
+                Layer: () => Ib,
+                RNN: () => Yx,
+                RNNCell: () => Xx,
+                activation: () => yk,
+                add: () => Ek,
+                alphaDropout: () => gS,
+                average: () => Nk,
+                averagePooling1d: () => Lk,
+                averagePooling2d: () => Bk,
+                averagePooling3d: () => Wk,
+                avgPool1d: () => Pk,
+                avgPool2d: () => zk,
+                avgPool3d: () => Vk,
+                avgPooling1d: () => $k,
+                avgPooling2d: () => Uk,
+                avgPooling3d: () => jk,
+                batchNormalization: () => Mk,
+                bidirectional: () => uS,
+                concatenate: () => Ck,
+                conv1d: () => uk,
+                conv2d: () => lk,
+                conv2dTranspose: () => ck,
+                conv3d: () => hk,
+                conv3dTranspose: () => pk,
+                convLstm2d: () => aS,
+                convLstm2dCell: () => sS,
+                cropping2D: () => fk,
+                dense: () => bk,
+                depthwiseConv2d: () => gk,
+                dot: () => Fk,
+                dropout: () => vk,
+                elu: () => nk,
+                embedding: () => Ak,
+                flatten: () => wk,
+                gaussianDropout: () => mS,
+                gaussianNoise: () => fS,
+                globalAveragePooling1d: () => Gk,
+                globalAveragePooling2d: () => Hk,
+                globalMaxPool1d: () => cS,
+                globalMaxPool2d: () => hS,
+                globalMaxPooling1d: () => qk,
+                globalMaxPooling2d: () => Kk,
+                gru: () => Zk,
+                gruCell: () => Jk,
+                input: () => Xv,
+                inputLayer: () => tk,
+                layerNormalization: () => Dk,
+                leakyReLU: () => ak,
+                lstm: () => eS,
+                lstmCell: () => tS,
+                masking: () => yS,
+                maxPool1d: () => pS,
+                maxPool2d: () => dS,
+                maxPooling1d: () => Yk,
+                maxPooling2d: () => Xk,
+                maxPooling3d: () => Qk,
+                maximum: () => Tk,
+                minimum: () => Rk,
+                multiply: () => _k,
+                permute: () => Ik,
+                prelu: () => sk,
+                reLU: () => rk,
+                repeatVector: () => kk,
+                reshape: () => Sk,
+                rnn: () => iS,
+                separableConv2d: () => dk,
+                simpleRNN: () => nS,
+                simpleRNNCell: () => rS,
+                softmax: () => ik,
+                spatialDropout1d: () => xk,
+                stackedRNNCells: () => oS,
+                thresholdedReLU: () => ok,
+                timeDistributed: () => lS,
+                upSampling2d: () => mk,
+                zeroPadding2d: () => Ok
+            });
+            var ob = 0;
+
+            function ub() {
+                return ob++
+            }
+            var lb = {};
+
+            function cb(e = "") {
+                return e in lb || (lb[e] = 0), lb[e] += 1, e + lb[e].toString()
+            }
+
+            function hb(e) {
+                return Array.isArray(e) && Array.isArray(e[0])
+            }
+
+            function pb(e) {
+                return 0 === e.length ? [] : Array.isArray(e[0]) ? e : [e]
+            }
+
+            function db(e) {
+                let t;
+                if (Array.isArray(e)) {
+                    if (1 !== e.length) throw new og("Expected Tensor length to be 1; got " + e.length);
+                    t = e[0]
+                } else t = e;
+                return t
+            }
+
+            function fb(e) {
+                if (Array.isArray(e) && Array.isArray(e[0])) {
+                    if (1 === e.length) return (e = e)[0];
+                    throw new og("Expected exactly 1 Shape; got " + e.length)
+                }
+                return e
+            }
+
+            function mb(e) {
+                let t = 0;
+                for (const n of e) 0 === n.shape.length ? t += 1 : t += n.shape.reduce((e, t) => e * t);
+                return t
+            }
+            var gb = class {
+                constructor(e, t = "float32", n = "Variable", r = !0, a = null) {
+                    this.dtype = null == t ? "float32" : t, this.shape = e.shape, this.id = ub(), n = null == n ? "Variable" : n, this.originalName = Jg(n), this.name = ey(this.originalName), this.trainable_ = r, this.constraint = a, this.val = sp(e, this.trainable_, this.name, this.dtype)
+                }
+                read() {
+                    return this.assertNotDisposed(), this.val
+                }
+                write(e) {
+                    return this.assertNotDisposed(),
+                        function (e, t) {
+                            if (e.shape.toString() !== t.shape.toString()) throw new Error("Shape mismatch: " + JSON.stringify(e.shape) + " vs. " + JSON.stringify(t.shape))
+                        }(this.val, e), this.val.id !== e.id && (this.val.assign(e), null != this.constraint && this.val.assign(this.constraint.apply(this.val))), this
+                }
+                dispose() {
+                    this.assertNotDisposed(), this.val.dispose()
+                }
+                assertNotDisposed() {
+                    if (this.val.isDisposed) throw new Error(`LayersVariable ${this.name} is already disposed.`)
+                }
+                get trainable() {
+                    return this.trainable_
+                }
+                set trainable(e) {
+                    this.trainable_ = e, this.val.trainable = e
+                }
+            };
+
+            function yb(e) {
+                return e.map(e => e.read())
+            }
+
+            function bb(e) {
+                e.forEach(e => {
+                    e[0].write(e[1])
+                })
+            }
+            var vb = class {
+                    constructor(e) {
+                        this.dtype = e.dtype, this.shape = e.shape, null != e.shape ? this.ndim = e.shape.length : this.ndim = e.ndim, this.maxNDim = e.maxNDim, this.minNDim = e.minNDim, this.axes = e.axes || {}
+                    }
+                },
+                xb = class {
+                    constructor(e, t, n, r, a, s, i) {
+                        this.dtype = e, this.shape = t, this.sourceLayer = n, this.inputs = r, this.callArgs = a, this.outputTensorIndex = i, this.id = ub(), null != s && (this.originalName = Jg(s), this.name = ey(this.originalName)), this.rank = t.length
+                    }
+                },
+                wb = 0,
+                kb = class {
+                    constructor(e, t) {
+                        this.callArgs = t, this.id = wb++, this.outboundLayer = e.outboundLayer, this.inboundLayers = e.inboundLayers, this.nodeIndices = e.nodeIndices, this.tensorIndices = e.tensorIndices, this.inputTensors = e.inputTensors, this.outputTensors = e.outputTensors, this.inputMasks = e.inputMasks, this.outputMasks = e.outputMasks, this.inputShapes = e.inputShapes, this.outputShapes = e.outputShapes;
+                        for (const t of e.inboundLayers) null != t && t.outboundNodes.push(this);
+                        e.outboundLayer.inboundNodes.push(this)
+                    }
+                    getConfig() {
+                        const e = [];
+                        for (const t of this.inboundLayers) null != t ? e.push(t.name) : e.push(null);
+                        return {
+                            outboundLayer: this.outboundLayer ? this.outboundLayer.name : null,
+                            inboundLayers: e,
+                            nodeIndices: this.nodeIndices,
+                            tensorIndices: this.tensorIndices
+                        }
+                    }
+                },
+                Sb = 0,
+                Ib = class extends Fo.Serializable {
+                    constructor(e = {}) {
+                        super(), this._callHook = null, this._addedWeightNames = [], this._stateful = !1, this.id = Sb++, this.activityRegularizer = null, this.inputSpec = null, this.supportsMasking = !1, this._trainableWeights = [], this._nonTrainableWeights = [], this._losses = [], this._updates = [], this._built = !1, this.inboundNodes = [], this.outboundNodes = [];
+                        let t = e.name;
+                        if (!t) {
+                            const e = this.getClassName();
+                            t = mg(e) + "_" + cb(e)
+                        }
+                        if (this.name = t, this.trainable_ = null == e.trainable || e.trainable, null != e.inputShape || null != e.batchInputShape) {
+                            let t;
+                            if (null != e.batchInputShape) t = e.batchInputShape;
+                            else if (null != e.inputShape) {
+                                let n = null;
+                                null != e.batchSize && (n = e.batchSize), t = [n].concat(e.inputShape)
+                            }
+                            this.batchInputShape = t;
+                            let n = e.dtype;
+                            null == n && (n = e.inputDType), null == n && (n = "float32"), this.dtype = n
+                        }
+                        null != e.weights ? this.initialWeights = e.weights : this.initialWeights = null, this._refCount = null, this.fastWeightInitDuringBuild = !1
+                    }
+                    static nodeKey(e, t) {
+                        return e.name + "_ib-" + t.toString()
+                    }
+                    getNodeAtIndex(e, t) {
+                        if (0 === this.inboundNodes.length) throw new ig(`The layer has never been called and thus has no defined ${t}.`);
+                        if (this.inboundNodes.length <= e) throw new og(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);
+                        return this.inboundNodes[e]
+                    }
+                    getInputAt(e) {
+                        return dg(this.getNodeAtIndex(e, "input").inputTensors)
+                    }
+                    getOutputAt(e) {
+                        return dg(this.getNodeAtIndex(e, "output").outputTensors)
+                    }
+                    get input() {
+                        if (this.inboundNodes.length > 1) throw new sg(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);
+                        if (0 === this.inboundNodes.length) throw new sg(`Layer ${this.name} is not connected, no input to return.`);
+                        return dg(this.getNodeAtIndex(0, "input").inputTensors)
+                    }
+                    get output() {
+                        if (0 === this.inboundNodes.length) throw new sg(`Layer ${this.name} has no inbound nodes.`);
+                        if (this.inboundNodes.length > 1) throw new sg(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);
+                        return dg(this.getNodeAtIndex(0, "output").outputTensors)
+                    }
+                    get losses() {
+                        return this._losses
+                    }
+                    calculateLosses() {
+                        return this.losses.map(e => e())
+                    }
+                    get updates() {
+                        return this._updates
+                    }
+                    get built() {
+                        return this._built
+                    }
+                    set built(e) {
+                        this._built = e
+                    }
+                    get trainable() {
+                        return this.trainable_
+                    }
+                    set trainable(e) {
+                        this._trainableWeights.forEach(t => t.trainable = e), this.trainable_ = e
+                    }
+                    get trainableWeights() {
+                        return this.trainable_ ? this._trainableWeights.filter(e => e.trainable) : []
+                    }
+                    set trainableWeights(e) {
+                        this._trainableWeights = e
+                    }
+                    get nonTrainableWeights() {
+                        return this.trainable ? this._trainableWeights.filter(e => !e.trainable).concat(this._nonTrainableWeights) : this._trainableWeights.concat(this._nonTrainableWeights)
+                    }
+                    set nonTrainableWeights(e) {
+                        this._nonTrainableWeights = e
+                    }
+                    get weights() {
+                        return this.trainableWeights.concat(this.nonTrainableWeights)
+                    }
+                    get stateful() {
+                        return this._stateful
+                    }
+                    resetStates() {
+                        if (!this.stateful) throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")
+                    }
+                    assertInputCompatibility(e) {
+                        if (e = fg(e), null == this.inputSpec || 0 === this.inputSpec.length) return;
+                        const t = fg(this.inputSpec);
+                        if (e.length !== t.length) throw new og(`Layer ${this.name} expects ${t.length} inputs, but it received ${e.length} input tensors. Input received: ${e}`);
+                        for (let n = 0; n < e.length; n++) {
+                            const r = e[n],
+                                a = t[n];
+                            if (null == a) continue;
+                            const s = r.rank;
+                            if (null != a.ndim && s !== a.ndim) throw new og(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${a.ndim}, found ndim=${s}`);
+                            if (null != a.maxNDim && s > a.maxNDim) throw new og(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${a.maxNDim}, found ndim=${s}`);
+                            if (null != a.minNDim && s < a.minNDim) throw new og(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${a.minNDim}, found ndim=${s}.`);
+                            if (null != a.dtype && r.dtype !== a.dtype) throw new og(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${a.dtype}, found dtype=${r.dtype}.`);
+                            if (a.axes) {
+                                const e = r.shape;
+                                for (const t in a.axes) {
+                                    const r = Number(t),
+                                        s = a.axes[t],
+                                        i = r >= 0 ? e[r] : e[e.length + r];
+                                    if (null != s && -1 === [s, null].indexOf(i)) throw new og(`Input ${n} is incompatible with layer ${this.name}: expected axis ${r} of input shape to have value ${s} but got shape ${e}.`)
+                                }
+                            }
+                            if (null != a.shape)
+                                for (let e = 0; e < a.shape.length; ++e) {
+                                    const t = a.shape[e],
+                                        s = r.shape[e];
+                                    if (null != t && null != s && t !== s) throw new og(`Input ${n} is incompatible with layer ${this.name}: expected shape=${a.shape}, found shape=${r.shape}.`)
+                                }
+                        }
+                    }
+                    call(e, t) {
+                        return e
+                    }
+                    invokeCallHook(e, t) {
+                        null != this._callHook && this._callHook(e, t)
+                    }
+                    setCallHook(e) {
+                        this._callHook = e
+                    }
+                    clearCallHook() {
+                        this._callHook = null
+                    }
+                    apply(e, t) {
+                        t = t || {}, this.assertNotDisposed();
+                        const n = fg(e);
+                        let r = !0;
+                        for (const e of n)
+                            if (!(e instanceof xb)) {
+                                r = !1;
+                                break
+                            } let a = !0;
+                        for (const e of n)
+                            if (e instanceof xb) {
+                                a = !1;
+                                break
+                            } if (r === a) throw new og("Arguments to apply() must be all SymbolicTensors or all Tensors");
+                        return Zg(this.name, () => {
+                            if (!this.built) {
+                                this.assertInputCompatibility(e);
+                                const t = [];
+                                for (const n of fg(e)) t.push(n.shape);
+                                this.build(dg(t)), this.built = !0, this.initialWeights && this.setWeights(this.initialWeights), null === this._refCount && a && (this._refCount = 1)
+                            }
+                            if (this.assertInputCompatibility(e), a) {
+                                let r = this.call(e, t);
+                                const a = fg(r),
+                                    s = [];
+                                for (let e of a) - 1 !== n.indexOf(e) && (e = e.clone()), s.push(e);
+                                if (r = dg(s), null != this.activityRegularizer) throw new ug("Layer invocation in the presence of activity regularizer(s) is not supported yet.");
+                                return r
+                            } {
+                                const n = function (e) {
+                                        e = fg(e);
+                                        const t = [];
+                                        for (const n of e) t.push(n.shape);
+                                        return dg(t)
+                                    }(e),
+                                    r = this.computeOutputShape(n);
+                                let a;
+                                const s = "float32";
+                                if (this.warnOnIncompatibleInputShape(Array.isArray(e) ? n[0] : n), a = null != r && r.length > 0 && Array.isArray(r[0]) ? r.map((n, r) => new xb(s, n, this, fg(e), t, this.name, r)) : new xb(s, r, this, fg(e), t, this.name), this.addInboundNode(e, a, null, null, n, r, t), this._refCount++, null != this.activityRegularizer) throw new ug("Layer invocation in the presence of activity regularizer(s) is not supported yet.");
+                                return a
+                            }
+                        })
+                    }
+                    warnOnIncompatibleInputShape(e) {
+                        if (null != this.batchInputShape)
+                            if (e.length !== this.batchInputShape.length) console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);
+                            else {
+                                let t = !1;
+                                this.batchInputShape.forEach((n, r) => {
+                                    null != n && null != e[r] && e[r] !== n && (t = !0)
+                                }), t && console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)
+                            }
+                    }
+                    get outputShape() {
+                        if (null == this.inboundNodes || 0 === this.inboundNodes.length) throw new sg(`The layer ${this.name} has never been called and thus has no defined output shape.`);
+                        const e = [];
+                        for (const t of this.inboundNodes) {
+                            const n = JSON.stringify(t.outputShapes); - 1 === e.indexOf(n) && e.push(n)
+                        }
+                        if (1 === e.length) {
+                            const e = this.inboundNodes[0].outputShapes;
+                            return Array.isArray(e) && Array.isArray(e[0]) && 1 === e.length ? e[0] : e
+                        }
+                        throw new sg(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)
+                    }
+                    countParams() {
+                        if (!this.built) throw new ig(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);
+                        return mb(this.weights)
+                    }
+                    build(e) {
+                        this.built = !0
+                    }
+                    getWeights(e = !1) {
+                        return yb(e ? this.trainableWeights : this.weights)
+                    }
+                    setWeights(e) {
+                        ru(() => {
+                            const t = this.weights;
+                            if (t.length !== e.length) throw new og(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);
+                            if (0 === t.length) return;
+                            const n = [],
+                                r = yb(t);
+                            for (let a = 0; a < r.length; ++a) {
+                                const s = r[a],
+                                    i = t[a],
+                                    o = e[a];
+                                if (!va.arraysEqual(s.shape, o.shape)) throw new og(`Layer weight shape ${s.shape} not compatible with provided weight shape ${o.shape}`);
+                                n.push([i, o])
+                            }
+                            bb(n)
+                        })
+                    }
+                    addWeight(e, t, n, r, a, s, i, o) {
+                        if (-1 !== this._addedWeightNames.indexOf(e)) throw new og(`Duplicate weight name ${e} for layer ${this.name}`);
+                        this._addedWeightNames.push(e), null == n && (n = "float32"), this.fastWeightInitDuringBuild && (r = null != o ? o() : jy("zeros"));
+                        const u = r.apply(t, n),
+                            l = new gb(u, n, e, s, i);
+                        return u.dispose(), null != a && this.addLoss(() => a.apply(l.read())), null == s && (s = !0), s ? this._trainableWeights.push(l) : this._nonTrainableWeights.push(l), l
+                    }
+                    setFastWeightInitDuringBuild(e) {
+                        this.fastWeightInitDuringBuild = e
+                    }
+                    addLoss(e) {
+                        null == e || Array.isArray(e) && 0 === e.length || (e = fg(e), void 0 !== this._losses && null !== this._losses && this.losses.push(...e))
+                    }
+                    computeOutputShape(e) {
+                        return e
+                    }
+                    computeMask(e, t) {
+                        if (!this.supportsMasking) {
+                            if (null != t) {
+                                if (!Array.isArray(t)) throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);
+                                t.forEach(e => {
+                                    if (null != e) throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)
+                                })
+                            }
+                            return null
+                        }
+                        return t
+                    }
+                    addInboundNode(e, t, n, r, a, s, i = null) {
+                        const o = fg(e);
+                        t = fg(t), n = fg(n), r = fg(r), a = pb(a), s = pb(s);
+                        const u = [],
+                            l = [],
+                            c = [];
+                        for (const e of o) u.push(e.sourceLayer), l.push(e.nodeIndex), c.push(e.tensorIndex);
+                        new kb({
+                            outboundLayer: this,
+                            inboundLayers: u,
+                            nodeIndices: l,
+                            tensorIndices: c,
+                            inputTensors: o,
+                            outputTensors: t,
+                            inputMasks: n,
+                            outputMasks: r,
+                            inputShapes: a,
+                            outputShapes: s
+                        }, i);
+                        for (let e = 0; e < t.length; e++) t[e].sourceLayer = this, t[e].nodeIndex = this.inboundNodes.length - 1, t[e].tensorIndex = e
+                    }
+                    getConfig() {
+                        const e = {
+                            name: this.name,
+                            trainable: this.trainable
+                        };
+                        return null != this.batchInputShape && (e.batchInputShape = this.batchInputShape), null != this.dtype && (e.dtype = this.dtype), e
+                    }
+                    disposeWeights() {
+                        return this.weights.forEach(e => e.dispose()), this.weights.length
+                    }
+                    assertNotDisposed() {
+                        if (0 === this._refCount) throw new Error(`Layer '${this.name}' is already disposed.`)
+                    }
+                    dispose() {
+                        if (!this.built) throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);
+                        if (null === this._refCount) throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);
+                        this.assertNotDisposed();
+                        let e = 0;
+                        return 0 == --this._refCount && (e = this.disposeWeights()), {
+                            refCountAfterDispose: this._refCount,
+                            numDisposedVariables: e
+                        }
+                    }
+                };
+            var Ab, Eb, Nb = class extends Ib {
+                constructor(e) {
+                    if (super({
+                            dtype: e.dtype,
+                            name: null != e.name ? e.name : cb("input").toString()
+                        }), null == e.batchSize && (e.batchSize = null), null == e.sparse && (e.sparse = !1), this.trainable = !1, this.built = !0, this.sparse = e.sparse, null != e.inputShape && null != e.batchInputShape) throw new og("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");
+                    let t = e.batchInputShape;
+                    if (null == t) {
+                        if (null == e.inputShape) throw new og("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");
+                        t = [e.batchSize].concat(e.inputShape)
+                    } else if (null != e.batchSize) throw new og("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");
+                    const n = e.dtype || "float32";
+                    this.batchInputShape = t, this.dtype = n, this.inputSpec = [{
+                        shape: t
+                    }];
+                    const r = new xb(this.dtype, this.batchInputShape, this, [], {}, this.name);
+                    r.nodeIndex = 0, r.tensorIndex = 0, new kb({
+                        outboundLayer: this,
+                        inboundLayers: [],
+                        nodeIndices: [],
+                        tensorIndices: [],
+                        inputTensors: [r],
+                        outputTensors: [r],
+                        inputMasks: [null],
+                        outputMasks: [null],
+                        inputShapes: [t],
+                        outputShapes: [t]
+                    })
+                }
+                apply(e, t) {
+                    throw new og("Cannot pass any input to an InputLayer's apply() method. InputLayer name: " + this.name)
+                }
+                dispose() {
+                    return {
+                        refCountAfterDispose: this._refCount,
+                        numDisposedVariables: 0
+                    }
+                }
+                getConfig() {
+                    return {
+                        batchInputShape: this.batchInputShape,
+                        dtype: this.dtype,
+                        sparse: this.sparse,
+                        name: this.name
+                    }
+                }
+            };
+
+            function Cb(e) {
+                if (null == e.batchShape && null == e.shape) throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");
+                if (null != e.batchShape && null != e.shape) throw new og("Please provide either a `shape` or `batchShape` argument to Input, but not both.");
+                let t = e.batchShape;
+                null != e.shape && null == t && (t = [null].concat(e.shape));
+                let n = e.dtype;
+                null == n && (n = "float32");
+                return new Nb({
+                    batchInputShape: t,
+                    name: e.name,
+                    dtype: n,
+                    sparse: e.sparse
+                }).inboundNodes[0].outputTensors[0]
+            }
+            async function Tb(e) {
+                if (null == e) return;
+                const t = [],
+                    n = [],
+                    r = [];
+                for (const a in e) {
+                    const s = e[a];
+                    if ("number" != typeof s) {
+                        const e = s;
+                        t.push(e.data()), n.push(a), r.push(e)
+                    }
+                }
+                if (t.length > 0) {
+                    const a = await Promise.all(t);
+                    for (let t = 0; t < a.length; ++t) e[n[t]] = a[t][0];
+                    au(r)
+                }
+            }
+
+            function Rb(e) {
+                if (null != e)
+                    for (const t in e) {
+                        const n = e[t];
+                        "number" != typeof n && n.dispose()
+                    }
+            }
+            Nb.className = "InputLayer", Fo.registerClass(Nb), (Eb = Ab || (Ab = {}))[Eb.SILENT = 0] = "SILENT", Eb[Eb.VERBOSE = 1] = "VERBOSE";
+            var _b = class {
+                    constructor() {
+                        this.validationData = null
+                    }
+                    setParams(e) {
+                        this.params = e
+                    }
+                    async onEpochBegin(e, t) {}
+                    async onEpochEnd(e, t) {}
+                    async onBatchBegin(e, t) {}
+                    async onBatchEnd(e, t) {}
+                    async onTrainBegin(e) {}
+                    async onTrainEnd(e) {}
+                    setModel(e) {}
+                },
+                Fb = class {
+                    constructor(e, t = 10) {
+                        null == e && (e = []), this.callbacks = e, this.queueLength = t
+                    }
+                    append(e) {
+                        this.callbacks.push(e)
+                    }
+                    setParams(e) {
+                        for (const t of this.callbacks) t.setParams(e)
+                    }
+                    setModel(e) {
+                        for (const t of this.callbacks) t.setModel(e)
+                    }
+                    async onEpochBegin(e, t) {
+                        null == t && (t = {});
+                        for (const n of this.callbacks) await n.onEpochBegin(e, t)
+                    }
+                    async onEpochEnd(e, t) {
+                        null == t && (t = {});
+                        for (const n of this.callbacks) await n.onEpochEnd(e, t)
+                    }
+                    async onBatchBegin(e, t) {
+                        null == t && (t = {});
+                        for (const n of this.callbacks) await n.onBatchBegin(e, t)
+                    }
+                    async onBatchEnd(e, t) {
+                        null == t && (t = {});
+                        for (const n of this.callbacks) await n.onBatchEnd(e, t)
+                    }
+                    async onTrainBegin(e) {
+                        null == e && (e = {});
+                        for (const t of this.callbacks) await t.onTrainBegin(e)
+                    }
+                    async onTrainEnd(e) {
+                        null == e && (e = {});
+                        for (const t of this.callbacks) await t.onTrainEnd(e)
+                    }
+                },
+                Mb = class extends _b {
+                    constructor() {
+                        super()
+                    }
+                    async onEpochBegin(e) {
+                        this.seen = 0, this.totals = {}
+                    }
+                    async onBatchEnd(e, t) {
+                        null == t && (t = {});
+                        const n = null == t.size ? 0 : t.size;
+                        this.seen += n;
+                        for (const e in t) {
+                            const r = t[e];
+                            if ("number" == typeof r) this.totals.hasOwnProperty(e) || (this.totals[e] = 0), this.totals[e] = this.totals[e] + r * n;
+                            else {
+                                let t;
+                                e in this.totals ? t = this.totals[e] : this.totals[e] = 0;
+                                const a = ru(() => gu(this.totals[e], vu(r, n)));
+                                this.totals[e] = a, null != t && t.dispose()
+                            }
+                        }
+                    }
+                    async onEpochEnd(e, t) {
+                        if (null != t)
+                            for (const e of this.params.metrics) null != this.totals[e] && ("number" == typeof this.totals[e] ? t[e] = this.totals[e] / this.seen : ru(() => {
+                                const n = vu(bu(1, this.seen), this.totals[e]);
+                                t[e] = n, this.totals[e].dispose(), su(t[e])
+                            }))
+                    }
+                },
+                Db = class extends _b {
+                    async onTrainBegin(e) {
+                        this.epoch = [], this.history = {}
+                    }
+                    async onEpochEnd(e, t) {
+                        null == t && (t = {}), this.epoch.push(e);
+                        for (const e in t) null == this.history[e] && (this.history[e] = []), this.history[e].push(t[e])
+                    }
+                    async syncData() {
+                        const e = [],
+                            t = [],
+                            n = [];
+                        for (const r in this.history) {
+                            const a = this.history[r];
+                            for (let s = 0; s < a.length; ++s)
+                                if ("number" != typeof a[s]) {
+                                    const i = a[s];
+                                    e.push(i.data()), t.push(r), n.push(s)
+                                }
+                        }
+                        const r = await Promise.all(e);
+                        for (let e = 0; e < r.length; ++e) {
+                            this.history[t[e]][n[e]].dispose(), this.history[t[e]][n[e]] = r[e][0]
+                        }
+                    }
+                },
+                Ob = class extends _b {
+                    constructor(e, t) {
+                        if (super(), this.currentEpoch = 0, this.nowFunc = e.nowFunc, this.nextFrameFunc = e.nextFrameFunc || Vd, this.yieldEvery = t || "auto", "auto" === this.yieldEvery && (this.yieldEvery = 125), "never" === this.yieldEvery && null != e.onYield) throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");
+                        va.isNumber(this.yieldEvery) && (this.maybeWait = function (e, t, n) {
+                            let r, a = null != n ? n() : va.now();
+                            return (...s) => {
+                                const i = null != n ? n() : va.now();
+                                return i - a < t || (a = i, r = e(...s)), r
+                            }
+                        }(this.maybeWait.bind(this), this.yieldEvery, this.nowFunc)), this.trainBegin = e.onTrainBegin, this.trainEnd = e.onTrainEnd, this.epochBegin = e.onEpochBegin, this.epochEnd = e.onEpochEnd, this.batchBegin = e.onBatchBegin, this.batchEnd = e.onBatchEnd, this.yield = e.onYield
+                    }
+                    async maybeWait(e, t, n) {
+                        const r = [];
+                        null != this.yield && (await Tb(n), r.push(this.yield(e, t, n))), r.push(this.nextFrameFunc()), await Promise.all(r)
+                    }
+                    async onEpochBegin(e, t) {
+                        this.currentEpoch = e, null != this.epochBegin && (await Tb(t), await this.epochBegin(e, t))
+                    }
+                    async onEpochEnd(e, t) {
+                        const n = [];
+                        null != this.epochEnd && (await Tb(t), n.push(this.epochEnd(e, t))), "epoch" === this.yieldEvery && n.push(this.nextFrameFunc()), await Promise.all(n)
+                    }
+                    async onBatchBegin(e, t) {
+                        null != this.batchBegin && (await Tb(t), await this.batchBegin(e, t))
+                    }
+                    async onBatchEnd(e, t) {
+                        const n = [];
+                        null != this.batchEnd && (await Tb(t), n.push(this.batchEnd(e, t))), "batch" === this.yieldEvery ? n.push(this.nextFrameFunc()) : va.isNumber(this.yieldEvery) && n.push(this.maybeWait(this.currentEpoch, e, t)), await Promise.all(n)
+                    }
+                    async onTrainBegin(e) {
+                        null != this.trainBegin && (await Tb(e), await this.trainBegin(e))
+                    }
+                    async onTrainEnd(e) {
+                        null != this.trainEnd && (await Tb(e), await this.trainEnd(e))
+                    }
+                };
+
+            function Lb(e, t) {
+                if (null == e && (e = {}), e instanceof _b) return [e];
+                if (Array.isArray(e) && e[0] instanceof _b) return e;
+                return fg(e).map(e => new Ob(e, t))
+            }
+            var Pb = class {
+                constructor() {}
+                static registerCallbackConstructor(e, t) {
+                    va.assert(e >= 0 && Number.isInteger(e), () => "Verbosity level is expected to be an integer >= 0, but got " + e), Pb.checkForDuplicate(t), null == Pb.constructors[e] && (Pb.constructors[e] = []), Pb.constructors[e].push(t)
+                }
+                static checkForDuplicate(e) {
+                    for (const t in Pb.constructors) {
+                        Pb.constructors[+t].forEach(t => {
+                            if (t === e) throw new og("Duplicate callback constructor.")
+                        })
+                    }
+                }
+                static clear() {
+                    Pb.constructors = {}
+                }
+                static createCallbacks(e) {
+                    const t = [];
+                    for (const n in Pb.constructors) {
+                        const r = +n;
+                        e >= r && t.push(...Pb.constructors[r])
+                    }
+                    return t.map(e => new e)
+                }
+            };
+
+            function $b(e, t, n, r, a, s, i, o, u) {
+                const l = new Db,
+                    c = [new Mb, ...Pb.createCallbacks(t)];
+                null != e && c.push(...e), c.push(l);
+                const h = new Fb(c);
+                return h.setParams({
+                    epochs: n,
+                    initialEpoch: r,
+                    samples: a,
+                    steps: s,
+                    batchSize: i,
+                    verbose: t,
+                    doValidation: o,
+                    metrics: u
+                }), {
+                    callbackList: h,
+                    history: l
+                }
+            }
+
+            function Bb(e, t = {}, n = !1) {
+                return vg(e, Fo.SerializationMap.getMap().classNameMap, t, "layer", n)
+            }
+
+            function zb(e, t) {
+                return ru(() => {
+                    "float32" !== e.dtype && (e = Ai(e, "float32"));
+                    const n = yc(by(e), t, !0),
+                        r = Vl(n.shape, ag()),
+                        a = Wh(Oc(n, r));
+                    return bu(e, a)
+                })
+            }
+
+            function Ub(e, t) {
+                return ru(() => Lc(by(gc(t, e)), -1))
+            }
+
+            function Wb(e, t) {
+                return ru(() => Lc(xu(gc(t, e)), -1))
+            }
+
+            function Vb(e, t) {
+                return ru(() => {
+                    const n = gc(e, t),
+                        r = cl(xu(e), ag(), Number.MAX_VALUE),
+                        a = xu(bu(n, r));
+                    return vu(100, Lc(a, -1))
+                })
+            }
+
+            function jb(e, t, n = !1) {
+                return ru(() => {
+                    if (n) t = Lh(t);
+                    else {
+                        const e = yc(t, t.shape.length - 1, !0);
+                        t = bu(t, e)
+                    }
+                    return t = cl(t, ag(), 1 - ag()), pc(yc(vu(Ai(e, "float32"), rc(t)), t.shape.length - 1))
+                })
+            }
+
+            function Gb(e, t, n = !1) {
+                return ru(() => {
+                    const r = Ai(jl(function (e) {
+                            const t = [ry(e.shape)];
+                            return qu(e, t)
+                        }(e)), "int32"),
+                        a = (t = cl(t, ag(), 1 - ag())).shape;
+                    return jb(qu(Hi(r, a[a.length - 1]), a), t, n)
+                })
+            }
+
+            function Hb(e, t) {
+                return ru(() => {
+                    let n;
+                    return n = cl(t, ag(), 1 - ag()), n = rc(bu(n, gc(1, n))), Lc(function (e, t) {
+                        if (!va.arraysEqual(e.shape, t.shape)) throw new og(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);
+                        return ru(() => {
+                            const n = gh(t),
+                                r = pc(xu(t));
+                            return gu(gc(n, vu(t, e)), ac($l(r)))
+                        })
+                    }(e, n), -1)
+                })
+            }
+
+            function qb(e, t) {
+                return ru(() => {
+                    const n = zb(e, -1),
+                        r = zb(t, -1),
+                        a = vu(n, r);
+                    return pc(yc(a, -1))
+                })
+            }
+            Pb.constructors = {};
+            var Kb = {
+                meanSquaredError: Ub,
+                meanAbsoluteError: Wb,
+                meanAbsolutePercentageError: Vb,
+                meanSquaredLogarithmicError: function (e, t) {
+                    return ru(() => {
+                        const n = cl(t, ag(), Number.MAX_VALUE),
+                            r = rc(gu(1, n)),
+                            a = cl(e, ag(), Number.MAX_VALUE),
+                            s = rc(gu(1, a));
+                        return Lc(by(gc(r, s)), -1)
+                    })
+                },
+                squaredHinge: function (e, t) {
+                    return ru(() => {
+                        const n = Oc(0, gc(1, vu(e, t)));
+                        return Lc(by(n), -1)
+                    })
+                },
+                hinge: function (e, t) {
+                    return ru(() => {
+                        const n = Oc(0, gc(1, vu(e, t)));
+                        return Lc(n, -1)
+                    })
+                },
+                categoricalHinge: function (e, t) {
+                    return ru(() => {
+                        const n = yc(vu(e, t), -1),
+                            r = mc(vu(gc(1, e), t), -1);
+                        return Oc(0, gu(1, gc(r, n)))
+                    })
+                },
+                logcosh: function (e, t) {
+                    return ru(() => {
+                        const n = Math.log(2),
+                            r = gc(t, e),
+                            a = gc(gu(r, dc(vu(-2, r))), n);
+                        return Lc(a, -1)
+                    })
+                },
+                categoricalCrossentropy: jb,
+                sparseCategoricalCrossentropy: Gb,
+                binaryCrossentropy: Hb,
+                kullbackLeiblerDivergence: function (e, t) {
+                    return ru(() => {
+                        const n = cl(e, ag(), 1),
+                            r = cl(t, ag(), 1);
+                        return yc(vu(e, rc(bu(n, r))), -1)
+                    })
+                },
+                poisson: function (e, t) {
+                    return ru(() => {
+                        const n = rc(gu(ag(), t));
+                        return Lc(gc(t, vu(e, n)), -1)
+                    })
+                },
+                cosineProximity: qb
+            };
+
+            function Yb(e) {
+                if ("string" == typeof e) {
+                    if (e in Kb) return Kb[e];
+                    let t = "Unknown loss " + e;
+                    throw e.toLowerCase().includes("softmaxcrossentropy") && (t = `Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`), new og(t)
+                }
+                return e
+            }
+
+            function Xb(e, t) {
+                return ru(() => {
+                    const n = vu(.5, Yc(t)),
+                        r = oy(Hl(t, n), e.dtype);
+                    return Lc(Rl(e, r), -1)
+                })
+            }
+
+            function Qb(e, t) {
+                return ru(() => oy(Rl(Eu(e, -1), Eu(t, -1)), "float32"))
+            }
+
+            function Zb(e, t) {
+                return ru(() => Ai(yc(Cc(Rl(e, 1), Rl(t, 1))), "float32"))
+            }
+
+            function Jb(e, t) {
+                return ru(() => {
+                    const n = Zb(e, t),
+                        r = function (e, t) {
+                            return ru(() => Ai(yc(Cc(Rl(e, 0), Rl(t, 1))), "float32"))
+                        }(e, t),
+                        a = gu(n, r);
+                    return Ai(_l(Hl(a, 0), bu(n, a), 0), "float32")
+                })
+            }
+
+            function ev(e, t) {
+                return ru(() => {
+                    const n = Zb(e, t),
+                        r = function (e, t) {
+                            return ru(() => Ai(yc(Cc(Rl(e, 1), Rl(t, 0))), "float32"))
+                        }(e, t),
+                        a = gu(n, r);
+                    return Ai(_l(Hl(a, 0), bu(n, a), 0), "float32")
+                })
+            }
+
+            function tv(e, t) {
+                return Hb(e, t)
+            }
+
+            function nv(e, t) {
+                return e.rank === t.rank && (e = jh(e, [e.rank - 1])), (t = Eu(t, -1)).dtype !== e.dtype && (t = Ai(t, e.dtype)), Ai(Rl(e, t), "float32")
+            }
+            var rv = jb,
+                av = Gb,
+                sv = {
+                    binaryAccuracy: Xb,
+                    categoricalAccuracy: Qb,
+                    precision: Jb,
+                    categoricalCrossentropy: rv,
+                    sparseCategoricalCrossentropy: av,
+                    mse: Ub,
+                    MSE: Ub,
+                    mae: Wb,
+                    MAE: Wb,
+                    mape: Vb,
+                    MAPE: Vb,
+                    cosine: qb
+                };
+
+            function iv(e) {
+                if ("string" == typeof e && e in sv) return sv[e];
+                if ("string" != typeof e && null != e) return e;
+                throw new og("Unknown metric " + e)
+            }
+
+            function ov(e) {
+                if (hg(null !== e, "Unknown LossOrMetricFn " + e), "string" == typeof e) return e; {
+                    let t;
+                    for (const n of Object.keys(Kb))
+                        if (Kb[n] === e) {
+                            t = n;
+                            break
+                        } if (void 0 !== t) return t;
+                    for (const n of Object.keys(sv))
+                        if (sv[n] === e) {
+                            t = n;
+                            break
+                        } return void 0 !== t ? t : e.name
+                }
+            }
+
+            function uv(e, t, n = !1) {
+                if (null == e || "object" != typeof e || Object.getPrototypeOf(e) !== Object.prototype || ! function e(t) {
+                        if (null === t) return !0;
+                        if ("object" == typeof t) {
+                            if (Object.getPrototypeOf(t) === Object.prototype) {
+                                const n = Object.keys(t);
+                                for (const r of n) {
+                                    if ("string" != typeof r) return !1;
+                                    if (!e(t[r])) return !1
+                                }
+                                return !0
+                            }
+                            if (Array.isArray(t)) {
+                                for (const n of t)
+                                    if (!e(n)) return !1;
+                                return !0
+                            }
+                            return !1
+                        } {
+                            const e = typeof t;
+                            return "string" === e || "number" === e || "boolean" === e
+                        }
+                    }(e)) throw new Error("User-defined metadata is expected to be a JSON object, but is not.");
+                if (n) {
+                    const n = JSON.stringify(e);
+                    n.length > 1048576 && console.warn(`User-defined metadata of model "${t}" is too large in size (length=${n.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= 1048576.`)
+                }
+            }
+
+            function lv(e, t, n, r = console.log) {
+                const a = function (e) {
+                        let t = !0;
+                        const n = [],
+                            r = [];
+                        for (const t in e.nodesByDepth) n.push(e.nodesByDepth[t]);
+                        for (const e of n) {
+                            if (e.length > 1 || 1 === e.length && e[0].inboundLayers.length > 1) {
+                                t = !1;
+                                break
+                            }
+                            r.push(...e)
+                        }
+                        if (t)
+                            for (const n of e.layers) {
+                                let e = !1;
+                                for (const a of n.inboundNodes)
+                                    if (-1 !== r.indexOf(a)) {
+                                        if (e) {
+                                            t = !1;
+                                            break
+                                        }
+                                        e = !0
+                                    } if (!t) break
+                            }
+                        return t
+                    }(e),
+                    s = ["Layer (type)", "Output shape", "Param #"];
+                let i;
+                if (a ? (t = t || 65, n = n || [.45, .85, 1]) : (t = t || 98, n = n || [.33, .55, .67, 1]), n[n.length - 1] <= 1 && (n = n.map(e => Math.floor(t * e))), !a) {
+                    s.push("Receives inputs"), i = [];
+                    for (const t in e.nodesByDepth) i.push(...e.nodesByDepth[t])
+                }
+                r("_".repeat(t)), cv(s, n, r), r("=".repeat(t));
+                const o = e.layers;
+                for (let e = 0; e < o.length; ++e) a ? hv(o[e], n, r) : pv(o[e], n, i, r), r((e === o.length - 1 ? "=" : "_").repeat(t));
+                e.checkTrainableWeightsConsistency();
+                const u = function (e) {
+                        let t;
+                        t = null != e.collectedTrainableWeights ? mb(e.collectedTrainableWeights) : mb(e.trainableWeights);
+                        return t
+                    }(e),
+                    l = mb(e.nonTrainableWeights);
+                r("Total params: " + (u + l)), r("Trainable params: " + u), r("Non-trainable params: " + l), r("_".repeat(t))
+            }
+
+            function cv(e, t, n = console.log) {
+                let r = "";
+                for (let n = 0; n < e.length; ++n) n > 0 && (r = r.slice(0, r.length - 1) + " "), r += e[n], r = r.slice(0, t[n]), r += " ".repeat(t[n] - r.length);
+                n(r)
+            }
+
+            function hv(e, t, n) {
+                let r;
+                try {
+                    r = JSON.stringify(e.outputShape)
+                } catch (e) {
+                    r = "multiple"
+                }
+                cv([`${e.name} (${e.getClassName()})`, r, e.countParams().toString()], t, n)
+            }
+
+            function pv(e, t, n, r) {
+                let a;
+                try {
+                    a = JSON.stringify(e.outputShape)
+                } catch (e) {
+                    a = "multiple"
+                }
+                const s = [];
+                for (const t of e.inboundNodes)
+                    if (!(null != n && n.length > 0 && -1 === n.indexOf(t)))
+                        for (let e = 0; e < t.inboundLayers.length; ++e) {
+                            const n = t.inboundLayers[e].name,
+                                r = t.nodeIndices[e],
+                                a = t.tensorIndices[e];
+                            s.push(`${n}[${r}][${a}]`)
+                        }
+                const i = e.name,
+                    o = e.getClassName(),
+                    u = 0 === s.length ? "" : s[0];
+                cv([`${i} (${o})`, a, e.countParams().toString(), u], t, r);
+                for (let e = 1; e < s.length; ++e) cv(["", "", "", s[e]], t, r)
+            }
+
+            function dv(e, t, n) {
+                return ("inboundNodes" === e || "outputLayers" === e || "inputLayers" === e) && 0 === t && "string" == typeof n
+            }
+
+            function fv(e, t) {
+                if (null === e) return null;
+                if ("string" == typeof e) return gg(e);
+                if ("number" == typeof e || "boolean" == typeof e) return e;
+                if (e instanceof Array) {
+                    const n = [],
+                        r = e.length;
+                    for (let a = 0; a < r; ++a) {
+                        const r = e[a];
+                        dv(t, a, r) ? n.push(r) : n.push(fv(r, t))
+                    }
+                    return n
+                } {
+                    const t = {};
+                    for (const n of Object.keys(e)) {
+                        const r = e[n];
+                        if ("name" === n && "string" == typeof r) t[n] = r;
+                        else {
+                            const e = gg(n);
+                            t[e] = fv(r, e)
+                        }
+                    }
+                    return t
+                }
+            }
+            var mv = "3.13.0";
+            var gv = class {
+                    constructor(e) {
+                        if (this.id2Value = {}, this.id2Mask = {}, this.name2Id = {}, e instanceof gv)
+                            for (const t in e.id2Value) this.id2Value[t] = e.id2Value[t], t in e.id2Mask && (this.id2Mask[t] = e.id2Mask[t]);
+                        else {
+                            if (null == e) return;
+                            for (const t of e) this.add(t.key, t.value)
+                        }
+                    }
+                    add(e, t, n) {
+                        if (null != this.id2Value[e.id]) throw new og(`Duplicate key: name=${e.name}, id=${e.id}`);
+                        return this.id2Value[e.id] = function (e, t) {
+                            if (null == e.dtype || e.dtype === t.dtype) return t;
+                            try {
+                                return Ai(t, e.dtype)
+                            } catch (n) {
+                                throw new og(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)
+                            }
+                        }(e, t), this.name2Id[e.name] = e.id, null != n && (this.id2Mask[e.id] = n), this
+                    }
+                    addFeed(e) {
+                        this.add(e.key, e.value)
+                    }
+                    hasKey(e) {
+                        return null != this.id2Value[e.id]
+                    }
+                    names() {
+                        return Object.keys(this.name2Id)
+                    }
+                    getValue(e) {
+                        if (e instanceof xb) {
+                            if (null == this.id2Value[e.id]) throw new og("Nonexistent key: " + e.name);
+                            return this.id2Value[e.id]
+                        } {
+                            const t = this.name2Id[e];
+                            if (null == t) throw new og("Feed dict has no SymbolicTensor name: " + e);
+                            return this.id2Value[t]
+                        }
+                    }
+                    getMask(e) {
+                        if (e instanceof xb) {
+                            if (null == this.id2Value[e.id]) throw new og("Nonexistent key: " + e.name);
+                            return this.id2Mask[e.id]
+                        } {
+                            const t = this.name2Id[e];
+                            if (null == t) throw new og("Feed dict has no SymbolicTensor name: " + e);
+                            return this.id2Mask[t]
+                        }
+                    }
+                    disposeMasks() {
+                        null != this.id2Mask && au(this.id2Mask)
+                    }
+                },
+                yv = {},
+                bv = {};
+
+            function vv(e, t, n, r) {
+                const a = null != n && n.training,
+                    s = Array.isArray(e),
+                    i = s ? e : [e],
+                    o = i.map(e => e.name),
+                    u = [],
+                    l = t.names();
+                for (const e of o) - 1 !== l.indexOf(e) ? u.push(t.getValue(e)) : u.push(null);
+                null != r && (r.maxNumTensors = -1 / 0, r.minNumTensors = 1 / 0);
+                const c = o.join(",") + "|" + t.names().join(",");
+                let h, p;
+                if (null == yv[c]) {
+                    const e = function (e, t) {
+                        va.assert(null != e && e.length > 0, () => "Expected at least one fetch, got none");
+                        let n = [],
+                            r = {};
+                        if (1 === e.length) {
+                            const a = wv(e[0], t);
+                            n = a.sorted, r = a.recipientMap
+                        } else {
+                            const a = new Set;
+                            for (const s of e) {
+                                const {
+                                    sorted: e,
+                                    recipientMap: i
+                                } = wv(s, t);
+                                for (const t of e) a.has(t.name) || (n.push(t), a.add(t.name));
+                                for (const e in i) null == r[e] && (r[e] = new Set), i[e].forEach(t => r[e].add(t))
+                            }
+                        }
+                        return {
+                            sorted: n,
+                            recipientCounts: xv(r)
+                        }
+                    }(i, t);
+                    h = e.sorted, p = e.recipientCounts, yv[c] = h, bv[c] = p
+                }
+                h = yv[c], p = {}, a || Object.assign(p, bv[c]);
+                const d = new gv(t);
+                for (let e = 0; e < h.length; ++e) {
+                    if (null != r) {
+                        const e = tu().numTensors;
+                        e > r.maxNumTensors && (r.maxNumTensors = e), e < r.minNumTensors && (r.minNumTensors = e)
+                    }
+                    const s = h[e],
+                        i = s.sourceLayer;
+                    if (i instanceof Nb) continue;
+                    const l = [],
+                        c = [],
+                        f = [];
+                    let m = !1;
+                    for (const e of s.inputs) {
+                        const n = d.getValue(e),
+                            r = d.getMask(e);
+                        l.push(n), c.push(r), null != r && (m = !0), a || (p[e.name]--, 0 !== p[e.name] || t.hasKey(e) || -1 !== o.indexOf(e.name) || n.isDisposed || !0 === e.sourceLayer.stateful || f.push(n))
+                    }
+                    m && ((n = n || {}).mask = c[0]);
+                    const g = fg(i.apply(l, n));
+                    let y = null;
+                    i.supportsMasking && (y = i.computeMask(l, c));
+                    const b = kv(s),
+                        v = Array.isArray(b) ? b : [b];
+                    for (let e = 0; e < v.length; ++e) {
+                        d.hasKey(v[e]) || d.add(v[e], g[e], Array.isArray(y) ? y[0] : y);
+                        const t = o.indexOf(v[e].name); - 1 !== t && (u[t] = g[e])
+                    }
+                    a || au(f)
+                }
+                return d.disposeMasks(), s ? u : u[0]
+            }
+
+            function xv(e) {
+                const t = {};
+                for (const n in e) t[n] = e[n].size;
+                return t
+            }
+
+            function wv(e, t) {
+                const n = new Set,
+                    r = [],
+                    a = {};
+                for (const e of t.names()) n.add(e);
+                const s = [],
+                    i = [];
+                for (s.push(e); s.length > 0;) {
+                    const e = s[s.length - 1];
+                    if (n.has(e.name)) {
+                        s.pop();
+                        continue
+                    }
+                    const t = i[i.length - 1] === s.length - 1;
+                    if (0 === e.inputs.length || t) s.pop(), r.push(e), n.add(e.name), t && i.pop();
+                    else {
+                        i.push(s.length - 1);
+                        for (const t of e.inputs) null == a[t.name] && (a[t.name] = new Set), a[t.name].add(e.name), n.has(t.name) || s.push(t)
+                    }
+                }
+                return {
+                    sorted: r,
+                    recipientMap: a
+                }
+            }
+
+            function kv(e) {
+                let t;
+                if (1 === e.sourceLayer.inboundNodes.length) t = e.sourceLayer.output;
+                else {
+                    let n = null;
+                    for (let t = 0; t < e.sourceLayer.inboundNodes.length; ++t)
+                        for (const r of e.sourceLayer.inboundNodes[t].outputTensors)
+                            if (r.id === e.id) {
+                                n = t;
+                                break
+                            } t = e.sourceLayer.getOutputAt(n)
+                }
+                return t
+            }
+            var Sv = class extends Ib {
+                constructor(e) {
+                    if (super({}), this.containerNodes = new Set, this.name = e.name, null == this.name) {
+                        const e = this.getClassName().toLowerCase();
+                        this.name = cb(e)
+                    }
+                    if (this.supportsMasking = !1, this.trainable_ = !0, Array.isArray(e.inputs) ? this.inputs = e.inputs.slice() : this.inputs = [e.inputs], Array.isArray(e.outputs) ? this.outputs = e.outputs.slice() : this.outputs = [e.outputs], wg(this.inputs).length !== this.inputs.length) throw new og("The list of inputs passed to the model is redundant. All inputs should only appear once. Found: " + this.inputs.map(e => e.name));
+                    wg(this.outputs).length !== this.outputs.length && console.warn("The list of outputs passed to the model is redundant. All outputs should only appear once. Found: " + this.outputs.map(e => e.name)), this.inputLayers = [], this.inputLayersNodeIndices = [], this.inputLayersTensorIndices = [], this.outputLayers = [], this.outputLayersNodeIndices = [], this.outputLayersTensorIndices = [], this.layers = [], this.internalContainerRefs = [];
+                    for (const e of this.outputs) {
+                        const t = e.sourceLayer,
+                            n = e.nodeIndex,
+                            r = e.tensorIndex;
+                        this.outputLayers.push(t), this.outputLayersNodeIndices.push(n), this.outputLayersTensorIndices.push(r)
+                    }
+                    for (const e of this.inputs) {
+                        const t = e.sourceLayer,
+                            n = e.nodeIndex,
+                            r = e.tensorIndex;
+                        hg(0 === n, "input layer has >1 nodes"), hg(0 === r, "input layer has >1 tensors"), this.inputLayers.push(t), this.inputLayersNodeIndices.push(n), this.inputLayersTensorIndices.push(r)
+                    }
+                    this.inputNames = [], this.outputNames = [], this.feedInputShapes = [], this.feedInputNames = [], this.feedOutputNames = [];
+                    for (let t = 0; t < this.inputLayers.length; t++) {
+                        const n = this.inputLayers[t];
+                        if (!(n instanceof Nb)) throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${e.inputs}. Input ${t} (0-based) originates from layer type ${n.getClassName()}.`);
+                        this.inputNames.push(n.name), this.feedInputShapes.push(n.batchInputShape), this.feedInputNames.push(n.name)
+                    }
+                    for (const e of this.outputLayers) this.outputNames.push(e.name);
+                    this.internalInputShapes = this.inputs.map(e => e.shape), this.internalOutputShapes = this.outputs.map(e => e.shape);
+                    const t = {},
+                        n = {},
+                        r = {},
+                        a = {},
+                        s = {},
+                        i = [],
+                        o = (e, t, n, r, a, u) => {
+                            null != r && null != a && null != u || (r = e.sourceLayer, a = e.nodeIndex, u = e.tensorIndex);
+                            const l = r.inboundNodes[a];
+                            if (-1 !== n.indexOf(l)) throw new ig(`The tensor ${e.name} at layer "${r.name}" is part of a cycle.`);
+                            if (-1 !== t.indexOf(l)) return;
+                            this.containerNodes.add(Sv.nodeKey(r, a)), r.id in s || (s[r.id] = Object.keys(s).length), -1 === n.indexOf(l) && n.push(l);
+                            const c = l.inboundLayers.length;
+                            for (let e = 0; e < c; e++) {
+                                const r = l.inputTensors[e],
+                                    a = l.inboundLayers[e],
+                                    s = l.nodeIndices[e],
+                                    i = l.tensorIndices[e];
+                                o(r, t, n, a, s, i)
+                            }
+                            for (t.push(l); n.indexOf(l) >= 0;) n.splice(n.indexOf(l), 1);
+                            i.push(l)
+                        },
+                        u = [],
+                        l = [];
+                    for (const e of this.outputs) o(e, u, l);
+                    const c = i.slice().reverse();
+                    for (const e of c) {
+                        n[e.id] = e, e.id in t || (t[e.id] = 0);
+                        let s = t[e.id];
+                        const i = null == r[e.outboundLayer.id] ? 0 : r[e.outboundLayer.id];
+                        s = Math.max(s, i), r[e.outboundLayer.id] = s, a[e.outboundLayer.id] = e.outboundLayer, t[e.id] = s;
+                        for (let r = 0; r < e.inboundLayers.length; r++) {
+                            const a = e.inboundLayers[r],
+                                i = e.nodeIndices[r],
+                                o = a.inboundNodes[i],
+                                u = null == t[o.id] ? 0 : t[o.id];
+                            t[o.id] = Math.max(s + 1, u), n[o.id] = o
+                        }
+                    }
+                    const h = {};
+                    for (const e in t) {
+                        const r = t[e];
+                        r in h || (h[r] = []), h[r].push(n[e])
+                    }
+                    const p = {};
+                    for (const e in r) {
+                        const t = r[e];
+                        t in p || (p[t] = []), p[t].push(a[e])
+                    }
+                    let d = Object.keys(p).map(e => parseInt(e, 10)).sort(xg);
+                    this.layers = [];
+                    for (const e of d) {
+                        const t = p[e];
+                        t.sort((e, t) => {
+                            const n = s[e.id],
+                                r = s[t.id];
+                            return n < r ? -1 : n > r ? 1 : 0
+                        });
+                        for (const e of t) e instanceof Sv && this.internalContainerRefs.push(e), this.layers.push(e)
+                    }
+                    this.layersByDepth = p, d = Object.keys(h).map(e => parseInt(e, 10)).sort(xg);
+                    const f = this.inputs.slice(),
+                        m = [];
+                    for (const e of d)
+                        for (const t of h[e]) {
+                            const e = t.outboundLayer;
+                            if (null != e) {
+                                for (const n of t.inputTensors)
+                                    if (-1 === f.indexOf(n)) throw new ig(`Graph disconnected: cannot obtain value for tensor ${n} at layer "${e.name}". The following previous layers were accessed without issue: ${m}`);
+                                for (const e of t.outputTensors) f.push(e);
+                                m.push(e.name)
+                            }
+                        }
+                    this.nodesByDepth = h;
+                    const g = this.layers.map(e => e.name);
+                    for (const e of g) {
+                        const t = g.filter(t => t === e).length;
+                        if (1 !== t) throw new ig(`The name "${e}" is used ${t} times in the model. All layer names should be unique. Layer names: ` + JSON.stringify(g))
+                    }
+                    this.outboundNodes = [], this.inboundNodes = [], new kb({
+                        outboundLayer: this,
+                        inboundLayers: [],
+                        nodeIndices: [],
+                        tensorIndices: [],
+                        inputTensors: this.inputs,
+                        outputTensors: this.outputs,
+                        inputMasks: this.inputs.map(e => null),
+                        outputMasks: this.outputs.map(e => null),
+                        inputShapes: this.inputs.map(e => e.shape),
+                        outputShapes: this.outputs.map(e => e.shape)
+                    }), this.built = !0, this._refCount = 1
+                }
+                assertNotDisposed() {
+                    if (0 === this._refCount) throw new Error(`Container '${this.name}' is already disposed.`)
+                }
+                dispose() {
+                    this.assertNotDisposed();
+                    const e = {
+                        refCountAfterDispose: null,
+                        numDisposedVariables: 0
+                    };
+                    if (0 == --this._refCount) {
+                        for (const t of this.layers) e.numDisposedVariables += t.dispose().numDisposedVariables;
+                        for (const t of this.internalContainerRefs) e.numDisposedVariables += t.dispose().numDisposedVariables
+                    }
+                    return e.refCountAfterDispose = this._refCount, e
+                }
+                get trainable() {
+                    return this.trainable_
+                }
+                set trainable(e) {
+                    this.layers.forEach(t => {
+                        t._trainableWeights.forEach(t => t.trainable = e)
+                    }), this.trainable_ = e
+                }
+                get trainableWeights() {
+                    if (this._trainableWeights.length > 0) throw new og("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");
+                    if (!this.trainable) return [];
+                    let e = [];
+                    for (const t of this.layers) e = e.concat(t.trainableWeights);
+                    return e
+                }
+                get nonTrainableWeights() {
+                    const e = [];
+                    for (const t of this.layers) e.push(...t.nonTrainableWeights);
+                    if (!this.trainable) {
+                        const t = [];
+                        for (const e of this.layers) t.push(...e.trainableWeights);
+                        return t.concat(e)
+                    }
+                    return e
+                }
+                get weights() {
+                    return this.trainableWeights.concat(this.nonTrainableWeights)
+                }
+                loadWeights(e, t = !0) {
+                    const n = {};
+                    let r = 0;
+                    for (const e of this.layers)
+                        for (const t of e.weights) {
+                            if (null != n[t.originalName]) throw new og("Duplicate weight name: " + t.originalName);
+                            n[t.originalName] = t, r++
+                        }
+                    const a = [];
+                    for (const r in e) {
+                        let s = r;
+                        if (null == n[r]) {
+                            const e = r.split("/");
+                            s = e.slice(0, -2).concat([e[e.length - 1]]).join("/")
+                        }
+                        if (null != n[s]) a.push([n[s], e[r]]);
+                        else if (t) throw new og("Provided weight data has no target variable: " + r);
+                        delete n[s]
+                    }
+                    if (t) {
+                        const e = [];
+                        for (const t in n) e.push(t);
+                        if (e.length > 0) throw new og(`${e.length} of ${r} weights are not set: ${e}`)
+                    }
+                    bb(a)
+                }
+                updatedConfig() {
+                    const e = this.getConfig(),
+                        t = {};
+                    return t.className = this.getClassName(), t.config = e, t.kerasVersion = "tfjs-layers " + mv, t.backend = "TensorFlow.js", t
+                }
+                toJSON(e, t = !0) {
+                    const n = function e(t, n) {
+                        if (null == t) return null;
+                        if ("string" == typeof t) return mg(t);
+                        if ("number" == typeof t || "boolean" == typeof t) return t;
+                        if (t instanceof Array) {
+                            const r = [],
+                                a = t.length;
+                            for (let s = 0; s < a; ++s) {
+                                const a = t[s];
+                                dv(n, s, a) ? r.push(a) : r.push(e(a, n))
+                            }
+                            return r
+                        } {
+                            const n = {};
+                            for (const r of Object.keys(t)) {
+                                const a = t[r],
+                                    s = mg(r);
+                                n[s] = "name" !== r && "className" !== r || "string" != typeof a ? e(a, r) : a
+                            }
+                            return n
+                        }
+                    }(this.updatedConfig());
+                    return t ? JSON.stringify(n) : n
+                }
+                call(e, t) {
+                    return ru(() => {
+                        e = fg(e);
+                        const n = new gv;
+                        for (let t = 0; t < this.inputs.length; ++t) n.add(this.inputs[t], e[t]);
+                        return vv(this.outputs, n, t)
+                    })
+                }
+                computeMask(e, t) {
+                    return ru(() => {
+                        let n;
+                        return e = fg(e), n = null == t ? cg(null, e.length) : fg(t), this.runInternalGraph(e, n)[1]
+                    })
+                }
+                computeOutputShape(e) {
+                    const t = pb(e);
+                    if (t.length !== this.inputLayers.length) throw new og(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);
+                    const n = {};
+                    for (let e = 0; e < t.length; e++) {
+                        const r = this.inputLayers[e],
+                            a = t[e];
+                        n[r.name + "_0_0"] = a
+                    }
+                    const r = Object.keys(this.nodesByDepth).map(e => parseInt(e, 10)).sort(xg);
+                    if (r.length > 1)
+                        for (const e of r) {
+                            const t = this.nodesByDepth[e];
+                            for (const e of t) {
+                                const t = e.outboundLayer;
+                                if (-1 !== this.inputLayers.map(e => e.id).indexOf(t.id)) continue;
+                                const r = [];
+                                for (let t = 0; t < e.inboundLayers.length; t++) {
+                                    const a = e.inboundLayers[t],
+                                        s = e.nodeIndices[t],
+                                        i = e.tensorIndices[t],
+                                        o = n[`${a.name}_${s}_${i}`];
+                                    r.push(o)
+                                }
+                                const a = pb(t.computeOutputShape(dg(r))),
+                                    s = t.inboundNodes.indexOf(e);
+                                for (let e = 0; e < a.length; e++) {
+                                    n[`${t.name}_${s}_${e}`] = a[e]
+                                }
+                            }
+                        }
+                    const a = [],
+                        s = [];
+                    for (let e = 0; e < this.outputLayers.length; e++) {
+                        const t = this.outputLayers[e],
+                            n = this.outputLayersNodeIndices[e],
+                            r = this.outputLayersTensorIndices[e],
+                            a = `${t.name}_${n}_${r}`;
+                        s.push(a)
+                    }
+                    for (let e = 0; e < s.length; e++) {
+                        const t = s[e];
+                        hg(t in n), a.push(n[t])
+                    }
+                    return dg(a)
+                }
+                runInternalGraph(e, t) {
+                    null == t && (t = cg(null, e.length));
+                    const n = {};
+                    for (let r = 0; r < this.inputs.length; ++r) {
+                        const a = this.inputs[r],
+                            s = e[r],
+                            i = t[r];
+                        n[a.id] = [s, i]
+                    }
+                    const r = Object.keys(this.nodesByDepth).map(e => parseInt(e, 10)).sort(xg);
+                    for (const e of r) {
+                        const t = this.nodesByDepth[e];
+                        for (const e of t) {
+                            const t = e.outboundLayer,
+                                r = e.inputTensors,
+                                a = e.outputTensors,
+                                s = new Array;
+                            for (const e of r) e.id in n && s.push(n[e.id]);
+                            if (s.length === r.length) {
+                                let r, i, o, u, l = {};
+                                if (null != e.callArgs && (l = e.callArgs), 1 === s.length) {
+                                    const [e, n] = s[0];
+                                    null == l.mask && (l.mask = n), o = fg(t.call(e, l)), u = fg(t.computeMask(e, n)), r = [e], i = [n]
+                                } else r = s.map(e => e[0]), i = s.map(e => e[1]), null == l.mask && (l.mask = i), o = fg(t.call(r, l)), u = fg(t.computeMask(r, i));
+                                if (t.activityRegularizer) throw new ug("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");
+                                for (let e = 0; e < a.length; ++e) {
+                                    const t = a[e],
+                                        r = o[e],
+                                        s = u[e];
+                                    n[t.id] = [r, s]
+                                }
+                            }
+                        }
+                    }
+                    const a = [],
+                        s = [],
+                        i = [];
+                    for (const e of this.outputs) {
+                        hg(e.id in n, `Could not compute output ${e.name} : ${e.id}`);
+                        const [t, r] = n[e.id];
+                        i.push(t.shape), a.push(t), s.push(r)
+                    }
+                    return [a, s, i]
+                }
+                buildNodeConversionMap(e) {
+                    const t = {};
+                    let n;
+                    for (const e of this.layers) {
+                        n = e instanceof Sv ? 1 : 0;
+                        for (let r = 0; r < e.inboundNodes.length; r++) {
+                            const a = Sv.nodeKey(e, r);
+                            this.containerNodes.has(a) && (t[a] = n, n += 1)
+                        }
+                    }
+                    return t
+                }
+                getLayer(e, t) {
+                    if (null != t) {
+                        if (this.layers.length <= t) throw new og(`Was asked to retrieve layer at index ${t}, but model only has ${this.layers.length} layer(s).`);
+                        return this.layers[t]
+                    }
+                    if (null == e) throw new og("Provide either a layer name or layer index");
+                    for (const t of this.layers)
+                        if (t.name === e) return t;
+                    throw new og("No such layer: " + e)
+                }
+                calculateLosses() {
+                    return ru(() => {
+                        const e = [];
+                        for (const t of this.layers)
+                            for (let n = 0; n < t.inboundNodes.length; ++n) {
+                                const r = Sv.nodeKey(t, n);
+                                this.containerNodes.has(r) && e.push(...t.calculateLosses())
+                            }
+                        return e
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            name: this.name
+                        },
+                        t = this.buildNodeConversionMap(this.layers),
+                        n = [];
+                    for (const e of this.layers) {
+                        const r = e.getClassName(),
+                            a = e.getConfig(),
+                            s = [];
+                        for (let n = 0; n < e.inboundNodes.length; n++) {
+                            const r = e.inboundNodes[n],
+                                a = Sv.nodeKey(e, n);
+                            let i = {};
+                            if (this.containerNodes.has(a)) {
+                                if (r.callArgs) try {
+                                    JSON.stringify(r.callArgs), i = r.callArgs
+                                } catch (t) {
+                                    console.warn(`Layer ${e.name} was passed non-serializable keyword arguments: ${r.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`), i = {}
+                                }
+                                if (r.inboundLayers.length > 0) {
+                                    const e = [];
+                                    for (let n = 0; n < r.inboundLayers.length; n++) {
+                                        const a = r.inboundLayers[n],
+                                            s = r.nodeIndices[n],
+                                            o = r.tensorIndices[n];
+                                        let u = t[Sv.nodeKey(a, s)];
+                                        null == u && (u = 0), e.push([a.name, u, o, i])
+                                    }
+                                    s.push(e)
+                                }
+                            }
+                        }
+                        const i = {};
+                        i.name = e.name, i.className = r, i.config = a, i.inboundNodes = s, n.push(i)
+                    }
+                    e.layers = n;
+                    const r = [];
+                    for (let e = 0; e < this.inputLayers.length; e++) {
+                        const n = this.inputLayers[e],
+                            a = this.inputLayersNodeIndices[e],
+                            s = Sv.nodeKey(n, a);
+                        if (!this.containerNodes.has(s)) continue;
+                        let i = t[s];
+                        null == i && (i = 0);
+                        const o = this.inputLayersTensorIndices[e];
+                        r.push([n.name, i, o])
+                    }
+                    e.inputLayers = r;
+                    const a = [];
+                    for (let e = 0; e < this.outputLayers.length; e++) {
+                        const n = this.outputLayers[e],
+                            r = this.outputLayersNodeIndices[e],
+                            s = Sv.nodeKey(n, r);
+                        if (!this.containerNodes.has(s)) continue;
+                        let i = t[s];
+                        null == i && (i = 0);
+                        const o = this.outputLayersTensorIndices[e];
+                        a.push([n.name, i, o])
+                    }
+                    return e.outputLayers = a, e
+                }
+                static fromConfig(e, t, n = {}, r = !1) {
+                    const a = {},
+                        s = {};
+
+                    function i(e, t) {
+                        e.name in s ? s[e.name].push(t) : s[e.name] = [t]
+                    }
+
+                    function o(e, t) {
+                        const n = [];
+                        let r;
+                        for (const s of t) {
+                            const o = s[0],
+                                u = s[1],
+                                l = s[2];
+                            if (r = null == s[3] ? {} : s[3], !(o in a)) return void i(e, t);
+                            const c = a[o];
+                            if (c.inboundNodes.length <= u) return void i(e, t);
+                            const h = c.inboundNodes[u];
+                            n.push(h.outputTensors[l])
+                        }
+                        n.length > 0 && e.apply(dg(n), r)
+                    }
+
+                    function u(e) {
+                        const n = e.name,
+                            s = Bb(e, null != t.customObjects ? t.customObjects : {});
+                        s.setFastWeightInitDuringBuild(r), a[n] = s;
+                        e.inboundNodes.forEach(e => {
+                            if (!(e instanceof Array)) throw new og("Corrupted configuration, expected array for nodeData: " + e);
+                            i(s, e)
+                        })
+                    }
+                    const l = t.name,
+                        c = t.layers;
+                    for (const e of c) u(e);
+                    for (; !kg(s);)
+                        for (const e of c) {
+                            const t = a[e.name];
+                            if (t.name in s) {
+                                const e = s[t.name];
+                                delete s[t.name];
+                                for (const n of e) o(t, n)
+                            }
+                        }
+                    const h = [],
+                        p = [],
+                        d = t.inputLayers;
+                    for (const e of d) {
+                        const t = e[0],
+                            n = e[1],
+                            r = e[2];
+                        hg(t in a);
+                        const s = a[t].inboundNodes[n].outputTensors;
+                        h.push(s[r])
+                    }
+                    const f = t.outputLayers;
+                    for (const e of f) {
+                        const t = e[0],
+                            n = e[1],
+                            r = e[2];
+                        hg(t in a);
+                        const s = a[t].inboundNodes[n].outputTensors;
+                        p.push(s[r])
+                    }
+                    return new e({
+                        inputs: h,
+                        outputs: p,
+                        name: l
+                    })
+                }
+                get stateful() {
+                    if (this._stateful) throw new og("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");
+                    for (const e of this.layers)
+                        if (e.stateful) return !0;
+                    return !1
+                }
+                resetStates() {
+                    ru(() => {
+                        this.layers.forEach(e => {
+                            e.stateful && e.resetStates()
+                        })
+                    })
+                }
+            };
+
+            function Iv(e, t) {
+                return function (e, t, n) {
+                    const r = t.length;
+                    if (null == e || Array.isArray(e) && 0 === e.length) return t.map(e => null);
+                    if (1 === r) return Array.isArray(e) && 1 === e.length ? e : "object" == typeof e && t[0] in e ? [e[t[0]]] : [e];
+                    if (Array.isArray(e)) {
+                        if (e.length !== r) throw new Error(`Provided ${n} is an array of ${e.length} element(s), but the model has ${r} outputs. Make sure a set of weights is provided for each model output.`);
+                        return e
+                    }
+                    if ("object" == typeof e && Object.keys(e).length > 0 && "object" == typeof e[Object.keys(e)[0]]) {
+                        const n = [];
+                        return t.forEach(t => {
+                            t in e ? n.push(e[t]) : n.push(null)
+                        }), n
+                    }
+                    throw new Error(`The model has multiple (${r}) outputs, so ${n} must be either an array with ${r} elements or an object with ${t} keys. Provided ${n} not understood: ${JSON.stringify(e)}`)
+                }(e, t, "classWeight")
+            }
+            async function Av(e, t, n, r) {
+                if (null != t || null != r) throw new Error("Support sampleWeight is not implemented yet");
+                if (null != n) {
+                    const t = ru(() => {
+                            if (1 === e.shape.length) return Ei(e);
+                            if (2 === e.shape.length) {
+                                if (e.shape[1] > 1) {
+                                    return Eu(e, 1)
+                                }
+                                if (1 === e.shape[1]) return qu(e, [e.shape[0]]);
+                                throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)
+                            }
+                            throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)
+                        }),
+                        r = Array.from(await t.data());
+                    au(t);
+                    const a = [];
+                    return r.forEach(e => {
+                        if (null == n[e]) throw new Error(`classWeight must contain all classes in the training data. The class ${e} exists in the data but not in classWeight`);
+                        a.push(n[e])
+                    }), Yh(a, "float32")
+                }
+                return null
+            }
+
+            function Ev(e, t) {
+                return vu(e, t)
+            }
+
+            function Nv(e, t) {
+                let n, r;
+                const a = t;
+                n = a.xs, r = a.ys, va.assert(null != n && null != r, () => "A Dataset iterator for fitDataset() is expected to generate objects of the form `{xs: xVal, ys: yVal}`, where the two values may be `tf.Tensor`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates " + t);
+                const s = Cv("input", e.inputNames, n),
+                    i = Cv("output", e.outputNames, r),
+                    o = s[0].shape[0];
+                va.assert(s.length === e.inputs.length, () => `LayersModel has ${e.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(e.inputNames)})`), va.assert(i.length === e.outputs.length, () => `LayersModel has ${e.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(e.outputNames)})`);
+                for (let t = 0; t < s.length; t++) va.assert(s[t].shape[0] === o, () => `Batch size mismatch: input ${e.inputNames[t]} has ${s[t].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);
+                for (let t = 0; t < i.length; t++) va.assert(i[t].shape[0] === o, () => `Batch size mismatch: output ${e.outputNames[t]} has ${i[t].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);
+                return {
+                    xs: s,
+                    ys: i
+                }
+            }
+
+            function Cv(e, t, n) {
+                if (n instanceof Ya) return [n];
+                if (Array.isArray(n)) return va.assert(n.length === t.length, () => `Received an array of ${n.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`), n; {
+                    const r = [];
+                    for (const a of t) {
+                        if (null == n[a]) throw new og(`The feature data generated by the dataset lacks the required ${e} key '${a}'.`);
+                        r.push(n[a])
+                    }
+                    return r
+                }
+            }
+            async function Tv(e, t, n) {
+                const r = null != n.batchesPerEpoch;
+                if (va.assert(null != e.optimizer, () => "You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."), va.assert(null != n, () => "For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."), va.assert(null != n.epochs && n.epochs > 0 && Number.isInteger(n.epochs), () => "For fitDataset(), config.epochs is expected to be a positive integer, but got " + n.epochs), va.assert(!r || n.batchesPerEpoch > 0 && Number.isInteger(n.batchesPerEpoch), () => "For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got " + n.batchesPerEpoch), va.assert(null == n.validationSplit, () => "`validationSplit` is not supported by `fitDataset()`. Use validationData instead."), e.isTraining) throw new Error("Cannot start training because another fit() call is ongoing.");
+                e.isTraining = !0;
+                try {
+                    const a = null != n.validationData;
+                    let s, i;
+                    if (a)
+                        if (Rv(n.validationData)) va.assert(null == n.validationBatches || n.validationBatches > 0 && Number.isInteger(n.validationBatches), () => "For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got " + n.validationBatches);
+                        else {
+                            const e = function (e) {
+                                if (3 === e.length) throw new ug("Validation with sample weights is not implemented yet.");
+                                return {
+                                    xs: e[0],
+                                    ys: e[1]
+                                }
+                            }(n.validationData);
+                            s = e.xs, i = e.ys
+                        } const o = e.makeTrainFunction(),
+                        u = e.getDedupedMetricsNames();
+                    let l;
+                    l = a ? u.slice().concat(u.map(e => "val_" + e)) : u.slice();
+                    const c = Lb(n.callbacks, n.yieldEvery),
+                        h = null == n.verbose ? 1 : n.verbose,
+                        {
+                            callbackList: p,
+                            history: d
+                        } = $b(c, h, n.epochs, null, null, function (e, t) {
+                            let n = null;
+                            null != t.batchesPerEpoch ? n = t.batchesPerEpoch : Number.isFinite(e.size) && (n = e.size);
+                            return n
+                        }(t, n), null, a, l);
+                    p.setModel(e), e.history = d, await p.onTrainBegin(), e.stopTraining_ = !1;
+                    let f = null == n.initialEpoch ? 0 : n.initialEpoch,
+                        m = await t.iterator();
+                    for (; f < n.epochs;) {
+                        const l = {};
+                        await p.onEpochBegin(f);
+                        let c = 0,
+                            h = 0;
+                        for (r || (m = await t.iterator()); !r || c < n.batchesPerEpoch;) {
+                            const t = await m.next();
+                            if (r && t.done) {
+                                console.warn(`You provided \`batchesPerEpoch\` as ${n.batchesPerEpoch}, but your dataset iterator ran out of data after ${c} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${n.batchesPerEpoch*n.epochs} batches). You may need to use the repeat() function when building your dataset.`);
+                                break
+                            }
+                            if (null != t.value) {
+                                const {
+                                    xs: r,
+                                    ys: a
+                                } = Nv(e, t.value), s = {};
+                                s.batch = h, s.size = r[0].shape[0], await p.onBatchBegin(h, s);
+                                const i = [];
+                                if (null != n.classWeight) {
+                                    const t = Iv(n.classWeight, e.outputNames);
+                                    for (let e = 0; e < t.length; ++e) i.push(await Av(a[e], null, t[e]))
+                                }
+                                const l = r.concat(a).concat(i),
+                                    d = o(l);
+                                au(l);
+                                for (let e = 0; e < u.length; ++e) {
+                                    const t = u[e],
+                                        n = d[e];
+                                    s[t] = n, su(n)
+                                }
+                                await p.onBatchEnd(h, s), Rb(s), h++, c++
+                            }
+                            if (r ? c >= n.batchesPerEpoch : t.done) {
+                                if (a) {
+                                    let t;
+                                    t = Rv(n.validationData) ? fg(await e.evaluateDataset(n.validationData, {
+                                        batches: n.validationBatches
+                                    })) : fg(e.evaluate(s, i, {
+                                        batchSize: null == n.validationBatchSize ? 32 : n.validationBatchSize,
+                                        verbose: 0
+                                    }));
+                                    for (let n = 0; n < e.metricsNames.length; ++n) l["val_" + e.metricsNames[n]] = t[n]
+                                }
+                                break
+                            }
+                            if (e.stopTraining_) break
+                        }
+                        if (await p.onEpochEnd(f, l), f++, e.stopTraining_) break
+                    }
+                    return await p.onTrainEnd(), await e.history.syncData(), e.history
+                } finally {
+                    e.isTraining = !1
+                }
+            }
+
+            function Rv(e) {
+                return "function" == typeof e.iterator
+            }
+
+            function _v(e) {
+                va.assert(e > 0 && Number.isInteger(e), () => "batchSize is required to be a positive integer, but got " + e)
+            }
+
+            function Fv(e, t, n) {
+                return null == e ? [null] : Array.isArray(e) ? e.map(e => ly(e, t, n - t)) : ly(e, t, n - t)
+            }
+
+            function Mv(e, t) {
+                return ru(() => null == e ? null : Array.isArray(e) ? e.map(e => Mv(e, t)) : yy(e, "int32" === t.dtype ? t : Ai(t, "int32")))
+            }
+
+            function Dv(e, t) {
+                const n = [];
+                let r = 0,
+                    a = null;
+                for (; r < e;) a = r + t, a >= e && (a = e), n.push([r, a]), r = a;
+                return n
+            }
+            async function Ov(e, t, n, r = {}) {
+                if (e.isTraining) throw new Error("Cannot start training because another fit() call is ongoing.");
+                let a, s, i, o, u, l, c, h, p;
+                e.isTraining = !0;
+                try {
+                    const d = null == r.batchSize ? 32 : r.batchSize;
+                    _v(d);
+                    const f = !1,
+                        m = await e.standardizeUserData(t, n, r.sampleWeight, r.classWeight, f, d);
+                    a = m[0], s = m[1], p = m[2];
+                    let g, y = !1;
+                    if (null != r.validationData && r.validationData.length > 0) {
+                        if (y = !0, 2 !== r.validationData.length) throw 3 === r.validationData.length ? new ug("validationData including sample weights is not supported yet.") : new og(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${r.validationData} is invalid.`);
+                        u = r.validationData[0], l = r.validationData[1];
+                        const t = !0,
+                            n = await e.standardizeUserData(u, l, null, null, t, d);
+                        c = n[0], h = n[1], g = c.concat(h)
+                    } else if (null != r.validationSplit && r.validationSplit > 0 && r.validationSplit < 1) {
+                        y = !0;
+                        const e = Math.floor(a[0].shape[0] * (1 - r.validationSplit)),
+                            t = a[0].shape[0];
+                        c = Fv(a, e, t), i = a, a = Fv(a, 0, e), h = Fv(s, e, t), o = s, s = Fv(s, 0, e), g = c.concat(h)
+                    } else null != r.validationSteps && (y = !0);
+                    const b = a.concat(s).concat(p);
+                    e.checkTrainableWeightsConsistency();
+                    const v = e.makeTrainFunction(),
+                        x = e.getDedupedMetricsNames();
+                    let w, k;
+                    y ? (e.makeTestFunction(), w = e.testFunction, k = x.slice().concat(x.map(e => "val_" + e))) : (w = null, g = [], k = x.slice());
+                    const S = Lb(r.callbacks, r.yieldEvery);
+                    return await async function (e, t, n, r, a, s, i, o, u, l, c, h, p, d, f) {
+                        null == a && (a = 32), null == s && (s = 1), null == c && (c = !0), null == p && (p = 0);
+                        let m = !1;
+                        if (null != u && null != l && (m = !0), null != f && (m = !0, null == d)) throw new og("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");
+                        const g = e.checkNumSamples(n, a, d, "steps_per_epoch");
+                        let y;
+                        null != g && (y = iy(0, g)), null == i && (i = 1);
+                        const {
+                            callbackList: b,
+                            history: v
+                        } = $b(o, i, s, p, g, d, a, m, h);
+                        b.setModel(e), e.history = v, await b.onTrainBegin(), e.stopTraining_ = !1;
+                        for (let i = p; i < s; ++i) {
+                            await b.onEpochBegin(i);
+                            const s = {};
+                            if (null != d) throw new ug("stepsPerEpoch mode is not implemented yet."); {
+                                if ("batch" === c) throw new ug("batch shuffling is not implemneted yet");
+                                c && va.shuffle(y);
+                                const i = Yh(y),
+                                    o = Dv(g, a);
+                                for (let c = 0; c < o.length; ++c) {
+                                    const h = {};
+                                    if (await b.onBatchBegin(c, h), ru(() => {
+                                            const p = o[c][0],
+                                                d = o[c][1],
+                                                f = ly(i, p, d - p);
+                                            h.batch = c, h.size = d - p;
+                                            const g = Mv(n, f),
+                                                y = t(g);
+                                            for (let e = 0; e < r.length; ++e) {
+                                                const t = r[e],
+                                                    n = y[e];
+                                                h[t] = n, su(n)
+                                            }
+                                            if (c === o.length - 1 && m) {
+                                                const t = e.testLoop(u, l, a);
+                                                for (let e = 0; e < r.length; ++e) {
+                                                    const n = r[e],
+                                                        a = t[e];
+                                                    su(a), s["val_" + n] = a
+                                                }
+                                            }
+                                        }), await b.onBatchEnd(c, h), Rb(h), e.stopTraining_) break
+                                }
+                                i.dispose()
+                            }
+                            if (await b.onEpochEnd(i, s), e.stopTraining_) break
+                        }
+                        return await b.onTrainEnd(), await e.history.syncData(), e.history
+                    }(e, v, b, x, d, r.epochs, r.verbose, S, w, g, r.shuffle, k, r.initialEpoch, null, null)
+                } finally {
+                    e.isTraining = !1, Pv(a, t), Pv(s, n), Pv(i, t), Pv(o, n), Pv(c, u), Pv(h, l), null != p && au(p)
+                }
+            }
+
+            function Lv(e) {
+                const t = [];
+                e instanceof Ya && (e = [e]);
+                for (let n = 0; n < e.length; ++n) {
+                    const r = e[n];
+                    if (1 === r.rank) t.push(uy(r, 1));
+                    else {
+                        if (0 === r.rank) throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");
+                        t.push(r)
+                    }
+                }
+                return t
+            }
+
+            function Pv(e, t) {
+                if (null == e) return;
+                const n = [];
+                if (t instanceof Ya) n.push(t.id);
+                else if (Array.isArray(t)) t.forEach(e => n.push(e.id));
+                else if (null != t)
+                    for (const e in t) {
+                        const r = t[e];
+                        n.push(r.id)
+                    }
+                const r = [];
+                if (e instanceof Ya) - 1 === n.indexOf(e.id) && r.push(e);
+                else if (Array.isArray(e)) e.forEach(e => {
+                    -1 === n.indexOf(e.id) && r.push(e)
+                });
+                else if (null != e)
+                    for (const t in e) {
+                        const a = e[t]; - 1 === n.indexOf(a.id) && r.push(a)
+                    }
+                r.forEach(e => {
+                    e.isDisposed || e.dispose()
+                })
+            }
+
+            function $v(e) {
+                return Array.isArray(e)
+            }
+
+            function Bv(e) {
+                return ! function (e) {
+                    return e instanceof Ya
+                }(e) && !$v(e)
+            }
+
+            function zv(e, t, n, r = !0, a = "") {
+                if (null == t || 0 === t.length) {
+                    if (null != e) {
+                        let t = !1;
+                        if ($v(e) && e.length > 0) t = !0;
+                        else if (Bv(e)) {
+                            for (const n in e)
+                                if (e.hasOwnProperty(n)) {
+                                    t = !0;
+                                    break
+                                }
+                        } else t = !0;
+                        if (t) throw new og(`Error when checking model ${a} expected no data, but got ${e}`)
+                    }
+                    return []
+                }
+                if (null == e) return t.map(e => null);
+                let s;
+                if (Bv(e)) {
+                    e = e, s = [];
+                    for (const n of t) {
+                        if (null == e[n]) throw new og(`No data provided for "${n}". Need data for each key in: ${t}`);
+                        s.push(e[n])
+                    }
+                } else if ($v(e)) {
+                    if ((e = e).length !== t.length) throw new og(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);
+                    s = e
+                } else {
+                    if (e = e, t.length > 1) throw new og(`The model ${a} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);
+                    s = [e]
+                }
+                if (s = Lv(s), null != n)
+                    for (let e = 0; e < t.length; ++e) {
+                        if (null == n[e]) continue;
+                        const i = s[e];
+                        if (i.shape.length !== n[e].length) throw new og(`Error when checking ${a}: expected ${t[e]} to have ${n[e].length} dimension(s). but got array with shape ${i.shape}`);
+                        for (let t = 0; t < n[e].length; ++t) {
+                            if (0 === t && !r) continue;
+                            const s = i.shape[t],
+                                o = n[e][t];
+                            if (null != o && o >= 0 && s !== o) throw new og(`${a} expected a batch of elements where each example has shape [${n[e].slice(1,n[e].length)}] (i.e.,tensor shape [*,${n[e].slice(1,n[e].length)}]) but the ${a} received an input with ${i.shape[0]} examples, each with shape [${i.shape.slice(1,i.shape.length)}] (tensor shape [${i.shape}])`)
+                        }
+                    }
+                return s
+            }
+
+            function Uv(e, t, n, r = !0, a = "") {
+                let s;
+                if (Array.isArray(e)) {
+                    if (e.length !== t.length) throw new og(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);
+                    s = e
+                } else {
+                    if (t.length > 1) throw new og(`The model expects ${t.length} ${a} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);
+                    s = [e]
+                }
+                if (null != n)
+                    for (let e = 0; e < t.length; ++e) {
+                        if (null == n[e]) continue;
+                        const i = s[e];
+                        if (i.shape.length !== n[e].length) throw new og(`Error when checking ${a}: expected ${t[e]} to have ${n[e].length} dimension(s), but got array with shape ${JSON.stringify(i.shape)}`);
+                        for (let s = 0; s < n[e].length; ++s) {
+                            if (0 === s && !r) continue;
+                            const o = i.shape[s],
+                                u = n[e][s];
+                            if (null != u && u !== o) throw new og(`Error when checking ${a}: expected ${t[e]} to have shape ${JSON.stringify(n[e])} but got array with shape ${JSON.stringify(i.shape)}.`)
+                        }
+                    }
+            }
+            var Wv = class extends Sv {
+                constructor(e) {
+                    super(e), this.isTraining = !1
+                }
+                summary(e, t, n = console.log) {
+                    if (!this.built) throw new og("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");
+                    lv(this, e, t, n)
+                }
+                compile(e) {
+                    if (null == e.loss && (e.loss = []), this.loss = e.loss, "string" == typeof e.optimizer) this.optimizer_ = function (e) {
+                        const t = {
+                            Adagrad: () => Ud.adagrad(.01),
+                            Adadelta: () => Ud.adadelta(1, .95, ag()),
+                            Adam: () => Ud.adam(.001, .9, .999, ag()),
+                            Adamax: () => Ud.adamax(.002, .9, .999, ag(), 0),
+                            RMSProp: () => Ud.rmsprop(.001, .9, 0, ag()),
+                            SGD: () => Ud.sgd(.01)
+                        };
+                        if (t.adagrad = t.Adagrad, t.adadelta = t.Adadelta, t.adam = t.Adam, t.adamax = t.Adamax, t.rmsprop = t.RMSProp, t.sgd = t.SGD, e in t) return t[e]();
+                        throw new og("Unknown Optimizer " + e)
+                    }(e.optimizer), this.isOptimizerOwned = !0;
+                    else {
+                        if (!(e.optimizer instanceof Fd)) throw new og("User-defined optimizer must be an instance of tf.Optimizer.");
+                        this.optimizer_ = e.optimizer, this.isOptimizerOwned = !1
+                    }
+                    let t = [];
+                    if (Array.isArray(e.loss) || "string" == typeof e.loss || "function" == typeof e.loss)
+                        if (Array.isArray(e.loss)) {
+                            if (e.loss.length !== this.outputs.length) throw new og(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);
+                            const n = e.loss;
+                            t = n.map(e => Yb(e))
+                        } else {
+                            const n = Yb(e.loss);
+                            this.outputs.forEach(e => {
+                                t.push(n)
+                            })
+                        }
+                    else {
+                        e.loss = e.loss;
+                        for (const t in e.loss)
+                            if (-1 === this.outputNames.indexOf(t)) throw new og(`Unknown entry in loss dictionary: "${t}". Only expected the following keys: ${this.outputNames}`);
+                        for (const n of this.outputNames) null == e.loss[n] && console.warn(`Output "${n}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${n} during training`), t.push(Yb(e.loss[n]))
+                    }
+                    this.lossFunctions = t, this.feedOutputNames = [], this.feedOutputShapes = [], this.feedLossFns = [];
+                    for (let e = 0; e < this.outputs.length; ++e) {
+                        const t = this.internalOutputShapes[e],
+                            n = this.outputNames[e];
+                        this.feedOutputNames.push(n), this.feedOutputShapes.push(t), this.feedLossFns.push(this.lossFunctions[e])
+                    }
+                    const n = [];
+                    this.metrics = e.metrics, this.metricsNames = ["loss"], this.metricsTensors = [], Zg("loss", () => {
+                        for (let e = 0; e < this.outputs.length; ++e) {
+                            if (-1 !== n.indexOf(e)) continue;
+                            const t = this.lossFunctions[e];
+                            this.outputs.length > 1 && (this.metricsTensors.push([t, e]), this.metricsNames.push(this.outputNames[e] + "_loss"))
+                        }
+                    });
+                    const r = function (e, t) {
+                            if (null == e || Array.isArray(e) && 0 === e.length) return t.map(e => []);
+                            let n;
+                            if ("string" == typeof e || "function" == typeof e) n = [e];
+                            else {
+                                if (!Array.isArray(e) && "object" != typeof e) throw new TypeError("Type of metrics argument not understood. Expected an string,function, Array, or Object, found: " + e);
+                                n = e
+                            }
+                            if (Array.isArray(n)) return t.map(e => n); {
+                                const e = [];
+                                for (const r of t) {
+                                    let t = n.hasOwnProperty(r) ? n[r] : [];
+                                    Array.isArray(t) || (t = [t]), e.push(t)
+                                }
+                                return e
+                            }
+                        }(e.metrics, this.outputNames),
+                        a = (e, t, n) => {
+                            this.outputNames.length > 1 && (t = this.outputNames[e] + "_" + t), this.metricsNames.push(t), this.metricsTensors.push([n, e])
+                        };
+                    Zg("metric", () => {
+                        for (let e = 0; e < this.outputs.length; ++e) {
+                            if (-1 !== n.indexOf(e)) continue;
+                            (t => {
+                                let n, r, s;
+                                for (const i of t) {
+                                    if ("string" == typeof i && -1 !== ["accuracy", "acc", "crossentropy", "ce"].indexOf(i)) {
+                                        const t = this.internalOutputShapes[e];
+                                        let a;
+                                        1 === t[t.length - 1] || this.lossFunctions[e] === Hb ? -1 !== ["accuracy", "acc"].indexOf(i) ? r = Xb : -1 !== ["crossentropy", "ce"].indexOf(i) && (r = tv) : this.lossFunctions[e] === Gb ? -1 !== ["accuracy", "acc"].indexOf(i) ? r = nv : -1 !== ["crossentropy", "ce"].indexOf(i) && (r = av) : -1 !== ["accuracy", "acc"].indexOf(i) ? r = Qb : -1 !== ["crossentropy", "ce"].indexOf(i) && (r = rv), -1 !== ["accuracy", "acc"].indexOf(i) ? a = "acc" : -1 !== ["crossentropy", "ce"].indexOf(i) && (a = "ce"), s = r, n = "" + a
+                                    } else {
+                                        const e = iv(i);
+                                        s = e, n = "" + ov(i)
+                                    }
+                                    let t;
+                                    Zg(n, () => {
+                                        t = s
+                                    }), a(e, n, t)
+                                }
+                            })(r[e])
+                        }
+                    }), this.collectedTrainableWeights = this.trainableWeights
+                }
+                checkTrainableWeightsConsistency() {
+                    null != this.collectedTrainableWeights && this.trainableWeights.length !== this.collectedTrainableWeights.length && console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")
+                }
+                evaluate(e, t, n = {}) {
+                    const r = null == n.batchSize ? 32 : n.batchSize;
+                    _v(r);
+                    const a = this.standardizeUserDataXY(e, t, !0, r);
+                    try {
+                        const s = a[0].concat(a[1]);
+                        this.makeTestFunction();
+                        const i = this.testFunction;
+                        return dg(this.testLoop(i, s, r, n.verbose, n.steps))
+                    } finally {
+                        Pv(a[0], e), Pv(a[1], t)
+                    }
+                }
+                async evaluateDataset(e, t) {
+                    return this.makeTestFunction(), async function (e, t, n) {
+                        const r = null != (n = n || {}).batches,
+                            a = e.testFunction;
+                        let s = [];
+                        if (n.verbose > 0) throw new ug("Verbose mode is not implemented yet.");
+                        va.assert(!r || n.batches > 0 && Number.isInteger(n.batches), () => "Test loop expects `batches` to be a positive integer, but received " + JSON.stringify(n.batches));
+                        const i = "function" == typeof t.next ? t : await t.iterator();
+                        let o = 0,
+                            u = 0;
+                        for (; !r || u < n.batches;) {
+                            const t = await i.next();
+                            if (s = ru(() => {
+                                    if (t.value) {
+                                        const {
+                                            xs: n,
+                                            ys: r
+                                        } = Nv(e, t.value), i = n.concat(r), l = ru(() => a(i));
+                                        if (au(i), 0 === u)
+                                            for (let e = 0; e < l.length; ++e) s.push(Ah(0));
+                                        const c = i[0].shape[0];
+                                        for (let e = 0; e < l.length; ++e) {
+                                            const t = l[e],
+                                                n = s[e];
+                                            s[e] = ru(() => gu(s[e], vu(c, t))), u > 0 && au(n)
+                                        }
+                                        au(l), o += c, ++u
+                                    }
+                                    return s
+                                }), t.done) {
+                                r && console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${n.batches} batches). You may need to use the repeat() function when building your dataset.`);
+                                break
+                            }
+                        }
+                        for (let e = 0; e < s.length; ++e) {
+                            const t = s[e];
+                            s[e] = bu(s[e], o), au(t)
+                        }
+                        return dg(s)
+                    }(this, e, t)
+                }
+                checkNumSamples(e, t, n, r = "steps") {
+                    let a;
+                    if (null != n) {
+                        if (a = null, null != t) throw new og(`If ${r} is set, batchSize must be null or undefined.Got batchSize = ${t}`)
+                    } else {
+                        if (null == e) throw new og(`Either the input data should have a defined shape, or ${r} shoud be specified.`);
+                        a = Array.isArray(e) ? e[0].shape[0] : e.shape[0]
+                    }
+                    return a
+                }
+                execute(e, t) {
+                    if (Array.isArray(t) && 0 === t.length) throw new og("`outputs` is an empty Array, which is not allowed.");
+                    const n = Array.isArray(t),
+                        r = n ? t : [t],
+                        a = this.retrieveSymbolicTensors(r),
+                        s = new gv;
+                    if (e instanceof Ya && (e = [e]), Array.isArray(e)) {
+                        if (e.length !== this.inputs.length) throw new og(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);
+                        for (let t = 0; t < this.inputs.length; ++t) s.add(this.inputs[t], e[t])
+                    } else
+                        for (const t of this.inputs) {
+                            const n = e[t.name];
+                            if (null == n) throw new og("No value is provided for the model's input " + t.name);
+                            s.add(t, n)
+                        }
+                    const i = vv(a, s);
+                    return n ? i : i[0]
+                }
+                retrieveSymbolicTensors(e) {
+                    const t = cg(null, e.length);
+                    let n = e.length;
+                    for (const r of this.layers) {
+                        const a = Array.isArray(r.output) ? r.output : [r.output],
+                            s = a.map(e => e.name);
+                        for (let r = 0; r < e.length; ++r) {
+                            const i = s.indexOf(e[r]);
+                            if (-1 !== i && (t[r] = a[i], n--), 0 === n) break
+                        }
+                        if (0 === n) break
+                    }
+                    if (n > 0) {
+                        const n = [];
+                        throw t.forEach((t, r) => {
+                            null == t && n.push(e[r])
+                        }), new og("Cannot find SymbolicTensors for output name(s): " + JSON.stringify(n))
+                    }
+                    return t
+                }
+                predictLoop(e, t = 32, n = !1) {
+                    return ru(() => {
+                        const r = this.checkNumSamples(e);
+                        if (n) throw new ug("Verbose predictLoop() is not implemented yet.");
+                        const a = Dv(r, t),
+                            s = this.outputs.map(e => []);
+                        for (let t = 0; t < a.length; ++t) {
+                            ru(() => {
+                                const n = a[t][0],
+                                    r = a[t][1],
+                                    s = Fv(e, n, r),
+                                    i = [];
+                                if (Array.isArray(s))
+                                    for (let e = 0; e < s.length; ++e) i.push({
+                                        key: this.inputs[e],
+                                        value: s[e]
+                                    });
+                                else i.push({
+                                    key: this.inputs[0],
+                                    value: s
+                                });
+                                const o = new gv(i);
+                                return vv(this.outputs, o)
+                            }).forEach((e, t) => s[t].push(e))
+                        }
+                        return dg(s.map(e => Xu(e, 0)))
+                    })
+                }
+                predict(e, t = {}) {
+                    const n = Lv(e);
+                    Uv(n, this.inputNames, this.feedInputShapes, !1);
+                    try {
+                        const r = null == t.batchSize ? 32 : t.batchSize;
+                        return _v(r), this.predictLoop(n, r)
+                    } finally {
+                        Pv(n, e)
+                    }
+                }
+                predictOnBatch(e) {
+                    Uv(e, this.inputNames, this.feedInputShapes, !0);
+                    const t = (Array.isArray(e) ? e[0] : e).shape[0];
+                    return this.predictLoop(e, t)
+                }
+                standardizeUserDataXY(e, t, n = !0, r) {
+                    if (null == this.optimizer_) throw new ig("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");
+                    const a = [];
+                    for (let e = 0; e < this.feedOutputShapes.length; ++e) {
+                        const t = this.feedOutputShapes[e];
+                        this.feedLossFns[e] === Gb ? a.push(t.slice(0, t.length - 1).concat([1])) : a.push(t)
+                    }
+                    if (function (e, t, n) {
+                            const r = wg(e.map(e => e.shape[0]));
+                            r.sort();
+                            const a = wg(t.map(e => e.shape[0]));
+                            if (a.sort(), r.length > 1) throw new og("All input Tensors (x) should have the same number of samples. Got array shapes: " + JSON.stringify(e.map(e => e.shape)));
+                            if (a.length > 1) throw new og("All target Tensors (y) should have the same number of samples. Got array shapes: " + JSON.stringify(t.map(e => e.shape)));
+                            if (r.length > 0 && a.length > 0 && !va.arraysEqual(r, a)) throw new og(`Input Tensors should have the same number of samples as target Tensors. Found ${r[0]} input sample(s) and ${a[0]} target sample(s).`)
+                        }(e = zv(e, this.feedInputNames, this.feedInputShapes, !1, "input"), t = zv(t, this.feedOutputNames, a, !1, "target")), function (e, t, n) {
+                            const r = [Ub, Hb, jb];
+                            for (let a = 0; a < e.length; ++a) {
+                                const s = e[a],
+                                    i = t[a],
+                                    o = n[a];
+                                if (null != i) {
+                                    if (i === jb && 1 === s.shape[s.shape.length - 1]) throw new og(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);
+                                    if (-1 !== r.indexOf(i)) {
+                                        const e = s.shape.slice(1),
+                                            t = o.slice(1);
+                                        for (let n = 0; n < e.length; ++n) {
+                                            const r = e[n],
+                                                a = t[n];
+                                            if (null != a && r !== a) throw new og(`A target Tensor with shape ${s.shape} was passed for an output of shape ${o}, while using a loss function that expects targets to have the same shape as the output.`)
+                                        }
+                                    }
+                                }
+                            }
+                        }(t, this.feedLossFns, this.feedOutputShapes), this.stateful && null != r && r > 0 && e[0].shape[0] % r != 0) throw new og(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${r}. Found: ${e[0].shape[0]} sample(s).`);
+                    return [e, t]
+                }
+                async standardizeUserData(e, t, n, r, a = !0, s) {
+                    const [i, o] = this.standardizeUserDataXY(e, t, a, s);
+                    if (null != n) throw new Error("sample weight is not supported yet.");
+                    let u = null;
+                    if (null != r) {
+                        const e = Iv(r, this.outputNames);
+                        u = [];
+                        for (let t = 0; t < e.length; ++t) u.push(await Av(o[t], null, e[t]))
+                    }
+                    return [i, o, u]
+                }
+                testLoop(e, t, n, r = 0, a) {
+                    return ru(() => {
+                        const s = this.checkNumSamples(t, n, a, "steps"),
+                            i = [];
+                        if (r > 0) throw new ug("Verbose mode is not implemented yet.");
+                        if (null != a) throw new ug("steps mode in testLoop() is not implemented yet"); {
+                            const r = Dv(s, n),
+                                a = Yh(iy(0, s));
+                            for (let n = 0; n < r.length; ++n) {
+                                const s = r[n][0],
+                                    o = r[n][1],
+                                    u = ly(a, s, o - s),
+                                    l = Mv(t, u),
+                                    c = e(l);
+                                if (0 === n)
+                                    for (let e = 0; e < c.length; ++e) i.push(Ah(0));
+                                for (let e = 0; e < c.length; ++e) {
+                                    const t = c[e];
+                                    i[e] = gu(i[e], vu(o - s, t))
+                                }
+                            }
+                            for (let e = 0; e < i.length; ++e) i[e] = bu(i[e], s)
+                        }
+                        return i
+                    })
+                }
+                getDedupedMetricsNames() {
+                    const e = this.metricsNames,
+                        t = [];
+                    for (let n = 0; n < e.length; ++n) {
+                        const r = e[n];
+                        let a = r;
+                        if (pg(e, r) > 1) {
+                            a += "_" + pg(e.slice(0, n), r)
+                        }
+                        t.push(a)
+                    }
+                    return t
+                }
+                makeTrainFunction() {
+                    return e => {
+                        const t = [],
+                            n = e.slice(0, this.inputs.length),
+                            r = e.slice(this.inputs.length, this.inputs.length + this.outputs.length),
+                            a = e.slice(this.inputs.length + this.outputs.length, this.inputs.length + 2 * this.outputs.length),
+                            s = [],
+                            i = this.collectedTrainableWeights.map(e => e.read());
+                        return [this.optimizer_.minimize(() => {
+                            const e = [];
+                            for (let t = 0; t < this.inputs.length; ++t) e.push({
+                                key: this.inputs[t],
+                                value: n[t]
+                            });
+                            const i = new gv(e),
+                                o = vv(this.outputs, i, {
+                                    training: !0
+                                });
+                            let u;
+                            for (let e = 0; e < this.lossFunctions.length; ++e) {
+                                let n = (0, this.lossFunctions[e])(r[e], o[e]);
+                                null != a[e] && (n = Ev(n, a[e]));
+                                const s = Lc(n);
+                                t.push(s), u = 0 === e ? n : gu(u, n)
+                            }
+                            for (let e = 0; e < this.metricsTensors.length; ++e) {
+                                let n;
+                                if (this.outputs.length > 1 && e < this.outputs.length) n = t[e];
+                                else {
+                                    const t = this.metricsTensors[e][0],
+                                        a = this.metricsTensors[e][1];
+                                    n = Lc(t(r[a], o[a]))
+                                }
+                                su(n), s.push(n)
+                            }
+                            return u = Lc(u), this.calculateLosses().forEach(e => {
+                                u = gu(u, e)
+                            }), u
+                        }, !0, i)].concat(s)
+                    }
+                }
+                makeTestFunction() {
+                    this.testFunction = e => ru(() => {
+                        const t = [];
+                        let n;
+                        const r = e.slice(0, this.inputs.length),
+                            a = e.slice(this.inputs.length, this.inputs.length + this.outputs.length),
+                            s = [];
+                        for (let e = 0; e < this.inputs.length; ++e) s.push({
+                            key: this.inputs[e],
+                            value: r[e]
+                        });
+                        const i = new gv(s),
+                            o = vv(this.outputs, i);
+                        for (let e = 0; e < this.lossFunctions.length; ++e) {
+                            const r = this.lossFunctions[e],
+                                s = Lc(r(a[e], o[e]));
+                            n = 0 === e ? s : gu(n, s), t.push(n)
+                        }
+                        for (let e = 0; e < this.metricsTensors.length; ++e) {
+                            const n = this.metricsTensors[e][0],
+                                r = this.metricsTensors[e][1],
+                                s = Lc(n(a[r], o[r]));
+                            t.push(s)
+                        }
+                        return t
+                    })
+                }
+                async fit(e, t, n = {}) {
+                    return Ov(this, e, t, n)
+                }
+                async fitDataset(e, t) {
+                    return Tv(this, e, t)
+                }
+                async trainOnBatch(e, t) {
+                    const n = await this.standardizeUserData(e, t),
+                        r = n[0],
+                        a = n[1],
+                        s = this.makeTrainFunction()(r.concat(a)),
+                        i = [];
+                    for (const e of s) {
+                        const t = await e.data();
+                        i.push(t[0])
+                    }
+                    return au(s), Pv(n[0], e), Pv(n[1], t), dg(i)
+                }
+                getNamedWeights(e) {
+                    const t = [],
+                        n = null != e && e.trainableOnly,
+                        r = n ? this.trainableWeights : this.weights,
+                        a = this.getWeights(n);
+                    for (let e = 0; e < r.length; ++e) n && !r[e].trainable || t.push({
+                        name: r[e].originalName,
+                        tensor: a[e]
+                    });
+                    return t
+                }
+                set stopTraining(e) {
+                    this.stopTraining_ = e
+                }
+                get stopTraining() {
+                    return this.stopTraining_
+                }
+                get optimizer() {
+                    return this.optimizer_
+                }
+                set optimizer(e) {
+                    this.optimizer_ !== e && (this.optimizer_ = e, this.isOptimizerOwned = !1)
+                }
+                dispose() {
+                    const e = super.dispose();
+                    if (0 === e.refCountAfterDispose && null != this.optimizer && this.isOptimizerOwned) {
+                        const t = tu().numTensors;
+                        this.optimizer_.dispose(), e.numDisposedVariables += t - tu().numTensors
+                    }
+                    return e
+                }
+                getLossIdentifiers() {
+                    let e;
+                    if ("string" == typeof this.loss) e = mg(this.loss);
+                    else if (Array.isArray(this.loss)) {
+                        for (const e of this.loss)
+                            if ("string" != typeof e) throw new Error("Serialization of non-string loss is not supported.");
+                        e = this.loss.map(e => mg(e))
+                    } else {
+                        const t = Object.keys(this.loss);
+                        e = {};
+                        const n = this.loss;
+                        for (const r of t) {
+                            if ("string" != typeof n[r]) throw new Error("Serialization of non-string loss is not supported.");
+                            e[r] = mg(n[r])
+                        }
+                    }
+                    return e
+                }
+                getMetricIdentifiers() {
+                    if ("string" == typeof this.metrics || "function" == typeof this.metrics) return [mg(ov(this.metrics))];
+                    if (Array.isArray(this.metrics)) return this.metrics.map(e => mg(ov(e))); {
+                        const e = {};
+                        for (const t in this.metrics) e[t] = mg(ov(this.metrics[t]));
+                        return e
+                    }
+                }
+                getTrainingConfig() {
+                    return {
+                        loss: this.getLossIdentifiers(),
+                        metrics: this.getMetricIdentifiers(),
+                        optimizer_config: {
+                            class_name: this.optimizer.getClassName(),
+                            config: this.optimizer.getConfig()
+                        }
+                    }
+                }
+                loadTrainingConfig(e) {
+                    if (null != e.weighted_metrics) throw new Error("Loading weight_metrics is not supported yet.");
+                    if (null != e.loss_weights) throw new Error("Loading loss_weights is not supported yet.");
+                    if (null != e.sample_weight_mode) throw new Error("Loading sample_weight_mode is not supported yet.");
+                    const t = Bb(fv(e.optimizer_config));
+                    let n, r;
+                    if ("string" == typeof e.loss) n = gg(e.loss);
+                    else if (Array.isArray(e.loss)) n = e.loss.map(e => gg(e));
+                    else if (null != e.loss) {
+                        n = {};
+                        for (const t in e.loss) n[t] = gg(e.loss[t])
+                    }
+                    if (Array.isArray(e.metrics)) r = e.metrics.map(e => gg(e));
+                    else if (null != e.metrics) {
+                        r = {};
+                        for (const t in e.metrics) r[t] = gg(e.metrics[t])
+                    }
+                    this.compile({
+                        loss: n,
+                        metrics: r,
+                        optimizer: t
+                    })
+                }
+                async save(e, t) {
+                    if ("string" == typeof e) {
+                        const t = Ci.getSaveHandlers(e);
+                        if (0 === t.length) throw new og(`Cannot find any save handlers for URL '${e}'`);
+                        if (t.length > 1) throw new og(`Found more than one (${t.length}) save handlers for URL '${e}'`);
+                        e = t[0]
+                    }
+                    if (null == e.save) throw new og("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");
+                    const n = await Ci.encodeWeights(this.getNamedWeights(t)),
+                        r = {
+                            modelTopology: this.toJSON(null, !1),
+                            format: "layers-model",
+                            generatedBy: "TensorFlow.js tfjs-layers v" + mv,
+                            convertedBy: null
+                        };
+                    if (null != t && t.includeOptimizer && null != this.optimizer) {
+                        r.trainingConfig = this.getTrainingConfig();
+                        const e = "optimizer",
+                            {
+                                data: t,
+                                specs: a
+                            } = await Ci.encodeWeights(await this.optimizer.getWeights(), e);
+                        n.specs.push(...a), n.data = Ci.concatenateArrayBuffers([n.data, t])
+                    }
+                    if (null != this.userDefinedMetadata) {
+                        const e = !0;
+                        uv(this.userDefinedMetadata, this.name, e), r.userDefinedMetadata = this.userDefinedMetadata
+                    }
+                    return r.weightData = n.data, r.weightSpecs = n.specs, e.save(r)
+                }
+                setUserDefinedMetadata(e) {
+                    uv(e, this.name), this.userDefinedMetadata = e
+                }
+                getUserDefinedMetadata() {
+                    return this.userDefinedMetadata
+                }
+            };
+            Wv.className = "Model", Fo.registerClass(Wv);
+            var Vv = class extends Wv {};
+            async function jv(e, t) {
+                "modelTopology" in e || (e = {
+                    modelTopology: e
+                });
+                let n = (e = e).modelTopology;
+                null != n.model_config && (n = n.model_config);
+                const r = Bb(fv(n), t);
+                if (null != e.weightsManifest) {
+                    const t = await Ci.loadWeights(e.weightsManifest, e.pathPrefix, r.weights.map(e => e.originalName)),
+                        n = {};
+                    for (const e of r.weights) n[e.originalName] = t[e.originalName];
+                    r.loadWeights(n), au(t)
+                }
+                return r
+            }
+            async function Gv(e, t) {
+                if (null == t && (t = {}), "string" == typeof e) {
+                    const n = Ci.getLoadHandlers(e, t);
+                    if (0 === n.length) n.push(Ci.browserHTTPRequest(e, t));
+                    else if (n.length > 1) throw new og(`Found more than one (${n.length}) load handlers for URL '${e}'`);
+                    e = n[0]
+                }
+                return async function (e, t, n) {
+                    null == n && (n = {});
+                    if (null == e.load) throw new og("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");
+                    const r = await e.load();
+                    let a = r.modelTopology;
+                    null != a.model_config && (a = a.model_config);
+                    const s = null == n.strict || n.strict,
+                        i = null != r.weightData && null != r.weightSpecs && s,
+                        o = Bb(fv(a), t, i),
+                        u = r.trainingConfig;
+                    null != u && o.loadTrainingConfig(u);
+                    null != r.userDefinedMetadata && o.setUserDefinedMetadata(r.userDefinedMetadata);
+                    if (null != r.weightData) {
+                        if (null == r.weightSpecs) throw new og("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");
+                        const {
+                            modelWeights: e,
+                            optimizerWeights: t
+                        } = function (e, t) {
+                            const n = Ci.decodeWeights(e, t),
+                                r = {},
+                                a = [];
+                            return t.forEach(e => {
+                                "optimizer" === e.group ? a.push({
+                                    name: e.name,
+                                    tensor: n[e.name]
+                                }) : r[e.name] = n[e.name]
+                            }), {
+                                modelWeights: r,
+                                optimizerWeights: a
+                            }
+                        }(r.weightData, r.weightSpecs);
+                        o.loadWeights(e, s), null != o.optimizer && t.length > 0 && await o.optimizer.setWeights(t), au(e), au(t.map(e => e.tensor))
+                    }
+                    return o
+                }(e, void 0, t)
+            }
+            Vv.className = "Functional", Fo.registerClass(Vv);
+            var Hv = class extends Wv {
+                constructor(e) {
+                    if (super({
+                            inputs: [],
+                            outputs: []
+                        }), e = e || {}, this.trainable = !0, this.built = !1, this.name = null != e.name ? e.name : cb("sequential_"), null != e.layers)
+                        for (const t of e.layers) this.add(t)
+                }
+                checkShape(e) {
+                    if (e.inboundNodes[0].outputTensors[0].shape.some(e => e < 0)) throw new og(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)
+                }
+                add(e) {
+                    const t = e instanceof Hv || e instanceof Wv;
+                    let n;
+                    if (t) {
+                        if (n = e, 1 !== n.outputs.length) throw new og("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+                        if (1 !== n.inputs.length) throw new og("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")
+                    }
+                    if (0 === this.outputs.length) {
+                        if (0 === e.inboundNodes.length) {
+                            if (null == e.batchInputShape) throw new og("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");
+                            const t = Cb({
+                                batchShape: e.batchInputShape,
+                                dtype: e.dtype,
+                                name: e.name + "_input"
+                            });
+                            e.apply(t)
+                        }
+                        if (t) this.outputs = n.outputs, this.inputs = n.inputs;
+                        else {
+                            if (1 !== e.inboundNodes.length) throw new og(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${e.name} which has ${e.inboundNodes.length} pre-existing inbound connections.`);
+                            if (1 !== e.inboundNodes[0].outputTensors.length) throw new og("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+                            this.checkShape(e), this.outputs = [e.inboundNodes[0].outputTensors[0]], this.inputs = function e(t, n, r) {
+                                if ((null == n || null != r && r > 0) && (n = t.sourceLayer, r = t.nodeIndex), 0 === n.inboundNodes.length) return [t]; {
+                                    const t = n.inboundNodes[r];
+                                    if (0 === t.inboundLayers.length) return t.inputTensors; {
+                                        const n = [];
+                                        for (let r = 0; r < t.inboundLayers.length; r++) {
+                                            const a = e(t.inputTensors[r], t.inboundLayers[r], t.nodeIndices[r]);
+                                            for (const e of a) - 1 === n.indexOf(e) && n.push(e)
+                                        }
+                                        return n
+                                    }
+                                }
+                            }(this.outputs[0])
+                        }
+                        this.inboundNodes = [], new kb({
+                            outboundLayer: this,
+                            inboundLayers: [],
+                            nodeIndices: [],
+                            tensorIndices: [],
+                            inputTensors: this.inputs,
+                            outputTensors: this.outputs,
+                            inputMasks: cg(null, this.inputs.length),
+                            outputMasks: [null],
+                            inputShapes: this.inputs.map(e => e.shape),
+                            outputShapes: this.outputs[0].shape
+                        })
+                    } else {
+                        const t = e.apply(this.outputs[0]);
+                        if (Array.isArray(t)) throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+                        this.checkShape(e), this.outputs = [t], this.inboundNodes[0].outputTensors = this.outputs, this.inboundNodes[0].outputShapes = [this.outputs[0].shape]
+                    }
+                    this.layers.push(e), this.built = !1
+                }
+                pop() {
+                    if (0 === this.layers.length) throw new TypeError("There are no layers in the model.");
+                    if (this.layers.pop(), 0 === this.layers.length) this.outputs = [], this.inboundNodes = [], this.outboundNodes = [];
+                    else {
+                        const e = this.layers.length - 1;
+                        this.layers[e].outboundNodes = [], this.outputs = [this.layers[e].output], this.inboundNodes[0].outputTensors = this.outputs, this.inboundNodes[0].outputShapes = [this.outputs[0].shape]
+                    }
+                }
+                call(e, t) {
+                    return null == this.model && this.build(), this.model.call(e, t)
+                }
+                build(e) {
+                    if (fb(e), 0 === this.inputs.length || 0 === this.outputs.length) throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");
+                    this.model = new Wv({
+                        inputs: this.inputs,
+                        outputs: this.outputs[0],
+                        name: this.name + "_model"
+                    }), this.model.trainable = this.trainable, this.supportsMasking = this.model.supportsMasking, this.inputLayers = this.model.inputLayers, this.inputLayersNodeIndices = this.model.inputLayersNodeIndices, this.inputLayersTensorIndices = this.model.inputLayersTensorIndices, this.outputLayers = this.model.outputLayers, this.outputLayersNodeIndices = this.model.outputLayersNodeIndices, this.outputLayersTensorIndices = this.model.outputLayersTensorIndices, this.nodesByDepth = this.model.nodesByDepth, this.containerNodes = this.model.containerNodes, this.outputNames = this.model.outputNames, this.inputNames = this.model.inputNames, this.built = !0
+                }
+                countParams() {
+                    return this.built || this.build(), super.countParams()
+                }
+                summary(e, t, n = console.log) {
+                    this.built || this.build(), super.summary(e, t, n)
+                }
+                setWeights(e) {
+                    null == this.model && this.build(), this.model.setWeights(e)
+                }
+                evaluate(e, t, n = {}) {
+                    if (!this.built) throw new ig("The model needs to be compiled before being used.");
+                    return this.model.evaluate(e, t, n)
+                }
+                async evaluateDataset(e, t) {
+                    if (!this.built) throw new ig("The model needs to be compiled before being used.");
+                    return this.model.evaluateDataset(e, t)
+                }
+                predict(e, t = {}) {
+                    return null == this.model && this.build(), this.model.predict(e, t)
+                }
+                predictOnBatch(e) {
+                    return null == this.model && this.build(), this.model.predictOnBatch(e)
+                }
+                compile(e) {
+                    this.build(), this.model.compile(e), this.optimizer_ = this.model.optimizer, this.isOptimizerOwned = this.model.isOptimizerOwned, this.loss = this.model.loss, this.metrics = this.model.metrics, this.metricsTensors = this.model.metricsTensors, this.metricsNames = this.model.metricsNames
+                }
+                get optimizer() {
+                    return null == this.model ? void 0 : this.model.optimizer
+                }
+                set optimizer(e) {
+                    this.model.optimizer = e
+                }
+                async fit(e, t, n = {}) {
+                    if (!this.built) throw new ig("The model needs to be compiled before being used.");
+                    return this.model.fit(e, t, n)
+                }
+                async fitDataset(e, t) {
+                    if (!this.built) throw new ig("The model needs to be compiled before being used.");
+                    return this.model.fitDataset(e, t)
+                }
+                async trainOnBatch(e, t) {
+                    return this.model.trainOnBatch(e, t)
+                }
+                static fromConfig(e, t, n = {}, r = !1) {
+                    let a, s = {};
+                    if (t instanceof Array) {
+                        if (null == t[0].className || "Merge" === t[0].className) throw new og("Legacy serialization format not supported yet.");
+                        a = t
+                    } else va.assert(null != t.layers, () => "When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."), a = t.layers, delete t.layers, s = t;
+                    const i = new e(s);
+                    if (!(i instanceof Hv)) throw new ug("Sequential.fromConfig called on non-Sequential input: " + i);
+                    for (const e of a) {
+                        const t = Bb(e, void 0, r);
+                        r && t.setFastWeightInitDuringBuild(!0), i.add(t)
+                    }
+                    return i
+                }
+                set stopTraining(e) {
+                    if (null == this.model) throw new og("Cannot set the stopTraining property of a sequential model before it is compiled.");
+                    this.model.stopTraining = e
+                }
+                get stopTraining() {
+                    if (null == this.model) throw new og("Cannot get the stopTraining property of a sequential model before it is compiled.");
+                    return this.model.stopTraining
+                }
+                getConfig() {
+                    const e = [];
+                    for (const t of this.layers) {
+                        const n = {};
+                        n.className = t.getClassName(), n.config = t.getConfig(), e.push(n)
+                    }
+                    return {
+                        name: this.name,
+                        layers: e
+                    }
+                }
+            };
+
+            function qv(e) {
+                return new Wv(e)
+            }
+
+            function Kv(e) {
+                return new Hv(e)
+            }
+
+            function Yv(e, t) {
+                return null == t && (t = {}), Gv(e, t)
+            }
+
+            function Xv(e) {
+                return Cb(e)
+            }
+
+            function Qv(e, t) {
+                Pb.registerCallbackConstructor(e, t)
+            }
+            Hv.className = "Sequential", Fo.registerClass(Hv);
+            var Zv = class extends Fo.Serializable {
+                    getConfig() {
+                        return {}
+                    }
+                },
+                Jv = class extends Zv {
+                    apply(e, t = 1) {
+                        return function (e, t = 1) {
+                            if (1 !== t) throw new ug(`Support for alpha values other than 1 (${t}) is not implemented yet.`);
+                            return Ll(e)
+                        }(e, t)
+                    }
+                };
+            Jv.className = "elu", Fo.registerClass(Jv);
+            var ex = class extends Zv {
+                apply(e) {
+                    return Eh(e)
+                }
+            };
+            ex.className = "selu", Fo.registerClass(ex);
+            var tx = class extends Zv {
+                apply(e) {
+                    return gh(e)
+                }
+            };
+            tx.className = "relu", Fo.registerClass(tx);
+            var nx = class extends Zv {
+                apply(e) {
+                    return ru(() => Uc(6, gh(e)))
+                }
+            };
+            nx.className = "relu6", Fo.registerClass(nx);
+            var rx = class extends Zv {
+                apply(e) {
+                    return e
+                }
+            };
+            rx.className = "linear", Fo.registerClass(rx);
+            var ax = class extends Zv {
+                apply(e) {
+                    return Qu(e)
+                }
+            };
+            ax.className = "sigmoid", Fo.registerClass(ax);
+            var sx = class extends Zv {
+                apply(e) {
+                    return function (e) {
+                        return ru(() => {
+                            const t = gu(.5, vu(.2, e));
+                            return cl(t, 0, 1)
+                        })
+                    }(e)
+                }
+            };
+            sx.className = "hardSigmoid", Fo.registerClass(sx);
+            var ix = class extends Zv {
+                apply(e) {
+                    return dc(e)
+                }
+            };
+            ix.className = "softplus", Fo.registerClass(ix);
+            var ox = class extends Zv {
+                apply(e) {
+                    return function (e) {
+                        return ru(() => bu(e, gu(xu(e), 1)))
+                    }(e)
+                }
+            };
+            ox.className = "softsign", Fo.registerClass(ox);
+            var ux = class extends Zv {
+                apply(e) {
+                    return Ju(e)
+                }
+            };
+            ux.className = "tanh", Fo.registerClass(ux);
+            var lx = class extends Zv {
+                apply(e, t = -1) {
+                    return Lh(e, t)
+                }
+            };
+            lx.className = "softmax", Fo.registerClass(lx);
+            var cx = class extends Zv {
+                apply(e, t = -1) {
+                    return bc(e, t)
+                }
+            };
+            cx.className = "logSoftmax", Fo.registerClass(cx);
+            var hx = class extends Zv {
+                apply(e, t = 1) {
+                    return ru(() => vu(Qu(vu(e, t)), e))
+                }
+            };
+            hx.className = "swish", Fo.registerClass(hx);
+            var px = class extends Zv {
+                apply(e) {
+                    return ru(() => vu(e, Ju(dc(e))))
+                }
+            };
+
+            function dx(e) {
+                return e.getClassName()
+            }
+
+            function fx(e, t = {}) {
+                return vg(e, Fo.SerializationMap.getMap().classNameMap, t, "activation")
+            }
+
+            function mx(e) {
+                if (null == e) {
+                    const e = {
+                        className: "linear",
+                        config: {}
+                    };
+                    return fx(e)
+                }
+                if ("string" == typeof e) {
+                    const t = {};
+                    return t.className = e, t.config = {}, fx(t)
+                }
+                return e instanceof Zv ? e : fx(e)
+            }
+
+            function gx(e) {
+                if (null != e && "object" != typeof e) throw new Error("Argument to L1L2 regularizer's constructor is expected to be an object, but received: " + e)
+            }
+            px.className = "mish", Fo.registerClass(px);
+            var yx = class extends Fo.Serializable {},
+                bx = class extends yx {
+                    constructor(e) {
+                        super(), gx(e), this.l1 = null == e || null == e.l1 ? .01 : e.l1, this.l2 = null == e || null == e.l2 ? .01 : e.l2, this.hasL1 = 0 !== this.l1, this.hasL2 = 0 !== this.l2
+                    }
+                    apply(e) {
+                        return ru(() => {
+                            let t = Pc([1]);
+                            return this.hasL1 && (t = gu(t, yc(vu(this.l1, xu(e))))), this.hasL2 && (t = gu(t, yc(vu(this.l2, by(e))))), qu(t, [])
+                        })
+                    }
+                    getConfig() {
+                        return {
+                            l1: this.l1,
+                            l2: this.l2
+                        }
+                    }
+                    static fromConfig(e, t) {
+                        return new e({
+                            l1: t.l1,
+                            l2: t.l2
+                        })
+                    }
+                };
+            bx.className = "L1L2", Fo.registerClass(bx);
+            var vx = {
+                l1l2: "L1L2"
+            };
+
+            function xx(e) {
+                return bg(e)
+            }
+
+            function wx(e, t = {}) {
+                return vg(e, Fo.SerializationMap.getMap().classNameMap, t, "regularizer")
+            }
+
+            function kx(e) {
+                if (null == e) return null;
+                if ("string" == typeof e) {
+                    return wx({
+                        className: e in vx ? vx[e] : e,
+                        config: {}
+                    })
+                }
+                return e instanceof yx ? e : wx(e)
+            }
+            var Sx = class extends Ib {
+                constructor(e) {
+                    super(null == e ? {} : e), this.supportsMasking = !0, null != e && (this.maxValue = e.maxValue)
+                }
+                call(e, t) {
+                    e = db(e);
+                    let n = gh(e);
+                    return null != this.maxValue && (n = cl(n, 0, this.maxValue)), n
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = {
+                            maxValue: this.maxValue
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Sx.className = "ReLU", Fo.registerClass(Sx);
+            var Ix = class extends Ib {
+                constructor(e) {
+                    super(null == e ? {} : e), this.DEFAULT_ALPHA = .3, null == e && (e = {}), this.alpha = null == e.alpha ? this.DEFAULT_ALPHA : e.alpha
+                }
+                call(e, t) {
+                    const n = db(e);
+                    return Zl(n, this.alpha)
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = {
+                            alpha: this.alpha
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Ix.className = "LeakyReLU", Fo.registerClass(Ix);
+            var Ax = class extends Ib {
+                constructor(e) {
+                    if (super(null == e ? {} : e), this.DEFAULT_ALPHA_INITIALIZER = "zeros", null == e && (e = {}), this.supportsMasking = !0, this.alphaInitializer = jy(e.alphaInitializer || this.DEFAULT_ALPHA_INITIALIZER), this.alphaRegularizer = kx(e.alphaRegularizer), this.alphaConstraint = Lg(e.alphaConstraint), null == e.sharedAxes) this.sharedAxes = null;
+                    else if (Array.isArray(e.sharedAxes)) this.sharedAxes = e.sharedAxes;
+                    else {
+                        if ("number" != typeof e.sharedAxes) throw new og("Expected sharedAxes to be a number or an array of numbers, but got " + e.sharedAxes);
+                        this.sharedAxes = [e.sharedAxes]
+                    }
+                }
+                build(e) {
+                    const t = (e = fb(e)).slice(1);
+                    if (null != this.sharedAxes)
+                        for (const e of this.sharedAxes) t[e - 1] = 1;
+                    this.alpha = this.addWeight("alpha", t, "float32", this.alphaInitializer, this.alphaRegularizer, !0, this.alphaConstraint);
+                    const n = {};
+                    if (null != this.sharedAxes)
+                        for (let t = 1; t < e.length; ++t) n[t] = e[t];
+                    this.inputSpec = [new vb({
+                        ndim: e.length,
+                        axes: n
+                    })], this.built = !0
+                }
+                call(e, t) {
+                    return e = db(e), sh(e, this.alpha.read())
+                }
+                getConfig() {
+                    const e = {
+                            alphaInitializer: Vy(this.alphaInitializer),
+                            alphaRegularizer: xx(this.alphaRegularizer),
+                            alphaConstraint: Dg(this.alphaConstraint),
+                            sharedAxes: this.sharedAxes
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Ax.className = "PReLU", Fo.registerClass(Ax);
+            var Ex = class extends Ib {
+                constructor(e) {
+                    if (super(null == e ? {} : e), this.DEFAULT_ALPHA = 1, null == e && (e = {}), null != e.alpha && e.alpha !== this.DEFAULT_ALPHA) throw new ug(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);
+                    this.alpha = null == e.alpha ? this.DEFAULT_ALPHA : e.alpha
+                }
+                call(e, t) {
+                    const n = db(e);
+                    return Ll(n)
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = {
+                            alpha: this.alpha
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Ex.className = "ELU", Fo.registerClass(Ex);
+            var Nx = class extends Ib {
+                constructor(e) {
+                    super(null == e ? {} : e), this.DEFAULT_THETA = 1, null == e && (e = {}), this.theta = null == e.theta ? this.DEFAULT_THETA : e.theta
+                }
+                call(e, t) {
+                    const n = db(e);
+                    return vu(n, Ai(Hl(n, this.theta), "float32"))
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = {
+                            theta: this.theta
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Nx.className = "ThresholdedReLU", Fo.registerClass(Nx);
+            var Cx = class extends Ib {
+                constructor(e) {
+                    super(null == e ? {} : e), this.DEFAULT_AXIS = 1, null == e && (e = {}), this.softmax = (new lx).apply, this.axis = null == e.axis ? this.DEFAULT_AXIS : e.axis
+                }
+                call(e, t) {
+                    const n = db(e);
+                    return this.softmax(n, this.axis)
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = {
+                            axis: this.axis
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+
+            function Tx(e, t, n) {
+                if ("number" == typeof e) return cg(e, t);
+                if (e.length !== t) throw new og(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);
+                for (let a = 0; a < t; ++a) {
+                    const s = e[a];
+                    if ((r = s) !== parseInt(r.toString(), 10)) throw new og(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${s}`)
+                }
+                return e;
+                var r
+            }
+
+            function Rx(e, t, n, r, a = 1) {
+                if (null == e) return e;
+                let s;
+                return s = "same" === n ? e : e - (t + (t - 1) * (a - 1)) + 1, Math.floor((s + r - 1) / r)
+            }
+
+            function _x(e, t, n, r) {
+                if (null == e) return null;
+                if ("valid" === r) e = e * t + sy([n - t, 0]);
+                else {
+                    if ("same" !== r) throw new og(`Unsupport padding mode: ${r}.`);
+                    e *= t
+                }
+                return e
+            }
+
+            function Fx(e, t) {
+                return ru(() => (Kg(t), "channelsFirst" === t ? qi(e, [0, 2, 3, 1]) : e))
+            }
+
+            function Mx(e, t) {
+                return ru(() => (Kg(t), "channelsFirst" === t ? qi(e, [0, 2, 3, 4, 1]) : e))
+            }
+
+            function Dx(e, t, n, r = [1, 1], a = "valid", s, i, o = null) {
+                return ru(() => {
+                    if (null == s && (s = "channelsLast"), Kg(s), 3 !== e.rank && 4 !== e.rank) throw new og(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);
+                    if (3 !== t.rank && 4 !== t.rank) throw new og(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);
+                    let u = Fx(e, s);
+                    if ("causal" === a) throw new ug("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");
+                    return u = bp.conv2d({
+                        x: u,
+                        filter: t,
+                        strides: r,
+                        pad: "same" === a ? "same" : "valid",
+                        dilations: i,
+                        dataFormat: "NHWC",
+                        bias: n,
+                        activation: o
+                    }), "channelsFirst" === s && (u = qi(u, [0, 3, 1, 2])), u
+                })
+            }
+            Cx.className = "Softmax", Fo.registerClass(Cx);
+            var Ox = class extends Ib {
+                    constructor(e, t) {
+                        if (super(t), this.bias = null, this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal", this.DEFAULT_BIAS_INITIALIZER = "zeros", Ox.verifyArgs(t), this.rank = e, Ag(this.rank, "rank"), 1 !== this.rank && 2 !== this.rank && 3 !== this.rank) throw new ug(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);
+                        if (this.kernelSize = Tx(t.kernelSize, e, "kernelSize"), this.strides = Tx(null == t.strides ? 1 : t.strides, e, "strides"), this.padding = null == t.padding ? "valid" : t.padding, Yg(this.padding), this.dataFormat = null == t.dataFormat ? "channelsLast" : t.dataFormat, Kg(this.dataFormat), this.activation = mx(t.activation), this.useBias = null == t.useBias || t.useBias, this.biasInitializer = jy(t.biasInitializer || this.DEFAULT_BIAS_INITIALIZER), this.biasConstraint = Lg(t.biasConstraint), this.biasRegularizer = kx(t.biasRegularizer), this.activityRegularizer = kx(t.activityRegularizer), this.dilationRate = Tx(null == t.dilationRate ? 1 : t.dilationRate, e, "dilationRate"), 1 === this.rank && Array.isArray(this.dilationRate) && 1 !== this.dilationRate.length) throw new og("dilationRate must be a number or an array of a single number for 1D convolution, but received " + JSON.stringify(this.dilationRate));
+                        if (2 === this.rank) {
+                            if ("number" == typeof this.dilationRate) this.dilationRate = [this.dilationRate, this.dilationRate];
+                            else if (2 !== this.dilationRate.length) throw new og("dilationRate must be a number or array of two numbers for 2D convolution, but received " + JSON.stringify(this.dilationRate))
+                        } else if (3 === this.rank)
+                            if ("number" == typeof this.dilationRate) this.dilationRate = [this.dilationRate, this.dilationRate, this.dilationRate];
+                            else if (3 !== this.dilationRate.length) throw new og("dilationRate must be a number or array of three numbers for 3D convolution, but received " + JSON.stringify(this.dilationRate))
+                    }
+                    static verifyArgs(e) {
+                        if (hg("kernelSize" in e, "required key 'kernelSize' not in config"), "number" != typeof e.kernelSize && !Ig(e.kernelSize, "number", 1, 3)) throw new og(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)
+                    }
+                    getConfig() {
+                        const e = {
+                                kernelSize: this.kernelSize,
+                                strides: this.strides,
+                                padding: this.padding,
+                                dataFormat: this.dataFormat,
+                                dilationRate: this.dilationRate,
+                                activation: dx(this.activation),
+                                useBias: this.useBias,
+                                biasInitializer: Vy(this.biasInitializer),
+                                biasRegularizer: xx(this.biasRegularizer),
+                                activityRegularizer: xx(this.activityRegularizer),
+                                biasConstraint: Dg(this.biasConstraint)
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                },
+                Lx = class extends Ox {
+                    constructor(e, t) {
+                        super(e, t), this.kernel = null, Lx.verifyArgs(t), this.filters = t.filters, Ag(this.filters, "filters"), this.kernelInitializer = jy(t.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER), this.kernelConstraint = Lg(t.kernelConstraint), this.kernelRegularizer = kx(t.kernelRegularizer)
+                    }
+                    build(e) {
+                        e = fb(e);
+                        const t = "channelsFirst" === this.dataFormat ? 1 : e.length - 1;
+                        if (null == e[t]) throw new og("The channel dimension of the input should be defined. Found " + e[t]);
+                        const n = e[t],
+                            r = this.kernelSize.concat([n, this.filters]);
+                        this.kernel = this.addWeight("kernel", r, null, this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.useBias && (this.bias = this.addWeight("bias", [this.filters], null, this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint)), this.inputSpec = [{
+                            ndim: this.rank + 2,
+                            axes: {
+                                [t]: n
+                            }
+                        }], this.built = !0
+                    }
+                    call(e, t) {
+                        return ru(() => {
+                            let t;
+                            e = db(e);
+                            const n = null == this.bias ? null : this.bias.read(),
+                                r = Eg(this.activation.getClassName());
+                            if (null != r && 2 === this.rank) t = Dx(e, this.kernel.read(), n, this.strides, this.padding, this.dataFormat, this.dilationRate, r);
+                            else {
+                                if (1 === this.rank) t = function (e, t, n, r = 1, a = "valid", s, i = 1) {
+                                    return ru(() => {
+                                        if (null == s && (s = "channelsLast"), Kg(s), 3 !== e.shape.length) throw new og(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);
+                                        if (3 !== t.shape.length) throw new og(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);
+                                        if (null != n && 1 !== n.shape.length) throw new og(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);
+                                        if ("channelsFirst" === s && (e = qi(e, [0, 2, 1])), "causal" === a) throw new ug("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");
+                                        let o = gl(e, t, r, "same" === a ? "same" : "valid", "NWC", i);
+                                        return null != n && (o = xy(o, n)), o
+                                    })
+                                }(e, this.kernel.read(), n, this.strides[0], this.padding, this.dataFormat, this.dilationRate[0]);
+                                else if (2 === this.rank) t = Dx(e, this.kernel.read(), n, this.strides, this.padding, this.dataFormat, this.dilationRate);
+                                else {
+                                    if (3 !== this.rank) throw new ug("convolutions greater than 3D are not implemented yet.");
+                                    t = function (e, t, n, r = [1, 1, 1], a = "valid", s, i) {
+                                        return ru(() => {
+                                            if (null == s && (s = "channelsLast"), Kg(s), 4 !== e.rank && 5 !== e.rank) throw new og(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);
+                                            if (4 !== t.rank && 5 !== t.rank) throw new og(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);
+                                            let o = Mx(e, s);
+                                            if ("causal" === a) throw new ug("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");
+                                            return o = vl(o, t, r, "same" === a ? "same" : "valid", "NDHWC", i), null != n && (o = xy(o, n)), "channelsFirst" === s && (o = qi(o, [0, 4, 1, 2, 3])), o
+                                        })
+                                    }(e, this.kernel.read(), n, this.strides, this.padding, this.dataFormat, this.dilationRate)
+                                }
+                                null != this.activation && (t = this.activation.apply(t))
+                            }
+                            return t
+                        })
+                    }
+                    computeOutputShape(e) {
+                        e = fb(e);
+                        const t = [],
+                            n = "channelsLast" === this.dataFormat ? e.slice(1, e.length - 1) : e.slice(2);
+                        for (let e = 0; e < n.length; ++e) {
+                            const r = Rx(n[e], this.kernelSize[e], this.padding, this.strides[e], "number" == typeof this.dilationRate ? this.dilationRate : this.dilationRate[e]);
+                            t.push(r)
+                        }
+                        let r = [e[0]];
+                        return "channelsLast" === this.dataFormat ? (r = r.concat(t), r.push(this.filters)) : (r.push(this.filters), r = r.concat(t)), r
+                    }
+                    getConfig() {
+                        const e = {
+                                filters: this.filters,
+                                kernelInitializer: Vy(this.kernelInitializer),
+                                kernelRegularizer: xx(this.kernelRegularizer),
+                                kernelConstraint: Dg(this.kernelConstraint)
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                    static verifyArgs(e) {
+                        if (!("filters" in e) || "number" != typeof e.filters || e.filters < 1) throw new og("Convolution layer expected config.filters to be a 'number' > 0 but got " + JSON.stringify(e.filters))
+                    }
+                },
+                Px = class extends Lx {
+                    constructor(e) {
+                        super(2, e), Px.verifyArgs(e)
+                    }
+                    getConfig() {
+                        const e = super.getConfig();
+                        return delete e.rank, e
+                    }
+                    static verifyArgs(e) {
+                        if ("number" != typeof e.kernelSize && !Ig(e.kernelSize, "number", 1, 2)) throw new og(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)
+                    }
+                };
+            Px.className = "Conv2D", Fo.registerClass(Px);
+            var $x = class extends Lx {
+                constructor(e) {
+                    super(3, e), $x.verifyArgs(e)
+                }
+                getConfig() {
+                    const e = super.getConfig();
+                    return delete e.rank, e
+                }
+                static verifyArgs(e) {
+                    if ("number" != typeof e.kernelSize && (!Array.isArray(e.kernelSize) || 1 !== e.kernelSize.length && 3 !== e.kernelSize.length)) throw new og(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)
+                }
+            };
+            $x.className = "Conv3D", Fo.registerClass($x);
+            var Bx = class extends Px {
+                constructor(e) {
+                    if (super(e), this.inputSpec = [new vb({
+                            ndim: 4
+                        })], "same" !== this.padding && "valid" !== this.padding) throw new og("Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode " + this.padding)
+                }
+                build(e) {
+                    if (4 !== (e = fb(e)).length) throw new og("Input should have rank 4; Received input shape: " + JSON.stringify(e));
+                    const t = "channelsFirst" === this.dataFormat ? 1 : e.length - 1;
+                    if (null == e[t]) throw new og("The channel dimension of the inputs should be defined. Found `None`.");
+                    const n = e[t],
+                        r = this.kernelSize.concat([this.filters, n]);
+                    this.kernel = this.addWeight("kernel", r, "float32", this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.useBias && (this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint)), this.inputSpec = [new vb({
+                        ndim: 4,
+                        axes: {
+                            [t]: n
+                        }
+                    })], this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        let t = db(e);
+                        if (4 !== t.shape.length) throw new og("Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-" + t.shape.length);
+                        const n = t.shape,
+                            r = n[0];
+                        let a, s;
+                        "channelsFirst" === this.dataFormat ? (a = 2, s = 3) : (a = 1, s = 2);
+                        const i = n[a],
+                            o = n[s],
+                            u = this.kernelSize[0],
+                            l = this.kernelSize[1],
+                            c = this.strides[0],
+                            h = this.strides[1],
+                            p = [r, _x(i, c, u, this.padding), _x(o, h, l, this.padding), this.filters];
+                        "channelsLast" !== this.dataFormat && (t = qi(t, [0, 2, 3, 1]));
+                        let d = bl(t, this.kernel.read(), p, this.strides, this.padding);
+                        return "channelsLast" !== this.dataFormat && (d = qi(d, [0, 3, 1, 2])), null != this.bias && (d = xy(d, this.bias.read(), this.dataFormat)), null != this.activation && (d = this.activation.apply(d)), d
+                    })
+                }
+                computeOutputShape(e) {
+                    const t = (e = fb(e)).slice();
+                    let n, r, a;
+                    "channelsFirst" === this.dataFormat ? (n = 1, r = 2, a = 3) : (n = 3, r = 1, a = 2);
+                    const s = this.kernelSize[0],
+                        i = this.kernelSize[1],
+                        o = this.strides[0],
+                        u = this.strides[1];
+                    return t[n] = this.filters, t[r] = _x(t[r], o, s, this.padding), t[a] = _x(t[a], u, i, this.padding), t
+                }
+                getConfig() {
+                    const e = super.getConfig();
+                    return delete e.dilationRate, e
+                }
+            };
+            Bx.className = "Conv2DTranspose", Fo.registerClass(Bx);
+            var zx = class extends $x {
+                constructor(e) {
+                    if (super(e), this.inputSpec = [new vb({
+                            ndim: 5
+                        })], "same" !== this.padding && "valid" !== this.padding) throw new og("Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode " + this.padding)
+                }
+                build(e) {
+                    if (5 !== (e = fb(e)).length) throw new og("Input should have rank 5; Received input shape: " + JSON.stringify(e));
+                    const t = "channelsFirst" === this.dataFormat ? 1 : e.length - 1;
+                    if (null == e[t]) throw new og("The channel dimension of the inputs should be defined. Found `None`.");
+                    const n = e[t],
+                        r = this.kernelSize.concat([this.filters, n]);
+                    this.kernel = this.addWeight("kernel", r, "float32", this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.useBias && (this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint)), this.inputSpec = [new vb({
+                        ndim: 5,
+                        axes: {
+                            [t]: n
+                        }
+                    })], this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        let t = db(e);
+                        if (5 !== t.shape.length) throw new og("Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-" + t.shape.length);
+                        const n = t.shape,
+                            r = n[0];
+                        let a, s, i;
+                        "channelsFirst" === this.dataFormat ? (i = 2, a = 3, s = 4) : (i = 1, a = 2, s = 3);
+                        const o = n[i],
+                            u = n[a],
+                            l = n[s],
+                            c = this.kernelSize[0],
+                            h = this.kernelSize[1],
+                            p = this.kernelSize[2],
+                            d = this.strides[0],
+                            f = this.strides[1],
+                            m = this.strides[2],
+                            g = [r, _x(o, d, c, this.padding), _x(u, f, h, this.padding), _x(l, m, p, this.padding), this.filters];
+                        "channelsLast" !== this.dataFormat && (t = qi(t, [0, 2, 3, 4, 1]));
+                        let y = wl(t, this.kernel.read(), g, this.strides, this.padding);
+                        return "channelsLast" !== this.dataFormat && (y = qi(y, [0, 4, 1, 2, 3])), null !== this.bias && (y = xy(y, this.bias.read(), this.dataFormat)), null !== this.activation && (y = this.activation.apply(y)), y
+                    })
+                }
+                computeOutputShape(e) {
+                    const t = (e = fb(e)).slice();
+                    let n, r, a, s;
+                    "channelsFirst" === this.dataFormat ? (n = 1, r = 2, a = 3, s = 4) : (n = 4, r = 1, a = 2, s = 3);
+                    const i = this.kernelSize[0],
+                        o = this.kernelSize[1],
+                        u = this.kernelSize[2],
+                        l = this.strides[0],
+                        c = this.strides[1],
+                        h = this.strides[2];
+                    return t[n] = this.filters, t[r] = _x(t[r], l, i, this.padding), t[a] = _x(t[a], c, o, this.padding), t[s] = _x(t[s], h, u, this.padding), t
+                }
+                getConfig() {
+                    const e = super.getConfig();
+                    return delete e.dilationRate, e
+                }
+            };
+            zx.className = "Conv3DTranspose", Fo.registerClass(zx);
+            var Ux = class extends Lx {
+                constructor(e, t) {
+                    if (super(e, t), this.DEFAULT_DEPTHWISE_INITIALIZER = "glorotUniform", this.DEFAULT_POINTWISE_INITIALIZER = "glorotUniform", this.depthwiseKernel = null, this.pointwiseKernel = null, null == t.filters) throw new og("The `filters` configuration field is required by SeparableConv, but is unspecified.");
+                    if (null != t.kernelInitializer || null != t.kernelRegularizer || null != t.kernelConstraint) throw new og("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");
+                    if (null != t.padding && "same" !== t.padding && "valid" !== t.padding) throw new og(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);
+                    this.depthMultiplier = null == t.depthMultiplier ? 1 : t.depthMultiplier, this.depthwiseInitializer = jy(t.depthwiseInitializer || this.DEFAULT_DEPTHWISE_INITIALIZER), this.depthwiseRegularizer = kx(t.depthwiseRegularizer), this.depthwiseConstraint = Lg(t.depthwiseConstraint), this.pointwiseInitializer = jy(t.depthwiseInitializer || this.DEFAULT_POINTWISE_INITIALIZER), this.pointwiseRegularizer = kx(t.pointwiseRegularizer), this.pointwiseConstraint = Lg(t.pointwiseConstraint)
+                }
+                build(e) {
+                    if ((e = fb(e)).length < this.rank + 2) throw new og(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);
+                    const t = "channelsFirst" === this.dataFormat ? 1 : e.length - 1;
+                    if (null == e[t] || e[t] < 0) throw new og("The channel dimension of the inputs should be defined, but found " + JSON.stringify(e[t]));
+                    const n = e[t],
+                        r = this.kernelSize.concat([n, this.depthMultiplier]),
+                        a = [];
+                    for (let e = 0; e < this.rank; ++e) a.push(1);
+                    a.push(n * this.depthMultiplier, this.filters);
+                    this.depthwiseKernel = this.addWeight("depthwise_kernel", r, "float32", this.depthwiseInitializer, this.depthwiseRegularizer, !0, this.depthwiseConstraint), this.pointwiseKernel = this.addWeight("pointwise_kernel", a, "float32", this.pointwiseInitializer, this.pointwiseRegularizer, !0, this.pointwiseConstraint), this.useBias ? this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint) : this.bias = null, this.inputSpec = [new vb({
+                        ndim: this.rank + 2,
+                        axes: {
+                            [t]: n
+                        }
+                    })], this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        let t;
+                        if (e = db(e), 1 === this.rank) throw new ug("1D separable convolution is not implemented yet.");
+                        return 2 === this.rank && ("channelsFirst" === this.dataFormat && (e = qi(e, [0, 2, 3, 1])), t = Nh(e, this.depthwiseKernel.read(), this.pointwiseKernel.read(), this.strides, this.padding, this.dilationRate, "NHWC")), this.useBias && (t = xy(t, this.bias.read(), this.dataFormat)), null != this.activation && (t = this.activation.apply(t)), "channelsFirst" === this.dataFormat && (t = qi(t, [0, 3, 1, 2])), t
+                    })
+                }
+                getConfig() {
+                    const e = super.getConfig();
+                    return delete e.rank, delete e.kernelInitializer, delete e.kernelRegularizer, delete e.kernelConstraint, e.depthwiseInitializer = Vy(this.depthwiseInitializer), e.pointwiseInitializer = Vy(this.pointwiseInitializer), e.depthwiseRegularizer = xx(this.depthwiseRegularizer), e.pointwiseRegularizer = xx(this.pointwiseRegularizer), e.depthwiseConstraint = Dg(this.depthwiseConstraint), e.pointwiseConstraint = Dg(this.pointwiseConstraint), e
+                }
+            };
+            Ux.className = "SeparableConv";
+            var Wx = class extends Ux {
+                constructor(e) {
+                    super(2, e)
+                }
+            };
+            Wx.className = "SeparableConv2D", Fo.registerClass(Wx);
+            var Vx = class extends Lx {
+                constructor(e) {
+                    super(1, e), Vx.verifyArgs(e), this.inputSpec = [{
+                        ndim: 3
+                    }]
+                }
+                getConfig() {
+                    const e = super.getConfig();
+                    return delete e.rank, delete e.dataFormat, e
+                }
+                static verifyArgs(e) {
+                    if ("number" != typeof e.kernelSize && !Ig(e.kernelSize, "number", 1, 1)) throw new og(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)
+                }
+            };
+            Vx.className = "Conv1D", Fo.registerClass(Vx);
+            var jx = class extends Ib {
+                constructor(e) {
+                    super(e), "number" == typeof e.cropping ? this.cropping = [
+                        [e.cropping, e.cropping],
+                        [e.cropping, e.cropping]
+                    ] : "number" == typeof e.cropping[0] ? this.cropping = [
+                        [e.cropping[0], e.cropping[0]],
+                        [e.cropping[1], e.cropping[1]]
+                    ] : this.cropping = e.cropping, this.dataFormat = void 0 === e.dataFormat ? "channelsLast" : e.dataFormat, this.inputSpec = [{
+                        ndim: 4
+                    }]
+                }
+                computeOutputShape(e) {
+                    return "channelsFirst" === this.dataFormat ? [e[0], e[1], e[2] - this.cropping[0][0] - this.cropping[0][1], e[3] - this.cropping[1][0] - this.cropping[1][1]] : [e[0], e[1] - this.cropping[0][0] - this.cropping[0][1], e[2] - this.cropping[1][0] - this.cropping[1][1], e[3]]
+                }
+                call(e, t) {
+                    return ru(() => {
+                        if (e = db(e), "channelsLast" === this.dataFormat) {
+                            const t = hy(e, this.cropping[0][0], e.shape[1] - this.cropping[0][0] - this.cropping[0][1], 2);
+                            return hy(t, this.cropping[1][0], e.shape[2] - this.cropping[1][1] - this.cropping[1][0], 3)
+                        } {
+                            const t = hy(e, this.cropping[0][0], e.shape[2] - this.cropping[0][0] - this.cropping[0][1], 3);
+                            return hy(t, this.cropping[1][0], e.shape[3] - this.cropping[1][1] - this.cropping[1][0], 4)
+                        }
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            cropping: this.cropping,
+                            dataFormat: this.dataFormat
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            jx.className = "Cropping2D", Fo.registerClass(jx);
+            var Gx = class extends Ib {
+                constructor(e) {
+                    var t;
+                    super(e), this.DEFAULT_SIZE = [2, 2], this.inputSpec = [{
+                        ndim: 4
+                    }], this.size = null == e.size ? this.DEFAULT_SIZE : e.size, this.dataFormat = null == e.dataFormat ? "channelsLast" : e.dataFormat, Kg(this.dataFormat), this.interpolation = null == e.interpolation ? "nearest" : e.interpolation, t = this.interpolation, Sg(Vg, "InterpolationFormat", t)
+                }
+                computeOutputShape(e) {
+                    if ("channelsFirst" === this.dataFormat) {
+                        const t = null == e[2] ? null : this.size[0] * e[2],
+                            n = null == e[3] ? null : this.size[1] * e[3];
+                        return [e[0], e[1], t, n]
+                    } {
+                        const t = null == e[1] ? null : this.size[0] * e[1],
+                            n = null == e[2] ? null : this.size[1] * e[2];
+                        return [e[0], t, n, e[3]]
+                    }
+                }
+                call(e, t) {
+                    return ru(() => {
+                        let t = db(e);
+                        const n = t.shape;
+                        if ("channelsFirst" === this.dataFormat) {
+                            t = qi(t, [0, 2, 3, 1]);
+                            const e = this.size[0] * n[2],
+                                r = this.size[1] * n[3],
+                                a = "nearest" === this.interpolation ? Nd.resizeNearestNeighbor(t, [e, r]) : Nd.resizeBilinear(t, [e, r]);
+                            return qi(a, [0, 3, 1, 2])
+                        } {
+                            const e = this.size[0] * n[1],
+                                r = this.size[1] * n[2];
+                            return "nearest" === this.interpolation ? Nd.resizeNearestNeighbor(t, [e, r]) : Nd.resizeBilinear(t, [e, r])
+                        }
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            size: this.size,
+                            dataFormat: this.dataFormat
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Gx.className = "UpSampling2D", Fo.registerClass(Gx);
+            var Hx = class extends Ox {
+                constructor(e) {
+                    super(2, e), this.depthwiseKernel = null, this.depthMultiplier = null == e.depthMultiplier ? 1 : e.depthMultiplier, this.depthwiseInitializer = jy(e.depthwiseInitializer || this.DEFAULT_KERNEL_INITIALIZER), this.depthwiseConstraint = Lg(e.depthwiseConstraint), this.depthwiseRegularizer = kx(e.depthwiseRegularizer)
+                }
+                build(e) {
+                    if ((e = fb(e)).length < 4) throw new og(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);
+                    const t = "channelsFirst" === this.dataFormat ? 1 : 3;
+                    if (null == e[t] || e[t] < 0) throw new og(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);
+                    const n = e[t],
+                        r = [this.kernelSize[0], this.kernelSize[1], n, this.depthMultiplier];
+                    this.depthwiseKernel = this.addWeight("depthwise_kernel", r, null, this.depthwiseInitializer, this.depthwiseRegularizer, !0, this.depthwiseConstraint), this.useBias ? this.bias = this.addWeight("bias", [n * this.depthMultiplier], null, this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint) : this.bias = null, this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        let t = function (e, t, n = [1, 1], r = "valid", a, s) {
+                            return ru(() => {
+                                null == a && (a = "channelsLast"), Kg(a);
+                                let i = Fx(e, a);
+                                if (4 !== e.rank) throw new og(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);
+                                if (4 !== t.rank) throw new og(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);
+                                return i = Nl(i, t, n, "same" === r ? "same" : "valid", "NHWC", s), "channelsFirst" === a && (i = qi(i, [0, 3, 1, 2])), i
+                            })
+                        }(e = db(e), this.depthwiseKernel.read(), this.strides, this.padding, this.dataFormat, null);
+                        return this.useBias && (t = xy(t, this.bias.read(), this.dataFormat)), null != this.activation && (t = this.activation.apply(t)), t
+                    })
+                }
+                computeOutputShape(e) {
+                    e = fb(e);
+                    const t = "channelsFirst" === this.dataFormat ? e[2] : e[1],
+                        n = "channelsFirst" === this.dataFormat ? e[3] : e[2],
+                        r = "channelsFirst" === this.dataFormat ? e[1] * this.depthMultiplier : e[3] * this.depthMultiplier,
+                        a = Rx(t, this.kernelSize[0], this.padding, this.strides[0]),
+                        s = Rx(n, this.kernelSize[1], this.padding, this.strides[1]);
+                    return "channelsFirst" === this.dataFormat ? [e[0], r, a, s] : [e[0], a, s, r]
+                }
+                getConfig() {
+                    const e = super.getConfig();
+                    return e.depthMultiplier = this.depthMultiplier, e.depthwiseInitializer = Vy(this.depthwiseInitializer), e.depthwiseRegularizer = xx(this.depthwiseRegularizer), e.depthwiseConstraint = Dg(this.depthwiseRegularizer), e
+                }
+            };
+
+            function qx(e, t, n, r) {
+                if (Array.isArray(e)) {
+                    if (null != t || null != n) throw new og("When inputs is an array, neither initialState or constants should be provided");
+                    null != r && (n = e.slice(e.length - r, e.length), e = e.slice(0, e.length - r)), e.length > 1 && (t = e.slice(1, e.length)), e = e[0]
+                }
+
+                function a(e) {
+                    return null == e || Array.isArray(e) ? e : [e]
+                }
+                return {
+                    inputs: e,
+                    initialState: t = a(t),
+                    constants: n = a(n)
+                }
+            }
+
+            function Kx(e, t, n, r = !1, a, s, i = !1, o = !1) {
+                return ru(() => {
+                    const u = t.shape.length;
+                    if (u < 3) throw new og(`Input should be at least 3D, but is ${u}D.`);
+                    const l = [1, 0].concat(iy(2, u));
+                    if (t = qi(t, l), null != s) throw new ug("The rnn() functoin of the deeplearn.js backend does not support constants yet.");
+                    i && console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."), null != a && ((a = Ai(Ai(a, "bool"), "float32")).rank === u - 1 && (a = Bl(a, -1)), a = qi(a, l)), r && (t = bh(t, 0), null != a && (a = bh(a, 0)));
+                    const c = [];
+                    let h, p = n;
+                    const d = t.shape[0],
+                        f = ap(t);
+                    let m, g;
+                    null != a && (m = ap(a));
+                    for (let t = 0; t < d; ++t) {
+                        const n = f[t],
+                            r = ru(() => e(n, p));
+                        if (null == a) h = r[0], p = r[1];
+                        else {
+                            const e = ru(() => {
+                                const e = m[t],
+                                    n = gc(Yc(e), e);
+                                return {
+                                    output: gu(vu(r[0], e), vu(p[0], n)),
+                                    newStates: p.map((t, a) => gu(vu(r[1][a], e), vu(t, n)))
+                                }
+                            });
+                            h = e.output, p = e.newStates
+                        }
+                        o && c.push(h)
+                    }
+                    if (o) {
+                        g = Gh(c, 1)
+                    }
+                    return [h, g, p]
+                })
+            }
+            Hx.className = "DepthwiseConv2D", Fo.registerClass(Hx);
+            var Yx = class extends Ib {
+                constructor(e) {
+                    let t;
+                    if (super(e), null == e.cell) throw new og("cell property is missing for the constructor of RNN.");
+                    if (t = Array.isArray(e.cell) ? new rw({
+                            cells: e.cell
+                        }) : e.cell, null == t.stateSize) throw new og("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");
+                    this.cell = t, this.returnSequences = null != e.returnSequences && e.returnSequences, this.returnState = null != e.returnState && e.returnState, this.goBackwards = null != e.goBackwards && e.goBackwards, this._stateful = null != e.stateful && e.stateful, this.unroll = null != e.unroll && e.unroll, this.supportsMasking = !0, this.inputSpec = [new vb({
+                        ndim: 3
+                    })], this.stateSpec = null, this.states_ = null, this.numConstants = null, this.keptStates = []
+                }
+                getStates() {
+                    if (null == this.states_) {
+                        return iy(0, Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1).map(e => null)
+                    }
+                    return this.states_
+                }
+                setStates(e) {
+                    this.states_ = e
+                }
+                computeOutputShape(e) {
+                    hb(e) && (e = e[0]), e = e;
+                    let t = this.cell.stateSize;
+                    Array.isArray(t) || (t = [t]);
+                    const n = t[0];
+                    let r;
+                    if (r = this.returnSequences ? [e[0], e[1], n] : [e[0], n], this.returnState) {
+                        const n = [];
+                        for (const r of t) n.push([e[0], r]);
+                        return [r].concat(n)
+                    }
+                    return r
+                }
+                computeMask(e, t) {
+                    return ru(() => {
+                        Array.isArray(t) && (t = t[0]);
+                        const e = this.returnSequences ? t : null;
+                        if (this.returnState) {
+                            const t = this.states.map(e => null);
+                            return [e].concat(t)
+                        }
+                        return e
+                    })
+                }
+                get states() {
+                    if (null == this.states_) {
+                        const e = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1,
+                            t = [];
+                        for (let n = 0; n < e; ++n) t.push(null);
+                        return t
+                    }
+                    return this.states_
+                }
+                set states(e) {
+                    this.states_ = e
+                }
+                build(e) {
+                    if (null != this.numConstants) throw new ug("Constants support is not implemented in RNN yet.");
+                    hb(e) && (e = e[0]), e = e;
+                    const t = this.stateful ? e[0] : null,
+                        n = e.slice(2);
+                    this.inputSpec[0] = new vb({
+                        shape: [t, null, ...n]
+                    });
+                    const r = [e[0]].concat(e.slice(2));
+                    let a;
+                    if (this.cell.build(r), a = Array.isArray(this.cell.stateSize) ? this.cell.stateSize : [this.cell.stateSize], null != this.stateSpec) {
+                        if (!va.arraysEqual(this.stateSpec.map(e => e.shape[e.shape.length - 1]), a)) throw new og(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)
+                    } else this.stateSpec = a.map(e => new vb({
+                        shape: [null, e]
+                    }));
+                    this.stateful && this.resetStates()
+                }
+                resetStates(e, t = !1) {
+                    ru(() => {
+                        if (!this.stateful) throw new sg("Cannot call resetStates() on an RNN Layer that is not stateful.");
+                        const n = this.inputSpec[0].shape[0];
+                        if (null == n) throw new og("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");
+                        if (null == this.states_) Array.isArray(this.cell.stateSize) ? this.states_ = this.cell.stateSize.map(e => Pc([n, e])) : this.states_ = [Pc([n, this.cell.stateSize])];
+                        else if (null == e) au(this.states_), null != this.keptStates && (au(this.keptStates), this.keptStates = []), Array.isArray(this.cell.stateSize) ? this.states_ = this.cell.stateSize.map(e => Pc([n, e])) : this.states_[0] = Pc([n, this.cell.stateSize]);
+                        else {
+                            if (Array.isArray(e) || (e = [e]), e.length !== this.states_.length) throw new og(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);
+                            !0 === t ? this.keptStates.push(this.states_.slice()) : au(this.states_);
+                            for (let t = 0; t < this.states_.length; ++t) {
+                                const r = e[t],
+                                    a = Array.isArray(this.cell.stateSize) ? this.cell.stateSize[t] : this.cell.stateSize,
+                                    s = [n, a];
+                                if (!va.arraysEqual(r.shape, s)) throw new og(`State ${t} is incompatible with layer ${this.name}: expected shape=${s}, received shape=${r.shape}`);
+                                this.states_[t] = r
+                            }
+                        }
+                        this.states_ = this.states_.map(e => su(e.clone()))
+                    })
+                }
+                apply(e, t) {
+                    let n = null == t ? null : t.initialState,
+                        r = null == t ? null : t.constants;
+                    null == t && (t = {});
+                    const a = qx(e, n, r, this.numConstants);
+                    e = a.inputs, n = a.initialState, r = a.constants;
+                    let s = [],
+                        i = [];
+                    if (null != n) {
+                        t.initialState = n, s = s.concat(n), this.stateSpec = [];
+                        for (const e of n) this.stateSpec.push(new vb({
+                            shape: e.shape
+                        }));
+                        i = i.concat(this.stateSpec)
+                    }
+                    null != r && (t.constants = r, s = s.concat(r), this.numConstants = r.length);
+                    if (s[0] instanceof xb) {
+                        const n = [e].concat(s),
+                            r = this.inputSpec.concat(i),
+                            a = this.inputSpec;
+                        this.inputSpec = r;
+                        const o = super.apply(n, t);
+                        return this.inputSpec = a, o
+                    }
+                    return super.apply(e, t)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        const n = null == t ? null : t.mask,
+                            r = null == t ? null : t.training;
+                        let a = null == t ? null : t.initialState;
+                        e = db(e), null == a && (a = this.stateful ? this.states_ : this.getInitialState(e));
+                        const s = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+                        if (a.length !== s) throw new og(`RNN Layer has ${s} state(s) but was passed ${a.length} initial state(s).`);
+                        this.unroll && console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");
+                        const i = {
+                                training: r
+                            },
+                            o = Kx((e, t) => {
+                                const n = this.cell.call([e].concat(t), i);
+                                return [n[0], n.slice(1)]
+                            }, e, a, this.goBackwards, n, null, this.unroll, this.returnSequences),
+                            u = o[0],
+                            l = o[1],
+                            c = o[2];
+                        this.stateful && this.resetStates(c, r);
+                        const h = this.returnSequences ? l : u;
+                        return this.returnState ? [h].concat(c) : h
+                    })
+                }
+                getInitialState(e) {
+                    return ru(() => {
+                        let t = Pc(e.shape);
+                        return t = yc(t, [1, 2]), t = uy(t), Array.isArray(this.cell.stateSize) ? this.cell.stateSize.map(e => e > 1 ? fy(t, [1, e]) : t) : this.cell.stateSize > 1 ? [fy(t, [1, this.cell.stateSize])] : [t]
+                    })
+                }
+                get trainableWeights() {
+                    return this.trainable ? this.cell.trainableWeights : []
+                }
+                get nonTrainableWeights() {
+                    return this.trainable ? this.cell.nonTrainableWeights : this.cell.weights
+                }
+                setFastWeightInitDuringBuild(e) {
+                    super.setFastWeightInitDuringBuild(e), null != this.cell && this.cell.setFastWeightInitDuringBuild(e)
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            returnSequences: this.returnSequences,
+                            returnState: this.returnState,
+                            goBackwards: this.goBackwards,
+                            stateful: this.stateful,
+                            unroll: this.unroll
+                        };
+                    null != this.numConstants && (t.numConstants = this.numConstants);
+                    const n = this.cell.getConfig();
+                    return this.getClassName() === Yx.className && (t.cell = {
+                        className: this.cell.getClassName(),
+                        config: n
+                    }), Object.assign({}, n, e, t)
+                }
+                static fromConfig(e, t, n = {}) {
+                    const r = Bb(t.cell, n);
+                    return new e(Object.assign(t, {
+                        cell: r
+                    }))
+                }
+            };
+            Yx.className = "RNN", Fo.registerClass(Yx);
+            var Xx = class extends Ib {},
+                Qx = class extends Xx {
+                    constructor(e) {
+                        super(e), this.DEFAULT_ACTIVATION = "tanh", this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal", this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal", this.DEFAULT_BIAS_INITIALIZER = "zeros", this.units = e.units, Ag(this.units, "units"), this.activation = mx(null == e.activation ? this.DEFAULT_ACTIVATION : e.activation), this.useBias = null == e.useBias || e.useBias, this.kernelInitializer = jy(e.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER), this.recurrentInitializer = jy(e.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER), this.biasInitializer = jy(e.biasInitializer || this.DEFAULT_BIAS_INITIALIZER), this.kernelRegularizer = kx(e.kernelRegularizer), this.recurrentRegularizer = kx(e.recurrentRegularizer), this.biasRegularizer = kx(e.biasRegularizer), this.kernelConstraint = Lg(e.kernelConstraint), this.recurrentConstraint = Lg(e.recurrentConstraint), this.biasConstraint = Lg(e.biasConstraint), this.dropout = ay([1, sy([0, null == e.dropout ? 0 : e.dropout])]), this.recurrentDropout = ay([1, sy([0, null == e.recurrentDropout ? 0 : e.recurrentDropout])]), this.dropoutFunc = e.dropoutFunc, this.stateSize = this.units, this.dropoutMask = null, this.recurrentDropoutMask = null
+                    }
+                    build(e) {
+                        e = fb(e), this.kernel = this.addWeight("kernel", [e[e.length - 1], this.units], null, this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units], null, this.recurrentInitializer, this.recurrentRegularizer, !0, this.recurrentConstraint), this.useBias ? this.bias = this.addWeight("bias", [this.units], null, this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint) : this.bias = null, this.built = !0
+                    }
+                    call(e, t) {
+                        return ru(() => {
+                            if (2 !== (e = e).length) throw new og(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);
+                            let n = e[1];
+                            e = e[0];
+                            const r = null != t.training && t.training;
+                            let a;
+                            0 < this.dropout && this.dropout < 1 && null == this.dropoutMask && (this.dropoutMask = aw({
+                                ones: () => Yc(e),
+                                rate: this.dropout,
+                                training: r,
+                                dropoutFunc: this.dropoutFunc
+                            })), 0 < this.recurrentDropout && this.recurrentDropout < 1 && null == this.recurrentDropoutMask && (this.recurrentDropoutMask = aw({
+                                ones: () => Yc(n),
+                                rate: this.recurrentDropout,
+                                training: r,
+                                dropoutFunc: this.dropoutFunc
+                            }));
+                            const s = this.dropoutMask,
+                                i = this.recurrentDropoutMask;
+                            a = gy(null != s ? vu(e, s) : e, this.kernel.read()), null != this.bias && (a = xy(a, this.bias.read())), null != i && (n = vu(n, i));
+                            let o = gu(a, gy(n, this.recurrentKernel.read()));
+                            return null != this.activation && (o = this.activation.apply(o)), [o, o]
+                        })
+                    }
+                    getConfig() {
+                        const e = super.getConfig(),
+                            t = {
+                                units: this.units,
+                                activation: dx(this.activation),
+                                useBias: this.useBias,
+                                kernelInitializer: Vy(this.kernelInitializer),
+                                recurrentInitializer: Vy(this.recurrentInitializer),
+                                biasInitializer: Vy(this.biasInitializer),
+                                kernelRegularizer: xx(this.kernelRegularizer),
+                                recurrentRegularizer: xx(this.recurrentRegularizer),
+                                biasRegularizer: xx(this.biasRegularizer),
+                                activityRegularizer: xx(this.activityRegularizer),
+                                kernelConstraint: Dg(this.kernelConstraint),
+                                recurrentConstraint: Dg(this.recurrentConstraint),
+                                biasConstraint: Dg(this.biasConstraint),
+                                dropout: this.dropout,
+                                recurrentDropout: this.recurrentDropout
+                            };
+                        return Object.assign({}, e, t)
+                    }
+                };
+            Qx.className = "SimpleRNNCell", Fo.registerClass(Qx);
+            var Zx = class extends Yx {
+                constructor(e) {
+                    e.cell = new Qx(e), super(e)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        null != this.cell.dropoutMask && (au(this.cell.dropoutMask), this.cell.dropoutMask = null), null != this.cell.recurrentDropoutMask && (au(this.cell.recurrentDropoutMask), this.cell.recurrentDropoutMask = null);
+                        const n = null == t ? null : t.mask,
+                            r = null == t ? null : t.training,
+                            a = null == t ? null : t.initialState;
+                        return super.call(e, {
+                            mask: n,
+                            training: r,
+                            initialState: a
+                        })
+                    })
+                }
+                static fromConfig(e, t) {
+                    return new e(t)
+                }
+            };
+            Zx.className = "SimpleRNN", Fo.registerClass(Zx);
+            var Jx = class extends Xx {
+                constructor(e) {
+                    if (super(e), this.DEFAULT_ACTIVATION = "tanh", this.DEFAULT_RECURRENT_ACTIVATION = "hardSigmoid", this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal", this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal", this.DEFAULT_BIAS_INITIALIZER = "zeros", e.resetAfter) throw new og("GRUCell does not support reset_after parameter set to true.");
+                    this.units = e.units, Ag(this.units, "units"), this.activation = mx(void 0 === e.activation ? this.DEFAULT_ACTIVATION : e.activation), this.recurrentActivation = mx(void 0 === e.recurrentActivation ? this.DEFAULT_RECURRENT_ACTIVATION : e.recurrentActivation), this.useBias = null == e.useBias || e.useBias, this.kernelInitializer = jy(e.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER), this.recurrentInitializer = jy(e.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER), this.biasInitializer = jy(e.biasInitializer || this.DEFAULT_BIAS_INITIALIZER), this.kernelRegularizer = kx(e.kernelRegularizer), this.recurrentRegularizer = kx(e.recurrentRegularizer), this.biasRegularizer = kx(e.biasRegularizer), this.kernelConstraint = Lg(e.kernelConstraint), this.recurrentConstraint = Lg(e.recurrentConstraint), this.biasConstraint = Lg(e.biasConstraint), this.dropout = ay([1, sy([0, null == e.dropout ? 0 : e.dropout])]), this.recurrentDropout = ay([1, sy([0, null == e.recurrentDropout ? 0 : e.recurrentDropout])]), this.dropoutFunc = e.dropoutFunc, this.implementation = e.implementation, this.stateSize = this.units, this.dropoutMask = null, this.recurrentDropoutMask = null
+                }
+                build(e) {
+                    const t = (e = fb(e))[e.length - 1];
+                    this.kernel = this.addWeight("kernel", [t, 3 * this.units], null, this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, 3 * this.units], null, this.recurrentInitializer, this.recurrentRegularizer, !0, this.recurrentConstraint), this.useBias ? this.bias = this.addWeight("bias", [3 * this.units], null, this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint) : this.bias = null, this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        if (2 !== (e = e).length) throw new og(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);
+                        const n = null != t.training && t.training;
+                        let r = e[1];
+                        e = e[0], 0 < this.dropout && this.dropout < 1 && null == this.dropoutMask && (this.dropoutMask = aw({
+                            ones: () => Yc(e),
+                            rate: this.dropout,
+                            training: n,
+                            count: 3,
+                            dropoutFunc: this.dropoutFunc
+                        })), 0 < this.recurrentDropout && this.recurrentDropout < 1 && null == this.recurrentDropoutMask && (this.recurrentDropoutMask = aw({
+                            ones: () => Yc(r),
+                            rate: this.recurrentDropout,
+                            training: n,
+                            count: 3,
+                            dropoutFunc: this.dropoutFunc
+                        }));
+                        const a = this.dropoutMask,
+                            s = this.recurrentDropoutMask;
+                        let i, o, u;
+                        0 < this.dropout && this.dropout < 1 && (e = vu(e, a[0]));
+                        let l = gy(e, this.kernel.read());
+                        this.useBias && (l = xy(l, this.bias.read())), 0 < this.recurrentDropout && this.recurrentDropout < 1 && (r = vu(r, s[0]));
+                        const c = this.recurrentKernel.read(),
+                            [h, p] = zh(c, [2 * this.units, this.units], c.rank - 1),
+                            d = gy(r, h),
+                            [f, m, g] = zh(l, 3, l.rank - 1),
+                            [y, b] = zh(d, 2, d.rank - 1);
+                        i = this.recurrentActivation.apply(gu(f, y)), o = this.recurrentActivation.apply(gu(m, b));
+                        const v = gy(vu(o, r), p);
+                        u = this.activation.apply(gu(g, v));
+                        const x = gu(vu(i, r), vu(gu(1, pc(i)), u));
+                        return [x, x]
+                    })
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            units: this.units,
+                            activation: dx(this.activation),
+                            recurrentActivation: dx(this.recurrentActivation),
+                            useBias: this.useBias,
+                            kernelInitializer: Vy(this.kernelInitializer),
+                            recurrentInitializer: Vy(this.recurrentInitializer),
+                            biasInitializer: Vy(this.biasInitializer),
+                            kernelRegularizer: xx(this.kernelRegularizer),
+                            recurrentRegularizer: xx(this.recurrentRegularizer),
+                            biasRegularizer: xx(this.biasRegularizer),
+                            activityRegularizer: xx(this.activityRegularizer),
+                            kernelConstraint: Dg(this.kernelConstraint),
+                            recurrentConstraint: Dg(this.recurrentConstraint),
+                            biasConstraint: Dg(this.biasConstraint),
+                            dropout: this.dropout,
+                            recurrentDropout: this.recurrentDropout,
+                            implementation: this.implementation,
+                            resetAfter: !1
+                        };
+                    return Object.assign({}, e, t)
+                }
+            };
+            Jx.className = "GRUCell", Fo.registerClass(Jx);
+            var ew = class extends Yx {
+                constructor(e) {
+                    0 === e.implementation && console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."), e.cell = new Jx(e), super(e)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        null != this.cell.dropoutMask && (au(this.cell.dropoutMask), this.cell.dropoutMask = null), null != this.cell.recurrentDropoutMask && (au(this.cell.recurrentDropoutMask), this.cell.recurrentDropoutMask = null);
+                        const n = null == t ? null : t.mask,
+                            r = null == t ? null : t.training,
+                            a = null == t ? null : t.initialState;
+                        return super.call(e, {
+                            mask: n,
+                            training: r,
+                            initialState: a
+                        })
+                    })
+                }
+                static fromConfig(e, t) {
+                    return 0 === t.implmentation && (t.implementation = 1), new e(t)
+                }
+            };
+            ew.className = "GRU", Fo.registerClass(ew);
+            var tw = class extends Xx {
+                constructor(e) {
+                    super(e), this.DEFAULT_ACTIVATION = "tanh", this.DEFAULT_RECURRENT_ACTIVATION = "hardSigmoid", this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal", this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal", this.DEFAULT_BIAS_INITIALIZER = "zeros", this.units = e.units, Ag(this.units, "units"), this.activation = mx(void 0 === e.activation ? this.DEFAULT_ACTIVATION : e.activation), this.recurrentActivation = mx(void 0 === e.recurrentActivation ? this.DEFAULT_RECURRENT_ACTIVATION : e.recurrentActivation), this.useBias = null == e.useBias || e.useBias, this.kernelInitializer = jy(e.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER), this.recurrentInitializer = jy(e.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER), this.biasInitializer = jy(e.biasInitializer || this.DEFAULT_BIAS_INITIALIZER), this.unitForgetBias = e.unitForgetBias, this.kernelRegularizer = kx(e.kernelRegularizer), this.recurrentRegularizer = kx(e.recurrentRegularizer), this.biasRegularizer = kx(e.biasRegularizer), this.kernelConstraint = Lg(e.kernelConstraint), this.recurrentConstraint = Lg(e.recurrentConstraint), this.biasConstraint = Lg(e.biasConstraint), this.dropout = ay([1, sy([0, null == e.dropout ? 0 : e.dropout])]), this.recurrentDropout = ay([1, sy([0, null == e.recurrentDropout ? 0 : e.recurrentDropout])]), this.dropoutFunc = e.dropoutFunc, this.implementation = e.implementation, this.stateSize = [this.units, this.units], this.dropoutMask = null, this.recurrentDropoutMask = null
+                }
+                build(e) {
+                    var t;
+                    const n = (e = fb(e))[e.length - 1];
+                    let r;
+                    if (this.kernel = this.addWeight("kernel", [n, 4 * this.units], null, this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, 4 * this.units], null, this.recurrentInitializer, this.recurrentRegularizer, !0, this.recurrentConstraint), this.useBias) {
+                        if (this.unitForgetBias) {
+                            const e = this.biasInitializer,
+                                n = this.units;
+                            r = new((t = class extends Ay {
+                                apply(t, r) {
+                                    const a = e.apply([n]),
+                                        s = (new Ny).apply([n]),
+                                        i = e.apply([2 * n]);
+                                    return dy(dy(a, s), i)
+                                }
+                            }).className = "CustomInit", t)
+                        } else r = this.biasInitializer;
+                        this.bias = this.addWeight("bias", [4 * this.units], null, r, this.biasRegularizer, !0, this.biasConstraint)
+                    } else this.bias = null;
+                    this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        const n = null != t.training && t.training;
+                        if (3 !== (e = e).length) throw new og(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);
+                        let r = e[1];
+                        const a = e[2];
+                        e = e[0], 0 < this.dropout && this.dropout < 1 && null == this.dropoutMask && (this.dropoutMask = aw({
+                            ones: () => Yc(e),
+                            rate: this.dropout,
+                            training: n,
+                            count: 4,
+                            dropoutFunc: this.dropoutFunc
+                        })), 0 < this.recurrentDropout && this.recurrentDropout < 1 && null == this.recurrentDropoutMask && (this.recurrentDropoutMask = aw({
+                            ones: () => Yc(r),
+                            rate: this.recurrentDropout,
+                            training: n,
+                            count: 4,
+                            dropoutFunc: this.dropoutFunc
+                        }));
+                        const s = this.dropoutMask,
+                            i = this.recurrentDropoutMask;
+                        let o, u, l, c;
+                        0 < this.dropout && this.dropout < 1 && (e = vu(e, s[0]));
+                        let h = gy(e, this.kernel.read());
+                        0 < this.recurrentDropout && this.recurrentDropout < 1 && (r = vu(r, i[0])), h = gu(h, gy(r, this.recurrentKernel.read())), this.useBias && (h = xy(h, this.bias.read()));
+                        const [p, d, f, m] = zh(h, 4, h.rank - 1);
+                        o = this.recurrentActivation.apply(p), u = this.recurrentActivation.apply(d), l = gu(vu(u, a), vu(o, this.activation.apply(f))), c = this.recurrentActivation.apply(m);
+                        const g = vu(c, this.activation.apply(l));
+                        return [g, g, l]
+                    })
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            units: this.units,
+                            activation: dx(this.activation),
+                            recurrentActivation: dx(this.recurrentActivation),
+                            useBias: this.useBias,
+                            kernelInitializer: Vy(this.kernelInitializer),
+                            recurrentInitializer: Vy(this.recurrentInitializer),
+                            biasInitializer: Vy(this.biasInitializer),
+                            unitForgetBias: this.unitForgetBias,
+                            kernelRegularizer: xx(this.kernelRegularizer),
+                            recurrentRegularizer: xx(this.recurrentRegularizer),
+                            biasRegularizer: xx(this.biasRegularizer),
+                            activityRegularizer: xx(this.activityRegularizer),
+                            kernelConstraint: Dg(this.kernelConstraint),
+                            recurrentConstraint: Dg(this.recurrentConstraint),
+                            biasConstraint: Dg(this.biasConstraint),
+                            dropout: this.dropout,
+                            recurrentDropout: this.recurrentDropout,
+                            implementation: this.implementation
+                        };
+                    return Object.assign({}, e, t)
+                }
+            };
+            tw.className = "LSTMCell", Fo.registerClass(tw);
+            var nw = class extends Yx {
+                constructor(e) {
+                    0 === e.implementation && console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."), e.cell = new tw(e), super(e)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        null != this.cell.dropoutMask && (au(this.cell.dropoutMask), this.cell.dropoutMask = null), null != this.cell.recurrentDropoutMask && (au(this.cell.recurrentDropoutMask), this.cell.recurrentDropoutMask = null);
+                        const n = null == t ? null : t.mask,
+                            r = null == t ? null : t.training,
+                            a = null == t ? null : t.initialState;
+                        return super.call(e, {
+                            mask: n,
+                            training: r,
+                            initialState: a
+                        })
+                    })
+                }
+                static fromConfig(e, t) {
+                    return 0 === t.implmentation && (t.implementation = 1), new e(t)
+                }
+            };
+            nw.className = "LSTM", Fo.registerClass(nw);
+            var rw = class extends Xx {
+                constructor(e) {
+                    super(e), this.cells = e.cells
+                }
+                get stateSize() {
+                    const e = [];
+                    for (const t of this.cells.slice().reverse()) Array.isArray(t.stateSize) ? e.push(...t.stateSize) : e.push(t.stateSize);
+                    return e
+                }
+                call(e, t) {
+                    return ru(() => {
+                        let n = (e = e).slice(1);
+                        const r = [];
+                        for (const e of this.cells.slice().reverse()) Array.isArray(e.stateSize) ? r.push(n.splice(0, e.stateSize.length)) : r.push(n.splice(0, 1));
+                        r.reverse();
+                        const a = [];
+                        let s;
+                        for (let i = 0; i < this.cells.length; ++i) {
+                            const o = this.cells[i];
+                            n = r[i], s = 0 === i ? [e[0]].concat(n) : [s[0]].concat(n), s = o.call(s, t), a.push(s.slice(1))
+                        }
+                        n = [];
+                        for (const e of a.slice().reverse()) n.push(...e);
+                        return [s[0]].concat(n)
+                    })
+                }
+                build(e) {
+                    let t;
+                    hb(e) && (e = e[0]), e = e, this.cells.forEach((n, r) => {
+                        Zg("RNNCell_" + r, () => {
+                            n.build(e), t = Array.isArray(n.stateSize) ? n.stateSize[0] : n.stateSize, e = [e[0], t]
+                        })
+                    }), this.built = !0
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            cells: this.cells.map(e => ({
+                                className: e.getClassName(),
+                                config: e.getConfig()
+                            }))
+                        };
+                    return Object.assign({}, e, t)
+                }
+                static fromConfig(e, t, n = {}) {
+                    const r = [];
+                    for (const e of t.cells) r.push(Bb(e, n));
+                    return new e({
+                        cells: r
+                    })
+                }
+                get trainableWeights() {
+                    if (!this.trainable) return [];
+                    const e = [];
+                    for (const t of this.cells) e.push(...t.trainableWeights);
+                    return e
+                }
+                get nonTrainableWeights() {
+                    const e = [];
+                    for (const t of this.cells) e.push(...t.nonTrainableWeights);
+                    if (!this.trainable) {
+                        const t = [];
+                        for (const e of this.cells) t.push(...e.trainableWeights);
+                        return t.concat(e)
+                    }
+                    return e
+                }
+                getWeights() {
+                    const e = [];
+                    for (const t of this.cells) e.push(...t.weights);
+                    return yb(e)
+                }
+                setWeights(e) {
+                    const t = [];
+                    for (const n of this.cells) {
+                        const r = n.weights.length,
+                            a = e.splice(r);
+                        for (let e = 0; e < n.weights.length; ++e) t.push([n.weights[e], a[e]])
+                    }
+                    bb(t)
+                }
+            };
+
+            function aw(e) {
+                const {
+                    ones: t,
+                    rate: n,
+                    training: r = !1,
+                    count: a = 1,
+                    dropoutFunc: s
+                } = e, i = () => null != s ? s(t(), n) : wy(t(), n), o = () => ky(i, t, r);
+                if (!a || a <= 1) return su(o().clone());
+                return Array(a).fill(void 0).map(o).map(e => su(e.clone()))
+            }
+            rw.className = "StackedRNNCells", Fo.registerClass(rw);
+            var sw = class extends Yx {
+                constructor(e) {
+                    if (e.unroll) throw new ug("Unrolling is not possible with convolutional RNNs.");
+                    if (Array.isArray(e.cell)) throw new ug("It is not possible at the moment to stack convolutional cells.");
+                    super(e), this.inputSpec = [new vb({
+                        ndim: 5
+                    })]
+                }
+                call(e, t) {
+                    return ru(() => {
+                        if (null != this.cell.dropoutMask && (au(this.cell.dropoutMask), this.cell.dropoutMask = null), null != this.cell.recurrentDropoutMask && (au(this.cell.recurrentDropoutMask), this.cell.recurrentDropoutMask = null), t && t.constants) throw new og("ConvRNN2D cell does not support constants");
+                        const n = null == t ? null : t.mask,
+                            r = null == t ? null : t.training,
+                            a = null == t ? null : t.initialState;
+                        return super.call(e, {
+                            mask: n,
+                            training: r,
+                            initialState: a
+                        })
+                    })
+                }
+                computeOutputShape(e) {
+                    let t = this.computeSingleOutputShape(e);
+                    return this.returnSequences || (t = [t[0], ...t.slice(2)]), this.returnState && (t = [t, ...Array(2).fill([e[0], ...t.slice(-3)])]), t
+                }
+                getInitialState(e) {
+                    return ru(() => {
+                        const {
+                            stateSize: t
+                        } = this.cell, n = e.shape, r = this.computeSingleOutputShape(n), a = Pc([r[0], ...r.slice(2)]);
+                        return Array.isArray(t) ? Array(t.length).fill(a) : [a]
+                    })
+                }
+                resetStates(e, t = !1) {
+                    ru(() => {
+                        if (!this.stateful) throw new sg("Cannot call resetStates() on an RNN Layer that is not stateful.");
+                        const n = this.inputSpec[0].shape,
+                            r = this.computeSingleOutputShape(n),
+                            a = [r[0], ...r.slice(2)];
+                        if (null == n[0]) throw new og("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");
+                        if (null == this.getStates()) Array.isArray(this.cell.stateSize) ? this.states_ = this.cell.stateSize.map(() => Pc(a)) : this.states_ = [Pc(a)];
+                        else if (null == e) au(this.states_), null != this.keptStates && (au(this.keptStates), this.keptStates = []), Array.isArray(this.cell.stateSize) ? this.states_ = this.cell.stateSize.map(() => Pc(a)) : this.states_[0] = Pc(a);
+                        else {
+                            if (Array.isArray(e) || (e = [e]), e.length !== this.states_.length) throw new og(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);
+                            t ? this.keptStates.push(this.states_.slice()) : au(this.states_);
+                            for (let t = 0; t < this.states_.length; ++t) {
+                                const n = e[t],
+                                    r = a;
+                                if (!va.arraysEqual(n.shape, r)) throw new og(`State ${t} is incompatible with layer ${this.name}: expected shape=${r}, received shape=${n.shape}`);
+                                this.states_[t] = n
+                            }
+                        }
+                        this.states_ = this.states_.map(e => su(e.clone()))
+                    })
+                }
+                computeSingleOutputShape(e) {
+                    const {
+                        dataFormat: t,
+                        filters: n,
+                        kernelSize: r,
+                        padding: a,
+                        strides: s,
+                        dilationRate: i
+                    } = this.cell, o = "channelsFirst" === t, u = e[o ? 3 : 2], l = e[o ? 4 : 3], c = Rx(u, r[0], a, s[0], i[0]), h = Rx(l, r[1], a, s[1], i[1]);
+                    return [...e.slice(0, 2), ...o ? [n, c, h] : [c, h, n]]
+                }
+            };
+            sw.className = "ConvRNN2D";
+            var iw = class extends tw {
+                constructor(e) {
+                    const {
+                        filters: t,
+                        kernelSize: n,
+                        strides: r,
+                        padding: a,
+                        dataFormat: s,
+                        dilationRate: i
+                    } = e;
+                    super(Object.assign({}, e, {
+                        units: t
+                    })), this.filters = t, Ag(this.filters, "filters"), this.kernelSize = Tx(n, 2, "kernelSize"), this.kernelSize.forEach(e => Ag(e, "kernelSize")), this.strides = Tx(r || 1, 2, "strides"), this.strides.forEach(e => Ag(e, "strides")), this.padding = a || "valid", Yg(this.padding), this.dataFormat = s || "channelsLast", Kg(this.dataFormat), this.dilationRate = Tx(i || 1, 2, "dilationRate"), this.dilationRate.forEach(e => Ag(e, "dilationRate"))
+                }
+                build(e) {
+                    var t;
+                    e = fb(e);
+                    const n = "channelsFirst" === this.dataFormat ? 1 : e.length - 1;
+                    if (null == e[n]) throw new og("The channel dimension of the input should be defined. Found " + e[n]);
+                    const r = e[n],
+                        a = this.kernelSize.concat([r, 4 * this.filters]);
+                    this.kernel = this.addWeight("kernel", a, null, this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint);
+                    const s = this.kernelSize.concat([this.filters, 4 * this.filters]);
+                    if (this.recurrentKernel = this.addWeight("recurrent_kernel", s, null, this.recurrentInitializer, this.recurrentRegularizer, !0, this.recurrentConstraint), this.useBias) {
+                        let e;
+                        if (this.unitForgetBias) {
+                            const n = this.biasInitializer,
+                                r = this.filters;
+                            e = new((t = class extends Ay {
+                                apply(e, t) {
+                                    return py([n.apply([r]), $c([r]), n.apply([2 * r])])
+                                }
+                            }).className = "CustomInit", t)
+                        } else e = this.biasInitializer;
+                        this.bias = this.addWeight("bias", [4 * this.filters], null, e, this.biasRegularizer, !0, this.biasConstraint)
+                    }
+                    this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        if (3 !== e.length) throw new og(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);
+                        const n = t.training || !1,
+                            r = e[0],
+                            a = e[1],
+                            s = e[2];
+                        0 < this.dropout && this.dropout < 1 && null == this.dropoutMask && (this.dropoutMask = aw({
+                            ones: () => Yc(r),
+                            rate: this.dropout,
+                            training: n,
+                            count: 4,
+                            dropoutFunc: this.dropoutFunc
+                        }));
+                        const i = this.dropoutMask,
+                            o = (e, t, n) => t && t[n] ? vu(t[n], e) : e;
+                        let u = o(r, i, 0),
+                            l = o(r, i, 1),
+                            c = o(r, i, 2),
+                            h = o(r, i, 3);
+                        0 < this.recurrentDropout && this.recurrentDropout < 1 && null == this.recurrentDropoutMask && (this.recurrentDropoutMask = aw({
+                            ones: () => Yc(a),
+                            rate: this.recurrentDropout,
+                            training: n,
+                            count: 4,
+                            dropoutFunc: this.dropoutFunc
+                        }));
+                        const p = this.recurrentDropoutMask;
+                        let d = o(a, p, 0),
+                            f = o(a, p, 1),
+                            m = o(a, p, 2),
+                            g = o(a, p, 3);
+                        const [y, b, v, x] = zh(this.kernel.read(), 4, 3), [w, k, S, I] = this.useBias ? zh(this.bias.read(), 4) : [null, null, null, null];
+                        u = this.inputConv(u, y, w, this.padding), l = this.inputConv(l, b, k, this.padding), c = this.inputConv(c, v, S, this.padding), h = this.inputConv(h, x, I, this.padding);
+                        const [A, E, N, C] = zh(this.recurrentKernel.read(), 4, 3);
+                        d = this.recurrentConv(d, A), f = this.recurrentConv(f, E), m = this.recurrentConv(m, N), g = this.recurrentConv(g, C);
+                        const T = this.recurrentActivation.apply(gu(u, d)),
+                            R = this.recurrentActivation.apply(gu(l, f)),
+                            _ = gu(vu(R, s), vu(T, this.activation.apply(gu(c, m)))),
+                            F = vu(this.recurrentActivation.apply(gu(h, g)), this.activation.apply(_));
+                        return [F, F, _]
+                    })
+                }
+                getConfig() {
+                    const e = function (e, t) {
+                            var n = {};
+                            for (var r in e) Object.prototype.hasOwnProperty.call(e, r) && t.indexOf(r) < 0 && (n[r] = e[r]);
+                            if (null != e && "function" == typeof Object.getOwnPropertySymbols) {
+                                var a = 0;
+                                for (r = Object.getOwnPropertySymbols(e); a < r.length; a++) t.indexOf(r[a]) < 0 && Object.prototype.propertyIsEnumerable.call(e, r[a]) && (n[r[a]] = e[r[a]])
+                            }
+                            return n
+                        }(super.getConfig(), ["units"]),
+                        t = {
+                            filters: this.filters,
+                            kernelSize: this.kernelSize,
+                            padding: this.padding,
+                            dataFormat: this.dataFormat,
+                            dilationRate: this.dilationRate,
+                            strides: this.strides
+                        };
+                    return Object.assign({}, e, t)
+                }
+                inputConv(e, t, n, r) {
+                    const a = ml(e, t, this.strides, r || "valid", "channelsFirst" === this.dataFormat ? "NCHW" : "NHWC", this.dilationRate);
+                    return n ? xy(a, n, this.dataFormat) : a
+                }
+                recurrentConv(e, t) {
+                    return ml(e, t, 1, "same", "channelsFirst" === this.dataFormat ? "NCHW" : "NHWC")
+                }
+            };
+            iw.className = "ConvLSTM2DCell", Fo.registerClass(iw);
+            var ow = class extends sw {
+                constructor(e) {
+                    const t = new iw(e);
+                    super(Object.assign({}, e, {
+                        cell: t
+                    }))
+                }
+                static fromConfig(e, t) {
+                    return new e(t)
+                }
+            };
+            ow.className = "ConvLSTM2D", Fo.registerClass(ow);
+            var uw = class extends Ib {
+                constructor(e) {
+                    super(e), this.rate = Math.max(Math.min(e.rate, 1), 0), this.noiseShape = e.noiseShape, this.seed = e.seed, this.supportsMasking = !0
+                }
+                getNoiseShape(e) {
+                    if (null == this.noiseShape) return this.noiseShape;
+                    const t = e.shape,
+                        n = [];
+                    for (let e = 0; e < this.noiseShape.length; ++e) n.push(null == this.noiseShape[e] ? t[e] : this.noiseShape[e]);
+                    return n
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e);
+                        if (0 < this.rate && this.rate < 1) {
+                            const e = null != t.training && t.training,
+                                r = this.getNoiseShape(n);
+                            return ky(() => wy(n, this.rate, r, this.seed), () => n, e)
+                        }
+                        return e
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            rate: this.rate,
+                            noiseShape: this.noiseShape,
+                            seed: this.seed
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+                dispose() {
+                    return super.dispose()
+                }
+            };
+            uw.className = "Dropout", Fo.registerClass(uw);
+            var lw = class extends uw {
+                constructor(e) {
+                    super(e), this.inputSpec = [{
+                        ndim: 3
+                    }]
+                }
+                getNoiseShape(e) {
+                    const t = e.shape;
+                    return [t[0], 1, t[2]]
+                }
+            };
+            lw.className = "SpatialDropout1D", Fo.registerClass(lw);
+            var cw = class extends Ib {
+                constructor(e) {
+                    if (super(e), this.activation = null, this.useBias = !0, this.kernel = null, this.bias = null, this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal", this.DEFAULT_BIAS_INITIALIZER = "zeros", null == e.batchInputShape && null == e.inputShape && null != e.inputDim) {
+                        let t = null;
+                        null != e.batchSize && (t = e.batchSize), this.batchInputShape = [t, e.inputDim]
+                    }
+                    this.units = e.units, Ag(this.units, "units"), this.activation = mx(e.activation), null != e.useBias && (this.useBias = e.useBias), this.kernelInitializer = jy(e.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER), this.biasInitializer = jy(e.biasInitializer || this.DEFAULT_BIAS_INITIALIZER), this.kernelConstraint = Lg(e.kernelConstraint), this.biasConstraint = Lg(e.biasConstraint), this.kernelRegularizer = kx(e.kernelRegularizer), this.biasRegularizer = kx(e.biasRegularizer), this.activityRegularizer = kx(e.activityRegularizer), this.supportsMasking = !0, this.inputSpec = [{
+                        minNDim: 2
+                    }]
+                }
+                build(e) {
+                    const t = (e = fb(e))[e.length - 1];
+                    null == this.kernel && (this.kernel = this.addWeight("kernel", [t, this.units], null, this.kernelInitializer, this.kernelRegularizer, !0, this.kernelConstraint), this.useBias && (this.bias = this.addWeight("bias", [this.units], null, this.biasInitializer, this.biasRegularizer, !0, this.biasConstraint))), this.inputSpec = [{
+                        minNDim: 2,
+                        axes: {
+                            [-1]: t
+                        }
+                    }], this.built = !0
+                }
+                computeOutputShape(e) {
+                    const t = (e = fb(e)).slice();
+                    return t[t.length - 1] = this.units, t
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e),
+                            r = Eg(this.activation.getClassName());
+                        let a;
+                        return null != r ? a = gy(n, this.kernel.read(), r, this.bias ? this.bias.read() : null) : (a = gy(n, this.kernel.read()), null != this.bias && (a = xy(a, this.bias.read())), null != this.activation && (a = this.activation.apply(a))), a
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            units: this.units,
+                            activation: dx(this.activation),
+                            useBias: this.useBias,
+                            kernelInitializer: Vy(this.kernelInitializer),
+                            biasInitializer: Vy(this.biasInitializer),
+                            kernelRegularizer: xx(this.kernelRegularizer),
+                            biasRegularizer: xx(this.biasRegularizer),
+                            activityRegularizer: xx(this.activityRegularizer),
+                            kernelConstraint: Dg(this.kernelConstraint),
+                            biasConstraint: Dg(this.biasConstraint)
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            cw.className = "Dense", Fo.registerClass(cw);
+            var hw = class extends Ib {
+                constructor(e) {
+                    super(e = e || {}), this.inputSpec = [{
+                        minNDim: 3
+                    }], this.dataFormat = e.dataFormat
+                }
+                computeOutputShape(e) {
+                    e = fb(e);
+                    for (const t of e.slice(1))
+                        if (null == t) throw new og(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);
+                    return [e[0], ry(e, 1)]
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        let n = db(e);
+                        if ("channelsFirst" === this.dataFormat && n.rank > 1) {
+                            const e = [0];
+                            for (let t = 2; t < n.rank; ++t) e.push(t);
+                            e.push(1), n = qi(n, e)
+                        }
+                        return function (e) {
+                            if (e.rank <= 1) throw new og(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);
+                            const t = [e.shape[0], ry(e.shape, 1)];
+                            return qu(e, t)
+                        }(n)
+                    })
+                }
+                getConfig() {
+                    const e = {};
+                    null != this.dataFormat && (e.dataFormat = this.dataFormat);
+                    const t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            hw.className = "Flatten", Fo.registerClass(hw);
+            var pw = class extends Ib {
+                constructor(e) {
+                    super(e), this.supportsMasking = !0, this.activation = mx(e.activation)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e);
+                        return this.activation.apply(n)
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            activation: dx(this.activation)
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            pw.className = "Activation", Fo.registerClass(pw);
+            var dw = class extends Ib {
+                constructor(e) {
+                    super(e), this.n = e.n, this.inputSpec = [{
+                        ndim: 2
+                    }]
+                }
+                computeOutputShape(e) {
+                    return [e[0], this.n, e[1]]
+                }
+                call(e, t) {
+                    return ru(() => {
+                        return e = db(e), t = e, n = this.n, ru(() => {
+                            if (2 !== t.shape.length) throw new og(`repeat() expects a rank-2 tensor, but received a rank-${t.shape.length} tensor.`);
+                            return fy(uy(t, 1), [1, n, 1])
+                        });
+                        var t, n
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            n: this.n
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            dw.className = "RepeatVector", Fo.registerClass(dw);
+            var fw = class extends Ib {
+                constructor(e) {
+                    super(e), this.targetShape = e.targetShape;
+                    for (let e = 0; e < this.targetShape.length; ++e) this.isUnknown(this.targetShape[e]) && (this.targetShape[e] = null)
+                }
+                isUnknown(e) {
+                    return e < 0 || null == e
+                }
+                fixUnknownDimension(e, t) {
+                    const n = "Total size of new array must be unchanged.",
+                        r = t.slice();
+                    let a = 1,
+                        s = null;
+                    for (let e = 0; e < r.length; ++e) {
+                        const t = r[e];
+                        if (this.isUnknown(t)) {
+                            if (null !== s) throw new og("Can only specifiy one unknown dimension.");
+                            s = e
+                        } else a *= t
+                    }
+                    const i = ry(e);
+                    if (null !== s) {
+                        if (0 === a || i % a != 0) throw new og(n);
+                        r[s] = i / a
+                    } else if (i !== a) throw new og(n);
+                    return r
+                }
+                computeOutputShape(e) {
+                    let t = !1;
+                    for (let n = 0; n < e.length; ++n)
+                        if (this.isUnknown(e[n])) {
+                            t = !0;
+                            break
+                        } return t ? e.slice(0, 1).concat(this.targetShape) : e.slice(0, 1).concat(this.fixUnknownDimension(e.slice(1), this.targetShape))
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e),
+                            r = n.shape,
+                            a = r.slice(0, 1).concat(this.fixUnknownDimension(r.slice(1), this.targetShape));
+                        return qu(n, a)
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            targetShape: this.targetShape
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            fw.className = "Reshape", Fo.registerClass(fw);
+            var mw = class extends Ib {
+                constructor(e) {
+                    if (super(e), null == e.dims) throw new Error("Required configuration field `dims` is missing during Permute constructor call.");
+                    if (!Array.isArray(e.dims)) throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);
+                    const t = iy(1, e.dims.length + 1);
+                    if (!va.arraysEqual(e.dims.slice().sort(), t)) throw new Error("Invalid permutation `dims`: " + JSON.stringify(e.dims) + " `dims` must contain consecutive integers starting from 1.");
+                    this.dims = e.dims, this.dimsIncludingBatch = [0].concat(this.dims), this.inputSpec = [new vb({
+                        ndim: this.dims.length + 1
+                    })]
+                }
+                computeOutputShape(e) {
+                    const t = (e = fb(e)).slice();
+                    return this.dims.forEach((n, r) => {
+                        t[r + 1] = e[n]
+                    }), t
+                }
+                call(e, t) {
+                    return qi(db(e), this.dimsIncludingBatch)
+                }
+                getConfig() {
+                    const e = {
+                            dims: this.dims
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            mw.className = "Permute", Fo.registerClass(mw);
+            var gw = class extends Ib {
+                constructor(e) {
+                    super(null == e ? {} : e), this.supportsMasking = !0, this.maskValue = null != e ? null == e.maskValue ? 0 : e.maskValue : 0
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            maskValue: this.maskValue
+                        };
+                    return Object.assign(t, e), t
+                }
+                computeMask(e, t) {
+                    const n = db(e);
+                    return Au(Kc(n, this.maskValue), -1)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e),
+                            r = Au(Kc(n, this.maskValue), -1, !0);
+                        return vu(n, Ai(r, n.dtype))
+                    })
+                }
+            };
+            gw.className = "Masking", Fo.registerClass(gw);
+            var yw = class extends Ib {
+                constructor(e) {
+                    if (super(e), this.embeddings = null, this.DEFAULT_EMBEDDINGS_INITIALIZER = "randomUniform", null == e.batchInputShape && null == e.inputShape) {
+                        let t = null;
+                        null != e.batchSize && (t = e.batchSize), null == e.inputLength ? this.batchInputShape = [t, null] : this.batchInputShape = [t].concat(fg(e.inputLength))
+                    }
+                    this.inputDim = e.inputDim, Ag(this.inputDim, "inputDim"), this.outputDim = e.outputDim, Ag(this.outputDim, "outputDim"), this.embeddingsInitializer = jy(e.embeddingsInitializer || this.DEFAULT_EMBEDDINGS_INITIALIZER), this.embeddingsRegularizer = kx(e.embeddingsRegularizer), this.activityRegularizer = kx(e.activityRegularizer), this.embeddingsConstraint = Lg(e.embeddingsConstraint), this.maskZero = e.maskZero, this.supportsMasking = e.maskZero, this.inputLength = e.inputLength
+                }
+                build(e) {
+                    this.embeddings = this.addWeight("embeddings", [this.inputDim, this.outputDim], this.dtype, this.embeddingsInitializer, this.embeddingsRegularizer, !0, this.embeddingsConstraint), this.built = !0
+                }
+                warnOnIncompatibleInputShape(e) {}
+                computeMask(e, t) {
+                    return ru(() => this.maskZero ? (e = db(e), Kc(e, Fl(e))) : null)
+                }
+                computeOutputShape(e) {
+                    if (e = fb(e), null == this.inputLength) return [...e, this.outputDim];
+                    const t = fg(this.inputLength);
+                    if (t.length !== e.length - 1) throw new og(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`); {
+                        let n = 0;
+                        for (let r = 0; r < t.length; ++r) {
+                            const a = t[r],
+                                s = e[r + 1];
+                            if (null != a && null != s && a !== s) throw new og(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);
+                            null == a && (t[n] = s), n++
+                        }
+                    }
+                    return [e[0], ...t, this.outputDim]
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        let n = db(e);
+                        "int32" !== n.dtype && (n = oy(n, "int32"));
+                        const r = yy(this.embeddings.read(), qu(n, [n.size]));
+                        return qu(r, fb(this.computeOutputShape(n.shape)))
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            inputDim: this.inputDim,
+                            outputDim: this.outputDim,
+                            embeddingsInitializer: Vy(this.embeddingsInitializer),
+                            embeddingsRegularizer: xx(this.embeddingsRegularizer),
+                            activityRegularizer: xx(this.activityRegularizer),
+                            embeddingsConstraint: Dg(this.embeddingsConstraint),
+                            maskZero: this.maskZero,
+                            inputLength: this.inputLength
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            yw.className = "Embedding", Fo.registerClass(yw);
+            var bw = class extends Ib {
+                    constructor(e) {
+                        super(e || {}), this.supportsMasking = !0
+                    }
+                    mergeFunction(e) {
+                        throw new ug
+                    }
+                    computeElementwiseOpOutputShape(e, t) {
+                        if (null == e || null == t) return null;
+                        if (e.length < t.length) return this.computeElementwiseOpOutputShape(t, e);
+                        if (0 === t.length) return e;
+                        const n = e.slice(0, e.length - t.length);
+                        for (let r = 0; r < t.length; ++r) {
+                            const a = e[e.length - t.length + r],
+                                s = t[r];
+                            if (null == a || null == s || a < 0 || s < 0) n.push(null);
+                            else if (1 === a) n.push(s);
+                            else if (1 === s) n.push(a);
+                            else {
+                                if (a !== s) throw new og("Operands could not be broadcast together with shapes " + JSON.stringify(e) + " " + JSON.stringify(t));
+                                n.push(a)
+                            }
+                        }
+                        return n
+                    }
+                    build(e) {
+                        if (Array.isArray(e) && !Array.isArray(e[0]) && (e = [fb(e)]), (e = e).length < 2) throw new og(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);
+                        let t = [];
+                        for (const n of e) null != n && null !== n[0] && t.push(n[0]);
+                        if (t = wg(t), t.length > 1) throw new og(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);
+                        let n = null == e[0] ? null : e[0].slice(1);
+                        for (let t = 1; t < e.length; ++t) {
+                            const r = null == e[t] ? null : e[t].slice(1);
+                            n = this.computeElementwiseOpOutputShape(n, r)
+                        }
+                        const r = e.map(e => e.length); - 1 === e.indexOf(null) && 1 === wg(r).length ? this.reshapeRequired = !1 : this.reshapeRequired = !0
+                    }
+                    call(e, t) {
+                        return ru(() => {
+                            if (e = e, this.reshapeRequired) {
+                                const t = [],
+                                    n = e.map(e => e.rank);
+                                if (-1 === n.indexOf(null)) {
+                                    const r = sy(n);
+                                    for (let n of e) {
+                                        const e = n.rank;
+                                        for (let t = 0; t < r - e; ++t) n = uy(n, 1);
+                                        t.push(n)
+                                    }
+                                    return this.mergeFunction(t)
+                                } {
+                                    let n = !1;
+                                    for (const r of e) {
+                                        const e = r.rank;
+                                        if (null == e) {
+                                            const e = r.shape,
+                                                a = e[0],
+                                                s = e.slice(1).concat([a]);
+                                            let i = qu(r, [a].concat(ry(e.slice(1))));
+                                            i = qi(i, [1, 0]), i = qu(i, s), t.push(i), n = !0
+                                        } else if (e > 1) {
+                                            const a = iy(1, e).concat([0]);
+                                            t.push(qi(r, a)), n = !0
+                                        } else t.push(r)
+                                    }
+                                    let r = this.mergeFunction(t);
+                                    const a = r.rank;
+                                    if (n)
+                                        if (null == a) {
+                                            const e = r.shape,
+                                                t = e[e.length - 1],
+                                                n = [t].concat(e.slice(0, e.length - 1));
+                                            r = qu(qi(qu(r, [-1, t]), [1, 0]), n)
+                                        } else if (a > 1) {
+                                        const e = [a - 1].concat(iy(0, a - 1));
+                                        r = qi(r, e)
+                                    }
+                                    return r
+                                }
+                            }
+                            return this.mergeFunction(e)
+                        })
+                    }
+                    computeOutputShape(e) {
+                        let t;
+                        t = null == (e = e)[0] ? null : e[0].slice(1);
+                        for (let n = 1; n < e.length; ++n) {
+                            const r = null == e[n] ? null : e[n].slice(1);
+                            t = this.computeElementwiseOpOutputShape(t, r)
+                        }
+                        let n = [];
+                        for (const t of e) null != t && null !== t[0] && n.push(t[0]);
+                        return n = wg(n), t = 1 === n.length ? n.concat(t) : [null].concat(t), t
+                    }
+                    computeMask(e, t) {
+                        return ru(() => {
+                            if (null == t) return null;
+                            if (!Array.isArray(t)) throw new og("`mask` should be an Array");
+                            if (!Array.isArray(e)) throw new og("`inputs` should be an Array");
+                            if (t.length !== e.length) throw new og(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);
+                            if (t.every(e => null == e)) return null;
+                            let n = (t = t.map(e => null == e ? e : Bl(e, 0)))[0];
+                            for (let e = 1; e < t.length - 1; ++e) n = Cc(n, t[e]);
+                            return n
+                        })
+                    }
+                },
+                vw = class extends bw {
+                    constructor(e) {
+                        super(e)
+                    }
+                    mergeFunction(e) {
+                        return ru(() => {
+                            let t = e[0].clone();
+                            for (let n = 1; n < e.length; ++n) t = gu(t, e[n]);
+                            return t
+                        })
+                    }
+                };
+            vw.className = "Add", Fo.registerClass(vw);
+            var xw = class extends bw {
+                constructor(e) {
+                    super(e)
+                }
+                mergeFunction(e) {
+                    return ru(() => {
+                        let t = e[0].clone();
+                        for (let n = 1; n < e.length; ++n) t = vu(t, e[n]);
+                        return t
+                    })
+                }
+            };
+            xw.className = "Multiply", Fo.registerClass(xw);
+            var ww = class extends bw {
+                constructor(e) {
+                    super(e)
+                }
+                mergeFunction(e) {
+                    return ru(() => {
+                        let t = e[0].clone();
+                        for (let n = 1; n < e.length; ++n) t = gu(t, e[n]);
+                        return vu(1 / e.length, t)
+                    })
+                }
+            };
+            ww.className = "Average", Fo.registerClass(ww);
+            var kw = class extends bw {
+                constructor(e) {
+                    super(e)
+                }
+                mergeFunction(e) {
+                    return ru(() => {
+                        let t = e[0];
+                        for (let n = 1; n < e.length; ++n) t = Oc(t, e[n]);
+                        return t
+                    })
+                }
+            };
+            kw.className = "Maximum", Fo.registerClass(kw);
+            var Sw = class extends bw {
+                constructor(e) {
+                    super(e)
+                }
+                mergeFunction(e) {
+                    return ru(() => {
+                        let t = e[0];
+                        for (let n = 1; n < e.length; ++n) t = Uc(t, e[n]);
+                        return t
+                    })
+                }
+            };
+            Sw.className = "Minimum", Fo.registerClass(Sw);
+            var Iw = class extends bw {
+                constructor(e) {
+                    super(e), this.DEFAULT_AXIS = -1, null == e && (e = {}), this.axis = null == e.axis ? this.DEFAULT_AXIS : e.axis, this.supportsMasking = !0, this.reshapeRequired = !1
+                }
+                build(e) {
+                    if (!Array.isArray(e) || !Array.isArray(e[0]) || 1 === e.length) throw new og("A `Concatenate` layer should be called on a list of at least 2 inputs");
+                    e = e;
+                    let t = !0;
+                    for (const n of e)
+                        if (null != n) {
+                            t = !1;
+                            break
+                        } if (t) return;
+                    const n = [];
+                    for (let t = 0; t < e.length; ++t) {
+                        const r = e[t].slice();
+                        r.splice(this.axis, 1);
+                        let a = !1;
+                        for (const e of n)
+                            if (va.arraysEqual(e, r)) {
+                                a = !0;
+                                break
+                            } a || n.push(r)
+                    }
+                    if (n.length > 1) throw new og("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: " + JSON.stringify(e))
+                }
+                mergeFunction(e) {
+                    return ru(() => py(e, this.axis))
+                }
+                computeOutputShape(e) {
+                    if (!Array.isArray(e) || !Array.isArray(e[0])) throw new og("A `Concatenate` layer should be called on a list of inputs.");
+                    const t = e,
+                        n = t[0].slice(),
+                        r = this.axis < 0 ? n.length + this.axis : this.axis;
+                    for (const e of t.slice(1)) {
+                        if (null == n[r] || null == e[r]) {
+                            n[r] = null;
+                            break
+                        }
+                        n[r] += e[r]
+                    }
+                    return n
+                }
+                computeMask(e, t) {
+                    if (null == t) return null;
+                    if (!Array.isArray(t)) throw new og("`mask` should be an array for Concatenate");
+                    if (!Array.isArray(e)) throw new og("`inputs` should be an array for Concatenate");
+                    if (t.length !== e.length) throw new og(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);
+                    return ru(() => {
+                        let n = !0;
+                        if (t.forEach(e => {
+                                null == e || (n = !1)
+                            }), n) return null;
+                        const r = [];
+                        for (let n = 0; n < e.length; ++n) null == t[n] ? r.push(Ai(Yc(e[n]), "bool")) : t[n].rank < e[n].rank ? r.push(Bl(t[n], -1)) : r.push(t[n]);
+                        const a = Xu(r, this.axis);
+                        return Iu(a, -1, !1)
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            axis: this.axis
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+
+            function Aw(e, t) {
+                for (; e < 0;) e += t;
+                return e
+            }
+            Iw.className = "Concatenate", Fo.registerClass(Iw);
+            var Ew = class extends bw {
+                constructor(e) {
+                    super(e), this.axes = e.axes, this.normalize = null != e.normalize && e.normalize, this.supportsMasking = !0, this.reshapeRequired = !1
+                }
+                build(e) {
+                    va.assert(Array.isArray(e) && 2 === e.length && Array.isArray(e[0]) && Array.isArray(e[1]), () => "A `Dot` layer should be called on a list of exactly 2 inputs.");
+                    const t = e[0],
+                        n = e[1];
+                    if (t.length > 3 || n.length > 3) throw new ug("Dot layer does not support tensors of 4D or higher rank yet.");
+                    const r = this.interpretAxes(t, n);
+                    if (t[r[0]] !== n[r[1]]) throw new og(`Dimension incompatibility: ${t[r[0]]} !== ${n[r[1]]}`)
+                }
+                mergeFunction(e) {
+                    if (2 !== e.length) throw new og(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);
+                    let t, n = e[0],
+                        r = e[1];
+                    return t = Array.isArray(this.axes) ? this.axes.map((t, n) => Aw(t, e[n].shape.length)) : [Aw(this.axes, n.shape.length), Aw(this.axes, r.shape.length)], this.normalize && (n = zb(n, t[0]), r = zb(r, t[1])),
+                        function (e, t, n) {
+                            if (e.shape.length > 3 || t.shape.length > 3) throw new ug("batchDot is not implemented for tensors of 4D or higher rank yet");
+                            if (va.assert(e.shape.length >= 2, () => "batchDot requires the rank of x to be >= 2, but got " + e.shape.length), va.assert(e.shape.length >= 2, () => "batchDot requires the rank of y to be >= 2, but got " + t.shape.length), "number" == typeof n && (n = [n, n]), "complex64" === e.dtype || "complex64" === t.dtype) throw new ug("batchDot is not implemented for complex64-type Tensors yet.");
+                            const r = e.shape.length,
+                                a = t.shape.length;
+                            null == n && (n = [r - 1, a - 2]);
+                            const s = n;
+                            return ru(() => {
+                                let n, i;
+                                if (r > a) {
+                                    n = r - a;
+                                    const e = [];
+                                    for (let t = 0; t < n; ++t) e.push(1);
+                                    t = qu(t, t.shape.concat(e))
+                                } else if (a > r) {
+                                    n = a - r;
+                                    const t = [];
+                                    for (let e = 0; e < n; ++e) t.push(1);
+                                    e = qu(e, e.shape.concat(t))
+                                } else n = 0;
+                                if (2 === e.shape.length && 2 === t.shape.length) i = s[0] === s[1] ? yc(vu(e, t), s[0]) : yc(vu(qi(e, [1, 0]), t), s[1]);
+                                else {
+                                    const n = s[0] !== e.shape.length - 1,
+                                        r = s[1] === t.shape.length - 1;
+                                    i = Gi(e, t, n, r)
+                                }
+                                if (n > 0) {
+                                    let e;
+                                    e = r > a ? r + a - 3 : r - 1;
+                                    const t = [];
+                                    for (let r = e; r < e + n; ++r) t.push(r);
+                                    i = jh(i, t)
+                                }
+                                return 1 === i.shape.length && (i = Bl(i, 1)), i
+                            })
+                        }(n, r, t)
+                }
+                interpretAxes(e, t) {
+                    let n;
+                    return n = Array.isArray(this.axes) ? this.axes : [Aw(this.axes, e.length), Aw(this.axes, t.length)], n
+                }
+                computeOutputShape(e) {
+                    va.assert(Array.isArray(e) && 2 === e.length && Array.isArray(e[0]) && Array.isArray(e[1]), () => "A `Dot` layer should be called on a list of exactly 2 inputs.");
+                    const t = e[0].slice(),
+                        n = e[1].slice();
+                    if (t.length > 3 || n.length > 3) throw new ug("Dot layer does not support tensors of 4D or higher rank yet.");
+                    const r = this.interpretAxes(t, n);
+                    t.splice(r[0], 1), n.splice(r[1], 1), n.splice(0, 1);
+                    const a = t.concat(n);
+                    return 1 === a.length && a.push(1), a
+                }
+                computeMask(e, t) {
+                    return null
+                }
+                getConfig() {
+                    const e = {
+                            axes: this.axes,
+                            normalize: this.normalize
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Ew.className = "Dot", Fo.registerClass(Ew);
+            var Nw = class extends Ib {
+                constructor(e) {
+                    super(e), this.supportsMasking = !0, this.stddev = e.stddev
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            stddev: this.stddev
+                        };
+                    return Object.assign(t, e), t
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e);
+                        return ky(() => gu(my(n.shape, 0, this.stddev), n), () => n, t.training || !1)
+                    })
+                }
+            };
+            Nw.className = "GaussianNoise", Fo.registerClass(Nw);
+            var Cw = class extends Ib {
+                constructor(e) {
+                    super(e), this.supportsMasking = !0, this.rate = e.rate
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            rate: this.rate
+                        };
+                    return Object.assign(t, e), t
+                }
+                call(e, t) {
+                    return ru(() => {
+                        this.invokeCallHook(e, t);
+                        const n = db(e);
+                        if (this.rate > 0 && this.rate < 1) {
+                            return ky(() => {
+                                const e = Math.sqrt(this.rate / (1 - this.rate));
+                                return vu(n, my(n.shape, 1, e))
+                            }, () => n, t.training || !1)
+                        }
+                        return n
+                    })
+                }
+            };
+            Cw.className = "GaussianDropout", Fo.registerClass(Cw);
+            var Tw = class extends Ib {
+                constructor(e) {
+                    super(e), this.supportsMasking = !0, this.rate = e.rate, this.noiseShape = e.noiseShape
+                }
+                _getNoiseShape(e) {
+                    return this.noiseShape || db(e).shape
+                }
+                computeOutputShape(e) {
+                    return e
+                }
+                getConfig() {
+                    const e = super.getConfig(),
+                        t = {
+                            rate: this.rate
+                        };
+                    return Object.assign(t, e), t
+                }
+                call(e, t) {
+                    return ru(() => {
+                        if (this.rate < 1 && this.rate > 0) {
+                            const n = this._getNoiseShape(e);
+                            return ky(() => {
+                                const t = db(e),
+                                    r = -1.7580993408473766;
+                                let a = ql(ph(n), this.rate);
+                                a = oy(a, "float32");
+                                const s = ((1 - this.rate) * (1 + this.rate * r ** 2)) ** -.5,
+                                    i = -s * r * this.rate,
+                                    o = gu(vu(t, a), vu(gu(a, -1), r));
+                                return gu(vu(o, s), i)
+                            }, () => db(e), t.training || !1)
+                        }
+                        return e
+                    })
+                }
+            };
+
+            function Rw(e, t, n, r, a, s = .001) {
+                let i;
+                if (2 === e.rank) i = rl(e, t, n, r, a, s);
+                else if (3 === e.rank) i = al(e, t, n, r, a, s);
+                else {
+                    if (4 !== e.rank) throw new ug(`batchNormalization is not implemented for array of rank ${e.rank} yet`);
+                    i = sl(e, t, n, r, a, s)
+                }
+                return i
+            }
+
+            function _w(e, t, n, r, a = .001) {
+                return va.arraysEqual(r.slice().sort(), iy(0, e.rank - 1)) ? function (e, t, n, r, a = .001) {
+                    return ru(() => {
+                        const s = Gc(e, r),
+                            i = s.mean,
+                            o = s.variance;
+                        return [Rw(e, i, o, n, t, a), i, o]
+                    })
+                }(e, t, n, r, a) : function (e, t, n, r, a = .001) {
+                    return ru(() => {
+                        const s = Gc(e, r),
+                            i = s.mean,
+                            o = s.variance,
+                            u = [];
+                        for (const t of iy(0, e.rank)) - 1 !== r.indexOf(t) ? u.push(1) : u.push(e.shape[t]);
+                        const l = qu(i, u),
+                            c = qu(o, u),
+                            h = null == t ? null : qu(t, u),
+                            p = null == n ? null : qu(n, u);
+                        return [Rw(e, l, c, p, h, a), i, o]
+                    })
+                }(e, t, n, r, a)
+            }
+            Tw.className = "AlphaDropout", Fo.registerClass(Tw);
+            var Fw = class extends Ib {
+                constructor(e) {
+                    null == e && (e = {}), super(e), this.supportsMasking = !0, this.axis = null == e.axis ? -1 : e.axis, this.momentum = null == e.momentum ? .99 : e.momentum, this.epsilon = null == e.epsilon ? .001 : e.epsilon, this.center = null == e.center || e.center, this.scale = null == e.scale || e.scale, this.betaInitializer = jy(e.betaInitializer || "zeros"), this.gammaInitializer = jy(e.gammaInitializer || "ones"), this.movingMeanInitializer = jy(e.movingMeanInitializer || "zeros"), this.movingVarianceInitializer = jy(e.movingVarianceInitializer || "ones"), this.betaConstraint = Lg(e.betaConstraint), this.gammaConstraint = Lg(e.gammaConstraint), this.betaRegularizer = kx(e.betaRegularizer), this.gammaRegularizer = kx(e.gammaRegularizer)
+                }
+                build(e) {
+                    e = fb(e);
+                    const t = this.axis >= 0 ? this.axis : this.axis + e.length,
+                        n = e[t];
+                    if (null == n) throw new og(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);
+                    this.inputSpec = [new vb({
+                        ndim: e.length,
+                        axes: {
+                            [t]: n
+                        }
+                    })];
+                    const r = [n];
+                    this.scale && (this.gamma = this.addWeight("gamma", r, null, this.gammaInitializer, this.gammaRegularizer, !0, this.gammaConstraint)), this.center && (this.beta = this.addWeight("beta", r, null, this.betaInitializer, this.betaRegularizer, !0, this.betaConstraint)), this.movingMean = this.addWeight("moving_mean", r, null, this.movingMeanInitializer, null, !1), this.movingVariance = this.addWeight("moving_variance", r, null, this.movingVarianceInitializer, null, !1), this.built = !0
+                }
+                call(e, t) {
+                    return ru(() => {
+                        const n = null != t.training && t.training,
+                            r = db(e),
+                            a = r.shape,
+                            s = a.length,
+                            i = iy(0, s),
+                            o = this.axis >= 0 ? this.axis : this.axis + s;
+                        i.splice(o, 1);
+                        const u = cg(1, s);
+                        u[o] = a[o];
+                        const l = i.slice();
+                        l.sort();
+                        const c = !va.arraysEqual(l, iy(0, s).slice(0, s - 1));
+                        if (!n) return (() => {
+                            if (c) {
+                                const e = qu(this.movingMean.read(), u),
+                                    t = qu(this.movingVariance.read(), u),
+                                    n = this.center ? qu(this.beta.read(), u) : null,
+                                    a = this.scale ? qu(this.gamma.read(), u) : null;
+                                return Rw(r, e, t, n, a, this.epsilon)
+                            }
+                            return Rw(r, this.movingMean.read(), this.movingVariance.read(), null == this.beta ? null : this.beta.read(), null == this.gamma ? null : this.gamma.read(), this.epsilon)
+                        })();
+                        const [h, p, d] = _w(r, this.gamma.read(), this.beta.read(), i, this.epsilon), f = (e, t, n) => {
+                            ru(() => {
+                                const r = 1 - n,
+                                    a = e.read(),
+                                    s = vu(gc(a, t), r);
+                                e.write(gc(a, s))
+                            })
+                        };
+                        return (() => {
+                            f(this.movingMean, p, this.momentum), f(this.movingVariance, d, this.momentum)
+                        })(), h
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            axis: this.axis,
+                            momentum: this.momentum,
+                            epsilon: this.epsilon,
+                            center: this.center,
+                            scale: this.scale,
+                            betaInitializer: Vy(this.betaInitializer),
+                            gammaInitializer: Vy(this.gammaInitializer),
+                            movingMeanInitializer: Vy(this.movingMeanInitializer),
+                            movingVarianceInitializer: Vy(this.movingVarianceInitializer),
+                            betaRegularizer: xx(this.betaRegularizer),
+                            gammaRegularizer: xx(this.gammaRegularizer),
+                            betaConstraint: Dg(this.betaConstraint),
+                            gammaConstraint: Dg(this.gammaConstraint)
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Fw.className = "BatchNormalization", Fo.registerClass(Fw);
+            var Mw = class extends Ib {
+                constructor(e) {
+                    if (null == e && (e = {}), super(e), this.axis = null == e.axis ? -1 : e.axis, "number" == typeof this.axis) {
+                        if (!Number.isInteger(this.axis)) throw new Error("Expected axis to be an integer, but received " + this.axis)
+                    } else {
+                        if (!Array.isArray(this.axis)) throw new Error("Expected axis to be an integer or an array of integers, but received " + JSON.stringify(this.axis));
+                        for (const e of this.axis)
+                            if (!Number.isInteger(e)) throw new Error("Expected axis to be an array of integers, but received " + JSON.stringify(this.axis))
+                    }
+                    this.epsilon = null == e.epsilon ? .001 : e.epsilon, this.center = null == e.center || e.center, this.scale = null == e.scale || e.scale, this.betaInitializer = jy(e.betaInitializer || "zeros"), this.gammaInitializer = jy(e.gammaInitializer || "ones"), this.betaRegularizer = kx(e.betaRegularizer), this.gammaRegularizer = kx(e.gammaRegularizer), this.supportsMasking = !0
+                }
+                build(e) {
+                    const t = (e = fb(e)).length;
+                    "number" == typeof this.axis && (this.axis = [this.axis]);
+                    for (let e = 0; e < this.axis.length; ++e) this.axis[e] < 0 && (this.axis[e] += t);
+                    for (const e of this.axis)
+                        if (e < 0 || e >= t) throw new Error("Invalid axis: " + e);
+                    if (this.axis.length !== wg(this.axis).length) throw new Error("Found duplicate axes in: " + this.axis);
+                    const n = this.axis.map(t => e[t]);
+                    this.scale ? this.gamma = this.addWeight("gamma", n, "float32", this.gammaInitializer, this.gammaRegularizer, !0) : this.gamma = null, this.center ? this.beta = this.addWeight("beta", n, "float32", this.betaInitializer, this.betaRegularizer, !0) : this.beta = null, this.built = !0
+                }
+                call(e, t) {
+                    const n = db(e),
+                        r = n.shape,
+                        a = r.length;
+                    return ru(() => {
+                        let {
+                            mean: e,
+                            variance: t
+                        } = Gc(n, this.axis, !0);
+                        const s = cg(1, a);
+                        for (const e of this.axis) s[e] = r[e];
+                        const i = e => null != e && e.shape.length !== a ? qu(e, s) : e;
+                        let o = i(this.gamma.read()),
+                            u = i(this.beta.read());
+                        const l = [],
+                            c = [];
+                        for (let e = 0; e < a; ++e) - 1 !== this.axis.indexOf(e) ? (l.push(r[e]), c.push(1)) : (l.push(1), c.push(r[e]));
+                        return e = Ul(e, l), t = Ul(t, l), o = Ul(o, c), u = Ul(u, c), Rw(n, e, t, u, o, this.epsilon)
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            axis: this.axis,
+                            epsilon: this.epsilon,
+                            center: this.center,
+                            scale: this.scale,
+                            betaInitializer: Vy(this.betaInitializer),
+                            gammaInitializer: Vy(this.gammaInitializer),
+                            betaRegularizer: xx(this.betaRegularizer),
+                            gammaRegularizer: xx(this.gammaRegularizer)
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+            Mw.className = "LayerNormalization", Fo.registerClass(Mw);
+            var Dw = class extends Ib {
+                constructor(e) {
+                    if (null == e && (e = {}), super(e), this.dataFormat = null == e.dataFormat ? "channelsLast" : e.dataFormat, null == e.padding) this.padding = [
+                        [1, 1],
+                        [1, 1]
+                    ];
+                    else if ("number" == typeof e.padding) this.padding = [
+                        [e.padding, e.padding],
+                        [e.padding, e.padding]
+                    ];
+                    else {
+                        if (e.padding = e.padding, 2 !== e.padding.length) throw new og(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);
+                        let t, n;
+                        if ("number" == typeof e.padding[0]) t = [e.padding[0], e.padding[0]], n = [e.padding[1], e.padding[1]];
+                        else {
+                            if (e.padding = e.padding, 2 !== e.padding[0].length) throw new og(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);
+                            if (t = e.padding[0], 2 !== e.padding[1].length) throw new og(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);
+                            n = e.padding[1]
+                        }
+                        this.padding = [t, n]
+                    }
+                    this.inputSpec = [new vb({
+                        ndim: 4
+                    })]
+                }
+                computeOutputShape(e) {
+                    let t, n;
+                    return e = fb(e), "channelsFirst" === this.dataFormat ? (t = null != e[2] && e[2] >= 0 ? e[2] + this.padding[0][0] + this.padding[0][1] : null, n = null != e[3] && e[3] >= 0 ? e[3] + this.padding[1][0] + this.padding[1][1] : null, [e[0], e[1], t, n]) : (t = null != e[1] && e[1] >= 0 ? e[1] + this.padding[0][0] + this.padding[0][1] : null, n = null != e[2] && e[2] >= 0 ? e[2] + this.padding[1][0] + this.padding[1][1] : null, [e[0], t, n, e[3]])
+                }
+                call(e, t) {
+                    return ru(() => {
+                        return t = db(e), n = this.padding, r = this.dataFormat, ru(() => {
+                            if (4 !== t.rank) throw new og(`temporalPadding expects input tensor to be 4-D, but received a ${t.rank}-D tensor.`);
+                            if (null == n && (n = [
+                                    [1, 1],
+                                    [1, 1]
+                                ]), 2 !== n.length || 2 !== n[0].length || 2 !== n[1].length) throw new og("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");
+                            if (null == r && (r = "channelsLast"), "channelsLast" !== r && "channelsFirst" !== r) throw new og(`Unknown data format: ${r}. Supported data formats are 'channelsLast' and 'channelsFirst.`);
+                            let e;
+                            return e = "channelsFirst" === r ? [
+                                [0, 0],
+                                [0, 0], n[0], n[1]
+                            ] : [
+                                [0, 0], n[0], n[1],
+                                [0, 0]
+                            ], Qc(t, e)
+                        });
+                        var t, n, r
+                    })
+                }
+                getConfig() {
+                    const e = {
+                            padding: this.padding,
+                            dataFormat: this.dataFormat
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+            };
+
+            function Ow(e, t, n, r, a, s) {
+                return ru(() => {
+                    let i;
+                    Kg(a), Xg(s), Yg(r), null == n && (n = [1, 1]), null == r && (r = "valid"), null == a && (a = "channelsLast"), null == s && (s = "max"), e = Fx(e, a);
+                    const o = "same" === r ? "same" : "valid";
+                    return i = "max" === s ? Fc(e, t, n, o) : Ku(e, t, n, o), "channelsFirst" === a && (i = qi(i, [0, 3, 1, 2])), i
+                })
+            }
+
+            function Lw(e, t, n, r, a, s) {
+                return ru(() => {
+                    let i;
+                    Kg(a), Xg(s), Yg(r), null == n && (n = [1, 1, 1]), null == r && (r = "valid"), null == a && (a = "channelsLast"), null == s && (s = "max"), e = Mx(e, a);
+                    const o = "same" === r ? "same" : "valid";
+                    return i = "max" === s ? Mc(e, t, n, o) : Yu(e, t, n, o), "channelsFirst" === a && (i = qi(i, [0, 4, 1, 2, 3])), i
+                })
+            }
+            Dw.className = "ZeroPadding2D", Fo.registerClass(Dw);
+            var Pw = class extends Ib {
+                    constructor(e) {
+                        if (null == e.poolSize && (e.poolSize = 2), super(e), "number" == typeof e.poolSize) this.poolSize = [e.poolSize];
+                        else {
+                            if (!Array.isArray(e.poolSize) || 1 !== e.poolSize.length || "number" != typeof e.poolSize[0]) throw new og("poolSize for 1D convolutional layer must be a number or an Array of a single number, but received " + JSON.stringify(e.poolSize));
+                            this.poolSize = e.poolSize
+                        }
+                        if (Ag(this.poolSize, "poolSize"), null == e.strides) this.strides = this.poolSize;
+                        else if ("number" == typeof e.strides) this.strides = [e.strides];
+                        else {
+                            if (!Array.isArray(e.strides) || 1 !== e.strides.length || "number" != typeof e.strides[0]) throw new og("strides for 1D convolutional layer must be a number or an Array of a single number, but received " + JSON.stringify(e.strides));
+                            this.strides = e.strides
+                        }
+                        Ag(this.strides, "strides"), this.padding = null == e.padding ? "valid" : e.padding, Yg(this.padding), this.inputSpec = [new vb({
+                            ndim: 3
+                        })]
+                    }
+                    computeOutputShape(e) {
+                        const t = Rx((e = fb(e))[1], this.poolSize[0], this.padding, this.strides[0]);
+                        return [e[0], t, e[2]]
+                    }
+                    call(e, t) {
+                        return ru(() => {
+                            this.invokeCallHook(e, t), e = uy(db(e), 2);
+                            const n = this.poolingFunction(db(e), [this.poolSize[0], 1], [this.strides[0], 1], this.padding, "channelsLast");
+                            return jh(n, [2])
+                        })
+                    }
+                    getConfig() {
+                        const e = {
+                                poolSize: this.poolSize,
+                                padding: this.padding,
+                                strides: this.strides
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                },
+                $w = class extends Pw {
+                    constructor(e) {
+                        super(e)
+                    }
+                    poolingFunction(e, t, n, r, a) {
+                        return Kg(a), Yg(r), Ow(e, t, n, r, a, "max")
+                    }
+                };
+            $w.className = "MaxPooling1D", Fo.registerClass($w);
+            var Bw = class extends Pw {
+                constructor(e) {
+                    super(e)
+                }
+                poolingFunction(e, t, n, r, a) {
+                    return Kg(a), Yg(r), Ow(e, t, n, r, a, "avg")
+                }
+            };
+            Bw.className = "AveragePooling1D", Fo.registerClass(Bw);
+            var zw = class extends Ib {
+                    constructor(e) {
+                        if (null == e.poolSize && (e.poolSize = [2, 2]), super(e), this.poolSize = Array.isArray(e.poolSize) ? e.poolSize : [e.poolSize, e.poolSize], null == e.strides) this.strides = this.poolSize;
+                        else if (Array.isArray(e.strides)) {
+                            if (2 !== e.strides.length) throw new og(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);
+                            this.strides = e.strides
+                        } else this.strides = [e.strides, e.strides];
+                        Ag(this.poolSize, "poolSize"), Ag(this.strides, "strides"), this.padding = null == e.padding ? "valid" : e.padding, this.dataFormat = null == e.dataFormat ? "channelsLast" : e.dataFormat, Kg(this.dataFormat), Yg(this.padding), this.inputSpec = [new vb({
+                            ndim: 4
+                        })]
+                    }
+                    computeOutputShape(e) {
+                        e = fb(e);
+                        let t = "channelsFirst" === this.dataFormat ? e[2] : e[1],
+                            n = "channelsFirst" === this.dataFormat ? e[3] : e[2];
+                        return t = Rx(t, this.poolSize[0], this.padding, this.strides[0]), n = Rx(n, this.poolSize[1], this.padding, this.strides[1]), "channelsFirst" === this.dataFormat ? [e[0], e[1], t, n] : [e[0], t, n, e[3]]
+                    }
+                    call(e, t) {
+                        return ru(() => (this.invokeCallHook(e, t), this.poolingFunction(db(e), this.poolSize, this.strides, this.padding, this.dataFormat)))
+                    }
+                    getConfig() {
+                        const e = {
+                                poolSize: this.poolSize,
+                                padding: this.padding,
+                                strides: this.strides,
+                                dataFormat: this.dataFormat
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                },
+                Uw = class extends zw {
+                    constructor(e) {
+                        super(e)
+                    }
+                    poolingFunction(e, t, n, r, a) {
+                        return Kg(a), Yg(r), Ow(e, t, n, r, a, "max")
+                    }
+                };
+            Uw.className = "MaxPooling2D", Fo.registerClass(Uw);
+            var Ww = class extends zw {
+                constructor(e) {
+                    super(e)
+                }
+                poolingFunction(e, t, n, r, a) {
+                    return Kg(a), Yg(r), Ow(e, t, n, r, a, "avg")
+                }
+            };
+            Ww.className = "AveragePooling2D", Fo.registerClass(Ww);
+            var Vw = class extends Ib {
+                    constructor(e) {
+                        if (null == e.poolSize && (e.poolSize = [2, 2, 2]), super(e), this.poolSize = Array.isArray(e.poolSize) ? e.poolSize : [e.poolSize, e.poolSize, e.poolSize], null == e.strides) this.strides = this.poolSize;
+                        else if (Array.isArray(e.strides)) {
+                            if (3 !== e.strides.length) throw new og(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);
+                            this.strides = e.strides
+                        } else this.strides = [e.strides, e.strides, e.strides];
+                        Ag(this.poolSize, "poolSize"), Ag(this.strides, "strides"), this.padding = null == e.padding ? "valid" : e.padding, this.dataFormat = null == e.dataFormat ? "channelsLast" : e.dataFormat, Kg(this.dataFormat), Yg(this.padding), this.inputSpec = [new vb({
+                            ndim: 5
+                        })]
+                    }
+                    computeOutputShape(e) {
+                        e = fb(e);
+                        let t = "channelsFirst" === this.dataFormat ? e[2] : e[1],
+                            n = "channelsFirst" === this.dataFormat ? e[3] : e[2],
+                            r = "channelsFirst" === this.dataFormat ? e[4] : e[3];
+                        return t = Rx(t, this.poolSize[0], this.padding, this.strides[0]), n = Rx(n, this.poolSize[1], this.padding, this.strides[1]), r = Rx(r, this.poolSize[2], this.padding, this.strides[2]), "channelsFirst" === this.dataFormat ? [e[0], e[1], t, n, r] : [e[0], t, n, r, e[4]]
+                    }
+                    call(e, t) {
+                        return ru(() => (this.invokeCallHook(e, t), this.poolingFunction(db(e), this.poolSize, this.strides, this.padding, this.dataFormat)))
+                    }
+                    getConfig() {
+                        const e = {
+                                poolSize: this.poolSize,
+                                padding: this.padding,
+                                strides: this.strides,
+                                dataFormat: this.dataFormat
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                },
+                jw = class extends Vw {
+                    constructor(e) {
+                        super(e)
+                    }
+                    poolingFunction(e, t, n, r, a) {
+                        return Kg(a), Yg(r), Lw(e, t, n, r, a, "max")
+                    }
+                };
+            jw.className = "MaxPooling3D", Fo.registerClass(jw);
+            var Gw = class extends Vw {
+                constructor(e) {
+                    super(e)
+                }
+                poolingFunction(e, t, n, r, a) {
+                    return Kg(a), Yg(r), Lw(e, t, n, r, a, "avg")
+                }
+            };
+            Gw.className = "AveragePooling3D", Fo.registerClass(Gw);
+            var Hw = class extends Ib {
+                    constructor(e) {
+                        super(e), this.inputSpec = [new vb({
+                            ndim: 3
+                        })]
+                    }
+                    computeOutputShape(e) {
+                        return [e[0], e[2]]
+                    }
+                    call(e, t) {
+                        throw new ug
+                    }
+                },
+                qw = class extends Hw {
+                    constructor(e) {
+                        super(e || {})
+                    }
+                    call(e, t) {
+                        return ru(() => {
+                            const t = db(e);
+                            return Lc(t, 1)
+                        })
+                    }
+                };
+            qw.className = "GlobalAveragePooling1D", Fo.registerClass(qw);
+            var Kw = class extends Hw {
+                constructor(e) {
+                    super(e || {})
+                }
+                call(e, t) {
+                    return ru(() => {
+                        const t = db(e);
+                        return mc(t, 1)
+                    })
+                }
+            };
+            Kw.className = "GlobalMaxPooling1D", Fo.registerClass(Kw);
+            var Yw = class extends Ib {
+                    constructor(e) {
+                        super(e), this.dataFormat = null == e.dataFormat ? "channelsLast" : e.dataFormat, Kg(this.dataFormat), this.inputSpec = [new vb({
+                            ndim: 4
+                        })]
+                    }
+                    computeOutputShape(e) {
+                        return e = e, "channelsLast" === this.dataFormat ? [e[0], e[3]] : [e[0], e[1]]
+                    }
+                    call(e, t) {
+                        throw new ug
+                    }
+                    getConfig() {
+                        const e = {
+                                dataFormat: this.dataFormat
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                },
+                Xw = class extends Yw {
+                    call(e, t) {
+                        return ru(() => {
+                            const t = db(e);
+                            return "channelsLast" === this.dataFormat ? Lc(t, [1, 2]) : Lc(t, [2, 3])
+                        })
+                    }
+                };
+            Xw.className = "GlobalAveragePooling2D", Fo.registerClass(Xw);
+            var Qw = class extends Yw {
+                call(e, t) {
+                    return ru(() => {
+                        const t = db(e);
+                        return "channelsLast" === this.dataFormat ? mc(t, [1, 2]) : mc(t, [2, 3])
+                    })
+                }
+            };
+            Qw.className = "GlobalMaxPooling2D", Fo.registerClass(Qw);
+            var Zw = class extends Ib {
+                    constructor(e) {
+                        super(e), this.layer = e.layer
+                    }
+                    build(e) {
+                        this.built = !0
+                    }
+                    get trainable() {
+                        return null != this.layer && this.layer.trainable
+                    }
+                    set trainable(e) {
+                        null != this.layer && (this.layer.trainable = e)
+                    }
+                    get trainableWeights() {
+                        return this.layer.trainableWeights
+                    }
+                    get nonTrainableWeights() {
+                        return this.layer.nonTrainableWeights
+                    }
+                    get updates() {
+                        return this.layer._updates
+                    }
+                    get losses() {
+                        return this.layer.losses
+                    }
+                    getWeights() {
+                        return this.layer.getWeights()
+                    }
+                    setWeights(e) {
+                        this.layer.setWeights(e)
+                    }
+                    getConfig() {
+                        const e = {
+                                layer: {
+                                    className: this.layer.getClassName(),
+                                    config: this.layer.getConfig()
+                                }
+                            },
+                            t = super.getConfig();
+                        return Object.assign(e, t), e
+                    }
+                    setFastWeightInitDuringBuild(e) {
+                        super.setFastWeightInitDuringBuild(e), null != this.layer && this.layer.setFastWeightInitDuringBuild(e)
+                    }
+                    static fromConfig(e, t, n = {}) {
+                        const r = Bb(t.layer, n);
+                        delete t.layer;
+                        const a = {
+                            layer: r
+                        };
+                        return Object.assign(a, t), new e(a)
+                    }
+                },
+                Jw = class extends Zw {
+                    constructor(e) {
+                        super(e), this.supportsMasking = !0
+                    }
+                    build(e) {
+                        if ((e = fb(e)).length < 3) throw new og("TimeDistributed layer expects an input shape >= 3D, but received input shape " + JSON.stringify(e));
+                        this.inputSpec = [{
+                            shape: e
+                        }];
+                        const t = [e[0]].concat(e.slice(2));
+                        this.layer.built || (this.layer.build(t), this.layer.built = !0), super.build(e)
+                    }
+                    computeOutputShape(e) {
+                        const t = [(e = fb(e))[0]].concat(e.slice(2)),
+                            n = this.layer.computeOutputShape(t),
+                            r = e[1];
+                        return [n[0], r].concat(n.slice(1))
+                    }
+                    call(e, t) {
+                        return ru(() => Kx((e, n) => [db(this.layer.call(e, t)), []], e = db(e), [], !1, null, null, !1, !0)[1])
+                    }
+                };
+            Jw.className = "TimeDistributed", Fo.registerClass(Jw);
+            var ek = class extends Zw {
+                constructor(e) {
+                    super(e);
+                    const t = e.layer.getConfig(),
+                        n = {};
+                    n.className = e.layer.getClassName(), n.config = t, this.forwardLayer = Bb(n), t.goBackwards = !0 !== t.goBackwards;
+                    const r = {};
+                    var a;
+                    if (r.className = e.layer.getClassName(), r.config = t, this.backwardLayer = Bb(r), this.forwardLayer.name = "forward_" + this.forwardLayer.name, this.backwardLayer.name = "backward_" + this.backwardLayer.name, this.mergeMode = void 0 === e.mergeMode ? "concat" : e.mergeMode, a = this.mergeMode, Sg(Hg, "BidirectionalMergeMode", a), e.weights) throw new ug("weights support is not implemented for Bidirectional layer yet.");
+                    this._stateful = e.layer.stateful, this.returnSequences = e.layer.returnSequences, this.returnState = e.layer.returnState, this.supportsMasking = !0, this._trainable = !0, this.inputSpec = e.layer.inputSpec, this.numConstants = null
+                }
+                get trainable() {
+                    return this._trainable
+                }
+                set trainable(e) {
+                    this._trainable = e, null != this.forwardLayer && (this.forwardLayer.trainable = e), null != this.backwardLayer && (this.backwardLayer.trainable = e)
+                }
+                getWeights() {
+                    return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())
+                }
+                setWeights(e) {
+                    const t = e.length,
+                        n = Math.floor(t / 2);
+                    this.forwardLayer.setWeights(e.slice(0, n)), this.backwardLayer.setWeights(e.slice(n))
+                }
+                computeOutputShape(e) {
+                    let t, n, r, a = this.forwardLayer.computeOutputShape(e);
+                    return Array.isArray(a) && Array.isArray(a[0]) || (a = [a]), a = a, this.returnState ? (r = a.slice(1), t = a[0]) : t = a[0], t = t, "concat" === this.mergeMode ? (t[t.length - 1] *= 2, n = [t]) : n = null == this.mergeMode ? [t, t.slice()] : [t], this.returnState ? null == this.mergeMode ? n.concat(r).concat(r.slice()) : [t].concat(r).concat(r.slice()) : dg(n)
+                }
+                apply(e, t) {
+                    let n = null == t ? null : t.initialState,
+                        r = null == t ? null : t.constants;
+                    null == t && (t = {});
+                    const a = qx(e, n, r, this.numConstants);
+                    if (e = a.inputs, n = a.initialState, r = a.constants, Array.isArray(e) && (n = e.slice(1), e = e[0]), (null == n || 0 === n.length) && null == r) return super.apply(e, t);
+                    const s = [],
+                        i = [];
+                    if (null != n) {
+                        const e = n.length;
+                        if (e % 2 > 0) throw new og("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");
+                        t.initialState = n, s.push(...n);
+                        const r = n.map(e => new vb({
+                            shape: e.shape
+                        }));
+                        this.forwardLayer.stateSpec = r.slice(0, e / 2), this.backwardLayer.stateSpec = r.slice(e / 2), i.push(...r)
+                    }
+                    if (null != r) throw new ug("Support for constants in Bidirectional layers is not implemented yet.");
+                    const o = s[0] instanceof xb;
+                    for (const e of s)
+                        if (e instanceof xb !== o) throw new og("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");
+                    if (o) {
+                        const n = [e].concat(s),
+                            r = this.inputSpec.concat(i),
+                            a = this.inputSpec;
+                        this.inputSpec = r;
+                        const o = super.apply(n, t);
+                        return this.inputSpec = a, o
+                    }
+                    return super.apply(e, t)
+                }
+                call(e, t) {
+                    return ru(() => {
+                        const n = t.initialState;
+                        let r, a, s, i;
+                        if (null == n) r = this.forwardLayer.call(e, t), a = this.backwardLayer.call(e, t);
+                        else {
+                            const s = n.slice(0, n.length / 2),
+                                i = n.slice(n.length / 2);
+                            r = this.forwardLayer.call(e, Object.assign(t, {
+                                initialState: s
+                            })), a = this.backwardLayer.call(e, Object.assign(t, {
+                                initialState: i
+                            }))
+                        }
+                        return this.returnState && (Array.isArray(r) && (s = r.slice(1).concat(a.slice(1))), r = r[0], a = a[0]), this.returnSequences && (a = bh(a, 1)), "concat" === this.mergeMode ? i = py([r, a]) : "sum" === this.mergeMode ? i = gu(r, a) : "ave" === this.mergeMode ? i = vu(.5, gu(r, a)) : "mul" === this.mergeMode ? i = vu(r, a) : null == this.mergeMode && (i = [r, a]), this.returnState ? null == this.mergeMode ? i.concat(s) : [i].concat(s) : i
+                    })
+                }
+                resetStates(e) {
+                    this.forwardLayer.resetStates(), this.backwardLayer.resetStates()
+                }
+                build(e) {
+                    Zg(this.forwardLayer.name, () => {
+                        this.forwardLayer.build(e)
+                    }), Zg(this.backwardLayer.name, () => {
+                        this.backwardLayer.build(e)
+                    }), this.built = !0
+                }
+                computeMask(e, t) {
+                    let n;
+                    if (Array.isArray(t) && (t = t[0]), n = this.returnSequences ? null == this.mergeMode ? [t, t] : t : null == this.mergeMode ? [null, null] : null, this.returnState) {
+                        const e = this.forwardLayer.states.map(e => null);
+                        return Array.isArray(n) ? n.concat(e).concat(e) : [n].concat(e).concat(e)
+                    }
+                    return n
+                }
+                get trainableWeights() {
+                    return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)
+                }
+                get nonTrainableWeights() {
+                    return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)
+                }
+                setFastWeightInitDuringBuild(e) {
+                    super.setFastWeightInitDuringBuild(e), null != this.forwardLayer && this.forwardLayer.setFastWeightInitDuringBuild(e), null != this.backwardLayer && this.backwardLayer.setFastWeightInitDuringBuild(e)
+                }
+                getConfig() {
+                    const e = {
+                            mergeMode: this.mergeMode
+                        },
+                        t = super.getConfig();
+                    return Object.assign(e, t), e
+                }
+                static fromConfig(e, t) {
+                    const n = Bb(t.layer);
+                    if (delete t.layer, null != t.numConstants) throw new ug("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");
+                    const r = t;
+                    return r.layer = n, new e(r)
+                }
+            };
+
+            function tk(e) {
+                return new Nb(e)
+            }
+
+            function nk(e) {
+                return new Ex(e)
+            }
+
+            function rk(e) {
+                return new Sx(e)
+            }
+
+            function ak(e) {
+                return new Ix(e)
+            }
+
+            function sk(e) {
+                return new Ax(e)
+            }
+
+            function ik(e) {
+                return new Cx(e)
+            }
+
+            function ok(e) {
+                return new Nx(e)
+            }
+
+            function uk(e) {
+                return new Vx(e)
+            }
+
+            function lk(e) {
+                return new Px(e)
+            }
+
+            function ck(e) {
+                return new Bx(e)
+            }
+
+            function hk(e) {
+                return new $x(e)
+            }
+
+            function pk(e) {
+                return new zx(e)
+            }
+
+            function dk(e) {
+                return new Wx(e)
+            }
+
+            function fk(e) {
+                return new jx(e)
+            }
+
+            function mk(e) {
+                return new Gx(e)
+            }
+
+            function gk(e) {
+                return new Hx(e)
+            }
+
+            function yk(e) {
+                return new pw(e)
+            }
+
+            function bk(e) {
+                return new cw(e)
+            }
+
+            function vk(e) {
+                return new uw(e)
+            }
+
+            function xk(e) {
+                return new lw(e)
+            }
+
+            function wk(e) {
+                return new hw(e)
+            }
+
+            function kk(e) {
+                return new dw(e)
+            }
+
+            function Sk(e) {
+                return new fw(e)
+            }
+
+            function Ik(e) {
+                return new mw(e)
+            }
+
+            function Ak(e) {
+                return new yw(e)
+            }
+
+            function Ek(e) {
+                return new vw(e)
+            }
+
+            function Nk(e) {
+                return new ww(e)
+            }
+
+            function Ck(e) {
+                return new Iw(e)
+            }
+
+            function Tk(e) {
+                return new kw(e)
+            }
+
+            function Rk(e) {
+                return new Sw(e)
+            }
+
+            function _k(e) {
+                return new xw(e)
+            }
+
+            function Fk(e) {
+                return new Ew(e)
+            }
+
+            function Mk(e) {
+                return new Fw(e)
+            }
+
+            function Dk(e) {
+                return new Mw(e)
+            }
+
+            function Ok(e) {
+                return new Dw(e)
+            }
+
+            function Lk(e) {
+                return new Bw(e)
+            }
+
+            function Pk(e) {
+                return Lk(e)
+            }
+
+            function $k(e) {
+                return Lk(e)
+            }
+
+            function Bk(e) {
+                return new Ww(e)
+            }
+
+            function zk(e) {
+                return Bk(e)
+            }
+
+            function Uk(e) {
+                return Bk(e)
+            }
+
+            function Wk(e) {
+                return new Gw(e)
+            }
+
+            function Vk(e) {
+                return Wk(e)
+            }
+
+            function jk(e) {
+                return Wk(e)
+            }
+
+            function Gk(e) {
+                return new qw(e)
+            }
+
+            function Hk(e) {
+                return new Xw(e)
+            }
+
+            function qk(e) {
+                return new Kw(e)
+            }
+
+            function Kk(e) {
+                return new Qw(e)
+            }
+
+            function Yk(e) {
+                return new $w(e)
+            }
+
+            function Xk(e) {
+                return new Uw(e)
+            }
+
+            function Qk(e) {
+                return new jw(e)
+            }
+
+            function Zk(e) {
+                return new ew(e)
+            }
+
+            function Jk(e) {
+                return new Jx(e)
+            }
+
+            function eS(e) {
+                return new nw(e)
+            }
+
+            function tS(e) {
+                return new tw(e)
+            }
+
+            function nS(e) {
+                return new Zx(e)
+            }
+
+            function rS(e) {
+                return new Qx(e)
+            }
+
+            function aS(e) {
+                return new ow(e)
+            }
+
+            function sS(e) {
+                return new iw(e)
+            }
+
+            function iS(e) {
+                return new Yx(e)
+            }
+
+            function oS(e) {
+                return new rw(e)
+            }
+
+            function uS(e) {
+                return new ek(e)
+            }
+
+            function lS(e) {
+                return new Jw(e)
+            }
+            ek.className = "Bidirectional", Fo.registerClass(ek);
+            var cS = qk,
+                hS = Kk,
+                pS = Yk,
+                dS = Xk;
+
+            function fS(e) {
+                return new Nw(e)
+            }
+
+            function mS(e) {
+                return new Cw(e)
+            }
+
+            function gS(e) {
+                return new Tw(e)
+            }
+
+            function yS(e) {
+                return new gw(e)
+            }
+            var bS = {};
+
+            function vS(e, t) {
+                return Xb(e, t)
+            }
+
+            function xS(e, t) {
+                return tv(e, t)
+            }
+
+            function wS(e, t) {
+                return nv(e, t)
+            }
+
+            function kS(e, t) {
+                return Qb(e, t)
+            }
+
+            function SS(e, t) {
+                return rv(e, t)
+            }
+
+            function IS(e, t) {
+                return Jb(e, t)
+            }
+
+            function AS(e, t) {
+                return ev(e, t)
+            }
+
+            function ES(e, t) {
+                return qb(e, t)
+            }
+
+            function NS(e, t) {
+                return Wb(e, t)
+            }
+
+            function CS(e, t) {
+                return Vb(e, t)
+            }
+
+            function TS(e, t) {
+                return Vb(e, t)
+            }
+
+            function RS(e, t) {
+                return Vb(e, t)
+            }
+
+            function _S(e, t) {
+                return Ub(e, t)
+            }
+
+            function FS(e, t) {
+                return Ub(e, t)
+            }
+
+            function MS(e, t) {
+                return Ub(e, t)
+            }
+            w(bS, {
+                MAPE: () => TS,
+                MSE: () => FS,
+                binaryAccuracy: () => vS,
+                binaryCrossentropy: () => xS,
+                categoricalAccuracy: () => kS,
+                categoricalCrossentropy: () => SS,
+                cosineProximity: () => ES,
+                mape: () => RS,
+                meanAbsoluteError: () => NS,
+                meanAbsolutePercentageError: () => CS,
+                meanSquaredError: () => _S,
+                mse: () => MS,
+                precision: () => IS,
+                recall: () => AS,
+                sparseCategoricalAccuracy: () => wS
+            });
+            var DS = {};
+            w(DS, {
+                modelFromJSON: () => jv
+            });
+            var OS = {};
+
+            function LS(e) {
+                return new bx(e)
+            }
+
+            function PS(e) {
+                return gx(t = e), new bx({
+                    l1: null != t ? t.l1 : null,
+                    l2: 0
+                });
+                var t
+            }
+
+            function $S(e) {
+                return gx(t = e), new bx({
+                    l2: null != t ? t.l2 : null,
+                    l1: 0
+                });
+                var t
+            }
+            w(OS, {
+                l1: () => PS,
+                l1l2: () => LS,
+                l2: () => $S
+            });
+            var BS = class extends _b {
+                constructor() {
+                    super(...arguments), this.model = null
+                }
+                setModel(e) {
+                    if (!(e instanceof Wv)) throw new Error("model must be a LayersModel, not some other Container");
+                    this.model = e
+                }
+            };
+
+            function zS(e, t) {
+                return e < t
+            }
+
+            function US(e, t) {
+                return e > t
+            }
+            var WS = class extends BS {
+                constructor(e) {
+                    if (super(), null == e && (e = {}), e.restoreBestWeights) throw new ug("restoreBestWeights = True is not implemented in EarlyStopping yet.");
+                    this.monitor = e.monitor || "val_loss", this.minDelta = Math.abs(e.minDelta || 0), this.patience = e.patience || 0, this.verbose = e.verbose || 0, this.mode = e.mode || "auto", this.baseline = e.baseline, -1 === ["auto", "min", "max"].indexOf(this.mode) && (console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`), this.mode = "auto"), "min" === this.mode ? this.monitorFunc = zS : "max" === this.mode || -1 !== this.monitor.indexOf("acc") ? this.monitorFunc = US : this.monitorFunc = zS, this.monitorFunc === zS && (this.minDelta *= -1)
+                }
+                async onTrainBegin(e) {
+                    this.wait = 0, this.stoppedEpoch = 0, null != this.baseline ? this.best = this.baseline : this.best = this.monitorFunc === zS ? 1 / 0 : -1 / 0
+                }
+                async onEpochEnd(e, t) {
+                    await Tb(t);
+                    const n = this.getMonitorValue(t);
+                    null != n && (this.monitorFunc(n - this.minDelta, this.best) ? (this.best = n, this.wait = 0) : (this.wait++, this.wait >= this.patience && (this.stoppedEpoch = e, this.model.stopTraining = !0)))
+                }
+                async onTrainEnd(e) {
+                    this.stoppedEpoch > 0 && this.verbose && console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)
+                }
+                getMonitorValue(e) {
+                    null == e && (e = {});
+                    const t = e[this.monitor];
+                    return null == t && console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`), t
+                }
+            };
+            var VS, jS, GS, HS, qS, KS = {
+                earlyStopping: function (e) {
+                    return new WS(e)
+                }
+            };
+            Ye().registerFlag("KEEP_INTERMEDIATE_TENSORS", () => !1, e => {
+                e && console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")
+            }), (jS = VS || (VS = {}))[jS.DT_INVALID = 0] = "DT_INVALID", jS[jS.DT_FLOAT = 1] = "DT_FLOAT", jS[jS.DT_DOUBLE = 2] = "DT_DOUBLE", jS[jS.DT_INT32 = 3] = "DT_INT32", jS[jS.DT_UINT8 = 4] = "DT_UINT8", jS[jS.DT_INT16 = 5] = "DT_INT16", jS[jS.DT_INT8 = 6] = "DT_INT8", jS[jS.DT_STRING = 7] = "DT_STRING", jS[jS.DT_COMPLEX64 = 8] = "DT_COMPLEX64", jS[jS.DT_INT64 = 9] = "DT_INT64", jS[jS.DT_BOOL = 10] = "DT_BOOL", jS[jS.DT_QINT8 = 11] = "DT_QINT8", jS[jS.DT_QUINT8 = 12] = "DT_QUINT8", jS[jS.DT_QINT32 = 13] = "DT_QINT32", jS[jS.DT_BFLOAT16 = 14] = "DT_BFLOAT16", jS[jS.DT_QINT16 = 15] = "DT_QINT16", jS[jS.DT_QUINT16 = 16] = "DT_QUINT16", jS[jS.DT_UINT16 = 17] = "DT_UINT16", jS[jS.DT_COMPLEX128 = 18] = "DT_COMPLEX128", jS[jS.DT_HALF = 19] = "DT_HALF", jS[jS.DT_RESOURCE = 20] = "DT_RESOURCE", jS[jS.DT_VARIANT = 21] = "DT_VARIANT", jS[jS.DT_UINT32 = 22] = "DT_UINT32", jS[jS.DT_UINT64 = 23] = "DT_UINT64", jS[jS.DT_FLOAT_REF = 101] = "DT_FLOAT_REF", jS[jS.DT_DOUBLE_REF = 102] = "DT_DOUBLE_REF", jS[jS.DT_INT32_REF = 103] = "DT_INT32_REF", jS[jS.DT_UINT8_REF = 104] = "DT_UINT8_REF", jS[jS.DT_INT16_REF = 105] = "DT_INT16_REF", jS[jS.DT_INT8_REF = 106] = "DT_INT8_REF", jS[jS.DT_STRING_REF = 107] = "DT_STRING_REF", jS[jS.DT_COMPLEX64_REF = 108] = "DT_COMPLEX64_REF", jS[jS.DT_INT64_REF = 109] = "DT_INT64_REF", jS[jS.DT_BOOL_REF = 110] = "DT_BOOL_REF", jS[jS.DT_QINT8_REF = 111] = "DT_QINT8_REF", jS[jS.DT_QUINT8_REF = 112] = "DT_QUINT8_REF", jS[jS.DT_QINT32_REF = 113] = "DT_QINT32_REF", jS[jS.DT_BFLOAT16_REF = 114] = "DT_BFLOAT16_REF", jS[jS.DT_QINT16_REF = 115] = "DT_QINT16_REF", jS[jS.DT_QUINT16_REF = 116] = "DT_QUINT16_REF", jS[jS.DT_UINT16_REF = 117] = "DT_UINT16_REF", jS[jS.DT_COMPLEX128_REF = 118] = "DT_COMPLEX128_REF", jS[jS.DT_HALF_REF = 119] = "DT_HALF_REF", jS[jS.DT_RESOURCE_REF = 120] = "DT_RESOURCE_REF", jS[jS.DT_VARIANT_REF = 121] = "DT_VARIANT_REF", jS[jS.DT_UINT32_REF = 122] = "DT_UINT32_REF", jS[jS.DT_UINT64_REF = 123] = "DT_UINT64_REF", HS = GS || (GS = {}), (qS = HS.CheckpointFormatVersion || (HS.CheckpointFormatVersion = {}))[qS.LEGACY = 0] = "LEGACY", qS[qS.V1 = 1] = "V1", qS[qS.V2 = 2] = "V2";
+            var YS = {};
+
+            function XS(e, t) {
+                const n = {
+                    tfOpName: e,
+                    category: "custom",
+                    inputs: [],
+                    attrs: [],
+                    customExecutor: t
+                };
+                YS[e] = n
+            }
+
+            function QS(e) {
+                return YS[e]
+            }
+
+            function ZS(e) {
+                delete YS[e]
+            }
+
+            function JS(e, t, n, r, a) {
+                const s = t.inputParams[e];
+                if (s && void 0 !== s.inputIndexStart) {
+                    const e = s.inputIndexStart,
+                        i = 0 === s.inputIndexEnd ? void 0 : void 0 === s.inputIndexEnd ? e + 1 : s.inputIndexEnd;
+                    if ("tensor" === s.type) return eI(t.inputNames[s.inputIndexStart], n, r, a);
+                    if ("tensors" === s.type) {
+                        return t.inputNames.slice(e, i).map(e => eI(e, n, r, a))
+                    }
+                    const o = eI(t.inputNames.slice(e)[0], n, r, a),
+                        u = o.dataSync();
+                    return "number" === s.type ? u[0] : va.toNestedArray(o.shape, u)
+                }
+                const i = t.attrParams[e];
+                return i && i.value
+            }
+
+            function eI(e, t, n, r) {
+                const [a, s] = rI(e);
+                if (null != r) {
+                    const e = r.getHashTableHandleByName(a);
+                    if (null != e) return e
+                }
+                const i = n.currentContextIds.find(e => !!t[nI(a, e)]);
+                return void 0 !== i ? t[nI(a, i)][s] : void 0
+            }
+
+            function tI(e, t) {
+                const [n, r, a] = rI(e);
+                return [nI(n, t && t.currentContextId), r, a]
+            }
+
+            function nI(e, t) {
+                return t ? `${e}-${t}` : e
+            }
+
+            function rI(e) {
+                const t = e.split(":");
+                if (1 === t.length) return [e, 0, void 0];
+                const n = t[0],
+                    r = 3 === t.length ? t[1] : void 0;
+                return [n, Number(t[t.length - 1]), r]
+            }
+
+            function aI(e, t, n) {
+                let r = JS("pad", e, t, n);
+                if ("explicit" === r) {
+                    r = JS("explicitPaddings", e, t, n);
+                    const a = [
+                        [0, 0],
+                        [0, 0],
+                        [0, 0],
+                        [0, 0]
+                    ];
+                    for (let e = 0; e < 4; e++) a[e][0] = r[2 * e], a[e][1] = r[2 * e + 1];
+                    return a
+                }
+                return r
+            }
+
+            function sI(e) {
+                return e.kept ? e : Ei(e)
+            }
+            var iI = {};
+            w(iI, {
+                json: () => oI
+            });
+            var oI = [{
+                    tfOpName: "Add",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "AddV2",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "AddN",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "tensors",
+                        type: "tensors"
+                    }]
+                }, {
+                    tfOpName: "BiasAdd",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Sub",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "RealDiv",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Div",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "DivNoNan",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "FloorDiv",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Mul",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Maximum",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Minimum",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Pow",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "SquaredDifference",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Mod",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "FloorMod",
+                    category: "arithmetic",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }],
+                uI = {};
+            w(uI, {
+                json: () => lI
+            });
+            var lI = [{
+                    tfOpName: "Abs",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Acos",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Asin",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Atan",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Atan2",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "y",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Ceil",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "ClipByValue",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "clipValueMin",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "clipValueMax",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Complex",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "real",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "imag",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "ComplexAbs",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Cos",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Cosh",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Elu",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Exp",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Floor",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Log",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Imag",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "Tout",
+                        name: "outputType",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Neg",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Real",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "Tout",
+                        name: "outputType",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Prelu",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "alpha",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Relu",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Relu6",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Selu",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Sigmoid",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Sin",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Sinh",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Sqrt",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Rsqrt",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Square",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Tan",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Tanh",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Sign",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Round",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Expm1",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Log1p",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Reciprocal",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Softplus",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Asinh",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Acosh",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Atanh",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Erf",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Prod",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axes",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool",
+                        notSupported: !0
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LeakyRelu",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "alpha",
+                        name: "alpha",
+                        type: "number",
+                        defaultValue: .2
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "IsNan",
+                    category: "basic_math",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }],
+                cI = {};
+            w(cI, {
+                json: () => hI
+            });
+            var hI = [{
+                    tfOpName: "EmptyTensorList",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "elementShape",
+                        type: "shape"
+                    }, {
+                        start: 1,
+                        name: "maxNumElements",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "LoopCond",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "pred",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "Switch",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "data",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "pred",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "Merge",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "tensors",
+                        type: "tensors"
+                    }]
+                }, {
+                    tfOpName: "Enter",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "frame_name",
+                        name: "frameName",
+                        type: "string"
+                    }, {
+                        tfName: "is_constant",
+                        name: "isConstant",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Exit",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "NextIteration",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "TensorArrayV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "size",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }, {
+                        tfName: "element_shape",
+                        name: "elementShape",
+                        type: "shape"
+                    }, {
+                        tfName: "dynamic_size",
+                        name: "dynamicSize",
+                        type: "bool"
+                    }, {
+                        tfName: "clear_after_read",
+                        name: "clearAfterRead",
+                        type: "bool"
+                    }, {
+                        tfName: "identical_element_shapes",
+                        name: "identicalElementShapes",
+                        type: "bool"
+                    }, {
+                        tfName: "tensor_array_name",
+                        name: "name",
+                        type: "string"
+                    }]
+                }, {
+                    tfOpName: "TensorArrayWriteV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "index",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "flowIn",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "TensorArrayReadV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "index",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "flowIn",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "TensorArrayGatherV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "flowIn",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }, {
+                        tfName: "element_shape",
+                        name: "elementShape",
+                        type: "shape"
+                    }]
+                }, {
+                    tfOpName: "TensorArrayScatterV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "flowIn",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorArrayConcatV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "flowIn",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }, {
+                        tfName: "element_shape_except0",
+                        name: "elementShapeExcept0",
+                        type: "shape",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "TensorArraySplitV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "lengths",
+                        type: "number[]"
+                    }, {
+                        start: 3,
+                        name: "flowIn",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorArraySizeV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "flowIn",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "TensorArrayCloseV3",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorArrayId",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "StatelessIf",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "cond",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "then_branch",
+                        name: "thenBranch",
+                        type: "func"
+                    }, {
+                        tfName: "else_branch",
+                        name: "elseBranch",
+                        type: "func"
+                    }]
+                }, {
+                    tfOpName: "If",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "cond",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "then_branch",
+                        name: "thenBranch",
+                        type: "func"
+                    }, {
+                        tfName: "else_branch",
+                        name: "elseBranch",
+                        type: "func"
+                    }]
+                }, {
+                    tfOpName: "StatelessWhile",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "cond",
+                        name: "cond",
+                        type: "func"
+                    }, {
+                        tfName: "body",
+                        name: "body",
+                        type: "func"
+                    }]
+                }, {
+                    tfOpName: "While",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "cond",
+                        name: "cond",
+                        type: "func"
+                    }, {
+                        tfName: "body",
+                        name: "body",
+                        type: "func"
+                    }]
+                }, {
+                    tfOpName: "TensorListScatter",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "elementShape",
+                        type: "shape"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListScatterV2",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "elementShape",
+                        type: "shape"
+                    }, {
+                        start: 3,
+                        name: "numElements",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListGather",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "elementShape",
+                        type: "shape"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListGetItem",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "index",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "elementShape",
+                        type: "shape"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListSetItem",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "index",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "tensor",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListReserve",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "elementShape",
+                        type: "shape"
+                    }, {
+                        start: 1,
+                        name: "numElements",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListFromTensor",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "elementShape",
+                        type: "shape"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListStack",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "elementShape",
+                        type: "shape"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }, {
+                        tfName: "num_elements",
+                        name: "numElements",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListSplit",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "elementShape",
+                        type: "shape"
+                    }, {
+                        start: 2,
+                        name: "lengths",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListConcat",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "element_shape",
+                        name: "elementShape",
+                        type: "shape"
+                    }, {
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListPopBack",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "elementShape",
+                        type: "shape"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TensorListPushBack",
+                    category: "control",
+                    inputs: [{
+                        start: 0,
+                        name: "tensorListId",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "tensor",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "element_dtype",
+                        name: "elementDType",
+                        type: "dtype"
+                    }]
+                }],
+                pI = {};
+            w(pI, {
+                json: () => dI
+            });
+            var dI = [{
+                    tfOpName: "AvgPool",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }, {
+                        tfName: "ksize",
+                        name: "kernelSize",
+                        type: "number[]"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "MaxPool",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }, {
+                        tfName: "ksize",
+                        name: "kernelSize",
+                        type: "number[]"
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: [],
+                        notSupported: !0
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "MaxPoolWithArgmax",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "ksize",
+                        name: "kernelSize",
+                        type: "number[]"
+                    }, {
+                        tfName: "include_batch_in_index",
+                        name: "includeBatchInIndex",
+                        type: "bool"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "AvgPool3D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }, {
+                        tfName: "ksize",
+                        name: "kernelSize",
+                        type: "number[]"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "MaxPool3D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }, {
+                        tfName: "ksize",
+                        name: "kernelSize",
+                        type: "number[]"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Conv1D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "stride",
+                        name: "stride",
+                        type: "number"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NWC"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "dilation",
+                        name: "dilation",
+                        type: "number",
+                        defaultValue: 1
+                    }]
+                }, {
+                    tfOpName: "Conv2D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "useCudnnOnGpu",
+                        name: "useCudnnOnGpu",
+                        type: "bool"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NHWC"
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "_FusedConv2D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "num_args",
+                        name: "numArgs",
+                        type: "number"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "use_cudnn_on_gpu",
+                        name: "useCudnnOnGpu",
+                        type: "bool",
+                        defaultValue: !0
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NHWC"
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]",
+                        defaultValue: [1, 1, 1, 1]
+                    }, {
+                        tfName: "fused_ops",
+                        name: "fusedOps",
+                        type: "string[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "epsilon",
+                        name: "epsilon",
+                        type: "number",
+                        defaultValue: 1e-4
+                    }, {
+                        tfName: "leakyrelu_alpha",
+                        name: "leakyreluAlpha",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "Conv2DBackpropInput",
+                    category: "convolution",
+                    inputs: [{
+                        start: 2,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }, {
+                        start: 0,
+                        name: "outputShape",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "DepthwiseConv2d",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "input",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NHWC"
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "DepthwiseConv2dNative",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "input",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NHWC"
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "FusedDepthwiseConv2dNative",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "num_args",
+                        name: "numArgs",
+                        type: "number"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NHWC"
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]",
+                        defaultValue: [1, 1, 1, 1]
+                    }, {
+                        tfName: "fused_ops",
+                        name: "fusedOps",
+                        type: "string[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "explicit_paddings",
+                        name: "explicitPaddings",
+                        type: "number[]",
+                        defaultValue: []
+                    }]
+                }, {
+                    tfOpName: "Conv3D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        defaultValue: "NHWC"
+                    }, {
+                        tfName: "dilations",
+                        name: "dilations",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "Dilation2D",
+                    category: "convolution",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "filter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "strides",
+                        name: "strides",
+                        type: "number[]"
+                    }, {
+                        tfName: "rates",
+                        name: "dilations",
+                        type: "number[]"
+                    }, {
+                        tfName: "padding",
+                        name: "pad",
+                        type: "string"
+                    }]
+                }],
+                fI = {};
+            w(fI, {
+                json: () => mI
+            });
+            var mI = [{
+                    tfOpName: "Fill",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "shape",
+                        type: "number[]"
+                    }, {
+                        start: 1,
+                        name: "value",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "LinSpace",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "start",
+                        type: "number"
+                    }, {
+                        start: 1,
+                        name: "stop",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "num",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "OneHot",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "indices",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "depth",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "onValue",
+                        type: "number",
+                        defaultValue: 1
+                    }, {
+                        start: 3,
+                        name: "offValue",
+                        type: "number",
+                        defaultValue: 0
+                    }],
+                    attrs: [{
+                        tfName: "axis",
+                        name: "axis",
+                        type: "number",
+                        notSupported: !0
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Ones",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "shape",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "OnesLike",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "RandomUniform",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "shape",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "minval",
+                        name: "minval",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "maxval",
+                        name: "maxval",
+                        type: "number",
+                        defaultValue: 1
+                    }, {
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }, {
+                        tfName: "seed",
+                        name: "seed",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "seed2",
+                        name: "seed2",
+                        type: "number",
+                        defaultValue: 0,
+                        notSupported: !0
+                    }, {
+                        tfName: "T",
+                        name: "T",
+                        type: "number",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Range",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "start",
+                        type: "number"
+                    }, {
+                        start: 1,
+                        name: "stop",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "step",
+                        type: "number",
+                        defaultValue: 0
+                    }],
+                    attrs: [{
+                        tfName: "Tidx",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "TruncatedNormal",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "shape",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "means",
+                        name: "mean",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "stddev",
+                        name: "stdDev",
+                        type: "number",
+                        defaultValue: 1
+                    }, {
+                        tfName: "seed",
+                        name: "seed",
+                        type: "number"
+                    }, {
+                        tfName: "seed2",
+                        name: "seed2",
+                        type: "number",
+                        defaultValue: 0,
+                        notSupported: !0
+                    }, {
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }, {
+                        tfName: "T",
+                        name: "T",
+                        type: "number",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Zeros",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "shape",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "ZerosLike",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "Multinomial",
+                    category: "creation",
+                    inputs: [{
+                        start: 0,
+                        name: "logits",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "numSamples",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "seed",
+                        name: "seed",
+                        type: "number"
+                    }, {
+                        tfName: "seed2",
+                        name: "seed2",
+                        type: "number"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }, {
+                        tfName: "output_dtype",
+                        name: "output_dtype",
+                        type: "dtype"
+                    }]
+                }],
+                gI = {};
+            w(gI, {
+                json: () => yI
+            });
+            var yI = [{
+                    tfOpName: "NonMaxSuppressionV2",
+                    category: "dynamic",
+                    inputs: [{
+                        start: 0,
+                        name: "boxes",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scores",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "maxOutputSize",
+                        type: "number"
+                    }, {
+                        start: 3,
+                        name: "iouThreshold",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "NonMaxSuppressionV3",
+                    category: "dynamic",
+                    inputs: [{
+                        start: 0,
+                        name: "boxes",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scores",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "maxOutputSize",
+                        type: "number"
+                    }, {
+                        start: 3,
+                        name: "iouThreshold",
+                        type: "number"
+                    }, {
+                        start: 4,
+                        name: "scoreThreshold",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "NonMaxSuppressionV4",
+                    category: "dynamic",
+                    inputs: [{
+                        start: 0,
+                        name: "boxes",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scores",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "maxOutputSize",
+                        type: "number"
+                    }, {
+                        start: 3,
+                        name: "iouThreshold",
+                        type: "number"
+                    }, {
+                        start: 4,
+                        name: "scoreThreshold",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "T_threshold",
+                        name: "threshold",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "pad_to_max_output_size",
+                        name: "padToMaxOutputSize",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "NonMaxSuppressionV5",
+                    category: "dynamic",
+                    inputs: [{
+                        start: 0,
+                        name: "boxes",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scores",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "maxOutputSize",
+                        type: "number"
+                    }, {
+                        start: 3,
+                        name: "iouThreshold",
+                        type: "number"
+                    }, {
+                        start: 4,
+                        name: "scoreThreshold",
+                        type: "number"
+                    }, {
+                        start: 5,
+                        name: "softNmsSigma",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "Where",
+                    category: "dynamic",
+                    inputs: [{
+                        start: 0,
+                        name: "condition",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "ListDiff",
+                    category: "dynamic",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "y",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }],
+                bI = {};
+            w(bI, {
+                json: () => vI
+            });
+            var vI = [{
+                    tfOpName: "TopKV2",
+                    category: "evaluation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "k",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "sorted",
+                        name: "sorted",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Unique",
+                    category: "evaluation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "UniqueV2",
+                    category: "evaluation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number"
+                    }]
+                }],
+                xI = {};
+            w(xI, {
+                json: () => wI
+            });
+            var wI = [{
+                    tfOpName: "PlaceholderWithDefault",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "default",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "shape",
+                        name: "shape",
+                        type: "shape"
+                    }, {
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "Placeholder",
+                    category: "graph",
+                    attrs: [{
+                        tfName: "shape",
+                        name: "shape",
+                        type: "shape"
+                    }, {
+                        tfName: "dtype",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "Const",
+                    category: "graph"
+                }, {
+                    tfOpName: "Identity",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "IdentityN",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "x",
+                        type: "tensors"
+                    }]
+                }, {
+                    tfOpName: "Snapshot",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "Rank",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "Size",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "Shape",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "ShapeN",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "x",
+                        type: "tensors"
+                    }]
+                }, {
+                    tfOpName: "Print",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "data",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "message",
+                        name: "message",
+                        type: "string"
+                    }, {
+                        tfName: "first_n",
+                        name: "firstN",
+                        type: "number",
+                        notSupported: !0
+                    }, {
+                        tfName: "summarize",
+                        name: "summarize",
+                        type: "number",
+                        defaultValue: 3
+                    }]
+                }, {
+                    tfOpName: "NoOp",
+                    category: "graph",
+                    inputs: []
+                }, {
+                    tfOpName: "StopGradient",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "FakeQuantWithMinMaxVars",
+                    category: "graph",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "min",
+                        name: "min",
+                        type: "number"
+                    }, {
+                        tfName: "max",
+                        name: "max",
+                        type: "number"
+                    }]
+                }],
+                kI = {};
+            w(kI, {
+                json: () => SI
+            });
+            var SI = [{
+                    tfOpName: "HashTable",
+                    category: "hash_table",
+                    inputs: [],
+                    attrs: [{
+                        tfName: "shared_name",
+                        name: "sharedName",
+                        type: "string"
+                    }, {
+                        tfName: "use_node_name_sharing",
+                        name: "useNodeNameSharing",
+                        type: "bool"
+                    }, {
+                        tfName: "key_dtype",
+                        name: "keyDType",
+                        type: "dtype"
+                    }, {
+                        tfName: "value_dtype",
+                        name: "valueDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "HashTableV2",
+                    category: "hash_table",
+                    inputs: [],
+                    attrs: [{
+                        tfName: "shared_name",
+                        name: "sharedName",
+                        type: "string"
+                    }, {
+                        tfName: "use_node_name_sharing",
+                        name: "useNodeNameSharing",
+                        type: "bool"
+                    }, {
+                        tfName: "key_dtype",
+                        name: "keyDType",
+                        type: "dtype"
+                    }, {
+                        tfName: "value_dtype",
+                        name: "valueDType",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "LookupTableImport",
+                    category: "hash_table",
+                    inputs: [{
+                        start: 0,
+                        name: "tableHandle",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "keys",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "values",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "Tin",
+                        name: "tIn",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "Tout",
+                        name: "tOut",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LookupTableImportV2",
+                    category: "hash_table",
+                    inputs: [{
+                        start: 0,
+                        name: "tableHandle",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "keys",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "values",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "Tin",
+                        name: "tIn",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "Tout",
+                        name: "tOut",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LookupTableFind",
+                    category: "hash_table",
+                    inputs: [{
+                        start: 0,
+                        name: "tableHandle",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "keys",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "defaultValue",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "Tin",
+                        name: "tIn",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "Tout",
+                        name: "tOut",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LookupTableFindV2",
+                    category: "hash_table",
+                    inputs: [{
+                        start: 0,
+                        name: "tableHandle",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "keys",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "defaultValue",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "Tin",
+                        name: "tIn",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "Tout",
+                        name: "tOut",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LookupTableSize",
+                    category: "hash_table",
+                    inputs: [{
+                        start: 0,
+                        name: "tableHandle",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "LookupTableSizeV2",
+                    category: "hash_table",
+                    inputs: [{
+                        start: 0,
+                        name: "tableHandle",
+                        type: "tensor"
+                    }]
+                }],
+                II = {};
+            w(II, {
+                json: () => AI
+            });
+            var AI = [{
+                    tfOpName: "ResizeBilinear",
+                    category: "image",
+                    inputs: [{
+                        start: 0,
+                        name: "images",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "size",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "align_corners",
+                        name: "alignCorners",
+                        type: "bool"
+                    }, {
+                        tfName: "half_pixel_centers",
+                        name: "halfPixelCenters",
+                        type: "bool"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "ResizeNearestNeighbor",
+                    category: "image",
+                    inputs: [{
+                        start: 0,
+                        name: "images",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "size",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "align_corners",
+                        name: "alignCorners",
+                        type: "bool"
+                    }, {
+                        tfName: "half_pixel_centers",
+                        name: "halfPixelCenters",
+                        type: "bool"
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "CropAndResize",
+                    category: "image",
+                    inputs: [{
+                        start: 0,
+                        name: "image",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "boxes",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "boxInd",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "cropSize",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "method",
+                        name: "method",
+                        type: "string"
+                    }, {
+                        tfName: "extrapolation_value",
+                        name: "extrapolationValue",
+                        type: "number"
+                    }]
+                }],
+                EI = {};
+            w(EI, {
+                json: () => NI
+            });
+            var NI = [{
+                    tfOpName: "Equal",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "NotEqual",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Greater",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "GreaterEqual",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Less",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LessEqual",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LogicalAnd",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LogicalNot",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LogicalOr",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Select",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "condition",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "SelectV2",
+                    category: "logical",
+                    inputs: [{
+                        start: 0,
+                        name: "condition",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }],
+                CI = {};
+            w(CI, {
+                json: () => TI
+            });
+            var TI = [{
+                    tfOpName: "_FusedMatMul",
+                    category: "matrices",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        end: 0,
+                        name: "args",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "num_args",
+                        name: "numArgs",
+                        type: "number"
+                    }, {
+                        tfName: "fused_ops",
+                        name: "fusedOps",
+                        type: "string[]",
+                        defaultValue: []
+                    }, {
+                        tfName: "epsilon",
+                        name: "epsilon",
+                        type: "number",
+                        defaultValue: 1e-4
+                    }, {
+                        tfName: "transpose_a",
+                        name: "transposeA",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "transpose_b",
+                        name: "transposeB",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "MatMul",
+                    category: "matrices",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "transpose_a",
+                        name: "transposeA",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "transpose_b",
+                        name: "transposeB",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "BatchMatMul",
+                    category: "matrices",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "adj_x",
+                        name: "transposeA",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "adj_y",
+                        name: "transposeB",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "BatchMatMulV2",
+                    category: "matrices",
+                    inputs: [{
+                        start: 0,
+                        name: "a",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "b",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "adj_x",
+                        name: "transposeA",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "adj_y",
+                        name: "transposeB",
+                        type: "bool",
+                        defaultValue: !1
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Transpose",
+                    category: "matrices",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "perm",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Einsum",
+                    category: "matrices",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "tensors",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "equation",
+                        name: "equation",
+                        type: "string"
+                    }, {
+                        tfName: "N",
+                        name: "n",
+                        type: "number",
+                        defaultValue: 2
+                    }, {
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }],
+                RI = {};
+            w(RI, {
+                json: () => _I
+            });
+            var _I = [{
+                    tfOpName: "FusedBatchNorm",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scale",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "offset",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "mean",
+                        type: "tensor"
+                    }, {
+                        start: 4,
+                        name: "variance",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "epsilon",
+                        name: "epsilon",
+                        type: "number",
+                        defaultValue: .001
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "FusedBatchNormV2",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scale",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "offset",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "mean",
+                        type: "tensor"
+                    }, {
+                        start: 4,
+                        name: "variance",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "epsilon",
+                        name: "epsilon",
+                        type: "number",
+                        defaultValue: .001
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "FusedBatchNormV3",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "scale",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "offset",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "mean",
+                        type: "tensor"
+                    }, {
+                        start: 4,
+                        name: "variance",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "epsilon",
+                        name: "epsilon",
+                        type: "number",
+                        defaultValue: .001
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "LRN",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "depth_radius",
+                        name: "radius",
+                        type: "number",
+                        defaultValue: 5
+                    }, {
+                        tfName: "bias",
+                        name: "bias",
+                        type: "number",
+                        defaultValue: 1
+                    }, {
+                        tfName: "alpha",
+                        name: "alpha",
+                        type: "number",
+                        defaultValue: 1
+                    }, {
+                        tfName: "beta",
+                        name: "beta",
+                        type: "number",
+                        defaultValue: .5
+                    }]
+                }, {
+                    tfOpName: "Softmax",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "LogSoftmax",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "SparseToDense",
+                    category: "normalization",
+                    inputs: [{
+                        start: 0,
+                        name: "sparseIndices",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "outputShape",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "sparseValues",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "defaultValue",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "validate_indices",
+                        name: "validateIndices",
+                        type: "bool",
+                        defaultValue: !0,
+                        notSupported: !0
+                    }]
+                }],
+                FI = {};
+            w(FI, {
+                json: () => MI
+            });
+            var MI = [{
+                    tfOpName: "Bincount",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "size",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "weights",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "DenseBincount",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "size",
+                        type: "number"
+                    }, {
+                        start: 2,
+                        name: "weights",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "binary_output",
+                        name: "binaryOutput",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Max",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Mean",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Min",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Sum",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "All",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Any",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "ArgMax",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "ArgMin",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "Prod",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "keep_dims",
+                        name: "keepDims",
+                        type: "bool"
+                    }]
+                }, {
+                    tfOpName: "Cumsum",
+                    category: "reduction",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "exclusive",
+                        name: "exclusive",
+                        type: "bool"
+                    }, {
+                        tfName: "reverse",
+                        name: "reverse",
+                        type: "bool"
+                    }]
+                }],
+                DI = {};
+            w(DI, {
+                json: () => OI
+            });
+            var OI = [{
+                    tfOpName: "ConcatV2",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        end: -1,
+                        name: "tensors",
+                        type: "tensors"
+                    }, {
+                        start: -1,
+                        name: "axis",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "N",
+                        name: "n",
+                        type: "number",
+                        defaultValue: 2
+                    }]
+                }, {
+                    tfOpName: "Concat",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 1,
+                        end: 0,
+                        name: "tensors",
+                        type: "tensors"
+                    }, {
+                        start: 0,
+                        name: "axis",
+                        type: "number"
+                    }],
+                    attrs: [{
+                        tfName: "N",
+                        name: "n",
+                        type: "number",
+                        defaultValue: 2
+                    }]
+                }, {
+                    tfOpName: "GatherV2",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "axis",
+                        type: "number",
+                        defaultValue: 0
+                    }],
+                    attrs: [{
+                        tfName: "batch_dims",
+                        name: "batchDims",
+                        type: "number",
+                        defaultValue: 0
+                    }]
+                }, {
+                    tfOpName: "Gather",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "validate_indices",
+                        name: "validateIndices",
+                        type: "bool",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Reverse",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "dims",
+                        type: "bool[]"
+                    }]
+                }, {
+                    tfOpName: "ReverseV2",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "Slice",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "begin",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "size",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "StridedSlice",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "begin",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "end",
+                        type: "number[]"
+                    }, {
+                        start: 3,
+                        name: "strides",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "begin_mask",
+                        name: "beginMask",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "end_mask",
+                        name: "endMask",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "new_axis_mask",
+                        name: "newAxisMask",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "ellipsis_mask",
+                        name: "ellipsisMask",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "shrink_axis_mask",
+                        name: "shrinkAxisMask",
+                        type: "number",
+                        defaultValue: 0
+                    }]
+                }, {
+                    tfOpName: "Pack",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        end: 0,
+                        name: "tensors",
+                        type: "tensors"
+                    }],
+                    attrs: [{
+                        tfName: "axis",
+                        name: "axis",
+                        type: "number",
+                        defaultValue: 0
+                    }]
+                }, {
+                    tfOpName: "Unpack",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "tensor",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "axis",
+                        name: "axis",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        tfName: "num",
+                        name: "num",
+                        type: "number",
+                        defaultValue: 0,
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "Tile",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "reps",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "Split",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "axis",
+                        type: "number",
+                        defaultValue: 0
+                    }, {
+                        start: 1,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "num_split",
+                        name: "numOrSizeSplits",
+                        type: "number",
+                        defaultValue: 1
+                    }]
+                }, {
+                    tfOpName: "SplitV",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "numOrSizeSplits",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "axis",
+                        type: "number",
+                        defaultValue: 0
+                    }]
+                }, {
+                    tfOpName: "ScatterNd",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "indices",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "values",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "shape",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "GatherNd",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "SparseToDense",
+                    category: "slice_join",
+                    inputs: [{
+                        start: 0,
+                        name: "sparseIndices",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "outputShape",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "sparseValues",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "defaultValue",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "validate_indices",
+                        name: "validateIndices",
+                        type: "bool",
+                        defaultValue: !1,
+                        notSupported: !0
+                    }]
+                }],
+                LI = {};
+            w(LI, {
+                json: () => PI
+            });
+            var PI = [{
+                    tfOpName: "SparseFillEmptyRows",
+                    category: "sparse",
+                    inputs: [{
+                        start: 0,
+                        name: "indices",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "values",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "denseShape",
+                        type: "tensor"
+                    }, {
+                        start: 3,
+                        name: "defaultValue",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "SparseReshape",
+                    category: "sparse",
+                    inputs: [{
+                        start: 0,
+                        name: "inputIndices",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "inputShape",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "newShape",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "T",
+                        name: "dtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "SparseSegmentMean",
+                    category: "sparse",
+                    inputs: [{
+                        start: 0,
+                        name: "data",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "segmentIds",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "SparseSegmentSum",
+                    category: "sparse",
+                    inputs: [{
+                        start: 0,
+                        name: "data",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "indices",
+                        type: "tensor"
+                    }, {
+                        start: 2,
+                        name: "segmentIds",
+                        type: "tensor"
+                    }]
+                }],
+                $I = {};
+            w($I, {
+                json: () => BI
+            });
+            var BI = [{
+                    tfOpName: "FFT",
+                    category: "spectral",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "IFFT",
+                    category: "spectral",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }]
+                }, {
+                    tfOpName: "RFFT",
+                    category: "spectral",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "fft_length",
+                        type: "number",
+                        notSupported: !0
+                    }]
+                }, {
+                    tfOpName: "IRFFT",
+                    category: "spectral",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "fft_length",
+                        type: "number",
+                        notSupported: !0
+                    }]
+                }],
+                zI = {};
+            w(zI, {
+                json: () => UI
+            });
+            var UI = [{
+                    tfOpName: "StringNGrams",
+                    category: "string",
+                    inputs: [{
+                        start: 0,
+                        name: "data",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "dataSplits",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "separator",
+                        name: "separator",
+                        type: "string"
+                    }, {
+                        tfName: "ngram_widths",
+                        name: "nGramWidths",
+                        type: "number[]"
+                    }, {
+                        tfName: "left_pad",
+                        name: "leftPad",
+                        type: "string"
+                    }, {
+                        tfName: "right_pad",
+                        name: "rightPad",
+                        type: "string"
+                    }, {
+                        tfName: "pad_width",
+                        name: "padWidth",
+                        type: "number"
+                    }, {
+                        tfName: "preserve_short_sequences",
+                        name: "preserveShortSequences",
+                        type: "bool"
+                    }],
+                    outputs: ["ngrams", "ngrams_splits"]
+                }, {
+                    tfOpName: "StringSplit",
+                    category: "string",
+                    inputs: [{
+                        start: 0,
+                        name: "input",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "delimiter",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "skip_empty",
+                        name: "skipEmpty",
+                        type: "bool"
+                    }],
+                    outputs: ["indices", "values", "shape"]
+                }, {
+                    tfOpName: "StringToHashBucketFast",
+                    category: "string",
+                    inputs: [{
+                        start: 0,
+                        name: "input",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "num_buckets",
+                        name: "numBuckets",
+                        type: "number"
+                    }]
+                }],
+                WI = {};
+            w(WI, {
+                json: () => VI
+            });
+            var VI = [{
+                    tfOpName: "Cast",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "SrcT",
+                        name: "sdtype",
+                        type: "dtype",
+                        notSupported: !0
+                    }, {
+                        tfName: "DstT",
+                        name: "dtype",
+                        type: "dtype"
+                    }]
+                }, {
+                    tfOpName: "ExpandDims",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "axis",
+                        type: "number"
+                    }]
+                }, {
+                    tfOpName: "MirrorPad",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "padding",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "mode",
+                        name: "mode",
+                        type: "string"
+                    }]
+                }, {
+                    tfOpName: "Pad",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "padding",
+                        type: "number[]"
+                    }],
+                    attrs: [{
+                        tfName: "constant_value",
+                        name: "constantValue",
+                        type: "number",
+                        defaultValue: 0
+                    }]
+                }, {
+                    tfOpName: "PadV2",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "padding",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "constantValue",
+                        type: "number",
+                        defaultValue: 0
+                    }]
+                }, {
+                    tfOpName: "Reshape",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "shape",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "Squeeze",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "axis",
+                        tfDeprecatedName: "squeeze_dims",
+                        name: "axis",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "SpaceToBatchND",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "blockShape",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "paddings",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "BatchToSpaceND",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "blockShape",
+                        type: "number[]"
+                    }, {
+                        start: 2,
+                        name: "crops",
+                        type: "number[]"
+                    }]
+                }, {
+                    tfOpName: "DepthToSpace",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }],
+                    attrs: [{
+                        tfName: "block_size",
+                        name: "blockSize",
+                        type: "number"
+                    }, {
+                        tfName: "data_format",
+                        name: "dataFormat",
+                        type: "string"
+                    }]
+                }, {
+                    tfOpName: "BroadcastTo",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "x",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "shape",
+                        type: "number[]"
+                    }],
+                    attrs: []
+                }, {
+                    tfOpName: "BroadcastArgs",
+                    category: "transformation",
+                    inputs: [{
+                        start: 0,
+                        name: "s0",
+                        type: "tensor"
+                    }, {
+                        start: 1,
+                        name: "s1",
+                        type: "tensor"
+                    }],
+                    attrs: []
+                }],
+                jI = class {
+                    static get Instance() {
+                        return this._instance || (this._instance = new this)
+                    }
+                    constructor() {
+                        const e = [].concat(...[iI, uI, cI, pI, fI, gI, bI, xI, kI, II, EI, CI, RI, FI, DI, LI, $I, zI, WI].map(e => e.json));
+                        this.opMappers = e.reduce((e, t) => (e[t.tfOpName] = t, e), {})
+                    }
+                    transformGraph(e, t = {}) {
+                        const n = e.node,
+                            r = [],
+                            a = [],
+                            s = [],
+                            i = n.reduce((e, t) => (e[t.name] = this.mapNode(t), t.op.startsWith("Placeholder") ? r.push(e[t.name]) : "Const" === t.op ? a.push(e[t.name]) : null != t.input && 0 !== t.input.length || s.push(e[t.name]), e), {});
+                        let o = [];
+                        const u = [];
+                        let l = {},
+                            c = {};
+                        null != t && (l = this.mapSignatureEntries(t.inputs), c = this.mapSignatureEntries(t.outputs));
+                        const h = Object.keys(i);
+                        h.forEach(e => {
+                            const t = i[e];
+                            t.inputNames.forEach((e, n) => {
+                                const [r, , a] = tI(e), s = i[r];
+                                if (null != s.outputs) {
+                                    const e = s.outputs.indexOf(a);
+                                    if (-1 !== e) {
+                                        const a = `${r}:${e}`;
+                                        t.inputNames[n] = a
+                                    }
+                                }
+                                t.inputs.push(s), s.children.push(t)
+                            })
+                        }), 0 === Object.keys(c).length ? h.forEach(e => {
+                            const t = i[e];
+                            0 === t.children.length && u.push(t)
+                        }) : Object.keys(c).forEach(e => {
+                            const [t] = tI(e), n = i[t];
+                            null != n && (n.signatureKey = c[e], u.push(n))
+                        }), Object.keys(l).length > 0 ? Object.keys(l).forEach(e => {
+                            const [t] = tI(e), n = i[t];
+                            n && (n.signatureKey = l[e], o.push(n))
+                        }) : o = r;
+                        let p = {};
+                        null != e.library && null != e.library.function && (p = e.library.function.reduce((e, t) => (e[t.signature.name] = this.mapFunction(t), e), {}));
+                        const d = {
+                            nodes: i,
+                            inputs: o,
+                            outputs: u,
+                            weights: a,
+                            placeholders: r,
+                            signature: t,
+                            functions: p
+                        };
+                        return s.length > 0 && (d.initNodes = s), d
+                    }
+                    mapSignatureEntries(e) {
+                        return Object.keys(e || {}).reduce((t, n) => (t[e[n].name] = n, t), {})
+                    }
+                    mapNode(e) {
+                        const t = QS(e.op) || this.opMappers[e.op] || {};
+                        null == e.attr && (e.attr = {});
+                        const n = {
+                            name: e.name,
+                            op: e.op,
+                            category: t.category,
+                            inputNames: (e.input || []).map(e => e.startsWith("^") ? e.substr(1) : e),
+                            inputs: [],
+                            children: [],
+                            inputParams: {},
+                            attrParams: {},
+                            rawAttrs: e.attr,
+                            outputs: t.outputs
+                        };
+                        return null != t.inputs && (n.inputParams = t.inputs.reduce((e, t) => (e[t.name] = {
+                            type: t.type,
+                            inputIndexStart: t.start,
+                            inputIndexEnd: t.end
+                        }, e), {})), null != t.attrs && (n.attrParams = t.attrs.reduce((t, n) => {
+                            const r = n.type;
+                            let a = void 0;
+                            switch (n.type) {
+                                case "string":
+                                    a = HI(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = HI(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "string[]":
+                                    a = nA(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = nA(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "number":
+                                    a = KI(e.attr, n.tfName, n.defaultValue || 0), void 0 === a && n.tfDeprecatedName && (a = KI(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "number[]":
+                                    a = tA(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = tA(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "bool":
+                                    a = qI(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = qI(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "bool[]":
+                                    a = aA(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = aA(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "shape":
+                                    a = eA(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = eA(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "shape[]":
+                                    a = rA(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = rA(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "dtype":
+                                    a = QI(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = QI(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "dtype[]":
+                                    a = ZI(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = ZI(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "func":
+                                    a = XI(e.attr, n.tfName, n.defaultValue), void 0 === a && n.tfDeprecatedName && (a = XI(e.attr, n.tfDeprecatedName, n.defaultValue));
+                                    break;
+                                case "tensor":
+                                case "tensors":
+                                    break;
+                                default:
+                                    throw new Error(`Unsupported param type: ${n.type} for op: ${e.op}`)
+                            }
+                            return t[n.name] = {
+                                value: a,
+                                type: r
+                            }, t
+                        }, {})), n
+                    }
+                    mapFunction(e) {
+                        const t = e.nodeDef,
+                            n = [];
+                        let r = {};
+                        null != t && (r = t.reduce((e, t) => (e[t.name] = this.mapNode(t), "Const" === t.op && n.push(e[t.name]), e), {}));
+                        const a = [],
+                            s = [];
+                        e.signature.inputArg.forEach(e => {
+                            const [t] = tI(e.name), n = {
+                                name: t,
+                                op: "Placeholder",
+                                inputs: [],
+                                inputNames: [],
+                                category: "graph",
+                                inputParams: {},
+                                attrParams: {
+                                    dtype: {
+                                        value: YI(e.type),
+                                        type: "dtype"
+                                    }
+                                },
+                                children: []
+                            };
+                            n.signatureKey = e.name, a.push(n), r[t] = n
+                        });
+                        Object.keys(r).forEach(e => {
+                            const t = r[e];
+                            t.inputNames.forEach((e, n) => {
+                                const [a, , s] = tI(e), i = r[a];
+                                if (null != i.outputs) {
+                                    const e = i.outputs.indexOf(s);
+                                    if (-1 !== e) {
+                                        const r = `${a}:${e}`;
+                                        t.inputNames[n] = r
+                                    }
+                                }
+                                t.inputs.push(i), i.children.push(t)
+                            })
+                        });
+                        const i = e.ret;
+                        e.signature.outputArg.forEach(e => {
+                            const [t, n] = tI(i[e.name]), a = r[t];
+                            null != a && (a.defaultOutput = n, s.push(a))
+                        });
+                        const o = this.mapArgsToSignature(e);
+                        return {
+                            nodes: r,
+                            inputs: a,
+                            outputs: s,
+                            weights: n,
+                            placeholders: [],
+                            signature: o
+                        }
+                    }
+                    mapArgsToSignature(e) {
+                        return {
+                            methodName: e.signature.name,
+                            inputs: e.signature.inputArg.reduce((e, t) => (e[t.name] = this.mapArgToTensorInfo(t), e), {}),
+                            outputs: e.signature.outputArg.reduce((t, n) => (t[n.name] = this.mapArgToTensorInfo(n, e.ret), t), {})
+                        }
+                    }
+                    mapArgToTensorInfo(e, t) {
+                        let n = e.name;
+                        return null != t && (n = t[n]), {
+                            name: n,
+                            dtype: e.type
+                        }
+                    }
+                };
+
+            function GI(e, t) {
+                const n = Array.isArray(e) ? String.fromCharCode.apply(null, e) : function (e) {
+                    const t = Ye().global;
+                    if (void 0 !== t.atob) return t.atob(e);
+                    if (void 0 !== i) return new i(e, "base64").toString();
+                    throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")
+                }(e);
+                return t ? n : n.toLowerCase()
+            }
+
+            function HI(e, t, n, r = !1) {
+                const a = e[t];
+                return null != a ? GI(a.s, r) : n
+            }
+
+            function qI(e, t, n) {
+                const r = e[t];
+                return r ? r.b : n
+            }
+
+            function KI(e, t, n) {
+                const r = e[t] || {},
+                    a = null != r.i ? r.i : null != r.f ? r.f : n;
+                return "number" == typeof a ? a : parseInt(a, 10)
+            }
+
+            function YI(e) {
+                switch ("string" == typeof e && (e = VS[e]), e) {
+                    case VS.DT_FLOAT:
+                    case VS.DT_HALF:
+                        return "float32";
+                    case VS.DT_INT32:
+                    case VS.DT_INT64:
+                    case VS.DT_INT8:
+                    case VS.DT_UINT8:
+                        return "int32";
+                    case VS.DT_BOOL:
+                        return "bool";
+                    case VS.DT_DOUBLE:
+                        return "float32";
+                    case VS.DT_STRING:
+                        return "string";
+                    default:
+                        return null
+                }
+            }
+
+            function XI(e, t, n) {
+                const r = e[t];
+                return r && r.func ? r.func.name : n
+            }
+
+            function QI(e, t, n) {
+                const r = e[t];
+                return r && r.type ? YI(r.type) : n
+            }
+
+            function ZI(e, t, n) {
+                const r = e[t];
+                return r && r.list && r.list.type ? r.list.type.map(e => YI(e)) : n
+            }
+
+            function JI(e) {
+                if (!e.unknownRank) return null != e.dim ? e.dim.map(e => "number" == typeof e.size ? e.size : parseInt(e.size, 10)) : []
+            }
+
+            function eA(e, t, n) {
+                const r = e[t];
+                return r && r.shape ? JI(r.shape) : n
+            }
+
+            function tA(e, t, n) {
+                const r = e[t];
+                return r ? ((r.list.f && r.list.f.length ? r.list.f : r.list.i) || []).map(e => "number" == typeof e ? e : parseInt(e, 10)) : n
+            }
+
+            function nA(e, t, n, r = !1) {
+                const a = e[t];
+                return a && a.list && a.list.s ? a.list.s.map(e => GI(e, r)) : n
+            }
+
+            function rA(e, t, n) {
+                const r = e[t];
+                return r && r.list && r.list.shape ? r.list.shape.map(e => JI(e)) : n
+            }
+
+            function aA(e, t, n) {
+                const r = e[t];
+                return r && r.list && r.list.b ? r.list.b : n
+            }
+
+            function sA(e, t, n = "") {
+                if ("number" != typeof e && "number" != typeof t) {
+                    va.assert(e.length === t.length, () => n + ` Shapes ${e} and ${t} must match`);
+                    for (let r = 0; r < e.length; r++) {
+                        const a = e[r],
+                            s = t[r];
+                        va.assert(a < 0 || s < 0 || a === s, () => n + ` Shapes ${e} and ${t} must match`)
+                    }
+                }
+            }
+
+            function iA(e) {
+                return "number" != typeof e && !e.some(e => e < 0)
+            }
+
+            function oA(e, t, n) {
+                let r = uA(e, n);
+                const a = !iA(r);
+                if (a && 0 === t.length) throw new Error("Tried to calculate elements of an empty list with non-fully-defined elementShape: " + r);
+                if (a && t.forEach(e => {
+                        r = uA(e.shape, r)
+                    }), !iA(r)) throw new Error("Non-fully-defined elementShape: " + r);
+                return r
+            }
+
+            function uA(e, t) {
+                if ("number" == typeof e) return t;
+                if ("number" == typeof t) return e;
+                if (e.length !== t.length) throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);
+                const n = [];
+                for (let r = 0; r < e.length; ++r) {
+                    const a = e[r],
+                        s = t[r];
+                    if (a >= 0 && s >= 0 && a !== s) throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);
+                    n[r] = a >= 0 ? a : s
+                }
+                return n
+            }
+            var lA = class {
+                constructor(e, t, n, r = -1) {
+                    this.tensors = e, this.elementShape = t, this.elementDtype = n, null != e && e.forEach(e => {
+                        if (n !== e.dtype) throw new Error(`Invalid data types; op elements ${n}, but list elements ${e.dtype}`);
+                        sA(t, e.shape, "TensorList shape mismatch: "), su(e)
+                    }), this.idTensor = Ah(0), this.maxNumElements = r, su(this.idTensor)
+                }
+                get id() {
+                    return this.idTensor.id
+                }
+                copy() {
+                    return new lA([...this.tensors], this.elementShape, this.elementDtype)
+                }
+                clearAndClose(e) {
+                    this.tensors.forEach(t => {
+                        null != e && e.has(t.id) || t.dispose()
+                    }), this.tensors.length = 0, this.idTensor.dispose()
+                }
+                size() {
+                    return this.tensors.length
+                }
+                stack(e, t, n = -1) {
+                    if (t !== this.elementDtype) throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);
+                    if (-1 !== n && this.tensors.length !== n) throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);
+                    sA(e, this.elementShape, "TensorList shape mismatch: ");
+                    const r = oA(this.elementShape, this.tensors, e);
+                    return ru(() => {
+                        const e = this.tensors.map(e => qu(e, r));
+                        return Gh(e, 0)
+                    })
+                }
+                popBack(e, t) {
+                    if (t !== this.elementDtype) throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);
+                    if (0 === this.size()) throw new Error("Trying to pop from an empty list.");
+                    const n = oA(this.elementShape, this.tensors, e),
+                        r = this.tensors.pop();
+                    return sA(r.shape, e, "TensorList shape mismatch: "), qu(r, n)
+                }
+                pushBack(e) {
+                    if (e.dtype !== this.elementDtype) throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);
+                    if (sA(e.shape, this.elementShape, "TensorList shape mismatch: "), this.maxNumElements === this.size()) throw new Error("Trying to push element into a full list.");
+                    su(e), this.tensors.push(e)
+                }
+                resize(e) {
+                    if (e < 0) throw new Error("TensorListResize expects size to be non-negative. Got: " + e);
+                    if (-1 !== this.maxNumElements && e > this.maxNumElements) throw new Error(`TensorListResize input size ${e} is greater maxNumElement ${this.maxNumElements}.`);
+                    this.tensors.length = e
+                }
+                getItem(e, t, n) {
+                    if (n !== this.elementDtype) throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);
+                    if (e < 0 || e > this.tensors.length) throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);
+                    if (null == this.tensors[e]) throw new Error(`element at index ${e} is null.`);
+                    sA(this.tensors[e].shape, t, "TensorList shape mismatch: ");
+                    const r = oA(this.elementShape, this.tensors, t);
+                    return qu(this.tensors[e], r)
+                }
+                setItem(e, t) {
+                    if (t.dtype !== this.elementDtype) throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);
+                    if (e < 0 || -1 !== this.maxNumElements && e >= this.maxNumElements) throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);
+                    sA(this.elementShape, t.shape, "TensorList shape mismatch: "), su(t), this.tensors[e] = t
+                }
+                gather(e, t, n) {
+                    if (t !== this.elementDtype) throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);
+                    sA(this.elementShape, n, "TensorList shape mismatch: "), e = e.slice(0, this.size());
+                    const r = oA(this.elementShape, this.tensors, n);
+                    return 0 === e.length ? Ls([], [0].concat(r)) : ru(() => {
+                        const t = e.map(e => qu(this.tensors[e], r));
+                        return Gh(t, 0)
+                    })
+                }
+                concat(e, t) {
+                    if (e && e !== this.elementDtype) throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);
+                    sA(this.elementShape, t, "TensorList shape mismatch: ");
+                    const n = oA(this.elementShape, this.tensors, t);
+                    return 0 === this.size() ? Ls([], [0].concat(n)) : ru(() => {
+                        const e = this.tensors.map(e => qu(e, n));
+                        return Xu(e, 0)
+                    })
+                }
+            };
+            var cA = async (e, t, n) => {
+                switch (e.op) {
+                    case "If":
+                    case "StatelessIf": {
+                        const r = JS("thenBranch", e, t, n),
+                            a = JS("elseBranch", e, t, n),
+                            s = JS("cond", e, t, n),
+                            i = JS("args", e, t, n);
+                        return (await s.data())[0] ? n.functionMap[r].executeFunctionAsync(i, n.tensorArrayMap, n.tensorListMap) : n.functionMap[a].executeFunctionAsync(i, n.tensorArrayMap, n.tensorListMap)
+                    }
+                    case "While":
+                    case "StatelessWhile": {
+                        const r = JS("body", e, t, n),
+                            a = JS("cond", e, t, n),
+                            s = JS("args", e, t, n),
+                            i = await n.functionMap[a].executeFunctionAsync(s, n.tensorArrayMap, n.tensorListMap),
+                            o = s.map(e => e.id);
+                        let u = await i[0].data();
+                        i.forEach(e => {
+                            e.kept || -1 !== o.indexOf(e.id) || e.dispose()
+                        });
+                        let l = s;
+                        for (; u[0];) {
+                            const e = l;
+                            l = await n.functionMap[r].executeFunctionAsync(l, n.tensorArrayMap, n.tensorListMap);
+                            const t = l.map(e => e.id);
+                            e.forEach(e => {
+                                e.kept || -1 !== o.indexOf(e.id) || -1 !== t.indexOf(e.id) || e.dispose()
+                            });
+                            const s = await n.functionMap[a].executeFunctionAsync(l, n.tensorArrayMap, n.tensorListMap);
+                            u = await s[0].data(), s.forEach(e => {
+                                e.kept || -1 !== o.indexOf(e.id) || -1 !== t.indexOf(e.id) || e.dispose()
+                            })
+                        }
+                        return l
+                    }
+                    case "LoopCond":
+                        return [sI(JS("pred", e, t, n))];
+                    case "Switch": {
+                        const r = JS("pred", e, t, n);
+                        let a = JS("data", e, t, n);
+                        return a.kept || (a = sI(a)), (await r.data())[0] ? [void 0, a] : [a, void 0]
+                    }
+                    case "Merge": {
+                        const r = e.inputNames.find(e => void 0 !== eI(e, t, n));
+                        if (r) {
+                            return [sI(eI(r, t, n))]
+                        }
+                        return
+                    }
+                    case "Enter": {
+                        const r = JS("frameName", e, t, n),
+                            a = JS("tensor", e, t, n);
+                        return n.enterFrame(r), [sI(a)]
+                    }
+                    case "Exit": {
+                        const r = JS("tensor", e, t, n);
+                        return n.exitFrame(), [sI(r)]
+                    }
+                    case "NextIteration": {
+                        const r = JS("tensor", e, t, n);
+                        return n.nextIteration(), [sI(r)]
+                    }
+                    case "TensorArrayV3": {
+                        const r = JS("size", e, t, n),
+                            a = JS("dtype", e, t, n),
+                            s = JS("elementShape", e, t, n),
+                            i = JS("dynamicSize", e, t, n),
+                            o = JS("clearAfterRead", e, t, n),
+                            u = JS("identicalElementShapes", e, t, n),
+                            l = JS("name", e, t, n),
+                            c = new class {
+                                constructor(e, t, n, r, a, s, i) {
+                                    this.name = e, this.dtype = t, this.maxSize = n, this.elementShape = r, this.identicalElementShapes = a, this.dynamicSize = s, this.clearAfterRead = i, this.tensors = [], this.closed_ = !1, this.idTensor = Ah(0), su(this.idTensor)
+                                }
+                                get id() {
+                                    return this.idTensor.id
+                                }
+                                get closed() {
+                                    return this.closed_
+                                }
+                                clearAndClose(e) {
+                                    this.tensors.forEach(t => {
+                                        null != e && e.has(t.tensor.id) || t.tensor.dispose()
+                                    }), this.tensors = [], this.closed_ = !0, this.idTensor.dispose()
+                                }
+                                size() {
+                                    return this.tensors.length
+                                }
+                                read(e) {
+                                    if (this.closed_) throw new Error(`TensorArray ${this.name} has already been closed.`);
+                                    if (e < 0 || e >= this.size()) throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);
+                                    const t = this.tensors[e];
+                                    if (t.cleared) throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);
+                                    return this.clearAfterRead && (t.cleared = !0), t.read = !0, t.tensor
+                                }
+                                readMany(e) {
+                                    return e.map(e => this.read(e))
+                                }
+                                write(e, t) {
+                                    if (this.closed_) throw new Error(`TensorArray ${this.name} has already been closed.`);
+                                    if (e < 0 || !this.dynamicSize && e >= this.maxSize) throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);
+                                    const n = this.tensors[e] || {};
+                                    if (t.dtype !== this.dtype) throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},\n          because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);
+                                    if (0 !== this.size() || null != this.elementShape && 0 !== this.elementShape.length || (this.elementShape = t.shape), sA(this.elementShape, t.shape, `TensorArray ${this.name}: Could not write to TensorArray index ${e}.`), n.read) throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);
+                                    if (n.written) throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);
+                                    n.tensor = t, su(t), n.written = !0, this.tensors[e] = n
+                                }
+                                writeMany(e, t) {
+                                    if (e.length !== t.length) throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);
+                                    e.forEach((e, n) => this.write(e, t[n]))
+                                }
+                                gather(e, t) {
+                                    if (t && t !== this.dtype) throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);
+                                    if (e) e = e.slice(0, this.size());
+                                    else {
+                                        e = [];
+                                        for (let t = 0; t < this.size(); t++) e.push(t)
+                                    }
+                                    if (0 === e.length) return Ls([], [0].concat(this.elementShape));
+                                    const n = this.readMany(e);
+                                    return sA(this.elementShape, n[0].shape, "TensorArray shape mismatch: "), Gh(n, 0)
+                                }
+                                concat(e) {
+                                    if (e && e !== this.dtype) throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);
+                                    if (0 === this.size()) return Ls([], [0].concat(this.elementShape));
+                                    const t = [];
+                                    for (let e = 0; e < this.size(); e++) t.push(e);
+                                    const n = this.readMany(t);
+                                    return sA(this.elementShape, n[0].shape, `TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`), Xu(n, 0)
+                                }
+                                scatter(e, t) {
+                                    if (t.dtype !== this.dtype) throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);
+                                    if (e.length !== t.shape[0]) throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);
+                                    const n = Math.max(...e);
+                                    if (!this.dynamicSize && n >= this.maxSize) throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);
+                                    this.writeMany(e, ap(t, 0))
+                                }
+                                split(e, t) {
+                                    if (t.dtype !== this.dtype) throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);
+                                    let n = 0;
+                                    const r = e.map(e => (n += e, n));
+                                    if (n !== t.shape[0]) throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${n}, and tensor's shape is: ${t.shape}`);
+                                    if (!this.dynamicSize && e.length !== this.maxSize) throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);
+                                    const a = 0 === n ? 0 : t.size / n,
+                                        s = [];
+                                    ru(() => {
+                                        t = qu(t, [1, n, a]);
+                                        for (let n = 0; n < e.length; ++n) {
+                                            const i = [0, 0 === n ? 0 : r[n - 1], 0],
+                                                o = [1, e[n], a];
+                                            s[n] = qu(Zu(t, i, o), this.elementShape)
+                                        }
+                                        return s
+                                    });
+                                    const i = [];
+                                    for (let t = 0; t < e.length; t++) i[t] = t;
+                                    this.writeMany(i, s)
+                                }
+                            }(l, a, r, s, u, i, o);
+                        return n.addTensorArray(c), [c.idTensor, Ah(1)]
+                    }
+                    case "TensorArrayWriteV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = JS("index", e, t, n),
+                            s = JS("tensor", e, t, n),
+                            i = n.getTensorArray(r.id);
+                        return i.write(a, s), [i.idTensor]
+                    }
+                    case "TensorArrayReadV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = JS("index", e, t, n);
+                        return [n.getTensorArray(r.id).read(a)]
+                    }
+                    case "TensorArrayGatherV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = JS("indices", e, t, n),
+                            s = JS("dtype", e, t, n);
+                        return [n.getTensorArray(r.id).gather(a, s)]
+                    }
+                    case "TensorArrayScatterV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = JS("indices", e, t, n),
+                            s = JS("tensor", e, t, n),
+                            i = n.getTensorArray(r.id);
+                        return i.scatter(a, s), [i.idTensor]
+                    }
+                    case "TensorArrayConcatV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = n.getTensorArray(r.id),
+                            s = JS("dtype", e, t, n);
+                        return [a.concat(s)]
+                    }
+                    case "TensorArraySplitV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = JS("tensor", e, t, n),
+                            s = JS("lengths", e, t, n),
+                            i = n.getTensorArray(r.id);
+                        return i.split(s, a), [i.idTensor]
+                    }
+                    case "TensorArraySizeV3": {
+                        const r = JS("tensorArrayId", e, t, n);
+                        return [Ah(n.getTensorArray(r.id).size(), "int32")]
+                    }
+                    case "TensorArrayCloseV3": {
+                        const r = JS("tensorArrayId", e, t, n),
+                            a = n.getTensorArray(r.id);
+                        return a.clearAndClose(), [a.idTensor]
+                    }
+                    case "TensorListSetItem": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = JS("index", e, t, n),
+                            s = JS("tensor", e, t, n),
+                            i = n.getTensorList(r.id);
+                        return i.setItem(a, s), [i.idTensor]
+                    }
+                    case "TensorListGetItem": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = JS("index", e, t, n),
+                            s = JS("elementShape", e, t, n),
+                            i = JS("elementDType", e, t, n);
+                        return [n.getTensorList(r.id).getItem(a, s, i)]
+                    }
+                    case "TensorListScatterV2":
+                    case "TensorListScatter": {
+                        const r = JS("indices", e, t, n),
+                            a = function (e, t, n, r) {
+                                if (t.length !== e.shape[0]) throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);
+                                const a = Math.max(...t);
+                                if (null != r && -1 !== r && a >= r) throw new Error(`Max index must be < array size (${a}  vs. ${r})`);
+                                const s = new lA([], n, e.dtype, r),
+                                    i = ap(e, 0);
+                                return t.forEach((e, t) => {
+                                    s.setItem(e, i[t])
+                                }), s
+                            }(JS("tensor", e, t, n), r, JS("elementShape", e, t, n), JS("numElements", e, t, n));
+                        return n.addTensorList(a), [a.idTensor]
+                    }
+                    case "TensorListReserve":
+                    case "EmptyTensorList": {
+                        const r = JS("elementShape", e, t, n),
+                            a = JS("elementDType", e, t, n);
+                        let s;
+                        s = "TensorListReserve" === e.op ? "numElements" : "maxNumElements";
+                        const i = function (e, t, n) {
+                            return new lA([], e, t, n)
+                        }(r, a, JS(s, e, t, n));
+                        return n.addTensorList(i), [i.idTensor]
+                    }
+                    case "TensorListGather": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = JS("indices", e, t, n),
+                            s = JS("elementShape", e, t, n),
+                            i = JS("elementDType", e, t, n);
+                        return [n.getTensorList(r.id).gather(a, i, s)]
+                    }
+                    case "TensorListStack": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = JS("elementShape", e, t, n),
+                            s = JS("elementDType", e, t, n),
+                            i = JS("numElements", e, t, n);
+                        return [n.getTensorList(r.id).stack(a, s, i)]
+                    }
+                    case "TensorListFromTensor": {
+                        const r = function (e, t, n) {
+                            const r = e.dtype;
+                            if (e.shape.length < 1) throw new Error("Tensor must be at least a vector, but saw shape: " + e.shape);
+                            if (e.dtype !== n) throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${n}`);
+                            sA(e.shape.slice(1), t, "TensorList shape mismatch: ");
+                            const a = ap(e);
+                            return new lA(a, t, r)
+                        }(JS("tensor", e, t, n), JS("elementShape", e, t, n), JS("elementDType", e, t, n));
+                        return n.addTensorList(r), [r.idTensor]
+                    }
+                    case "TensorListConcat": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = n.getTensorList(r.id),
+                            s = JS("dtype", e, t, n),
+                            i = JS("elementShape", e, t, n);
+                        return [a.concat(s, i)]
+                    }
+                    case "TensorListPushBack": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = JS("tensor", e, t, n),
+                            s = n.getTensorList(r.id);
+                        return s.pushBack(a), [s.idTensor]
+                    }
+                    case "TensorListPopBack": {
+                        const r = JS("tensorListId", e, t, n),
+                            a = JS("elementShape", e, t, n),
+                            s = JS("elementDType", e, t, n);
+                        return [n.getTensorList(r.id).popBack(a, s)]
+                    }
+                    case "TensorListSplit": {
+                        const r = JS("tensor", e, t, n),
+                            a = JS("elementShape", e, t, n),
+                            s = function (e, t, n) {
+                                let r = 0;
+                                const a = t.map(e => (r += e, r));
+                                if (r !== e.shape[0]) throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${r}, and tensor's shape is: ${e.shape}`);
+                                const s = uA(e.shape.slice(1), n),
+                                    i = 0 === r ? 0 : e.size / r,
+                                    o = ru(() => {
+                                        const n = [];
+                                        e = qu(e, [1, r, i]);
+                                        for (let r = 0; r < t.length; ++r) {
+                                            const o = [0, 0 === r ? 0 : a[r - 1], 0],
+                                                u = [1, t[r], i];
+                                            n[r] = qu(Zu(e, o, u), s)
+                                        }
+                                        return e.dispose(), n
+                                    }),
+                                    u = new lA([], n, e.dtype, t.length);
+                                for (let e = 0; e < o.length; e++) u.setItem(e, o[e]);
+                                return u
+                            }(r, JS("lengths", e, t, n), a);
+                        return n.addTensorList(s), [s.idTensor]
+                    }
+                    default:
+                        throw TypeError(`Node type ${e.op} is not implemented`)
+                }
+            };
+
+            function hA(e, t, n) {
+                const [r, a] = JS("fusedOps", e, t, n), s = "biasadd" === r, i = !s, o = "prelu" === a, u = "fusedbatchnorm" === r, l = JS("numArgs", e, t, n);
+                if (s) {
+                    if (o && 2 !== l) throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");
+                    if (!o && s && 1 !== l) throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")
+                }
+                if (u) throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");
+                const c = JS("strides", e, t, n),
+                    h = aI(e, t, n),
+                    p = JS("dataFormat", e, t, n).toUpperCase(),
+                    d = JS("dilations", e, t, n);
+                let [f, m] = JS("args", e, t, n);
+                i && (m = f, f = void 0);
+                return {
+                    stride: c,
+                    pad: h,
+                    dataFormat: p,
+                    dilations: d,
+                    biasArg: f,
+                    preluArg: m,
+                    activationFunc: a,
+                    leakyreluAlpha: JS("leakyreluAlpha", e, t, n)
+                }
+            }
+
+            function pA(e, t, n) {
+                return {
+                    boxes: JS("boxes", e, t, n),
+                    scores: JS("scores", e, t, n),
+                    maxOutputSize: JS("maxOutputSize", e, t, n),
+                    iouThreshold: JS("iouThreshold", e, t, n),
+                    scoreThreshold: JS("scoreThreshold", e, t, n),
+                    softNmsSigma: JS("softNmsSigma", e, t, n)
+                }
+            }
+            var dA = async (e, t, n, r) => {
+                switch (e.op) {
+                    case "HashTable":
+                    case "HashTableV2": {
+                        const a = JS("keyDType", e, t, n),
+                            s = JS("valueDType", e, t, n),
+                            i = new class {
+                                constructor(e, t) {
+                                    this.keyDType = e, this.valueDType = t, this.handle = Ah(0), this.tensorMap = new Map, su(this.handle)
+                                }
+                                get id() {
+                                    return this.handle.id
+                                }
+                                clearAndClose() {
+                                    this.tensorMap.forEach(e => e.dispose()), this.tensorMap.clear(), this.handle.dispose()
+                                }
+                                size() {
+                                    return this.tensorMap.size
+                                }
+                                tensorSize() {
+                                    return Ah(this.size(), "int32")
+                                }
+                                async import(e, t) {
+                                    this.checkKeyAndValueTensor(e, t);
+                                    const n = await e.data();
+                                    return this.tensorMap.forEach(e => e.dispose()), this.tensorMap.clear(), ru(() => {
+                                        const e = ap(t),
+                                            r = n.length,
+                                            a = e.length;
+                                        va.assert(r === a, () => `The number of elements doesn't match, keys has ${r} elements, the values has ${a} elements.`);
+                                        for (let t = 0; t < r; t++) {
+                                            const r = n[t],
+                                                a = e[t];
+                                            su(a), this.tensorMap.set(r, a)
+                                        }
+                                        return this.handle
+                                    })
+                                }
+                                async find(e, t) {
+                                    this.checkKeyAndValueTensor(e, t);
+                                    const n = await e.data();
+                                    return ru(() => {
+                                        const e = [];
+                                        for (let r = 0; r < n.length; r++) {
+                                            const a = n[r],
+                                                s = this.findWithDefault(a, t);
+                                            e.push(s)
+                                        }
+                                        return Gh(e)
+                                    })
+                                }
+                                findWithDefault(e, t) {
+                                    const n = this.tensorMap.get(e);
+                                    return null != n ? n : t
+                                }
+                                checkKeyAndValueTensor(e, t) {
+                                    if (e.dtype !== this.keyDType) throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);
+                                    if (t.dtype !== this.valueDType) throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)
+                                }
+                            }(a, s);
+                        return r.addHashTable(e.name, i), [i.handle]
+                    }
+                    case "LookupTableImport":
+                    case "LookupTableImportV2": {
+                        const a = JS("tableHandle", e, t, n, r),
+                            s = JS("keys", e, t, n),
+                            i = JS("values", e, t, n),
+                            o = r.getHashTableById(a.id);
+                        return [await o.import(s, i)]
+                    }
+                    case "LookupTableFind":
+                    case "LookupTableFindV2": {
+                        const a = JS("tableHandle", e, t, n, r),
+                            s = JS("keys", e, t, n),
+                            i = JS("defaultValue", e, t, n),
+                            o = r.getHashTableById(a.id);
+                        return [await o.find(s, i)]
+                    }
+                    case "LookupTableSize":
+                    case "LookupTableSizeV2": {
+                        const a = JS("tableHandle", e, t, n, r);
+                        return [r.getHashTableById(a.id).tensorSize()]
+                    }
+                    default:
+                        throw TypeError(`Node type ${e.op} is not implemented`)
+                }
+            };
+
+            function fA(e, t, n, r) {
+                const a = ((e, t, n) => {
+                    switch (e.category) {
+                        case "arithmetic":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "BiasAdd":
+                                    case "AddV2":
+                                    case "Add":
+                                        return [gu(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "AddN":
+                                        return [Su(JS("tensors", e, t, n))];
+                                    case "FloorMod":
+                                    case "Mod":
+                                        return [Vc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Mul":
+                                        return [vu(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "RealDiv":
+                                    case "Div":
+                                        return [bu(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "DivNoNan":
+                                        return [Ml(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "FloorDiv":
+                                        return [yu(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Sub":
+                                        return [gc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Minimum":
+                                        return [Uc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Maximum":
+                                        return [Oc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Pow":
+                                        return [ah(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "SquaredDifference":
+                                        return [Vh(JS("a", e, t, n), JS("b", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "basic_math":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Abs":
+                                    case "ComplexAbs":
+                                        return [xu(JS("x", e, t, n))];
+                                    case "Acos":
+                                        return [wu(JS("x", e, t, n))];
+                                    case "Acosh":
+                                        return [ku(JS("x", e, t, n))];
+                                    case "Asin":
+                                        return [Cu(JS("x", e, t, n))];
+                                    case "Asinh":
+                                        return [Tu(JS("x", e, t, n))];
+                                    case "Atan":
+                                        return [Ru(JS("x", e, t, n))];
+                                    case "Atan2":
+                                        return [_u(JS("x", e, t, n), JS("y", e, t, n))];
+                                    case "Atanh":
+                                        return [Fu(JS("x", e, t, n))];
+                                    case "Ceil":
+                                        return [ll(JS("x", e, t, n))];
+                                    case "Complex":
+                                        return [Ds(JS("real", e, t, n), JS("imag", e, t, n))];
+                                    case "Cos":
+                                        return [kl(JS("x", e, t, n))];
+                                    case "Cosh":
+                                        return [Sl(JS("x", e, t, n))];
+                                    case "Elu":
+                                        return [Ll(JS("x", e, t, n))];
+                                    case "Erf":
+                                        return [Pl(JS("x", e, t, n))];
+                                    case "Exp":
+                                        return [$l(JS("x", e, t, n))];
+                                    case "Expm1":
+                                        return [zl(JS("x", e, t, n))];
+                                    case "Floor":
+                                        return [jl(JS("x", e, t, n))];
+                                    case "Log":
+                                        return [rc(JS("x", e, t, n))];
+                                    case "Log1p":
+                                        return [ac(JS("x", e, t, n))];
+                                    case "Imag":
+                                        return [Kl(JS("x", e, t, n))];
+                                    case "Neg":
+                                        return [pc(JS("x", e, t, n))];
+                                    case "Reciprocal":
+                                        return [mh(JS("x", e, t, n))];
+                                    case "Real":
+                                        return [fh(JS("x", e, t, n))];
+                                    case "Relu":
+                                        return [gh(JS("x", e, t, n))];
+                                    case "Round":
+                                        return [Sh(JS("x", e, t, n))];
+                                    case "Selu":
+                                        return [Eh(JS("x", e, t, n))];
+                                    case "Sigmoid":
+                                        return [Qu(JS("x", e, t, n))];
+                                    case "Sin":
+                                        return [Rh(JS("x", e, t, n))];
+                                    case "Sign":
+                                        return [Th(JS("x", e, t, n))];
+                                    case "Sinh":
+                                        return [_h(JS("x", e, t, n))];
+                                    case "Softplus":
+                                        return [dc(JS("x", e, t, n))];
+                                    case "Sqrt":
+                                        return [Wh(JS("x", e, t, n))];
+                                    case "Square":
+                                        return [jc(JS("x", e, t, n))];
+                                    case "Tanh":
+                                        return [Ju(JS("x", e, t, n))];
+                                    case "Tan":
+                                        return [Kh(JS("x", e, t, n))];
+                                    case "ClipByValue":
+                                        return [cl(JS("x", e, t, n), JS("clipValueMin", e, t, n), JS("clipValueMax", e, t, n))];
+                                    case "Relu6":
+                                        return [yh(JS("x", e, t, n))];
+                                    case "Rsqrt":
+                                        return [Ih(eI(e.inputNames[0], t, n))];
+                                    case "Prod":
+                                        return [ih(JS("x", e, t, n), JS("axes", e, t, n))];
+                                    case "LeakyRelu":
+                                        return [Zl(JS("x", e, t, n), JS("alpha", e, t, n))];
+                                    case "Prelu":
+                                        return [sh(JS("x", e, t, n), JS("alpha", e, t, n))];
+                                    case "IsNan":
+                                        return [Ql(eI(e.inputNames[0], t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "control":
+                            return cA(e, t, n);
+                        case "convolution":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Conv1D": {
+                                        const r = JS("stride", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("dataFormat", e, t, n).toUpperCase(),
+                                            i = JS("dilation", e, t, n);
+                                        return [gl(JS("x", e, t, n), JS("filter", e, t, n), r, a, s, i)]
+                                    }
+                                    case "Conv2D": {
+                                        const r = JS("strides", e, t, n),
+                                            a = aI(e, t, n),
+                                            s = JS("dataFormat", e, t, n).toUpperCase(),
+                                            i = JS("dilations", e, t, n);
+                                        return [ml(JS("x", e, t, n), JS("filter", e, t, n), [r[1], r[2]], a, s, [i[1], i[2]])]
+                                    }
+                                    case "_FusedConv2D": {
+                                        const {
+                                            stride: r,
+                                            pad: a,
+                                            dataFormat: s,
+                                            dilations: i,
+                                            biasArg: o,
+                                            preluArg: u,
+                                            activationFunc: l,
+                                            leakyreluAlpha: c
+                                        } = hA(e, t, n);
+                                        return [bp.conv2d({
+                                            x: JS("x", e, t, n),
+                                            filter: JS("filter", e, t, n),
+                                            strides: [r[1], r[2]],
+                                            pad: a,
+                                            dataFormat: s,
+                                            dilations: [i[1], i[2]],
+                                            bias: o,
+                                            activation: l,
+                                            preluActivationWeights: u,
+                                            leakyreluAlpha: c
+                                        })]
+                                    }
+                                    case "FusedDepthwiseConv2dNative": {
+                                        const {
+                                            stride: r,
+                                            pad: a,
+                                            dataFormat: s,
+                                            dilations: i,
+                                            biasArg: o,
+                                            preluArg: u,
+                                            activationFunc: l,
+                                            leakyreluAlpha: c
+                                        } = hA(e, t, n);
+                                        return [bp.depthwiseConv2d({
+                                            x: JS("x", e, t, n),
+                                            filter: JS("filter", e, t, n),
+                                            strides: [r[1], r[2]],
+                                            pad: a,
+                                            dataFormat: s,
+                                            dilations: [i[1], i[2]],
+                                            bias: o,
+                                            activation: l,
+                                            preluActivationWeights: u,
+                                            leakyreluAlpha: c
+                                        })]
+                                    }
+                                    case "Conv2DBackpropInput":
+                                    case "Conv2dTranspose": {
+                                        const r = JS("outputShape", e, t, n),
+                                            a = JS("strides", e, t, n),
+                                            s = aI(e, t, n);
+                                        return [bl(JS("x", e, t, n), JS("filter", e, t, n), r, [a[1], a[2]], s)]
+                                    }
+                                    case "DepthwiseConv2dNative":
+                                    case "DepthwiseConv2d": {
+                                        const r = JS("strides", e, t, n),
+                                            a = aI(e, t, n),
+                                            s = JS("dilations", e, t, n),
+                                            i = JS("dataFormat", e, t, n).toUpperCase();
+                                        return [Nl(JS("input", e, t, n), JS("filter", e, t, n), [r[1], r[2]], a, i, [s[1], s[2]])]
+                                    }
+                                    case "Conv3D": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("dataFormat", e, t, n).toUpperCase(),
+                                            i = JS("dilations", e, t, n);
+                                        return [vl(JS("x", e, t, n), JS("filter", e, t, n), [r[1], r[2], r[3]], a, s, [i[1], i[2], i[3]])]
+                                    }
+                                    case "AvgPool": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("kernelSize", e, t, n);
+                                        return [Ku(JS("x", e, t, n), [s[1], s[2]], [r[1], r[2]], a)]
+                                    }
+                                    case "MaxPool": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("kernelSize", e, t, n);
+                                        return [Fc(JS("x", e, t, n), [s[1], s[2]], [r[1], r[2]], a)]
+                                    }
+                                    case "MaxPoolWithArgmax": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("kernelSize", e, t, n),
+                                            i = JS("includeBatchInIndex", e, t, n),
+                                            {
+                                                result: o,
+                                                indexes: u
+                                            } = Dc(JS("x", e, t, n), [s[1], s[2]], [r[1], r[2]], a, i);
+                                        return [o, u]
+                                    }
+                                    case "AvgPool3D": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("kernelSize", e, t, n);
+                                        return [Yu(JS("x", e, t, n), [s[1], s[2], s[3]], [r[1], r[2], r[3]], a)]
+                                    }
+                                    case "MaxPool3D": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("kernelSize", e, t, n);
+                                        return [Mc(JS("x", e, t, n), [s[1], s[2], s[3]], [r[1], r[2], r[3]], a)]
+                                    }
+                                    case "Dilation2D": {
+                                        const r = JS("strides", e, t, n),
+                                            a = JS("pad", e, t, n),
+                                            s = JS("dilations", e, t, n),
+                                            i = r[1],
+                                            o = r[2],
+                                            u = s[1],
+                                            l = s[2];
+                                        return [Tl(JS("x", e, t, n), JS("filter", e, t, n), [i, o], a, [u, l], "NHWC")]
+                                    }
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "creation":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Fill": {
+                                        const r = JS("shape", e, t, n),
+                                            a = JS("dtype", e, t, n);
+                                        return [Vl(r, JS("value", e, t, n), a)]
+                                    }
+                                    case "LinSpace":
+                                        return [tc(JS("start", e, t, n), JS("stop", e, t, n), JS("num", e, t, n))];
+                                    case "Multinomial": {
+                                        const r = JS("logits", e, t, n),
+                                            a = JS("numSamples", e, t, n),
+                                            s = JS("seed", e, t, n);
+                                        return [qc(r, a, s)]
+                                    }
+                                    case "OneHot": {
+                                        const r = JS("indices", e, t, n),
+                                            a = JS("depth", e, t, n),
+                                            s = JS("onValue", e, t, n),
+                                            i = JS("offValue", e, t, n);
+                                        return [Hi(r, a, s, i)]
+                                    }
+                                    case "Ones":
+                                        return [$c(JS("shape", e, t, n), JS("dtype", e, t, n))];
+                                    case "OnesLike":
+                                        return [Yc(JS("x", e, t, n))];
+                                    case "RandomUniform":
+                                        return [ph(JS("shape", e, t, n), JS("minval", e, t, n), JS("maxval", e, t, n), JS("dtype", e, t, n))];
+                                    case "Range":
+                                        return [dh(JS("start", e, t, n), JS("stop", e, t, n), JS("step", e, t, n), JS("dtype", e, t, n))];
+                                    case "TruncatedNormal": {
+                                        const r = JS("shape", e, t, n),
+                                            a = JS("mean", e, t, n),
+                                            s = JS("stdDev", e, t, n),
+                                            i = JS("seed", e, t, n);
+                                        return [tp(r, a, s, JS("dtype", e, t, n), i)]
+                                    }
+                                    case "Zeros":
+                                        return [Pc(JS("shape", e, t, n), JS("dtype", e, t, n))];
+                                    case "ZerosLike":
+                                        return [Fl(JS("x", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "dynamic":
+                            return (async (e, t, n) => {
+                                switch (e.op) {
+                                    case "NonMaxSuppressionV5": {
+                                        const {
+                                            boxes: r,
+                                            scores: a,
+                                            maxOutputSize: s,
+                                            iouThreshold: i,
+                                            scoreThreshold: o,
+                                            softNmsSigma: u
+                                        } = pA(e, t, n), l = await Nd.nonMaxSuppressionWithScoreAsync(r, a, s, i, o, u);
+                                        return [l.selectedIndices, l.selectedScores]
+                                    }
+                                    case "NonMaxSuppressionV4": {
+                                        const {
+                                            boxes: r,
+                                            scores: a,
+                                            maxOutputSize: s,
+                                            iouThreshold: i,
+                                            scoreThreshold: o
+                                        } = pA(e, t, n), u = JS("padToMaxOutputSize", e, t, n), l = await Nd.nonMaxSuppressionPaddedAsync(r, a, s, i, o, u);
+                                        return [l.selectedIndices, l.validOutputs]
+                                    }
+                                    case "NonMaxSuppressionV3":
+                                    case "NonMaxSuppressionV2": {
+                                        const {
+                                            boxes: r,
+                                            scores: a,
+                                            maxOutputSize: s,
+                                            iouThreshold: i,
+                                            scoreThreshold: o
+                                        } = pA(e, t, n);
+                                        return [await Nd.nonMaxSuppressionAsync(r, a, s, i, o)]
+                                    }
+                                    case "Where": {
+                                        const r = Ai(JS("condition", e, t, n), "bool"),
+                                            a = [await op(r)];
+                                        return r.dispose(), a
+                                    }
+                                    case "ListDiff":
+                                        return Ch(JS("x", e, t, n), JS("y", e, t, n));
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n);
+                        case "evaluation":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "TopKV2": {
+                                        const r = JS("x", e, t, n),
+                                            a = JS("k", e, t, n),
+                                            s = JS("sorted", e, t, n),
+                                            i = ep(r, a, s);
+                                        return [i.values, i.indices]
+                                    }
+                                    case "Unique": {
+                                        const r = JS("x", e, t, n),
+                                            a = np(r);
+                                        return [a.values, a.indices]
+                                    }
+                                    case "UniqueV2": {
+                                        const r = JS("x", e, t, n),
+                                            a = JS("axis", e, t, n),
+                                            s = np(r, a);
+                                        return [s.values, s.indices]
+                                    }
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "image":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "ResizeBilinear": {
+                                        const r = JS("images", e, t, n),
+                                            a = JS("size", e, t, n),
+                                            s = JS("alignCorners", e, t, n),
+                                            i = JS("halfPixelCenters", e, t, n);
+                                        return [Nd.resizeBilinear(r, [a[0], a[1]], s, i)]
+                                    }
+                                    case "ResizeNearestNeighbor": {
+                                        const r = JS("images", e, t, n),
+                                            a = JS("size", e, t, n),
+                                            s = JS("alignCorners", e, t, n),
+                                            i = JS("halfPixelCenters", e, t, n);
+                                        return [Nd.resizeNearestNeighbor(r, [a[0], a[1]], s, i)]
+                                    }
+                                    case "CropAndResize": {
+                                        const r = JS("image", e, t, n),
+                                            a = JS("boxes", e, t, n),
+                                            s = JS("boxInd", e, t, n),
+                                            i = JS("cropSize", e, t, n),
+                                            o = JS("method", e, t, n),
+                                            u = JS("extrapolationValue", e, t, n);
+                                        return [Nd.cropAndResize(r, a, s, i, o, u)]
+                                    }
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "graph":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Const":
+                                        return t[e.name];
+                                    case "PlaceholderWithDefault":
+                                        const r = JS("default", e, t, n);
+                                        return [eI(e.name, t, n) || r];
+                                    case "Placeholder":
+                                        return [eI(e.name, t, n)];
+                                    case "Identity":
+                                    case "StopGradient":
+                                    case "FakeQuantWithMinMaxVars":
+                                        return [sI(JS("x", e, t, n))];
+                                    case "IdentityN":
+                                        return JS("x", e, t, n).map(e => sI(e));
+                                    case "Snapshot":
+                                        return [sI(JS("x", e, t, n))];
+                                    case "Shape":
+                                        return [Yh(JS("x", e, t, n).shape, "int32")];
+                                    case "ShapeN":
+                                        return JS("x", e, t, n).map(e => Yh(e.shape));
+                                    case "Size":
+                                        return [Ah(JS("x", e, t, n).size, "int32")];
+                                    case "Rank":
+                                        return [Ah(JS("x", e, t, n).rank, "int32")];
+                                    case "NoOp":
+                                        return [Ah(1)];
+                                    case "Print":
+                                        const a = JS("x", e, t, n),
+                                            s = JS("data", e, t, n),
+                                            i = JS("message", e, t, n),
+                                            o = JS("summarize", e, t, n);
+                                        console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."), console.log(i);
+                                        for (let e = 0; e < s.length; e++) console.log(Array.prototype.slice.call(s[e].dataSync()).slice(0, o));
+                                        return [a];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "logical":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Equal":
+                                        return [Rl(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "NotEqual":
+                                        return [Kc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Greater":
+                                        return [Hl(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "GreaterEqual":
+                                        return [ql(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Less":
+                                        return [Jl(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "LessEqual":
+                                        return [ec(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "LogicalAnd":
+                                        return [Cc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "LogicalNot":
+                                        return [Tc(JS("a", e, t, n))];
+                                    case "LogicalOr":
+                                        return [Rc(JS("a", e, t, n), JS("b", e, t, n))];
+                                    case "Select":
+                                    case "SelectV2":
+                                        return [_l(JS("condition", e, t, n), JS("a", e, t, n), JS("b", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "matrices":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "BatchMatMul":
+                                    case "BatchMatMulV2":
+                                    case "MatMul":
+                                        return [Gi(JS("a", e, t, n), JS("b", e, t, n), JS("transposeA", e, t, n), JS("transposeB", e, t, n))];
+                                    case "Einsum":
+                                        return [Ol(JS("equation", e, t, n), ...JS("tensors", e, t, n))];
+                                    case "Transpose":
+                                        return [qi(JS("x", e, t, n), JS("perm", e, t, n))];
+                                    case "_FusedMatMul":
+                                        const [r, a] = JS("fusedOps", e, t, n), s = "biasadd" === r, i = "prelu" === a, o = JS("numArgs", e, t, n), u = JS("leakyreluAlpha", e, t, n);
+                                        if (s) {
+                                            if (i && 2 !== o) throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");
+                                            if (!i && 1 !== o) throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")
+                                        }
+                                        const [l, c] = JS("args", e, t, n);
+                                        return [bp.matMul({
+                                            a: JS("a", e, t, n),
+                                            b: JS("b", e, t, n),
+                                            transposeA: JS("transposeA", e, t, n),
+                                            transposeB: JS("transposeB", e, t, n),
+                                            bias: l,
+                                            activation: a,
+                                            preluActivationWeights: c,
+                                            leakyreluAlpha: u
+                                        })];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "normalization":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "FusedBatchNorm":
+                                    case "FusedBatchNormV2":
+                                    case "FusedBatchNormV3":
+                                        return [nl(JS("x", e, t, n), JS("mean", e, t, n), JS("variance", e, t, n), JS("offset", e, t, n), JS("scale", e, t, n), JS("epsilon", e, t, n))];
+                                    case "LRN":
+                                        return [nc(JS("x", e, t, n), JS("radius", e, t, n), JS("bias", e, t, n), JS("alpha", e, t, n), JS("beta", e, t, n))];
+                                    case "Softmax":
+                                        return [Lh(JS("x", e, t, n))];
+                                    case "LogSoftmax":
+                                        return [bc(JS("x", e, t, n))];
+                                    case "SparseToDense":
+                                        return [pp(JS("sparseIndices", e, t, n), JS("outputShape", e, t, n), JS("sparseValues", e, t, n), JS("defaultValue", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "reduction":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Max": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [mc(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "Mean": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [Lc(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "Min": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [zc(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "Sum": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [yc(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "All": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [Iu(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "Any": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [Au(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "ArgMax": {
+                                        const r = JS("axis", e, t, n);
+                                        return [Eu(JS("x", e, t, n), r)]
+                                    }
+                                    case "ArgMin": {
+                                        const r = JS("axis", e, t, n);
+                                        return [Nu(JS("x", e, t, n), r)]
+                                    }
+                                    case "Prod": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("keepDims", e, t, n);
+                                        return [ih(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "Cumsum": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("exclusive", e, t, n),
+                                            s = JS("reverse", e, t, n);
+                                        return [Il(JS("x", e, t, n), r, a, s)]
+                                    }
+                                    case "Bincount":
+                                        const r = JS("x", e, t, n),
+                                            a = JS("weights", e, t, n),
+                                            s = JS("size", e, t, n);
+                                        return [il(r, a, s)];
+                                    case "DenseBincount": {
+                                        const r = JS("x", e, t, n),
+                                            a = JS("weights", e, t, n),
+                                            s = JS("size", e, t, n),
+                                            i = JS("binaryOutput", e, t, n);
+                                        return [Al(r, a, s, i)]
+                                    }
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "slice_join":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "ConcatV2":
+                                    case "Concat": {
+                                        const r = JS("n", e, t, n),
+                                            a = JS("axis", e, t, n);
+                                        let s = JS("tensors", e, t, n);
+                                        return s = s.slice(0, r), [Xu(s, a)]
+                                    }
+                                    case "Gather": {
+                                        const r = JS("x", e, t, n),
+                                            a = JS("indices", e, t, n);
+                                        return [Gl(r, Ai(a, "int32"), 0)]
+                                    }
+                                    case "GatherV2": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("batchDims", e, t, n),
+                                            s = JS("x", e, t, n),
+                                            i = JS("indices", e, t, n);
+                                        return [Gl(s, Ai(i, "int32"), r, a)]
+                                    }
+                                    case "Reverse": {
+                                        const r = JS("dims", e, t, n),
+                                            a = [];
+                                        for (let e = 0; e < r.length; e++) r[e] && a.push(e);
+                                        const s = JS("x", e, t, n);
+                                        return [bh(s, a)]
+                                    }
+                                    case "ReverseV2": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("x", e, t, n);
+                                        return [bh(a, r)]
+                                    }
+                                    case "Slice": {
+                                        const r = JS("begin", e, t, n),
+                                            a = JS("size", e, t, n);
+                                        return [Zu(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "StridedSlice": {
+                                        const r = JS("begin", e, t, n),
+                                            a = JS("end", e, t, n),
+                                            s = JS("strides", e, t, n),
+                                            i = JS("beginMask", e, t, n),
+                                            o = JS("endMask", e, t, n),
+                                            u = JS("ellipsisMask", e, t, n),
+                                            l = JS("newAxisMask", e, t, n),
+                                            c = JS("shrinkAxisMask", e, t, n),
+                                            h = JS("x", e, t, n);
+                                        return [qh(h, r, a, s, i, o, u, l, c)]
+                                    }
+                                    case "Pack":
+                                        return ru(() => {
+                                            const r = JS("axis", e, t, n),
+                                                a = JS("tensors", e, t, n),
+                                                s = a[0].shape,
+                                                i = jh(a[0]).shape,
+                                                o = a.map(e => {
+                                                    const t = va.arraysEqual(e.shape, s);
+                                                    if (!t && !va.arraysEqual(jh(e).shape, i)) throw new Error("the input tensors shape does not match");
+                                                    return t ? e : qu(e, s)
+                                                });
+                                            return [Gh(o, r)]
+                                        });
+                                    case "Unpack": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("tensor", e, t, n);
+                                        return ap(a, r)
+                                    }
+                                    case "Tile": {
+                                        const r = JS("reps", e, t, n);
+                                        return [Ul(JS("x", e, t, n), r)]
+                                    }
+                                    case "Split":
+                                    case "SplitV": {
+                                        const r = JS("axis", e, t, n),
+                                            a = JS("numOrSizeSplits", e, t, n),
+                                            s = JS("x", e, t, n);
+                                        return zh(s, a, r)
+                                    }
+                                    case "ScatterNd": {
+                                        const r = JS("indices", e, t, n),
+                                            a = JS("values", e, t, n),
+                                            s = JS("shape", e, t, n);
+                                        return [hp(r, a, s)]
+                                    }
+                                    case "GatherNd": {
+                                        const r = JS("x", e, t, n),
+                                            a = JS("indices", e, t, n);
+                                        return [dp(r, a)]
+                                    }
+                                    case "SparseToDense": {
+                                        const r = JS("sparseIndices", e, t, n),
+                                            a = JS("outputShape", e, t, n),
+                                            s = JS("sparseValues", e, t, n),
+                                            i = JS("defaultValue", e, t, n);
+                                        return [pp(r, s, a, s.dtype === i.dtype ? i : Ai(i, s.dtype))]
+                                    }
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "sparse":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "SparseFillEmptyRows": {
+                                        const {
+                                            outputIndices: r,
+                                            outputValues: a,
+                                            emptyRowIndicator: s,
+                                            reverseIndexMap: i
+                                        } = Rd.sparseFillEmptyRows(JS("indices", e, t, n), JS("values", e, t, n), JS("denseShape", e, t, n), JS("defaultValue", e, t, n));
+                                        return [r, a, s, i]
+                                    }
+                                    case "SparseReshape": {
+                                        const {
+                                            outputIndices: r,
+                                            outputShape: a
+                                        } = Rd.sparseReshape(JS("inputIndices", e, t, n), JS("inputShape", e, t, n), JS("newShape", e, t, n));
+                                        return [r, a]
+                                    }
+                                    case "SparseSegmentMean":
+                                        return [Rd.sparseSegmentMean(JS("data", e, t, n), JS("indices", e, t, n), JS("segmentIds", e, t, n))];
+                                    case "SparseSegmentSum":
+                                        return [Rd.sparseSegmentSum(JS("data", e, t, n), JS("indices", e, t, n), JS("segmentIds", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "spectral":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "FFT":
+                                        return [Ph(JS("x", e, t, n))];
+                                    case "IFFT":
+                                        return [$h(JS("x", e, t, n))];
+                                    case "RFFT":
+                                        return [Uh(JS("x", e, t, n))];
+                                    case "IRFFT":
+                                        return [Bh(JS("x", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "string":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "StringNGrams": {
+                                        const {
+                                            nGrams: r,
+                                            nGramsSplits: a
+                                        } = _d.stringNGrams(JS("data", e, t, n), JS("dataSplits", e, t, n), JS("separator", e, t, n), JS("nGramWidths", e, t, n), JS("leftPad", e, t, n), JS("rightPad", e, t, n), JS("padWidth", e, t, n), JS("preserveShortSequences", e, t, n));
+                                        return [r, a]
+                                    }
+                                    case "StringSplit": {
+                                        const {
+                                            indices: r,
+                                            values: a,
+                                            shape: s
+                                        } = _d.stringSplit(JS("input", e, t, n), JS("delimiter", e, t, n), JS("skipEmpty", e, t, n));
+                                        return [r, a, s]
+                                    }
+                                    case "StringToHashBucketFast":
+                                        return [_d.stringToHashBucketFast(JS("input", e, t, n), JS("numBuckets", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "transformation":
+                            return ru(() => ((e, t, n) => {
+                                switch (e.op) {
+                                    case "Cast":
+                                        return [Ai(JS("x", e, t, n), JS("dtype", e, t, n))];
+                                    case "ExpandDims": {
+                                        const r = JS("axis", e, t, n);
+                                        return [Bl(JS("x", e, t, n), r)]
+                                    }
+                                    case "Squeeze": {
+                                        const r = JS("axis", e, t, n);
+                                        return [jh(JS("x", e, t, n), r)]
+                                    }
+                                    case "Reshape":
+                                        return [qu(JS("x", e, t, n), JS("shape", e, t, n))];
+                                    case "MirrorPad":
+                                        return [Wc(JS("x", e, t, n), JS("padding", e, t, n), JS("mode", e, t, n))];
+                                    case "PadV2":
+                                    case "Pad":
+                                        return [Qc(JS("x", e, t, n), JS("padding", e, t, n), JS("constantValue", e, t, n))];
+                                    case "SpaceToBatchND": {
+                                        const r = JS("blockShape", e, t, n),
+                                            a = JS("paddings", e, t, n);
+                                        return [nh(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "BatchToSpaceND": {
+                                        const r = JS("blockShape", e, t, n),
+                                            a = JS("crops", e, t, n);
+                                        return [tl(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "DepthToSpace": {
+                                        const r = JS("blockSize", e, t, n),
+                                            a = JS("dataFormat", e, t, n).toUpperCase();
+                                        return [El(JS("x", e, t, n), r, a)]
+                                    }
+                                    case "BroadcastTo":
+                                        return [ul(JS("x", e, t, n), JS("shape", e, t, n))];
+                                    case "BroadcastArgs":
+                                        return [ol(JS("s0", e, t, n), JS("s1", e, t, n))];
+                                    default:
+                                        throw TypeError(`Node type ${e.op} is not implemented`)
+                                }
+                            })(e, t, n));
+                        case "hash_table":
+                            return dA(e, t, n, r);
+                        case "custom":
+                            const a = QS(e.op);
+                            if (a && a.customExecutor) return a.customExecutor(new class {
+                                constructor(e, t, n) {
+                                    this.node = e, this.tensorMap = t, this.context = n, this.inputs = [], this.attrs = {}, this.inputs = e.inputNames.map(e => this.getInput(e)), null != e.rawAttrs && (this.attrs = Object.keys(e.rawAttrs).reduce((e, t) => (e[t] = this.getAttr(t), e), {}))
+                                }
+                                getInput(e) {
+                                    return eI(e, this.tensorMap, this.context)
+                                }
+                                getAttr(e, t) {
+                                    const n = this.node.rawAttrs[e];
+                                    if (null != n.tensor) return eI(e, this.tensorMap, this.context);
+                                    if (null != n.i || null != n.f) return KI(this.node.rawAttrs, e, t);
+                                    if (null != n.s) return HI(this.node.rawAttrs, e, t);
+                                    if (null != n.b) return qI(this.node.rawAttrs, e, t);
+                                    if (null != n.shape) return eA(this.node.rawAttrs, e, t);
+                                    if (null != n.type) return QI(this.node.rawAttrs, e, t);
+                                    if (null != n.list) {
+                                        if (null != n.list.i || null != n.list.f) return tA(this.node.rawAttrs, e, t);
+                                        if (null != n.list.s) return nA(this.node.rawAttrs, e, t);
+                                        if (null != n.list.shape) return rA(this.node.rawAttrs, e, t);
+                                        if (null != n.list.b) return aA(this.node.rawAttrs, e, t);
+                                        if (null != n.list.type) return ZI(this.node.rawAttrs, e, t)
+                                    }
+                                    return t
+                                }
+                            }(e, t, n));
+                            throw TypeError(`Custom op ${e.op} is not registered.`);
+                        default:
+                            throw TypeError(`Unknown op '${e.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)
+                    }
+                })(e, t, n);
+                return va.isPromise(a) ? a.then(e => [].concat(e)) : [].concat(a)
+            }
+            var mA = class {
+                constructor(e = {}, t = {}, n = {}, r = {}) {
+                    this.weightMap = e, this.tensorArrayMap = t, this.tensorListMap = n, this.functionMap = r, this.rootContext = {
+                        id: 0,
+                        frameName: "",
+                        iterationId: 0
+                    }, this.contexts = [this.rootContext], this.lastId = 0, this.generateCurrentContextIds()
+                }
+                newFrame(e, t) {
+                    return {
+                        id: e,
+                        frameName: t,
+                        iterationId: 0
+                    }
+                }
+                set currentContext(e) {
+                    this.contexts !== e && (this.contexts = e, this.generateCurrentContextIds())
+                }
+                get currentContext() {
+                    return this.contexts
+                }
+                get currentContextId() {
+                    return this._currentContextIds[0]
+                }
+                get currentContextIds() {
+                    return this._currentContextIds
+                }
+                generateCurrentContextIds() {
+                    const e = [];
+                    for (let t = 0; t < this.contexts.length - 1; t++) {
+                        const n = this.contexts.slice(0, this.contexts.length - t);
+                        e.push(this.contextIdforContexts(n))
+                    }
+                    e.push(""), this._currentContextIds = e
+                }
+                contextIdforContexts(e) {
+                    return e ? e.map(e => 0 === e.id && 0 === e.iterationId ? "" : `${e.frameName}-${e.iterationId}`).join("/") : ""
+                }
+                enterFrame(e) {
+                    this.contexts && (this.lastId++, this.contexts = this.contexts.slice(), this.contexts.push(this.newFrame(this.lastId, e)), this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))
+                }
+                exitFrame() {
+                    if (!(this.contexts && this.contexts.length > 1)) throw new Error("Cannot exit frame, the context is empty");
+                    this.contexts = this.contexts.slice(), this.contexts.splice(-1), this.currentContextIds.shift()
+                }
+                nextIteration() {
+                    if (!(this.contexts && this.contexts.length > 0)) throw new Error("Cannot increase frame iteration, the context is empty"); {
+                        this.contexts = this.contexts.slice(), this.lastId++;
+                        const e = Object.assign({}, this.contexts[this.contexts.length - 1]);
+                        e.iterationId += 1, e.id = this.lastId, this.contexts.splice(-1, 1, e), this._currentContextIds.splice(0, 1, this.contextIdforContexts(this.contexts))
+                    }
+                }
+                getWeight(e) {
+                    return this.weightMap[e]
+                }
+                addTensorArray(e) {
+                    this.tensorArrayMap[e.id] = e
+                }
+                getTensorArray(e) {
+                    return this.tensorArrayMap[e]
+                }
+                addTensorList(e) {
+                    this.tensorListMap[e.id] = e
+                }
+                getTensorList(e) {
+                    return this.tensorListMap[e]
+                }
+                dispose(e) {
+                    for (const t in this.tensorArrayMap) this.tensorArrayMap[t].clearAndClose(e);
+                    for (const t in this.tensorListMap) this.tensorListMap[t].clearAndClose(e)
+                }
+            };
+
+            function gA(e, t, n, r) {
+                const a = new Set,
+                    s = [];
+                let i = null,
+                    o = null;
+                const u = new Set,
+                    l = Object.keys(e).map(e => rI(e)[0]);
+                let c = [];
+                null != r && (c = r.map(e => rI(e.name)[0]));
+                const h = [...t];
+                for (; h.length > 0;) {
+                    const e = h.pop();
+                    (xA(e) || wA(e) || kA(e)) && null == i && (i = e, o = i.children.map(e => e.name).filter(e => a.has(e))), a.add(e.name), null == n[e.name] && (-1 === l.indexOf(e.name) && -1 === c.indexOf(e.name) && (0 !== e.inputs.length ? e.inputs.forEach(e => {
+                        u.has(e.name) || (u.add(e.name), h.push(e))
+                    }) : s.push(e.name)))
+                }
+                return {
+                    inputs: e,
+                    outputs: t,
+                    usedNodes: a,
+                    missingInputs: s,
+                    dynamicNode: i,
+                    syncInputs: o
+                }
+            }
+            var yA = ["Switch", "Merge", "Enter", "Exit", "NextIteration", "StatelessIf", "StatelessWhile", "if", "While"],
+                bA = ["NonMaxSuppressionV2", "NonMaxSuppressionV3", "NonMaxSuppressionV5", "Where"],
+                vA = ["HashTable", "HashTableV2", "LookupTableImport", "LookupTableImportV2", "LookupTableFind", "LookupTableFindV2", "LookupTableSize", "LookupTableSizeV2"];
+
+            function xA(e) {
+                return yA.indexOf(e.op) >= 0
+            }
+
+            function wA(e) {
+                return bA.indexOf(e.op) >= 0
+            }
+
+            function kA(e) {
+                return vA.indexOf(e.op) >= 0
+            }
+            var SA = class {
+                    constructor(e, t) {
+                        this.graph = e, this.parent = t, this.compiledMap = new Map, this._weightMap = {}, this.SEPERATOR = ",", this._functions = {}, this._functionExecutorMap = {}, this.intermediateTensors = {}, this.keepTensorForDebug = !1, this._outputs = e.outputs, this._inputs = e.inputs, this._initNodes = e.initNodes, this._signature = e.signature, this._functions = e.functions, null != e.functions && Object.keys(e.functions).forEach(t => {
+                            this._functionExecutorMap[t] = new SA(e.functions[t], this)
+                        })
+                    }
+                    get weightIds() {
+                        return this.parent ? this.parent.weightIds : this._weightIds
+                    }
+                    get functionExecutorMap() {
+                        return this.parent ? this.parent.functionExecutorMap : this._functionExecutorMap
+                    }
+                    get weightMap() {
+                        return this.parent ? this.parent.weightMap : this._weightMap
+                    }
+                    set weightMap(e) {
+                        const t = Object.keys(e).map(t => e[t].map(e => e.id));
+                        this._weightIds = [].concat(...t), this._weightMap = e
+                    }
+                    set resourceManager(e) {
+                        this._resourceManager = e
+                    }
+                    get inputs() {
+                        return this._inputs.map(e => ({
+                            name: e.name,
+                            shape: e.attrParams.shape ? e.attrParams.shape.value : void 0,
+                            dtype: e.attrParams.dtype ? e.attrParams.dtype.value : void 0
+                        }))
+                    }
+                    get outputs() {
+                        return this._outputs.map(e => ({
+                            name: e.name,
+                            shape: e.attrParams.shape ? e.attrParams.shape.value : void 0,
+                            dtype: e.attrParams.dtype ? e.attrParams.dtype.value : void 0
+                        }))
+                    }
+                    get inputNodes() {
+                        return this._inputs.map(e => e.signatureKey || e.name)
+                    }
+                    get outputNodes() {
+                        return this._outputs.map(e => {
+                            const t = e.signatureKey || e.name;
+                            return e.defaultOutput ? `${t}:${e.defaultOutput}` : t
+                        })
+                    }
+                    get functions() {
+                        return Object.keys(this._functions).reduce((e, t) => (e[t] = this._functions[t].signature, e), {})
+                    }
+                    getCompilationKey(e, t) {
+                        const n = e.map(e => e.name).sort(),
+                            r = t.map(e => e.name).sort();
+                        return n.join(this.SEPERATOR) + "--" + r.join(this.SEPERATOR)
+                    }
+                    compile(e, t) {
+                        const n = gA(e, t, this.weightMap, this._initNodes),
+                            {
+                                missingInputs: r,
+                                dynamicNode: a,
+                                syncInputs: s
+                            } = n;
+                        if (null != a) throw new Error(`This execution contains the node '${a.name}', which has the dynamic op '${a.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${s}]`);
+                        if (r.length > 0) {
+                            const n = t.map(e => e.name),
+                                a = Object.keys(e);
+                            throw new Error(`Cannot compute the outputs [${n}] from the provided inputs [${a}]. Missing the following inputs: [${r}]`)
+                        }
+                        return function (e, t, n) {
+                            const {
+                                usedNodes: r,
+                                inputs: a
+                            } = n, s = [], i = Object.keys(a).map(e => rI(e)[0]).map(t => e.nodes[t]), o = e.initNodes;
+                            i.forEach(e => {
+                                r.has(e.name) && s.push(e)
+                            }), e.weights.forEach(e => {
+                                r.has(e.name) && s.push(e)
+                            }), null != o && o.forEach(e => {
+                                r.has(e.name) && s.push(e)
+                            });
+                            const u = new Set,
+                                l = [];
+                            for (; s.length > 0;) {
+                                const e = s.pop();
+                                u.add(e.name), t[e.name] || l.push(e), e.children.forEach(e => {
+                                    !u.has(e.name) && r.has(e.name) && e.inputs.every(e => u.has(e.name)) && s.push(e)
+                                })
+                            }
+                            return l
+                        }(this.graph, this.weightMap, n)
+                    }
+                    execute(e, t) {
+                        e = this.mapInputs(e);
+                        const n = Object.keys(e).sort();
+                        this.checkInputs(e), this.checkInputShapeAndType(e), t = this.mapOutputs(t), this.checkOutputs(t);
+                        const r = n.map(e => this.graph.nodes[rI(e)[0]]),
+                            a = t.map(e => rI(e)[0]);
+                        let s = a.map(e => this.graph.nodes[e]);
+                        this.resetIntermediateTensors(), 0 === s.length && (s = this._outputs);
+                        const i = this.getCompilationKey(r, s);
+                        let o = this.compiledMap.get(i);
+                        null == o && (o = this.compile(e, s), this.compiledMap.set(i, o));
+                        const u = {},
+                            l = {};
+                        return ru(() => {
+                            const n = new mA(this.weightMap, u, l, this.functionExecutorMap),
+                                r = Object.assign({}, this.weightMap);
+                            Object.keys(e).forEach(t => {
+                                const [n, a] = rI(t), s = [];
+                                s[a] = e[t], r[n] = s
+                            });
+                            const s = this.getFrozenTensorIds(r),
+                                i = {};
+                            for (let e = 0; e < o.length; e++) {
+                                const t = o[e];
+                                if (!r[t.name]) {
+                                    const e = fA(t, r, n, this._resourceManager);
+                                    if (va.isPromise(e)) throw new Error(`The execution of the op '${t.op}' returned a promise. Please use model.executeAsync() instead.`);
+                                    r[t.name] = e, this.checkTensorForDisposal(t.name, t, r, n, s, a, i)
+                                }
+                            }
+                            return null == this.parent && n.dispose(s), t.map(e => eI(e, r, n))
+                        })
+                    }
+                    getFrozenTensorIds(e) {
+                        const t = [].concat.apply([], Object.keys(e).map(t => e[t]).map(e => e.map(e => e.id)));
+                        return new Set(t)
+                    }
+                    checkTensorForDisposal(e, t, n, r, a, s, i) {
+                        "control" !== t.category && -1 === s.indexOf(e) && (n[e].forEach(e => {
+                            null != e && (i[e.id] = (i[e.id] || 0) + t.children.length)
+                        }), t.inputs.forEach(e => {
+                            if ("control" !== e.category) {
+                                const s = function (e, t, n) {
+                                    return t[nI(e, n.currentContextId)]
+                                }(e.name, n, r);
+                                null != s && s.forEach(e => {
+                                    if (e && !e.kept && !a.has(e.id)) {
+                                        const n = i[e.id];
+                                        if (1 === n) {
+                                            if (this.keepTensorForDebug) {
+                                                const [n, a] = tI(t.name, r);
+                                                this.intermediateTensors[n] || (this.intermediateTensors[n] = []), this.intermediateTensors[n][a] = e
+                                            } else e.dispose();
+                                            delete i[e.id]
+                                        } else null != n && i[e.id]--
+                                    }
+                                })
+                            }
+                        }))
+                    }
+                    async executeAsync(e, t) {
+                        return this._executeAsync(e, t)
+                    }
+                    disposeIntermediateTensors() {
+                        this.intermediateTensors && (Object.keys(this.intermediateTensors).forEach(e => this.intermediateTensors[e].forEach(e => e.dispose())), this.disposeTensorsMap())
+                    }
+                    disposeTensorsMap() {
+                        this.tensorsMap && Object.keys(this.tensorsMap).forEach(e => {
+                            this.tensorsMap[e].forEach(e => {
+                                !e || e.kept || e.isDisposed || this.keepIds.has(e.id) || e.dispose()
+                            })
+                        })
+                    }
+                    getIntermediateTensors() {
+                        return this.tensorsMap
+                    }
+                    resetIntermediateTensors() {
+                        for (const e in this.intermediateTensors) this.intermediateTensors[e].forEach(e => e.dispose()), delete this.intermediateTensors[e]
+                    }
+                    async _executeAsync(e, t, n = !1, r = {}, a = {}) {
+                        n || (e = this.mapInputs(e), this.checkInputs(e), this.checkInputShapeAndType(e), t = this.mapOutputs(t), this.checkOutputs(t));
+                        try {
+                            this.keepTensorForDebug = Ye().getBool("KEEP_INTERMEDIATE_TENSORS")
+                        } catch (e) {
+                            console.warn(e.message)
+                        }
+                        this.resetIntermediateTensors();
+                        const s = new mA(this.weightMap, r, a, this.functionExecutorMap);
+                        this.tensorsMap = await this.executeWithControlFlow(e, s, t, n);
+                        const i = t.map(e => eI(e, this.tensorsMap, s)),
+                            o = i.map(e => e.id),
+                            u = Object.keys(e).map(t => e[t].id);
+                        return this.keepIds = new Set([...o, ...u, ...this.weightIds]), this.keepTensorForDebug || this.disposeTensorsMap(), null == this.parent && s.dispose(this.keepIds), i
+                    }
+                    async executeFunctionAsync(e, t, n) {
+                        const r = e.reduce((e, t, n) => (e[this.inputs[n].name] = t, e), {});
+                        return this._executeAsync(r, this.outputNodes, !0, t, n)
+                    }
+                    async executeWithControlFlow(e, t, n, r) {
+                        const a = Object.keys(e),
+                            s = a.map(e => this.graph.nodes[rI(e)[0]]),
+                            i = n.map(e => rI(e)[0]);
+                        let o = i.map(e => this.graph.nodes[e]);
+                        0 === o.length && (o = this._outputs);
+                        const {
+                            usedNodes: u,
+                            missingInputs: l,
+                            dynamicNode: c,
+                            syncInputs: h
+                        } = gA(e, o, this.weightMap, this._initNodes), p = [...s, ...this.graph.weights, ...this._initNodes || []].map(e => ({
+                            node: e,
+                            contexts: t.currentContext
+                        })), d = Object.assign({}, this.weightMap);
+                        Object.keys(e).forEach(t => {
+                            const [n, r] = rI(t), a = [];
+                            a[r] = e[t], d[n] = a
+                        });
+                        const f = {},
+                            m = this.getFrozenTensorIds(d),
+                            g = {};
+                        for (; p.length > 0;) {
+                            const e = this.processStack(s, p, t, d, g, m, i, f, u);
+                            await Promise.all(e)
+                        }
+                        null != c || r || console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");
+                        const y = o.filter(e => !xA(e) && !eI(e.name, d, t)).map(e => e.name);
+                        if (y.length > 0) {
+                            let e = "";
+                            throw null != c && (e = `Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${h}]`), new Error(`Cannot compute the outputs [${y}] from the provided inputs [${a}]. Consider providing the following inputs: [${l}]. ${e}`)
+                        }
+                        return d
+                    }
+                    processStack(e, t, n, r, a, s, i, o, u) {
+                        const l = [];
+                        for (; t.length > 0;) {
+                            const e = t.pop();
+                            n.currentContext = e.contexts;
+                            let c = "";
+                            if ("Enter" === e.node.op && JS("isConstant", e.node, r, n) && ([c] = tI(e.node.name, n)), null == r[e.node.name]) {
+                                const h = fA(e.node, r, n, this._resourceManager);
+                                c || ([c] = tI(e.node.name, n));
+                                const p = n.currentContext;
+                                va.isPromise(h) ? l.push(h.then(l => (r[c] = l, n.currentContext = p, this.checkTensorForDisposal(c, e.node, r, n, s, i, o), this.processChildNodes(e.node, t, n, r, a, u), l))) : (r[c] = h, this.checkTensorForDisposal(c, e.node, r, n, s, i, o), this.processChildNodes(e.node, t, n, r, a, u))
+                            } else this.processChildNodes(e.node, t, n, r, a, u)
+                        }
+                        return l
+                    }
+                    processChildNodes(e, t, n, r, a, s) {
+                        e.children.forEach(e => {
+                            const [i] = tI(e.name, n);
+                            !a[i] && s.has(e.name) && ("Merge" === e.op ? e.inputNames.some(e => !!eI(e, r, n)) && (a[i] = !0, t.push({
+                                contexts: n.currentContext,
+                                node: e
+                            })) : e.inputNames.every(e => !!eI(e, r, n)) && (a[i] = !0, t.push({
+                                contexts: n.currentContext,
+                                node: e
+                            })))
+                        })
+                    }
+                    dispose() {
+                        Object.keys(this.weightMap).forEach(e => this.weightMap[e].forEach(e => e.dispose()))
+                    }
+                    checkInputShapeAndType(e) {
+                        Object.keys(e).forEach(t => {
+                            const n = e[t],
+                                [r] = rI(t),
+                                a = this.graph.nodes[r];
+                            if (a.attrParams.shape && a.attrParams.shape.value) {
+                                const e = a.attrParams.shape.value,
+                                    t = e.length === n.shape.length && n.shape.every((t, n) => -1 === e[n] || e[n] === t);
+                                va.assert(t, () => `The shape of dict['${a.name}'] provided in model.execute(dict) must be [${e}], but was [${n.shape}]`)
+                            }
+                            a.attrParams.dtype && a.attrParams.dtype.value && va.assert(n.dtype === a.attrParams.dtype.value, () => `The dtype of dict['${a.name}'] provided in model.execute(dict) must be ${a.attrParams.dtype.value}, but was ${n.dtype}`)
+                        })
+                    }
+                    mapInputs(e) {
+                        const t = {};
+                        for (const n in e)
+                            if (null != this._signature && null != this._signature.inputs && null != this._signature.inputs[n]) {
+                                t[this._signature.inputs[n].name] = e[n]
+                            } else t[n] = e[n];
+                        return t
+                    }
+                    checkInputs(e) {
+                        const t = Object.keys(e).filter(e => {
+                            const [t] = rI(e);
+                            return null == this.graph.nodes[t]
+                        });
+                        if (t.length > 0) throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)
+                    }
+                    mapOutputs(e) {
+                        return e.map(e => {
+                            if (null != this._signature && null != this._signature.outputs && null != this._signature.outputs[e]) {
+                                return this._signature.outputs[e].name
+                            }
+                            return e
+                        }, {})
+                    }
+                    checkOutputs(e) {
+                        e.forEach(e => {
+                            const [t] = rI(e);
+                            if (!this.graph.nodes[t]) throw new Error(`The output '${e}' is not found in the graph`)
+                        })
+                    }
+                },
+                IA = class {
+                    constructor(e, t = {}) {
+                        this.modelUrl = e, this.loadOptions = t, this.version = "n/a", null == t && (this.loadOptions = {}), this.resourceManager = new class {
+                            constructor(e = {}, t = {}) {
+                                this.hashTableNameToHandle = e, this.hashTableMap = t
+                            }
+                            addHashTable(e, t) {
+                                this.hashTableNameToHandle[e] = t.handle, this.hashTableMap[t.id] = t
+                            }
+                            getHashTableHandleByName(e) {
+                                return this.hashTableNameToHandle[e]
+                            }
+                            getHashTableById(e) {
+                                return this.hashTableMap[e]
+                            }
+                            dispose() {
+                                for (const e in this.hashTableMap) this.hashTableMap[e].clearAndClose(), delete this.hashTableMap[e];
+                                for (const e in this.hashTableNameToHandle) this.hashTableNameToHandle[e].dispose(), delete this.hashTableNameToHandle[e]
+                            }
+                        }
+                    }
+                    get modelVersion() {
+                        return this.version
+                    }
+                    get inputNodes() {
+                        return this.executor.inputNodes
+                    }
+                    get outputNodes() {
+                        return this.executor.outputNodes
+                    }
+                    get inputs() {
+                        return this.executor.inputs
+                    }
+                    get outputs() {
+                        return this.executor.outputs
+                    }
+                    get weights() {
+                        return this.executor.weightMap
+                    }
+                    get metadata() {
+                        return this.artifacts.userDefinedMetadata
+                    }
+                    get modelSignature() {
+                        return this.signature
+                    }
+                    findIOHandler() {
+                        const e = this.modelUrl;
+                        if (null != e.load) this.handler = e;
+                        else if (null != this.loadOptions.requestInit) this.handler = Ci.browserHTTPRequest(e, this.loadOptions);
+                        else {
+                            const t = Ci.getLoadHandlers(e, this.loadOptions);
+                            if (0 === t.length) t.push(Ci.browserHTTPRequest(e, this.loadOptions));
+                            else if (t.length > 1) throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);
+                            this.handler = t[0]
+                        }
+                    }
+                    async load() {
+                        if (this.findIOHandler(), null == this.handler.load) throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");
+                        const e = await this.handler.load();
+                        return this.loadSync(e)
+                    }
+                    loadSync(e) {
+                        this.artifacts = e;
+                        const t = this.artifacts.modelTopology;
+                        let n;
+                        n = null != this.artifacts.userDefinedMetadata && null != this.artifacts.userDefinedMetadata.signature ? this.artifacts.userDefinedMetadata.signature : this.artifacts.signature, this.signature = n, this.version = `${t.versions.producer}.${t.versions.minConsumer}`;
+                        const r = Ci.decodeWeights(this.artifacts.weightData, this.artifacts.weightSpecs);
+                        if (this.executor = new SA(jI.Instance.transformGraph(t, this.signature)), this.executor.weightMap = this.convertTensorMapToTensorsMap(r), this.executor.resourceManager = this.resourceManager, null != e.modelInitializer && null != e.modelInitializer.node) {
+                            const t = jI.Instance.transformGraph(e.modelInitializer);
+                            this.initializer = new SA(t), this.initializer.weightMap = this.executor.weightMap, this.initializer.resourceManager = this.resourceManager, this.initializer.executeAsync({}, [])
+                        }
+                        return !0
+                    }
+                    async save(e, t) {
+                        if ("string" == typeof e) {
+                            const t = Ci.getSaveHandlers(e);
+                            if (0 === t.length) throw new Error(`Cannot find any save handlers for URL '${e}'`);
+                            if (t.length > 1) throw new Error(`Found more than one (${t.length}) save handlers for URL '${e}'`);
+                            e = t[0]
+                        }
+                        if (null == e.save) throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");
+                        return e.save(this.artifacts)
+                    }
+                    predict(e, t) {
+                        return this.execute(e, this.outputNodes)
+                    }
+                    normalizeInputs(e) {
+                        if (!(e instanceof Ya || Array.isArray(e))) return e;
+                        if ((e = Array.isArray(e) ? e : [e]).length !== this.inputNodes.length) throw new Error(`Input tensor count mismatch,the graph model has ${this.inputNodes.length} placeholders, while there are ${e.length} input tensors.`);
+                        return this.inputNodes.reduce((t, n, r) => (t[n] = e[r], t), {})
+                    }
+                    normalizeOutputs(e) {
+                        return e = e || this.outputNodes, Array.isArray(e) ? e : [e]
+                    }
+                    execute(e, t) {
+                        e = this.normalizeInputs(e), t = this.normalizeOutputs(t);
+                        const n = this.executor.execute(e, t);
+                        return n.length > 1 ? n : n[0]
+                    }
+                    async executeAsync(e, t) {
+                        e = this.normalizeInputs(e), t = this.normalizeOutputs(t);
+                        const n = await this.executor.executeAsync(e, t);
+                        return n.length > 1 ? n : n[0]
+                    }
+                    getIntermediateTensors() {
+                        return this.executor.getIntermediateTensors()
+                    }
+                    disposeIntermediateTensors() {
+                        this.executor.disposeIntermediateTensors()
+                    }
+                    convertTensorMapToTensorsMap(e) {
+                        return Object.keys(e).reduce((t, n) => (t[n] = [e[n]], t), {})
+                    }
+                    dispose() {
+                        this.executor.dispose(), this.initializer && this.initializer.dispose(), this.resourceManager.dispose()
+                    }
+                };
+            async function AA(e, t = {}) {
+                if (null == e) throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");
+                null == t && (t = {}), t.fromTFHub && null == e.load && (e.endsWith("/") || (e += "/"), e += "model.json?tfjs-format=file");
+                const n = new IA(e, t);
+                return await n.load(), n
+            }
+            var EA = "3.13.0",
+                NA = {};
+            w(NA, {
+                CSVDataset: () => yE,
+                Dataset: () => iE,
+                FileDataSource: () => RE,
+                TextLineDataset: () => hE,
+                URLDataSource: () => _E,
+                array: () => uE,
+                csv: () => FE,
+                func: () => ME,
+                generator: () => DE,
+                microphone: () => LE,
+                version_data: () => PE,
+                webcam: () => OE,
+                zip: () => lE
+            });
+            var CA = k(W()),
+                TA = k(W());
+
+            function RA(e, t, n = new Map, r = new Set) {
+                if (null == e) return null;
+                if ("function" == typeof Blob && e instanceof Blob) return e.slice();
+                if (r.has(e)) throw new Error("Circular references are not supported.");
+                if (n.has(e)) return n.get(e);
+                const a = t(e);
+                if (a.recurse && null !== a.value) throw new Error("A deep map function may not return both a value and recurse=true.");
+                if (a.recurse) {
+                    if (DA(e)) {
+                        const a = Array.isArray(e) ? [] : {};
+                        r.add(e);
+                        for (const s in e) {
+                            const i = RA(e[s], t, n, r);
+                            a[s] = i
+                        }
+                        return r.delete(e), e.__proto__ && (a.__proto__ = e.__proto__), a
+                    }
+                    throw new Error("Can't recurse into non-iterable type: " + e)
+                }
+                return n.set(e, a.value), a.value
+            }
+
+            function _A(e, t = FA) {
+                return function e(t, n, r = new Set) {
+                    const a = t[0];
+                    if (r.has(a)) throw new Error("Circular references are not supported.");
+                    const s = n(t);
+                    if (s.recurse && null !== s.value) throw new Error("A deep zip function may not return both a value and recurse=true.");
+                    if (s.recurse) {
+                        if (DA(a)) {
+                            const s = Array.isArray(a) ? [] : {};
+                            r.add(a);
+                            for (const i in a) {
+                                const a = t.map(e => e[i]),
+                                    o = e(a, n, r);
+                                s[i] = o
+                            }
+                            return r.delete(a), s
+                        }
+                        throw new Error("Can't recurse into non-iterable type: " + a)
+                    }
+                    return s.value
+                }(e, t)
+            }
+
+            function FA(e) {
+                return null === e ? null : DA(e[0]) ? {
+                    value: null,
+                    recurse: !0
+                } : {
+                    value: e,
+                    recurse: !1
+                }
+            }
+            async function MA(e, t) {
+                const n = new Map;
+                RA(e, t, n);
+                for (const e of Array.from(n.keys())) {
+                    const t = n.get(e);
+                    if (va.isPromise(t)) {
+                        const r = await t;
+                        n.set(e, r)
+                    }
+                }
+                return RA(e, t, n)
+            }
+
+            function DA(e) {
+                let t = !1;
+                if (Ye().get("IS_BROWSER")) t = e instanceof TextDecoder;
+                else {
+                    const {
+                        StringDecoder: n
+                    } = V();
+                    t = e instanceof n
+                }
+                return null != e && !ArrayBuffer.isView(e) && (Array.isArray(e) || "object" == typeof e && !(e instanceof Ya) && !(e instanceof Promise) && !t)
+            }
+
+            function OA(e) {
+                return RA(e, LA)
+            }
+
+            function LA(e) {
+                return e instanceof Ya ? {
+                    value: e.clone(),
+                    recurse: !1
+                } : DA(e) ? {
+                    value: null,
+                    recurse: !0
+                } : {
+                    value: e,
+                    recurse: !1
+                }
+            }
+            var PA = class {
+                    constructor(e) {
+                        if (this.capacity = e, this.begin = 0, this.end = 0, null == e) throw new RangeError("Can't create a ring buffer of unknown capacity.");
+                        if (e < 1) throw new RangeError("Can't create ring buffer of capacity < 1.");
+                        this.data = new Array(e), this.doubledCapacity = 2 * e
+                    }
+                    wrap(e) {
+                        for (; e < 0;) e += this.doubledCapacity;
+                        return e % this.doubledCapacity
+                    }
+                    get(e) {
+                        if (e < 0) throw new RangeError("Can't get item at a negative index.");
+                        return this.data[e % this.capacity]
+                    }
+                    set(e, t) {
+                        if (e < 0) throw new RangeError("Can't set item at a negative index.");
+                        this.data[e % this.capacity] = t
+                    }
+                    length() {
+                        let e = this.end - this.begin;
+                        return e < 0 && (e = this.doubledCapacity + e), e
+                    }
+                    isFull() {
+                        return this.length() === this.capacity
+                    }
+                    isEmpty() {
+                        return 0 === this.length()
+                    }
+                    push(e) {
+                        if (this.isFull()) throw new RangeError("Ring buffer is full.");
+                        this.set(this.end, e), this.end = this.wrap(this.end + 1)
+                    }
+                    pushAll(e) {
+                        for (const t of e) this.push(t)
+                    }
+                    pop() {
+                        if (this.isEmpty()) throw new RangeError("Ring buffer is empty.");
+                        this.end = this.wrap(this.end - 1);
+                        const e = this.get(this.end);
+                        return this.set(this.end, void 0), e
+                    }
+                    unshift(e) {
+                        if (this.isFull()) throw new RangeError("Ring buffer is full.");
+                        this.begin = this.wrap(this.begin - 1), this.set(this.begin, e)
+                    }
+                    shift() {
+                        if (this.isEmpty()) throw new RangeError("Ring buffer is empty.");
+                        const e = this.get(this.begin);
+                        return this.set(this.begin, void 0), this.begin = this.wrap(this.begin + 1), e
+                    }
+                    shuffleExcise(e) {
+                        if (this.isEmpty()) throw new RangeError("Ring buffer is empty.");
+                        const t = this.wrap(this.begin + e),
+                            n = this.get(t);
+                        return this.set(t, this.pop()), n
+                    }
+                },
+                $A = class extends PA {
+                    constructor() {
+                        super($A.INITIAL_CAPACITY)
+                    }
+                    isFull() {
+                        return !1
+                    }
+                    push(e) {
+                        super.isFull() && this.expand(), super.push(e)
+                    }
+                    unshift(e) {
+                        super.isFull() && this.expand(), super.unshift(e)
+                    }
+                    expand() {
+                        const e = 2 * this.capacity,
+                            t = new Array(e),
+                            n = this.length();
+                        for (let e = 0; e < n; e++) t[e] = this.get(this.wrap(this.begin + e));
+                        this.data = t, this.capacity = e, this.doubledCapacity = 2 * this.capacity, this.begin = 0, this.end = n
+                    }
+                };
+
+            function BA(e) {
+                return new jA(e)
+            }
+
+            function zA(e) {
+                return new GA(e)
+            }
+            $A.INITIAL_CAPACITY = 32;
+            var UA, WA, VA = class {
+                    async toArray() {
+                        const e = [];
+                        let t = await this.next();
+                        for (; !t.done;) e.push(t.value), t = await this.next();
+                        return e
+                    }
+                    async toArrayForTest() {
+                        const e = this.prefetch(100),
+                            t = [];
+                        let n = await e.next();
+                        for (; !n.done;) t.push(n.value), n = await e.next();
+                        return t
+                    }
+                    async resolveFully() {
+                        let e = await this.next();
+                        for (; !e.done;) e = await this.next()
+                    }
+                    async resolveWhile(e) {
+                        let t = await this.next(),
+                            n = e(t.value);
+                        for (; !t.done && n;) t = await this.next(), n = e(t.value)
+                    }
+                    handleErrors(e) {
+                        return new ZA(this, e)
+                    }
+                    filter(e) {
+                        return new XA(this, e)
+                    }
+                    map(e) {
+                        return new QA(this, e)
+                    }
+                    mapAsync(e) {
+                        return new JA(this, e)
+                    }
+                    serialMapAsync(e) {
+                        return new JA(this, e).serial()
+                    }
+                    flatmap(e) {
+                        return new tE(this, e)
+                    }
+                    async forEachAsync(e) {
+                        return this.map(e).resolveFully()
+                    }
+                    async serialForEach(e) {
+                        return this.serialMapAsync(e).resolveWhile(e => !0 === e)
+                    }
+                    rowMajorBatch(e, t = !0) {
+                        return new YA(this, e, t)
+                    }
+                    columnMajorBatch(e, t = !0, n = FA) {
+                        return this.rowMajorBatch(e, t).map(e => _A(e, n))
+                    }
+                    concatenate(e, t) {
+                        return new nE(BA([this, e]), t)
+                    }
+                    take(e) {
+                        return e < 0 || null == e ? this : new KA(this, e)
+                    }
+                    skip(e) {
+                        return e < 0 || null == e ? this : new qA(this, e)
+                    }
+                    prefetch(e) {
+                        return new aE(this, e)
+                    }
+                    shuffle(e, t) {
+                        return new sE(this, e, t)
+                    }
+                    serial() {
+                        return new HA(this)
+                    }
+                },
+                jA = class extends VA {
+                    constructor(e) {
+                        super(), this.items = e, this.trav = 0
+                    }
+                    summary() {
+                        return `Array of ${this.items.length} items`
+                    }
+                    async next() {
+                        if (this.trav >= this.items.length) return {
+                            value: null,
+                            done: !0
+                        };
+                        const e = this.items[this.trav];
+                        return this.trav++, {
+                            value: OA(e),
+                            done: !1
+                        }
+                    }
+                },
+                GA = class extends VA {
+                    constructor(e) {
+                        super(), this.nextFn = e
+                    }
+                    summary() {
+                        return "Function call"
+                    }
+                    async next() {
+                        try {
+                            return this.nextFn()
+                        } catch (e) {
+                            throw e.message = "Error thrown while iterating through a dataset: " + e.message, e
+                        }
+                    }
+                },
+                HA = class extends VA {
+                    constructor(e) {
+                        super(), this.upstream = e, this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Serial"
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    async serialNext() {
+                        return this.upstream.next()
+                    }
+                },
+                qA = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.maxCount = t, this.count = 0, this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Skip"
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    async serialNext() {
+                        for (; this.count++ < this.maxCount;) {
+                            const e = await this.upstream.next();
+                            if (e.done) return e;
+                            au(e.value)
+                        }
+                        return this.upstream.next()
+                    }
+                },
+                KA = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.maxCount = t, this.count = 0
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Take"
+                    }
+                    async next() {
+                        return this.count++ >= this.maxCount ? {
+                            value: null,
+                            done: !0
+                        } : this.upstream.next()
+                    }
+                },
+                YA = class extends VA {
+                    constructor(e, t, n = !0) {
+                        super(), this.upstream = e, this.batchSize = t, this.enableSmallLastBatch = n, this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> RowMajorBatch"
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    async serialNext() {
+                        const e = [];
+                        for (; e.length < this.batchSize;) {
+                            const t = await this.upstream.next();
+                            if (t.done) return this.enableSmallLastBatch && e.length > 0 ? {
+                                value: e,
+                                done: !1
+                            } : {
+                                value: null,
+                                done: !0
+                            };
+                            e.push(t.value)
+                        }
+                        return {
+                            value: e,
+                            done: !1
+                        }
+                    }
+                },
+                XA = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.predicate = t, this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Filter"
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    async serialNext() {
+                        for (;;) {
+                            const e = await this.upstream.next();
+                            if (e.done || this.predicate(e.value)) return e;
+                            au(e.value)
+                        }
+                    }
+                },
+                QA = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.transform = t
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Map"
+                    }
+                    async next() {
+                        const e = await this.upstream.next();
+                        if (e.done) return {
+                            value: null,
+                            done: !0
+                        };
+                        const t = us.getTensorsInContainer(e.value),
+                            n = this.transform(e.value),
+                            r = us.getTensorsInContainer(n);
+                        for (const e of t) us.isTensorInList(e, r) || e.dispose();
+                        return {
+                            value: n,
+                            done: !1
+                        }
+                    }
+                },
+                ZA = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.handler = t, this.count = 0, this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> handleErrors"
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    async serialNext() {
+                        for (;;) try {
+                            return await this.upstream.next()
+                        } catch (e) {
+                            if (!this.handler(e)) return {
+                                value: null,
+                                done: !0
+                            }
+                        }
+                    }
+                },
+                JA = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.transform = t
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> AsyncMap"
+                    }
+                    async next() {
+                        const e = await this.upstream.next();
+                        if (e.done) return {
+                            value: null,
+                            done: !0
+                        };
+                        const t = us.getTensorsInContainer(e.value),
+                            n = await this.transform(e.value),
+                            r = us.getTensorsInContainer(n);
+                        for (const e of t) us.isTensorInList(e, r) || e.dispose();
+                        return {
+                            value: n,
+                            done: !1
+                        }
+                    }
+                },
+                eE = class extends VA {
+                    constructor() {
+                        super(), this.outputQueue = new $A, this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    async serialNext() {
+                        for (; 0 === this.outputQueue.length();)
+                            if (!await this.pump()) return {
+                                value: null,
+                                done: !0
+                            };
+                        return {
+                            value: this.outputQueue.shift(),
+                            done: !1
+                        }
+                    }
+                },
+                tE = class extends eE {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.transform = t
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Flatmap"
+                    }
+                    async pump() {
+                        const e = await this.upstream.next();
+                        if (e.done) return !1;
+                        const t = us.getTensorsInContainer(e.value),
+                            n = this.transform(e.value),
+                            r = us.getTensorsInContainer(n);
+                        this.outputQueue.pushAll(n);
+                        for (const e of t) us.isTensorInList(e, r) || e.dispose();
+                        return !0
+                    }
+                },
+                nE = class extends VA {
+                    constructor(e, t) {
+                        super(), this.baseErrorHandler = t, this.lastRead = null, this.iterator = null, this.moreIterators = e
+                    }
+                    summary() {
+                        return "TODO: fill in upstream of chained summaries -> Chained"
+                    }
+                    async next() {
+                        return this.lastRead = this.readFromChain(this.lastRead), this.lastRead
+                    }
+                    async readFromChain(e) {
+                        if (await e, null == this.iterator) {
+                            const e = await this.moreIterators.next();
+                            if (e.done) return {
+                                value: null,
+                                done: !0
+                            };
+                            this.iterator = e.value, null != this.baseErrorHandler && (this.iterator = this.iterator.handleErrors(this.baseErrorHandler))
+                        }
+                        const t = await this.iterator.next();
+                        return t.done ? (this.iterator = null, this.readFromChain(e)) : t
+                    }
+                };
+            (WA = UA || (UA = {}))[WA.FAIL = 0] = "FAIL", WA[WA.SHORTEST = 1] = "SHORTEST", WA[WA.LONGEST = 2] = "LONGEST";
+            var rE = class extends VA {
+                    constructor(e, t = UA.FAIL) {
+                        super(), this.iterators = e, this.mismatchMode = t, this.count = 0, this.currentPromise = null
+                    }
+                    summary() {
+                        return "{TODO: fill in upstream of zip summaries} -> Zip"
+                    }
+                    async nextState(e) {
+                        await e;
+                        let t = 0,
+                            n = 0;
+                        const r = await MA(this.iterators, (function (e) {
+                            if (e instanceof VA) {
+                                return {
+                                    value: e.next().then(e => (t++, e.done && n++, e.value)),
+                                    recurse: !1
+                                }
+                            }
+                            return {
+                                value: null,
+                                recurse: !0
+                            }
+                        }));
+                        if (t === n) return {
+                            value: null,
+                            done: !0
+                        };
+                        if (n > 0) switch (this.mismatchMode) {
+                            case UA.FAIL:
+                                throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);
+                            case UA.SHORTEST:
+                                return {
+                                    value: null, done: !0
+                                }
+                        }
+                        return this.count++, {
+                            value: r,
+                            done: !1
+                        }
+                    }
+                    async next() {
+                        return this.currentPromise = this.nextState(this.currentPromise), this.currentPromise
+                    }
+                },
+                aE = class extends VA {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.bufferSize = t, this.buffer = new PA(t)
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Prefetch"
+                    }
+                    refill() {
+                        for (; !this.buffer.isFull();) {
+                            const e = this.upstream.next();
+                            this.buffer.push(e)
+                        }
+                    }
+                    next() {
+                        return this.refill(), this.buffer.shift()
+                    }
+                },
+                sE = class extends aE {
+                    constructor(e, t, n) {
+                        super(e, t), this.upstream = e, this.windowSize = t, this.upstreamExhausted = !1, this.random = TA.alea(n || va.now().toString()), this.lastRead = Promise.resolve({
+                            value: null,
+                            done: !1
+                        })
+                    }
+                    async next() {
+                        return this.lastRead = this.lastRead.then(() => this.serialNext()), this.lastRead
+                    }
+                    randomInt(e) {
+                        return Math.floor(this.random() * e)
+                    }
+                    chooseIndex() {
+                        return this.randomInt(this.buffer.length())
+                    }
+                    async serialNext() {
+                        for (this.upstreamExhausted || this.refill(); !this.buffer.isEmpty();) {
+                            const e = this.chooseIndex(),
+                                t = await this.buffer.shuffleExcise(e);
+                            if (!t.done) return this.refill(), t;
+                            this.upstreamExhausted = !0
+                        }
+                        return {
+                            value: null,
+                            done: !0
+                        }
+                    }
+                },
+                iE = class {
+                    constructor() {
+                        this.size = null
+                    }
+                    batch(e, t = !0) {
+                        const n = this;
+                        let r;
+                        return va.assert(e > 0, () => "batchSize needs to be positive, but it is\n      " + e), r = this.size === 1 / 0 || null == this.size ? this.size : t ? Math.ceil(this.size / e) : Math.floor(this.size / e), oE(async () => (await n.iterator()).columnMajorBatch(e, t, cE), r)
+                    }
+                    concatenate(e) {
+                        const t = this;
+                        let n;
+                        return n = this.size === 1 / 0 || e.size === 1 / 0 ? 1 / 0 : null != this.size && null != e.size ? this.size + e.size : null, oE(async () => (await t.iterator()).concatenate(await e.iterator()), n)
+                    }
+                    filter(e) {
+                        const t = this;
+                        let n;
+                        return n = this.size === 1 / 0 ? 1 / 0 : null, oE(async () => (await t.iterator()).filter(t => ru(() => e(t))), n)
+                    }
+                    async forEachAsync(e) {
+                        return (await this.iterator()).forEachAsync(e)
+                    }
+                    map(e) {
+                        const t = this;
+                        return oE(async () => (await t.iterator()).map(t => ru(() => e(t))), this.size)
+                    }
+                    mapAsync(e) {
+                        const t = this;
+                        return oE(async () => (await t.iterator()).mapAsync(e), this.size)
+                    }
+                    prefetch(e) {
+                        if (null == e) throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");
+                        const t = this;
+                        return oE(async () => (await t.iterator()).prefetch(e), this.size)
+                    }
+                    repeat(e) {
+                        const t = this;
+                        let n;
+                        return n = null != this.size && e > 0 ? this.size * e : 0 === e ? 0 : null != this.size && (void 0 === e || e < 0) ? 1 / 0 : null, oE(async () => {
+                            const n = zA(async () => ({
+                                value: await t.iterator(),
+                                done: !1
+                            }));
+                            return r = n.take(e), new nE(r, a);
+                            var r, a
+                        }, n)
+                    }
+                    skip(e) {
+                        const t = this;
+                        let n;
+                        return n = null != this.size && e >= 0 && this.size >= e ? this.size - e : null != this.size && (this.size < e || void 0 === e || e < 0) ? 0 : null, oE(async () => (await t.iterator()).skip(e), n)
+                    }
+                    shuffle(e, t, n = !0) {
+                        if (null == e || e < 0) throw null == this.size ? new RangeError("`Dataset.shuffle()` requires bufferSize to be specified.") : new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);
+                        const r = this,
+                            a = CA.alea(t || va.now().toString());
+                        return oE(async () => {
+                            let t = a.int32();
+                            return n && (t += a.int32()), (await r.iterator()).shuffle(e, t.toString())
+                        }, this.size)
+                    }
+                    take(e) {
+                        const t = this;
+                        let n;
+                        return n = null != this.size && this.size > e ? e : null != this.size && this.size <= e ? this.size : null, oE(async () => (await t.iterator()).take(e), n)
+                    }
+                    async toArray() {
+                        if (this.size === 1 / 0) throw new Error("Can not convert infinite data stream to array.");
+                        return (await this.iterator()).toArray()
+                    }
+                    async toArrayForTest() {
+                        if (this.size === 1 / 0) throw new Error("Can not convert infinite data stream to array.");
+                        return (await this.iterator()).toArrayForTest()
+                    }
+                };
+
+            function oE(e, t = null) {
+                return new class extends iE {
+                    constructor() {
+                        super(...arguments), this.size = t
+                    }
+                    async iterator() {
+                        return e()
+                    }
+                }
+            }
+
+            function uE(e) {
+                return oE(async () => BA(e), e.length)
+            }
+
+            function lE(e) {
+                if (!DA(e)) throw new Error("The argument to zip() must be an object or array.");
+                let t;
+                if (Array.isArray(e))
+                    for (let n = 0; n < e.length; n++) t = null == t ? e[n].size : Math.min(t, e[n].size);
+                else if (e instanceof Object)
+                    for (const n in e) t = null == t ? e[n].size : Math.min(t, e[n].size);
+                return oE(async () => function (e, t = UA.FAIL) {
+                    return new rE(e, t)
+                }(await MA(e, e => {
+                    if (e instanceof iE) return {
+                        value: e.iterator(),
+                        recurse: !1
+                    };
+                    if (DA(e)) return {
+                        value: null,
+                        recurse: !0
+                    };
+                    throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")
+                }), UA.SHORTEST), t)
+            }
+
+            function cE(e) {
+                if (null === e) return null;
+                const t = e[0];
+                if (null == (n = t) || null === (r = n) || "object" != typeof r && "function" != typeof r || Array.isArray(n) || "object" == typeof n && n instanceof Ya || va.isTypedArray(n)) {
+                    return {
+                        value: function (e) {
+                            if (0 === e.length) throw new Error("Can't make a batch of zero elements.");
+                            return e[0] instanceof Ya ? Gh(e) : Ls(e)
+                        }(e),
+                        recurse: !1
+                    }
+                }
+                var n, r;
+                return {
+                    value: null,
+                    recurse: !0
+                }
+            }
+            iE.MAX_BUFFER_SIZE = 1e4;
+            var hE = class extends iE {
+                    constructor(e) {
+                        super(), this.input = e
+                    }
+                    async iterator() {
+                        return (await this.input.iterator()).decodeUTF8().split("\n").map(e => (e.endsWith("\r") && (e = e.slice(0, -1)), e))
+                    }
+                },
+                pE = Symbol("out"),
+                dE = Symbol("field"),
+                fE = Symbol("quote"),
+                mE = Symbol("quoteafterquote"),
+                gE = Symbol("quoteinquote"),
+                yE = class extends iE {
+                    constructor(e, t) {
+                        super(), this.input = e, this.hasHeader = !0, this.fullColumnNames = null, this.columnNamesValidated = !1, this.columnConfigs = null, this.configuredColumnsOnly = !1, this.delimiter = ",", this.delimWhitespace = !1, this.base = new hE(e), t || (t = {}), this.hasHeader = !1 !== t.hasHeader, this.fullColumnNames = t.columnNames, this.columnConfigs = t.columnConfigs, this.configuredColumnsOnly = t.configuredColumnsOnly, t.delimWhitespace ? (va.assert(null == t.delimiter, () => "Delimiter should not be provided when delimWhitespace is true."), this.delimWhitespace = !0, this.delimiter = " ") : this.delimiter = t.delimiter ? t.delimiter : ","
+                    }
+                    async columnNames() {
+                        return this.columnNamesValidated || await this.setColumnNames(), this.configuredColumnsOnly ? Object.keys(this.columnConfigs) : this.fullColumnNames
+                    }
+                    async setColumnNames() {
+                        const e = await this.maybeReadHeaderLine();
+                        if (!this.fullColumnNames && !e) throw new Error("Column names must be provided if there is no header line.");
+                        this.fullColumnNames && e && va.assert(e.length === this.fullColumnNames.length, () => "The length of provided columnNames (" + this.fullColumnNames.length.toString() + ") does not match the length of the header line read from file (" + e.length.toString() + ")."), this.fullColumnNames || (this.fullColumnNames = e);
+                        const t = this.fullColumnNames.reduce((e, t) => (e[t] = e[t] + 1 || 1, e), {}),
+                            n = Object.keys(t).filter(e => t[e] > 1);
+                        if (va.assert(0 === n.length, () => "Duplicate column names found: " + n.toString()), this.columnConfigs)
+                            for (const e of Object.keys(this.columnConfigs)) {
+                                if (-1 === this.fullColumnNames.indexOf(e)) throw new Error('The key "' + e + '" provided in columnConfigs does not match any of the column names (' + this.fullColumnNames.toString() + ").")
+                            }
+                        this.columnNamesValidated = !0
+                    }
+                    async maybeReadHeaderLine() {
+                        if (this.hasHeader) {
+                            const e = await this.base.iterator(),
+                                t = await e.next();
+                            if (t.done) throw new Error("No data was found for CSV parsing.");
+                            const n = t.value;
+                            return this.parseRow(n, !1)
+                        }
+                        return null
+                    }
+                    async iterator() {
+                        this.columnNamesValidated || await this.setColumnNames();
+                        let e = await this.base.iterator();
+                        return this.hasHeader && (e = e.skip(1)), e.map(e => this.makeDataElement(e))
+                    }
+                    makeDataElement(e) {
+                        const t = this.parseRow(e),
+                            n = {},
+                            r = {};
+                        for (let a = 0; a < this.fullColumnNames.length; a++) {
+                            const s = this.fullColumnNames[a],
+                                i = this.columnConfigs ? this.columnConfigs[s] : null;
+                            if (!this.configuredColumnsOnly || i) {
+                                const o = t[a];
+                                let u = null;
+                                if ("" === o)
+                                    if (i && void 0 !== i.default) u = i.default;
+                                    else {
+                                        if (i && (i.required || i.isLabel)) throw new Error(`Required column ${s} is empty in this line: ${e}`);
+                                        u = void 0
+                                    }
+                                else {
+                                    const e = Number(o);
+                                    if (isNaN(e)) u = i && "bool" === i.dtype ? this.getBoolean(o) : o;
+                                    else if (i && i.dtype) switch (i.dtype) {
+                                        case "float32":
+                                            u = e;
+                                            break;
+                                        case "int32":
+                                            u = Math.floor(e);
+                                            break;
+                                        case "bool":
+                                            u = this.getBoolean(o);
+                                            break;
+                                        default:
+                                            u = e
+                                    } else u = e
+                                }
+                                i && i.isLabel ? r[s] = u : n[s] = u
+                            }
+                        }
+                        return 0 === Object.keys(r).length ? n : {
+                            xs: n,
+                            ys: r
+                        }
+                    }
+                    getBoolean(e) {
+                        return "1" === e || "true" === e.toLowerCase() ? 1 : 0
+                    }
+                    parseRow(e, t = !0) {
+                        const n = [];
+                        let r = 0;
+                        const a = e.length;
+                        let s = pE;
+                        for (let t = 0; t < a; t++) switch (s) {
+                            case pE:
+                                switch (e.charAt(t)) {
+                                    case '"':
+                                        r = t + 1, s = fE;
+                                        break;
+                                    case this.delimiter:
+                                        if (r = t + 1, " " === this.delimiter && this.delimWhitespace) break;
+                                        n.push(""), s = pE;
+                                        break;
+                                    default:
+                                        s = dE, r = t
+                                }
+                                break;
+                            case dE:
+                                switch (e.charAt(t)) {
+                                    case this.delimiter:
+                                        n.push(e.substring(r, t)), s = pE, r = t + 1
+                                }
+                                break;
+                            case fE:
+                                switch (e.charAt(t)) {
+                                    case '"':
+                                        s = mE
+                                }
+                                break;
+                            case mE:
+                                switch (e.charAt(t)) {
+                                    case this.delimiter:
+                                        n.push(e.substring(r, t - 1)), s = pE, r = t + 1;
+                                        break;
+                                    case '"':
+                                        s = fE;
+                                        break;
+                                    default:
+                                        s = gE
+                                }
+                                break;
+                            case gE:
+                                switch (e.charAt(t)) {
+                                    case '"':
+                                        s = fE
+                                }
+                        }
+                        if (s === mE ? n.push(e.substring(r, a - 1)) : n.push(e.substring(r)), t && n.length !== this.fullColumnNames.length) throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);
+                        return n
+                    }
+                },
+                bE = class extends VA {
+                    constructor(e) {
+                        super(), this.microphoneConfig = e, this.isClosed = !1, this.fftSize = e.fftSize || 1024;
+                        const t = Math.log2(this.fftSize);
+                        if (this.fftSize < 0 || t < 4 || t > 14 || !Number.isInteger(t)) throw new Error("Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got " + this.fftSize);
+                        if (this.numFrames = e.numFramesPerSpectrogram || 43, this.sampleRateHz = e.sampleRateHz, this.columnTruncateLength = e.columnTruncateLength || this.fftSize, this.audioTrackConstraints = e.audioTrackConstraints, this.smoothingTimeConstant = e.smoothingTimeConstant || 0, this.includeSpectrogram = !1 !== e.includeSpectrogram, this.includeWaveform = !0 === e.includeWaveform, !this.includeSpectrogram && !this.includeWaveform) throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")
+                    }
+                    summary() {
+                        return "microphone"
+                    }
+                    static async create(e = {}) {
+                        if (Ye().get("IS_NODE")) throw new Error("microphone API is only supported in browser environment.");
+                        const t = new bE(e);
+                        return await t.start(), t
+                    }
+                    async start() {
+                        try {
+                            this.stream = await navigator.mediaDevices.getUserMedia({
+                                audio: null == this.audioTrackConstraints || this.audioTrackConstraints,
+                                video: !1
+                            })
+                        } catch (e) {
+                            throw new Error("Error thrown while initializing video stream: " + e.message)
+                        }
+                        if (!this.stream) throw new Error("Could not obtain audio from microphone.");
+                        const e = window.AudioContext || window.webkitAudioContext;
+                        if (this.audioContext = new e, this.sampleRateHz) {
+                            if (this.audioContext.sampleRate !== this.sampleRateHz) throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`)
+                        } else this.sampleRateHz = this.audioContext.sampleRate;
+                        const t = this.audioContext.createMediaStreamSource(this.stream);
+                        this.analyser = this.audioContext.createAnalyser(), this.analyser.fftSize = 2 * this.fftSize, this.analyser.smoothingTimeConstant = this.smoothingTimeConstant, t.connect(this.analyser), this.freqData = new Float32Array(this.fftSize), this.timeData = new Float32Array(this.fftSize)
+                    }
+                    async next() {
+                        if (this.isClosed) return {
+                            value: null,
+                            done: !0
+                        };
+                        let e, t;
+                        const n = await this.getAudioData();
+                        if (this.includeSpectrogram) {
+                            const t = this.flattenQueue(n.freqDataQueue);
+                            e = this.getTensorFromAudioDataArray(t, [this.numFrames, this.columnTruncateLength, 1])
+                        }
+                        if (this.includeWaveform) {
+                            const e = this.flattenQueue(n.timeDataQueue);
+                            t = this.getTensorFromAudioDataArray(e, [this.numFrames * this.fftSize, 1])
+                        }
+                        return {
+                            value: {
+                                spectrogram: e,
+                                waveform: t
+                            },
+                            done: !1
+                        }
+                    }
+                    async capture() {
+                        return (await this.next()).value
+                    }
+                    async getAudioData() {
+                        const e = [],
+                            t = [];
+                        let n = 0;
+                        return new Promise(r => {
+                            const a = setInterval(() => {
+                                this.includeSpectrogram && (this.analyser.getFloatFrequencyData(this.freqData), this.freqData[0] === -1 / 0 && r({
+                                    freqDataQueue: e,
+                                    timeDataQueue: t
+                                }), e.push(this.freqData.slice(0, this.columnTruncateLength))), this.includeWaveform && (this.analyser.getFloatTimeDomainData(this.timeData), t.push(this.timeData.slice())), ++n === this.numFrames && (clearInterval(a), r({
+                                    freqDataQueue: e,
+                                    timeDataQueue: t
+                                }))
+                            }, this.fftSize / this.sampleRateHz * 1e3)
+                        })
+                    }
+                    stop() {
+                        this.isClosed || (this.isClosed = !0, this.analyser.disconnect(), this.audioContext.close(), null != this.stream && this.stream.getTracks().length > 0 && this.stream.getTracks()[0].stop())
+                    }
+                    toArray() {
+                        throw new Error("Can not convert infinite audio stream to array.")
+                    }
+                    getSampleRate() {
+                        return this.sampleRateHz
+                    }
+                    flattenQueue(e) {
+                        const t = e[0].length,
+                            n = new Float32Array(e.length * t);
+                        return e.forEach((e, r) => n.set(e, r * t)), n
+                    }
+                    getTensorFromAudioDataArray(e, t) {
+                        const n = new Float32Array(va.sizeFromShape(t));
+                        return n.set(e, n.length - e.length), Ls(n, t)
+                    }
+                },
+                vE = class extends VA {
+                    constructor(e, t) {
+                        if (super(), this.webcamVideoElement = e, this.webcamConfig = t, this.isClosed = !0, this.resize = !1, this.needToResize())
+                            if (this.resize = !0, this.cropSize = [this.webcamConfig.resizeHeight, this.webcamConfig.resizeWidth], this.cropBoxInd = Yh([0], "int32"), this.webcamConfig.centerCrop) {
+                                const e = 1 * this.webcamConfig.resizeWidth / this.webcamVideoElement.width,
+                                    t = 1 * this.webcamConfig.resizeHeight / this.webcamVideoElement.height,
+                                    n = (1 - e) / 2,
+                                    r = (1 - t) / 2,
+                                    a = n + e,
+                                    s = t + r;
+                                this.cropBox = Xh([r, n, s, a], [1, 4])
+                            } else this.cropBox = Xh([0, 0, 1, 1], [1, 4])
+                    }
+                    summary() {
+                        return "webcam"
+                    }
+                    static async create(e, t = {}) {
+                        if (Ye().get("IS_NODE")) throw new Error("tf.data.webcam is only supported in browser environment.");
+                        if (!e) {
+                            if (e = document.createElement("video"), !t.resizeWidth || !t.resizeHeight) throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");
+                            e.width = t.resizeWidth, e.height = t.resizeHeight
+                        }
+                        const n = new vE(e, t);
+                        return await n.start(), n
+                    }
+                    async start() {
+                        this.webcamConfig.facingMode && va.assert("user" === this.webcamConfig.facingMode || "environment" === this.webcamConfig.facingMode, () => `Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);
+                        try {
+                            this.stream = await navigator.mediaDevices.getUserMedia({
+                                video: {
+                                    deviceId: this.webcamConfig.deviceId,
+                                    facingMode: this.webcamConfig.facingMode ? this.webcamConfig.facingMode : "user",
+                                    width: this.webcamVideoElement.width,
+                                    height: this.webcamVideoElement.height
+                                }
+                            })
+                        } catch (e) {
+                            throw e.message = "Error thrown while initializing video stream: " + e.message, e
+                        }
+                        if (!this.stream) throw new Error("Could not obtain video from webcam.");
+                        try {
+                            this.webcamVideoElement.srcObject = this.stream
+                        } catch (e) {
+                            console.log(e), this.webcamVideoElement.src = window.URL.createObjectURL(this.stream)
+                        }
+                        return this.webcamVideoElement.play(), this.isClosed = !1, new Promise(e => {
+                            this.webcamVideoElement.onloadedmetadata = () => {
+                                e()
+                            }
+                        })
+                    }
+                    async next() {
+                        if (this.isClosed) return {
+                            value: null,
+                            done: !0
+                        };
+                        let e;
+                        try {
+                            e = eo.fromPixels(this.webcamVideoElement)
+                        } catch (e) {
+                            throw new Error("Error thrown converting video to pixels: " + JSON.stringify(e))
+                        }
+                        if (!this.resize) return {
+                            value: e,
+                            done: !1
+                        };
+                        try {
+                            return {
+                                value: this.cropAndResizeFrame(e),
+                                done: !1
+                            }
+                        } catch (e) {
+                            throw new Error("Error thrown cropping the video: " + e.message)
+                        } finally {
+                            e.dispose()
+                        }
+                    }
+                    needToResize() {
+                        return !(!this.webcamConfig.resizeWidth || !this.webcamConfig.resizeHeight || this.webcamVideoElement.width === this.webcamConfig.resizeWidth && this.webcamVideoElement.height === this.webcamConfig.resizeHeight)
+                    }
+                    cropAndResizeFrame(e) {
+                        return ru(() => {
+                            const t = Bl(Ai(e, "float32"), 0);
+                            let n;
+                            n = Nd.cropAndResize(t, this.cropBox, this.cropBoxInd, this.cropSize, "bilinear");
+                            const r = n.shape;
+                            return qu(n, r.slice(1))
+                        })
+                    }
+                    async capture() {
+                        return (await this.next()).value
+                    }
+                    stop() {
+                        this.stream.getTracks().forEach(e => e.stop());
+                        try {
+                            this.webcamVideoElement.srcObject = null
+                        } catch (e) {
+                            console.log(e), this.webcamVideoElement.src = null
+                        }
+                        this.isClosed = !0
+                    }
+                    toArray() {
+                        throw new Error("Can not convert infinite video stream to array.")
+                    }
+                },
+                xE = class {},
+                wE = class extends VA {
+                    split(e) {
+                        return new kE(this, e)
+                    }
+                },
+                kE = class extends wE {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.impl = new SE(e, t)
+                    }
+                    summary() {
+                        return this.impl.summary()
+                    }
+                    async next() {
+                        return this.impl.next()
+                    }
+                },
+                SE = class extends eE {
+                    constructor(e, t) {
+                        super(), this.upstream = e, this.separator = t, this.carryover = ""
+                    }
+                    summary() {
+                        return `${this.upstream.summary()} -> Split('${this.separator}')`
+                    }
+                    async pump() {
+                        const e = await this.upstream.next();
+                        if (e.done) return "" !== this.carryover && (this.outputQueue.push(this.carryover), this.carryover = "", !0);
+                        const t = e.value.split(this.separator);
+                        t[0] = this.carryover + t[0];
+                        for (const e of t.slice(0, -1)) this.outputQueue.push(e);
+                        return this.carryover = t[t.length - 1], !0
+                    }
+                },
+                IE = class extends VA {
+                    decodeUTF8() {
+                        return new AE(this)
+                    }
+                },
+                AE = class extends wE {
+                    constructor(e) {
+                        super(), this.upstream = e, this.impl = new EE(e)
+                    }
+                    summary() {
+                        return this.impl.summary()
+                    }
+                    async next() {
+                        return this.impl.next()
+                    }
+                },
+                EE = class extends eE {
+                    constructor(e) {
+                        if (super(), this.upstream = e, Ye().get("IS_BROWSER")) this.decoder = new TextDecoder("utf-8");
+                        else {
+                            const {
+                                StringDecoder: e
+                            } = V();
+                            this.decoder = new e("utf8")
+                        }
+                    }
+                    summary() {
+                        return this.upstream.summary() + " -> Utf8"
+                    }
+                    async pump() {
+                        const e = await this.upstream.next();
+                        let t, n;
+                        return !e.done && (t = e.value, n = Ye().get("IS_BROWSER") ? this.decoder.decode(t, {
+                            stream: !0
+                        }) : this.decoder.write(i.from(t.buffer)), this.outputQueue.push(n), !0)
+                    }
+                },
+                NE = class extends IE {
+                    constructor(e, t = {}) {
+                        super(), this.file = e, this.options = t, va.assert(e instanceof Uint8Array || !!Ye().get("IS_BROWSER") && (e instanceof File || e instanceof Blob), () => "FileChunkIterator only supports File, Blob and Uint8Array right now."), this.offset = t.offset || 0, this.chunkSize = t.chunkSize || 1048576
+                    }
+                    summary() {
+                        return "FileChunks " + this.file
+                    }
+                    async next() {
+                        if (this.offset >= (this.file instanceof Uint8Array ? this.file.byteLength : this.file.size)) return {
+                            value: null,
+                            done: !0
+                        };
+                        const e = new Promise((e, t) => {
+                            const n = this.offset + this.chunkSize;
+                            if (this.file instanceof Uint8Array) e(new Uint8Array(this.file.slice(this.offset, n)));
+                            else {
+                                const r = new FileReader;
+                                r.onload = n => {
+                                    let a = r.result;
+                                    if (a instanceof ArrayBuffer && (a = new Uint8Array(a)), !(a instanceof Uint8Array)) return t(new TypeError("FileReader returned unknown type."));
+                                    e(a)
+                                }, r.onabort = e => t(new Error("Aborted")), r.onerror = e => t(new Error(e.type));
+                                const a = this.file.slice(this.offset, n);
+                                r.readAsArrayBuffer(a)
+                            }
+                            this.offset = n
+                        });
+                        return {
+                            value: await e,
+                            done: !1
+                        }
+                    }
+                };
+            var CE = e => ({
+                method: e.method,
+                headers: e.headers,
+                body: e.body,
+                mode: e.mode,
+                credentials: e.credentials,
+                cache: e.cache,
+                redirect: e.redirect,
+                referrer: e.referrer,
+                integrity: e.integrity
+            });
+
+            function TE(e) {
+                return "string" == typeof e && "file://" === e.substr(0, 7)
+            }
+            var RE = class extends xE {
+                    constructor(e, t = {}) {
+                        super(), this.input = e, this.options = t
+                    }
+                    async iterator() {
+                        if (TE(this.input) && Ye().get("IS_NODE")) {
+                            const e = j();
+                            this.input = e.readFileSync(this.input.substr(7))
+                        }
+                        return new NE(this.input, this.options)
+                    }
+                },
+                _E = class extends xE {
+                    constructor(e, t = {}) {
+                        super(), this.url = e, this.fileOptions = t
+                    }
+                    async iterator() {
+                        return TE(this.url) ? new RE(this.url, this.fileOptions).iterator() : async function (e, t = {}, n) {
+                            let r, a;
+                            "string" == typeof e ? r = e : (r = e.url, a = CE(e));
+                            const s = await (n || va.fetch)(r, a);
+                            if (s.ok) {
+                                const e = new Uint8Array(await s.arrayBuffer());
+                                return new NE(e, t)
+                            }
+                            throw new Error(s.statusText)
+                        }(this.url, this.fileOptions)
+                    }
+                };
+
+            function FE(e, t = {}) {
+                return new yE(new _E(e), t)
+            }
+
+            function ME(e) {
+                const t = zA(e);
+                return oE(async () => t)
+            }
+
+            function DE(e) {
+                return oE(async () => {
+                    const t = await e();
+                    return zA(() => t.next())
+                })
+            }
+            async function OE(e, t) {
+                return vE.create(e, t)
+            }
+            async function LE(e) {
+                return bE.create(e)
+            }
+            var PE = "3.13.0";
+
+            function $E(e, t) {
+                Array.isArray(e) || (e = [e]), e.forEach(e => {
+                    null != e && va.assert("complex64" !== e.dtype, () => t + " does not support complex64 tensors in the CPU backend.")
+                })
+            }
+            var BE = jf.whereImpl,
+                zE = class extends Z {
+                    constructor() {
+                        super(), this.blockSize = 48, this.firstUse = !0, this.data = new Q(this, eu())
+                    }
+                    nextDataId() {
+                        return zE.nextDataId++
+                    }
+                    write(e, t, n) {
+                        this.firstUse && (this.firstUse = !1, Ye().get("IS_NODE") && jd.warn("\n============================\nHi there 👋. Looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, which binds to TensorFlow C++, by running npm i @tensorflow/tfjs-node, or npm i @tensorflow/tfjs-node-gpu if you have CUDA. Then call require('@tensorflow/tfjs-node'); (-gpu suffix for CUDA) at the start of your program. Visit https://github.com/tensorflow/tfjs-node for more details.\n============================"));
+                        const r = {
+                            id: this.nextDataId()
+                        };
+                        return this.data.set(r, {
+                            values: e,
+                            dtype: n,
+                            refCount: 1
+                        }), r
+                    }
+                    makeTensorInfo(e, t, n) {
+                        let r;
+                        if ("string" === t && null != n && n.length > 0 && va.isString(n[0])) {
+                            const a = n.map(e => va.encodeString(e));
+                            r = this.write(a, e, t)
+                        } else r = this.write(n, e, t);
+                        return {
+                            dataId: r,
+                            shape: e,
+                            dtype: t
+                        }
+                    }
+                    refCount(e) {
+                        if (this.data.has(e)) {
+                            return this.data.get(e).refCount
+                        }
+                        return 0
+                    }
+                    incRef(e) {
+                        this.data.get(e).refCount++
+                    }
+                    decRef(e) {
+                        if (this.data.has(e)) {
+                            this.data.get(e).refCount--
+                        }
+                    }
+                    move(e, t, n, r, a) {
+                        this.data.set(e, {
+                            values: t,
+                            dtype: r,
+                            refCount: a
+                        })
+                    }
+                    numDataIds() {
+                        return this.data.numDataIds()
+                    }
+                    async read(e) {
+                        return this.readSync(e)
+                    }
+                    readSync(e) {
+                        const {
+                            dtype: t,
+                            complexTensorInfos: n
+                        } = this.data.get(e);
+                        if ("complex64" === t) {
+                            const e = this.readSync(n.real.dataId),
+                                t = this.readSync(n.imag.dataId);
+                            return jd.mergeRealAndImagArrays(e, t)
+                        }
+                        return this.data.get(e).values
+                    }
+                    bufferSync(e) {
+                        const t = this.readSync(e.dataId);
+                        let n = t;
+                        if ("string" === e.dtype) try {
+                            n = t.map(e => va.decodeString(e))
+                        } catch (e) {
+                            throw new Error("Failed to decode encoded string bytes into utf-8")
+                        }
+                        return Ii(e.shape, e.dtype, n)
+                    }
+                    makeOutput(e, t, n) {
+                        const r = this.write(e, t, n);
+                        return eu().makeTensorFromDataId(r, t, n, this)
+                    }
+                    disposeData(e, t = !1) {
+                        if (this.data.has(e)) {
+                            if (this.data.get(e).refCount--, !t && this.data.get(e).refCount > 0) return !1;
+                            const {
+                                complexTensorInfos: n
+                            } = this.data.get(e);
+                            null != n && (this.disposeData(n.real.dataId, !0), this.disposeData(n.imag.dataId, !0)), this.data.delete(e)
+                        }
+                        return !0
+                    }
+                    disposeIntermediateTensorInfo(e) {
+                        this.disposeData(e.dataId)
+                    }
+                    async time(e) {
+                        const t = va.now();
+                        e();
+                        return {
+                            kernelMs: va.now() - t
+                        }
+                    }
+                    memory() {
+                        return {
+                            unreliable: !0,
+                            reasons: ["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]
+                        }
+                    }
+                    where(e) {
+                        $E([e], "where");
+                        const t = this.readSync(e.dataId);
+                        return BE(e.shape, t)
+                    }
+                    dispose() {}
+                    floatPrecision() {
+                        return 32
+                    }
+                    epsilon() {
+                        return super.epsilon()
+                    }
+                };
+            zE.nextDataId = 0;
+            var UE = {};
+
+            function WE(e) {
+                const t = new Float32Array(e.length);
+                for (let n = 0; n < e.length; ++n) t[n] = Math.abs(e[n]);
+                return t
+            }
+            w(UE, {
+                addImpl: () => nN,
+                bincountImpl: () => iN,
+                bincountReduceImpl: () => oN,
+                ceilImpl: () => hN,
+                concatImpl: () => fN,
+                equalImpl: () => mN,
+                expImpl: () => bN,
+                expm1Impl: () => wN,
+                floorImpl: () => IN,
+                gatherNdImpl: () => NN,
+                gatherV2Impl: () => CN,
+                greaterEqualImpl: () => FN,
+                greaterImpl: () => TN,
+                lessEqualImpl: () => $N,
+                lessImpl: () => ON,
+                linSpaceImpl: () => UN,
+                logImpl: () => WN,
+                maxImpl: () => GN,
+                maximumImpl: () => HN,
+                minimumImpl: () => YN,
+                multiplyImpl: () => ZN,
+                negImpl: () => nC,
+                notEqualImpl: () => aC,
+                prodImpl: () => cC,
+                rangeImpl: () => pC,
+                rsqrtImpl: () => dC,
+                sigmoidImpl: () => gC,
+                simpleAbsImpl: () => WE,
+                sliceImpl: () => vC,
+                sparseFillEmptyRowsImpl: () => kC,
+                sparseReshapeImpl: () => SC,
+                sparseSegmentReductionImpl: () => IC,
+                sqrtImpl: () => AC,
+                squaredDifferenceImpl: () => CC,
+                stridedSliceImpl: () => _C,
+                stringNGramsImpl: () => FC,
+                stringSplitImpl: () => DC,
+                stringToHashBucketFastImpl: () => OC,
+                subImpl: () => LC,
+                tileImpl: () => zC,
+                topKImpl: () => VC,
+                transposeImpl: () => oC,
+                uniqueImpl: () => jC
+            });
+            var VE = {
+                kernelName: et,
+                backendName: "cpu",
+                kernelFunc: e => {
+                    const {
+                        x: t
+                    } = e.inputs, n = e.backend;
+                    $E(t, "abs");
+                    let r = new Float32Array(va.sizeFromShape(t.shape));
+                    return r = WE(n.data.get(t.dataId).values), n.makeOutput(r, t.shape, t.dtype)
+                }
+            };
+
+            function jE(e) {
+                return (t, n, r, a, s) => {
+                    const i = jd.assertAndGetBroadcastShape(t, n),
+                        o = i.length,
+                        u = va.computeStrides(i),
+                        l = va.sizeFromShape(i),
+                        c = va.getTypedArrayFromDType(s, l),
+                        h = t.length,
+                        p = n.length,
+                        d = va.computeStrides(t),
+                        f = va.computeStrides(n),
+                        m = jd.getBroadcastDims(t, i),
+                        g = jd.getBroadcastDims(n, i);
+                    if (m.length + g.length === 0)
+                        for (let t = 0; t < c.length; ++t) c[t] = e(r[t % r.length], a[t % a.length]);
+                    else
+                        for (let t = 0; t < c.length; ++t) {
+                            const n = va.indexToLoc(t, o, u),
+                                s = n.slice(-h);
+                            m.forEach(e => s[e] = 0);
+                            const i = va.locToIndex(s, h, d),
+                                l = n.slice(-p);
+                            g.forEach(e => l[e] = 0);
+                            const y = va.locToIndex(l, p, f);
+                            c[t] = e(r[i], a[y])
+                        }
+                    return [c, i]
+                }
+            }
+
+            function GE(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    real: r,
+                    imag: a
+                } = t, s = n.data.get(r.dataId).values, i = n.data.get(a.dataId).values, o = n.makeTensorInfo(r.shape, "complex64");
+                return n.data.get(o.dataId).complexTensorInfos = {
+                    real: n.makeTensorInfo(r.shape, "float32", s),
+                    imag: n.makeTensorInfo(a.shape, "float32", i)
+                }, o
+            }
+            var HE = {
+                kernelName: Et,
+                backendName: "cpu",
+                kernelFunc: GE
+            };
+
+            function qE(e, t, n = "float32") {
+                if ("complex64" === n) {
+                    return GE({
+                        inputs: {
+                            real: qE(e, t, "float32"),
+                            imag: qE(e, t, "float32")
+                        },
+                        backend: e
+                    })
+                }
+                const r = va.makeZerosTypedArray(va.sizeFromShape(t), n);
+                return e.makeTensorInfo(t, n, r)
+            }
+
+            function KE(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    x: r
+                } = t;
+                return n.incRef(r.dataId), {
+                    dataId: r.dataId,
+                    shape: r.shape,
+                    dtype: r.dtype
+                }
+            }
+            var YE = {
+                kernelName: fn,
+                backendName: "cpu",
+                kernelFunc: KE
+            };
+
+            function XE(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    input: r
+                } = t, a = n.data.get(r.dataId).complexTensorInfos.real, s = n.data.get(a.dataId).values;
+                return n.makeTensorInfo(a.shape, a.dtype, s)
+            }
+            var QE = {
+                kernelName: ir,
+                backendName: "cpu",
+                kernelFunc: XE
+            };
+
+            function ZE(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    dtype: s
+                } = r;
+                if ("complex64" === s) {
+                    if ("complex64" === a.dtype) return KE({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    const e = qE(n, a.shape, a.dtype),
+                        t = ZE({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                dtype: "float32"
+                            }
+                        }),
+                        r = GE({
+                            inputs: {
+                                real: t,
+                                imag: e
+                            },
+                            backend: n
+                        });
+                    return n.disposeIntermediateTensorInfo(e), n.disposeIntermediateTensorInfo(t), r
+                }
+                if ("complex64" === a.dtype) {
+                    const e = XE({
+                            inputs: {
+                                input: a
+                            },
+                            backend: n
+                        }),
+                        t = ZE({
+                            inputs: {
+                                x: e
+                            },
+                            backend: n,
+                            attrs: {
+                                dtype: s
+                            }
+                        });
+                    return n.disposeIntermediateTensorInfo(e), t
+                }
+                if (!va.hasEncodingLoss(a.dtype, s)) {
+                    const e = KE({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    return {
+                        dataId: e.dataId,
+                        shape: e.shape,
+                        dtype: s
+                    }
+                }
+                if ("int32" === s) {
+                    const e = n.data.get(a.dataId).values,
+                        t = Int32Array.from(e);
+                    return n.makeTensorInfo(a.shape, "int32", t)
+                }
+                if ("bool" === s) {
+                    const e = n.data.get(a.dataId).values,
+                        t = va.toTypedArray([0], a.dtype),
+                        [r, s] = jE((e, t) => e !== t ? 1 : 0)(a.shape, [], e, t, "bool");
+                    return n.makeTensorInfo(s, "bool", r)
+                }
+                throw new Error(`Error in Cast: failed to cast ${a.dtype} to ${s}`)
+            }
+            var JE = {
+                kernelName: St,
+                backendName: "cpu",
+                kernelFunc: ZE
+            };
+
+            function eN(e, t, n, r) {
+                return null == n ? ({
+                    inputs: n,
+                    backend: a
+                }) => {
+                    const {
+                        a: s,
+                        b: i
+                    } = n, o = a;
+                    $E([s, i], e);
+                    const u = o.data.get(s.dataId).values,
+                        l = o.data.get(i.dataId).values,
+                        c = "string" === s.dtype ? jd.fromUint8ToStringArray(u) : u,
+                        h = "string" === s.dtype ? jd.fromUint8ToStringArray(l) : l,
+                        p = r || s.dtype,
+                        [d, f] = t(s.shape, i.shape, c, h, p);
+                    return o.makeTensorInfo(f, p, d)
+                } : ({
+                    inputs: e,
+                    backend: a
+                }) => {
+                    const {
+                        a: s,
+                        b: i
+                    } = e, o = a;
+                    if ("complex64" === s.dtype || "complex64" === i.dtype) {
+                        const e = ZE({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: o,
+                                attrs: {
+                                    dtype: "complex64"
+                                }
+                            }),
+                            t = o.data.get(e.dataId),
+                            r = t.complexTensorInfos.real,
+                            a = t.complexTensorInfos.imag,
+                            u = o.data.get(r.dataId).values,
+                            l = o.data.get(a.dataId).values,
+                            c = ZE({
+                                inputs: {
+                                    x: i
+                                },
+                                backend: o,
+                                attrs: {
+                                    dtype: "complex64"
+                                }
+                            }),
+                            h = o.data.get(c.dataId),
+                            p = h.complexTensorInfos.real,
+                            d = h.complexTensorInfos.imag,
+                            f = o.data.get(p.dataId).values,
+                            m = o.data.get(d.dataId).values,
+                            [g, y, b] = n(s.shape, i.shape, u, l, f, m),
+                            v = o.makeTensorInfo(b, "float32", g),
+                            x = o.makeTensorInfo(b, "float32", y),
+                            w = GE({
+                                inputs: {
+                                    real: v,
+                                    imag: x
+                                },
+                                backend: o
+                            });
+                        return o.disposeIntermediateTensorInfo(e), o.disposeIntermediateTensorInfo(c), o.disposeIntermediateTensorInfo(v), o.disposeIntermediateTensorInfo(x), w
+                    } {
+                        const e = o.data.get(s.dataId).values,
+                            n = o.data.get(i.dataId).values,
+                            a = r || s.dtype,
+                            [u, l] = t(s.shape, i.shape, e, n, a);
+                        return o.makeTensorInfo(l, a, u)
+                    }
+                }
+            }
+
+            function tN(e) {
+                return (t, n, r, a, s, i) => {
+                    const o = jd.assertAndGetBroadcastShape(t, n),
+                        u = va.sizeFromShape(o),
+                        l = o.length,
+                        c = va.computeStrides(o),
+                        h = va.getTypedArrayFromDType("float32", u),
+                        p = va.getTypedArrayFromDType("float32", u),
+                        d = jd.getBroadcastDims(t, o),
+                        f = jd.getBroadcastDims(n, o),
+                        m = jd.mergeRealAndImagArrays(r, a),
+                        g = jd.mergeRealAndImagArrays(s, i),
+                        y = t.length,
+                        b = va.computeStrides(t),
+                        v = n.length,
+                        x = va.computeStrides(n);
+                    if (d.length + f.length === 0)
+                        for (let t = 0; t < h.length; t++) {
+                            const n = t % m.length,
+                                r = t % g.length,
+                                a = e(m[2 * n], m[2 * n + 1], g[2 * r], g[2 * r + 1]);
+                            h[t] = a.real, p[t] = a.imag
+                        } else
+                            for (let t = 0; t < h.length; t++) {
+                                const n = va.indexToLoc(t, l, c),
+                                    r = n.slice(-y);
+                                d.forEach(e => r[e] = 0);
+                                const a = va.locToIndex(r, y, b),
+                                    s = n.slice(-v);
+                                f.forEach(e => s[e] = 0);
+                                const i = va.locToIndex(s, v, x),
+                                    o = e(m[2 * a], m[2 * a + 1], g[2 * i], g[2 * i + 1]);
+                                h[t] = o.real, p[t] = o.imag
+                            }
+                    return [h, p, o]
+                }
+            }
+            var nN = jE((e, t) => e + t),
+                rN = tN((e, t, n, r) => ({
+                    real: e + n,
+                    imag: t + r
+                })),
+                aN = eN(rt, nN, rN),
+                sN = {
+                    kernelName: rt,
+                    backendName: "cpu",
+                    kernelFunc: aN
+                };
+
+            function iN(e, t, n, r, a) {
+                const s = va.sizeFromShape(r),
+                    i = va.makeZerosTypedArray(a, n);
+                for (let n = 0; n < e.length; n++) {
+                    const r = e[n];
+                    if (r < 0) throw new Error("Input x must be non-negative!");
+                    r >= a || (i[r] += s > 0 ? t[n] : 1)
+                }
+                return i
+            }
+
+            function oN(e, t, n, r = !1) {
+                const a = e.shape[0],
+                    s = e.shape[1],
+                    i = Ii([a, n], t.dtype);
+                for (let o = 0; o < a; o++)
+                    for (let a = 0; a < s; a++) {
+                        const s = e.get(o, a);
+                        if (s < 0) throw new Error("Input x must be non-negative!");
+                        s >= n || (r ? i.set(1, o, s) : t.size > 0 ? i.set(i.get(o, s) + t.get(o, a), o, s) : i.set(i.get(o, s) + 1, o, s))
+                    }
+                return i
+            }
+
+            function uN(e) {
+                return (t, n, r) => {
+                    const a = va.getTypedArrayFromDType(n, t.length);
+                    for (let n = 0; n < t.length; ++n) a[n] = e(t[n], r);
+                    return a
+                }
+            }
+
+            function lN(e, t, n) {
+                return ({
+                    inputs: r,
+                    attrs: a,
+                    backend: s
+                }) => {
+                    const {
+                        x: i
+                    } = r;
+                    if ($E(i, e), "string" === i.dtype || "string" === n) throw new Error("unaryKernelFunc does not support string input/output");
+                    const o = s,
+                        u = o.data.get(i.dataId).values,
+                        l = va.sizeFromShape(i.shape),
+                        c = n || i.dtype,
+                        h = va.getArrayFromDType(c, l);
+                    for (let e = 0; e < l; ++e) h[e] = t(u[e], a);
+                    return o.makeTensorInfo(i.shape, c, h)
+                }
+            }
+
+            function cN(e, t, n) {
+                return ({
+                    inputs: r,
+                    attrs: a,
+                    backend: s
+                }) => {
+                    const {
+                        x: i
+                    } = r;
+                    if ($E(i, e), "string" === i.dtype || "string" === n) throw new Error("unaryKernelFunc does not support string input/output");
+                    const o = s,
+                        u = o.data.get(i.dataId).values,
+                        l = n || i.dtype,
+                        c = t(u, l, a);
+                    return o.makeTensorInfo(i.shape, l, c)
+                }
+            }
+            var hN = uN(e => Math.ceil(e)),
+                pN = cN(It, hN),
+                dN = {
+                    kernelName: It,
+                    backendName: "cpu",
+                    kernelFunc: pN
+                };
+
+            function fN(e, t, n, r) {
+                const a = va.getArrayFromDType(n, va.sizeFromShape(t));
+                if (r && "string" !== n) {
+                    let t = 0;
+                    e.forEach(e => {
+                        const n = va.sizeFromShape(e.shape);
+                        a.set(e.vals, t), t += n
+                    })
+                } else {
+                    let r = 0;
+                    e.forEach(e => {
+                        const s = "string" === n ? jd.fromUint8ToStringArray(e.vals) : e.vals;
+                        let i = 0;
+                        for (let n = 0; n < e.shape[0]; ++n) {
+                            const o = n * t[1] + r;
+                            for (let t = 0; t < e.shape[1]; ++t) a[o + t] = s[i++]
+                        }
+                        r += e.shape[1]
+                    })
+                }
+                return a
+            }
+            var mN = jE((e, t) => e === t ? 1 : 0),
+                gN = eN(Jt, mN, null, "bool"),
+                yN = {
+                    kernelName: Jt,
+                    backendName: "cpu",
+                    kernelFunc: gN
+                },
+                bN = uN(e => Math.exp(e)),
+                vN = cN(en, bN, "float32"),
+                xN = {
+                    kernelName: en,
+                    backendName: "cpu",
+                    kernelFunc: vN
+                },
+                wN = uN(e => Math.expm1(e)),
+                kN = cN(nn, wN),
+                SN = {
+                    kernelName: nn,
+                    backendName: "cpu",
+                    kernelFunc: kN
+                },
+                IN = uN(e => Math.floor(e)),
+                AN = cN(on, IN),
+                EN = {
+                    kernelName: on,
+                    backendName: "cpu",
+                    kernelFunc: AN
+                };
+
+            function NN(e, t, n, r, a, s, i, o, u) {
+                const l = Ii([r, s], n);
+                for (let n = 0; n < r; n++) {
+                    const r = [];
+                    let c = 0;
+                    for (let t = 0; t < a; t++) {
+                        const s = e[n * a + t];
+                        c += s * i[t], r.push(s)
+                    }
+                    if (c < 0 || c >= u / s) throw new Error(`Invalid indices: ${r} does not index into ${o}`);
+                    for (let e = 0; e < s; e++) l.values[n * s + e] = t.get(...t.indexToLoc(c * s + e))
+                }
+                return l
+            }
+
+            function CN(e, t, n) {
+                const r = Ii(n, e.dtype);
+                for (let n = 0; n < r.size; ++n) {
+                    const a = r.indexToLoc(n).slice(),
+                        s = a[0],
+                        i = a[2],
+                        o = t.locToIndex([s, i]);
+                    a[2] = t.values[o];
+                    const u = e.locToIndex(a);
+                    0 <= u && u < e.values.length && (r.values[n] = e.values[u])
+                }
+                return r
+            }
+            var TN = jE((e, t) => e > t ? 1 : 0),
+                RN = eN(pn, TN, null, "bool"),
+                _N = {
+                    kernelName: pn,
+                    backendName: "cpu",
+                    kernelFunc: RN
+                },
+                FN = jE((e, t) => e >= t ? 1 : 0),
+                MN = eN(dn, FN, null, "bool"),
+                DN = {
+                    kernelName: dn,
+                    backendName: "cpu",
+                    kernelFunc: MN
+                },
+                ON = jE((e, t) => e < t ? 1 : 0),
+                LN = eN(wn, ON, null, "bool"),
+                PN = {
+                    kernelName: wn,
+                    backendName: "cpu",
+                    kernelFunc: LN
+                },
+                $N = jE((e, t) => e <= t ? 1 : 0),
+                BN = eN(kn, $N, null, "bool"),
+                zN = {
+                    kernelName: kn,
+                    backendName: "cpu",
+                    kernelFunc: BN
+                };
+
+            function UN(e, t, n) {
+                const r = (t - e) / (n - 1),
+                    a = va.makeZerosTypedArray(n, "float32");
+                a[0] = e;
+                for (let e = 1; e < a.length; e++) a[e] = a[e - 1] + r;
+                return a
+            }
+            var WN = uN(e => Math.log(e)),
+                VN = cN(In, WN),
+                jN = {
+                    kernelName: In,
+                    backendName: "cpu",
+                    kernelFunc: VN
+                };
+
+            function GN(e, t, n, r) {
+                const a = va.getTypedArrayFromDType(r, va.sizeFromShape(n));
+                for (let n = 0; n < a.length; ++n) {
+                    const r = n * t;
+                    let s = e[r];
+                    for (let n = 0; n < t; ++n) {
+                        const t = e[r + n];
+                        (Number.isNaN(t) || t > s) && (s = t)
+                    }
+                    a[n] = s
+                }
+                return a
+            }
+            var HN = jE((e, t) => Math.max(e, t)),
+                qN = eN(Mn, HN),
+                KN = {
+                    kernelName: Mn,
+                    backendName: "cpu",
+                    kernelFunc: qN
+                },
+                YN = jE((e, t) => Math.min(e, t)),
+                XN = eN(Un, YN),
+                QN = {
+                    kernelName: Un,
+                    backendName: "cpu",
+                    kernelFunc: XN
+                },
+                ZN = jE((e, t) => e * t),
+                JN = tN((e, t, n, r) => ({
+                    real: e * n - t * r,
+                    imag: e * r + t * n
+                })),
+                eC = eN(Gn, ZN, JN),
+                tC = {
+                    kernelName: Gn,
+                    backendName: "cpu",
+                    kernelFunc: eC
+                };
+
+            function nC(e, t, n) {
+                const r = va.createScalarValue(-1, n);
+                return ZN([], t, r, e, n)
+            }
+            var rC = {
+                    kernelName: Hn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            x: r
+                        } = t;
+                        $E(r, "neg");
+                        const a = n.data.get(r.dataId).values,
+                            [s, i] = nC(a, r.shape, r.dtype);
+                        return n.makeTensorInfo(i, r.dtype, s)
+                    }
+                },
+                aC = jE((e, t) => e !== t ? 1 : 0),
+                sC = eN(qn, aC, null, "bool"),
+                iC = {
+                    kernelName: qn,
+                    backendName: "cpu",
+                    kernelFunc: sC
+                };
+
+            function oC(e, t, n, r, a) {
+                const s = t.length,
+                    i = va.sizeFromShape(t),
+                    o = va.computeStrides(t),
+                    u = va.computeStrides(a),
+                    l = va.getTypedArrayFromDType(n, va.sizeFromShape(a));
+                for (let t = 0; t < i; ++t) {
+                    const n = va.indexToLoc(t, s, o),
+                        a = new Array(n.length);
+                    for (let e = 0; e < a.length; e++) a[e] = n[r[e]];
+                    l[va.locToIndex(a, s, u)] = e[t]
+                }
+                return l
+            }
+
+            function uC(e) {
+                const {
+                    inputs: t,
+                    attrs: n,
+                    backend: r
+                } = e, {
+                    x: a
+                } = t, {
+                    perm: s
+                } = n;
+                $E(a, "transpose");
+                const i = a.shape.length,
+                    o = new Array(i);
+                for (let e = 0; e < o.length; e++) o[e] = a.shape[s[e]];
+                const u = oC(r.data.get(a.dataId).values, a.shape, a.dtype, s, o);
+                return {
+                    dataId: r.write(u, o, a.dtype),
+                    shape: o,
+                    dtype: a.dtype
+                }
+            }
+            var lC = {
+                kernelName: Yr,
+                backendName: "cpu",
+                kernelFunc: uC
+            };
+
+            function cC(e, t, n, r) {
+                const [a, s] = jd.computeOutAndReduceShapes(e, r), i = cs(t, "int32"), o = va.makeZerosTypedArray(va.sizeFromShape(a), i), u = va.sizeFromShape(s);
+                for (let e = 0; e < o.length; ++e) {
+                    const t = e * u;
+                    let r = 1;
+                    for (let e = 0; e < u; ++e) r *= n[t + e];
+                    o[e] = r
+                }
+                return {
+                    outVals: o,
+                    outShape: a,
+                    outDtype: i
+                }
+            }
+            var hC = {
+                kernelName: ar,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r;
+                    $E(a, "prod");
+                    const o = a.shape.length,
+                        u = va.parseAxisParam(s, a.shape),
+                        l = jd.getAxesPermutation(u, o);
+                    let c = u,
+                        h = a;
+                    const p = [];
+                    null != l && (h = uC({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: l
+                        }
+                    }), p.push(h), c = jd.getInnerMostAxes(c.length, o));
+                    const d = n.data.get(h.dataId).values,
+                        {
+                            outVals: f,
+                            outShape: m,
+                            outDtype: g
+                        } = cC(h.shape, h.dtype, d, c);
+                    let y = m;
+                    return i && (y = jd.expandShapeToKeepDim(m, u)), p.forEach(e => n.disposeIntermediateTensorInfo(e)), n.makeTensorInfo(y, g, f)
+                }
+            };
+
+            function pC(e, t, n, r) {
+                if (e === t || e < t && n < 0 || t < e && n > 1) return va.makeZerosTypedArray(0, r);
+                const a = Math.abs(Math.ceil((t - e) / n)),
+                    s = va.makeZerosTypedArray(a, r);
+                t < e && 1 === n && (n = -1), s[0] = e;
+                for (let e = 1; e < s.length; e++) s[e] = s[e - 1] + n;
+                return s
+            }
+            var dC = uN(e => 1 / Math.sqrt(e)),
+                fC = cN(yr, dC),
+                mC = {
+                    kernelName: yr,
+                    backendName: "cpu",
+                    kernelFunc: fC
+                },
+                gC = uN(e => 1 / (1 + Math.exp(-e))),
+                yC = lN(Ar, e => 1 / (1 + Math.exp(-e))),
+                bC = {
+                    kernelName: Ar,
+                    backendName: "cpu",
+                    kernelFunc: yC
+                };
+
+            function vC(e, t, n, r, a) {
+                const s = fo.isSliceContinous(r, t, n),
+                    i = va.sizeFromShape(n),
+                    o = va.computeStrides(r);
+                if (s) {
+                    const n = fo.computeFlatOffset(t, o);
+                    return "string" === a ? e.slice(n, n + i) : e.subarray(n, n + i)
+                }
+                const u = Ii(r, a, "string" === a ? jd.fromUint8ToStringArray(e) : e),
+                    l = Ii(n, a);
+                for (let e = 0; e < l.size; ++e) {
+                    const n = l.indexToLoc(e),
+                        r = n.map((e, n) => e + t[n]);
+                    l.set(u.get(...r), ...n)
+                }
+                return "string" === a ? jd.fromStringArrayToUint8(l.values) : l.values
+            }
+
+            function xC(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    begin: s,
+                    size: i
+                } = r;
+                $E(a, "slice");
+                const [o, u] = fo.parseSliceParams(a, s, i);
+                fo.assertParamsValid(a, o, u);
+                const l = vC(n.data.get(a.dataId).values, o, u, a.shape, a.dtype);
+                return n.makeTensorInfo(u, a.dtype, l)
+            }
+            var wC = {
+                kernelName: wr,
+                backendName: "cpu",
+                kernelFunc: xC
+            };
+
+            function kC(e, t, n, r, a, s, i) {
+                const o = t[0],
+                    u = s[0],
+                    l = new Array(u),
+                    c = new Array(o),
+                    h = t[1];
+                if (0 === u) {
+                    if (0 !== o) throw new Error(jd.getSparseFillEmptyRowsIndicesDenseShapeMismatch(o));
+                    return [va.getArrayFromDType(n, 0), [0, h], va.getArrayFromDType(a, 0), l, c]
+                }
+                let p = !0,
+                    d = 0;
+                const f = new Array(u).fill(0);
+                for (let t = 0; t < o; ++t) {
+                    const n = e[t * h];
+                    if (n < 0) throw new Error(jd.getSparseFillEmptyRowsNegativeIndexErrorMessage(t, n));
+                    if (n >= u) throw new Error(jd.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(t, n, u));
+                    ++f[n], p = p && n >= d, d = n
+                }
+                let m = !0;
+                for (let e = 0; e < u; ++e) {
+                    const t = 0 === f[e];
+                    l[e] = t, m = m && !t, f[e] = Math.max(f[e], 1), e > 0 && (f[e] += f[e - 1])
+                }
+                if (m && p) {
+                    const t = e,
+                        n = r;
+                    for (let e = 0; e < o; ++e) c[e] = e;
+                    return [t, [o, h], n, l, c]
+                } {
+                    const t = f[u - 1],
+                        s = va.getArrayFromDType(n, t * h),
+                        p = va.getArrayFromDType(a, t),
+                        d = new Array(u).fill(0);
+                    for (let t = 0; t < o; ++t) {
+                        const n = e[t * h],
+                            a = d[n],
+                            i = (0 === n ? 0 : f[n - 1]) + a;
+                        d[n]++;
+                        for (let n = 0; n < h; ++n) s[i * h + n] = e[t * h + n];
+                        p[i] = r[t], c[t] = i
+                    }
+                    for (let e = 0; e < u; ++e) {
+                        if (0 === d[e]) {
+                            const t = 0 === e ? 0 : f[e - 1];
+                            s[t * h + 0] = e;
+                            for (let e = 1; e < h; ++e) s[t * h + e] = 0;
+                            p[t] = i
+                        }
+                    }
+                    return [s, [t, h], p, l, c]
+                }
+            }
+
+            function SC(e, t, n, r, a) {
+                const s = va.sizeFromShape(r),
+                    i = t[0],
+                    o = a.length,
+                    u = [];
+                let l = 1,
+                    c = -1;
+                for (let e = 0; e < o; ++e) {
+                    const t = a[e];
+                    if (-1 === t) {
+                        if (-1 !== c) throw new Error(jd.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(c, e));
+                        c = e, u.push(1)
+                    } else {
+                        if (t < 0) throw new Error(jd.getSparseReshapeNegativeOutputDimErrorMessage(e, t));
+                        l *= t, u.push(t)
+                    }
+                }
+                if (-1 !== c) {
+                    if (l <= 0) throw new Error(jd.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());
+                    const e = Math.trunc(s / l);
+                    if (l * e !== s) throw new Error(jd.getSparseReshapeInputOutputMultipleErrorMessage(r, u));
+                    u[c] = e
+                }
+                if (va.sizeFromShape(u) !== s) throw new Error(jd.getSparseReshapeInputOutputMismatchErrorMessage(r, u));
+                const h = r.length,
+                    p = [];
+                if (h > 0) {
+                    p[h - 1] = 1;
+                    for (let e = h - 2; e >= 0; --e) p[e] = p[e + 1] * r[e + 1]
+                }
+                const d = [];
+                if (o > 0) {
+                    d[o - 1] = 1;
+                    for (let e = o - 2; e >= 0; --e) d[e] = d[e + 1] * u[e + 1]
+                }
+                const f = va.getArrayFromDType(n, i * o);
+                for (let t = 0; t < i; ++t) {
+                    let n = 0;
+                    for (let r = 0; r < h; ++r) n += e[t * h + r] * p[r];
+                    for (let e = 0; e < o; ++e) f[t * o + e] = Math.trunc(n / d[e]), n %= d[e]
+                }
+                return [f, [i, o], u]
+            }
+
+            function IC(e, t, n, r, a, s = !1, i = 0) {
+                const o = r.length,
+                    u = [t[0], e.length / t[0]],
+                    l = u[1],
+                    c = o > 0 ? a[o - 1] + 1 : 0;
+                if (c < 0) throw new Error(jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+                const h = t.slice();
+                h[0] = c;
+                const p = h.reduce((e, t) => e * t, 1),
+                    d = va.getArrayFromDType(n, p);
+                if (0 === o) return c > 0 && d.fill(i), [d, h];
+                if (c <= 0) throw new Error(jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+                let f = 0,
+                    m = 1,
+                    g = 0,
+                    y = a[f];
+                for (;;) {
+                    let t = 0;
+                    if (m < o) {
+                        if (t = a[m], y === t) {
+                            ++m;
+                            continue
+                        }
+                        if (y >= t) throw new Error(jd.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())
+                    }
+                    if (y < 0 || y >= c) throw new Error(jd.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y, c));
+                    y > g && d.fill(i, g * l, y * l);
+                    for (let t = f; t < m; ++t) {
+                        const n = r[t];
+                        if (n < 0 || n >= u[0]) throw new Error(jd.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(t, r[t], u[0]));
+                        for (let t = 0; t < l; t++) d[y * l + t] += e[n * l + t]
+                    }
+                    if (s)
+                        for (let e = 0; e < l; e++) d[y * l + e] /= m - f;
+                    if (f = m, ++m, g = y + 1, y = t, m > o) break
+                }
+                return g < c && d.fill(i, g * l, c * l), [d, h]
+            }
+            var AC = uN(e => Math.sqrt(e)),
+                EC = lN(Nr, e => Math.sqrt(e)),
+                NC = {
+                    kernelName: Nr,
+                    backendName: "cpu",
+                    kernelFunc: EC
+                },
+                CC = jE((e, t) => {
+                    const n = e - t;
+                    return n * n
+                }),
+                TC = eN(Pr, CC),
+                RC = {
+                    kernelName: Pr,
+                    backendName: "cpu",
+                    kernelFunc: TC
+                };
+
+            function _C(e, t, n, r) {
+                const a = Ii(e, t.dtype);
+                for (let e = 0; e < a.size; e++) {
+                    const s = a.indexToLoc(e),
+                        i = new Array(s.length);
+                    for (let e = 0; e < i.length; e++) i[e] = s[e] * n[e] + r[e];
+                    a.set(t.get(...i), ...s)
+                }
+                return a
+            }
+
+            function FC(e, t, n, r, a, s, i, o) {
+                return new class {
+                    constructor(e, t, n, r, a, s) {
+                        this.separator = va.encodeString(e), this.nGramWidths = t, this.leftPad = va.encodeString(n), this.rightPad = va.encodeString(r), this.padWidth = a, this.preserveShort = s
+                    }
+                    getPadWidth(e) {
+                        return Math.min(this.padWidth < 0 ? e - 1 : this.padWidth, e - 1)
+                    }
+                    getNumNGrams(e, t) {
+                        const n = this.getPadWidth(t);
+                        return Math.max(0, e + 2 * n - t + 1)
+                    }
+                    createNGrams(e, t, n, r, a, s) {
+                        for (let i = 0; i < a; ++i) {
+                            const o = this.getPadWidth(s),
+                                u = Math.max(0, o - i),
+                                l = Math.max(0, o - (a - (i + 1))),
+                                c = s - (u + l),
+                                h = t + (u > 0 ? 0 : i - o);
+                            let p = 0;
+                            p += u * this.leftPad.length;
+                            for (let t = 0; t < c; ++t) p += e[h + t].length;
+                            p += l * this.rightPad.length;
+                            p += (u + l + c - 1) * this.separator.length, n[r + i] = new Uint8Array(p);
+                            const d = n[r + i];
+                            let f = 0;
+                            const m = e => e.forEach(e => d[f++] = e);
+                            for (let e = 0; e < u; ++e) m(this.leftPad), m(this.separator);
+                            for (let t = 0; t < c - 1; ++t) m(e[h + t]), m(this.separator);
+                            if (c > 0) {
+                                m(e[h + c - 1]);
+                                for (let e = 0; e < l; ++e) m(this.separator), m(this.rightPad)
+                            } else {
+                                for (let e = 0; e < l - 1; ++e) m(this.rightPad), m(this.separator);
+                                m(this.rightPad)
+                            }
+                        }
+                    }
+                    compute(e, t) {
+                        const n = e.length,
+                            r = t.length;
+                        if (r > 0) {
+                            let e = t[0];
+                            if (0 !== e) throw new Error("First split value must be 0, got " + e);
+                            for (let a = 1; a < r; ++a) {
+                                let r = t[a] >= e;
+                                if (r = r && t[a] <= n, !r) throw new Error(`Invalid split value ${t[a]}, must be in [${e}, ${n}]`);
+                                e = t[a]
+                            }
+                            if (e !== n) throw new Error(`Last split value must be data size. Expected ${n}, got ${e}`)
+                        }
+                        const a = r - 1,
+                            s = va.getArrayFromDType("int32", r);
+                        if (0 === n || 0 === r) {
+                            const e = new Array(n);
+                            for (let e = 0; e <= a; ++e) s[e] = 0;
+                            return [e, s]
+                        }
+                        s[0] = 0;
+                        for (let e = 1; e <= a; ++e) {
+                            const n = t[e] - t[e - 1];
+                            let r = 0;
+                            this.nGramWidths.forEach(e => {
+                                r += this.getNumNGrams(n, e)
+                            }), this.preserveShort && n > 0 && 0 === r && (r = 1), s[e] = s[e - 1] + r
+                        }
+                        const i = new Array(s[a]);
+                        for (let n = 0; n < a; ++n) {
+                            const r = t[n];
+                            let a = s[n];
+                            if (this.nGramWidths.forEach(s => {
+                                    const o = t[n + 1] - t[n],
+                                        u = this.getNumNGrams(o, s);
+                                    this.createNGrams(e, r, i, a, u, s), a += u
+                                }), this.preserveShort && a === s[n]) {
+                                const s = t[n + 1] - t[n];
+                                if (0 === s) continue;
+                                const o = s + 2 * this.padWidth,
+                                    u = 1;
+                                this.createNGrams(e, r, i, a, u, o)
+                            }
+                        }
+                        return [i, s]
+                    }
+                }(n, r, a, s, i, o).compute(e, t)
+            }
+
+            function MC(e, t, n, r) {
+                if (!e.length) return;
+                if (0 === t.length) {
+                    for (let t = 0; t < e.length; ++t) r.push(e.subarray(t, t + 1));
+                    return
+                }
+                if (1 === t.length) {
+                    const a = t[0];
+                    let s = e.indexOf(a);
+                    for (; - 1 !== s;) {
+                        const t = e.subarray(0, s);
+                        n && 0 === t.length || r.push(t), s = (e = e.subarray(s + 1)).indexOf(a)
+                    }
+                    return void(n && 0 === e.length || r.push(e))
+                }
+                let a = 0;
+                for (let s = 0; s < e.length + 1; s++)
+                    if (s === e.length || -1 !== t.indexOf(e[s])) {
+                        const t = e.subarray(a, s);
+                        n && 0 === t.length || r.push(t), a = s + 1
+                    }
+            }
+
+            function DC(e, t, n) {
+                const r = e.length,
+                    a = [];
+                let s = 0,
+                    i = 0;
+                const o = new Array(r);
+                for (let u = 0; u < r; ++u) {
+                    const r = a.length;
+                    MC(e[u], t, n, a);
+                    const l = a.length - r;
+                    o[u] = l, s += l, i = Math.max(i, l)
+                }
+                const u = va.getArrayFromDType("int32", 2 * s),
+                    l = new Array(s),
+                    c = [r, i];
+                let h = 0;
+                for (let e = 0; e < r; ++e)
+                    for (let t = 0; t < o[e]; ++t) u[2 * h] = e, u[2 * h + 1] = t, l[h] = a[h], ++h;
+                return [u, l, c]
+            }
+
+            function OC(e, t) {
+                const n = va.getArrayFromDType("int32", e.length);
+                for (let r = 0; r < e.length; ++r) n[r] = va.fingerPrint64(e[r]).modulo(t).getLowBitsUnsigned();
+                return n
+            }
+            var LC = jE((e, t) => e - t),
+                PC = tN((e, t, n, r) => ({
+                    real: e - n,
+                    imag: t - r
+                })),
+                $C = eN(Vr, LC, PC),
+                BC = {
+                    kernelName: Vr,
+                    backendName: "cpu",
+                    kernelFunc: $C
+                };
+
+            function zC(e, t) {
+                const n = new Array(e.rank);
+                for (let r = 0; r < n.length; r++) n[r] = e.shape[r] * t[r];
+                const r = Ii(n, e.dtype);
+                for (let t = 0; t < r.values.length; ++t) {
+                    const n = r.indexToLoc(t),
+                        a = new Array(e.rank);
+                    for (let t = 0; t < a.length; t++) a[t] = n[t] % e.shape[t];
+                    const s = e.locToIndex(a);
+                    r.values[t] = e.values[s]
+                }
+                return r
+            }
+            var UC = (e, t) => {
+                const n = t.value - e.value;
+                return 0 === n ? e.index - t.index : n
+            };
+
+            function WC(e, t, n = 0, r = e.length - 1) {
+                for (; r > n;) {
+                    if (r - n > 600) {
+                        const a = r - n + 1,
+                            s = t - n + 1,
+                            i = Math.log(a),
+                            o = .5 * Math.exp(2 * i / 3),
+                            u = .5 * Math.sqrt(i * o * (a - o) / a) * Math.sign(s - a / 2);
+                        WC(e, t, Math.max(n, Math.floor(t - s * o / a + u)), Math.min(r, Math.floor(t + (a - s) * o / a + u)))
+                    }
+                    const a = e[t];
+                    let s = n,
+                        i = r;
+                    for (va.swap(e, n, t), UC(e[r], a) > 0 && va.swap(e, n, r); s < i;) {
+                        for (va.swap(e, s, i), s++, i--; UC(e[s], a) < 0;) s += 1;
+                        for (; UC(e[i], a) > 0;) i -= 1
+                    }
+                    0 === UC(e[n], a) ? va.swap(e, n, i) : (i += 1, va.swap(e, i, r)), i <= t && (n = i + 1), t <= i && (r = i - 1)
+                }
+            }
+
+            function VC(e, t, n, r, a) {
+                const s = t[t.length - 1],
+                    [i, o] = [e.length / s, s],
+                    u = va.getTypedArrayFromDType(n, i * r),
+                    l = va.getTypedArrayFromDType("int32", i * r);
+                for (let t = 0; t < i; t++) {
+                    const n = t * o,
+                        s = e.subarray(n, n + o);
+                    let i = new Array(s.length);
+                    s.forEach((e, t) => i[t] = {
+                        value: e,
+                        index: t
+                    }), r < i.length && (WC(i, r), i = i.slice(0, r)), a && i.sort(UC);
+                    const c = t * r,
+                        h = u.subarray(c, c + r),
+                        p = l.subarray(c, c + r);
+                    for (let e = 0; e < r; e++) h[e] = i[e].value, p[e] = i[e].index
+                }
+                const c = t.slice();
+                return c[c.length - 1] = r, [Ii(c, n, u), Ii(c, "int32", l)]
+            }
+
+            function jC(e, t, n, r) {
+                const a = va.parseAxisParam(t, n)[0],
+                    s = [1, n[0], 1];
+                for (let e = 0; e < a; e++) s[0] *= n[e];
+                s[1] = n[a];
+                for (let e = a + 1; e < n.length; e++) s[2] *= n[e];
+                const i = {},
+                    o = new Int32Array(n[a]),
+                    u = new Ha(s, r, e),
+                    l = [],
+                    c = 1 === s[0] && 1 === s[2];
+                for (let t = 0; t < n[a]; t++) {
+                    let n;
+                    if (c) n = e[t].toString();
+                    else {
+                        const e = [];
+                        for (let n = 0; n < s[0]; n++)
+                            for (let r = 0; r < s[2]; r++) e.push(u.get(n, t, r));
+                        n = e.join(",")
+                    }
+                    if (void 0 !== i[n]) o[t] = i[n];
+                    else {
+                        const e = Object.keys(i).length;
+                        i[n] = e, o[t] = e, l.push(t)
+                    }
+                }
+                const h = s.slice();
+                h[1] = Object.keys(i).length;
+                const p = new Ha(h, r);
+                l.forEach((e, t) => {
+                    for (let n = 0; n < s[0]; n++)
+                        for (let r = 0; r < s[2]; r++) p.set(u.get(n, e, r), n, t, r)
+                });
+                const d = n.slice();
+                return d[a] = h[1], {
+                    outputValues: p.values,
+                    outputShape: d,
+                    indices: o
+                }
+            }
+            du("cpu", () => new zE, 1);
+            var GC = lN(Xt, e => e >= 0 ? e : Math.exp(e) - 1),
+                HC = {
+                    kernelName: Xt,
+                    backendName: "cpu",
+                    kernelFunc: GC
+                };
+
+            function qC(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    alpha: s
+                } = r;
+                $E([a], "leakyRelu");
+                const i = va.sizeFromShape(a.shape),
+                    o = n.data.get(a.dataId).values,
+                    u = va.getTypedArrayFromDType("float32", i);
+                for (let e = 0; e < o.length; e++) u[e] = o[e] < 0 ? s * o[e] : o[e];
+                return n.makeTensorInfo(a.shape, "float32", u)
+            }
+            var KC = {
+                    kernelName: xn,
+                    backendName: "cpu",
+                    kernelFunc: qC
+                },
+                YC = jE((e, t) => e < 0 ? t * e : e);
+
+            function XC(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    x: r,
+                    alpha: a
+                } = t;
+                $E([r, a], "prelu");
+                const s = n.data.get(r.dataId).values,
+                    i = n.data.get(a.dataId).values,
+                    [o, u] = YC(r.shape, a.shape, s, i, "float32");
+                return n.makeTensorInfo(u, "float32", o)
+            }
+            var QC = {
+                    kernelName: rr,
+                    backendName: "cpu",
+                    kernelFunc: XC
+                },
+                ZC = lN(ur, e => Math.max(0, e)),
+                JC = {
+                    kernelName: ur,
+                    backendName: "cpu",
+                    kernelFunc: ZC
+                },
+                eT = lN(fr, e => Math.min(Math.max(0, e), 6)),
+                tT = {
+                    kernelName: fr,
+                    backendName: "cpu",
+                    kernelFunc: eT
+                };
+
+            function nT(e, t, n, r, a) {
+                if ("linear" === n) return KE({
+                    inputs: {
+                        x: t
+                    },
+                    backend: e
+                });
+                if ("relu" === n) return ZC({
+                    inputs: {
+                        x: t
+                    },
+                    backend: e
+                });
+                if ("elu" === n) return GC({
+                    inputs: {
+                        x: t
+                    },
+                    backend: e
+                });
+                if ("relu6" === n) return eT({
+                    inputs: {
+                        x: t
+                    },
+                    backend: e
+                });
+                if ("prelu" === n) return XC({
+                    inputs: {
+                        x: t,
+                        alpha: r
+                    },
+                    backend: e
+                });
+                if ("leakyrelu" === n) return qC({
+                    inputs: {
+                        x: t
+                    },
+                    backend: e,
+                    attrs: {
+                        alpha: a
+                    }
+                });
+                if ("sigmoid" === n) return yC({
+                    inputs: {
+                        x: t
+                    },
+                    backend: e
+                });
+                throw new Error(`Activation ${n} has not been implemented for the CPU backend.`)
+            }
+
+            function rT(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    shape: s
+                } = r, i = va.sizeFromShape(a.shape), o = va.inferFromImplicitShape(s, i), u = va.sizeFromShape(o);
+                va.assert(i === u, () => `The new shape (${o}) has ${u} elements and the old shape (${a.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`), n.incRef(a.dataId);
+                const l = n.data.get(a.dataId);
+                if (null != l.complexTensorInfos) {
+                    const e = l.complexTensorInfos.real,
+                        t = l.complexTensorInfos.imag;
+                    e.shape = o, t.shape = o
+                }
+                return {
+                    dataId: a.dataId,
+                    shape: o,
+                    dtype: a.dtype
+                }
+            }
+            var aT = {
+                kernelName: lr,
+                backendName: "cpu",
+                kernelFunc: rT
+            };
+
+            function sT(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    a: a,
+                    b: s
+                } = t, {
+                    transposeA: i,
+                    transposeB: o
+                } = r;
+                $E([a, s], "matMul");
+                const u = a.shape.length,
+                    l = s.shape.length,
+                    c = i ? a.shape[u - 2] : a.shape[u - 1],
+                    h = o ? s.shape[l - 1] : s.shape[l - 2],
+                    p = i ? a.shape[u - 1] : a.shape[u - 2],
+                    d = o ? s.shape[l - 2] : s.shape[l - 1],
+                    f = a.shape.slice(0, -2),
+                    m = s.shape.slice(0, -2),
+                    g = va.sizeFromShape(f),
+                    y = va.sizeFromShape(m),
+                    b = Yi.assertAndGetBroadcastShape(a.shape.slice(0, -2), s.shape.slice(0, -2)).concat([p, d]);
+                va.assert(c === h, () => `Error in matMul: inner shapes (${c}) and (${h}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);
+                const v = o ? [y, d, h] : [y, h, d],
+                    x = rT({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: i ? [g, c, p] : [g, p, c]
+                        }
+                    }),
+                    w = rT({
+                        inputs: {
+                            x: s
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: v
+                        }
+                    }),
+                    k = i ? x.shape[1] : x.shape[2],
+                    S = i ? x.shape[2] : x.shape[1],
+                    I = o ? w.shape[1] : w.shape[2],
+                    A = Math.max(g, y),
+                    E = n.data.get(x.dataId).values,
+                    N = n.data.get(w.dataId).values,
+                    C = va.computeStrides(x.shape),
+                    T = va.computeStrides(w.shape),
+                    [R, _, F] = i ? [C[0], 1, C[1]] : [C[0], C[1], 1],
+                    [M, D, O] = o ? [1, T[1], T[0]] : [T[1], 1, T[0]],
+                    L = S * I,
+                    P = Ii([A, S, I], x.dtype),
+                    $ = P.values,
+                    B = n.blockSize;
+                for (let e = 0; e < A; e++)
+                    for (let t = 0; t < S; t += B)
+                        for (let n = 0; n < I; n += B)
+                            for (let r = 0; r < k; r += B) {
+                                const a = Math.min(t + B, S),
+                                    s = Math.min(n + B, I),
+                                    i = Math.min(r + B, k);
+                                for (let o = t; o < a; o++)
+                                    for (let t = n; t < s; t++) {
+                                        let n = 0;
+                                        for (let a = r; a < i; a++) {
+                                            const r = Math.min(e, g - 1) * R,
+                                                s = Math.min(e, y - 1) * O;
+                                            n += E[r + o * _ + a * F] * N[a * M + t * D + s]
+                                        }
+                                        $[e * L + (o * I + t)] += n
+                                    }
+                            }
+                return n.disposeIntermediateTensorInfo(x), n.disposeIntermediateTensorInfo(w), n.makeTensorInfo(b, P.dtype, P.values)
+            }
+            var iT = {
+                kernelName: bt,
+                backendName: "cpu",
+                kernelFunc: sT
+            };
+            var oT = {
+                    kernelName: ra,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            a: a,
+                            b: s,
+                            bias: i,
+                            preluActivationWeights: o
+                        } = t, {
+                            transposeA: u,
+                            transposeB: l,
+                            activation: c,
+                            leakyreluAlpha: h
+                        } = r;
+                        let p, d, f;
+                        const m = [];
+                        p = sT({
+                            inputs: {
+                                a: a,
+                                b: s
+                            },
+                            attrs: {
+                                transposeA: u,
+                                transposeB: l
+                            },
+                            backend: n
+                        }), i && (d = aN({
+                            inputs: {
+                                a: p,
+                                b: i
+                            },
+                            backend: n
+                        }), m.push(p), p = d), c && (f = nT(n, p, c, o, h), m.push(p), p = f);
+                        for (const e of m) n.disposeIntermediateTensorInfo(e);
+                        return p
+                    }
+                },
+                uT = lN(tt, e => Math.acos(e)),
+                lT = {
+                    kernelName: tt,
+                    backendName: "cpu",
+                    kernelFunc: uT
+                },
+                cT = lN(nt, e => Math.acosh(e)),
+                hT = {
+                    kernelName: nt,
+                    backendName: "cpu",
+                    kernelFunc: cT
+                };
+            var pT = {
+                kernelName: at,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, r = t;
+                    $E(t, "addN");
+                    const a = r.map(e => n.data.get(e.dataId).values),
+                        s = Ii(r[0].shape, r[0].dtype),
+                        i = s.values;
+                    for (let e = 0; e < r.length; e++) {
+                        const t = a[e];
+                        for (let e = 0; e < i.length; e++) i[e] += t[e]
+                    }
+                    return n.makeTensorInfo(s.shape, s.dtype, s.values)
+                }
+            };
+            var dT = {
+                kernelName: st,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r;
+                    $E(a, "all");
+                    const o = va.parseAxisParam(s, a.shape);
+                    let u = o;
+                    const l = jd.getAxesPermutation(u, a.shape.length);
+                    let c = a;
+                    null != l && (c = uC({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: l
+                        }
+                    }), u = jd.getInnerMostAxes(u.length, a.shape.length)), jd.assertAxesAreInnerMostDims("all", u, c.shape.length);
+                    const [h, p] = jd.computeOutAndReduceShapes(c.shape, u), d = va.sizeFromShape(p), f = va.makeZerosTypedArray(va.sizeFromShape(h), c.dtype), m = n.data.get(c.dataId).values;
+                    for (let e = 0; e < f.length; ++e) {
+                        const t = e * d;
+                        let n = m[t];
+                        for (let e = 0; e < d; ++e) {
+                            const r = m[t + e];
+                            n = n && r
+                        }
+                        f[e] = n
+                    }
+                    null != l && n.disposeIntermediateTensorInfo(c);
+                    const g = n.makeTensorInfo(h, c.dtype, f);
+                    if (i) {
+                        const e = rT({
+                            inputs: {
+                                x: g
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: jd.expandShapeToKeepDim(h, o)
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(g), e
+                    }
+                    return g
+                }
+            };
+            var fT = {
+                kernelName: it,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r;
+                    $E(a, "any");
+                    const o = va.parseAxisParam(s, a.shape);
+                    let u = o;
+                    const l = jd.getAxesPermutation(u, a.shape.length);
+                    let c = a;
+                    null != l && (c = uC({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: l
+                        }
+                    }), u = jd.getInnerMostAxes(u.length, a.shape.length)), jd.assertAxesAreInnerMostDims("any", u, c.shape.length);
+                    const [h, p] = jd.computeOutAndReduceShapes(c.shape, u), d = va.sizeFromShape(p), f = va.makeZerosTypedArray(va.sizeFromShape(h), c.dtype), m = n.data.get(c.dataId).values;
+                    for (let e = 0; e < f.length; ++e) {
+                        const t = e * d;
+                        let n = m[t];
+                        for (let e = 0; e < d; ++e) {
+                            const r = m[t + e];
+                            n = n || r
+                        }
+                        f[e] = n
+                    }
+                    null != l && n.disposeIntermediateTensorInfo(c);
+                    const g = n.makeTensorInfo(h, c.dtype, f);
+                    if (i) {
+                        const e = rT({
+                            inputs: {
+                                x: g
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: jd.expandShapeToKeepDim(h, o)
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(g), e
+                    }
+                    return g
+                }
+            };
+            var mT = {
+                kernelName: ot,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s
+                    } = r;
+                    $E(a, "argMax");
+                    let i = va.parseAxisParam(s, a.shape);
+                    const o = jd.getAxesPermutation(i, a.shape.length);
+                    let u = a;
+                    const l = [];
+                    null != o && (u = uC({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: o
+                        }
+                    }), l.push(u), i = jd.getInnerMostAxes(i.length, u.shape.length)), i = [i[0]], jd.assertAxesAreInnerMostDims("argMax", i, u.shape.length);
+                    const [c, h] = jd.computeOutAndReduceShapes(u.shape, i), p = va.sizeFromShape(c), d = va.makeZerosTypedArray(p, "int32"), f = va.sizeFromShape(h), m = n.data.get(u.dataId).values;
+                    for (let e = 0; e < d.length; ++e) {
+                        const t = e * f;
+                        let n = m[t],
+                            r = 0;
+                        for (let e = 0; e < f; ++e) {
+                            const a = m[t + e];
+                            a > n && (n = a, r = e)
+                        }
+                        d[e] = r
+                    }
+                    return l.forEach(e => n.disposeIntermediateTensorInfo(e)), n.makeTensorInfo(c, "int32", d)
+                }
+            };
+            var gT = {
+                    kernelName: ut,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            axis: s
+                        } = r;
+                        $E(a, "argMin");
+                        let i = va.parseAxisParam(s, a.shape);
+                        const o = jd.getAxesPermutation(i, a.shape.length);
+                        let u = a;
+                        const l = [];
+                        null != o && (u = uC({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: o
+                            }
+                        }), l.push(u), i = jd.getInnerMostAxes(i.length, u.shape.length)), i = [i[0]], jd.assertAxesAreInnerMostDims("argMin", i, u.shape.length);
+                        const [c, h] = jd.computeOutAndReduceShapes(u.shape, i), p = va.sizeFromShape(c), d = va.makeZerosTypedArray(p, "int32"), f = va.sizeFromShape(h), m = n.data.get(u.dataId).values;
+                        for (let e = 0; e < d.length; ++e) {
+                            const t = e * f;
+                            let n = m[t],
+                                r = 0;
+                            for (let e = 0; e < f; ++e) {
+                                const a = m[t + e];
+                                a < n && (n = a, r = e)
+                            }
+                            d[e] = r
+                        }
+                        return l.forEach(e => n.disposeIntermediateTensorInfo(e)), n.makeTensorInfo(c, "int32", d)
+                    }
+                },
+                yT = lN(lt, e => Math.asin(e)),
+                bT = {
+                    kernelName: lt,
+                    backendName: "cpu",
+                    kernelFunc: yT
+                },
+                vT = lN(ct, e => Math.asinh(e)),
+                xT = {
+                    kernelName: ct,
+                    backendName: "cpu",
+                    kernelFunc: vT
+                },
+                wT = lN(ht, e => Math.atan(e)),
+                kT = {
+                    kernelName: ht,
+                    backendName: "cpu",
+                    kernelFunc: wT
+                },
+                ST = jE((e, t) => Math.atan2(e, t)),
+                IT = eN(dt, ST),
+                AT = {
+                    kernelName: dt,
+                    backendName: "cpu",
+                    kernelFunc: IT
+                },
+                ET = lN(pt, e => Math.atanh(e)),
+                NT = {
+                    kernelName: pt,
+                    backendName: "cpu",
+                    kernelFunc: ET
+                };
+
+            function CT(e, t, n, r, a, s) {
+                const i = a.strideHeight,
+                    o = a.strideWidth,
+                    u = a.dilationHeight,
+                    l = a.dilationWidth,
+                    c = a.effectiveFilterHeight,
+                    h = a.effectiveFilterWidth,
+                    p = a.padInfo.top,
+                    d = a.padInfo.left,
+                    f = "max" === s ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY,
+                    m = Ii(a.outShape, n),
+                    g = m.values,
+                    y = a.outShape[1] * a.outShape[2] * a.outShape[3],
+                    b = a.outShape[2] * a.outShape[3],
+                    v = a.outShape[3];
+                for (let t = 0; t < a.batchSize; ++t) {
+                    const n = t * y,
+                        m = t * r[0];
+                    for (let t = 0; t < a.inChannels; ++t)
+                        for (let y = 0; y < a.outHeight; ++y) {
+                            const x = y * i - p,
+                                w = Math.max(0, x),
+                                k = Math.min(a.inHeight, c + x),
+                                S = n + y * b;
+                            for (let n = 0; n < a.outWidth; ++n) {
+                                const i = n * o - d,
+                                    c = Math.max(0, i),
+                                    p = Math.min(a.inWidth, h + i);
+                                let y = f,
+                                    b = 0,
+                                    x = 0;
+                                for (let n = w; n < k; n += u) {
+                                    const a = m + n * r[1];
+                                    for (let n = c; n < p; n += l) {
+                                        const i = e[a + n * r[2] + t];
+                                        "max" === s && i > y ? y = i : "avg" === s && (b += i, x++)
+                                    }
+                                    if (isNaN(y)) break
+                                }
+                                g[S + n * v + t] = "avg" === s ? b / x : y
+                            }
+                        }
+                }
+                return m
+            }
+
+            function TT(e, t, n, r, a = !1, s = !1) {
+                const i = Ii(r.outShape, "int32"),
+                    o = r.strideHeight,
+                    u = r.strideWidth,
+                    l = r.dilationHeight,
+                    c = r.dilationWidth,
+                    h = r.effectiveFilterHeight,
+                    p = r.effectiveFilterWidth,
+                    d = r.padInfo.top,
+                    f = r.padInfo.left,
+                    m = Ii(t, n, e);
+                for (let e = 0; e < r.batchSize; ++e)
+                    for (let t = 0; t < r.inChannels; ++t)
+                        for (let n = 0; n < r.outHeight; ++n) {
+                            const g = n * o - d;
+                            let y = g;
+                            for (; y < 0;) y += l;
+                            const b = Math.min(r.inHeight, h + g);
+                            for (let o = 0; o < r.outWidth; ++o) {
+                                const h = o * u - f;
+                                let d = h;
+                                for (; d < 0;) d += c;
+                                const v = Math.min(r.inWidth, p + h);
+                                let x = Number.NEGATIVE_INFINITY,
+                                    w = -1;
+                                for (let n = y; n < b; n += l) {
+                                    const i = n - g;
+                                    for (let o = d; o < v; o += c) {
+                                        const u = o - h,
+                                            l = m.get(e, n, o, t);
+                                        l > x && (x = l, w = a ? s ? ((e * r.inHeight + n) * r.inWidth + o) * r.inChannels + t : (n * r.inWidth + o) * r.inChannels + t : i * p + u)
+                                    }
+                                }
+                                i.set(w, e, n, o, t)
+                            }
+                        }
+                return i
+            }
+
+            function RT(e, t, n, r, a, s) {
+                const i = a.strideDepth,
+                    o = a.strideHeight,
+                    u = a.strideWidth,
+                    l = a.dilationDepth,
+                    c = a.dilationHeight,
+                    h = a.dilationWidth,
+                    p = a.effectiveFilterDepth,
+                    d = a.effectiveFilterHeight,
+                    f = a.effectiveFilterWidth,
+                    m = a.padInfo.front,
+                    g = a.padInfo.top,
+                    y = a.padInfo.left,
+                    b = "max" === s ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY,
+                    v = Ii(a.outShape, n),
+                    x = v.values,
+                    w = a.outShape[1] * a.outShape[2] * a.outShape[3] * a.outShape[4],
+                    k = a.outShape[2] * a.outShape[3] * a.outShape[4],
+                    S = a.outShape[3] * a.outShape[4],
+                    I = a.outShape[4];
+                for (let t = 0; t < a.batchSize; ++t) {
+                    const n = t * w,
+                        v = t * r[0];
+                    for (let t = 0; t < a.inChannels; ++t)
+                        for (let w = 0; w < a.outDepth; ++w) {
+                            const A = w * i - m;
+                            let E = A;
+                            for (; E < 0;) E += l;
+                            const N = Math.min(a.inDepth, p + A),
+                                C = n + w * k;
+                            for (let n = 0; n < a.outHeight; ++n) {
+                                const i = n * o - g;
+                                let p = i;
+                                for (; p < 0;) p += c;
+                                const m = Math.min(a.inHeight, d + i),
+                                    w = C + n * S;
+                                for (let n = 0; n < a.outWidth; ++n) {
+                                    const i = n * u - y;
+                                    let o = i;
+                                    for (; o < 0;) o += h;
+                                    const d = Math.min(a.inWidth, f + i),
+                                        g = w + n * I;
+                                    let k = b,
+                                        S = 0,
+                                        A = 0;
+                                    for (let n = E; n < N; n += l) {
+                                        const a = v + n * r[1];
+                                        for (let n = p; n < m; n += c) {
+                                            const i = a + n * r[2];
+                                            for (let n = o; n < d; n += h) {
+                                                const a = e[i + n * r[3] + t];
+                                                if ("max" === s && a > k ? k = a : "avg" === s && (S += a, A++), isNaN(k)) break
+                                            }
+                                            if (isNaN(k)) break
+                                        }
+                                        if (isNaN(k)) break
+                                    }
+                                    x[g + t] = "avg" === s ? S / A : k
+                                }
+                            }
+                        }
+                }
+                return v
+            }
+            var _T = {
+                kernelName: ft,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t;
+                    $E(a, "avgPool");
+                    const {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u
+                    } = r;
+                    va.assert(jd.eitherStridesOrDilationsAreOne(i, 1), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);
+                    const l = jd.computePool2DInfo(a.shape, s, i, 1, o, u);
+                    let c;
+                    if (1 === l.filterWidth && 1 === l.filterHeight && va.arraysEqual(l.inShape, l.outShape)) c = KE({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    else {
+                        const e = n.data.get(a.dataId).values,
+                            t = va.computeStrides(a.shape),
+                            r = CT(e, a.shape, a.dtype, t, l, "avg");
+                        c = n.makeTensorInfo(l.outShape, a.dtype, r.values)
+                    }
+                    return c
+                }
+            };
+            var FT = {
+                kernelName: gt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u,
+                        dataFormat: l
+                    } = r;
+                    $E(a, "avgPool3d");
+                    const c = jd.computePool3DInfo(a.shape, s, i, 1, o, u, l),
+                        h = RT(n.data.get(a.dataId).values, a.shape, a.dtype, va.computeStrides(a.shape), c, "avg");
+                    return n.makeTensorInfo(h.shape, "float32", h.values)
+                }
+            };
+            var MT = {
+                kernelName: yt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s
+                    } = t, {
+                        filterSize: i,
+                        strides: o,
+                        pad: u,
+                        dimRoundingMode: l
+                    } = r;
+                    $E([a, s], "avgPool3DGrad");
+                    const c = jd.computePool3DInfo(s.shape, i, o, 1, u, l),
+                        h = c.strideDepth,
+                        p = c.strideHeight,
+                        d = c.strideWidth,
+                        f = c.filterDepth,
+                        m = c.filterHeight,
+                        g = c.filterWidth,
+                        y = c.dilationDepth,
+                        b = c.dilationHeight,
+                        v = c.dilationWidth,
+                        x = c.effectiveFilterDepth,
+                        w = c.effectiveFilterHeight,
+                        k = c.effectiveFilterWidth,
+                        S = x - 1 - c.padInfo.front,
+                        I = k - 1 - c.padInfo.left,
+                        A = w - 1 - c.padInfo.top,
+                        E = Ii(s.shape, "float32"),
+                        N = 1 / (f * m * g),
+                        C = n.bufferSync(a);
+                    for (let e = 0; e < c.batchSize; ++e)
+                        for (let t = 0; t < c.inChannels; ++t)
+                            for (let n = 0; n < c.inDepth; ++n)
+                                for (let r = 0; r < c.inHeight; ++r)
+                                    for (let a = 0; a < c.inWidth; ++a) {
+                                        const s = n - S,
+                                            i = r - A,
+                                            o = a - I;
+                                        let u = 0;
+                                        for (let n = 0; n < x; n += y) {
+                                            const r = (s + n) / h;
+                                            if (!(r < 0 || r >= c.outDepth || Math.floor(r) !== r))
+                                                for (let n = 0; n < w; n += b) {
+                                                    const a = (i + n) / p;
+                                                    if (!(a < 0 || a >= c.outHeight || Math.floor(a) !== a))
+                                                        for (let n = 0; n < k; n += v) {
+                                                            const s = (o + n) / d;
+                                                            if (s < 0 || s >= c.outWidth || Math.floor(s) !== s) continue;
+                                                            u += C.get(e, r, a, s, t)
+                                                        }
+                                                }
+                                        }
+                                        E.set(u * N, e, n, r, a, t)
+                                    }
+                    return n.makeTensorInfo(E.shape, E.dtype, E.values)
+                }
+            };
+            var DT = {
+                kernelName: mt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s
+                    } = t, i = s;
+                    $E([a, s], "avgPoolGrad");
+                    const {
+                        filterSize: o,
+                        strides: u,
+                        pad: l
+                    } = r, c = jd.computePool2DInfo(i.shape, o, u, 1, l), h = c.strideHeight, p = c.strideWidth, d = c.filterHeight, f = c.filterWidth, m = c.dilationHeight, g = c.dilationWidth, y = c.effectiveFilterHeight, b = c.effectiveFilterWidth, v = b - 1 - c.padInfo.left, x = y - 1 - c.padInfo.top, w = Ii(i.shape, "float32"), k = 1 / (d * f), S = n.data.get(a.dataId).values, I = Ii(a.shape, "float32", S);
+                    for (let e = 0; e < c.batchSize; ++e)
+                        for (let t = 0; t < c.inChannels; ++t)
+                            for (let n = 0; n < c.inHeight; ++n)
+                                for (let r = 0; r < c.inWidth; ++r) {
+                                    const a = n - x,
+                                        s = r - v;
+                                    let i = 0;
+                                    for (let n = 0; n < y; n += m) {
+                                        const r = (a + n) / h;
+                                        if (!(r < 0 || r >= c.outHeight || Math.floor(r) !== r))
+                                            for (let n = 0; n < b; n += g) {
+                                                const a = (s + n) / p;
+                                                if (a < 0 || a >= c.outWidth || Math.floor(a) !== a) continue;
+                                                i += I.get(e, r, a, t)
+                                            }
+                                    }
+                                    w.set(i * k, e, n, r, t)
+                                }
+                    return n.makeTensorInfo(w.shape, w.dtype, w.values)
+                }
+            };
+            var OT = {
+                kernelName: ln,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        scale: s,
+                        offset: i,
+                        mean: o,
+                        variance: u
+                    } = t;
+                    va.assert(o.shape.length === u.shape.length, () => "Batch normalization gradient requires mean and variance to have equal ranks."), va.assert(null == i || o.shape.length === i.shape.length, () => "Batch normalization gradient requires mean and offset to have equal ranks."), va.assert(null == s || o.shape.length === s.shape.length, () => "Batch normalization gradient requires mean and scale to have equal ranks."), $E([a, o, u, s, i], "batchNorm");
+                    let {
+                        varianceEpsilon: l
+                    } = r;
+                    null == l && (l = .001);
+                    const c = n.data.get(a.dataId).values,
+                        h = n.data.get(o.dataId).values,
+                        p = n.data.get(u.dataId).values,
+                        d = s ? n.data.get(s.dataId).values : new Float32Array([1]),
+                        f = i ? n.data.get(i.dataId).values : new Float32Array([0]),
+                        m = new Float32Array(c.length),
+                        g = f.length,
+                        y = d.length,
+                        b = p.length,
+                        v = h.length;
+                    let x = 0,
+                        w = 0,
+                        k = 0,
+                        S = 0;
+                    for (let e = 0; e < c.length; ++e) m[e] = f[x++] + (c[e] - h[w++]) * d[k++] / Math.sqrt(p[S++] + l), x >= g && (x = 0), w >= v && (w = 0), k >= y && (k = 0), S >= b && (S = 0);
+                    return n.makeTensorInfo(a.shape, a.dtype, m)
+                }
+            };
+            var LT = {
+                kernelName: vt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        blockShape: s,
+                        crops: i
+                    } = r;
+                    $E([a], "batchToSpaceND");
+                    const o = s.reduce((e, t) => e * t),
+                        u = jd.getReshaped(a.shape, s, o),
+                        l = jd.getPermuted(u.length, s.length),
+                        c = jd.getReshapedPermuted(a.shape, s, o),
+                        h = jd.getSliceBeginCoords(i, s.length),
+                        p = jd.getSliceSize(c, i, s.length),
+                        d = rT({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: u
+                            }
+                        }),
+                        f = uC({
+                            inputs: {
+                                x: d
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: l
+                            }
+                        }),
+                        m = rT({
+                            inputs: {
+                                x: f
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: c
+                            }
+                        }),
+                        g = xC({
+                            inputs: {
+                                x: m
+                            },
+                            backend: n,
+                            attrs: {
+                                begin: h,
+                                size: p
+                            }
+                        });
+                    return n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(m), g
+                }
+            };
+            var PT = {
+                kernelName: xt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        weights: s
+                    } = t, {
+                        size: i
+                    } = r, o = iN(n.data.get(a.dataId).values, n.data.get(s.dataId).values, s.dtype, s.shape, i);
+                    return n.makeTensorInfo([i], s.dtype, o)
+                }
+            };
+            var $T = {
+                    kernelName: kt,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            s0: r,
+                            s1: a
+                        } = t, s = n.data.get(r.dataId).values, i = n.data.get(a.dataId).values, o = jd.assertAndGetBroadcastShape(Array.from(s), Array.from(i));
+                        return n.makeTensorInfo([o.length], "int32", Int32Array.from(o))
+                    }
+                },
+                BT = lN(At, (e, t) => {
+                    const n = t;
+                    return e > n.clipValueMax ? n.clipValueMax : e < n.clipValueMin ? n.clipValueMin : e
+                }),
+                zT = {
+                    kernelName: At,
+                    backendName: "cpu",
+                    kernelFunc: BT
+                },
+                UT = {
+                    kernelName: Nt,
+                    backendName: "cpu",
+                    kernelFunc: e => {
+                        const {
+                            x: t
+                        } = e.inputs, n = e.backend, r = new Float32Array(va.sizeFromShape(t.shape)), a = n.data.get(t.dataId), s = a.complexTensorInfos.real, i = a.complexTensorInfos.imag, o = n.data.get(s.dataId).values, u = n.data.get(i.dataId).values;
+                        for (let e = 0; e < o.length; e++) {
+                            const t = o[e],
+                                n = u[e];
+                            r[e] = Math.hypot(t, n)
+                        }
+                        return n.makeOutput(r, t.shape, "float32")
+                    }
+                };
+
+            function WT(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    input: r
+                } = t, a = n.data.get(r.dataId).complexTensorInfos.imag, s = n.data.get(a.dataId).values;
+                return n.makeTensorInfo(a.shape, a.dtype, s)
+            }
+            var VT = {
+                kernelName: gn,
+                backendName: "cpu",
+                kernelFunc: WT
+            };
+
+            function jT(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    axis: a
+                } = r, s = va.parseAxisParam(a, t[0].shape)[0];
+                let i = jd.computeOutShape(t.map(e => e.shape), s);
+                if (0 === va.sizeFromShape(i)) return n.makeTensorInfo(i, t[0].dtype, []);
+                const o = t.filter(e => va.sizeFromShape(e.shape) > 0);
+                if (1 === o.length) return KE({
+                    inputs: {
+                        x: o[0]
+                    },
+                    backend: n
+                });
+                const u = o.map(e => e.shape);
+                if (jd.assertParamsConsistent(u, s), "complex64" === o[0].dtype) {
+                    const e = o.map(e => XE({
+                            inputs: {
+                                input: e
+                            },
+                            backend: n
+                        })),
+                        t = o.map(e => WT({
+                            inputs: {
+                                input: e
+                            },
+                            backend: n
+                        })),
+                        r = jT({
+                            inputs: e,
+                            backend: n,
+                            attrs: {
+                                axis: s
+                            }
+                        }),
+                        a = jT({
+                            inputs: t,
+                            backend: n,
+                            attrs: {
+                                axis: s
+                            }
+                        }),
+                        i = GE({
+                            inputs: {
+                                real: r,
+                                imag: a
+                            },
+                            backend: n
+                        });
+                    return e.forEach(e => n.disposeIntermediateTensorInfo(e)), t.forEach(e => n.disposeIntermediateTensorInfo(e)), n.disposeIntermediateTensorInfo(r), n.disposeIntermediateTensorInfo(a), i
+                }
+                const l = o.map(e => {
+                        const t = va.sizeFromShape(e.shape.slice(s));
+                        return rT({
+                            inputs: {
+                                x: e
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [-1, t]
+                            }
+                        })
+                    }),
+                    c = l.map(e => ({
+                        vals: n.data.get(e.dataId).values,
+                        shape: e.shape
+                    }));
+                i = jd.computeOutShape(l.map(e => e.shape), 1);
+                const h = 1 === l[0].shape[0],
+                    p = fN(c, i, t[0].dtype, h),
+                    d = jd.computeOutShape(o.map(e => e.shape), s),
+                    f = n.makeTensorInfo(d, t[0].dtype, p);
+                return l.forEach(e => n.disposeIntermediateTensorInfo(e)), f
+            }
+            var GT = {
+                kernelName: Ct,
+                backendName: "cpu",
+                kernelFunc: jT
+            };
+
+            function HT(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a,
+                    filter: s
+                } = t, {
+                    strides: i,
+                    pad: o,
+                    dataFormat: u,
+                    dilations: l,
+                    dimRoundingMode: c
+                } = r;
+                $E([a, s], "conv2d");
+                const h = jd.convertConv2DDataFormat(u),
+                    p = jd.computeConv2DInfo(a.shape, s.shape, i, l, o, c, !1, h),
+                    d = p.filterHeight,
+                    f = p.filterWidth,
+                    m = p.dilationHeight,
+                    g = p.dilationWidth,
+                    y = p.padInfo.left,
+                    b = p.padInfo.top,
+                    v = "channelsLast" === p.dataFormat,
+                    x = new Ha(p.outShape, a.dtype),
+                    w = va.computeStrides(a.shape),
+                    k = va.computeStrides(s.shape),
+                    S = w[0],
+                    I = v ? w[1] : w[2],
+                    A = v ? w[2] : 1,
+                    E = v ? 1 : w[1],
+                    N = x.strides[0],
+                    C = v ? x.strides[1] : x.strides[2],
+                    T = v ? x.strides[2] : 1,
+                    R = v ? 1 : x.strides[1],
+                    _ = n.data.get(a.dataId).values,
+                    F = n.data.get(s.dataId).values,
+                    M = x.values;
+                for (let e = 0; e < p.batchSize; ++e) {
+                    const t = e * S,
+                        n = e * N;
+                    for (let e = 0; e < p.outHeight; ++e) {
+                        const r = n + e * C,
+                            a = e * p.strideHeight - b;
+                        for (let e = 0; e < d; ++e) {
+                            const n = a + e * m;
+                            if (n < 0 || n >= p.inHeight) continue;
+                            const s = e * k[0],
+                                i = t + n * I;
+                            for (let e = 0; e < p.outWidth; ++e) {
+                                const t = r + e * T,
+                                    n = e * p.strideWidth - y;
+                                for (let e = 0; e < f; ++e) {
+                                    const r = n + e * g;
+                                    if (r < 0 || r >= p.inWidth) continue;
+                                    const a = i + r * A;
+                                    let o = s + e * k[1];
+                                    for (let e = 0; e < p.inChannels; ++e) {
+                                        const n = _[a + e * E];
+                                        for (let e = 0; e < p.outChannels; ++e) M[t + e * R] += n * F[o + e];
+                                        o += p.outChannels
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+                return n.makeTensorInfo(x.shape, x.dtype, M)
+            }
+            var qT = {
+                kernelName: Tt,
+                backendName: "cpu",
+                kernelFunc: HT
+            };
+            var KT = {
+                kernelName: Rt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        dy: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dataFormat: u,
+                        dimRoundingMode: l,
+                        filterShape: c
+                    } = r;
+                    $E([a, s], "conv2dBackpropFilter");
+                    const h = jd.convertConv2DDataFormat(u),
+                        p = jd.computeConv2DInfo(a.shape, c, i, 1, o, l, !1, h),
+                        {
+                            strideHeight: d,
+                            strideWidth: f,
+                            filterHeight: m,
+                            filterWidth: g
+                        } = p,
+                        y = "channelsLast" === p.dataFormat,
+                        b = new Ha(p.filterShape, "float32"),
+                        v = p.padInfo.left,
+                        x = p.padInfo.top,
+                        w = n.data.get(a.dataId).values,
+                        k = n.data.get(s.dataId).values,
+                        S = new Ha(a.shape, a.dtype, w),
+                        I = new Ha(s.shape, s.dtype, k);
+                    for (let e = 0; e < m; ++e) {
+                        const t = Math.max(0, Math.ceil((x - e) / d)),
+                            n = Math.min(p.outHeight, (p.inHeight + x - e) / d);
+                        for (let r = 0; r < g; ++r) {
+                            const a = Math.max(0, Math.ceil((v - r) / f)),
+                                s = Math.min(p.outWidth, (p.inWidth + v - r) / f);
+                            for (let i = 0; i < p.inChannels; ++i)
+                                for (let o = 0; o < p.outChannels; ++o) {
+                                    let u = 0;
+                                    for (let l = 0; l < p.batchSize; ++l)
+                                        for (let c = t; c < n; ++c) {
+                                            const t = e + c * d - x;
+                                            for (let e = a; e < s; ++e) {
+                                                const n = r + e * f - v;
+                                                u += y ? S.get(l, t, n, i) * I.get(l, c, e, o) : S.get(l, i, t, n) * I.get(l, o, c, e)
+                                            }
+                                        }
+                                    b.set(u, e, r, i, o)
+                                }
+                        }
+                    }
+                    return n.makeTensorInfo(b.shape, b.dtype, b.values)
+                }
+            };
+            var YT = {
+                kernelName: _t,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        filter: s
+                    } = t, {
+                        inputShape: i,
+                        strides: o,
+                        pad: u,
+                        dataFormat: l,
+                        dimRoundingMode: c
+                    } = r;
+                    $E([a, s], "conv2dBackpropInput");
+                    const h = va.computeStrides(s.shape),
+                        p = va.computeStrides(a.shape);
+                    let d = jd.convertConv2DDataFormat(l);
+                    const f = jd.computeConv2DInfo(i, s.shape, o, 1, u, c, !1, d),
+                        m = new Ha(f.inShape, "float32"),
+                        g = m.values,
+                        y = n.data.get(a.dataId).values,
+                        b = n.data.get(s.dataId).values,
+                        [v, x, w] = h,
+                        {
+                            batchSize: k,
+                            filterHeight: S,
+                            filterWidth: I,
+                            inChannels: A,
+                            inHeight: E,
+                            inWidth: N,
+                            outChannels: C,
+                            outHeight: T,
+                            outWidth: R,
+                            strideHeight: _,
+                            strideWidth: F
+                        } = f;
+                    d = f.dataFormat;
+                    const M = S - 1 - f.padInfo.top,
+                        D = I - 1 - f.padInfo.left,
+                        O = "channelsLast" === d,
+                        L = m.strides[0],
+                        P = O ? m.strides[1] : m.strides[2],
+                        $ = O ? m.strides[2] : 1,
+                        B = O ? 1 : m.strides[1],
+                        z = p[0],
+                        U = O ? p[1] : p[2],
+                        W = O ? p[2] : 1,
+                        V = O ? 1 : p[1];
+                    for (let e = 0; e < k; ++e)
+                        for (let t = 0; t < A; ++t)
+                            for (let n = 0; n < E; ++n) {
+                                const r = n - M,
+                                    a = Math.max(0, Math.ceil(r / _)),
+                                    s = Math.min(T, (S + r) / _);
+                                for (let i = 0; i < N; ++i) {
+                                    const o = i - D,
+                                        u = Math.max(0, Math.ceil(o / F)),
+                                        l = Math.min(R, (I + o) / F);
+                                    let c = 0;
+                                    for (let n = a; n < s; ++n) {
+                                        const a = n * _ - r;
+                                        for (let r = u; r < l; ++r) {
+                                            const s = z * e + U * n + W * r,
+                                                i = v * (S - 1 - a) + x * (I - 1 - (r * F - o)) + w * t;
+                                            for (let e = 0; e < C; ++e) {
+                                                c += y[s + V * e] * b[i + e]
+                                            }
+                                        }
+                                    }
+                                    g[L * e + P * n + $ * i + B * t] = c
+                                }
+                            }
+                    return n.makeTensorInfo(m.shape, m.dtype, m.values)
+                }
+            };
+            var XT = {
+                kernelName: Ft,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dilations: u
+                    } = r;
+                    $E([a, s], "conv3d");
+                    const l = jd.computeConv3DInfo(a.shape, s.shape, i, u, o),
+                        {
+                            filterDepth: c,
+                            filterHeight: h,
+                            filterWidth: p,
+                            dilationDepth: d,
+                            dilationHeight: f,
+                            dilationWidth: m,
+                            padInfo: g
+                        } = l,
+                        y = g.front,
+                        b = g.left,
+                        v = g.top,
+                        x = new Ha(l.outShape, a.dtype),
+                        w = n.data.get(a.dataId).values,
+                        k = n.data.get(s.dataId).values,
+                        S = x.values,
+                        I = va.computeStrides(a.shape),
+                        A = va.computeStrides(s.shape);
+                    for (let e = 0; e < l.batchSize; ++e) {
+                        const t = e * I[0],
+                            n = e * x.strides[0];
+                        for (let e = 0; e < l.outDepth; ++e) {
+                            const r = n + e * x.strides[1],
+                                a = e * l.strideDepth - y;
+                            for (let e = 0; e < c; ++e) {
+                                const n = a + e * d;
+                                if (n < 0 || n >= l.inDepth) continue;
+                                const s = e * A[0],
+                                    i = t + n * I[1];
+                                for (let e = 0; e < l.outHeight; ++e) {
+                                    const t = r + e * x.strides[2],
+                                        n = e * l.strideHeight - v;
+                                    for (let e = 0; e < h; ++e) {
+                                        const r = n + e * f;
+                                        if (r < 0 || r >= l.inHeight) continue;
+                                        const a = s + e * A[1],
+                                            o = i + r * I[2];
+                                        for (let e = 0; e < l.outWidth; ++e) {
+                                            const n = t + e * l.outChannels,
+                                                r = e * l.strideWidth - b;
+                                            for (let e = 0; e < p; ++e) {
+                                                const t = r + e * m;
+                                                if (t < 0 || t >= l.inWidth) continue;
+                                                const s = a + e * A[2],
+                                                    i = o + t * l.inChannels;
+                                                let u = s;
+                                                for (let e = 0; e < l.inChannels; ++e) {
+                                                    const t = w[i + e];
+                                                    for (let e = 0; e < l.outChannels; ++e) S[n + e] += t * k[u + e];
+                                                    u += l.outChannels
+                                                }
+                                            }
+                                        }
+                                    }
+                                }
+                            }
+                        }
+                    }
+                    return n.makeTensorInfo(x.shape, x.dtype, x.values)
+                }
+            };
+            var QT = {
+                kernelName: Mt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        dy: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        filterShape: u
+                    } = r;
+                    $E([a, s], "conv3dBackpropFilterV2");
+                    const l = va.computeStrides(a.shape),
+                        c = va.computeStrides(s.shape),
+                        h = jd.computeConv3DInfo(a.shape, u, i, 1, o),
+                        p = h.strideDepth,
+                        d = h.strideHeight,
+                        f = h.strideWidth,
+                        m = h.filterDepth,
+                        g = h.filterHeight,
+                        y = h.filterWidth,
+                        b = new Ha(h.filterShape, "float32"),
+                        v = b.values,
+                        [x, w, k, S] = b.strides,
+                        I = n.data.get(s.dataId).values,
+                        [A, E, N, C] = c,
+                        T = n.data.get(a.dataId).values,
+                        [R, _, F, M] = l,
+                        D = h.padInfo.front,
+                        O = h.padInfo.left,
+                        L = h.padInfo.top;
+                    for (let e = 0; e < m; ++e) {
+                        const t = Math.max(0, Math.ceil((D - e) / p)),
+                            n = Math.min(h.outDepth, (h.inDepth + D - e) / p),
+                            r = e * x;
+                        for (let a = 0; a < g; ++a) {
+                            const s = Math.max(0, Math.ceil((L - a) / d)),
+                                i = Math.min(h.outHeight, (h.inHeight + L - a) / d),
+                                o = a * w + r;
+                            for (let r = 0; r < y; ++r) {
+                                const u = Math.max(0, Math.ceil((O - r) / f)),
+                                    l = Math.min(h.outWidth, (h.inWidth + O - r) / f),
+                                    c = r * k + o;
+                                for (let o = 0; o < h.inChannels; ++o) {
+                                    const m = o * S + c;
+                                    for (let c = 0; c < h.outChannels; ++c) {
+                                        let g = 0;
+                                        for (let m = 0; m < h.batchSize; ++m) {
+                                            const h = m * R,
+                                                y = m * A;
+                                            for (let m = t; m < n; ++m) {
+                                                const t = (e + m * p - D) * _ + h,
+                                                    n = m * E + y;
+                                                for (let e = s; e < i; ++e) {
+                                                    const s = (a + e * d - L) * F + t,
+                                                        i = e * N + n;
+                                                    for (let e = u; e < l; ++e) {
+                                                        const t = e * C + i;
+                                                        g += T[(r + e * f - O) * M + s + o] * I[t + c]
+                                                    }
+                                                }
+                                            }
+                                        }
+                                        v[m + c] = g
+                                    }
+                                }
+                            }
+                        }
+                    }
+                    return n.makeTensorInfo(b.shape, b.dtype, b.values)
+                }
+            };
+            var ZT = {
+                    kernelName: Dt,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            dy: a,
+                            filter: s
+                        } = t, {
+                            pad: i,
+                            strides: o,
+                            inputShape: u
+                        } = r;
+                        $E([a], "conv3dBackpropInputV2");
+                        const l = va.computeStrides(a.shape),
+                            c = va.computeStrides(s.shape),
+                            h = jd.computeConv3DInfo(u, s.shape, o, 1, i),
+                            p = new Ha(h.inShape, "float32"),
+                            d = p.values,
+                            [f, m, g, y] = p.strides,
+                            b = n.data.get(a.dataId).values,
+                            [v, x, w, k] = l,
+                            S = n.data.get(s.dataId).values,
+                            [I, A, E, N] = c,
+                            {
+                                batchSize: C,
+                                filterDepth: T,
+                                filterHeight: R,
+                                filterWidth: _,
+                                inChannels: F,
+                                inDepth: M,
+                                inHeight: D,
+                                inWidth: O,
+                                outChannels: L,
+                                outDepth: P,
+                                outHeight: $,
+                                outWidth: B,
+                                strideDepth: z,
+                                strideHeight: U,
+                                strideWidth: W
+                            } = h,
+                            V = T - 1 - h.padInfo.front,
+                            j = R - 1 - h.padInfo.top,
+                            G = _ - 1 - h.padInfo.left;
+                        for (let e = 0; e < C; ++e)
+                            for (let t = 0; t < F; ++t)
+                                for (let n = 0; n < M; ++n) {
+                                    const r = n - V,
+                                        a = Math.max(0, Math.ceil(r / z)),
+                                        s = Math.min(P, (T + r) / z);
+                                    for (let i = 0; i < D; ++i) {
+                                        const o = i - j,
+                                            u = Math.max(0, Math.ceil(o / U)),
+                                            l = Math.min($, (R + o) / U);
+                                        for (let c = 0; c < O; ++c) {
+                                            const h = c - G,
+                                                p = Math.max(0, Math.ceil(h / W)),
+                                                C = Math.min(B, (_ + h) / W);
+                                            let F = 0;
+                                            for (let n = a; n < s; ++n) {
+                                                const a = n * z - r;
+                                                for (let r = u; r < l; ++r) {
+                                                    const s = r * U - o;
+                                                    for (let i = p; i < C; ++i) {
+                                                        const o = v * e + x * n + w * r + k * i,
+                                                            u = I * (T - 1 - a) + A * (R - 1 - s) + E * (_ - 1 - (i * W - h)) + N * t;
+                                                        for (let e = 0; e < L; ++e) {
+                                                            F += b[o + e] * S[u + e]
+                                                        }
+                                                    }
+                                                }
+                                            }
+                                            d[f * e + m * n + g * i + y * c + t] = F
+                                        }
+                                    }
+                                }
+                        return n.makeTensorInfo(p.shape, p.dtype, p.values)
+                    }
+                },
+                JT = lN(Ot, e => Math.cos(e)),
+                eR = {
+                    kernelName: Ot,
+                    backendName: "cpu",
+                    kernelFunc: JT
+                },
+                tR = lN(Lt, e => Math.cosh(e)),
+                nR = {
+                    kernelName: Lt,
+                    backendName: "cpu",
+                    kernelFunc: tR
+                };
+            var rR = {
+                kernelName: $t,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        image: a,
+                        boxes: s,
+                        boxInd: i
+                    } = t, {
+                        cropSize: o,
+                        method: u,
+                        extrapolationValue: l
+                    } = r, [c, h, p, d] = a.shape, f = s.shape[0], [m, g] = o, y = Ii([f, m, g, d], "float32"), b = n.data.get(s.dataId).values, v = n.data.get(i.dataId).values, x = n.data.get(a.dataId).values, w = va.computeStrides(a.shape), k = va.computeStrides(y.shape);
+                    for (let e = 0; e < f; e++) {
+                        const t = 4 * e,
+                            n = b[t],
+                            r = b[t + 1],
+                            a = b[t + 2],
+                            s = b[t + 3],
+                            i = v[e];
+                        if (i >= c) continue;
+                        const o = m > 1 ? (a - n) * (h - 1) / (m - 1) : 0,
+                            f = g > 1 ? (s - r) * (p - 1) / (g - 1) : 0;
+                        for (let t = 0; t < m; t++) {
+                            const c = m > 1 ? n * (h - 1) + t * o : .5 * (n + a) * (h - 1);
+                            if (c < 0 || c > h - 1)
+                                for (let n = 0; n < g; n++)
+                                    for (let r = 0; r < d; r++) {
+                                        const a = r + n * k[2] + t * k[1] + e * k[0];
+                                        y.values[a] = l
+                                    } else if ("bilinear" === u) {
+                                        const n = Math.floor(c),
+                                            a = Math.ceil(c),
+                                            o = c - n;
+                                        for (let u = 0; u < g; u++) {
+                                            const c = g > 1 ? r * (p - 1) + u * f : .5 * (r + s) * (p - 1);
+                                            if (c < 0 || c > p - 1) {
+                                                for (let n = 0; n < d; n++) {
+                                                    const r = n + u * k[2] + t * k[1] + e * k[0];
+                                                    y.values[r] = l
+                                                }
+                                                continue
+                                            }
+                                            const h = Math.floor(c),
+                                                m = Math.ceil(c),
+                                                b = c - h;
+                                            for (let r = 0; r < d; r++) {
+                                                let s = r + h * w[2] + n * w[1] + i * w[0];
+                                                const l = x[s];
+                                                s = r + m * w[2] + n * w[1] + i * w[0];
+                                                const c = x[s];
+                                                s = r + h * w[2] + a * w[1] + i * w[0];
+                                                const p = x[s];
+                                                s = r + m * w[2] + a * w[1] + i * w[0];
+                                                const d = l + (c - l) * b,
+                                                    f = p + (x[s] - p) * b;
+                                                s = r + u * k[2] + t * k[1] + e * k[0], y.values[s] = d + (f - d) * o
+                                            }
+                                        }
+                                    } else
+                                        for (let n = 0; n < g; ++n) {
+                                            const a = g > 1 ? r * (p - 1) + n * f : .5 * (r + s) * (p - 1);
+                                            if (a < 0 || a > p - 1) {
+                                                for (let r = 0; r < d; r++) {
+                                                    const a = r + n * k[2] + t * k[1] + e * k[0];
+                                                    y.values[a] = l
+                                                }
+                                                continue
+                                            }
+                                            const o = Math.round(a),
+                                                u = Math.round(c);
+                                            for (let r = 0; r < d; r++) {
+                                                const a = r + o * w[2] + u * w[1] + i * w[0],
+                                                    s = r + n * k[2] + t * k[1] + e * k[0];
+                                                y.values[s] = x[a]
+                                            }
+                                        }
+                        }
+                    }
+                    return n.makeTensorInfo(y.shape, y.dtype, y.values)
+                }
+            };
+            var aR = {
+                kernelName: Pt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        exclusive: i,
+                        reverse: o
+                    } = r;
+                    $E(a, "cumsum");
+                    const u = jd.getAxesPermutation([s], a.shape.length);
+                    let l = a;
+                    null != u && (l = uC({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: u
+                        }
+                    }));
+                    const c = jd.getInnerMostAxes(1, a.shape.length)[0];
+                    if (c !== l.shape.length - 1) throw new Error(`backend.cumsum in CPU expects an inner-most axis=${l.shape.length-1} but got axis=${c}`);
+                    const h = cs(l.dtype, "int32"),
+                        p = va.makeZerosTypedArray(va.sizeFromShape(l.shape), h),
+                        d = n.data.get(l.dataId).values,
+                        f = l.shape[l.shape.length - 1],
+                        m = o ? (e, t) => e + f - t - 1 : (e, t) => e + t;
+                    for (let e = 0; e < d.length; e += f)
+                        for (let t = 0; t < f; t++) {
+                            const n = m(e, t);
+                            if (0 === t) p[n] = i ? 0 : d[n];
+                            else {
+                                const r = m(e, t - 1);
+                                p[n] = i ? d[r] + p[r] : d[n] + p[r]
+                            }
+                        }
+                    const g = n.makeTensorInfo(l.shape, h, p);
+                    if (null != u) {
+                        const e = uC({
+                            inputs: {
+                                x: g
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: jd.getUndoAxesPermutation(u)
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(g), n.disposeIntermediateTensorInfo(l), e
+                    }
+                    return g
+                }
+            };
+            var sR = {
+                kernelName: Bt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        weights: s
+                    } = t, {
+                        size: i,
+                        binaryOutput: o
+                    } = r;
+                    if (1 === a.shape.length) {
+                        const e = iN(n.data.get(a.dataId).values, n.data.get(s.dataId).values, s.dtype, s.shape, i);
+                        return n.makeTensorInfo([i], s.dtype, e)
+                    }
+                    if (2 === a.shape.length) {
+                        const e = oN(n.bufferSync(a), n.bufferSync(s), i, o);
+                        return n.makeTensorInfo(e.shape, s.dtype, e.values)
+                    }
+                    throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)
+                }
+            };
+            var iR = {
+                kernelName: zt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        blockSize: s,
+                        dataFormat: i
+                    } = r;
+                    va.assert("NHWC" === i, () => "Only NHWC dataFormat supported on CPU for depthToSpace. Got " + i);
+                    const o = a.shape[0],
+                        u = a.shape[1],
+                        l = a.shape[2],
+                        c = a.shape[3],
+                        h = u * s,
+                        p = l * s,
+                        d = c / (s * s),
+                        f = n.data.get(a.dataId).values,
+                        m = new Float32Array(o * h * p * d);
+                    let g = 0;
+                    for (let e = 0; e < o; ++e)
+                        for (let t = 0; t < h; ++t) {
+                            const n = Math.floor(t / s),
+                                r = t % s;
+                            for (let t = 0; t < p; ++t) {
+                                const a = Math.floor(t / s),
+                                    i = (r * s + t % s) * d;
+                                for (let t = 0; t < d; ++t) {
+                                    const r = t + i + c * (a + l * (n + u * e));
+                                    m[g++] = f[r]
+                                }
+                            }
+                        }
+                    return n.makeTensorInfo([o, h, p, d], a.dtype, m)
+                }
+            };
+
+            function oR(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a,
+                    filter: s
+                } = t, {
+                    strides: i,
+                    pad: o,
+                    dilations: u,
+                    dimRoundingMode: l
+                } = r;
+                $E([a, s], "depthwiseConv2DNative");
+                const c = va.computeStrides(a.shape),
+                    h = va.computeStrides(s.shape);
+                let p = u;
+                null == p && (p = [1, 1]), va.assert(jd.eitherStridesOrDilationsAreOne(i, p), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${p}'`);
+                const d = jd.computeConv2DInfo(a.shape, s.shape, i, p, o, l, !0),
+                    {
+                        filterHeight: f,
+                        filterWidth: m,
+                        dilationHeight: g,
+                        dilationWidth: y,
+                        padInfo: b
+                    } = d,
+                    v = b.left,
+                    x = b.top,
+                    w = d.outChannels / d.inChannels,
+                    k = new Ha(d.outShape, a.dtype),
+                    S = n.data.get(a.dataId).values,
+                    I = n.data.get(s.dataId).values,
+                    A = k.values;
+                for (let e = 0; e < d.batchSize; ++e) {
+                    const t = e * c[0],
+                        n = e * k.strides[0];
+                    for (let e = 0; e < d.outHeight; ++e) {
+                        const r = n + e * k.strides[1],
+                            a = e * d.strideHeight - x;
+                        for (let e = 0; e < f; ++e) {
+                            const n = a + e * g;
+                            if (n < 0 || n >= d.inHeight) continue;
+                            const s = e * h[0],
+                                i = t + n * c[1];
+                            for (let e = 0; e < d.outWidth; ++e) {
+                                const t = r + e * k.strides[2],
+                                    n = e * d.strideWidth - v;
+                                for (let e = 0; e < m; ++e) {
+                                    const r = n + e * y;
+                                    if (r < 0 || r >= d.inWidth) continue;
+                                    const a = s + e * h[1],
+                                        o = i + r * d.inChannels;
+                                    let u = t,
+                                        l = a;
+                                    for (let e = 0; e < d.inChannels; ++e) {
+                                        const t = S[o + e];
+                                        for (let e = 0; e < w; ++e) A[u + e] += t * I[l + e];
+                                        u += w, l += w
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+                return n.makeTensorInfo(k.shape, k.dtype, k.values)
+            }
+            var uR = {
+                kernelName: Ut,
+                backendName: "cpu",
+                kernelFunc: oR
+            };
+            var lR = {
+                kernelName: Wt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        dy: s
+                    } = t, {
+                        strides: i,
+                        dilations: o,
+                        pad: u,
+                        dimRoundingMode: l,
+                        filterShape: c
+                    } = r;
+                    $E([a, s], "depthwiseConv2dNativeBackpropFilter");
+                    const h = jd.computeConv2DInfo(a.shape, c, i, o, u, l, !0),
+                        {
+                            strideHeight: p,
+                            strideWidth: d,
+                            filterHeight: f,
+                            filterWidth: m
+                        } = h,
+                        g = new Ha(h.filterShape, "float32"),
+                        y = h.padInfo.left,
+                        b = h.padInfo.top,
+                        v = h.outChannels / h.inChannels,
+                        x = n.data.get(a.dataId).values,
+                        w = new Ha(a.shape, a.dtype, x),
+                        k = n.data.get(s.dataId).values,
+                        S = new Ha(s.shape, s.dtype, k);
+                    for (let e = 0; e < f; ++e) {
+                        const t = Math.max(0, Math.ceil((b - e) / p)),
+                            n = Math.min(h.outHeight, (h.inHeight + b - e) / p);
+                        for (let r = 0; r < m; ++r) {
+                            const a = Math.max(0, Math.ceil((y - r) / d)),
+                                s = Math.min(h.outWidth, (h.inWidth + y - r) / d);
+                            for (let i = 0; i < h.outChannels; ++i) {
+                                const o = Math.trunc(i / v),
+                                    u = i % v;
+                                let l = 0;
+                                for (let u = 0; u < h.batchSize; ++u)
+                                    for (let c = t; c < n; ++c) {
+                                        const t = e + c * p - b;
+                                        for (let e = a; e < s; ++e) {
+                                            const n = r + e * d - y;
+                                            l += w.get(u, t, n, o) * S.get(u, c, e, i)
+                                        }
+                                    }
+                                g.set(l, e, r, o, u)
+                            }
+                        }
+                    }
+                    return n.makeTensorInfo(g.shape, g.dtype, g.values)
+                }
+            };
+            var cR = {
+                kernelName: Vt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        dilations: o,
+                        pad: u,
+                        dimRoundingMode: l,
+                        inputShape: c
+                    } = r;
+                    $E([a, s], "depthwiseConv2DNativeBackpropInput");
+                    const h = va.computeStrides(a.shape),
+                        p = va.computeStrides(s.shape),
+                        d = jd.computeConv2DInfo(c, s.shape, i, o, u, l, !0),
+                        f = new Ha(d.inShape, "float32"),
+                        m = f.values,
+                        [g, y, b] = f.strides,
+                        v = n.data.get(a.dataId).values,
+                        [x, w, k] = h,
+                        S = n.data.get(s.dataId).values,
+                        [I, A, E] = p,
+                        {
+                            batchSize: N,
+                            filterHeight: C,
+                            filterWidth: T,
+                            inChannels: R,
+                            inHeight: _,
+                            inWidth: F,
+                            outChannels: M,
+                            outHeight: D,
+                            outWidth: O,
+                            strideHeight: L,
+                            strideWidth: P
+                        } = d,
+                        $ = C - 1 - d.padInfo.top,
+                        B = T - 1 - d.padInfo.left,
+                        z = M / R;
+                    for (let e = 0; e < N; ++e)
+                        for (let t = 0; t < R; ++t)
+                            for (let n = 0; n < _; ++n) {
+                                const r = n - $,
+                                    a = Math.max(0, Math.ceil(r / L)),
+                                    s = Math.min(D, (C + r) / L);
+                                for (let i = 0; i < F; ++i) {
+                                    const o = i - B,
+                                        u = Math.max(0, Math.ceil(o / P)),
+                                        l = Math.min(O, (T + o) / P);
+                                    let c = 0;
+                                    for (let n = a; n < s; ++n) {
+                                        const a = n * L - r;
+                                        for (let r = u; r < l; ++r) {
+                                            const s = x * e + w * n + k * r,
+                                                i = I * (C - 1 - a) + A * (T - 1 - (r * P - o)) + E * t;
+                                            for (let e = 0; e < z; ++e) {
+                                                c += v[s + (t * z + e)] * S[i + e]
+                                            }
+                                        }
+                                    }
+                                    m[g * e + y * n + b * i + t] = c
+                                }
+                            }
+                    return n.makeTensorInfo(f.shape, f.dtype, f.values)
+                }
+            };
+            var hR = {
+                    kernelName: jt,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            x: r
+                        } = t, a = va.sizeFromShape(r.shape), s = n.data.get(r.dataId).values, i = Ii([a, a], r.dtype), o = i.values;
+                        for (let e = 0; e < s.length; e++) o[e * a + e] = s[e];
+                        const u = [...r.shape, ...r.shape];
+                        return n.makeTensorInfo(u, i.dtype, i.values)
+                    }
+                },
+                pR = {
+                    kernelName: Gt,
+                    backendName: "cpu",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t,
+                        attrs: n
+                    }) => {
+                        const {
+                            x: r,
+                            filter: a
+                        } = e, {
+                            strides: s,
+                            pad: i,
+                            dilations: o
+                        } = n, u = t, l = u.data.get(r.dataId).values, c = r.shape.length, h = u.data.get(a.dataId).values, p = a.shape.length, {
+                            batchSize: d,
+                            inHeight: f,
+                            inWidth: m,
+                            inChannels: g,
+                            outHeight: y,
+                            outWidth: b,
+                            padInfo: v,
+                            strideHeight: x,
+                            strideWidth: w,
+                            filterHeight: k,
+                            filterWidth: S,
+                            dilationHeight: I,
+                            dilationWidth: A,
+                            outShape: E
+                        } = jd.computeDilation2DInfo(r.shape, a.shape, s, i, "NHWC", o), N = va.sizeFromShape(E), C = E.length, T = va.getArrayFromDType(r.dtype, N);
+                        for (let e = 0; e < d; ++e)
+                            for (let t = 0; t < y; ++t) {
+                                const n = t * x - v.top;
+                                for (let s = 0; s < b; ++s) {
+                                    const i = s * w - v.left;
+                                    for (let o = 0; o < g; ++o) {
+                                        let u = Number.MIN_SAFE_INTEGER;
+                                        for (let t = 0; t < k; ++t) {
+                                            const s = n + t * I;
+                                            if (s >= 0 && s < f)
+                                                for (let n = 0; n < S; ++n) {
+                                                    const d = i + n * A;
+                                                    if (d >= 0 && d < m) {
+                                                        const i = va.locToIndex([e, s, d, o], c, va.computeStrides(r.shape)),
+                                                            f = va.locToIndex([t, n, o], p, va.computeStrides(a.shape)),
+                                                            m = l[i] + h[f];
+                                                        m > u && (u = m)
+                                                    }
+                                                }
+                                        }
+                                        T[va.locToIndex([e, t, s, o], C, va.computeStrides(E))] = u
+                                    }
+                                }
+                            }
+                        return {
+                            dataId: u.write(va.toTypedArray(T, r.dtype), E, r.dtype),
+                            shape: E,
+                            dtype: r.dtype
+                        }
+                    }
+                },
+                dR = {
+                    kernelName: qt,
+                    backendName: "cpu",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t,
+                        attrs: n
+                    }) => {
+                        const {
+                            x: r,
+                            filter: a,
+                            dy: s
+                        } = e, {
+                            strides: i,
+                            pad: o,
+                            dilations: u
+                        } = n, l = t, c = va.toNestedArray(r.shape, l.data.get(r.dataId).values), h = va.toNestedArray(a.shape, l.data.get(a.dataId).values), {
+                            batchSize: p,
+                            inHeight: d,
+                            inWidth: f,
+                            inChannels: m,
+                            outHeight: g,
+                            outWidth: y,
+                            padInfo: b,
+                            strideHeight: v,
+                            strideWidth: x,
+                            filterHeight: w,
+                            filterWidth: k,
+                            dilationHeight: S,
+                            dilationWidth: I,
+                            outShape: A
+                        } = jd.computeDilation2DInfo(r.shape, a.shape, i, o, "NHWC", u);
+                        va.assert(s.rank === A.length, () => `Error in ${qt}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);
+                        const E = va.toNestedArray(A, l.data.get(s.dataId).values),
+                            N = va.makeZerosNestedTypedArray(a.shape, a.dtype);
+                        for (let e = 0; e < p; ++e)
+                            for (let t = 0; t < g; ++t) {
+                                const n = t * v - b.top;
+                                for (let r = 0; r < y; ++r) {
+                                    const a = r * x - b.left;
+                                    for (let s = 0; s < m; ++s) {
+                                        let i = Number.MIN_SAFE_INTEGER,
+                                            o = 0,
+                                            u = 0;
+                                        for (let t = 0; t < w; ++t) {
+                                            const r = n + t * S;
+                                            if (r >= 0 && r < d)
+                                                for (let n = 0; n < k; ++n) {
+                                                    const l = a + n * I;
+                                                    if (l >= 0 && l < f) {
+                                                        const a = c[e][r][l][s] + h[t][n][s];
+                                                        a > i && (i = a, o = t, u = n)
+                                                    }
+                                                }
+                                        }
+                                        N[o][u][s] += E[e][t][r][s]
+                                    }
+                                }
+                            }
+                        return {
+                            dataId: l.write(va.toTypedArray(N, r.dtype), a.shape, a.dtype),
+                            shape: a.shape,
+                            dtype: a.dtype
+                        }
+                    }
+                },
+                fR = {
+                    kernelName: Ht,
+                    backendName: "cpu",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t,
+                        attrs: n
+                    }) => {
+                        const {
+                            x: r,
+                            filter: a,
+                            dy: s
+                        } = e, {
+                            strides: i,
+                            pad: o,
+                            dilations: u
+                        } = n, l = t, c = va.toNestedArray(r.shape, l.data.get(r.dataId).values), h = va.toNestedArray(a.shape, l.data.get(a.dataId).values), {
+                            batchSize: p,
+                            inHeight: d,
+                            inWidth: f,
+                            inChannels: m,
+                            outHeight: g,
+                            outWidth: y,
+                            padInfo: b,
+                            strideHeight: v,
+                            strideWidth: x,
+                            filterHeight: w,
+                            filterWidth: k,
+                            dilationHeight: S,
+                            dilationWidth: I,
+                            outShape: A
+                        } = jd.computeDilation2DInfo(r.shape, a.shape, i, o, "NHWC", u);
+                        va.assert(s.rank === A.length, () => `Error in ${Ht}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);
+                        const E = va.toNestedArray(A, l.data.get(s.dataId).values),
+                            N = va.makeZerosNestedTypedArray(r.shape, r.dtype);
+                        for (let e = 0; e < p; ++e)
+                            for (let t = 0; t < g; ++t) {
+                                const n = t * v - b.top;
+                                for (let r = 0; r < y; ++r) {
+                                    const a = r * x - b.left;
+                                    for (let s = 0; s < m; ++s) {
+                                        let i = Number.MIN_SAFE_INTEGER,
+                                            o = n < 0 ? 0 : n,
+                                            u = a < 0 ? 0 : a;
+                                        for (let t = 0; t < w; ++t) {
+                                            const r = n + t * S;
+                                            if (r >= 0 && r < d)
+                                                for (let n = 0; n < k; ++n) {
+                                                    const l = a + n * I;
+                                                    if (l >= 0 && l < f) {
+                                                        const a = c[e][r][l][s] + h[t][n][s];
+                                                        a > i && (i = a, o = r, u = l)
+                                                    }
+                                                }
+                                        }
+                                        N[e][o][u][s] += E[e][t][r][s]
+                                    }
+                                }
+                            }
+                        return {
+                            dataId: l.write(va.toTypedArray(N, r.dtype), r.shape, r.dtype),
+                            shape: r.shape,
+                            dtype: r.dtype
+                        }
+                    }
+                };
+
+            function mR(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    axis: s,
+                    keepDims: i
+                } = r;
+                let o;
+                $E(a, "sum"), o = "bool" === a.dtype ? ZE({
+                    inputs: {
+                        x: a
+                    },
+                    backend: n,
+                    attrs: {
+                        dtype: "int32"
+                    }
+                }) : KE({
+                    inputs: {
+                        x: a
+                    },
+                    backend: n
+                });
+                const u = o.shape.length,
+                    l = va.parseAxisParam(s, o.shape),
+                    c = jd.getAxesPermutation(l, u);
+                let h = l,
+                    p = o;
+                null != c && (p = uC({
+                    inputs: {
+                        x: o
+                    },
+                    backend: n,
+                    attrs: {
+                        perm: c
+                    }
+                }), h = jd.getInnerMostAxes(h.length, u)), jd.assertAxesAreInnerMostDims("sum", h, p.shape.length);
+                const [d, f] = jd.computeOutAndReduceShapes(p.shape, h);
+                let m = qE(n, d, jd.upcastType(p.dtype, "int32"));
+                const g = va.sizeFromShape(f),
+                    y = n.data.get(m.dataId).values,
+                    b = n.data.get(p.dataId).values;
+                for (let e = 0; e < y.length; ++e) {
+                    const t = e * g;
+                    let n = 0;
+                    for (let e = 0; e < g; ++e) n += b[t + e];
+                    y[e] = n
+                }
+                if (i) {
+                    const e = m;
+                    m = rT({
+                        inputs: {
+                            x: m
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: jd.expandShapeToKeepDim(m.shape, l)
+                        }
+                    }), n.disposeIntermediateTensorInfo(e)
+                }
+                return n.disposeIntermediateTensorInfo(o), null != c && n.disposeIntermediateTensorInfo(p), m
+            }
+            var gR = {
+                kernelName: Cr,
+                backendName: "cpu",
+                kernelFunc: mR
+            };
+            var yR = {
+                kernelName: Yt,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        equation: a
+                    } = r, s = t, {
+                        allDims: i,
+                        summedDims: o,
+                        idDims: u
+                    } = jd.decodeEinsumEquation(a, s.length);
+                    jd.checkEinsumDimSizes(i.length, u, s);
+                    const {
+                        path: l,
+                        steps: c
+                    } = jd.getEinsumComputePath(o, u), h = c.length;
+                    let p = null,
+                        d = i.length;
+                    const f = [];
+                    for (let e = 0; e < h; ++e) {
+                        for (const t of c[e]) {
+                            const {
+                                permutationIndices: e,
+                                expandDims: r
+                            } = jd.getEinsumPermutation(d, u[t]);
+                            let a;
+                            jd.isIdentityPermutation(e) ? a = s[t] : (a = uC({
+                                inputs: {
+                                    x: s[t]
+                                },
+                                backend: n,
+                                attrs: {
+                                    perm: e
+                                }
+                            }), f.push(a));
+                            const i = a.shape.slice();
+                            for (let e = 0; e < r.length; ++e) i.splice(r[e], 0, 1);
+                            va.arraysEqual(a.shape, i) || (a = rT({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: i
+                                }
+                            }), f.push(a)), null === p ? p = a : (p = eC({
+                                inputs: {
+                                    a: a,
+                                    b: p
+                                },
+                                backend: n
+                            }), f.push(p))
+                        }
+                        e < h - 1 && (l[e] >= 0 && (p = mR({
+                            inputs: {
+                                x: p
+                            },
+                            backend: n,
+                            attrs: {
+                                axis: l[e] - (i.length - d),
+                                keepDims: !1
+                            }
+                        }), f.push(p)), d--)
+                    }
+                    for (const e of f) e !== p && n.disposeIntermediateTensorInfo(e);
+                    return p
+                }
+            };
+            var bR = {
+                    kernelName: Qt,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            dy: r,
+                            y: a
+                        } = t;
+                        $E([r, a], "eluGrad");
+                        const s = new Float32Array(va.sizeFromShape(a.shape)),
+                            i = n.data.get(a.dataId).values,
+                            o = n.data.get(r.dataId).values;
+                        for (let e = 0; e < i.length; ++e) {
+                            const t = i[e];
+                            s[e] = t >= 1 ? o[e] : o[e] * (t + 1)
+                        }
+                        return n.makeTensorInfo(a.shape, "float32", s)
+                    }
+                },
+                vR = jd.ERF_P,
+                xR = jd.ERF_A1,
+                wR = jd.ERF_A2,
+                kR = jd.ERF_A3,
+                SR = jd.ERF_A4,
+                IR = jd.ERF_A5,
+                AR = lN(Zt, e => {
+                    const t = Math.sign(e),
+                        n = Math.abs(e),
+                        r = 1 / (1 + vR * n);
+                    return t * (1 - ((((IR * r + SR) * r + kR) * r + wR) * r + xR) * r * Math.exp(-n * n))
+                }),
+                ER = {
+                    kernelName: Zt,
+                    backendName: "cpu",
+                    kernelFunc: AR
+                };
+
+            function NR(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    input: a
+                } = t, {
+                    dim: s
+                } = r, i = a.shape.length, o = a.shape.slice();
+                let u = s;
+                return s < 0 && (va.assert(-(i + 1) <= s, () => `Axis must be in the interval [${-(i+1)}, ${i}]`), u = i + s + 1), o.splice(u, 0, 1), rT({
+                    inputs: {
+                        x: a
+                    },
+                    backend: n,
+                    attrs: {
+                        shape: o
+                    }
+                })
+            }
+            var CR = {
+                    kernelName: tn,
+                    backendName: "cpu",
+                    kernelFunc: NR
+                },
+                TR = jE((e, t) => e / t),
+                RR = eN(Kt, TR),
+                _R = {
+                    kernelName: Kt,
+                    backendName: "cpu",
+                    kernelFunc: RR
+                };
+
+            function FR(e, t, n) {
+                const r = e.shape,
+                    a = r[0],
+                    s = r[1],
+                    i = n.data.get(e.dataId),
+                    o = i.complexTensorInfos.real,
+                    u = i.complexTensorInfos.imag,
+                    l = [a, s],
+                    c = va.sizeFromShape(l),
+                    h = va.getTypedArrayFromDType("float32", c),
+                    p = va.getTypedArrayFromDType("float32", c);
+                for (let e = 0; e < a; e++) {
+                    const r = xC({
+                            inputs: {
+                                x: o
+                            },
+                            backend: n,
+                            attrs: {
+                                begin: [e, 0],
+                                size: [1, s]
+                            }
+                        }),
+                        a = xC({
+                            inputs: {
+                                x: u
+                            },
+                            backend: n,
+                            attrs: {
+                                begin: [e, 0],
+                                size: [1, s]
+                            }
+                        }),
+                        i = GE({
+                            inputs: {
+                                real: r,
+                                imag: a
+                            },
+                            backend: n
+                        }),
+                        {
+                            real: l,
+                            imag: c
+                        } = MR(i, t, n),
+                        d = jd.mergeRealAndImagArrays(l, c);
+                    for (let t = 0; t < s; t++) {
+                        const n = jd.getComplexWithIndex(d, t);
+                        h[e * s + t] = n.real, p[e * s + t] = n.imag
+                    }
+                    n.disposeIntermediateTensorInfo(r), n.disposeIntermediateTensorInfo(a), n.disposeIntermediateTensorInfo(i)
+                }
+                const d = n.makeTensorInfo(l, "float32", h),
+                    f = n.makeTensorInfo(l, "float32", p),
+                    m = GE({
+                        inputs: {
+                            real: d,
+                            imag: f
+                        },
+                        backend: n
+                    });
+                return n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(f), m
+            }
+
+            function MR(e, t, n) {
+                const r = va.sizeFromShape(e.shape),
+                    a = n.data.get(e.dataId),
+                    s = n.data.get(a.complexTensorInfos.real.dataId).values,
+                    i = n.data.get(a.complexTensorInfos.imag.dataId).values;
+                if (0 == ((o = r) & o - 1)) {
+                    const a = function e(t, n, r, a, s) {
+                            if (1 === r) return {
+                                real: t,
+                                imag: n
+                            };
+                            const i = jd.mergeRealAndImagArrays(t, n),
+                                o = r / 2,
+                                u = jd.complexWithEvenIndex(i),
+                                l = u.real,
+                                c = u.imag,
+                                h = [l.length],
+                                p = s.makeTensorInfo(h, "float32", l),
+                                d = s.makeTensorInfo(h, "float32", c),
+                                f = GE({
+                                    inputs: {
+                                        real: p,
+                                        imag: d
+                                    },
+                                    backend: s
+                                }),
+                                m = jd.complexWithOddIndex(i),
+                                g = m.real,
+                                y = m.imag,
+                                b = [g.length],
+                                v = s.makeTensorInfo(b, "float32", g),
+                                x = s.makeTensorInfo(b, "float32", y),
+                                w = GE({
+                                    inputs: {
+                                        real: v,
+                                        imag: x
+                                    },
+                                    backend: s
+                                }),
+                                k = e(l, c, o, a, s),
+                                S = k.real,
+                                I = k.imag,
+                                A = [S.length],
+                                E = s.makeTensorInfo(A, "float32", S),
+                                N = s.makeTensorInfo(A, "float32", I),
+                                C = GE({
+                                    inputs: {
+                                        real: E,
+                                        imag: N
+                                    },
+                                    backend: s
+                                }),
+                                T = e(g, y, o, a, s),
+                                R = T.real,
+                                _ = T.imag,
+                                F = [R.length],
+                                M = s.makeTensorInfo(F, "float32", R),
+                                D = s.makeTensorInfo(F, "float32", _),
+                                O = GE({
+                                    inputs: {
+                                        real: M,
+                                        imag: D
+                                    },
+                                    backend: s
+                                }),
+                                L = jd.exponents(r, a),
+                                P = [L.real.length],
+                                $ = s.makeTensorInfo(P, "float32", L.real),
+                                B = s.makeTensorInfo(P, "float32", L.imag),
+                                z = GE({
+                                    inputs: {
+                                        real: $,
+                                        imag: B
+                                    },
+                                    backend: s
+                                }),
+                                U = eC({
+                                    inputs: {
+                                        a: z,
+                                        b: O
+                                    },
+                                    backend: s
+                                }),
+                                W = aN({
+                                    inputs: {
+                                        a: C,
+                                        b: U
+                                    },
+                                    backend: s
+                                }),
+                                V = $C({
+                                    inputs: {
+                                        a: C,
+                                        b: U
+                                    },
+                                    backend: s
+                                }),
+                                j = XE({
+                                    inputs: {
+                                        input: W
+                                    },
+                                    backend: s
+                                }),
+                                G = XE({
+                                    inputs: {
+                                        input: V
+                                    },
+                                    backend: s
+                                }),
+                                H = WT({
+                                    inputs: {
+                                        input: W
+                                    },
+                                    backend: s
+                                }),
+                                q = WT({
+                                    inputs: {
+                                        input: V
+                                    },
+                                    backend: s
+                                }),
+                                K = jT({
+                                    inputs: [j, G],
+                                    backend: s,
+                                    attrs: {
+                                        axis: 0
+                                    }
+                                }),
+                                Y = jT({
+                                    inputs: [H, q],
+                                    backend: s,
+                                    attrs: {
+                                        axis: 0
+                                    }
+                                }),
+                                X = s.data.get(K.dataId).values,
+                                Q = s.data.get(Y.dataId).values;
+                            return s.disposeIntermediateTensorInfo(p), s.disposeIntermediateTensorInfo(d), s.disposeIntermediateTensorInfo(f), s.disposeIntermediateTensorInfo(v), s.disposeIntermediateTensorInfo(x), s.disposeIntermediateTensorInfo(w), s.disposeIntermediateTensorInfo(E), s.disposeIntermediateTensorInfo(N), s.disposeIntermediateTensorInfo(C), s.disposeIntermediateTensorInfo(M), s.disposeIntermediateTensorInfo(D), s.disposeIntermediateTensorInfo(O), s.disposeIntermediateTensorInfo($), s.disposeIntermediateTensorInfo(B), s.disposeIntermediateTensorInfo(z), s.disposeIntermediateTensorInfo(U), s.disposeIntermediateTensorInfo(W), s.disposeIntermediateTensorInfo(V), s.disposeIntermediateTensorInfo(j), s.disposeIntermediateTensorInfo(H), s.disposeIntermediateTensorInfo(G), s.disposeIntermediateTensorInfo(q), s.disposeIntermediateTensorInfo(K), s.disposeIntermediateTensorInfo(Y), {
+                                real: X,
+                                imag: Q
+                            }
+                        }(s, i, r, t, n),
+                        o = [e.shape[0], e.shape[1]];
+                    if (t) {
+                        const e = n.makeTensorInfo(o, "float32", a.real),
+                            t = n.makeTensorInfo(o, "float32", a.imag),
+                            s = n.makeTensorInfo([], "float32", va.createScalarValue(r, "float32")),
+                            i = KE({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: n
+                            }),
+                            u = _R.kernelFunc({
+                                inputs: {
+                                    a: e,
+                                    b: s
+                                },
+                                backend: n
+                            }),
+                            l = _R.kernelFunc({
+                                inputs: {
+                                    a: t,
+                                    b: i
+                                },
+                                backend: n
+                            }),
+                            c = n.data.get(u.dataId).values,
+                            h = n.data.get(l.dataId).values;
+                        return n.disposeIntermediateTensorInfo(e), n.disposeIntermediateTensorInfo(t), n.disposeIntermediateTensorInfo(s), n.disposeIntermediateTensorInfo(i), n.disposeIntermediateTensorInfo(u), n.disposeIntermediateTensorInfo(l), {
+                            real: c,
+                            imag: h
+                        }
+                    }
+                    return a
+                } {
+                    const e = function (e, t, n) {
+                        const r = new Float32Array(2 * t);
+                        for (let a = 0; a < t; a++) {
+                            let s = 0,
+                                i = 0;
+                            for (let r = 0; r < t; r++) {
+                                const o = jd.exponent(a * r, t, n),
+                                    u = jd.getComplexWithIndex(e, r);
+                                s += u.real * o.real - u.imag * o.imag, i += u.real * o.imag + u.imag * o.real
+                            }
+                            n && (s /= t, i /= t), jd.assignToTypedArray(r, s, i, a)
+                        }
+                        return r
+                    }(jd.mergeRealAndImagArrays(s, i), r, t);
+                    return jd.splitRealAndImagArrays(e)
+                }
+                var o
+            }
+            var DR = {
+                kernelName: rn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        input: r
+                    } = t, a = va.sizeFromShape(r.shape), s = r.shape[r.shape.length - 1], i = rT({
+                        inputs: {
+                            x: r
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [a / s, s]
+                        }
+                    }), o = FR(i, !1, n), u = rT({
+                        inputs: {
+                            x: o
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: r.shape
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(i), n.disposeIntermediateTensorInfo(o), u
+                }
+            };
+
+            function OR(e) {
+                const {
+                    backend: t,
+                    attrs: n
+                } = e, {
+                    shape: r,
+                    value: a,
+                    dtype: s
+                } = n, i = s || va.inferDtype(a), o = va.getArrayFromDType(i, va.sizeFromShape(r));
+                return function (e, t, n) {
+                    e.fill(t)
+                }(o, a), t.makeTensorInfo(r, i, o)
+            }
+            var LR = {
+                kernelName: an,
+                backendName: "cpu",
+                kernelFunc: OR
+            };
+            var PR = {
+                    kernelName: sn,
+                    backendName: "cpu",
+                    kernelFunc: ({
+                        inputs: e,
+                        attrs: t,
+                        backend: n
+                    }) => {
+                        const {
+                            image: r
+                        } = e, a = n, s = va.getTypedArrayFromDType(r.dtype, va.sizeFromShape(r.shape)), [i, o, u, l] = r.shape, c = a.data.get(r.dataId).values;
+                        for (let e = 0; e < i; e++) {
+                            const t = e * u * o * l;
+                            for (let e = 0; e < o; e++) {
+                                const n = e * (u * l);
+                                for (let e = 0; e < u; e++) {
+                                    const r = e * l;
+                                    for (let a = 0; a < l; a++) {
+                                        const i = Math.round(u - e - 1),
+                                            o = t + n + r + a;
+                                        let h = c[o];
+                                        if (i >= 0 && i < u) {
+                                            h = c[t + n + i * l + a]
+                                        }
+                                        s[o] = h
+                                    }
+                                }
+                            }
+                        }
+                        return {
+                            dataId: a.write(s, r.shape, r.dtype),
+                            shape: r.shape,
+                            dtype: r.dtype
+                        }
+                    }
+                },
+                $R = jE((e, t) => Math.floor(e / t)),
+                BR = eN(un, $R, null, "int32"),
+                zR = {
+                    kernelName: un,
+                    backendName: "cpu",
+                    kernelFunc: BR
+                };
+            var UR = {
+                kernelName: aa,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t, {
+                        strides: u,
+                        pad: l,
+                        dataFormat: c,
+                        dilations: h,
+                        dimRoundingMode: p,
+                        activation: d,
+                        leakyreluAlpha: f
+                    } = r;
+                    let m = HT({
+                        inputs: {
+                            x: a,
+                            filter: s
+                        },
+                        backend: n,
+                        attrs: {
+                            strides: u,
+                            pad: l,
+                            dataFormat: c,
+                            dilations: h,
+                            dimRoundingMode: p
+                        }
+                    });
+                    if (i) {
+                        const e = m;
+                        m = aN({
+                            inputs: {
+                                a: m,
+                                b: i
+                            },
+                            backend: n
+                        }), n.disposeIntermediateTensorInfo(e)
+                    }
+                    if (d) {
+                        const e = m;
+                        m = nT(n, m, d, o, f), n.disposeIntermediateTensorInfo(e)
+                    }
+                    return m
+                }
+            };
+            var WR = {
+                kernelName: sa,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t, {
+                        strides: u,
+                        pad: l,
+                        dataFormat: c,
+                        dilations: h,
+                        dimRoundingMode: p,
+                        activation: d,
+                        leakyreluAlpha: f
+                    } = r;
+                    let m = oR({
+                        inputs: {
+                            x: a,
+                            filter: s
+                        },
+                        backend: n,
+                        attrs: {
+                            strides: u,
+                            pad: l,
+                            dataFormat: c,
+                            dilations: h,
+                            dimRoundingMode: p
+                        }
+                    });
+                    if (i) {
+                        const e = m;
+                        m = aN({
+                            inputs: {
+                                a: m,
+                                b: i
+                            },
+                            backend: n
+                        }), n.disposeIntermediateTensorInfo(e)
+                    }
+                    if (d) {
+                        const e = m;
+                        m = nT(n, m, d, o, f), n.disposeIntermediateTensorInfo(e)
+                    }
+                    return m
+                }
+            };
+            var VR = {
+                kernelName: hn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        params: r,
+                        indices: a
+                    } = t, s = va.sizeFromShape(r.shape), i = a.shape, o = i[i.length - 1], [u, l, c, h] = jd.prepareAndValidate(r, a);
+                    if (0 === l) return n.makeTensorInfo(u, r.dtype, []);
+                    const p = NN(n.data.get(a.dataId).values, n.bufferSync(r), r.dtype, l, o, c, h, r.shape, s);
+                    return n.makeTensorInfo(u, r.dtype, p.values)
+                }
+            };
+            var jR = {
+                kernelName: cn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        indices: s
+                    } = t, {
+                        axis: i,
+                        batchDims: o
+                    } = r;
+                    $E([a, s], "gatherV2");
+                    const u = va.parseAxisParam(i, a.shape)[0],
+                        l = n.data.get(s.dataId).values,
+                        c = a.shape[u];
+                    for (let e = 0; e < l.length; ++e) {
+                        const t = l[e];
+                        va.assert(t <= c - 1 && t >= 0, () => `GatherV2: the index value ${t} is not in [0, ${c-1}]`)
+                    }
+                    let h = o;
+                    null == o && (h = 0);
+                    const p = va.sizeFromShape(s.shape),
+                        d = jd.segment_util.collectGatherOpShapeInfo(a, s, u, h),
+                        f = rT({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [d.batchSize, d.outerSize, d.dimSize, d.sliceSize]
+                            }
+                        }),
+                        m = rT({
+                            inputs: {
+                                x: s
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [d.batchSize, p / d.batchSize]
+                            }
+                        }),
+                        g = [d.batchSize, d.outerSize, p / d.batchSize, d.sliceSize],
+                        y = n.bufferSync(m),
+                        b = CN(n.bufferSync(f), y, g);
+                    return n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(m), n.makeTensorInfo(d.outputShape, b.dtype, b.values)
+                }
+            };
+            var GR = {
+                    kernelName: mn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            input: r
+                        } = t, a = va.sizeFromShape(r.shape), s = r.shape[r.shape.length - 1], i = rT({
+                            inputs: {
+                                x: r
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [a / s, s]
+                            }
+                        }), o = FR(i, !0, n), u = rT({
+                            inputs: {
+                                x: o
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: r.shape
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(i), n.disposeIntermediateTensorInfo(o), u
+                    }
+                },
+                HR = lN(yn, e => Number.isFinite(e) ? 1 : 0, "bool"),
+                qR = {
+                    kernelName: yn,
+                    backendName: "cpu",
+                    kernelFunc: HR
+                },
+                KR = lN(bn, e => Math.abs(e) === 1 / 0 ? 1 : 0, "bool"),
+                YR = {
+                    kernelName: bn,
+                    backendName: "cpu",
+                    kernelFunc: KR
+                },
+                XR = lN(vn, e => Number.isNaN(e) ? 1 : 0, "bool"),
+                QR = {
+                    kernelName: vn,
+                    backendName: "cpu",
+                    kernelFunc: XR
+                };
+            var ZR = {
+                    kernelName: Sn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            attrs: n
+                        } = e, {
+                            start: r,
+                            stop: a,
+                            num: s
+                        } = n, i = UN(r, a, s);
+                        return t.makeTensorInfo([i.length], "float32", i)
+                    }
+                },
+                JR = lN(An, e => Math.log1p(e)),
+                e_ = {
+                    kernelName: An,
+                    backendName: "cpu",
+                    kernelFunc: JR
+                },
+                t_ = jE((e, t) => e && t),
+                n_ = eN(En, t_, null, "bool"),
+                r_ = {
+                    kernelName: En,
+                    backendName: "cpu",
+                    kernelFunc: n_
+                },
+                a_ = lN(Nn, e => e ? 0 : 1, "bool"),
+                s_ = {
+                    kernelName: Nn,
+                    backendName: "cpu",
+                    kernelFunc: a_
+                },
+                i_ = jE((e, t) => e || t),
+                o_ = eN(Cn, i_, null, "bool"),
+                u_ = {
+                    kernelName: Cn,
+                    backendName: "cpu",
+                    kernelFunc: o_
+                };
+            var l_ = {
+                kernelName: Rn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        depthRadius: s,
+                        bias: i,
+                        alpha: o,
+                        beta: u
+                    } = r;
+                    $E(a, "LRN");
+                    const l = a.shape[3],
+                        c = l - 1,
+                        h = n.data.get(a.dataId).values,
+                        p = va.sizeFromShape(a.shape),
+                        d = new Float32Array(p);
+
+                    function f(e) {
+                        const t = e % l;
+                        let n = e - t + Math.max(0, t - s);
+                        const r = e - t + Math.min(t + s, c);
+                        let a = 0;
+                        for (; n <= r; n++) {
+                            const e = h[n];
+                            a += e * e
+                        }
+                        return a
+                    }
+                    for (let e = 0; e < p; e++) {
+                        const t = f(e),
+                            n = h[e] * Math.pow(i + o * t, -u);
+                        d[e] = n
+                    }
+                    return n.makeTensorInfo(a.shape, a.dtype, d)
+                }
+            };
+            var c_ = {
+                kernelName: _n,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        y: s,
+                        dy: i
+                    } = t, {
+                        depthRadius: o,
+                        bias: u,
+                        alpha: l,
+                        beta: c
+                    } = r;
+                    $E(i, "LRNGrad");
+                    const h = va.sizeFromShape(i.shape),
+                        p = i.shape[3],
+                        d = n.data.get(i.dataId).values,
+                        f = n.data.get(a.dataId).values,
+                        m = n.data.get(s.dataId).values,
+                        g = new Float32Array(h),
+                        y = h;
+                    for (let e = 0; e < y; e++) {
+                        const t = e % p,
+                            n = e - t + Math.max(0, t - o),
+                            r = e - t + Math.min(p, t + o + 1);
+                        let a = 0;
+                        for (let e = n; e < r; e++) a += Math.pow(f[e], 2);
+                        a = l * a + u;
+                        for (let t = n; t < r; t++) {
+                            let n = -2 * l * c * f[t] * m[e] / a;
+                            e === t && (n += Math.pow(a, -c)), n *= d[e], g[t] += n
+                        }
+                    }
+                    return n.makeTensorInfo(i.shape, a.dtype, g)
+                }
+            };
+
+            function h_(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    reductionIndices: s,
+                    keepDims: i
+                } = r, o = n;
+                let u = a.shape;
+                const l = u.length,
+                    c = va.parseAxisParam(s, u);
+                let h = c;
+                const p = jd.getAxesPermutation(h, l);
+                let d = o.data.get(a.dataId).values;
+                if (null != p) {
+                    const e = new Array(l);
+                    for (let t = 0; t < e.length; t++) e[t] = u[p[t]];
+                    d = oC(d, u, a.dtype, p, e), h = jd.getInnerMostAxes(h.length, l), u = e
+                }
+                $E(a, "max"), jd.assertAxesAreInnerMostDims("max", h, l);
+                const [f, m] = jd.computeOutAndReduceShapes(u, h), g = GN(d, va.sizeFromShape(m), f, a.dtype), y = o.write(g, f, a.dtype);
+                let b = f;
+                if (i) {
+                    b = jd.expandShapeToKeepDim(f, c)
+                }
+                return {
+                    dataId: y,
+                    shape: b,
+                    dtype: a.dtype
+                }
+            }
+            var p_ = {
+                kernelName: Fn,
+                backendName: "cpu",
+                kernelFunc: h_
+            };
+            var d_ = {
+                kernelName: Dn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t;
+                    $E(a, "maxPool");
+                    const {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u
+                    } = r;
+                    va.assert(jd.eitherStridesOrDilationsAreOne(i, 1), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);
+                    const l = jd.computePool2DInfo(a.shape, s, i, 1, o, u);
+                    let c;
+                    if (1 === l.filterWidth && 1 === l.filterHeight && va.arraysEqual(l.inShape, l.outShape)) c = KE({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    else {
+                        const e = n.data.get(a.dataId).values,
+                            t = va.computeStrides(a.shape),
+                            r = CT(e, a.shape, a.dtype, t, l, "max");
+                        c = n.makeTensorInfo(l.outShape, a.dtype, r.values)
+                    }
+                    return c
+                }
+            };
+            var f_ = {
+                kernelName: Ln,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u,
+                        dataFormat: l
+                    } = r;
+                    $E(a, "maxPool3d");
+                    const c = jd.computePool3DInfo(a.shape, s, i, 1, o, u, l),
+                        h = RT(n.data.get(a.dataId).values, a.shape, a.dtype, va.computeStrides(a.shape), c, "max");
+                    return n.makeTensorInfo(h.shape, "float32", h.values)
+                }
+            };
+            var m_ = {
+                kernelName: Pn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s
+                    } = t, {
+                        filterSize: i,
+                        strides: o,
+                        pad: u,
+                        dimRoundingMode: l
+                    } = r;
+                    $E([a, s], "maxPool3DGrad");
+                    const c = jd.computePool3DInfo(s.shape, i, o, 1, u, l),
+                        h = function (e, t) {
+                            const n = Ii(t.outShape, "int32"),
+                                r = t.strideDepth,
+                                a = t.strideHeight,
+                                s = t.strideWidth,
+                                i = t.dilationDepth,
+                                o = t.dilationHeight,
+                                u = t.dilationWidth,
+                                l = t.effectiveFilterDepth,
+                                c = t.effectiveFilterHeight,
+                                h = t.effectiveFilterWidth,
+                                p = t.padInfo.front,
+                                d = t.padInfo.top,
+                                f = t.padInfo.left;
+                            for (let m = 0; m < t.batchSize; ++m)
+                                for (let g = 0; g < t.inChannels; ++g)
+                                    for (let y = 0; y < t.outDepth; ++y) {
+                                        const b = y * r - p;
+                                        let v = b;
+                                        for (; v < 0;) v += i;
+                                        const x = Math.min(t.inDepth, l + b);
+                                        for (let r = 0; r < t.outHeight; ++r) {
+                                            const l = r * a - d;
+                                            let p = l;
+                                            for (; p < 0;) p += o;
+                                            const w = Math.min(t.inHeight, c + l);
+                                            for (let a = 0; a < t.outWidth; ++a) {
+                                                const d = a * s - f;
+                                                let k = d;
+                                                for (; k < 0;) k += u;
+                                                const S = Math.min(t.inWidth, h + d);
+                                                let I = Number.NEGATIVE_INFINITY,
+                                                    A = -1;
+                                                for (let t = v; t < x; t += i) {
+                                                    const n = t - b;
+                                                    for (let r = p; r < w; r += o) {
+                                                        const a = r - l;
+                                                        for (let s = k; s < S; s += u) {
+                                                            const i = s - d,
+                                                                o = e.get(m, t, r, s, g);
+                                                            o >= I && (I = o, A = n * c * h + a * c + i)
+                                                        }
+                                                    }
+                                                }
+                                                n.set(A, m, y, r, a, g)
+                                            }
+                                        }
+                                    }
+                            return n
+                        }(n.bufferSync(s), c),
+                        p = c.strideDepth,
+                        d = c.strideHeight,
+                        f = c.strideWidth,
+                        m = c.dilationDepth,
+                        g = c.dilationHeight,
+                        y = c.dilationWidth,
+                        b = c.effectiveFilterDepth,
+                        v = c.effectiveFilterHeight,
+                        x = c.effectiveFilterWidth,
+                        w = b - 1 - c.padInfo.front,
+                        k = x - 1 - c.padInfo.left,
+                        S = v - 1 - c.padInfo.top,
+                        I = Ii(s.shape, "float32"),
+                        A = n.bufferSync(a);
+                    for (let e = 0; e < c.batchSize; ++e)
+                        for (let t = 0; t < c.inChannels; ++t)
+                            for (let n = 0; n < c.inDepth; ++n)
+                                for (let r = 0; r < c.inHeight; ++r)
+                                    for (let a = 0; a < c.inWidth; ++a) {
+                                        const s = n - w,
+                                            i = r - S,
+                                            o = a - k;
+                                        let u = 0;
+                                        for (let n = 0; n < b; n += m) {
+                                            const r = (s + n) / p;
+                                            if (!(r < 0 || r >= c.outDepth || Math.floor(r) !== r))
+                                                for (let a = 0; a < v; a += g) {
+                                                    const s = (i + a) / d;
+                                                    if (!(s < 0 || s >= c.outHeight || Math.floor(s) !== s))
+                                                        for (let i = 0; i < x; i += y) {
+                                                            const l = (o + i) / f;
+                                                            if (l < 0 || l >= c.outWidth || Math.floor(l) !== l) continue;
+                                                            const p = b * v * x - 1 - h.get(e, r, s, l, t) === n * v * x + a * x + i ? 1 : 0;
+                                                            if (0 === p) continue;
+                                                            u += A.get(e, r, s, l, t) * p
+                                                        }
+                                                }
+                                        }
+                                        I.set(u, e, n, r, a, t)
+                                    }
+                    return n.makeTensorInfo(I.shape, I.dtype, I.values)
+                }
+            };
+            var g_ = {
+                kernelName: On,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s,
+                        output: i
+                    } = t, o = s;
+                    $E([s, i], "maxPoolGrad");
+                    const {
+                        filterSize: u,
+                        strides: l,
+                        pad: c,
+                        dimRoundingMode: h
+                    } = r, p = jd.computePool2DInfo(o.shape, u, l, 1, c, h), d = n.data.get(o.dataId).values, f = Ii(p.outShape, o.dtype, TT(d, o.shape, o.dtype, p).values), m = p.strideHeight, g = p.strideWidth, y = p.dilationHeight, b = p.dilationWidth, v = p.effectiveFilterHeight, x = p.effectiveFilterWidth, w = x - 1 - p.padInfo.left, k = v - 1 - p.padInfo.top, S = Ii(o.shape, "float32"), I = n.data.get(a.dataId).values, A = Ii(a.shape, "float32", I);
+                    for (let e = 0; e < p.batchSize; ++e)
+                        for (let t = 0; t < p.inChannels; ++t)
+                            for (let n = 0; n < p.inHeight; ++n)
+                                for (let r = 0; r < p.inWidth; ++r) {
+                                    const a = n - k,
+                                        s = r - w;
+                                    let i = 0;
+                                    for (let n = 0; n < v; n += y) {
+                                        const r = (a + n) / m;
+                                        if (!(r < 0 || r >= p.outHeight || Math.floor(r) !== r))
+                                            for (let a = 0; a < x; a += b) {
+                                                const o = (s + a) / g;
+                                                if (o < 0 || o >= p.outWidth || Math.floor(o) !== o) continue;
+                                                const u = v * x - 1 - f.get(e, r, o, t) === n * x + a ? 1 : 0;
+                                                if (0 === u) continue;
+                                                i += A.get(e, r, o, t) * u
+                                            }
+                                    }
+                                    S.set(i, e, n, r, t)
+                                }
+                    return n.makeTensorInfo(S.shape, S.dtype, S.values)
+                }
+            };
+            var y_ = {
+                kernelName: $n,
+                backendName: "cpu",
+                kernelFunc: ({
+                    inputs: e,
+                    attrs: t,
+                    backend: n
+                }) => {
+                    const {
+                        x: r
+                    } = e, {
+                        filterSize: a,
+                        strides: s,
+                        pad: i,
+                        includeBatchInIndex: o
+                    } = t, u = n;
+                    $E(r, "MaxPoolWithArgmax");
+                    const l = u.data.get(r.dataId).values,
+                        c = jd.computePool2DInfo(r.shape, a, s, [1, 1], i),
+                        [h, p] = function (e, t, n, r, a) {
+                            const s = CT(e, 0, n, va.computeStrides(t), a, "max"),
+                                i = TT(e, t, n, a, !0, r);
+                            return [s.values, i.values]
+                        }(l, r.shape, r.dtype, o, c),
+                        d = u.write(h, c.outShape, r.dtype),
+                        f = u.write(p, c.outShape, r.dtype);
+                    return [{
+                        dataId: d,
+                        shape: c.outShape,
+                        dtype: r.dtype
+                    }, {
+                        dataId: f,
+                        shape: c.outShape,
+                        dtype: "int32"
+                    }]
+                }
+            };
+            var b_ = {
+                kernelName: Bn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r, o = va.parseAxisParam(s, a.shape), u = jd.computeOutAndReduceShapes(a.shape, o)[1], l = va.sizeFromShape(u), c = [], h = n.makeTensorInfo([], "float32", new Float32Array([l]));
+                    c.push(h);
+                    const p = ZE({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            dtype: "float32"
+                        }
+                    });
+                    c.push(p);
+                    const d = RR({
+                        inputs: {
+                            a: p,
+                            b: h
+                        },
+                        backend: n
+                    });
+                    c.push(d);
+                    const f = mR({
+                        inputs: {
+                            x: d
+                        },
+                        backend: n,
+                        attrs: {
+                            axis: s,
+                            keepDims: i
+                        }
+                    });
+                    return c.forEach(e => n.disposeIntermediateTensorInfo(e)), f
+                }
+            };
+            var v_ = {
+                kernelName: zn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r;
+                    $E(a, "min");
+                    const o = va.parseAxisParam(s, a.shape);
+                    let u = o;
+                    const l = jd.getAxesPermutation(u, a.shape.length);
+                    let c = a;
+                    null != l && (c = uC({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: l
+                        }
+                    }), u = jd.getInnerMostAxes(u.length, a.shape.length)), jd.assertAxesAreInnerMostDims("min", u, c.shape.length);
+                    const [h, p] = jd.computeOutAndReduceShapes(c.shape, u), d = va.sizeFromShape(p), f = va.makeZerosTypedArray(va.sizeFromShape(h), c.dtype), m = n.data.get(c.dataId).values;
+                    for (let e = 0; e < f.length; ++e) {
+                        const t = e * d;
+                        let n = m[t];
+                        for (let e = 0; e < d; ++e) {
+                            const r = m[t + e];
+                            (Number.isNaN(r) || r < n) && (n = r)
+                        }
+                        f[e] = n
+                    }
+                    null != l && n.disposeIntermediateTensorInfo(c);
+                    const g = n.makeTensorInfo(h, c.dtype, f);
+                    if (i) {
+                        const e = rT({
+                            inputs: {
+                                x: g
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: jd.expandShapeToKeepDim(h, o)
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(g), e
+                    }
+                    return g
+                }
+            };
+            var x_ = {
+                    kernelName: Wn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            paddings: s,
+                            mode: i
+                        } = r;
+                        $E(a, "mirrorPad");
+                        const o = s.map((e, t) => e[0] + a.shape[t] + e[1]),
+                            u = s.map(e => e[0]),
+                            l = s.map((e, t) => e[0] + a.shape[t]),
+                            c = "reflect" === i ? 0 : 1,
+                            h = n.data.get(a.dataId).values,
+                            p = a.shape.length,
+                            d = va.computeStrides(a.shape),
+                            f = va.sizeFromShape(o),
+                            m = o.length,
+                            g = va.computeStrides(o),
+                            y = va.getTypedArrayFromDType(a.dtype, f);
+                        for (let e = 0; e < f; e++) {
+                            let t = va.indexToLoc(e, m, g);
+                            for (let e = 0; e < m; e++) t[e] < u[e] ? t[e] = 2 * u[e] - t[e] - c : t[e] >= l[e] && (t[e] = 2 * (l[e] - 1) - t[e] + c);
+                            t = t.map((e, t) => e - u[t]);
+                            const n = va.locToIndex(t, p, d);
+                            y[e] = h[n]
+                        }
+                        return {
+                            dataId: n.write(y, o, a.dtype),
+                            shape: o,
+                            dtype: a.dtype
+                        }
+                    }
+                },
+                w_ = jE((e, t) => {
+                    const n = e % t;
+                    return e < 0 && t < 0 || e >= 0 && t >= 0 ? n : (n + t) % t
+                }),
+                k_ = eN(Vn, w_),
+                S_ = {
+                    kernelName: Vn,
+                    backendName: "cpu",
+                    kernelFunc: k_
+                },
+                I_ = k(D());
+
+            function A_(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    logits: a
+                } = t, {
+                    dim: s
+                } = r, i = a.shape.length;
+                let o = s;
+                if (-1 === o && (o = i - 1), o !== i - 1) throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${o}`);
+                const u = va.parseAxisParam([o], a.shape),
+                    l = h_({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            reductionIndices: u,
+                            keepDims: !1
+                        }
+                    }),
+                    c = jd.expandShapeToKeepDim(l.shape, u),
+                    h = rT({
+                        inputs: {
+                            x: l
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: c
+                        }
+                    }),
+                    p = $C({
+                        inputs: {
+                            a: a,
+                            b: h
+                        },
+                        backend: n
+                    }),
+                    d = vN({
+                        inputs: {
+                            x: p
+                        },
+                        backend: n
+                    }),
+                    f = mR({
+                        inputs: {
+                            x: d
+                        },
+                        backend: n,
+                        attrs: {
+                            axis: u,
+                            keepDims: !1
+                        }
+                    }),
+                    m = rT({
+                        inputs: {
+                            x: f
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: c
+                        }
+                    }),
+                    g = RR({
+                        inputs: {
+                            a: d,
+                            b: m
+                        },
+                        backend: n
+                    });
+                return n.disposeIntermediateTensorInfo(l), n.disposeIntermediateTensorInfo(h), n.disposeIntermediateTensorInfo(p), n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(m), g
+            }
+            var E_ = {
+                kernelName: _r,
+                backendName: "cpu",
+                kernelFunc: A_
+            };
+            var N_ = {
+                    kernelName: jn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            logits: a
+                        } = t, {
+                            numSamples: s,
+                            seed: i,
+                            normalized: o
+                        } = r;
+                        $E(a, "multinomial");
+                        const u = o ? a : A_({
+                                inputs: {
+                                    logits: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    dim: -1
+                                }
+                            }),
+                            l = u.shape[0],
+                            c = u.shape[1],
+                            h = n.data.get(u.dataId).values,
+                            p = [l, s],
+                            d = va.makeZerosTypedArray(va.sizeFromShape(p), "int32");
+                        for (let e = 0; e < l; ++e) {
+                            const t = e * c,
+                                n = new Float32Array(c - 1);
+                            n[0] = h[t];
+                            for (let e = 1; e < n.length; ++e) n[e] = n[e - 1] + h[t + e];
+                            const r = I_.alea(i.toString()),
+                                a = e * s;
+                            for (let e = 0; e < s; ++e) {
+                                const t = r();
+                                d[a + e] = n.length;
+                                for (let r = 0; r < n.length; r++)
+                                    if (t < n[r]) {
+                                        d[a + e] = r;
+                                        break
+                                    }
+                            }
+                        }
+                        return o || n.disposeIntermediateTensorInfo(u), n.makeTensorInfo(p, "int32", d)
+                    }
+                },
+                C_ = jf.nonMaxSuppressionV3Impl;
+            var T_ = {
+                    kernelName: Kn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            boxes: a,
+                            scores: s
+                        } = t, {
+                            maxOutputSize: i,
+                            iouThreshold: o,
+                            scoreThreshold: u
+                        } = r;
+                        $E(a, "NonMaxSuppression");
+                        const l = n.data.get(a.dataId).values,
+                            c = n.data.get(s.dataId).values,
+                            {
+                                selectedIndices: h
+                            } = C_(l, c, i, o, u);
+                        return n.makeTensorInfo([h.length], "int32", new Int32Array(h))
+                    }
+                },
+                R_ = jf.nonMaxSuppressionV4Impl;
+            var __ = {
+                    kernelName: Yn,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            boxes: a,
+                            scores: s
+                        } = t, {
+                            maxOutputSize: i,
+                            iouThreshold: o,
+                            scoreThreshold: u,
+                            padToMaxOutputSize: l
+                        } = r;
+                        $E(a, "NonMaxSuppressionPadded");
+                        const c = n.data.get(a.dataId).values,
+                            h = n.data.get(s.dataId).values,
+                            {
+                                selectedIndices: p,
+                                validOutputs: d
+                            } = R_(c, h, i, o, u, l);
+                        return [n.makeTensorInfo([p.length], "int32", new Int32Array(p)), n.makeTensorInfo([], "int32", new Int32Array([d]))]
+                    }
+                },
+                F_ = jf.nonMaxSuppressionV5Impl;
+            var M_ = {
+                kernelName: Xn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        boxes: a,
+                        scores: s
+                    } = t, {
+                        maxOutputSize: i,
+                        iouThreshold: o,
+                        scoreThreshold: u,
+                        softNmsSigma: l
+                    } = r;
+                    $E(a, "NonMaxSuppressionWithScore");
+                    const c = n.data.get(a.dataId).values,
+                        h = n.data.get(s.dataId).values,
+                        p = i,
+                        d = o,
+                        f = u,
+                        m = l,
+                        {
+                            selectedIndices: g,
+                            selectedScores: y
+                        } = F_(c, h, p, d, f, m);
+                    return [n.makeTensorInfo([g.length], "int32", new Int32Array(g)), n.makeTensorInfo([y.length], "float32", new Float32Array(y))]
+                }
+            };
+            var D_ = {
+                kernelName: Zn,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        indices: a
+                    } = t, {
+                        depth: s,
+                        onValue: i,
+                        offValue: o
+                    } = r;
+                    $E(a, "oneHot");
+                    const u = va.sizeFromShape(a.shape),
+                        l = new Float32Array(u * s);
+                    l.fill(o);
+                    const c = n.data.get(a.dataId).values;
+                    for (let e = 0; e < u; ++e) c[e] >= 0 && c[e] < s && (l[e * s + c[e]] = i);
+                    return n.makeTensorInfo([...a.shape, s], "int32", l)
+                }
+            };
+
+            function O_(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    x: r
+                } = t;
+                if ("string" === r.dtype) throw new Error("zerosLike is not supported for string tensors");
+                if ("complex64" === r.dtype) {
+                    const e = XE({
+                            inputs: {
+                                input: r
+                            },
+                            backend: n
+                        }),
+                        t = O_({
+                            inputs: {
+                                x: e
+                            },
+                            backend: n
+                        }),
+                        a = WT({
+                            inputs: {
+                                input: r
+                            },
+                            backend: n
+                        }),
+                        s = O_({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n
+                        }),
+                        i = GE({
+                            inputs: {
+                                real: t,
+                                imag: s
+                            },
+                            backend: n
+                        });
+                    return n.disposeIntermediateTensorInfo(e), n.disposeIntermediateTensorInfo(t), n.disposeIntermediateTensorInfo(a), n.disposeIntermediateTensorInfo(s), i
+                }
+                return OR({
+                    backend: n,
+                    attrs: {
+                        shape: r.shape,
+                        value: 0,
+                        dtype: r.dtype
+                    }
+                })
+            }
+            var L_ = {
+                kernelName: Jr,
+                backendName: "cpu",
+                kernelFunc: O_
+            };
+            var P_ = {
+                kernelName: Qn,
+                backendName: "cpu",
+                kernelFunc: function e(t) {
+                    const {
+                        inputs: n,
+                        backend: r
+                    } = t, {
+                        x: a
+                    } = n;
+                    if ("string" === a.dtype) throw new Error("onesLike is not supported for string tensors");
+                    if ("complex64" === a.dtype) {
+                        const t = XE({
+                                inputs: {
+                                    input: a
+                                },
+                                backend: r
+                            }),
+                            n = e({
+                                inputs: {
+                                    x: t
+                                },
+                                backend: r
+                            }),
+                            s = WT({
+                                inputs: {
+                                    input: a
+                                },
+                                backend: r
+                            }),
+                            i = O_({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: r
+                            }),
+                            o = GE({
+                                inputs: {
+                                    real: n,
+                                    imag: i
+                                },
+                                backend: r
+                            });
+                        return r.disposeIntermediateTensorInfo(t), r.disposeIntermediateTensorInfo(n), r.disposeIntermediateTensorInfo(s), r.disposeIntermediateTensorInfo(i), o
+                    }
+                    return OR({
+                        backend: r,
+                        attrs: {
+                            shape: a.shape,
+                            value: 1,
+                            dtype: a.dtype
+                        }
+                    })
+                }
+            };
+
+            function $_(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    axis: a
+                } = r;
+                if (1 === t.length) return NR({
+                    inputs: {
+                        input: t[0]
+                    },
+                    backend: n,
+                    attrs: {
+                        dim: a
+                    }
+                });
+                const s = t[0].shape,
+                    i = t[0].dtype;
+                t.forEach(e => {
+                    va.assertShapesMatch(s, e.shape, "All tensors passed to stack must have matching shapes"), va.assert(i === e.dtype, () => "All tensors passed to stack must have matching dtypes")
+                });
+                const o = [],
+                    u = jT({
+                        inputs: t.map(e => {
+                            const t = NR({
+                                inputs: {
+                                    input: e
+                                },
+                                backend: n,
+                                attrs: {
+                                    dim: a
+                                }
+                            });
+                            return o.push(t), t
+                        }),
+                        backend: n,
+                        attrs: {
+                            axis: a
+                        }
+                    });
+                return o.forEach(e => n.disposeIntermediateTensorInfo(e)), u
+            }
+            var B_ = {
+                kernelName: Jn,
+                backendName: "cpu",
+                kernelFunc: $_
+            };
+            var z_ = {
+                    kernelName: er,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            paddings: s,
+                            constantValue: i
+                        } = r;
+                        $E(a, "pad");
+                        const o = s.map((e, t) => e[0] + a.shape[t] + e[1]),
+                            u = s.map(e => e[0]),
+                            l = n.data.get(a.dataId).values,
+                            c = va.sizeFromShape(a.shape),
+                            h = a.shape.length,
+                            p = va.computeStrides(a.shape),
+                            d = va.sizeFromShape(o),
+                            f = o.length,
+                            m = va.computeStrides(o),
+                            g = va.getTypedArrayFromDType(a.dtype, d);
+                        0 !== i && g.fill(i);
+                        for (let e = 0; e < c; e++) {
+                            const t = va.indexToLoc(e, h, p).map((e, t) => e + u[t]);
+                            g[va.locToIndex(t, f, m)] = l[e]
+                        }
+                        return {
+                            dataId: n.write(g, o, a.dtype),
+                            shape: o,
+                            dtype: a.dtype
+                        }
+                    }
+                },
+                U_ = jE((e, t) => Math.pow(e, t)),
+                W_ = eN(nr, U_),
+                V_ = {
+                    kernelName: nr,
+                    backendName: "cpu",
+                    kernelFunc: W_
+                };
+            var j_ = {
+                    kernelName: sr,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            attrs: n
+                        } = e, {
+                            start: r,
+                            stop: a,
+                            dtype: s,
+                            step: i
+                        } = n, o = pC(r, a, i, s);
+                        return t.makeTensorInfo([o.length], s, o)
+                    }
+                },
+                G_ = lN(or, e => 1 / e),
+                H_ = {
+                    kernelName: or,
+                    backendName: "cpu",
+                    kernelFunc: G_
+                };
+            var q_ = {
+                kernelName: pr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a
+                    } = t, {
+                        alignCorners: s,
+                        halfPixelCenters: i,
+                        size: o
+                    } = r;
+                    $E(a, "resizeBilinear");
+                    const u = va.computeStrides(a.shape),
+                        [l, c] = o,
+                        [h, p, d, f] = a.shape,
+                        m = n.data.get(a.dataId).values,
+                        g = new Float32Array(va.sizeFromShape([h, l, c, f])),
+                        y = [s && l > 1 ? p - 1 : p, s && c > 1 ? d - 1 : d],
+                        b = [s && l > 1 ? l - 1 : l, s && c > 1 ? c - 1 : c];
+                    let v = 0;
+                    const x = y[0] / b[0],
+                        w = y[1] / b[1];
+                    for (let e = 0; e < h; e++)
+                        for (let t = 0; t < l; t++) {
+                            let n;
+                            n = i ? x * (t + .5) - .5 : x * t;
+                            const r = Math.max(0, Math.floor(n)),
+                                a = n - r,
+                                s = Math.min(p - 1, Math.ceil(n)),
+                                o = e * u[0] + r * u[1],
+                                l = e * u[0] + s * u[1];
+                            for (let e = 0; e < c; e++) {
+                                let t;
+                                t = i ? w * (e + .5) - .5 : w * e;
+                                const n = Math.max(0, Math.floor(t)),
+                                    r = t - n,
+                                    s = Math.min(d - 1, Math.ceil(t)),
+                                    c = o + n * u[2],
+                                    h = l + n * u[2],
+                                    p = o + s * u[2],
+                                    y = l + s * u[2];
+                                for (let e = 0; e < f; e++) {
+                                    const t = m[c + e],
+                                        n = m[h + e],
+                                        s = t + (m[p + e] - t) * r,
+                                        i = s + (n + (m[y + e] - n) * r - s) * a;
+                                    g[v++] = i
+                                }
+                            }
+                        }
+                    return n.makeTensorInfo([h, l, c, f], "float32", g)
+                }
+            };
+            var K_ = {
+                kernelName: dr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a,
+                        dy: s
+                    } = t, {
+                        alignCorners: i
+                    } = r;
+                    $E([s, a], "resizeBilinearGrad");
+                    const o = va.computeStrides(a.shape),
+                        [u, l, c, h] = a.shape,
+                        [, p, d] = s.shape,
+                        f = new Float32Array(u * l * c * h),
+                        m = [i && p > 1 ? l - 1 : l, i && d > 1 ? c - 1 : c],
+                        g = [i && p > 1 ? p - 1 : p, i && d > 1 ? d - 1 : d],
+                        y = m[0] / g[0],
+                        b = m[1] / g[1],
+                        v = n.data.get(s.dataId).values;
+                    let x = 0;
+                    for (let e = 0; e < u; e++) {
+                        const t = e * o[0];
+                        for (let e = 0; e < p; e++) {
+                            const n = e * y,
+                                r = Math.floor(n),
+                                a = Math.min(Math.ceil(n), l - 1),
+                                s = t + r * o[1],
+                                i = t + a * o[1],
+                                u = n - r,
+                                p = 1 - u;
+                            for (let e = 0; e < d; e++) {
+                                const t = e * b,
+                                    n = Math.floor(t),
+                                    r = Math.min(Math.ceil(t), c - 1),
+                                    a = t - n,
+                                    l = 1 - a,
+                                    d = s + n * o[2],
+                                    m = s + r * o[2],
+                                    g = i + n * o[2],
+                                    y = i + r * o[2],
+                                    w = p * l,
+                                    k = p * a,
+                                    S = u * l,
+                                    I = u * a;
+                                for (let e = 0; e < h; e++) {
+                                    const t = v[x++];
+                                    f[d + e] += t * w, f[m + e] += t * k, f[g + e] += t * S, f[y + e] += t * I
+                                }
+                            }
+                        }
+                    }
+                    return n.makeTensorInfo([u, c, l, h], "float32", f)
+                }
+            };
+            var Y_ = {
+                kernelName: cr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a
+                    } = t, {
+                        alignCorners: s,
+                        halfPixelCenters: i,
+                        size: o
+                    } = r;
+                    $E(a, "resizeNearestNeighbor");
+                    const u = va.computeStrides(a.shape),
+                        [l, c] = o,
+                        [h, p, d, f] = a.shape,
+                        m = n.data.get(a.dataId).values,
+                        g = new Float32Array(h * l * c * f),
+                        y = [s && l > 1 ? p - 1 : p, s && c > 1 ? d - 1 : d],
+                        b = [s && l > 1 ? l - 1 : l, s && c > 1 ? c - 1 : c],
+                        v = y[0] / b[0],
+                        x = y[1] / b[1];
+                    let w = 0;
+                    for (let e = 0; e < h; e++) {
+                        const t = e * u[0];
+                        for (let e = 0; e < l; e++) {
+                            const n = i ? v * (e + .5) : v * e;
+                            let r = Math.min(p - 1, s ? Math.round(n) : Math.floor(n));
+                            i && (r = Math.max(0, r));
+                            const a = t + r * u[1];
+                            for (let e = 0; e < c; e++) {
+                                const t = i ? x * (e + .5) : x * e;
+                                let n = Math.min(d - 1, s ? Math.round(t) : Math.floor(t));
+                                i && (n = Math.max(0, n));
+                                const r = a + n * u[2];
+                                for (let e = 0; e < f; e++) {
+                                    const t = m[r + e];
+                                    g[w++] = t
+                                }
+                            }
+                        }
+                    }
+                    return n.makeTensorInfo([h, l, c, f], a.dtype, g)
+                }
+            };
+            var X_ = {
+                kernelName: hr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a,
+                        dy: s
+                    } = t, {
+                        alignCorners: i
+                    } = r;
+                    $E([s, a], "resizeNearestNeighborGrad");
+                    const o = va.computeStrides(a.shape),
+                        u = va.computeStrides(s.shape),
+                        [l, c, h, p] = a.shape,
+                        [, d, f] = s.shape,
+                        m = new Float32Array(l * c * h * p),
+                        g = n.data.get(s.dataId).values,
+                        y = [i && d > 1 ? c - 1 : c, i && f > 1 ? h - 1 : h],
+                        b = [i && d > 1 ? d - 1 : d, i && f > 1 ? f - 1 : f],
+                        v = y[0] / b[0],
+                        x = y[1] / b[1],
+                        w = 1 / v,
+                        k = 1 / x,
+                        S = 2 * Math.ceil(w) + 2,
+                        I = 2 * Math.ceil(k) + 2;
+                    for (let e = 0; e < l; e++) {
+                        const t = e * o[0];
+                        for (let e = 0; e < c; e++) {
+                            const n = t + e * o[1],
+                                r = Math.floor(e * w),
+                                a = Math.floor(r - S / 2);
+                            for (let r = 0; r < h; r++) {
+                                const s = n + r * o[2],
+                                    l = Math.floor(r * k),
+                                    y = Math.floor(l - I / 2);
+                                for (let n = 0; n < p; n++) {
+                                    let o = 0;
+                                    for (let s = 0; s < S; s++) {
+                                        const l = s + a;
+                                        if (l < 0 || l >= d) continue;
+                                        const p = t + l * u[1],
+                                            m = l * v;
+                                        if (e === Math.min(c - 1, i ? Math.round(m) : Math.floor(m)))
+                                            for (let e = 0; e < I; e++) {
+                                                const t = e + y;
+                                                if (t < 0 || t >= f) continue;
+                                                const a = p + t * u[2],
+                                                    s = t * x;
+                                                r === Math.min(h - 1, i ? Math.round(s) : Math.floor(s)) && (o += g[a + n])
+                                            }
+                                    }
+                                    m[s + n] = o
+                                }
+                            }
+                        }
+                    }
+                    return n.makeTensorInfo(a.shape, a.dtype, m)
+                }
+            };
+            var Q_ = {
+                    kernelName: mr,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            dims: s
+                        } = r;
+                        $E(a, "reverse");
+                        const i = a.shape.length,
+                            o = va.parseAxisParam(s, a.shape);
+                        if (0 === i) return KE({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n
+                        });
+                        const u = new Ha(a.shape, a.dtype),
+                            l = n.bufferSync(a);
+                        for (let e = 0; e < u.size; e++) {
+                            const t = u.indexToLoc(e),
+                                n = t.slice();
+                            o.forEach(e => n[e] = a.shape[e] - 1 - n[e]), u.set(l.get(...n), ...t)
+                        }
+                        return n.makeTensorInfo(u.shape, u.dtype, u.values)
+                    }
+                },
+                Z_ = {
+                    kernelName: na,
+                    backendName: "cpu",
+                    kernelFunc: ({
+                        inputs: e,
+                        attrs: t,
+                        backend: n
+                    }) => {
+                        const {
+                            image: r
+                        } = e, {
+                            radians: a,
+                            fillValue: s,
+                            center: i
+                        } = t, o = n, u = va.getTypedArrayFromDType(r.dtype, va.sizeFromShape(r.shape)), [l, c, h, p] = r.shape, [d, f] = jd.getImageCenter(i, c, h), m = Math.sin(a), g = Math.cos(a), y = o.data.get(r.dataId).values;
+                        for (let e = 0; e < l; e++) {
+                            const t = e * h * c * p;
+                            for (let e = 0; e < c; e++) {
+                                const n = e * (h * p);
+                                for (let r = 0; r < h; r++) {
+                                    const a = r * p;
+                                    for (let i = 0; i < p; i++) {
+                                        const o = [l, e, r, i],
+                                            b = o[2],
+                                            v = o[1];
+                                        let x = (b - d) * g - (v - f) * m,
+                                            w = (b - d) * m + (v - f) * g;
+                                        x = Math.round(x + d), w = Math.round(w + f);
+                                        let k = s;
+                                        if ("number" != typeof s && (k = 3 === i ? 255 : s[i]), x >= 0 && x < h && w >= 0 && w < c) {
+                                            k = y[t + w * (h * p) + x * p + i]
+                                        }
+                                        u[t + n + a + i] = k
+                                    }
+                                }
+                            }
+                        }
+                        return {
+                            dataId: o.write(u, r.shape, r.dtype),
+                            shape: r.shape,
+                            dtype: r.dtype
+                        }
+                    }
+                },
+                J_ = lN(gr, e => {
+                    const t = Math.floor(e);
+                    return e - t < .5 ? Math.floor(e) : e - t > .5 ? Math.ceil(e) : t % 2 == 0 ? t : t + 1
+                }),
+                eF = {
+                    kernelName: gr,
+                    backendName: "cpu",
+                    kernelFunc: J_
+                };
+
+            function tF(e, t, n, r, a, s, i, o, u, l) {
+                const c = [r / a, a],
+                    h = e.values,
+                    p = t.values;
+                if (0 === r) return Ii(n, t.dtype);
+                const d = Ii(c, t.dtype);
+                d.values.fill(u);
+                for (let e = 0; e < s; e++) {
+                    const s = [];
+                    let u = 0;
+                    for (let t = 0; t < i; t++) {
+                        const n = h[e * i + t];
+                        s.push(n), u += n * o[t]
+                    }
+                    if (u < 0 || u >= r / a) throw new Error(`Invalid indices: ${s} does not index into ${n}`);
+                    for (let n = 0; n < a; n++) l ? d.values[u * a + n] += p[e * a + n] : d.values[u * a + n] = 0 === t.rank ? p[0] : p[e * a + n]
+                }
+                return d
+            }
+            var nF = {
+                kernelName: br,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        indices: a,
+                        updates: s
+                    } = t, {
+                        shape: i
+                    } = r, {
+                        sliceRank: o,
+                        numUpdates: u,
+                        sliceSize: l,
+                        strides: c,
+                        outputSize: h
+                    } = jd.calculateShapes(s, a, i), p = tF(n.bufferSync(a), n.bufferSync(s), i, h, l, u, o, c, 0, !0);
+                    return n.makeTensorInfo(i, p.dtype, p.values)
+                }
+            };
+            var rF = {
+                    kernelName: vr,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            condition: r,
+                            t: a,
+                            e: s
+                        } = t;
+                        $E([r, a, s], "select");
+                        const i = r.shape.length,
+                            o = n.data.get(r.dataId).values,
+                            u = n.data.get(a.dataId).values,
+                            l = n.data.get(s.dataId).values,
+                            c = cs(a.dtype, s.dtype),
+                            h = va.makeZerosTypedArray(va.sizeFromShape(a.shape), c);
+                        let p = 0;
+                        const d = 0 === i || i > 1 || 1 === a.shape.length ? 1 : va.sizeFromShape(a.shape.slice(1));
+                        for (let e = 0; e < o.length; e++)
+                            for (let t = 0; t < d; t++) 1 === o[e] ? h[p++] = u[e] : h[p++] = l[e];
+                        return n.makeTensorInfo(a.shape, c, h)
+                    }
+                },
+                aF = jd.SELU_SCALEALPHA,
+                sF = jd.SELU_SCALE,
+                iF = lN(xr, e => e >= 0 ? sF * e : aF * (Math.exp(e) - 1)),
+                oF = {
+                    kernelName: xr,
+                    backendName: "cpu",
+                    kernelFunc: iF
+                },
+                uF = lN(Ir, e => e < 0 ? -1 : e > 0 ? 1 : 0),
+                lF = {
+                    kernelName: Ir,
+                    backendName: "cpu",
+                    kernelFunc: uF
+                },
+                cF = lN(kr, e => Math.sin(e)),
+                hF = {
+                    kernelName: kr,
+                    backendName: "cpu",
+                    kernelFunc: cF
+                },
+                pF = lN(Sr, e => Math.sinh(e)),
+                dF = {
+                    kernelName: Sr,
+                    backendName: "cpu",
+                    kernelFunc: pF
+                },
+                fF = Math.log(1.1920928955078125e-7) + 2,
+                mF = lN(Er, e => {
+                    const t = e > -fF,
+                        n = e < fF,
+                        r = Math.exp(e);
+                    let a;
+                    return a = n ? r : t ? e : Math.log(1 + r), a
+                }),
+                gF = {
+                    kernelName: Er,
+                    backendName: "cpu",
+                    kernelFunc: mF
+                };
+            var yF = {
+                kernelName: Tr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        blockShape: s,
+                        paddings: i
+                    } = r;
+                    $E([a], "spaceToBatchND");
+                    const o = va.sizeFromShape(s),
+                        u = [
+                            [0, 0]
+                        ];
+                    u.push(...i);
+                    for (let e = 1 + s.length; e < a.shape.length; ++e) u.push([0, 0]);
+                    const l = z_.kernelFunc({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                paddings: u,
+                                constantValue: 0
+                            }
+                        }),
+                        c = jd.getReshaped(l.shape, s, o, !1),
+                        h = jd.getPermuted(c.length, s.length, !1),
+                        p = jd.getReshapedPermuted(l.shape, s, o, !1),
+                        d = rT({
+                            inputs: {
+                                x: l
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: c
+                            }
+                        }),
+                        f = uC({
+                            inputs: {
+                                x: d
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: h
+                            }
+                        }),
+                        m = rT({
+                            inputs: {
+                                x: f
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: p
+                            }
+                        });
+                    return n.disposeIntermediateTensorInfo(l), n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(f), m
+                }
+            };
+            var bF = {
+                kernelName: Fr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        indices: r,
+                        values: a,
+                        denseShape: s,
+                        defaultValue: i
+                    } = t;
+                    if (1 !== s.shape.length) throw new Error("Dense shape must be a vector, saw:\n        " + s.shape);
+                    if (2 !== r.shape.length) throw new Error("Indices must be a matrix, saw:\n        " + r.shape);
+                    if (1 !== a.shape.length) throw new Error("Values must be a vector, saw:\n        " + a.shape);
+                    if (0 !== i.shape.length) throw new Error("Default value must be a scalar, saw:\n        " + i.shape);
+                    const o = n.data.get(r.dataId).values,
+                        u = n.data.get(a.dataId).values,
+                        l = n.data.get(s.dataId).values,
+                        c = n.data.get(i.dataId).values[0],
+                        [h, p, d, f, m] = kC(o, r.shape, r.dtype, u, a.dtype, l, c);
+                    return [n.makeTensorInfo(p, r.dtype, h), n.makeTensorInfo([p[0]], a.dtype, d), n.makeTensorInfo([f.length], "bool", new Uint8Array(f.map(e => Number(e)))), n.makeTensorInfo([m.length], r.dtype, new Int32Array(m))]
+                }
+            };
+            var vF = {
+                kernelName: Mr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        inputIndices: r,
+                        inputShape: a,
+                        newShape: s
+                    } = t;
+                    if (2 !== r.shape.length) throw new Error("Input indices should be a matrix but received shape\n        " + r.shape);
+                    if (1 !== a.shape.length) throw new Error("Input shape should be a vector but received shape\n        " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Target shape should be a vector but received shape " + s.shape);
+                    const i = Array.from(n.data.get(a.dataId).values),
+                        o = n.data.get(r.dataId).values,
+                        u = Array.from(n.data.get(s.dataId).values),
+                        [l, c, h] = SC(o, r.shape, r.dtype, i, u);
+                    return [n.makeTensorInfo(c, r.dtype, l), n.makeTensorInfo([h.length], s.dtype, new Int32Array(h))]
+                }
+            };
+            var xF = {
+                kernelName: Dr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        data: r,
+                        indices: a,
+                        segmentIds: s
+                    } = t;
+                    if (r.shape.length < 1) throw new Error("Data should be at least 1 dimensional but received scalar");
+                    if (1 !== a.shape.length) throw new Error("Indices should be a vector but received shape\n          " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Segment ids should be a vector but received shape\n          " + s.shape);
+                    if (a.shape[0] !== s.shape[0]) throw new Error("segmentIds and indices should have same size.");
+                    const i = n.data.get(r.dataId).values,
+                        o = n.data.get(a.dataId).values,
+                        u = n.data.get(s.dataId).values,
+                        [l, c] = IC(i, r.shape, r.dtype, o, u, !0);
+                    return n.makeTensorInfo(c, r.dtype, l)
+                }
+            };
+            var wF = {
+                kernelName: Or,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        data: r,
+                        indices: a,
+                        segmentIds: s
+                    } = t;
+                    if (r.shape.length < 1) throw new Error("Data should be at least 1 dimensional but received scalar");
+                    if (1 !== a.shape.length) throw new Error("Indices should be a vector but received shape\n         " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Segment ids should be a vector but received shape\n         " + s.shape);
+                    if (a.shape[0] !== s.shape[0]) throw new Error("segmentIds and indices should have same size.");
+                    const i = n.data.get(r.dataId).values,
+                        o = n.data.get(a.dataId).values,
+                        u = n.data.get(s.dataId).values,
+                        [l, c] = IC(i, r.shape, r.dtype, o, u);
+                    return n.makeTensorInfo(c, r.dtype, l)
+                }
+            };
+            var kF = {
+                kernelName: Lr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        sparseIndices: a,
+                        sparseValues: s,
+                        defaultValue: i
+                    } = t, {
+                        outputShape: o
+                    } = r, {
+                        sliceRank: u,
+                        numUpdates: l,
+                        sliceSize: c,
+                        strides: h,
+                        outputSize: p
+                    } = jd.calculateShapes(s, a, o), d = tF(n.bufferSync(a), n.bufferSync(s), o, p, c, l, u, h, n.data.get(i.dataId).values[0], !1);
+                    return n.makeTensorInfo(o, d.dtype, d.values)
+                }
+            };
+            var SF = {
+                    kernelName: Rr,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            numOrSizeSplits: s,
+                            axis: i
+                        } = r, o = va.parseAxisParam(i, a.shape)[0], u = jd.prepareSplitSize(a, s, o), l = new Array(a.shape.length).fill(0), c = a.shape.slice();
+                        return u.map(e => {
+                            const t = [...c];
+                            t[o] = e;
+                            const r = xC({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    begin: l,
+                                    size: t
+                                }
+                            });
+                            return l[o] += e, r
+                        })
+                    }
+                },
+                IF = {
+                    kernelName: $r,
+                    backendName: "cpu",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t
+                    }) => {
+                        const {
+                            x: n
+                        } = e, r = t;
+                        $E(n, "square");
+                        const a = r.data.get(n.dataId).values,
+                            s = new Float32Array(a.length);
+                        for (let e = 0; e < a.length; ++e) {
+                            const t = a[e];
+                            s[e] = t * t
+                        }
+                        return {
+                            dataId: r.write(s, n.shape, n.dtype),
+                            shape: n.shape,
+                            dtype: n.dtype
+                        }
+                    }
+                },
+                AF = lN(ea, (e, t) => {
+                    const n = t;
+                    return isNaN(e) ? NaN : e > 0 ? 1 : n.alpha
+                }),
+                EF = {
+                    kernelName: ea,
+                    backendName: "cpu",
+                    kernelFunc: AF
+                };
+            var NF = {
+                kernelName: Br,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        begin: s,
+                        end: i,
+                        strides: o,
+                        beginMask: u,
+                        endMask: l,
+                        ellipsisMask: c,
+                        newAxisMask: h,
+                        shrinkAxisMask: p
+                    } = r;
+                    $E(a, "stridedSlice");
+                    const {
+                        finalShapeSparse: d,
+                        finalShape: f,
+                        isIdentity: m,
+                        sliceDim0: g,
+                        isSimpleSlice: y,
+                        begin: b,
+                        end: v,
+                        strides: x
+                    } = fo.sliceInfo(a.shape, s, i, o, u, l, c, h, p);
+                    let w;
+                    if (m) w = rT({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: f
+                        }
+                    });
+                    else if (g || y) {
+                        va.assert(a.shape.length >= 1, () => "Input must have rank at least 1, got: " + a.shape.length);
+                        const e = fo.computeOutShape(b, v, x),
+                            t = xC({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    begin: b,
+                                    size: e
+                                }
+                            });
+                        w = rT({
+                            inputs: {
+                                x: t
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: f
+                            }
+                        }), n.disposeIntermediateTensorInfo(t)
+                    } else {
+                        const e = _C(d, n.bufferSync(a), x, b);
+                        w = n.makeTensorInfo(f, e.dtype, e.values)
+                    }
+                    return w
+                }
+            };
+            var CF = {
+                kernelName: zr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        separator: a,
+                        nGramWidths: s,
+                        leftPad: i,
+                        rightPad: o,
+                        padWidth: u,
+                        preserveShortSequences: l
+                    } = r, {
+                        data: c,
+                        dataSplits: h
+                    } = t, p = n.data.get(c.dataId).values, d = n.data.get(h.dataId).values, [f, m] = FC(p, d, a, s, i, o, u, l);
+                    return [n.makeTensorInfo([f.length], "string", f), n.makeTensorInfo(h.shape, "int32", m)]
+                }
+            };
+            var TF = {
+                kernelName: Ur,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        skipEmpty: a
+                    } = r, {
+                        input: s,
+                        delimiter: i
+                    } = t;
+                    if ("string" !== s.dtype) throw new Error("Input must be of datatype string");
+                    if (1 !== s.shape.length) throw new Error("Input must be a vector, got shape: " + s.shape);
+                    if (0 !== i.shape.length) throw new Error("Delimiter must be a scalar, got shape: " + i.shape);
+                    const o = n.data.get(s.dataId).values,
+                        u = n.data.get(i.dataId).values[0],
+                        [l, c, h] = DC(o, u, a),
+                        p = c.length;
+                    return [n.makeTensorInfo([p, 2], "int32", l), n.makeTensorInfo([p], "string", c), n.makeTensorInfo([2], "int32", new Int32Array(h))]
+                }
+            };
+            var RF = {
+                    kernelName: Wr,
+                    backendName: "cpu",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            numBuckets: a
+                        } = r, {
+                            input: s
+                        } = t;
+                        if ("string" !== s.dtype) throw new Error("Input must be of datatype string");
+                        if (a <= 0) throw new Error("Number of buckets must be at least 1");
+                        const i = OC(n.data.get(s.dataId).values, a);
+                        return n.makeTensorInfo(s.shape, "int32", i)
+                    }
+                },
+                _F = lN(jr, e => Math.tan(e)),
+                FF = {
+                    kernelName: jr,
+                    backendName: "cpu",
+                    kernelFunc: _F
+                },
+                MF = lN(Gr, e => Math.tanh(e));
+
+            function DF(e, t, n) {
+                switch (n) {
+                    case "reflect":
+                        return function (e, t) {
+                            let n = e;
+                            if (n < 0)
+                                if (t <= 1) n = 0;
+                                else {
+                                    const e = 2 * t;
+                                    n < e && (n = e * Math.trunc(-n / e) + n), n = n < -t ? n + e : -n - 1
+                                }
+                            else if (n > t - 1)
+                                if (t <= 1) n = 0;
+                                else {
+                                    const e = 2 * t;
+                                    n -= e * Math.trunc(n / e), n >= t && (n = e - n - 1)
+                                } return va.clamp(0, n, t - 1)
+                        }(e, t);
+                    case "wrap":
+                        return function (e, t) {
+                            let n = e;
+                            if (n < 0)
+                                if (t <= 1) n = 0;
+                                else {
+                                    const e = t - 1;
+                                    n += t * (Math.trunc(-n / e) + 1)
+                                }
+                            else if (n > t - 1)
+                                if (t <= 1) n = 0;
+                                else {
+                                    const e = t - 1;
+                                    n -= t * Math.trunc(n / e)
+                                } return va.clamp(0, n, t - 1)
+                        }(e, t);
+                    case "nearest":
+                        return function (e, t) {
+                            return va.clamp(0, e, t - 1)
+                        }(e, t);
+                    case "constant":
+                    default:
+                        return function (e, t) {
+                            return e
+                        }(e)
+                }
+            }
+
+            function OF(e, t, n, r, a, s, i, o, u, l, c) {
+                return 0 <= o && o < t && 0 <= u && u < n ? e[i * r + o * a + u * s + l] : c
+            }
+
+            function LF(e, t, n, r, a, s, i, o, u, l, c) {
+                return OF(e, t, n, r, a, s, i, Math.round(o), Math.round(u), l, c)
+            }
+
+            function PF(e, t, n, r, a, s, i, o, u, l, c) {
+                const h = Math.floor(o),
+                    p = Math.floor(u),
+                    d = h + 1,
+                    f = p + 1;
+                return (d - o) * ((f - u) * OF(e, t, n, r, a, s, i, h, p, l, c) + (u - p) * OF(e, t, n, r, a, s, i, h, f, l, c)) + (o - h) * ((f - u) * OF(e, t, n, r, a, s, i, d, p, l, c) + (u - p) * OF(e, t, n, r, a, s, i, d, f, l, c))
+            }
+            var $F = [oT, VE, lT, hT, sN, pT, dT, fT, mT, gT, bT, xT, kT, AT, NT, _T, FT, MT, DT, iT, OT, LT, PT, $T, JE, dN, zT, HE, UT, GT, qT, KT, YT, XT, QT, ZT, eR, nR, rR, aR, sR, iR, uR, lR, cR, hR, pR, dR, fR, yR, HC, bR, yN, ER, xN, CR, SN, DR, LR, PR, EN, zR, UR, WR, VR, jR, _N, DN, YE, GR, VT, qR, YR, QR, KC, PN, zN, ZR, jN, e_, r_, s_, u_, l_, c_, p_, KN, d_, f_, m_, g_, y_, b_, v_, QN, x_, S_, N_, tC, rC, T_, __, M_, iC, D_, P_, B_, z_, V_, QC, hC, j_, QE, _R, H_, JC, tT, aT, q_, K_, Y_, X_, Q_, Z_, eF, mC, nF, rF, oF, bC, lF, hF, dF, wC, E_, gF, yF, bF, vF, xF, wF, kF, SF, NC, IF, RC, EF, NF, CF, TF, RF, BC, gR, FF, {
+                kernelName: Gr,
+                backendName: "cpu",
+                kernelFunc: MF
+            }, {
+                kernelName: Hr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        reps: s
+                    } = r;
+                    $E(a, "tile");
+                    const i = zC(n.bufferSync(a), s);
+                    return n.makeTensorInfo(i.shape, i.dtype, i.values)
+                }
+            }, {
+                kernelName: qr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        k: s,
+                        sorted: i
+                    } = r;
+                    $E(a, "topk");
+                    const o = n.data.get(a.dataId).values,
+                        [u, l] = VC(o, a.shape, a.dtype, s, i);
+                    return [n.makeTensorInfo(u.shape, u.dtype, u.values), n.makeTensorInfo(l.shape, l.dtype, l.values)]
+                }
+            }, {
+                kernelName: Kr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, {
+                        image: a,
+                        transforms: s
+                    } = t, {
+                        interpolation: i,
+                        fillMode: o,
+                        fillValue: u,
+                        outputShape: l
+                    } = n, [c, h, p, d] = a.shape, [f, m] = null != l ? l : [h, p], g = [c, f, m, d], y = va.computeStrides(a.shape), b = y[0], v = y[1], x = y[2], w = va.getTypedArrayFromDType(a.dtype, va.sizeFromShape(g));
+                    w.fill(u);
+                    const k = r.data.get(a.dataId).values,
+                        S = r.data.get(s.dataId).values;
+                    for (let e = 0; e < c; ++e) {
+                        const t = 1 === s.shape[0] ? S : S.subarray(8 * e, 8 * e + 8);
+                        for (let n = 0; n < f; ++n)
+                            for (let r = 0; r < m; ++r)
+                                for (let a = 0; a < d; ++a) {
+                                    let s;
+                                    const l = t[6] * r + t[7] * n + 1;
+                                    if (0 === l) continue;
+                                    const c = (t[0] * r + t[1] * n + t[2]) / l,
+                                        d = (t[3] * r + t[4] * n + t[5]) / l,
+                                        f = DF(c, p, o),
+                                        m = DF(d, h, o);
+                                    switch (i) {
+                                        case "nearest":
+                                            s = LF(k, h, p, b, v, x, e, m, f, a, u);
+                                            break;
+                                        case "bilinear":
+                                            s = PF(k, h, p, b, v, x, e, m, f, a, u);
+                                            break;
+                                        default:
+                                            throw new Error("Error in Transform: Expect 'nearest' or 'bilinear', but got " + i)
+                                    }
+                                    w[e * b + n * v + r * x + a] = s
+                                }
+                        return r.makeTensorInfo(g, a.dtype, w)
+                    }
+                    return {
+                        dataId: r.write(w, g, a.dtype),
+                        shape: a.shape,
+                        dtype: a.dtype
+                    }
+                }
+            }, lC, {
+                kernelName: Xr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, {
+                        axis: a
+                    } = n, {
+                        x: s
+                    } = t;
+                    $E(s, "unique");
+                    const i = r.data.get(s.dataId).values,
+                        {
+                            outputValues: o,
+                            outputShape: u,
+                            indices: l
+                        } = jC(i, a, s.shape, s.dtype);
+                    return [r.makeTensorInfo(u, s.dtype, o), r.makeTensorInfo([l.length], "int32", l)]
+                }
+            }, {
+                kernelName: Qr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        value: a
+                    } = t;
+                    let {
+                        axis: s
+                    } = r;
+                    s < 0 && (s += a.shape.length);
+                    const i = a.shape.length,
+                        o = a.shape[s],
+                        u = new Array(i - 1);
+                    let l = 0;
+                    for (let e = 0; e < i; e++) e !== s && (u[l++] = a.shape[e]);
+                    const c = new Array(i).fill(0),
+                        h = a.shape.slice();
+                    h[s] = 1;
+                    const p = new Array(o);
+                    for (let e = 0; e < p.length; e++) {
+                        c[s] = e;
+                        const t = xC({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                begin: c,
+                                size: h
+                            }
+                        });
+                        p[e] = rT({
+                            inputs: {
+                                x: t
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: u
+                            }
+                        }), n.disposeIntermediateTensorInfo(t)
+                    }
+                    return p
+                }
+            }, {
+                kernelName: Zr,
+                backendName: "cpu",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        segmentIds: s
+                    } = t, {
+                        numSegments: i
+                    } = r;
+                    $E(a, "unsortedSegmentSum");
+                    const o = [],
+                        u = [],
+                        l = a.shape.length - s.shape.length;
+                    let c = s;
+                    for (let e = 0; e < l; ++e) {
+                        const t = NR({
+                            inputs: {
+                                input: c
+                            },
+                            backend: n,
+                            attrs: {
+                                dim: e + 1
+                            }
+                        });
+                        c = t, u.push(t)
+                    }
+                    for (let e = 0; e < i; ++e) {
+                        const t = va.createScalarValue(e, "int32"),
+                            r = n.makeTensorInfo([], "int32", t),
+                            s = gN({
+                                inputs: {
+                                    a: r,
+                                    b: c
+                                },
+                                backend: n
+                            }),
+                            i = ZE({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: n,
+                                attrs: {
+                                    dtype: "float32"
+                                }
+                            }),
+                            l = eC({
+                                inputs: {
+                                    a: i,
+                                    b: a
+                                },
+                                backend: n
+                            }),
+                            h = mR({
+                                inputs: {
+                                    x: l
+                                },
+                                backend: n,
+                                attrs: {
+                                    axis: 0,
+                                    keepDims: !1
+                                }
+                            });
+                        o.push(h), u.push(r), u.push(s), u.push(i), u.push(l), u.push(h)
+                    }
+                    const h = $_({
+                        inputs: o,
+                        backend: n,
+                        attrs: {
+                            axis: 0
+                        }
+                    });
+                    return u.forEach(e => n.disposeIntermediateTensorInfo(e)), h
+                }
+            }, L_];
+            for (const e of $F) da(e);
+            var BF = {};
+            w(BF, {
+                assertNotComplex: () => KM,
+                bindCanvasToFramebuffer: () => IM,
+                bindColorTextureToFramebuffer: () => AM,
+                bindTextureToProgramUniformSampler: () => SM,
+                bindTextureUnit: () => vM,
+                bindVertexBufferToProgramAttribute: () => bM,
+                callAndCheck: () => eM,
+                canBeRepresented: () => tM,
+                createFragmentShader: () => sM,
+                createFramebuffer: () => yM,
+                createProgram: () => lM,
+                createStaticIndexBuffer: () => dM,
+                createStaticVertexBuffer: () => pM,
+                createTexture: () => mM,
+                createVertexShader: () => aM,
+                getBatchDim: () => _M,
+                getExtensionOrThrow: () => rM,
+                getFramebufferErrorMessage: () => CM,
+                getMaxTexturesInShader: () => zM,
+                getNumChannels: () => fM,
+                getProgramUniformLocation: () => kM,
+                getProgramUniformLocationOrThrow: () => wM,
+                getRowsCols: () => FM,
+                getShapeAs3D: () => MM,
+                getTextureShapeFromLogicalShape: () => DM,
+                getWebGLDisjointQueryTimerVersion: () => UM,
+                getWebGLErrorMessage: () => nM,
+                getWebGLMaxTextureSize: () => PM,
+                hasExtension: () => WM,
+                isCapableOfRenderingToFloatTexture: () => jM,
+                isDownloadFloatTextureEnabled: () => GM,
+                isReshapeFree: () => LM,
+                isWebGLFenceEnabled: () => qM,
+                isWebGLVersionEnabled: () => VM,
+                linkProgram: () => cM,
+                resetMaxTextureSize: () => $M,
+                resetMaxTexturesInShader: () => BM,
+                unbindColorTextureFromFramebuffer: () => EM,
+                unbindTextureUnit: () => xM,
+                validateFramebuffer: () => NM,
+                validateProgram: () => hM,
+                validateTextureSize: () => gM
+            });
+            var zF, UF, WF, VF, jF, GF, HF = {},
+                qF = {
+                    alpha: !1,
+                    antialias: !1,
+                    premultipliedAlpha: !1,
+                    preserveDrawingBuffer: !1,
+                    depth: !1,
+                    stencil: !1,
+                    failIfMajorPerformanceCaveat: !0
+                };
+
+            function KF(e, t) {
+                HF[e] = t
+            }
+
+            function YF(e, t) {
+                if (!(e in HF)) {
+                    const n = function (e, t) {
+                        if (1 !== e && 2 !== e) throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");
+                        const n = null == t ? function (e) {
+                            if ("undefined" != typeof OffscreenCanvas && 2 === e) return new OffscreenCanvas(300, 150);
+                            if ("undefined" != typeof document) return document.createElement("canvas");
+                            throw new Error("Cannot create a canvas in this context")
+                        }(e) : t;
+                        if (n.addEventListener("webglcontextlost", t => {
+                                t.preventDefault(), delete HF[e]
+                            }, !1), 1 === e) return n.getContext("webgl", qF) || n.getContext("experimental-webgl", qF);
+                        return n.getContext("webgl2", qF)
+                    }(e, t);
+                    if (null === n) return console.log("Could not get context for WebGL version", e), null;
+                    HF[e] = n
+                }
+                const n = HF[e];
+                return null == n || n.isContextLost() ? (delete HF[e], YF(e)) : (n.disable(n.DEPTH_TEST), n.disable(n.STENCIL_TEST), n.disable(n.BLEND), n.disable(n.DITHER), n.disable(n.POLYGON_OFFSET_FILL), n.disable(n.SAMPLE_COVERAGE), n.enable(n.SCISSOR_TEST), n.enable(n.CULL_FACE), n.cullFace(n.BACK), HF[e])
+            }
+
+            function XF(e, t) {
+                return [t, e]
+            }
+
+            function QF(e) {
+                const t = va.sizeFromShape(e),
+                    n = Math.ceil(t / 4);
+                return va.sizeToSquarishShape(n)
+            }
+
+            function ZF(e, t) {
+                return [Math.max(1, Math.ceil(t / 2)), Math.max(1, Math.ceil(e / 2))]
+            }
+
+            function JF(e, t) {
+                const n = e;
+                let r, a, s, i, o, u, l, c, h, p;
+                return 2 === Ye().getNumber("WEBGL_VERSION") ? (r = n.R32F, a = n.R16F, s = n.RGBA16F, i = n.RGBA32F, o = n.RED, l = 4, c = 1, h = n.HALF_FLOAT, p = n.FLOAT, u = n.RGBA8) : (r = e.RGBA, a = e.RGBA, s = e.RGBA, i = n.RGBA, o = e.RGBA, l = 4, c = 4, h = null != t ? t.HALF_FLOAT_OES : null, p = e.FLOAT, u = e.RGBA), {
+                    internalFormatFloat: r,
+                    internalFormatHalfFloat: a,
+                    internalFormatPackedHalfFloat: s,
+                    internalFormatPackedFloat: i,
+                    textureFormatFloat: o,
+                    downloadTextureFormat: u,
+                    downloadUnpackNumChannels: l,
+                    defaultNumChannels: c,
+                    textureTypeHalfFloat: h,
+                    textureTypeFloat: p
+                }
+            }
+
+            function eM(e, t) {
+                const n = t();
+                return Ye().getBool("DEBUG") && function (e) {
+                    const t = e.getError();
+                    if (t !== e.NO_ERROR) throw new Error("WebGL Error: " + nM(e, t))
+                }(e), n
+            }(UF = zF || (zF = {}))[UF.DENSE = 0] = "DENSE", UF[UF.SHARED_BATCH = 1] = "SHARED_BATCH", (VF = WF || (WF = {}))[VF.RENDER = 0] = "RENDER", VF[VF.UPLOAD = 1] = "UPLOAD", VF[VF.PIXELS = 2] = "PIXELS", VF[VF.DOWNLOAD = 3] = "DOWNLOAD", (GF = jF || (jF = {}))[GF.UNPACKED_FLOAT16 = 0] = "UNPACKED_FLOAT16", GF[GF.UNPACKED_FLOAT32 = 1] = "UNPACKED_FLOAT32", GF[GF.PACKED_4X1_UNSIGNED_BYTE = 2] = "PACKED_4X1_UNSIGNED_BYTE", GF[GF.PACKED_2X2_FLOAT32 = 3] = "PACKED_2X2_FLOAT32", GF[GF.PACKED_2X2_FLOAT16 = 4] = "PACKED_2X2_FLOAT16";
+
+            function tM(e) {
+                return !!(Ye().getBool("WEBGL_RENDER_FLOAT32_ENABLED") || 0 === e || 5.96e-8 < Math.abs(e) && Math.abs(e) < 65504)
+            }
+
+            function nM(e, t) {
+                switch (t) {
+                    case e.NO_ERROR:
+                        return "NO_ERROR";
+                    case e.INVALID_ENUM:
+                        return "INVALID_ENUM";
+                    case e.INVALID_VALUE:
+                        return "INVALID_VALUE";
+                    case e.INVALID_OPERATION:
+                        return "INVALID_OPERATION";
+                    case e.INVALID_FRAMEBUFFER_OPERATION:
+                        return "INVALID_FRAMEBUFFER_OPERATION";
+                    case e.OUT_OF_MEMORY:
+                        return "OUT_OF_MEMORY";
+                    case e.CONTEXT_LOST_WEBGL:
+                        return "CONTEXT_LOST_WEBGL";
+                    default:
+                        return "Unknown error code " + t
+                }
+            }
+
+            function rM(e, t) {
+                return TM(e, () => e.getExtension(t), 'Extension "' + t + '" not supported on this browser.')
+            }
+
+            function aM(e, t) {
+                const n = TM(e, () => e.createShader(e.VERTEX_SHADER), "Unable to create vertex WebGLShader.");
+                if (eM(e, () => e.shaderSource(n, t)), eM(e, () => e.compileShader(n)), !1 === e.getShaderParameter(n, e.COMPILE_STATUS)) throw console.log(e.getShaderInfoLog(n)), new Error("Failed to compile vertex shader.");
+                return n
+            }
+
+            function sM(e, t) {
+                const n = TM(e, () => e.createShader(e.FRAGMENT_SHADER), "Unable to create fragment WebGLShader.");
+                if (eM(e, () => e.shaderSource(n, t)), eM(e, () => e.compileShader(n)), !1 === e.getShaderParameter(n, e.COMPILE_STATUS)) throw function (e, t) {
+                    const n = uM.exec(t);
+                    if (null == n) return console.log("Couldn't parse line number in error: " + t), void console.log(e);
+                    const r = +n[1],
+                        a = e.split("\n"),
+                        s = a.length.toString().length + 2,
+                        i = a.map((e, t) => va.rightPad((t + 1).toString(), s) + e);
+                    let o = 0;
+                    for (let e = 0; e < i.length; e++) o = Math.max(i[e].length, o);
+                    const u = i.slice(0, r - 1),
+                        l = i.slice(r - 1, r),
+                        c = i.slice(r);
+                    console.log(u.join("\n")), console.log(t.split("\n")[0]), console.log("%c " + va.rightPad(l[0], o), "border:1px solid red; background-color:#e3d2d2; color:#a61717"), console.log(c.join("\n"))
+                }(t, e.getShaderInfoLog(n)), new Error("Failed to compile fragment shader.");
+                return n
+            }
+            var iM, oM, uM = /ERROR: [0-9]+:([0-9]+):/g;
+
+            function lM(e) {
+                return TM(e, () => e.createProgram(), "Unable to create WebGLProgram.")
+            }
+
+            function cM(e, t) {
+                if (eM(e, () => e.linkProgram(t)), !1 === e.getProgramParameter(t, e.LINK_STATUS)) throw console.log(e.getProgramInfoLog(t)), new Error("Failed to link vertex and fragment shaders.")
+            }
+
+            function hM(e, t) {
+                if (eM(e, () => e.validateProgram(t)), !1 === e.getProgramParameter(t, e.VALIDATE_STATUS)) throw console.log(e.getProgramInfoLog(t)), new Error("Shader program validation failed.")
+            }
+
+            function pM(e, t) {
+                const n = TM(e, () => e.createBuffer(), "Unable to create WebGLBuffer");
+                return eM(e, () => e.bindBuffer(e.ARRAY_BUFFER, n)), eM(e, () => e.bufferData(e.ARRAY_BUFFER, t, e.STATIC_DRAW)), n
+            }
+
+            function dM(e, t) {
+                const n = TM(e, () => e.createBuffer(), "Unable to create WebGLBuffer");
+                return eM(e, () => e.bindBuffer(e.ELEMENT_ARRAY_BUFFER, n)), eM(e, () => e.bufferData(e.ELEMENT_ARRAY_BUFFER, t, e.STATIC_DRAW)), n
+            }
+
+            function fM() {
+                return 2 === Ye().getNumber("WEBGL_VERSION") ? 1 : 4
+            }
+
+            function mM(e) {
+                return TM(e, () => e.createTexture(), "Unable to create WebGLTexture.")
+            }
+
+            function gM(e, t) {
+                const n = Ye().getNumber("WEBGL_MAX_TEXTURE_SIZE");
+                if (e <= 0 || t <= 0) {
+                    throw new Error("Requested texture size " + `[${e}x${t}]` + " is invalid.")
+                }
+                if (e > n || t > n) {
+                    throw new Error("Requested texture size " + `[${e}x${t}]` + " greater than WebGL maximum on this browser / GPU " + `[${n}x${n}]` + ".")
+                }
+            }
+
+            function yM(e) {
+                return TM(e, () => e.createFramebuffer(), "Unable to create WebGLFramebuffer.")
+            }
+
+            function bM(e, t, n, r, a, s, i) {
+                const o = e.getAttribLocation(t, n);
+                return -1 !== o && (eM(e, () => e.bindBuffer(e.ARRAY_BUFFER, r)), eM(e, () => e.vertexAttribPointer(o, a, e.FLOAT, !1, s, i)), eM(e, () => e.enableVertexAttribArray(o)), !0)
+            }
+
+            function vM(e, t, n) {
+                RM(e, n), eM(e, () => e.activeTexture(e.TEXTURE0 + n)), eM(e, () => e.bindTexture(e.TEXTURE_2D, t))
+            }
+
+            function xM(e, t) {
+                RM(e, t), eM(e, () => e.activeTexture(e.TEXTURE0 + t)), eM(e, () => e.bindTexture(e.TEXTURE_2D, null))
+            }
+
+            function wM(e, t, n) {
+                return TM(e, () => e.getUniformLocation(t, n), 'uniform "' + n + '" not present in program.')
+            }
+
+            function kM(e, t, n) {
+                return e.getUniformLocation(t, n)
+            }
+
+            function SM(e, t, n, r) {
+                eM(e, () => vM(e, t, r)), eM(e, () => e.uniform1i(n, r))
+            }
+
+            function IM(e) {
+                eM(e, () => e.bindFramebuffer(e.FRAMEBUFFER, null)), eM(e, () => e.viewport(0, 0, e.canvas.width, e.canvas.height)), eM(e, () => e.scissor(0, 0, e.canvas.width, e.canvas.height))
+            }
+
+            function AM(e, t, n) {
+                eM(e, () => e.bindFramebuffer(e.FRAMEBUFFER, n)), eM(e, () => e.framebufferTexture2D(e.FRAMEBUFFER, e.COLOR_ATTACHMENT0, e.TEXTURE_2D, t, 0))
+            }
+
+            function EM(e, t) {
+                eM(e, () => e.bindFramebuffer(e.FRAMEBUFFER, t)), eM(e, () => e.framebufferTexture2D(e.FRAMEBUFFER, e.COLOR_ATTACHMENT0, e.TEXTURE_2D, null, 0))
+            }
+
+            function NM(e) {
+                const t = e.checkFramebufferStatus(e.FRAMEBUFFER);
+                if (t !== e.FRAMEBUFFER_COMPLETE) throw new Error("Error binding framebuffer: " + CM(e, t))
+            }
+
+            function CM(e, t) {
+                switch (t) {
+                    case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:
+                        return "FRAMEBUFFER_INCOMPLETE_ATTACHMENT";
+                    case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:
+                        return "FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";
+                    case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:
+                        return "FRAMEBUFFER_INCOMPLETE_DIMENSIONS";
+                    case e.FRAMEBUFFER_UNSUPPORTED:
+                        return "FRAMEBUFFER_UNSUPPORTED";
+                    default:
+                        return "unknown error " + t
+                }
+            }
+
+            function TM(e, t, n) {
+                const r = eM(e, () => t());
+                if (null == r) throw new Error(n);
+                return r
+            }
+
+            function RM(e, t) {
+                const n = e.MAX_COMBINED_TEXTURE_IMAGE_UNITS - 1,
+                    r = t + e.TEXTURE0;
+                if (r < e.TEXTURE0 || r > n) {
+                    throw new Error(`textureUnit must be in ${`[gl.TEXTURE0, gl.TEXTURE${n}]`}.`)
+                }
+            }
+
+            function _M(e, t = 2) {
+                return va.sizeFromShape(e.slice(0, e.length - t))
+            }
+
+            function FM(e) {
+                if (0 === e.length) throw Error("Cannot get rows and columns of an empty shape array.");
+                return [e.length > 1 ? e[e.length - 2] : 1, e[e.length - 1]]
+            }
+
+            function MM(e) {
+                let t = [1, 1, 1];
+                return 0 === e.length || 1 === e.length && 1 === e[0] || (t = [_M(e), ...FM(e)]), t
+            }
+
+            function DM(e, t = !1) {
+                let n = Ye().getNumber("WEBGL_MAX_TEXTURE_SIZE");
+                if (t && (n *= 2, 1 === (e = e.map((t, n) => n >= e.length - 2 ? va.nearestLargerEven(e[n]) : e[n])).length && (e = [2, e[0]])), 2 !== e.length) {
+                    const t = va.squeezeShape(e);
+                    e = t.newShape
+                }
+                let r = va.sizeFromShape(e);
+                if (e.length <= 1 && r <= n) return [1, r];
+                if (2 === e.length && e[0] <= n && e[1] <= n) return e;
+                if (3 === e.length && e[0] * e[1] <= n && e[2] <= n) return [e[0] * e[1], e[2]];
+                if (3 === e.length && e[0] <= n && e[1] * e[2] <= n) return [e[0], e[1] * e[2]];
+                if (4 === e.length && e[0] * e[1] * e[2] <= n && e[3] <= n) return [e[0] * e[1] * e[2], e[3]];
+                if (4 === e.length && e[0] <= n && e[1] * e[2] * e[3] <= n) return [e[0], e[1] * e[2] * e[3]];
+                if (t) {
+                    const t = _M(e);
+                    let n = 2,
+                        a = 2;
+                    return e.length && ([n, a] = FM(e)), r = t * (n / 2) * (a / 2), va.sizeToSquarishShape(r).map(e => 2 * e)
+                }
+                return va.sizeToSquarishShape(r)
+            }
+
+            function OM(e) {
+                return e % 2 == 0
+            }
+
+            function LM(e, t) {
+                if (e = e.slice(-2), t = t.slice(-2), va.arraysEqual(e, t)) return !0;
+                if (!e.length || !t.length) return !0;
+                if (0 === e[0] || 0 === e[1] || 0 === t[0] || 0 === t[1]) return !0;
+                if (e.length !== t.length) {
+                    const n = e.slice(-1)[0],
+                        r = t.slice(-1)[0];
+                    if (n === r) return !0;
+                    if (OM(n) && OM(r) && (1 === e[0] || 1 === t[0])) return !0
+                }
+                return e[1] === t[1] && OM(e[0]) && OM(t[0])
+            }
+
+            function PM(e) {
+                if (null == iM) {
+                    const t = YF(e);
+                    iM = t.getParameter(t.MAX_TEXTURE_SIZE)
+                }
+                return iM
+            }
+
+            function $M() {
+                iM = null
+            }
+
+            function BM() {
+                oM = null
+            }
+
+            function zM(e) {
+                if (null == oM) {
+                    const t = YF(e);
+                    oM = t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)
+                }
+                return Math.min(16, oM)
+            }
+
+            function UM(e) {
+                if (0 === e) return 0;
+                let t;
+                const n = YF(e);
+                return t = WM(n, "EXT_disjoint_timer_query_webgl2") && 2 === e ? 2 : WM(n, "EXT_disjoint_timer_query") ? 1 : 0, t
+            }
+
+            function WM(e, t) {
+                return null != e.getExtension(t)
+            }
+
+            function VM(e) {
+                try {
+                    if (null != YF(e)) return !0
+                } catch (e) {
+                    return console.log("Error when getting WebGL context: ", e), !1
+                }
+                return !1
+            }
+
+            function jM(e) {
+                if (0 === e) return !1;
+                const t = YF(e);
+                if (1 === e) {
+                    if (!WM(t, "OES_texture_float")) return !1
+                } else if (!WM(t, "EXT_color_buffer_float")) return !1;
+                return HM(t)
+            }
+
+            function GM(e) {
+                if (0 === e) return !1;
+                const t = YF(e);
+                if (1 !== e) {
+                    if (WM(t, "EXT_color_buffer_float")) return HM(t);
+                    const e = "EXT_color_buffer_half_float";
+                    if (WM(t, e)) {
+                        const n = t.getExtension(e);
+                        return function (e, t) {
+                            const n = JF(e, t),
+                                r = e.createTexture();
+                            e.bindTexture(e.TEXTURE_2D, r);
+                            e.texImage2D(e.TEXTURE_2D, 0, n.internalFormatHalfFloat, 1, 1, 0, n.textureFormatFloat, n.textureTypeHalfFloat, null);
+                            const a = e.createFramebuffer();
+                            e.bindFramebuffer(e.FRAMEBUFFER, a), e.framebufferTexture2D(e.FRAMEBUFFER, e.COLOR_ATTACHMENT0, e.TEXTURE_2D, r, 0);
+                            const s = e.checkFramebufferStatus(e.FRAMEBUFFER) === e.FRAMEBUFFER_COMPLETE;
+                            return e.bindTexture(e.TEXTURE_2D, null), e.bindFramebuffer(e.FRAMEBUFFER, null), e.deleteTexture(r), e.deleteFramebuffer(a), s
+                        }(t, n)
+                    }
+                    return !1
+                }
+                if (!WM(t, "OES_texture_float")) return !1;
+                if (!WM(t, "WEBGL_color_buffer_float")) return !1;
+                return HM(t)
+            }
+
+            function HM(e) {
+                const t = JF(e),
+                    n = e.createTexture();
+                e.bindTexture(e.TEXTURE_2D, n);
+                e.texImage2D(e.TEXTURE_2D, 0, t.internalFormatFloat, 1, 1, 0, t.textureFormatFloat, t.textureTypeFloat, null);
+                const r = e.createFramebuffer();
+                e.bindFramebuffer(e.FRAMEBUFFER, r), e.framebufferTexture2D(e.FRAMEBUFFER, e.COLOR_ATTACHMENT0, e.TEXTURE_2D, n, 0);
+                const a = e.checkFramebufferStatus(e.FRAMEBUFFER) === e.FRAMEBUFFER_COMPLETE;
+                return e.bindTexture(e.TEXTURE_2D, null), e.bindFramebuffer(e.FRAMEBUFFER, null), e.deleteTexture(n), e.deleteFramebuffer(r), a
+            }
+
+            function qM(e) {
+                if (2 !== e) return !1;
+                return null != YF(e).fenceSync
+            }
+
+            function KM(e, t) {
+                Array.isArray(e) || (e = [e]), e.forEach(e => {
+                    null != e && va.assert("complex64" !== e.dtype, () => t + " does not support complex64 tensors in the WebGL backend.")
+                })
+            }
+            var YM = Ye();
+
+            function XM() {
+                let e, t, n, r, a, s, i, o, u, l;
+                return 2 === Ye().getNumber("WEBGL_VERSION") ? (e = "#version 300 es", t = "in", n = "out", r = "in", a = "texture", s = "outputColor", i = "out vec4 outputColor;", o = "\n      bool isnan_custom(float val) {\n        return (val > 0.0 || val < 0.0) ? false : val != 0.0;\n      }\n\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan_custom(val.x),\n          isnan_custom(val.y), isnan_custom(val.z), isnan_custom(val.w));\n      }\n\n      #define isnan(value) isnan_custom(value)\n    ", u = "", l = "\n      #define round(value) newRound(value)\n      int newRound(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 newRound(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    ") : (e = "", t = "attribute", n = "varying", r = "varying", a = "texture2D", s = "gl_FragColor", i = "", o = "\n      #define isnan(value) isnan_custom(value)\n      bool isnan_custom(float val) {\n        return (val > 0. || val < 1. || val == 0.) ? false : true;\n      }\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));\n      }\n    ", u = "\n      uniform float INFINITY;\n\n      bool isinf(float val) {\n        return abs(val) == INFINITY;\n      }\n      bvec4 isinf(vec4 val) {\n        return equal(abs(val), vec4(INFINITY));\n      }\n    ", l = "\n      int round(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 round(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    "), {
+                    version: e,
+                    attribute: t,
+                    varyingVs: n,
+                    varyingFs: r,
+                    texture2D: a,
+                    output: s,
+                    defineOutput: i,
+                    defineSpecialNaN: o,
+                    defineSpecialInf: u,
+                    defineRound: l
+                }
+            }
+
+            function QM(e, t, n = "index") {
+                const r = va.computeStrides(t);
+                return r.map((t, a) => `${`int ${e[a]} = ${n} / ${t}`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${t}`:`index -= ${e[a]} * ${t}`};`).join("")
+            }
+
+            function ZM(e, t, n = "index") {
+                const r = va.computeStrides(t);
+                return r.map((t, a) => `${`int ${e[a]} = ${n} / outShapeStrides[${a}]`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * outShapeStrides[${a}]`:`index -= ${e[a]} * outShapeStrides[${a}]`};`).join("")
+            }
+
+            function JM(e, t, n = "index") {
+                const r = function (e, t) {
+                    const n = e.length,
+                        r = e.map(e => `${t}[${e}]`),
+                        a = new Array(n - 1);
+                    a[n - 2] = r[n - 1];
+                    for (let e = n - 3; e >= 0; --e) a[e] = `(${a[e+1]} * ${r[e+1]})`;
+                    return a
+                }(e.map((e, t) => t), t);
+                return r.map((t, a) => `${`int ${e[a]} = ${n} / ${r[a]}`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${r[a]}`:`index -= ${e[a]} * ${r[a]}`};`).join("")
+            }
+
+            function eD(e) {
+                const t = va.computeStrides(e).map(e => e.toString());
+                return `\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;\n  }\n`
+            }
+            YM.registerFlag("HAS_WEBGL", () => YM.getNumber("WEBGL_VERSION") > 0), YM.registerFlag("WEBGL_VERSION", () => VM(2) ? 2 : VM(1) ? 1 : 0), YM.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS", () => !1), YM.registerFlag("WEBGL_BUFFER_SUPPORTED", () => 2 === YM.get("WEBGL_VERSION")), YM.registerFlag("WEBGL_CPU_FORWARD", () => !0), YM.registerFlag("WEBGL_FORCE_F16_TEXTURES", () => !1), YM.registerFlag("WEBGL_PACK", () => YM.getBool("HAS_WEBGL")), YM.registerFlag("WEBGL_PACK_NORMALIZATION", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_CLIP", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_DEPTHWISECONV", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_BINARY_OPERATIONS", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_UNARY_OPERATIONS", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_PACK_REDUCE", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_LAZILY_UNPACK", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_CONV_IM2COL", () => YM.getBool("WEBGL_PACK")), YM.registerFlag("WEBGL_MAX_TEXTURE_SIZE", () => PM(YM.getNumber("WEBGL_VERSION"))), YM.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER", () => zM(YM.getNumber("WEBGL_VERSION"))), YM.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION", () => {
+                const e = YM.getNumber("WEBGL_VERSION");
+                return 0 === e ? 0 : UM(e)
+            }), YM.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE", () => YM.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") > 0 && !Ss.isMobile()), YM.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE", () => jM(YM.getNumber("WEBGL_VERSION"))), YM.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED", () => !YM.getBool("WEBGL_FORCE_F16_TEXTURES") && YM.getBool("WEBGL_RENDER_FLOAT32_CAPABLE")), YM.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED", () => GM(YM.getNumber("WEBGL_VERSION"))), YM.registerFlag("WEBGL_FENCE_API_ENABLED", () => qM(YM.getNumber("WEBGL_VERSION"))), YM.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM", () => YM.getBool("WEBGL_RENDER_FLOAT32_ENABLED") ? 4 : 0), YM.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD", () => -1, e => {
+                if (e < 0 && -1 !== e) throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)
+            }), YM.registerFlag("WEBGL_FLUSH_THRESHOLD", () => Ss.isMobile() ? 1 : -1, e => {
+                if (e < 0 && -1 !== e) throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)
+            }), YM.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD", () => 128), YM.registerFlag("WEBGL_USE_SHAPES_UNIFORMS", () => !1), YM.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD", () => 1e5), YM.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD", () => 128);
+            var tD = "\n  const float FLOAT_MAX = 1.70141184e38;\n  const float FLOAT_MIN = 1.17549435e-38;\n\n  lowp vec4 encode_float(highp float v) {\n    if (isnan(v)) {\n      return vec4(255, 255, 255, 255);\n    }\n\n    highp float av = abs(v);\n\n    if(av < FLOAT_MIN) {\n      return vec4(0.0, 0.0, 0.0, 0.0);\n    } else if(v > FLOAT_MAX) {\n      return vec4(0.0, 0.0, 128.0, 127.0) / 255.0;\n    } else if(v < -FLOAT_MAX) {\n      return vec4(0.0, 0.0,  128.0, 255.0) / 255.0;\n    }\n\n    highp vec4 c = vec4(0,0,0,0);\n\n    highp float e = floor(log2(av));\n    highp float m = exp2(fract(log2(av))) - 1.0;\n\n    c[2] = floor(128.0 * m);\n    m -= c[2] / 128.0;\n    c[1] = floor(32768.0 * m);\n    m -= c[1] / 32768.0;\n    c[0] = floor(8388608.0 * m);\n\n    highp float ebias = e + 127.0;\n    c[3] = floor(ebias / 2.0);\n    ebias -= c[3] * 2.0;\n    c[2] += floor(ebias) * 128.0;\n\n    c[3] += 128.0 * step(0.0, -v);\n\n    return c / 255.0;\n  }\n",
+                {
+                    getBroadcastDims: nD
+                } = jd;
+
+            function rD(e, t, n) {
+                const r = [];
+                if (e.forEach(e => {
+                        const t = va.sizeFromShape(e.shapeInfo.logicalShape);
+                        if (e.shapeInfo.isUniform ? r.push(`uniform float ${e.name}${t>1?`[${t}]`:""};`) : (r.push(`uniform sampler2D ${e.name};`), r.push(`uniform int offset${e.name};`)), n.enableShapeUniforms) {
+                            const {
+                                uniformShape: t
+                            } = fD(n.packedInputs, e.shapeInfo.logicalShape, e.shapeInfo.texShape);
+                            switch (t.length) {
+                                case 1:
+                                    r.push(`uniform int ${e.name}Shape;`);
+                                    break;
+                                case 2:
+                                    r.push(`uniform ivec2 ${e.name}Shape;`);
+                                    break;
+                                case 3:
+                                    r.push(`uniform ivec3 ${e.name}Shape;`);
+                                    break;
+                                case 4:
+                                    r.push(`uniform ivec4 ${e.name}Shape;`)
+                            }
+                            r.push(`uniform ivec2 ${e.name}TexShape;`)
+                        }
+                    }), n.enableShapeUniforms) {
+                    switch (t.logicalShape.length) {
+                        case 1:
+                            r.push("uniform int outShape;");
+                            break;
+                        case 2:
+                            r.push("uniform ivec2 outShape;"), r.push("uniform int outShapeStrides;");
+                            break;
+                        case 3:
+                            r.push("uniform ivec3 outShape;"), r.push("uniform ivec2 outShapeStrides;");
+                            break;
+                        case 4:
+                            r.push("uniform ivec4 outShape;"), r.push("uniform ivec3 outShapeStrides;")
+                    }
+                    r.push("uniform ivec2 outTexShape;")
+                }
+                n.customUniforms && n.customUniforms.forEach(e => {
+                    r.push(`uniform ${e.type} ${e.name}${e.arrayIndex?`[${e.arrayIndex}]`:""};`)
+                });
+                const a = r.join("\n"),
+                    s = e.map(e => function (e, t, n = !1, r) {
+                        let a = "";
+                        a += n ? sD(e, r) : aD(e, r);
+                        const s = e.shapeInfo.logicalShape,
+                            i = t.logicalShape;
+                        s.length <= i.length && (a += n ? function (e, t) {
+                            const n = e.name,
+                                r = n.charAt(0).toUpperCase() + n.slice(1),
+                                a = "get" + r + "AtOutCoords",
+                                s = e.shapeInfo.logicalShape.length,
+                                i = t.logicalShape.length,
+                                o = nD(e.shapeInfo.logicalShape, t.logicalShape),
+                                u = dD(i),
+                                l = i - s;
+                            let c;
+                            const h = ["x", "y", "z", "w", "u", "v"];
+                            c = 0 === s ? "" : i < 2 && o.length >= 1 ? "coords = 0;" : o.map(e => `coords.${h[e+l]} = 0;`).join("\n");
+                            let p = "";
+                            p = i < 2 && s > 0 ? "coords" : e.shapeInfo.logicalShape.map((e, t) => "coords." + h[t + l]).join(", ");
+                            let d = "return outputValue;";
+                            const f = 1 === va.sizeFromShape(e.shapeInfo.logicalShape),
+                                m = 1 === va.sizeFromShape(t.logicalShape);
+                            if (1 !== s || f || m) {
+                                if (f && !m) d = 1 === i ? "\n        return vec4(outputValue.x, outputValue.x, 0., 0.);\n      " : "\n        return vec4(outputValue.x);\n      ";
+                                else if (o.length) {
+                                    const e = s - 2,
+                                        t = s - 1;
+                                    o.indexOf(e) > -1 && o.indexOf(t) > -1 ? d = "return vec4(outputValue.x);" : o.indexOf(e) > -1 ? d = "return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);" : o.indexOf(t) > -1 && (d = "return vec4(outputValue.xx, outputValue.zz);")
+                                }
+                            } else d = "\n      return vec4(outputValue.xy, outputValue.xy);\n    ";
+                            return `\n    vec4 ${a}() {\n      ${u} coords = getOutputCoords();\n      ${c}\n      vec4 outputValue = get${r}(${p});\n      ${d}\n    }\n  `
+                        }(e, t) : function (e, t) {
+                            const n = e.name,
+                                r = n.charAt(0).toUpperCase() + n.slice(1),
+                                a = "get" + r + "AtOutCoords",
+                                s = t.texShape,
+                                i = e.shapeInfo.texShape,
+                                o = e.shapeInfo.logicalShape.length,
+                                u = t.logicalShape.length;
+                            if (!e.shapeInfo.isUniform && o === u && null == e.shapeInfo.flatOffset && va.arraysEqual(i, s)) return `\n      float ${a}() {\n        return sampleTexture(${n}, resultUV);\n      }\n    `;
+                            const l = dD(u),
+                                c = nD(e.shapeInfo.logicalShape, t.logicalShape),
+                                h = u - o;
+                            let p;
+                            const d = ["x", "y", "z", "w", "u", "v"];
+                            p = 0 === o ? "" : u < 2 && c.length >= 1 ? "coords = 0;" : c.map(e => `coords.${d[e+h]} = 0;`).join("\n");
+                            let f = "";
+                            f = u < 2 && o > 0 ? "coords" : e.shapeInfo.logicalShape.map((e, t) => "coords." + d[t + h]).join(", ");
+                            return `\n    float ${a}() {\n      ${l} coords = getOutputCoords();\n      ${p}\n      return get${r}(${f});\n    }\n  `
+                        }(e, t));
+                        return a
+                    }(e, t, n.packedInputs, n.enableShapeUniforms)).join("\n"),
+                    i = t.texShape,
+                    o = XM(),
+                    u = function (e) {
+                        return `\n    float sampleTexture(sampler2D textureSampler, vec2 uv) {\n      return ${e.texture2D}(textureSampler, uv).r;\n    }\n  `
+                    }(o);
+                let l, c, h = function (e) {
+                    return `${e.version}\n    precision highp float;\n    precision highp int;\n    precision highp sampler2D;\n    ${e.varyingFs} vec2 resultUV;\n    ${e.defineOutput}\n    const vec2 halfCR = vec2(0.5, 0.5);\n\n    struct ivec5\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n    };\n\n    struct ivec6\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n      int v;\n    };\n\n    uniform float NAN;\n    ${e.defineSpecialNaN}\n    ${e.defineSpecialInf}\n    ${e.defineRound}\n\n    int imod(int x, int y) {\n      return x - y * (x / y);\n    }\n\n    int idiv(int a, int b, float sign) {\n      int res = a / b;\n      int mod = imod(a, b);\n      if (sign < 0. && mod != 0) {\n        res -= 1;\n      }\n      return res;\n    }\n\n    //Based on the work of Dave Hoskins\n    //https://www.shadertoy.com/view/4djSRW\n    #define HASHSCALE1 443.8975\n    float random(float seed){\n      vec2 p = resultUV * seed;\n      vec3 p3  = fract(vec3(p.xyx) * HASHSCALE1);\n      p3 += dot(p3, p3.yzx + 19.19);\n      return fract((p3.x + p3.y) * p3.z);\n    }\n\n    ${iD}\n    ${oD}\n    ${uD}\n  `
+                }(o);
+                t.isPacked ? (l = function (e, t, n) {
+                    switch (e.length) {
+                        case 0:
+                            return cD();
+                        case 1:
+                            return function (e, t, n) {
+                                const r = [Math.ceil(t[0] / 2), Math.ceil(t[1] / 2)];
+                                if (1 === r[0]) return n ? "\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));\n      }\n    " : `\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ${r[1]}.0);\n      }\n    `;
+                                if (1 === r[1]) return n ? "\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));\n      }\n    " : `\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ${r[0]}.0);\n      }\n    `;
+                                if (n) return "\n    int getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);\n    }\n  ";
+                                return `\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      return 2 * (resTexRC.x * ${r[1]} + resTexRC.y);\n    }\n  `
+                            }(0, t, n);
+                        case 2:
+                            return function (e, t, n) {
+                                const r = [Math.ceil(t[0] / 2), Math.ceil(t[1] / 2)];
+                                if (va.arraysEqual(e, t)) return n ? "\n      ivec2 getOutputCoords() {\n        ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n        return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));\n      }\n    " : `\n      ivec2 getOutputCoords() {\n        return 2 * ivec2(resultUV.yx * vec2(${r[0]}, ${r[1]}));\n      }\n    `;
+                                const a = Math.ceil(e[1] / 2);
+                                if (n) return "\n    ivec2 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec2(r, c);\n    }\n  ";
+                                return `\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec2(r, c);\n    }\n  `
+                            }(e, t, n);
+                        case 3:
+                            return function (e, t, n) {
+                                if (n) return "\n    ivec3 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec3(b, r, c);\n    }\n  ";
+                                const r = [Math.ceil(t[0] / 2), Math.ceil(t[1] / 2)],
+                                    a = Math.ceil(e[2] / 2),
+                                    s = a * Math.ceil(e[1] / 2);
+                                return `\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `
+                            }(e, t, n);
+                        default:
+                            return function (e, t, n) {
+                                if (n) return "\n    ivec4 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int texelsInLogicalRow = int(ceil(float(outShape[3]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatchN = texelsInBatch * outShape[1];\n\n      int b2 = index / texelsInBatchN;\n      index -= b2 * texelsInBatchN;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec4(b2, b, r, c);\n    }\n  ";
+                                const r = [Math.ceil(t[0] / 2), Math.ceil(t[1] / 2)],
+                                    a = Math.ceil(e[e.length - 1] / 2),
+                                    s = a * Math.ceil(e[e.length - 2] / 2);
+                                let i = s,
+                                    o = "",
+                                    u = "b, r, c";
+                                for (let t = 2; t < e.length - 1; t++) i *= e[e.length - t - 1], o = `\n      int b${t} = index / ${i};\n      index -= b${t} * ${i};\n    ` + o, u = `b${t}, ` + u;
+                                return `\n    ivec${e.length} getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n\n      ${o}\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec${e.length}(${u});\n    }\n  `
+                            }(e, t, n)
+                    }
+                }(t.logicalShape, i, n.enableShapeUniforms), c = function (e) {
+                    return `\n    void setOutput(vec4 val) {\n      ${e.output} = val;\n    }\n  `
+                }(o)) : (l = function (e, t, n) {
+                    switch (e.length) {
+                        case 0:
+                            return cD();
+                        case 1:
+                            return function (e, t, n) {
+                                if (1 === t[0]) return n ? "\n      int getOutputCoords() {\n        return int(resultUV.x * float(outTexShape[1]));\n      }\n    " : `\n      int getOutputCoords() {\n        return int(resultUV.x * ${t[1]}.0);\n      }\n    `;
+                                if (1 === t[1]) return n ? "\n      int getOutputCoords() {\n        return int(resultUV.y * float(outTexShape[0]));\n      }\n    " : `\n      int getOutputCoords() {\n        return int(resultUV.y * ${t[0]}.0);\n      }\n    `;
+                                if (n) return "\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      return resTexRC.x * outTexShape[1] + resTexRC.y;\n    }\n  ";
+                                return `\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      return resTexRC.x * ${t[1]} + resTexRC.y;\n    }\n  `
+                            }(0, t, n);
+                        case 2:
+                            return function (e, t, n) {
+                                if (va.arraysEqual(e, t)) return n ? "\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));\n      }\n    " : `\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));\n      }\n    `;
+                                if (1 === e[1]) return n ? "\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    " : `\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `;
+                                if (1 === e[0]) return n ? "\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(0, index);\n      }\n    " : `\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `;
+                                if (n) return "\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      int r = index / outShape[1];\n      int c = index - r * outShape[1];\n      return ivec2(r, c);\n    }\n  ";
+                                return `\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      int r = index / ${e[1]};\n      int c = index - r * ${e[1]};\n      return ivec2(r, c);\n    }\n  `
+                            }(e, t, n);
+                        case 3:
+                            return function (e, t, n) {
+                                if (n) {
+                                    return `\n  ivec3 getOutputCoords() {\n    ivec2 resTexRC = ivec2(resultUV.yx *\n                           vec2(outTexShape[0], outTexShape[1]));\n    int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n    ${ZM(["r","c","d"],e)}\n    return ivec3(r, c, d);\n  }\n`
+                                }
+                                const r = QM(["r", "c", "d"], e);
+                                return `\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${r}\n      return ivec3(r, c, d);\n    }\n  `
+                            }(e, t, n);
+                        case 4:
+                            return function (e, t, n) {
+                                if (n) {
+                                    return `\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      ${ZM(["r","c","d","d2"],e)}\n      return ivec4(r, c, d, d2);\n    }\n  `
+                                }
+                                const r = QM(["r", "c", "d", "d2"], e);
+                                return `\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${r}\n      return ivec4(r, c, d, d2);\n    }\n  `
+                            }(e, t, n);
+                        case 5:
+                            return function (e, t) {
+                                const n = QM(["r", "c", "d", "d2", "d3"], e);
+                                return `\n    ivec5 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},\n                             ${t[1]}));\n\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${n}\n\n      ivec5 outShape = ivec5(r, c, d, d2, d3);\n      return outShape;\n    }\n  `
+                            }(e, t);
+                        case 6:
+                            return function (e, t) {
+                                const n = QM(["r", "c", "d", "d2", "d3", "d4"], e);
+                                return `\n    ivec6 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${n}\n\n      ivec6 result = ivec6(r, c, d, d2, d3, d4);\n      return result;\n    }\n  `
+                            }(e, t);
+                        default:
+                            throw new Error(e.length + "-D output sampling is not yet supported")
+                    }
+                }(t.logicalShape, i, n.enableShapeUniforms), c = function (e) {
+                    return `\n    void setOutput(float val) {\n      ${e.output} = vec4(val, 0, 0, 0);\n    }\n  `
+                }(o)), n.packedInputs && (h += lD);
+                return [h, u, c, a, l, s, n.userCode].join("\n")
+            }
+
+            function aD(e, t = !1) {
+                const n = e.shapeInfo.logicalShape;
+                switch (n.length) {
+                    case 0:
+                        return function (e, t) {
+                            const n = e.name,
+                                r = "get" + n.charAt(0).toUpperCase() + n.slice(1);
+                            if (e.shapeInfo.isUniform) return `float ${r}() {return ${n};}`;
+                            const [a, s] = e.shapeInfo.texShape;
+                            if (1 === a && 1 === s) return `\n      float ${r}() {\n        return sampleTexture(${n}, halfCR);\n      }\n    `;
+                            const i = hD(n);
+                            if (t) return `\n    float ${r}() {\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], ${i});\n      return sampleTexture(${n}, uv);\n    }\n  `;
+                            const [o, u] = e.shapeInfo.texShape;
+                            return `\n    float ${r}() {\n      vec2 uv = uvFromFlat(${o}, ${u}, ${i});\n      return sampleTexture(${n}, uv);\n    }\n  `
+                        }(e, t);
+                    case 1:
+                        return function (e, t) {
+                            const n = e.name,
+                                r = "get" + n.charAt(0).toUpperCase() + n.slice(1);
+                            if (e.shapeInfo.isUniform) return `\n      float ${r}(int index) {\n        ${pD(e)}\n      }\n    `;
+                            const a = e.shapeInfo.texShape,
+                                s = a[0],
+                                i = a[1];
+                            if (1 === i && 1 === s) return `\n      float ${r}(int index) {\n        return sampleTexture(${n}, halfCR);\n      }\n    `;
+                            const o = hD(n);
+                            if (1 === i) return t ? `\n      float ${r}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / float(${n}TexShape[0]));\n        return sampleTexture(${n}, uv);\n      }\n    ` : `\n      float ${r}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / ${s}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;
+                            if (1 === s) return t ? `\n      float ${r}(int index) {\n        vec2 uv = vec2((float(index + ${o}) + 0.5) / float(${n}TexShape[1]), 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    ` : `\n      float ${r}(int index) {\n        vec2 uv = vec2((float(index + ${o}) + 0.5) / ${i}.0, 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    `;
+                            if (t) return `\n    float ${r}(int index) {\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${o});\n      return sampleTexture(${n}, uv);\n    }\n  `;
+                            return `\n    float ${r}(int index) {\n      vec2 uv = uvFromFlat(${s}, ${i}, index + ${o});\n      return sampleTexture(${n}, uv);\n    }\n  `
+                        }(e, t);
+                    case 2:
+                        return function (e, t) {
+                            const n = e.shapeInfo.logicalShape,
+                                r = e.name,
+                                a = "get" + r.charAt(0).toUpperCase() + r.slice(1),
+                                s = e.shapeInfo.texShape;
+                            if (null != s && va.arraysEqual(n, s)) {
+                                if (t) return `\n      float ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `;
+                                const e = s[0],
+                                    n = s[1];
+                                return `\n    float ${a}(int row, int col) {\n      vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}.0, ${e}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `
+                            }
+                            const {
+                                newShape: i,
+                                keptDims: o
+                            } = va.squeezeShape(n), u = i;
+                            if (u.length < n.length) {
+                                const n = mD(e, u),
+                                    r = ["row", "col"];
+                                return `\n      ${aD(n,t)}\n      float ${a}(int row, int col) {\n        return ${a}(${gD(r,o)});\n      }\n    `
+                            }
+                            if (e.shapeInfo.isUniform) return `\n      float ${a}(int row, int col) {\n        int index = round(dot(vec2(row, col), vec2(${n[1]}, 1)));\n        ${pD(e)}\n      }\n    `;
+                            const l = s[0],
+                                c = s[1],
+                                h = hD(r);
+                            if (1 === c) return t ? `\n      float ${a}(int row, int col) {\n        float index = dot(vec3(row, col, ${h}), vec3(${r}Shape[1], 1, 1));\n        vec2 uv = vec2(0.5, (index + 0.5) / float(${r}TexShape[0]));\n        return sampleTexture(${r}, uv);\n      }\n    ` : `\n    float ${a}(int row, int col) {\n      float index = dot(vec3(row, col, ${h}), vec3(${n[1]}, 1, 1));\n      vec2 uv = vec2(0.5, (index + 0.5) / ${l}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `;
+                            if (1 === l) return t ? `\n      float ${a}(int row, int col) {\n        float index = dot(vec3(row, col, ${h}), vec3(${r}Shape[1], 1, 1));\n        vec2 uv = vec2((index + 0.5) / float(${r}TexShape[1]), 0.5);\n        return sampleTexture(${r}, uv);\n      }\n    ` : `\n    float ${a}(int row, int col) {\n      float index = dot(vec3(row, col, ${h}), vec3(${n[1]}, 1, 1));\n      vec2 uv = vec2((index + 0.5) / ${c}.0, 0.5);\n      return sampleTexture(${r}, uv);\n    }\n  `;
+                            if (t) return `\n      float ${a}(int row, int col) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${r}Shape[1] + col + ${h};\n        vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);\n        return sampleTexture(${r}, uv);\n      }\n    `;
+                            return `\n  float ${a}(int row, int col) {\n    // Explicitly use integer operations as dot() only works on floats.\n    int index = row * ${n[1]} + col + ${h};\n    vec2 uv = uvFromFlat(${l}, ${c}, index);\n    return sampleTexture(${r}, uv);\n  }\n`
+                        }(e, t);
+                    case 3:
+                        return function (e, t) {
+                            const n = e.shapeInfo.logicalShape,
+                                r = e.name,
+                                a = "get" + r.charAt(0).toUpperCase() + r.slice(1),
+                                s = n[1] * n[2],
+                                i = n[2],
+                                {
+                                    newShape: o,
+                                    keptDims: u
+                                } = va.squeezeShape(n),
+                                l = o;
+                            if (l.length < n.length) {
+                                const n = mD(e, l),
+                                    r = ["row", "col", "depth"];
+                                return `\n        ${aD(n,t)}\n        float ${a}(int row, int col, int depth) {\n          return ${a}(${gD(r,u)});\n        }\n      `
+                            }
+                            if (e.shapeInfo.isUniform) return `\n      float ${a}(int row, int col, int depth) {\n        int index = round(dot(vec3(row, col, depth),\n                          vec3(${s}, ${i}, 1)));\n        ${pD(e)}\n      }\n    `;
+                            const c = e.shapeInfo.texShape,
+                                h = c[0],
+                                p = c[1],
+                                d = e.shapeInfo.flatOffset;
+                            if (p === s && null == d) return t ? `\n      float ${a}(int row, int col, int depth) {\n        int stride1 = ${r}Shape[2];\n        float texR = float(row);\n        float texC = dot(vec2(col, depth), vec2(stride1, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    ` : `\n        float ${a}(int row, int col, int depth) {\n          float texR = float(row);\n          float texC = dot(vec2(col, depth), vec2(${i}, 1));\n          vec2 uv = (vec2(texC, texR) + halfCR) /\n                     vec2(${p}.0, ${h}.0);\n          return sampleTexture(${r}, uv);\n        }\n      `;
+                            if (p === i && null == d) return t ? `\n      float ${a}(int row, int col, int depth) {\n        float texR = dot(vec2(row, col), vec2(${r}Shape[1], 1));\n        float texC = float(depth);\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    ` : `\n    float ${a}(int row, int col, int depth) {\n      float texR = dot(vec2(row, col), vec2(${n[1]}, 1));\n      float texC = float(depth);\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${p}.0, ${h}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `;
+                            const f = hD(r);
+                            if (t) return `\n    float ${a}(int row, int col, int depth) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int stride0 = ${r}Shape[1] * ${r}Shape[2];\n      int stride1 = ${r}Shape[2];\n      int index = row * ${s} + col * ${i} + depth + ${f};\n      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);\n      return sampleTexture(${r}, uv);\n    }\n    `;
+                            return `\n      float ${a}(int row, int col, int depth) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${s} + col * ${i} + depth + ${f};\n        vec2 uv = uvFromFlat(${h}, ${p}, index);\n        return sampleTexture(${r}, uv);\n      }\n  `
+                        }(e, t);
+                    case 4:
+                        return function (e, t) {
+                            const n = e.shapeInfo.logicalShape,
+                                r = e.name,
+                                a = "get" + r.charAt(0).toUpperCase() + r.slice(1),
+                                s = n[3],
+                                i = n[2] * s,
+                                o = n[1] * i,
+                                {
+                                    newShape: u,
+                                    keptDims: l
+                                } = va.squeezeShape(n);
+                            if (u.length < n.length) {
+                                const n = mD(e, u),
+                                    r = ["row", "col", "depth", "depth2"];
+                                return `\n      ${aD(n,t)}\n      float ${a}(int row, int col, int depth, int depth2) {\n        return ${a}(${gD(r,l)});\n      }\n    `
+                            }
+                            if (e.shapeInfo.isUniform) return `\n      float ${a}(int row, int col, int depth, int depth2) {\n        int index = round(dot(vec4(row, col, depth, depth2),\n                          vec4(${o}, ${i}, ${s}, 1)));\n        ${pD(e)}\n      }\n    `;
+                            const c = e.shapeInfo.flatOffset,
+                                h = e.shapeInfo.texShape,
+                                p = h[0],
+                                d = h[1],
+                                f = `int stride2 = ${r}Shape[3];`,
+                                m = `int stride1 = ${r}Shape[2] * stride2;`,
+                                g = `int stride0 = ${r}Shape[1] * stride1;`;
+                            if (d === o && null == c) return t ? `\n      float ${a}(int row, int col, int depth, int depth2) {\n        ${f}\n        ${m}\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(stride1, stride2, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    ` : `\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(${i}, ${s}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${d}.0, ${p}.0);\n        return sampleTexture(${r}, uv);\n      }\n    `;
+                            if (d === s && null == c) return t ? `\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${r}Shape[1] * ${r}Shape[2], ${r}Shape[2], 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    ` : `\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${n[1]*n[2]}, ${n[2]}, 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${d}.0, ${p}.0);\n        return sampleTexture(${r}, uv);\n      }\n    `;
+                            const y = hD(r);
+                            if (t) return `\n    float ${a}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      ${f}\n      ${m}\n      ${g}\n      int index = row * stride0 + col * stride1 +\n          depth * stride2 + depth2;\n      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index + ${y});\n      return sampleTexture(${r}, uv);\n    }\n  `;
+                            return `\n    float ${a}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${o} + col * ${i} +\n          depth * ${s} + depth2;\n      vec2 uv = uvFromFlat(${p}, ${d}, index + ${y});\n      return sampleTexture(${r}, uv);\n    }\n  `
+                        }(e, t);
+                    case 5:
+                        return function (e) {
+                            const t = e.shapeInfo.logicalShape,
+                                n = e.name,
+                                r = "get" + n.charAt(0).toUpperCase() + n.slice(1),
+                                a = t[4],
+                                s = t[3] * a,
+                                i = t[2] * s,
+                                o = t[1] * i,
+                                {
+                                    newShape: u,
+                                    keptDims: l
+                                } = va.squeezeShape(t);
+                            if (u.length < t.length) {
+                                const t = mD(e, u),
+                                    n = ["row", "col", "depth", "depth2", "depth3"];
+                                return `\n      ${aD(t)}\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        return ${r}(${gD(n,l)});\n      }\n    `
+                            }
+                            if (e.shapeInfo.isUniform) return `\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        float index = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${o}, ${i}, ${s}, ${a})) +\n          depth3;\n        ${pD(e)}\n      }\n    `;
+                            const c = e.shapeInfo.flatOffset,
+                                h = e.shapeInfo.texShape,
+                                p = h[0],
+                                d = h[1];
+                            if (d === o && null == c) return `\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n                         vec4(${i}, ${s}, ${a}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${d}.0, ${p}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;
+                            if (d === a && null == c) return `\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        float texR = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]},\n               ${t[2]*t[3]}, ${t[3]}, 1));\n        int texC = depth3;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${d}.0, ${p}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;
+                            const f = hD(n);
+                            return `\n    float ${r}(int row, int col, int depth, int depth2, int depth3) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${o} + col * ${i} + depth * ${s} +\n          depth2 * ${a} + depth3 + ${f};\n      vec2 uv = uvFromFlat(${p}, ${d}, index);\n      return sampleTexture(${n}, uv);\n    }\n  `
+                        }(e);
+                    case 6:
+                        return function (e) {
+                            const t = e.shapeInfo.logicalShape,
+                                n = e.name,
+                                r = "get" + n.charAt(0).toUpperCase() + n.slice(1),
+                                {
+                                    newShape: a,
+                                    keptDims: s
+                                } = va.squeezeShape(t);
+                            if (a.length < t.length) {
+                                const t = mD(e, a),
+                                    n = ["row", "col", "depth", "depth2", "depth3", "depth4"];
+                                return `\n      ${aD(t)}\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        return ${r}(${gD(n,s)});\n      }\n    `
+                            }
+                            const i = t[5],
+                                o = t[4] * i,
+                                u = t[3] * o,
+                                l = t[2] * u,
+                                c = t[1] * l;
+                            if (e.shapeInfo.isUniform) return `\n      float ${r}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n        int index = round(dot(\n          vec4(row, col, depth, depth2),\n          vec4(${c}, ${l}, ${u}, ${o})) +\n          dot(\n            vec2(depth3, depth4),\n            vec2(${i}, 1)));\n        ${pD(e)}\n      }\n    `;
+                            const h = e.shapeInfo.flatOffset,
+                                p = e.shapeInfo.texShape,
+                                d = p[0],
+                                f = p[1];
+                            if (f === c && null == h) return `\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n          vec4(${l}, ${u}, ${o}, ${i})) +\n               float(depth4);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${f}.0, ${d}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;
+                            if (f === i && null == h) return `\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        float texR = dot(vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]*t[4]},\n               ${t[2]*t[3]*t[4]},\n               ${t[3]*t[4]},\n               ${t[4]})) + float(depth3);\n        int texC = depth4;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${f}.0, ${d}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;
+                            const m = hD(n);
+                            return `\n    float ${r}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${c} + col * ${l} + depth * ${u} +\n          depth2 * ${o} + depth3 * ${i} + depth4 + ${m};\n      vec2 uv = uvFromFlat(${d}, ${f}, index);\n      return sampleTexture(${n}, uv);\n    }\n  `
+                        }(e);
+                    default:
+                        throw new Error(n.length + "-D input sampling is not yet supported")
+                }
+            }
+
+            function sD(e, t) {
+                switch (e.shapeInfo.logicalShape.length) {
+                    case 0:
+                        return function (e) {
+                            const t = e.name,
+                                n = "get" + t.charAt(0).toUpperCase() + t.slice(1),
+                                r = XM();
+                            return `\n    vec4 ${n}() {\n      return ${r.texture2D}(${t}, halfCR);\n    }\n  `
+                        }(e);
+                    case 1:
+                        return function (e, t) {
+                            const n = e.name,
+                                r = "get" + n.charAt(0).toUpperCase() + n.slice(1),
+                                a = e.shapeInfo.texShape,
+                                s = XM();
+                            if (t) return `\n    vec4 ${r}(int index) {\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      vec2 uv = packedUVfrom1D(\n        packedTexShape[0], packedTexShape[1], index);\n      return ${s.texture2D}(${n}, uv);\n    }\n  `;
+                            const i = [Math.ceil(a[0] / 2), Math.ceil(a[1] / 2)];
+                            return `\n    vec4 ${r}(int index) {\n      vec2 uv = packedUVfrom1D(\n        ${i[0]}, ${i[1]}, index);\n      return ${s.texture2D}(${n}, uv);\n    }\n  `
+                        }(e, t);
+                    case 2:
+                        return function (e, t) {
+                            const n = e.shapeInfo.logicalShape,
+                                r = e.name,
+                                a = "get" + r.charAt(0).toUpperCase() + r.slice(1),
+                                s = e.shapeInfo.texShape,
+                                i = s[0],
+                                o = s[1],
+                                u = XM();
+                            if (null != s && va.arraysEqual(n, s)) return t ? `\n      vec4 ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n\n        return ${u.texture2D}(${r}, uv);\n      }\n    ` : `\n      vec4 ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${o}.0, ${i}.0);\n\n        return ${u.texture2D}(${r}, uv);\n      }\n    `;
+                            if (t) return `\n    vec4 ${a}(int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${r}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);\n      return ${u.texture2D}(${r}, uv);\n    }\n  `;
+                            const l = [Math.ceil(s[0] / 2), Math.ceil(s[1] / 2)],
+                                c = Math.ceil(n[1] / 2);
+                            return `\n    vec4 ${a}(int row, int col) {\n      vec2 uv = packedUVfrom2D(${c}, ${l[0]}, ${l[1]}, row, col);\n      return ${u.texture2D}(${r}, uv);\n    }\n  `
+                        }(e, t);
+                    case 3:
+                        return function (e, t) {
+                            const n = e.shapeInfo.logicalShape,
+                                r = e.name,
+                                a = "get" + r.charAt(0).toUpperCase() + r.slice(1),
+                                s = e.shapeInfo.texShape,
+                                i = [Math.ceil(s[0] / 2), Math.ceil(s[1] / 2)];
+                            if (1 === n[0]) {
+                                const r = n.slice(1),
+                                    s = [1, 2],
+                                    i = mD(e, r),
+                                    o = ["b", "row", "col"];
+                                return `\n        ${sD(i,t)}\n        vec4 ${a}(int b, int row, int col) {\n          return ${a}(${gD(o,s)});\n        }\n      `
+                            }
+                            const o = XM();
+                            if (t) return `\n    vec4 ${a}(int b, int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${r}Shape[2]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${r}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom3D(\n        packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);\n      return ${o.texture2D}(${r}, uv);\n    }\n  `;
+                            const u = i[0],
+                                l = i[1],
+                                c = Math.ceil(n[2] / 2),
+                                h = c * Math.ceil(n[1] / 2);
+                            return `\n    vec4 ${a}(int b, int row, int col) {\n      vec2 uv = packedUVfrom3D(\n        ${u}, ${l}, ${h}, ${c}, b, row, col);\n      return ${o.texture2D}(${r}, uv);\n    }\n  `
+                        }(e, t);
+                    default:
+                        return function (e, t) {
+                            const n = e.name,
+                                r = "get" + n.charAt(0).toUpperCase() + n.slice(1),
+                                a = XM();
+                            if (t) return `\n    vec4 ${r}(int b2, int b, int row, int col) {\n      int valuesPerRow = int(ceil(float(${n}Shape[3]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[2]) / 2.0));\n      int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);\n      texelsInBatch *= ${n}Shape[1];\n      index = b2 * texelsInBatch + index;\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      int texR = index / packedTexShape[1];\n      int texC = index - texR * packedTexShape[1];\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${a.texture2D}(${n}, uv);\n    }\n  `;
+                            const s = e.shapeInfo.logicalShape,
+                                i = s.length,
+                                o = e.shapeInfo.texShape,
+                                u = [Math.ceil(o[0] / 2), Math.ceil(o[1] / 2)],
+                                l = u[0],
+                                c = u[1],
+                                h = Math.ceil(s[i - 1] / 2);
+                            let p = h * Math.ceil(s[i - 2] / 2),
+                                d = "int b, int row, int col",
+                                f = `b * ${p} + (row / 2) * ${h} + (col / 2)`;
+                            for (let e = 2; e < i - 1; e++) d = `int b${e}, ` + d, p *= s[i - e - 1], f = `b${e} * ${p} + ` + f;
+                            return `\n    vec4 ${r}(${d}) {\n      int index = ${f};\n      int texR = index / ${c};\n      int texC = index - texR * ${c};\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${c}, ${l});\n      return ${a.texture2D}(${n}, uv);\n    }\n  `
+                        }(e, t)
+                }
+            }
+            var iD = "\nvec2 uvFromFlat(int texNumR, int texNumC, int index) {\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\nvec2 packedUVfrom1D(int texNumR, int texNumC, int index) {\n  int texelIndex = index / 2;\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",
+                oD = "\nvec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,\n  int texNumC, int row, int col) {\n  int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",
+                uD = "\nvec2 packedUVfrom3D(int texNumR, int texNumC,\n    int texelsInBatch, int texelsInLogicalRow, int b,\n    int row, int col) {\n  int index = b * texelsInBatch + (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",
+                lD = "\n  float getChannel(vec4 frag, vec2 innerDims) {\n    vec2 modCoord = mod(innerDims, 2.);\n    return modCoord.x == 0. ?\n      (modCoord.y == 0. ? frag.r : frag.g) :\n      (modCoord.y == 0. ? frag.b : frag.a);\n  }\n  float getChannel(vec4 frag, int dim) {\n    float modCoord = mod(float(dim), 2.);\n    return modCoord == 0. ? frag.r : frag.g;\n  }\n";
+
+            function cD() {
+                return "\n    int getOutputCoords() {\n      return 0;\n    }\n  "
+            }
+
+            function hD(e) {
+                return "offset" + e
+            }
+
+            function pD(e) {
+                const t = e.name,
+                    n = va.sizeFromShape(e.shapeInfo.logicalShape);
+                return n < 2 ? `return ${t};` : `\n    for (int i = 0; i < ${n}; i++) {\n      if (i == index) {\n        return ${t}[i];\n      }\n    }\n  `
+            }
+
+            function dD(e) {
+                if (e <= 1) return "int";
+                if (2 === e) return "ivec2";
+                if (3 === e) return "ivec3";
+                if (4 === e) return "ivec4";
+                if (5 === e) return "ivec5";
+                if (6 === e) return "ivec6";
+                throw Error(`GPU for rank ${e} is not yet supported`)
+            }
+
+            function fD(e, t, n) {
+                const {
+                    newShape: r,
+                    keptDims: a
+                } = va.squeezeShape(t), s = t.length, i = e && 3 === s && 1 === t[0], o = i ? t.slice(1) : r, u = !e && s > 1 && !va.arraysEqual(t, n) && r.length < s || i;
+                return {
+                    useSqueezeShape: u,
+                    uniformShape: u ? o : t,
+                    keptDims: a
+                }
+            }
+
+            function mD(e, t) {
+                const n = JSON.parse(JSON.stringify(e));
+                return n.shapeInfo.logicalShape = t, n
+            }
+
+            function gD(e, t) {
+                return t.map(t => e[t]).join(", ")
+            }
+
+            function yD(e, t) {
+                if (e.length !== t.length) throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);
+                e.forEach((e, n) => {
+                    const r = e.logicalShape,
+                        a = t[n],
+                        s = a.shape;
+                    if (!va.arraysEqual(r, s)) throw Error(`Binary was compiled with different shapes than the current args. Shapes ${r} and ${s} must match`);
+                    if (e.isUniform && a.isUniform) return;
+                    const i = e.texShape,
+                        o = a.isUniform ? null : a.texData.texShape;
+                    if (!va.arraysEqual(i, o)) throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${i} and ${o} must match`)
+                })
+            }
+
+            function bD(e) {
+                return Ye().getBool("WEBGL_USE_SHAPES_UNIFORMS") && e <= 4
+            }
+            var vD = {};
+
+            function xD(e) {
+                const t = XM();
+                return aM(e, `${t.version}\n    precision highp float;\n    ${t.attribute} vec3 clipSpacePos;\n    ${t.attribute} vec2 uv;\n    ${t.varyingVs} vec2 resultUV;\n\n    void main() {\n      gl_Position = vec4(clipSpacePos, 1);\n      resultUV = uv;\n    }`)
+            }
+
+            function wD(e) {
+                return pM(e, new Float32Array([-1, 1, 0, 0, 1, -1, -1, 0, 0, 0, 1, 1, 0, 1, 1, 1, -1, 0, 1, 0]))
+            }
+
+            function kD(e) {
+                return dM(e, new Uint16Array([0, 1, 2, 2, 1, 3]))
+            }
+
+            function SD(e, t, n, r, a, s) {
+                gM(t, n);
+                const i = mM(e),
+                    o = e.TEXTURE_2D;
+                return eM(e, () => e.bindTexture(o, i)), eM(e, () => e.texParameteri(o, e.TEXTURE_WRAP_S, e.CLAMP_TO_EDGE)), eM(e, () => e.texParameteri(o, e.TEXTURE_WRAP_T, e.CLAMP_TO_EDGE)), eM(e, () => e.texParameteri(o, e.TEXTURE_MIN_FILTER, e.NEAREST)), eM(e, () => e.texParameteri(o, e.TEXTURE_MAG_FILTER, e.NEAREST)), 1 === Ye().getNumber("WEBGL_VERSION") ? eM(e, () => e.texImage2D(o, 0, r, t, n, 0, a, s, null)) : eM(e, () => e.texStorage2D(o, 1, r, t, n)), eM(e, () => e.bindTexture(e.TEXTURE_2D, null)), {
+                    texture: i,
+                    texShape: [n, t]
+                }
+            }
+
+            function ID(e) {
+                return e.internalFormatFloat
+            }
+
+            function AD(e, t, n, r) {
+                const [a, s] = XF(t, n);
+                return SD(e, a, s, ID(r), r.textureFormatFloat, e.FLOAT)
+            }
+
+            function ED(e) {
+                return e.internalFormatHalfFloat
+            }
+
+            function ND(e, t, n, r) {
+                const [a, s] = XF(t, n);
+                return SD(e, a, s, ED(r), r.textureFormatFloat, r.textureTypeHalfFloat)
+            }
+
+            function CD(e) {
+                return e.downloadTextureFormat
+            }
+
+            function TD(e, t, n, r) {
+                const [a, s] = XF(t, n);
+                return SD(e, a, s, CD(r), e.RGBA, e.UNSIGNED_BYTE)
+            }
+
+            function RD(e) {
+                return e.internalFormatPackedFloat
+            }
+
+            function _D(e, t, n, r) {
+                const [a, s] = ZF(t, n);
+                return SD(e, a, s, RD(r), e.RGBA, e.FLOAT)
+            }
+
+            function FD(e) {
+                return e.internalFormatPackedHalfFloat
+            }
+
+            function MD(e, t, n, r) {
+                const [a, s] = ZF(t, n);
+                return SD(e, a, s, FD(r), e.RGBA, r.textureTypeHalfFloat)
+            }
+
+            function DD(e, t, n) {
+                eM(e, () => e.bindBuffer(e.ARRAY_BUFFER, n));
+                return bM(e, t, "clipSpacePos", n, 3, 20, 0) && bM(e, t, "uv", n, 2, 20, 12)
+            }
+
+            function OD(e, t, n, r, a, s) {
+                let i, o, u;
+                eM(e, () => e.bindTexture(e.TEXTURE_2D, t)), a instanceof Uint8Array ? (i = new Uint8Array(n * r * 4), o = e.UNSIGNED_BYTE, u = e.RGBA) : (i = new Float32Array(n * r * 4), o = e.FLOAT, u = s.internalFormatPackedFloat), i.set(a), 2 === Ye().getNumber("WEBGL_VERSION") ? eM(e, () => e.texSubImage2D(e.TEXTURE_2D, 0, 0, 0, n, r, e.RGBA, o, i)) : eM(e, () => e.texImage2D(e.TEXTURE_2D, 0, u, n, r, 0, e.RGBA, o, i)), eM(e, () => e.bindTexture(e.TEXTURE_2D, null))
+            }
+
+            function LD(e, t, n) {
+                eM(e, () => e.bindTexture(e.TEXTURE_2D, t)), n.data instanceof Uint8Array ? 2 === Ye().getNumber("WEBGL_VERSION") ? eM(e, () => e.texSubImage2D(e.TEXTURE_2D, 0, 0, 0, n.width, n.height, e.RGBA, e.UNSIGNED_BYTE, n.data)) : eM(e, () => e.texImage2D(e.TEXTURE_2D, 0, e.RGBA, n.width, n.height, 0, e.RGBA, e.UNSIGNED_BYTE, n.data)) : 2 === Ye().getNumber("WEBGL_VERSION") ? eM(e, () => e.texSubImage2D(e.TEXTURE_2D, 0, 0, 0, e.RGBA, e.UNSIGNED_BYTE, n)) : eM(e, () => e.texImage2D(e.TEXTURE_2D, 0, e.RGBA, e.RGBA, e.UNSIGNED_BYTE, n)), eM(e, () => e.bindTexture(e.TEXTURE_2D, null))
+            }
+
+            function PD(e, t, n, r) {
+                const a = e.createBuffer();
+                eM(e, () => e.bindBuffer(e.PIXEL_PACK_BUFFER, a));
+                const s = 16 * t * n;
+                return eM(e, () => e.bufferData(e.PIXEL_PACK_BUFFER, s, e.STREAM_READ)), eM(e, () => e.readPixels(0, 0, n, t, e.RGBA, e.FLOAT, 0)), eM(e, () => e.bindBuffer(e.PIXEL_PACK_BUFFER, null)), a
+            }
+
+            function $D(e, t, n) {
+                const r = e,
+                    a = new Float32Array(n);
+                return r.bindBuffer(r.PIXEL_PACK_BUFFER, t), r.getBufferSubData(r.PIXEL_PACK_BUFFER, 0, a), r.bindBuffer(r.PIXEL_PACK_BUFFER, null), a
+            }
+
+            function BD(e, t, n, r) {
+                const [a, s] = XF(t, n), i = new Uint8Array(t * n * 4);
+                return eM(e, () => e.readPixels(0, 0, a, s, r.downloadTextureFormat, e.UNSIGNED_BYTE, i)), new Float32Array(i.buffer)
+            }
+
+            function zD(e, t, n, r, a, s, i, o) {
+                const u = e,
+                    l = new Float32Array(function (e, t) {
+                        const [n, r] = ZF(e, t);
+                        return n * r * 4
+                    }(s, i));
+                return u.bindBuffer(u.PIXEL_PACK_BUFFER, t), u.getBufferSubData(u.PIXEL_PACK_BUFFER, 0, l), u.bindBuffer(u.PIXEL_PACK_BUFFER, null), l
+            }
+
+            function UD(e, t, n) {
+                const r = new Float32Array(t * n * 4);
+                return eM(e, () => e.readPixels(0, 0, n, t, e.RGBA, e.FLOAT, r)), r
+            }
+            w(vD, {
+                bindVertexProgramAttributeStreams: () => DD,
+                createBufferFromOutputTexture: () => PD,
+                createFloat16MatrixTexture: () => ND,
+                createFloat16PackedMatrixTexture: () => MD,
+                createFloat32MatrixTexture: () => AD,
+                createIndexBuffer: () => kD,
+                createPackedMatrixTexture: () => _D,
+                createUnsignedBytesMatrixTexture: () => TD,
+                createVertexBuffer: () => wD,
+                createVertexShader: () => xD,
+                downloadByteEncodedFloatMatrixFromOutputTexture: () => BD,
+                downloadFloat32MatrixFromBuffer: () => $D,
+                downloadMatrixFromPackedOutputTexture: () => UD,
+                downloadPackedMatrixFromBuffer: () => zD,
+                getInternalFormatForFloat16MatrixTexture: () => ED,
+                getInternalFormatForFloat16PackedMatrixTexture: () => FD,
+                getInternalFormatForFloat32MatrixTexture: () => ID,
+                getInternalFormatForPackedMatrixTexture: () => RD,
+                getInternalFormatForUnsignedBytesMatrixTexture: () => CD,
+                uploadDenseMatrixToTexture: () => OD,
+                uploadPixelDataToTexture: () => LD
+            });
+            var WD = class {
+                constructor(e) {
+                    this.outputTexture = null, this.program = null, this.disposed = !1, this.vertexAttrsAreBound = !1, this.itemsToPoll = [];
+                    const t = Ye().getNumber("WEBGL_VERSION");
+                    null != e ? (this.gl = e, KF(t, e)) : this.gl = YF(t);
+                    let n = "WEBGL_color_buffer_float";
+                    if (1 === Ye().getNumber("WEBGL_VERSION")) {
+                        const e = "OES_texture_float",
+                            t = "OES_texture_half_float";
+                        if (this.textureFloatExtension = rM(this.gl, e), WM(this.gl, t)) this.textureHalfFloatExtension = rM(this.gl, t);
+                        else if (Ye().get("WEBGL_FORCE_F16_TEXTURES")) throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");
+                        if (this.colorBufferFloatExtension = this.gl.getExtension(n), WM(this.gl, "EXT_color_buffer_half_float")) this.colorBufferHalfFloatExtension = rM(this.gl, "EXT_color_buffer_half_float");
+                        else if (Ye().get("WEBGL_FORCE_F16_TEXTURES")) throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")
+                    } else if (n = "EXT_color_buffer_float", WM(this.gl, n)) this.colorBufferFloatExtension = this.gl.getExtension(n);
+                    else {
+                        if (!WM(this.gl, "EXT_color_buffer_half_float")) throw new Error("GL context does not support color renderable floats");
+                        this.colorBufferHalfFloatExtension = this.gl.getExtension("EXT_color_buffer_half_float")
+                    }
+                    this.vertexBuffer = wD(this.gl), this.indexBuffer = kD(this.gl), this.framebuffer = yM(this.gl), this.textureConfig = JF(this.gl, this.textureHalfFloatExtension)
+                }
+                get debug() {
+                    return Ye().getBool("DEBUG")
+                }
+                dispose() {
+                    if (this.disposed) return;
+                    null != this.program && console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."), null != this.outputTexture && console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");
+                    const e = this.gl;
+                    eM(e, () => e.finish()), eM(e, () => e.bindFramebuffer(e.FRAMEBUFFER, null)), eM(e, () => e.deleteFramebuffer(this.framebuffer)), eM(e, () => e.bindBuffer(e.ARRAY_BUFFER, null)), eM(e, () => e.bindBuffer(e.ELEMENT_ARRAY_BUFFER, null)), eM(e, () => e.deleteBuffer(this.indexBuffer)), this.disposed = !0
+                }
+                createFloat32MatrixTexture(e, t) {
+                    return this.throwIfDisposed(), AD(this.gl, e, t, this.textureConfig)
+                }
+                createFloat16MatrixTexture(e, t) {
+                    return this.throwIfDisposed(), ND(this.gl, e, t, this.textureConfig)
+                }
+                createUnsignedBytesMatrixTexture(e, t) {
+                    return this.throwIfDisposed(), TD(this.gl, e, t, this.textureConfig)
+                }
+                uploadPixelDataToTexture(e, t) {
+                    this.throwIfDisposed(), LD(this.gl, e, t)
+                }
+                uploadDenseMatrixToTexture(e, t, n, r) {
+                    this.throwIfDisposed(), OD(this.gl, e, t, n, r, this.textureConfig)
+                }
+                createFloat16PackedMatrixTexture(e, t) {
+                    return this.throwIfDisposed(), MD(this.gl, e, t, this.textureConfig)
+                }
+                createPackedMatrixTexture(e, t) {
+                    return this.throwIfDisposed(), _D(this.gl, e, t, this.textureConfig)
+                }
+                deleteMatrixTexture(e) {
+                    this.throwIfDisposed(), this.outputTexture === e && (EM(this.gl, this.framebuffer), this.outputTexture = null), eM(this.gl, () => this.gl.deleteTexture(e))
+                }
+                downloadByteEncodedFloatMatrixFromOutputTexture(e, t, n) {
+                    return this.downloadMatrixDriver(e, () => BD(this.gl, t, n, this.textureConfig))
+                }
+                downloadPackedMatrixFromBuffer(e, t, n, r, a, s) {
+                    return zD(this.gl, e, 0, 0, 0, a, s)
+                }
+                downloadFloat32MatrixFromBuffer(e, t) {
+                    return $D(this.gl, e, t)
+                }
+                createBufferFromTexture(e, t, n) {
+                    this.bindTextureToFrameBuffer(e);
+                    const r = PD(this.gl, t, n);
+                    return this.unbindTextureToFrameBuffer(), r
+                }
+                createAndWaitForFence() {
+                    const e = this.createFence(this.gl);
+                    return this.pollFence(e)
+                }
+                createFence(e) {
+                    let t, n;
+                    if (Ye().getBool("WEBGL_FENCE_API_ENABLED")) {
+                        const r = e,
+                            a = r.fenceSync(r.SYNC_GPU_COMMANDS_COMPLETE, 0);
+                        e.flush(), n = () => {
+                            const e = r.clientWaitSync(a, 0, 0);
+                            return e === r.ALREADY_SIGNALED || e === r.CONDITION_SATISFIED
+                        }, t = a
+                    } else Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") > 0 ? (t = this.beginQuery(), this.endQuery(), n = () => this.isQueryAvailable(t, Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))) : n = () => !0;
+                    return {
+                        query: t,
+                        isFencePassed: n
+                    }
+                }
+                downloadMatrixFromPackedTexture(e, t, n) {
+                    return this.downloadMatrixDriver(e, () => UD(this.gl, t, n))
+                }
+                createProgram(e) {
+                    this.throwIfDisposed();
+                    const t = this.gl;
+                    null == this.vertexShader && (this.vertexShader = xD(t));
+                    const n = lM(t);
+                    return eM(t, () => t.attachShader(n, this.vertexShader)), eM(t, () => t.attachShader(n, e)), cM(t, n), this.debug && hM(t, n), this.vertexAttrsAreBound || (this.setProgram(n), this.vertexAttrsAreBound = DD(t, this.program, this.vertexBuffer)), n
+                }
+                deleteProgram(e) {
+                    this.throwIfDisposed(), e === this.program && (this.program = null), null != e && eM(this.gl, () => this.gl.deleteProgram(e))
+                }
+                setProgram(e) {
+                    this.throwIfDisposed(), this.program = e, null != this.program && this.debug && hM(this.gl, this.program), eM(this.gl, () => this.gl.useProgram(e))
+                }
+                getUniformLocation(e, t, n = !0) {
+                    return this.throwIfDisposed(), n ? wM(this.gl, e, t) : kM(this.gl, e, t)
+                }
+                getAttributeLocation(e, t) {
+                    return this.throwIfDisposed(), eM(this.gl, () => this.gl.getAttribLocation(e, t))
+                }
+                getUniformLocationNoThrow(e, t) {
+                    return this.throwIfDisposed(), this.gl.getUniformLocation(e, t)
+                }
+                setInputMatrixTexture(e, t, n) {
+                    this.throwIfDisposed(), this.throwIfNoProgram(), SM(this.gl, e, t, n)
+                }
+                setOutputMatrixTexture(e, t, n) {
+                    this.setOutputMatrixTextureDriver(e, n, t)
+                }
+                setOutputPackedMatrixTexture(e, t, n) {
+                    this.throwIfDisposed();
+                    const [r, a] = ZF(t, n);
+                    this.setOutputMatrixTextureDriver(e, r, a)
+                }
+                setOutputMatrixWriteRegion(e, t, n, r) {
+                    this.setOutputMatrixWriteRegionDriver(n, e, r, t)
+                }
+                setOutputPackedMatrixWriteRegion(e, t, n, r) {
+                    throw new Error("setOutputPackedMatrixWriteRegion not implemented.")
+                }
+                debugValidate() {
+                    null != this.program && hM(this.gl, this.program), NM(this.gl)
+                }
+                executeProgram() {
+                    this.throwIfDisposed(), this.throwIfNoProgram();
+                    const e = this.gl;
+                    this.debug && this.debugValidate(), eM(e, () => e.drawElements(e.TRIANGLES, 6, e.UNSIGNED_SHORT, 0))
+                }
+                blockUntilAllProgramsCompleted() {
+                    this.throwIfDisposed(), eM(this.gl, () => this.gl.finish())
+                }
+                getQueryTimerExtension() {
+                    return null == this.disjointQueryTimerExtension && (this.disjointQueryTimerExtension = rM(this.gl, 2 === Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") ? "EXT_disjoint_timer_query_webgl2" : "EXT_disjoint_timer_query")), this.disjointQueryTimerExtension
+                }
+                getQueryTimerExtensionWebGL2() {
+                    return this.getQueryTimerExtension()
+                }
+                getQueryTimerExtensionWebGL1() {
+                    return this.getQueryTimerExtension()
+                }
+                beginQuery() {
+                    if (2 === Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")) {
+                        const e = this.gl,
+                            t = this.getQueryTimerExtensionWebGL2(),
+                            n = e.createQuery();
+                        return e.beginQuery(t.TIME_ELAPSED_EXT, n), n
+                    }
+                    const e = this.getQueryTimerExtensionWebGL1(),
+                        t = e.createQueryEXT();
+                    return e.beginQueryEXT(e.TIME_ELAPSED_EXT, t), t
+                }
+                endQuery() {
+                    if (2 === Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")) {
+                        const e = this.gl,
+                            t = this.getQueryTimerExtensionWebGL2();
+                        return void e.endQuery(t.TIME_ELAPSED_EXT)
+                    }
+                    const e = this.getQueryTimerExtensionWebGL1();
+                    e.endQueryEXT(e.TIME_ELAPSED_EXT)
+                }
+                async waitForQueryAndGetTime(e) {
+                    return await va.repeatedTry(() => this.disposed || this.isQueryAvailable(e, Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))), this.getQueryTime(e, Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))
+                }
+                getQueryTime(e, t) {
+                    if (0 === t) return null;
+                    if (2 === t) {
+                        const t = this.gl;
+                        return t.getQueryParameter(e, t.QUERY_RESULT) / 1e6
+                    } {
+                        const t = this.getQueryTimerExtensionWebGL1();
+                        return t.getQueryObjectEXT(e, t.QUERY_RESULT_EXT) / 1e6
+                    }
+                }
+                isQueryAvailable(e, t) {
+                    if (0 === t) return !0;
+                    if (2 === t) {
+                        const t = this.gl,
+                            n = this.getQueryTimerExtensionWebGL2(),
+                            r = t.getQueryParameter(e, t.QUERY_RESULT_AVAILABLE);
+                        return null == this.disjoint && (this.disjoint = this.gl.getParameter(n.GPU_DISJOINT_EXT)), r && !this.disjoint
+                    } {
+                        const t = this.getQueryTimerExtensionWebGL1(),
+                            n = t.getQueryObjectEXT(e, t.QUERY_RESULT_AVAILABLE_EXT);
+                        return null == this.disjoint && (this.disjoint = this.gl.getParameter(t.GPU_DISJOINT_EXT)), n && !this.disjoint
+                    }
+                }
+                pollFence(e) {
+                    return new Promise(t => {
+                        this.addItemToPoll(() => e.isFencePassed(), () => t())
+                    })
+                }
+                pollItems() {
+                    const e = function (e) {
+                        let t = 0;
+                        for (; t < e.length; ++t) {
+                            if (!e[t]()) break
+                        }
+                        return t - 1
+                    }(this.itemsToPoll.map(e => e.isDoneFn));
+                    for (let t = 0; t <= e; ++t) {
+                        const {
+                            resolveFn: e
+                        } = this.itemsToPoll[t];
+                        e()
+                    }
+                    this.itemsToPoll = this.itemsToPoll.slice(e + 1)
+                }
+                addItemToPoll(e, t) {
+                    this.itemsToPoll.push({
+                        isDoneFn: e,
+                        resolveFn: t
+                    }), this.itemsToPoll.length > 1 || va.repeatedTry(() => (this.pollItems(), 0 === this.itemsToPoll.length))
+                }
+                bindTextureToFrameBuffer(e) {
+                    this.throwIfDisposed(), AM(this.gl, e, this.framebuffer), this.debug && NM(this.gl)
+                }
+                unbindTextureToFrameBuffer() {
+                    null != this.outputTexture ? (AM(this.gl, this.outputTexture, this.framebuffer), this.debug && NM(this.gl)) : EM(this.gl, this.framebuffer)
+                }
+                downloadMatrixDriver(e, t) {
+                    this.bindTextureToFrameBuffer(e);
+                    const n = t();
+                    return this.unbindTextureToFrameBuffer(), n
+                }
+                setOutputMatrixTextureDriver(e, t, n) {
+                    this.throwIfDisposed();
+                    const r = this.gl;
+                    AM(r, e, this.framebuffer), this.debug && NM(r), this.outputTexture = e, eM(r, () => r.viewport(0, 0, t, n)), eM(r, () => r.scissor(0, 0, t, n))
+                }
+                setOutputMatrixWriteRegionDriver(e, t, n, r) {
+                    this.throwIfDisposed(), eM(this.gl, () => this.gl.scissor(e, t, n, r))
+                }
+                throwIfDisposed() {
+                    if (this.disposed) throw new Error("Attempted to use disposed GPGPUContext.")
+                }
+                throwIfNoProgram() {
+                    if (null == this.program) throw new Error("No GPU program is currently set.")
+                }
+            };
+            var {
+                addImpl: VD,
+                bincountImpl: jD,
+                bincountReduceImpl: GD,
+                ceilImpl: HD,
+                concatImpl: qD,
+                equalImpl: KD,
+                expImpl: YD,
+                expm1Impl: XD,
+                floorImpl: QD,
+                gatherNdImpl: ZD,
+                gatherV2Impl: JD,
+                greaterImpl: eO,
+                greaterEqualImpl: tO,
+                lessImpl: nO,
+                lessEqualImpl: rO,
+                linSpaceImpl: aO,
+                logImpl: sO,
+                maxImpl: iO,
+                maximumImpl: oO,
+                minimumImpl: uO,
+                multiplyImpl: lO,
+                negImpl: cO,
+                notEqualImpl: hO,
+                prodImpl: pO,
+                rangeImpl: dO,
+                rsqrtImpl: fO,
+                sigmoidImpl: mO,
+                simpleAbsImpl: gO,
+                sliceImpl: yO,
+                sparseFillEmptyRowsImpl: bO,
+                sparseReshapeImpl: vO,
+                sparseSegmentReductionImpl: xO,
+                sqrtImpl: wO,
+                stridedSliceImpl: kO,
+                stringNGramsImpl: SO,
+                stringSplitImpl: IO,
+                stringToHashBucketFastImpl: AO,
+                subImpl: EO,
+                tileImpl: NO,
+                topKImpl: CO,
+                transposeImpl: TO,
+                uniqueImpl: RO
+            } = UE;
+
+            function _O(e, t) {
+                return ["x", "y", "z", "w", "u", "v"].slice(0, t).map(t => `${e}.${t}`)
+            }
+
+            function FO(e, t) {
+                return 1 === t ? [e] : _O(e, t)
+            }
+            var MO = class {
+                constructor(e, t) {
+                    this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.customUniforms = [{
+                        name: "inputShape",
+                        type: "ivec3"
+                    }], this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length);
+                    let n = "";
+                    for (let e = 0; e < 4; e++) {
+                        let t = "thisRC = rc;";
+                        e % 2 == 1 && (t += "thisRC.z += 1;"), e > 1 && (t += "thisRC.y += 1;"), n += `\n        ${t}\n        ${e>0?"if(thisRC.y < rows && thisRC.z < cols){":""}\n          int flatIndex = getFlatIndex(thisRC);\n\n          ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);\n          vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));\n\n          result[${e}] =\n            getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);\n        ${e>0?"}":""}\n      `
+                    }
+                    var r, a;
+                    this.userCode = `\n      ${r=t,a=this.enableShapeUniforms,`\n    ivec3 inputCoordsFromReshapedOutCoords(int index) {\n      ${a?JM(["r","c","d"],"inputShape"):QM(["r","c","d"],r)}\n      return ivec3(r, c, d);\n    }\n  `}\n      ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":eD(e)}\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0.);\n\n        ivec3 thisRC;\n        int rows = ${this.enableShapeUniforms?"outShape[1]":e[1]};\n        int cols = ${this.enableShapeUniforms?"outShape[2]":e[2]};\n\n        ${n}\n\n        setOutput(result);\n      }\n    `
+                }
+            };
+
+            function DO(e, t, n, r, a) {
+                const s = function (e, t) {
+                    switch (e) {
+                        case jF.PACKED_2X2_FLOAT32:
+                            return RD(t);
+                        case jF.PACKED_2X2_FLOAT16:
+                            return FD(t);
+                        case jF.UNPACKED_FLOAT32:
+                            return ID(t);
+                        case jF.UNPACKED_FLOAT16:
+                            return ED(t);
+                        case jF.PACKED_4X1_UNSIGNED_BYTE:
+                            return CD(t);
+                        default:
+                            throw new Error("Unknown physical texture type " + e)
+                    }
+                }(t, r);
+                let i;
+                if (a) {
+                    const [t, n] = ZF(e[0], e[1]);
+                    i = t * n
+                } else {
+                    const [t, n] = XF(e[0], e[1]);
+                    i = t * n
+                }
+                return i * function (e, t) {
+                    const n = e;
+                    if (t === n.R32F) return 4;
+                    if (t === n.R16F) return 2;
+                    if (t === n.RGBA32F) return 16;
+                    if (t === e.RGBA) return 16;
+                    if (t === n.RGBA16F) return 8;
+                    if (t === n.RGBA8) return 4;
+                    throw new Error("Unknown internal format " + t)
+                }(n, s)
+            }
+
+            function OO(e, t) {
+                if (e === WF.UPLOAD) return jF.PACKED_2X2_FLOAT32;
+                if (e === WF.RENDER || null == e) return function (e) {
+                    return Ye().getBool("WEBGL_RENDER_FLOAT32_ENABLED") ? e ? jF.PACKED_2X2_FLOAT32 : jF.UNPACKED_FLOAT32 : e ? jF.PACKED_2X2_FLOAT16 : jF.UNPACKED_FLOAT16
+                }(t);
+                if (e === WF.DOWNLOAD || e === WF.PIXELS) return jF.PACKED_4X1_UNSIGNED_BYTE;
+                throw new Error("Unknown logical texture type " + e)
+            }
+
+            function LO(e, t, n) {
+                return `${e[0]}_${e[1]}_${t}_${n}`
+            }
+            var PO = class {
+                    constructor(e, t) {
+                        this.variableNames = ["A"], this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length), this.userCode = `\n      float unaryOperation(float x) {\n        ${t}\n      }\n\n      void main() {\n        float x = getAAtOutCoords();\n        float y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `
+                    }
+                },
+                $O = "return abs(x);",
+                BO = "return x;",
+                zO = class {
+                    constructor(e, t) {
+                        this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length), this.userCode = `\n      vec4 unaryOperation(vec4 x) {\n        ${t}\n      }\n\n      void main() {\n        vec4 x = getAAtOutCoords();\n        vec4 y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `
+                    }
+                },
+                UO = jf.whereImpl,
+                WO = {};
+            var VO = Ye().getNumber("CPU_HANDOFF_SIZE_THRESHOLD");
+            var jO = class extends Z {
+                constructor(e) {
+                    if (super(), this.pendingRead = new WeakMap, this.pendingDisposal = new WeakSet, this.dataRefCount = new WeakMap, this.numBytesInGPU = 0, this.uploadWaitMs = 0, this.downloadWaitMs = 0, this.lastGlFlushTime = 0, this.warnedAboutMemory = !1, this.pendingDeletes = 0, this.disposed = !1, !Ye().getBool("HAS_WEBGL")) throw new Error("WebGL is not supported on this device");
+                    let t;
+                    if (null != e) {
+                        if (e instanceof WD) t = e;
+                        else {
+                            const n = YF(Ye().getNumber("WEBGL_VERSION"), e);
+                            t = new WD(n)
+                        }
+                        this.binaryCache = {}, this.gpgpuCreatedLocally = !1
+                    } else {
+                        const e = YF(Ye().getNumber("WEBGL_VERSION"));
+                        t = new WD(e), this.binaryCache = ((n = Ye().getNumber("WEBGL_VERSION")) in WO || (WO[n] = {}), WO[n]), this.gpgpuCreatedLocally = !0
+                    }
+                    var n;
+                    this.gpgpu = t, this.canvas = this.gpgpu.gl.canvas, this.textureManager = new class {
+                        constructor(e) {
+                            this.gpgpu = e, this.numUsedTextures = 0, this.numFreeTextures = 0, this._numBytesAllocated = 0, this._numBytesFree = 0, this.freeTextures = {}, this.logEnabled = !1, this.usedTextures = {}
+                        }
+                        acquireTexture(e, t, n) {
+                            const r = OO(t, n),
+                                a = LO(e, r, n);
+                            a in this.freeTextures || (this.freeTextures[a] = []), a in this.usedTextures || (this.usedTextures[a] = []);
+                            const s = DO(e, r, this.gpgpu.gl, this.gpgpu.textureConfig, n);
+                            if (this.freeTextures[a].length > 0) {
+                                this.numFreeTextures--, this.numUsedTextures++, this._numBytesFree -= s, this.log();
+                                const e = this.freeTextures[a].shift();
+                                return this.usedTextures[a].push(e), e
+                            }
+                            let i;
+                            return r === jF.PACKED_2X2_FLOAT32 ? i = this.gpgpu.createPackedMatrixTexture(e[0], e[1]) : r === jF.PACKED_2X2_FLOAT16 ? i = this.gpgpu.createFloat16PackedMatrixTexture(e[0], e[1]) : r === jF.UNPACKED_FLOAT32 ? i = this.gpgpu.createFloat32MatrixTexture(e[0], e[1]) : r === jF.UNPACKED_FLOAT16 ? i = this.gpgpu.createFloat16MatrixTexture(e[0], e[1]) : r === jF.PACKED_4X1_UNSIGNED_BYTE && (i = this.gpgpu.createUnsignedBytesMatrixTexture(e[0], e[1])), this.usedTextures[a].push(i), this.numUsedTextures++, this._numBytesAllocated += s, this.log(), i
+                        }
+                        releaseTexture(e, t, n, r) {
+                            if (null == this.freeTextures) return;
+                            const a = OO(n, r),
+                                s = LO(t, a, r);
+                            s in this.freeTextures || (this.freeTextures[s] = []);
+                            const i = DO(t, a, this.gpgpu.gl, this.gpgpu.textureConfig, r),
+                                o = Ye().get("WEBGL_DELETE_TEXTURE_THRESHOLD"); - 1 !== o && this._numBytesAllocated > o ? (this.gpgpu.deleteMatrixTexture(e.texture), this._numBytesAllocated -= i) : (this.freeTextures[s].push(e), this.numFreeTextures++, this._numBytesFree += i), this.numUsedTextures--;
+                            const u = this.usedTextures[s],
+                                l = u.indexOf(e);
+                            if (l < 0) throw new Error("Cannot release a texture that was never provided by this texture manager");
+                            u.splice(l, 1), this.log()
+                        }
+                        log() {
+                            if (!this.logEnabled) return;
+                            const e = this.numFreeTextures + this.numUsedTextures;
+                            console.log("Free/Used", `${this.numFreeTextures} / ${this.numUsedTextures}`, `(${e})`);
+                            const t = this._numBytesFree / this._numBytesAllocated;
+                            console.log("Bytes allocated: " + this._numBytesAllocated), console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)
+                        }
+                        get numBytesAllocated() {
+                            return this._numBytesAllocated
+                        }
+                        get numBytesFree() {
+                            return this._numBytesFree
+                        }
+                        getNumUsedTextures() {
+                            return this.numUsedTextures
+                        }
+                        getNumFreeTextures() {
+                            return this.numFreeTextures
+                        }
+                        dispose() {
+                            if (null != this.freeTextures) {
+                                for (const e in this.freeTextures) this.freeTextures[e].forEach(e => {
+                                    this.gpgpu.deleteMatrixTexture(e.texture)
+                                });
+                                for (const e in this.usedTextures) this.usedTextures[e].forEach(e => {
+                                    this.gpgpu.deleteMatrixTexture(e.texture)
+                                });
+                                this.freeTextures = null, this.usedTextures = null, this.numUsedTextures = 0, this.numFreeTextures = 0, this._numBytesAllocated = 0, this._numBytesFree = 0
+                            }
+                        }
+                    }(this.gpgpu), this.numMBBeforeWarning = null == Ye().global.screen ? 1024 : Ye().global.screen.height * Ye().global.screen.width * window.devicePixelRatio * 600 / 1024 / 1024, this.texData = new Q(this, eu())
+                }
+                nextDataId() {
+                    return jO.nextDataId++
+                }
+                numDataIds() {
+                    return this.texData.numDataIds() - this.pendingDeletes
+                }
+                write(e, t, n) {
+                    if ((Ye().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS") || Ye().getBool("DEBUG")) && this.checkNumericalProblems(e), "complex64" === n && null != e) throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");
+                    const r = {
+                        id: this.nextDataId()
+                    };
+                    return this.texData.set(r, {
+                        shape: t,
+                        dtype: n,
+                        values: e,
+                        usage: WF.UPLOAD,
+                        refCount: 1
+                    }), r
+                }
+                refCount(e) {
+                    if (this.texData.has(e)) {
+                        return this.texData.get(e).refCount
+                    }
+                    return 0
+                }
+                incRef(e) {
+                    this.texData.get(e).refCount++
+                }
+                decRef(e) {
+                    if (this.texData.has(e)) {
+                        this.texData.get(e).refCount--
+                    }
+                }
+                move(e, t, n, r, a) {
+                    if (Ye().getBool("DEBUG") && this.checkNumericalProblems(t), "complex64" === r) throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");
+                    this.texData.set(e, {
+                        shape: n,
+                        dtype: r,
+                        values: t,
+                        usage: WF.UPLOAD,
+                        refCount: a
+                    })
+                }
+                disposeIntermediateTensorInfo(e) {
+                    this.disposeData(e.dataId)
+                }
+                readSync(e) {
+                    const t = this.texData.get(e),
+                        {
+                            values: n,
+                            dtype: r,
+                            complexTensorInfos: a,
+                            slice: s,
+                            shape: i,
+                            isPacked: o
+                        } = t;
+                    if (null != s) {
+                        let t;
+                        t = o ? new zO(i, BO) : new PO(i, BO);
+                        const n = this.runWebGLProgram(t, [{
+                                dataId: e,
+                                shape: i,
+                                dtype: r
+                            }], r),
+                            a = this.readSync(n.dataId);
+                        return this.disposeIntermediateTensorInfo(n), a
+                    }
+                    if (null != n) return this.convertAndCacheOnCPU(e);
+                    if ("string" === r) return n;
+                    const u = null != this.activeTimers;
+                    let l, c;
+                    if (u && (l = va.now()), "complex64" === r) {
+                        const e = this.readSync(a.real.dataId),
+                            t = this.readSync(a.imag.dataId);
+                        c = jd.mergeRealAndImagArrays(e, t)
+                    } else c = this.getValuesFromTexture(e);
+                    return u && (this.downloadWaitMs += va.now() - l), this.convertAndCacheOnCPU(e, c)
+                }
+                async read(e) {
+                    if (this.pendingRead.has(e)) {
+                        const t = this.pendingRead.get(e);
+                        return new Promise(e => t.push(e))
+                    }
+                    const t = this.texData.get(e),
+                        {
+                            values: n,
+                            shape: r,
+                            slice: a,
+                            dtype: s,
+                            complexTensorInfos: i,
+                            isPacked: o
+                        } = t;
+                    if (null != a) {
+                        let t;
+                        t = o ? new zO(r, BO) : new PO(r, BO);
+                        const n = this.runWebGLProgram(t, [{
+                                dataId: e,
+                                shape: r,
+                                dtype: s
+                            }], s),
+                            a = this.read(n.dataId);
+                        return this.disposeIntermediateTensorInfo(n), a
+                    }
+                    if (null != n) return this.convertAndCacheOnCPU(e);
+                    if (Ye().getBool("DEBUG") && !Ye().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED") && 2 === Ye().getNumber("WEBGL_VERSION")) throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");
+                    let u, l, c = null;
+                    if ("complex64" !== s && Ye().get("WEBGL_BUFFER_SUPPORTED")) {
+                        u = this.decode(e);
+                        const t = this.texData.get(u.dataId);
+                        c = this.gpgpu.createBufferFromTexture(t.texture.texture, ...QF(r))
+                    }
+                    if (this.pendingRead.set(e, []), "complex64" !== s && await this.gpgpu.createAndWaitForFence(), "complex64" === s) {
+                        const e = await Promise.all([this.read(i.real.dataId), this.read(i.imag.dataId)]),
+                            t = e[0],
+                            n = e[1];
+                        l = jd.mergeRealAndImagArrays(t, n)
+                    } else if (null == c) l = this.getValuesFromTexture(e);
+                    else {
+                        const e = va.sizeFromShape(r);
+                        l = this.gpgpu.downloadFloat32MatrixFromBuffer(c, e)
+                    }
+                    if (null != u && this.disposeIntermediateTensorInfo(u), null != c) {
+                        const e = this.gpgpu.gl;
+                        eM(e, () => e.deleteBuffer(c))
+                    }
+                    const h = this.convertAndCacheOnCPU(e, l),
+                        p = this.pendingRead.get(e);
+                    return this.pendingRead.delete(e), p.forEach(e => e(h)), this.pendingDisposal.has(e) && (this.pendingDisposal.delete(e), this.disposeData(e) && eu().removeDataId(e, this), this.pendingDeletes--), h
+                }
+                readToGPU(e, t = {}) {
+                    const n = this.texData.get(e),
+                        {
+                            values: r,
+                            shape: a,
+                            slice: s,
+                            dtype: i,
+                            isPacked: o,
+                            texture: u
+                        } = n;
+                    if ("complex64" === i) throw new Error("Does not support reading texture for complex64 dtype.");
+                    if (null != s) {
+                        let n;
+                        n = o ? new zO(a, BO) : new PO(a, BO);
+                        const r = this.runWebGLProgram(n, [{
+                                dataId: e,
+                                shape: a,
+                                dtype: i
+                            }], i),
+                            s = this.readToGPU(r, t);
+                        return this.disposeIntermediateTensorInfo(r), s
+                    }
+                    if (null == u) throw null != r ? new Error("Data is not on GPU but on CPU.") : new Error("There is no data on GPU or CPU.");
+                    const l = this.decode(e, t.customTexShape),
+                        c = eu().makeTensorFromDataId(l.dataId, l.shape, l.dtype),
+                        h = this.texData.get(l.dataId);
+                    return Object.assign({
+                        tensorRef: c
+                    }, h.texture)
+                }
+                bufferSync(e) {
+                    const t = this.readSync(e.dataId);
+                    let n = t;
+                    if ("string" === e.dtype) try {
+                        n = t.map(e => va.decodeString(e))
+                    } catch (e) {
+                        throw new Error("Failed to decode encoded string bytes into utf-8")
+                    }
+                    return Ii(e.shape, e.dtype, n)
+                }
+                checkNumericalProblems(e) {
+                    if (null != e)
+                        for (let t = 0; t < e.length; t++) {
+                            const n = e[t];
+                            if (!tM(n)) {
+                                if (Ye().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")) throw Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`);
+                                throw Error(`The value ${n} cannot be represented on this device.`)
+                            }
+                        }
+                }
+                getValuesFromTexture(e) {
+                    const {
+                        shape: t,
+                        dtype: n,
+                        isPacked: r
+                    } = this.texData.get(e), a = va.sizeFromShape(t);
+                    if (Ye().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")) {
+                        const n = this.decode(e),
+                            r = this.texData.get(n.dataId),
+                            s = this.gpgpu.downloadMatrixFromPackedTexture(r.texture.texture, ...QF(t)).subarray(0, a);
+                        return this.disposeIntermediateTensorInfo(n), s
+                    }
+                    const s = Ye().getBool("WEBGL_PACK") && !0 === r,
+                        i = s ? MM(t) : t,
+                        o = s ? new class {
+                            constructor(e) {
+                                this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !1, this.outTexUsage = WF.DOWNLOAD;
+                                const t = XM();
+                                this.outputShape = e, this.userCode = `\n      ${tD}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));\n        ${t.output} = encode_float(x);\n      }\n    `
+                            }
+                        }(i) : new class {
+                            constructor(e) {
+                                this.variableNames = ["A"], this.outTexUsage = WF.DOWNLOAD;
+                                const t = XM();
+                                this.outputShape = e, this.userCode = `\n      ${tD}\n\n      void main() {\n        float x = getAAtOutCoords();\n        ${t.output} = encode_float(x);\n      }\n    `
+                            }
+                        }(i),
+                        u = this.runWebGLProgram(o, [{
+                            shape: i,
+                            dtype: n,
+                            dataId: e
+                        }], "float32"),
+                        l = this.texData.get(u.dataId),
+                        c = this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(l.texture.texture, l.texShape[0], l.texShape[1]).subarray(0, a);
+                    return this.disposeIntermediateTensorInfo(u), c
+                }
+                timerAvailable() {
+                    return Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0
+                }
+                time(e) {
+                    const t = this.activeTimers,
+                        n = [];
+                    let r = !1;
+                    null == this.programTimersStack ? (this.programTimersStack = n, r = !0) : this.activeTimers.push(n), this.activeTimers = n, e();
+                    const a = va.flatten(this.activeTimers.map(e => e.query)).filter(e => null != e),
+                        s = va.flatten(this.activeTimers.map(e => e.name)).filter(e => null != e);
+                    this.activeTimers = t, r && (this.programTimersStack = null);
+                    const i = {
+                        uploadWaitMs: this.uploadWaitMs,
+                        downloadWaitMs: this.downloadWaitMs,
+                        kernelMs: null,
+                        wallMs: null
+                    };
+                    return (async () => {
+                        if (Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+                            const e = await Promise.all(a);
+                            i.kernelMs = va.sum(e), i.getExtraProfileInfo = () => e.map((e, t) => ({
+                                name: s[t],
+                                ms: e
+                            })).map(e => `${e.name}: ${e.ms}`).join(", ")
+                        } else i.kernelMs = {
+                            error: "WebGL query timers are not supported in this environment."
+                        };
+                        return this.uploadWaitMs = 0, this.downloadWaitMs = 0, i
+                    })()
+                }
+                memory() {
+                    return {
+                        unreliable: !1,
+                        numBytesInGPU: this.numBytesInGPU,
+                        numBytesInGPUAllocated: this.textureManager.numBytesAllocated,
+                        numBytesInGPUFree: this.textureManager.numBytesFree
+                    }
+                }
+                startTimer() {
+                    return Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0 ? this.gpgpu.beginQuery() : {
+                        startMs: va.now(),
+                        endMs: null
+                    }
+                }
+                endTimer(e) {
+                    return Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0 ? (this.gpgpu.endQuery(), e) : (e.endMs = va.now(), e)
+                }
+                async getQueryTime(e) {
+                    if (Ye().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) return this.gpgpu.waitForQueryAndGetTime(e);
+                    const t = e;
+                    return t.endMs - t.startMs
+                }
+                disposeData(e, t = !1) {
+                    if (this.pendingDisposal.has(e)) return !1;
+                    if (!this.texData.has(e)) return !0;
+                    if (t ? this.texData.get(e).refCount = 0 : this.texData.get(e).refCount--, !t && this.texData.get(e).refCount > 0) return !1;
+                    if (this.pendingRead.has(e)) return this.pendingDisposal.add(e), this.pendingDeletes++, !1;
+                    this.releaseGPUData(e);
+                    const {
+                        complexTensorInfos: n
+                    } = this.texData.get(e);
+                    return null != n && (this.disposeData(n.real.dataId, t), this.disposeData(n.imag.dataId, t)), this.texData.delete(e), !0
+                }
+                releaseGPUData(e) {
+                    const {
+                        texture: t,
+                        dtype: n,
+                        texShape: r,
+                        usage: a,
+                        isPacked: s,
+                        slice: i
+                    } = this.texData.get(e), o = i && i.origDataId || e, u = this.dataRefCount.get(o);
+                    u > 1 ? this.dataRefCount.set(o, u - 1) : (this.dataRefCount.delete(o), null != t && (this.numBytesInGPU -= this.computeBytes(r, n), this.textureManager.releaseTexture(t, r, a, s)));
+                    const l = this.texData.get(e);
+                    l.texture = null, l.texShape = null, l.isPacked = !1, l.slice = null
+                }
+                getTexture(e) {
+                    return this.uploadToGPU(e), this.texData.get(e).texture.texture
+                }
+                getDataInfo(e) {
+                    return this.texData.get(e)
+                }
+                shouldExecuteOnCPU(e, t = VO) {
+                    return Ye().getBool("WEBGL_CPU_FORWARD") && e.every(e => null == this.texData.get(e.dataId).texture && va.sizeFromShape(e.shape) < t)
+                }
+                getGPGPUContext() {
+                    return this.gpgpu
+                }
+                where(e) {
+                    jd.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");
+                    const t = e.dataSync();
+                    return UO(e.shape, t)
+                }
+                packedUnaryOp(e, t, n) {
+                    const r = new zO(e.shape, t),
+                        a = this.compileAndRun(r, [e], n);
+                    return eu().makeTensorFromDataId(a.dataId, a.shape, a.dtype)
+                }
+                abs(e) {
+                    if (this.shouldExecuteOnCPU([e]) && "complex64" !== e.dtype) {
+                        const t = gO(this.texData.get(e.dataId).values);
+                        return this.makeOutput(e.shape, e.dtype, t)
+                    }
+                    if (Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS")) return this.packedUnaryOp(e, $O, e.dtype);
+                    const t = new PO(e.shape, $O),
+                        n = this.compileAndRun(t, [e]);
+                    return eu().makeTensorFromDataId(n.dataId, n.shape, n.dtype)
+                }
+                makeTensorInfo(e, t, n) {
+                    let r;
+                    if ("string" === t && null != n && n.length > 0 && va.isString(n[0])) {
+                        const a = n.map(e => va.encodeString(e));
+                        r = this.write(a, e, t)
+                    } else r = this.write(n, e, t);
+                    return this.texData.get(r).usage = null, {
+                        dataId: r,
+                        shape: e,
+                        dtype: t
+                    }
+                }
+                makeOutput(e, t, n) {
+                    const {
+                        dataId: r
+                    } = this.makeTensorInfo(e, t, n);
+                    return eu().makeTensorFromDataId(r, e, t, this)
+                }
+                unpackTensor(e) {
+                    const t = new class {
+                        constructor(e) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !1, this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length);
+                            const t = e.length,
+                                n = FO("rc", t),
+                                r = dD(t),
+                                a = function (e, t) {
+                                    if (1 === e) return "rc";
+                                    let n = "";
+                                    for (let r = 0; r < e; r++) n += t[r], r < e - 1 && (n += ",");
+                                    return n
+                                }(t, n),
+                                s = n.slice(-2),
+                                i = t <= 1 ? "rc" : `vec2(${s.join(",")})`;
+                            this.userCode = `\n      void main() {\n        ${r} rc = getOutputCoords();\n        vec4 packedInput = getA(${a});\n\n        setOutput(getChannel(packedInput, ${i}));\n      }\n    `
+                        }
+                    }(e.shape);
+                    return this.runWebGLProgram(t, [e], e.dtype)
+                }
+                packTensor(e) {
+                    const t = new class {
+                        constructor(e) {
+                            if (this.variableNames = ["A"], this.packedInputs = !1, this.packedOutput = !0, this.outputShape = e, this.rank = e.length, this.enableShapeUniforms = bD(this.outputShape.length), 0 === this.rank) this.userCode = "\n        void main() {\n          setOutput(vec4(getA(), 0., 0., 0.));\n        }\n      ";
+                            else {
+                                const e = FO("rc", this.rank),
+                                    t = dD(this.rank),
+                                    n = this.getOutOfBoundsCondition(e),
+                                    r = this.getSetup(e),
+                                    a = this.getOutput(e);
+                                this.userCode = `\n        void main() {\n          ${t} rc = getOutputCoords();\n\n          if(${n}) {\n            setOutput(vec4(0));\n          } else {\n            ${r}\n\n            setOutput(vec4(${a}));\n          }\n        }\n      `
+                            }
+                        }
+                        getSourceCoordsArr(e) {
+                            const t = [];
+                            for (let n = 0; n <= 1; n++)
+                                for (let r = 0; r <= 1; r++) {
+                                    let a = `${0===n?"r":"rp1"}, ${0===r?"c":"cp1"}`;
+                                    for (let t = 2; t < this.rank; t++) a = e[e.length - 1 - t] + "," + a;
+                                    t.push(a)
+                                }
+                            return t
+                        }
+                        getOutOfBoundsCondition(e) {
+                            if (1 === this.rank) return "rc > " + (this.enableShapeUniforms ? "outShape" : this.outputShape[0]);
+                            let t = "";
+                            for (let n = this.rank - 2; n < this.rank; n++) t += `${e[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`, n < this.rank - 1 && (t += "||");
+                            return t
+                        }
+                        getSetup(e) {
+                            if (1 === this.rank) return "";
+                            const t = e.slice(-2),
+                                n = this.enableShapeUniforms ? `outShape[${this.rank} - 1]` : this.outputShape[this.rank - 1],
+                                r = this.enableShapeUniforms ? `outShape[${this.rank} - 2]` : this.outputShape[this.rank - 2];
+                            return `\n      int r = ${t[0]};\n      int c = ${t[1]};\n      int rp1 = r + 1;\n      int cp1 = c + 1;\n\n      bool cEdge = cp1 >= ${n};\n      bool rEdge = rp1 >= ${r};\n    `
+                        }
+                        getOutput(e) {
+                            const t = this.getSourceCoordsArr(e);
+                            if (1 === this.rank) {
+                                return `getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`
+                            }
+                            return `getA(${t[0]}),\n            cEdge ? 0. : getA(${t[1]}),\n            rEdge ? 0. : getA(${t[2]}),\n            rEdge || cEdge ? 0. : getA(${t[3]})`
+                        }
+                    }(e.shape);
+                    return this.runWebGLProgram(t, [e], e.dtype, null, !0)
+                }
+                packedReshape(e, t) {
+                    const n = [_M(e.shape), ...FM(e.shape)],
+                        r = {
+                            dtype: e.dtype,
+                            shape: n,
+                            dataId: e.dataId
+                        },
+                        a = [_M(t), ...FM(t)],
+                        s = new MO(a, n),
+                        i = [n],
+                        o = this.runWebGLProgram(s, [r], e.dtype, i, !0);
+                    return {
+                        dataId: o.dataId,
+                        shape: t,
+                        dtype: o.dtype
+                    }
+                }
+                decode(e, t) {
+                    const n = this.texData.get(e),
+                        {
+                            isPacked: r,
+                            shape: a,
+                            dtype: s
+                        } = n;
+                    if (null != t) {
+                        const e = va.sizeFromShape(a),
+                            n = t[0] * t[1] * 4;
+                        va.assert(e <= n, () => "customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")
+                    }
+                    const i = MM(a);
+                    let o;
+                    o = r ? new class {
+                        constructor(e) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.outPackingScheme = zF.DENSE, this.customUniforms = [{
+                                name: "texShape",
+                                type: "ivec2"
+                            }];
+                            const t = XM();
+                            this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length), this.userCode = `\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?ZM(["r","c","d"],e):QM(["r","c","d"],e)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));\n        }\n\n        ${t.output} = result;\n      }\n    `
+                        }
+                    }(i) : new class {
+                        constructor(e) {
+                            this.variableNames = ["A"], this.packedInputs = !1, this.packedOutput = !0, this.outPackingScheme = zF.DENSE, this.customUniforms = [{
+                                name: "texShape",
+                                type: "ivec2"
+                            }];
+                            const t = XM();
+                            this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length), this.userCode = `\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?ZM(["r","c","d"],e):QM(["r","c","d"],e)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getA(rc.x, rc.y, rc.z);\n        }\n\n        ${t.output} = result;\n      }\n    `
+                        }
+                    }(i);
+                    const u = [null != t ? t : QF(i)];
+                    return {
+                        dtype: s,
+                        shape: a,
+                        dataId: this.runWebGLProgram(o, [{
+                            shape: i,
+                            dtype: s,
+                            dataId: e
+                        }], s, u, !0, t).dataId
+                    }
+                }
+                runWebGLProgram(e, t, n, r, a = !1, s) {
+                    const i = this.makeTensorInfo(e.outputShape, n),
+                        o = this.texData.get(i.dataId);
+                    if (e.packedOutput && (o.isPacked = !0), e.outPackingScheme === zF.DENSE) {
+                        const t = null != s ? s : QF(e.outputShape);
+                        o.texShape = t.map(e => 2 * e)
+                    }
+                    if (null != e.outTexUsage && (o.usage = e.outTexUsage), 0 === va.sizeFromShape(i.shape)) return o.values = va.getTypedArrayFromDType(i.dtype, 0), i;
+                    const u = [],
+                        l = t.map(t => {
+                            if ("complex64" === t.dtype) throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");
+                            let n = this.texData.get(t.dataId);
+                            if (null == n.texture) {
+                                if (!e.packedInputs && va.sizeFromShape(t.shape) <= Ye().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM")) return {
+                                    shape: t.shape,
+                                    texData: null,
+                                    isUniform: !0,
+                                    uniformValues: n.values
+                                };
+                                e.packedInputs && (n.isPacked = !0, n.shape = t.shape)
+                            }
+                            if (this.uploadToGPU(t.dataId), !!n.isPacked != !!e.packedInputs) t = n.isPacked ? this.unpackTensor(t) : this.packTensor(t), u.push(t), n = this.texData.get(t.dataId);
+                            else if (n.isPacked && !LM(n.shape, t.shape)) {
+                                const e = t,
+                                    r = t.shape;
+                                t.shape = n.shape, t = this.packedReshape(t, r), u.push(t), n = this.texData.get(t.dataId), e.shape = r
+                            }
+                            return {
+                                shape: t.shape,
+                                texData: n,
+                                isUniform: !1
+                            }
+                        });
+                    this.uploadToGPU(i.dataId);
+                    const c = {
+                            shape: i.shape,
+                            texData: o,
+                            isUniform: !1
+                        },
+                        h = function (e, t, n) {
+                            let r = "";
+                            t.concat(n).forEach(t => {
+                                const a = null != t.texData && null != t.texData.slice && t.texData.slice.flatOffset > 0;
+                                if (e.enableShapeUniforms && !t.isUniform) {
+                                    const s = t.texData.texShape,
+                                        {
+                                            useSqueezeShape: i,
+                                            uniformShape: o,
+                                            keptDims: u
+                                        } = fD(e.packedInputs, t.shape, s);
+                                    let l = "",
+                                        c = "",
+                                        h = "";
+                                    if (1 === o.length && e.packedInputs) {
+                                        const e = [Math.ceil(s[0] / 2), Math.ceil(s[1] / 2)];
+                                        l = `${e[0]>1}_${e[1]>1}`
+                                    } else if (2 !== o.length || e.packedInputs) {
+                                        if (o.length > 2 && !e.packedInputs) {
+                                            const e = va.computeStrides(o);
+                                            h = `${e[0]===s[1]}_${e[e.length-1]===s[1]}`
+                                        }
+                                    } else c = `${o[0]>1}_${o[1]>1}`;
+                                    const p = t.shape.length,
+                                        d = 2 === o.length && va.arraysEqual(t.shape, s),
+                                        f = 1 === va.sizeFromShape(t.shape),
+                                        m = jd.getBroadcastDims(t.shape, n.shape),
+                                        g = !e.packedInputs && p === n.shape.length && va.arraysEqual(s, n.texData.texShape),
+                                        y = e.packedInputs || o.length > 2 ? "" : `${s[0]>1}_${s[1]>1}`;
+                                    r += `${p}_${g}_${i?u:""}_${o.length}_${f}_${m}_${d}_${l}_${c}_${h}_${y}_${a}`
+                                } else {
+                                    const e = t.isUniform ? "uniform" : t.texData.texShape;
+                                    r += `${t.shape}_${e}_${a}`
+                                }
+                            });
+                            const a = e.userCode;
+                            let s = e.constructor.name;
+                            return s += "_" + r + "_" + a + Ye().getNumber("WEBGL_VERSION"), s
+                        }(e, l, c),
+                        p = this.getAndSaveBinary(h, () => function (e, t, n, r) {
+                            const a = n.map((e, n) => {
+                                    const r = {
+                                        logicalShape: e.shape,
+                                        texShape: e.isUniform ? null : e.texData.texShape,
+                                        isUniform: e.isUniform,
+                                        isPacked: !e.isUniform && e.texData.isPacked,
+                                        flatOffset: null
+                                    };
+                                    return null != e.texData && null != e.texData.slice && e.texData.slice.flatOffset > 0 && (r.flatOffset = e.texData.slice.flatOffset), {
+                                        name: t.variableNames[n],
+                                        shapeInfo: r
+                                    }
+                                }),
+                                s = a.map(e => e.shapeInfo),
+                                i = {
+                                    logicalShape: r.shape,
+                                    texShape: r.texData.texShape,
+                                    isUniform: !1,
+                                    isPacked: r.texData.isPacked,
+                                    flatOffset: null
+                                },
+                                o = rD(a, i, t),
+                                u = sM(e.gl, o),
+                                l = e.createProgram(u);
+                            let c = null;
+                            const h = e.getUniformLocation(l, "NAN", !1);
+                            1 === Ye().getNumber("WEBGL_VERSION") && (c = e.getUniformLocation(l, "INFINITY", !1));
+                            const p = {},
+                                d = {},
+                                f = {};
+                            for (let n = 0; n < t.variableNames.length; n++) {
+                                const r = t.variableNames[n];
+                                p[r] = e.getUniformLocation(l, r, !1), p["offset" + r] = e.getUniformLocation(l, "offset" + r, !1), t.enableShapeUniforms && (d[r + "Shape"] = e.getUniformLocation(l, r + "Shape", !1), f[r + "TexShape"] = e.getUniformLocation(l, r + "TexShape", !1))
+                            }
+                            let m, g, y;
+                            t.enableShapeUniforms && (m = e.getUniformLocation(l, "outShape", !1), y = e.getUniformLocation(l, "outShapeStrides", !1), g = e.getUniformLocation(l, "outTexShape", !1));
+                            const b = [];
+                            return t.customUniforms && t.customUniforms.forEach((t, n) => {
+                                b[n] = e.getUniformLocation(l, t.name, !1)
+                            }), {
+                                program: t,
+                                fragmentShader: u,
+                                source: o,
+                                webGLProgram: l,
+                                uniformLocations: p,
+                                customUniformLocations: b,
+                                inShapeInfos: s,
+                                outShapeInfo: i,
+                                infLoc: c,
+                                nanLoc: h,
+                                inShapesLocations: d,
+                                inTexShapesLocations: f,
+                                outShapeLocation: m,
+                                outShapeStridesLocation: y,
+                                outTexShapeLocation: g
+                            }
+                        }(this.gpgpu, e, l, c)),
+                        d = null != this.activeTimers;
+                    let f;
+                    d && (f = this.startTimer()),
+                        function (e, t, n, r, a) {
+                            t.program.enableShapeUniforms || (yD(t.inShapeInfos, n), yD([t.outShapeInfo], [r]));
+                            const s = r.texData.texture,
+                                i = r.texData.texShape;
+                            r.texData.isPacked ? e.setOutputPackedMatrixTexture(s.texture, i[0], i[1]) : e.setOutputMatrixTexture(s.texture, i[0], i[1]), e.setProgram(t.webGLProgram), 1 === Ye().getNumber("WEBGL_VERSION") && null !== t.infLoc && e.gl.uniform1f(t.infLoc, 1 / 0), null !== t.nanLoc && e.gl.uniform1f(t.nanLoc, NaN), n.forEach((n, r) => {
+                                const a = t.program.variableNames[r],
+                                    s = t.uniformLocations[a],
+                                    i = t.uniformLocations["offset" + a],
+                                    o = t.inShapesLocations[a + "Shape"],
+                                    u = t.inTexShapesLocations[a + "TexShape"];
+                                if (o) {
+                                    const {
+                                        uniformShape: r
+                                    } = fD(t.program.packedInputs, n.shape, n.texData.texShape);
+                                    switch (r.length) {
+                                        case 1:
+                                            e.gl.uniform1iv(o, new Int32Array(r));
+                                            break;
+                                        case 2:
+                                            e.gl.uniform2iv(o, new Int32Array(r));
+                                            break;
+                                        case 3:
+                                            e.gl.uniform3iv(o, new Int32Array(r));
+                                            break;
+                                        case 4:
+                                            e.gl.uniform4iv(o, new Int32Array(r))
+                                    }
+                                }
+                                if (u && e.gl.uniform2i(u, n.texData.texShape[0], n.texData.texShape[1]), null != s)
+                                    if (n.isUniform)
+                                        if (va.sizeFromShape(n.shape) < 2) e.gl.uniform1f(s, n.uniformValues[0]);
+                                        else {
+                                            let t = n.uniformValues;
+                                            t instanceof Float32Array || (t = new Float32Array(t)), e.gl.uniform1fv(s, t)
+                                        }
+                                else null != n.texData.slice && null != i && e.gl.uniform1i(i, n.texData.slice.flatOffset), e.setInputMatrixTexture(n.texData.texture.texture, s, r)
+                            });
+                            const o = t.outShapeLocation;
+                            if (o) switch (r.shape.length) {
+                                case 1:
+                                    e.gl.uniform1iv(o, new Int32Array(r.shape));
+                                    break;
+                                case 2:
+                                    e.gl.uniform2iv(o, new Int32Array(r.shape));
+                                    break;
+                                case 3:
+                                    e.gl.uniform3iv(o, new Int32Array(r.shape));
+                                    break;
+                                case 4:
+                                    e.gl.uniform4iv(o, new Int32Array(r.shape))
+                            }
+                            if (t.outShapeStridesLocation) {
+                                const n = va.computeStrides(r.shape);
+                                switch (r.shape.length) {
+                                    case 2:
+                                        e.gl.uniform1iv(t.outShapeStridesLocation, new Int32Array(n));
+                                        break;
+                                    case 3:
+                                        e.gl.uniform2iv(t.outShapeStridesLocation, new Int32Array(n));
+                                        break;
+                                    case 4:
+                                        e.gl.uniform3iv(t.outShapeStridesLocation, new Int32Array(n))
+                                }
+                            }
+                            t.outTexShapeLocation && e.gl.uniform2i(t.outTexShapeLocation, r.texData.texShape[0], r.texData.texShape[1]), t.program.customUniforms && a && t.program.customUniforms.forEach((n, r) => {
+                                const s = t.customUniformLocations[r],
+                                    i = a[r];
+                                if ("float" === n.type) e.gl.uniform1fv(s, i);
+                                else if ("vec2" === n.type) e.gl.uniform2fv(s, i);
+                                else if ("vec3" === n.type) e.gl.uniform3fv(s, i);
+                                else if ("vec4" === n.type) e.gl.uniform4fv(s, i);
+                                else if ("int" === n.type) e.gl.uniform1iv(s, i);
+                                else if ("ivec2" === n.type) e.gl.uniform2iv(s, i);
+                                else if ("ivec3" === n.type) e.gl.uniform3iv(s, i);
+                                else {
+                                    if ("ivec4" !== n.type) throw Error(`uniform type ${n.type} is not supported yet.`);
+                                    e.gl.uniform4iv(s, i)
+                                }
+                            }), e.executeProgram()
+                        }(this.gpgpu, p, l, c, r), u.forEach(e => this.disposeIntermediateTensorInfo(e)), d && (f = this.endTimer(f), this.activeTimers.push({
+                            name: e.constructor.name,
+                            query: this.getQueryTime(f)
+                        }));
+                    const m = Ye().get("WEBGL_FLUSH_THRESHOLD");
+                    if (m > 0) {
+                        const e = va.now();
+                        e - this.lastGlFlushTime > m && (this.gpgpu.gl.flush(), this.lastGlFlushTime = e)
+                    }
+                    if (!Ye().getBool("WEBGL_LAZILY_UNPACK") && o.isPacked && !1 === a) {
+                        const e = this.unpackTensor(i);
+                        return this.disposeIntermediateTensorInfo(i), e
+                    }
+                    return i
+                }
+                compileAndRun(e, t, n, r, a = !1) {
+                    n = n || t[0].dtype;
+                    return this.runWebGLProgram(e, t, n, r, a)
+                }
+                getAndSaveBinary(e, t) {
+                    return e in this.binaryCache || (this.binaryCache[e] = t()), this.binaryCache[e]
+                }
+                getTextureManager() {
+                    return this.textureManager
+                }
+                dispose() {
+                    if (!this.disposed) {
+                        if (!Ye().getBool("IS_TEST")) {
+                            Object.keys(this.binaryCache).forEach(e => {
+                                this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram), delete this.binaryCache[e]
+                            })
+                        }
+                        this.textureManager.dispose(), null != this.canvas && "undefined" != typeof HTMLCanvasElement && this.canvas instanceof HTMLCanvasElement ? this.canvas.remove() : this.canvas = null, this.gpgpuCreatedLocally && (this.gpgpu.program = null, this.gpgpu.dispose()), this.disposed = !0
+                    }
+                }
+                floatPrecision() {
+                    return null == this.floatPrecisionValue && (this.floatPrecisionValue = ru(() => {
+                        if (!Ye().get("WEBGL_RENDER_FLOAT32_ENABLED")) {
+                            const e = Ye().getBool("DEBUG");
+                            Ye().set("DEBUG", !1);
+                            const t = this.abs(Ah(1e-8)).dataSync()[0];
+                            if (Ye().set("DEBUG", e), t > 0) return 32
+                        }
+                        return 16
+                    })), this.floatPrecisionValue
+                }
+                epsilon() {
+                    return 32 === this.floatPrecision() ? 1e-7 : 1e-4
+                }
+                uploadToGPU(e) {
+                    const t = this.texData.get(e),
+                        {
+                            shape: n,
+                            dtype: r,
+                            values: a,
+                            texture: s,
+                            usage: i,
+                            isPacked: o
+                        } = t;
+                    if (null != s) return;
+                    const u = null != this.activeTimers;
+                    let l;
+                    u && (l = va.now());
+                    let c = t.texShape;
+                    if (null == c && (c = DM(n, o), t.texShape = c), null != a) {
+                        const e = MM(n);
+                        let s, i = c[1],
+                            h = c[0];
+                        const p = a instanceof Uint8Array || a instanceof Uint8ClampedArray;
+                        !o && p || ([i, h] = ZF(c[0], c[1])), s = o ? new class {
+                            constructor(e, t = !1) {
+                                this.variableNames = ["A"], this.packedInputs = !1, this.packedOutput = !0, this.customUniforms = [{
+                                    name: "texShape",
+                                    type: "ivec2"
+                                }];
+                                const n = XM();
+                                this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length);
+                                let r = "",
+                                    a = "result";
+                                t && (a = "floor(result * 255. + 0.5)");
+                                for (let t = 0; t <= 1; t++)
+                                    for (let a = 0; a <= 1; a++) {
+                                        const s = 2 * t + a;
+                                        r += `\n          localCoords = coords;\n          if(localCoords[2] + ${a} < ${this.enableShapeUniforms?"outShape[2]":""+e[2]}) {\n          localCoords[2] += ${a};\n          if (localCoords[1] + ${t} < ${this.enableShapeUniforms?"outShape[1]":""+e[1]}) {\n            localCoords[1] += ${t};\n\n            flatIndex = getFlatIndex(localCoords);\n            offset = imod(flatIndex, 4);\n\n            flatIndex = idiv(flatIndex, 4, 1.);\n\n            int r = flatIndex / texShape[1];\n            int c = imod(flatIndex, texShape[1]);\n            vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n            values = ${n.texture2D}(A, uv);\n\n            if (offset == 0) {\n              result[${s}] = values[0];\n            } else if (offset == 1) {\n              result[${s}] = values[1];\n            } else if (offset == 2) {\n              result[${s}] = values[2];\n            } else {\n              result[${s}] = values[3];\n            }\n          }\n        }\n        `
+                                    }
+                                this.userCode = `\n        ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":eD(e)}\n\n        void main() {\n          ivec3 coords = getOutputCoords();\n\n          vec4 result = vec4(0.);\n          int flatIndex, r, c, offset;\n          ivec3 localCoords;\n          vec2 uv;\n          vec4 values;\n\n          ${r}\n\n          ${n.output} = ${a};\n        }\n    `
+                            }
+                        }(e, p) : new class {
+                            constructor(e, t = !1) {
+                                this.variableNames = ["A"], this.customUniforms = [{
+                                    name: "texShape",
+                                    type: "ivec2"
+                                }];
+                                const n = XM();
+                                this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length);
+                                let r = "result";
+                                t && (r = "floor(result * 255. + 0.5)"), this.userCode = `\n      ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":eD(e)}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n\n        int flatIndex = getFlatIndex(coords);\n        int offset = imod(flatIndex, 4);\n\n        flatIndex = idiv(flatIndex, 4, 1.);\n\n        int r = flatIndex / texShape[1];\n        int c = imod(flatIndex, texShape[1]);\n        vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n        vec4 values = ${n.texture2D}(A, uv);\n\n        float result;\n\n        if(offset == 0) {\n          result = values[0];\n        } else if(offset == 1) {\n          result = values[1];\n        } else if(offset == 2) {\n          result = values[2];\n        } else {\n          result = values[3];\n        }\n\n        ${n.output} = vec4(${r}, 0., 0., 0.);\n      }\n    `
+                            }
+                        }(e, p);
+                        const d = p ? [h, i] : c,
+                            f = this.makeTensorInfo(d, r),
+                            m = this.texData.get(f.dataId);
+                        m.usage = p ? WF.PIXELS : WF.UPLOAD, m.texShape = d, this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(f.dataId), i, h, a);
+                        const g = [
+                                [h, i]
+                            ],
+                            y = !0,
+                            b = this.runWebGLProgram(s, [f], r, g, y),
+                            v = this.texData.get(b.dataId);
+                        t.texture = v.texture, t.texShape = v.texShape, t.isPacked = v.isPacked, t.usage = v.usage, this.disposeIntermediateTensorInfo(f), this.texData.delete(b.dataId), t.values = null, u && (this.uploadWaitMs += va.now() - l)
+                    } else {
+                        const e = this.acquireTexture(c, i, r, o);
+                        t.texture = e
+                    }
+                }
+                convertAndCacheOnCPU(e, t) {
+                    const n = this.texData.get(e),
+                        {
+                            dtype: r
+                        } = n;
+                    return this.releaseGPUData(e), null != t && (n.values = function (e, t) {
+                        if ("float32" === t || "complex64" === t) return e;
+                        if ("int32" === t || "bool" === t) {
+                            const n = "int32" === t ? new Int32Array(e.length) : new Uint8Array(e.length);
+                            for (let t = 0; t < n.length; ++t) n[t] = Math.round(e[t]);
+                            return n
+                        }
+                        throw new Error("Unknown dtype " + t)
+                    }(t, r)), n.values
+                }
+                acquireTexture(e, t, n, r) {
+                    if (this.numBytesInGPU += this.computeBytes(e, n), !this.warnedAboutMemory && this.numBytesInGPU > 1024 * this.numMBBeforeWarning * 1024) {
+                        const e = (this.numBytesInGPU / 1024 / 1024).toFixed(2);
+                        this.warnedAboutMemory = !0, console.warn(`High memory usage in GPU: ${e} MB, most likely due to a memory leak`)
+                    }
+                    return this.textureManager.acquireTexture(e, t, r)
+                }
+                computeBytes(e, t) {
+                    return e[0] * e[1] * va.bytesPerElement(t)
+                }
+            };
+            jO.nextDataId = 0;
+            var GO = "3.13.0";
+
+            function HO() {
+                Ye().set("WEBGL_FORCE_F16_TEXTURES", !0)
+            }
+            Ss.isBrowser() && du("webgl", () => new jO, 2);
+            var qO = {
+                    forceHalfFloat: HO
+                },
+                KO = class {
+                    constructor(e, t, n) {
+                        this.variableNames = ["A", "B"], this.outputShape = jd.assertAndGetBroadcastShape(t, n), this.enableShapeUniforms = bD(this.outputShape.length), this.userCode = `\n      float binaryOperation(float a, float b) {\n        ${e}\n      }\n\n      void main() {\n        float a = getAAtOutCoords();\n        float b = getBAtOutCoords();\n        setOutput(binaryOperation(a, b));\n      }\n    `
+                    }
+                },
+                YO = class {
+                    constructor(e, t, n, r = !1) {
+                        this.variableNames = ["A", "B"], this.supportsBroadcasting = !0, this.packedInputs = !0, this.packedOutput = !0, this.outputShape = jd.assertAndGetBroadcastShape(t, n);
+                        const a = this.outputShape.length;
+                        this.enableShapeUniforms = bD(a);
+                        let s = "";
+                        if (r)
+                            if (0 === a || 1 === va.sizeFromShape(this.outputShape)) s = "\n          result.y = 0.;\n          result.z = 0.;\n          result.w = 0.;\n        ";
+                            else {
+                                if (s = `\n          ${dD(a)} coords = getOutputCoords();\n        `, 1 === a) this.enableShapeUniforms ? s += "\n            result.y = (coords + 1) >= outShape ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          " : s += `\n            result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `;
+                                else {
+                                    const e = FO("coords", a);
+                                    this.enableShapeUniforms ? s += `\n            bool nextRowOutOfBounds =\n              (${e[a-2]} + 1) >= outShape[${a} - 2];\n            bool nextColOutOfBounds =\n              (${e[a-1]} + 1) >= outShape[${a} - 1];\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          ` : s += `\n            bool nextRowOutOfBounds =\n              (${e[a-2]} + 1) >= ${this.outputShape[a-2]};\n            bool nextColOutOfBounds =\n              (${e[a-1]} + 1) >= ${this.outputShape[a-1]};\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `
+                                }
+                            } this.userCode = `\n      vec4 binaryOperation(vec4 a, vec4 b) {\n        ${e}\n      }\n\n      void main() {\n        vec4 a = getAAtOutCoords();\n        vec4 b = getBAtOutCoords();\n\n        vec4 result = binaryOperation(a, b);\n        ${s}\n\n        setOutput(result);\n      }\n    `
+                    }
+                };
+
+            function XO(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    x: r
+                } = t;
+                return n.incRef(r.dataId), {
+                    dataId: r.dataId,
+                    shape: r.shape,
+                    dtype: r.dtype
+                }
+            }
+            var QO = {
+                kernelName: fn,
+                backendName: "webgl",
+                kernelFunc: XO
+            };
+
+            function ZO(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    real: r,
+                    imag: a
+                } = t, s = n.makeTensorInfo(r.shape, "complex64"), i = n.texData.get(s.dataId), o = XO({
+                    inputs: {
+                        x: r
+                    },
+                    backend: n
+                }), u = XO({
+                    inputs: {
+                        x: a
+                    },
+                    backend: n
+                });
+                return i.complexTensorInfos = {
+                    real: o,
+                    imag: u
+                }, s
+            }
+            var JO = {
+                    kernelName: Et,
+                    backendName: "webgl",
+                    kernelFunc: ZO
+                },
+                eL = "return (a < 0.) ? b * a : a;",
+                tL = "\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n";
+            var nL = {
+                    kernelName: xn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            alpha: s
+                        } = r, i = n.makeTensorInfo([], "float32", va.createScalarValue(s, "float32")), o = Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new YO(tL, a.shape, i.shape) : new KO(eL, a.shape, i.shape), u = n.runWebGLProgram(o, [a, i], "float32");
+                        return n.disposeIntermediateTensorInfo(i), u
+                    }
+                },
+                rL = "return (a < 0.) ? b * a : a;",
+                aL = "\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n";
+            var sL = {
+                kernelName: rr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        x: r,
+                        alpha: a
+                    } = t, s = Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new YO(aL, r.shape, a.shape) : new KO(rL, r.shape, a.shape);
+                    return n.runWebGLProgram(s, [r, a], "float32")
+                }
+            };
+
+            function iL({
+                opSnippet: e,
+                packedOpSnippet: t,
+                cpuKernelImpl: n,
+                dtype: r
+            }) {
+                return ({
+                    inputs: a,
+                    backend: s
+                }) => {
+                    const {
+                        x: i
+                    } = a, o = s, u = r || i.dtype;
+                    if (o.shouldExecuteOnCPU([i]) && null != n) {
+                        const e = o.texData.get(i.dataId),
+                            t = n(e.values, u);
+                        return o.makeTensorInfo(i.shape, u, t)
+                    }
+                    let l;
+                    return l = Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS") && null != t ? new zO(i.shape, t) : new PO(i.shape, e), o.runWebGLProgram(l, [i], u)
+                }
+            }
+
+            function oL({
+                opSnippet: e,
+                packedOpSnippet: t,
+                checkOutOfBounds: n = !1,
+                supportsComplex: r = !1,
+                cpuKernelImpl: a,
+                dtype: s
+            }) {
+                return ({
+                    inputs: i,
+                    backend: o
+                }) => {
+                    const {
+                        a: u,
+                        b: l
+                    } = i, c = o;
+                    if (r && "complex64" === u.dtype) {
+                        const t = c.texData.get(u.dataId),
+                            n = c.texData.get(l.dataId),
+                            [r, a] = [
+                                [t.complexTensorInfos.real, n.complexTensorInfos.real],
+                                [t.complexTensorInfos.imag, n.complexTensorInfos.imag]
+                            ].map(t => {
+                                const [n, r] = t, a = {
+                                    dataId: n.dataId,
+                                    dtype: n.dtype,
+                                    shape: u.shape
+                                }, s = {
+                                    dataId: r.dataId,
+                                    dtype: r.dtype,
+                                    shape: l.shape
+                                }, i = new KO(e, u.shape, l.shape);
+                                return c.runWebGLProgram(i, [a, s], cs(n.dtype, r.dtype))
+                            }),
+                            s = ZO({
+                                inputs: {
+                                    real: r,
+                                    imag: a
+                                },
+                                backend: c
+                            });
+                        return c.disposeIntermediateTensorInfo(r), c.disposeIntermediateTensorInfo(a), s
+                    }
+                    const h = s || cs(u.dtype, l.dtype);
+                    if (("string" === u.dtype || "string" === l.dtype || c.shouldExecuteOnCPU([u, l])) && null != a) {
+                        const e = c.texData.get(u.dataId).values,
+                            t = c.texData.get(l.dataId).values,
+                            n = "string" === u.dtype ? jd.fromUint8ToStringArray(e) : e,
+                            r = "string" === u.dtype ? jd.fromUint8ToStringArray(t) : t,
+                            [s, i] = a(u.shape, l.shape, n, r, h),
+                            o = c.makeTensorInfo(i, h);
+                        return c.texData.get(o.dataId).values = s, o
+                    }
+                    let p;
+                    return p = Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS") && null != t ? new YO(t, u.shape, l.shape, n) : new KO(e, u.shape, l.shape), c.runWebGLProgram(p, [u, l], h)
+                }
+            }
+
+            function uL(e, t = !1) {
+                if ("linear" === e) return "return x;";
+                if ("relu" === e) return t ? "\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n" : "if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : x;\n";
+                if ("elu" === e) return t ? "\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n" : "return (x >= 0.0) ? x : (exp(x) - 1.0);";
+                if ("relu6" === e) return t ? "\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n" : "if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n";
+                if ("prelu" === e) return t ? aL : rL;
+                if ("leakyrelu" === e) return t ? tL : eL;
+                if ("sigmoid" === e) return "return 1.0 / (1.0 + exp(-1.0 * x));";
+                throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)
+            }
+            var lL = class {
+                    constructor(e, t, n, r = !1, a = !1, s = !1, i = null, o = !1, u = !1) {
+                        this.variableNames = ["matrixA", "matrixB"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = n, this.enableShapeUniforms = bD(this.outputShape.length);
+                        const l = r ? e[1] : e[2],
+                            c = Math.ceil(l / 2),
+                            h = r ? "i * 2, rc.y" : "rc.y, i * 2",
+                            p = a ? "rc.z, i * 2" : "i * 2, rc.z",
+                            d = r ? ["a.xxyy", "a.zzww"] : ["a.xxzz", "a.yyww"],
+                            f = a ? ["b.xzxz", "b.ywyw"] : ["b.xyxy", "b.zwzw"];
+                        let m = "",
+                            g = "";
+                        i && (m = o ? `vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${i}\n        }` : u ? `vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${i}\n        }` : `vec4 activation(vec4 x) {\n          ${i}\n        }`, g = "result = activation(result);");
+                        const y = s ? "result += getBiasAtOutCoords();" : "";
+                        s && this.variableNames.push("bias"), o && this.variableNames.push("preluActivationWeights"), u && this.variableNames.push("leakyreluAlpha");
+                        let b = "rc.x",
+                            v = "rc.x";
+                        e[0] < t[0] ? b = `int(min(float(rc.x), ${e[0]-1}.))` : t[0] < e[0] && (v = `int(min(float(rc.x), ${t[0]-1}.))`), this.userCode = `\n      ${m}\n      // Don't use uniform for sharedDimensionPacked for performance.\n      const float sharedDimension = ${c}.0;\n\n      vec4 dot2x2ARowBCol(ivec3 rc) {\n        vec4 result = vec4(0);\n        for (int i = 0; i < ${c}; i++) {\n          int batchA = ${b};\n          int batchB = ${v};\n          vec4 a = getMatrixA(batchA, ${h});\n          vec4 b = getMatrixB(batchB, ${p});\n\n          // These swizzled products need to be separately added.\n          // See: https://github.com/tensorflow/tfjs/issues/1735\n          result += (${d[0]} * ${f[0]});\n          result += (${d[1]} * ${f[1]});\n        }\n        return result;\n      }\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n        vec4 result = dot2x2ARowBCol(rc);\n\n        ${y}\n\n        ${g}\n\n        setOutput(result);\n      }\n    `
+                    }
+                },
+                cL = "return areal * breal - aimag * bimag;",
+                hL = "return areal * bimag + aimag * breal;",
+                pL = class {
+                    constructor(e, t, n) {
+                        this.variableNames = ["AReal", "AImag", "BReal", "BImag"], this.outputShape = jd.assertAndGetBroadcastShape(t, n), this.userCode = `\n      float binaryOpComplex(\n          float areal, float aimag, float breal, float bimag) {\n        ${e}\n      }\n\n      void main() {\n        float areal = getARealAtOutCoords();\n        float aimag = getAImagAtOutCoords();\n        float breal = getBRealAtOutCoords();\n        float bimag = getBImagAtOutCoords();\n        setOutput(binaryOpComplex(areal, aimag, breal, bimag));\n      }\n    `
+                    }
+                },
+                dL = "return a * b;";
+
+            function fL(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    a: r,
+                    b: a
+                } = t, s = jd.upcastType(r.dtype, a.dtype);
+                if ("complex64" === r.dtype) {
+                    const e = n.texData.get(r.dataId),
+                        t = n.texData.get(a.dataId),
+                        s = new pL(cL, r.shape, a.shape),
+                        i = new pL(hL, r.shape, a.shape),
+                        o = [{
+                            dataId: e.complexTensorInfos.real.dataId,
+                            dtype: e.complexTensorInfos.real.dtype,
+                            shape: r.shape
+                        }, {
+                            dataId: e.complexTensorInfos.imag.dataId,
+                            dtype: e.complexTensorInfos.imag.dtype,
+                            shape: r.shape
+                        }, {
+                            dataId: t.complexTensorInfos.real.dataId,
+                            dtype: t.complexTensorInfos.real.dtype,
+                            shape: a.shape
+                        }, {
+                            dataId: t.complexTensorInfos.imag.dataId,
+                            dtype: t.complexTensorInfos.imag.dtype,
+                            shape: a.shape
+                        }],
+                        u = n.runWebGLProgram(s, o, "float32"),
+                        l = n.runWebGLProgram(i, o, "float32"),
+                        c = ZO({
+                            inputs: {
+                                real: u,
+                                imag: l
+                            },
+                            backend: n
+                        });
+                    return n.disposeIntermediateTensorInfo(u), n.disposeIntermediateTensorInfo(l), c
+                }
+                if (n.shouldExecuteOnCPU([r, a])) {
+                    const e = n.texData.get(r.dataId),
+                        t = n.texData.get(a.dataId),
+                        [i, o] = lO(r.shape, a.shape, e.values, t.values, s),
+                        u = n.makeTensorInfo(o, s);
+                    return n.texData.get(u.dataId).values = i, u
+                }
+                let i;
+                return i = Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new YO(dL, r.shape, a.shape) : new KO(dL, r.shape, a.shape), n.runWebGLProgram(i, [r, a], s)
+            }
+            var mL = {
+                kernelName: Gn,
+                backendName: "webgl",
+                kernelFunc: fL
+            };
+
+            function gL(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    shape: s
+                } = r, i = n, o = va.sizeFromShape(a.shape), u = va.inferFromImplicitShape(s, o), l = va.sizeFromShape(u);
+                va.assert(o === l, () => `The new shape (${u}) has ${l} elements and the old shape (${a.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`);
+                const c = i.texData.get(a.dataId);
+                return !c.isPacked || LM(a.shape, u) || null !== c.texture && LM(c.shape, u) ? (i.incRef(a.dataId), {
+                    dataId: a.dataId,
+                    shape: u,
+                    dtype: a.dtype
+                }) : function (e, t, n) {
+                    const r = [_M(e.shape), ...FM(e.shape)],
+                        a = {
+                            dtype: e.dtype,
+                            shape: r,
+                            dataId: e.dataId
+                        },
+                        s = [_M(t), ...FM(t)],
+                        i = new MO(s, r),
+                        o = [r],
+                        u = n.runWebGLProgram(i, [a], e.dtype, o, !0);
+                    return {
+                        dataId: u.dataId,
+                        shape: t,
+                        dtype: u.dtype
+                    }
+                }(a, u, i)
+            }
+            var yL = {
+                    kernelName: lr,
+                    backendName: "webgl",
+                    kernelFunc: gL
+                },
+                bL = class {
+                    constructor(e, t) {
+                        this.variableNames = ["x"];
+                        const {
+                            windowSize: n,
+                            batchSize: r,
+                            inSize: a,
+                            outSize: s
+                        } = e;
+                        this.outputShape = [r, s];
+                        const i = 4 * Math.floor(n / 4),
+                            o = n % 4;
+                        let u = "sumValue += dot(values, ones);";
+                        if (null != t) {
+                            const e = 1 / t;
+                            u = `sumValue += dot(values * ${va.isInt(e)?e.toPrecision(2):e}, ones);`
+                        }
+                        let l = "";
+                        a % n > 0 && (l = `\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return 0.0;\n        }\n      `), this.userCode = `\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${l}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${i}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${u}\n        }\n\n        int inIdx = inOffset + ${i};\n        if (${1===o}) {\n          vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);\n\n          ${u}\n        } else if (${2===o}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1), 0.0, 0.0);\n\n          ${u}\n        } else if (${3===o}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2), 0.0);\n\n          ${u}\n        }\n        setOutput(sumValue);\n      }\n    `
+                    }
+                },
+                vL = class {
+                    constructor(e, t) {
+                        this.variableNames = ["x"];
+                        const {
+                            windowSize: n,
+                            batchSize: r,
+                            inSize: a,
+                            outSize: s
+                        } = e;
+                        this.outputShape = [r, s];
+                        let i = "0.0",
+                            o = "";
+                        "prod" === t ? i = "1.0" : "min" === t ? (i = "1.0 / 1e-20", o = "min") : "max" === t && (i = "-1.0 / 1e-20", o = "max");
+                        let u = `${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+                        "sum" === t ? u = "sumValue" : "prod" === t ? u = "prodValue" : "all" === t ? u = "allValue" : "any" === t && (u = "anyValue");
+                        const l = 4 * Math.floor(n / 4),
+                            c = n % 4;
+                        let h = `\n      if (${"sum"===t}) {\n        sumValue += dot(values, ones);\n      } else if (${"prod"===t}) {\n        vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);\n        prodValue *= tmp[0] * tmp[1];\n      } else {\n        minMaxValue = ${o}(values, minMaxValue);\n        if (${"min"===t} || ${"max"===t}) {\n          minMaxValue = ${o}(values, minMaxValue);\n          bvec4 isNaN = isnan(values);\n          if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {\n            minMaxValue = vec4(NAN);\n          }\n        }\n      }\n    `,
+                            p = "vec4";
+                        "all" === t ? (i = "1.0", h = "\n        bool reducedAllValue = all(values);\n        float floatedReducedAllValue = float(reducedAllValue);\n        allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);\n      ", p = "bvec4") : "any" === t && (i = "0.0", h = "\n        bool reducedAnyValue = any(values);\n        float floatedReducedAnyValue = float(reducedAnyValue);\n        anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);\n      ", p = "bvec4");
+                        let d = "";
+                        a % n > 0 && (d = `\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return initializationValue;\n        }\n      `), this.userCode = `\n      const float initializationValue = ${i};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${d}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        vec4 minMaxValue = vec4(${i});\n        float prodValue = 1.0;\n        float sumValue = 0.0;\n        float allValue = 1.0;\n        float anyValue = 0.0;\n\n        for (int i = 0; i < ${l}; i += 4) {\n          int inIdx = inOffset + i;\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${h}\n        }\n\n        int inIdx = inOffset + ${l};\n        if (${1===c}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          ${h}\n        } else if (${2===c}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          ${h}\n        } else if (${3===c}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          ${h}\n        }\n        setOutput(${u});\n      }\n    `
+                    }
+                };
+
+            function xL(e, t, n, r) {
+                const a = function (e) {
+                    const t = [];
+                    for (; 0 === t.length || 1 !== t[t.length - 1].outSize;) {
+                        const n = t.length ? t[t.length - 1].outSize : e[1],
+                            r = jd.computeOptimalWindowSize(n);
+                        t.push({
+                            inSize: n,
+                            windowSize: r,
+                            outSize: Math.ceil(n / r)
+                        })
+                    }
+                    return t
+                }(e.shape);
+                let s = e;
+                for (let i = 0; i < a.length; i++) {
+                    const {
+                        inSize: o,
+                        windowSize: u,
+                        outSize: l
+                    } = a[i];
+                    let c, h;
+                    c = "mean" === n ? 0 === i ? new bL({
+                        windowSize: u,
+                        inSize: o,
+                        batchSize: e.shape[0],
+                        outSize: l
+                    }, o) : new bL({
+                        windowSize: u,
+                        inSize: o,
+                        batchSize: e.shape[0],
+                        outSize: l
+                    }) : new vL({
+                        windowSize: u,
+                        inSize: o,
+                        batchSize: e.shape[0],
+                        outSize: l
+                    }, n), h = s, s = r.runWebGLProgram(c, [s], t), h.dataId !== e.dataId && r.disposeIntermediateTensorInfo(h)
+                }
+                return s
+            }
+
+            function wL(e, t, n) {
+                const r = Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new class {
+                    constructor(e, t) {
+                        this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0;
+                        const n = new Array(e.length);
+                        for (let r = 0; r < n.length; r++) n[r] = e[t[r]];
+                        if (this.outputShape = n, this.rank = n.length, this.rank > 6) throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);
+                        const r = dD(this.rank),
+                            a = _O("rc", this.rank),
+                            s = new Array(this.rank);
+                        for (let e = 0; e < t.length; e++) s[t[e]] = a[e];
+                        const i = `vec2(${s.slice(-2).join()})`,
+                            o = `++${a[this.rank-1]} < ${n[this.rank-1]}`,
+                            u = `getChannel(getA(${s.join()}), ${i})`;
+                        this.userCode = `\n    void main() {\n      ${r} rc = getOutputCoords();\n      vec4 result = vec4(0.);\n      result[0] = ${u};\n      if(${o}) {\n        result[1] = ${u};\n      }\n      --${a[this.rank-1]};\n      if(++${a[this.rank-2]} < ${n[this.rank-2]}) {\n        result[2] = ${u};\n        if(${o}) {\n          result[3] = ${u};\n        }\n      }\n      setOutput(result);\n    }\n    `
+                    }
+                }(e.shape, t) : new class {
+                    constructor(e, t) {
+                        this.variableNames = ["A"];
+                        const n = new Array(e.length);
+                        for (let r = 0; r < n.length; r++) n[r] = e[t[r]];
+                        this.outputShape = n, this.rank = n.length;
+                        const r = dD(this.rank),
+                            a = function (e) {
+                                const t = e.length;
+                                if (t > 6) throw Error(`Transpose for rank ${t} is not yet supported`);
+                                const n = ["resRC.x", "resRC.y", "resRC.z", "resRC.w", "resRC.u", "resRC.v"],
+                                    r = new Array(t);
+                                for (let t = 0; t < e.length; t++) r[e[t]] = n[t];
+                                return r.join()
+                            }(t);
+                        this.userCode = `\n    void main() {\n      ${r} resRC = getOutputCoords();\n      setOutput(getA(${a}));\n    }\n    `
+                    }
+                }(e.shape, t);
+                return n.runWebGLProgram(r, [e], e.dtype)
+            }
+
+            function kL(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    axis: s,
+                    keepDims: i
+                } = r;
+                return function (e, t, n, r) {
+                    const a = t,
+                        s = e.shape.length,
+                        i = va.parseAxisParam(a, e.shape);
+                    let o = i;
+                    const u = jd.getAxesPermutation(o, s),
+                        l = null != u;
+                    let c = e;
+                    l && (c = wL(e, u, r), o = jd.getInnerMostAxes(o.length, s)), jd.assertAxesAreInnerMostDims("sum", o, s);
+                    const [h, p] = jd.computeOutAndReduceShapes(c.shape, o);
+                    let d = h;
+                    n && (d = jd.expandShapeToKeepDim(h, i));
+                    const f = va.sizeFromShape(p),
+                        m = gL({
+                            inputs: {
+                                x: c
+                            },
+                            attrs: {
+                                shape: [va.sizeFromShape(e.shape) / f, f]
+                            },
+                            backend: r
+                        }),
+                        g = xL(m, hs(e.dtype), "sum", r),
+                        y = gL({
+                            inputs: {
+                                x: g
+                            },
+                            attrs: {
+                                shape: d
+                            },
+                            backend: r
+                        });
+                    return r.disposeIntermediateTensorInfo(m), r.disposeIntermediateTensorInfo(g), l && r.disposeIntermediateTensorInfo(c), y
+                }(a, s, i, n)
+            }
+            var SL = {
+                kernelName: Cr,
+                backendName: "webgl",
+                kernelFunc: kL
+            };
+
+            function IL(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    perm: s
+                } = r, i = n, o = a.shape.length, u = new Array(o);
+                for (let e = 0; e < u.length; e++) u[e] = a.shape[s[e]];
+                let l;
+                if (i.shouldExecuteOnCPU([a])) {
+                    const e = i.texData.get(a.dataId).values,
+                        t = TO(e, a.shape, a.dtype, s, u);
+                    l = i.makeTensorInfo(u, a.dtype);
+                    i.texData.get(l.dataId).values = t
+                } else l = wL(a, s, i);
+                return l
+            }
+            var AL = {
+                kernelName: Yr,
+                backendName: "webgl",
+                kernelFunc: IL
+            };
+
+            function EL({
+                a: e,
+                b: t,
+                transposeA: n,
+                transposeB: r,
+                backend: a,
+                bias: s = null,
+                preluActivationWeights: i = null,
+                leakyreluAlpha: o = 0,
+                activation: u = null
+            }) {
+                const l = e.shape.length,
+                    c = t.shape.length,
+                    h = n ? e.shape[l - 2] : e.shape[l - 1],
+                    p = r ? t.shape[c - 1] : t.shape[c - 2],
+                    d = n ? e.shape[l - 1] : e.shape[l - 2],
+                    f = r ? t.shape[c - 2] : t.shape[c - 1],
+                    m = e.shape.slice(0, -2),
+                    g = t.shape.slice(0, -2),
+                    y = va.sizeFromShape(m),
+                    b = va.sizeFromShape(g),
+                    v = Yi.assertAndGetBroadcastShape(e.shape.slice(0, -2), t.shape.slice(0, -2)).concat([d, f]);
+                va.assert(h === p, () => `Error in matMul: inner shapes (${h}) and (${p}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${n} and transposeB=${r} must match.`);
+                const x = n ? [y, h, d] : [y, d, h],
+                    w = r ? [b, f, p] : [b, p, f],
+                    k = gL({
+                        inputs: {
+                            x: e
+                        },
+                        backend: a,
+                        attrs: {
+                            shape: x
+                        }
+                    }),
+                    S = gL({
+                        inputs: {
+                            x: t
+                        },
+                        backend: a,
+                        attrs: {
+                            shape: w
+                        }
+                    }),
+                    I = [k, S],
+                    A = Math.max(y, b),
+                    E = n ? k.shape[1] : k.shape[2],
+                    N = null != s,
+                    C = null != i,
+                    T = "leakyrelu" === u,
+                    R = null != u ? uL(u, !0) : null;
+                let _;
+                if ((1 === d || 1 === f) && E > 1e3 && !1 === (N || C || T || null != R)) {
+                    let e = k,
+                        t = S;
+                    n && (e = IL({
+                        inputs: {
+                            x: k
+                        },
+                        backend: a,
+                        attrs: {
+                            perm: [0, 2, 1]
+                        }
+                    }), I.push(e)), r && (t = IL({
+                        inputs: {
+                            x: S
+                        },
+                        backend: a,
+                        attrs: {
+                            perm: [0, 2, 1]
+                        }
+                    }), I.push(t));
+                    const s = 1 === f;
+                    let i = e;
+                    1 !== f && (i = gL({
+                        inputs: {
+                            x: e
+                        },
+                        backend: a,
+                        attrs: {
+                            shape: [A, E, 1]
+                        }
+                    }), I.push(i));
+                    const o = 1 === f ? 2 : 1;
+                    let u = t;
+                    s && (u = gL({
+                        inputs: {
+                            x: t
+                        },
+                        backend: a,
+                        attrs: {
+                            shape: [A, 1, E]
+                        }
+                    }), I.push(u));
+                    const l = fL({
+                        inputs: {
+                            a: i,
+                            b: u
+                        },
+                        backend: a
+                    });
+                    _ = kL({
+                        inputs: {
+                            x: l
+                        },
+                        backend: a,
+                        attrs: {
+                            axis: o,
+                            keepDims: !0
+                        }
+                    }), I.push(l)
+                } else {
+                    const u = cs(e.dtype, t.dtype),
+                        l = new lL(x, w, [A, d, f], n, r, N, R, C, T),
+                        c = [k, S];
+                    if (null != s && c.push(s), C && c.push(i), T) {
+                        const e = a.makeTensorInfo([], "float32", va.createScalarValue(o, "float32"));
+                        c.push(e), I.push(e)
+                    }
+                    _ = a.runWebGLProgram(l, c, u)
+                }
+                const F = gL({
+                    inputs: {
+                        x: _
+                    },
+                    backend: a,
+                    attrs: {
+                        shape: v
+                    }
+                });
+                I.push(_);
+                for (const e of I) a.disposeIntermediateTensorInfo(e);
+                return F
+            }
+            var NL = {
+                    kernelName: ra,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            a: a,
+                            b: s,
+                            bias: i,
+                            preluActivationWeights: o
+                        } = t, {
+                            transposeA: u,
+                            transposeB: l,
+                            activation: c,
+                            leakyreluAlpha: h
+                        } = r;
+                        return EL({
+                            a: a,
+                            b: s,
+                            transposeA: u,
+                            transposeB: l,
+                            backend: n,
+                            bias: i,
+                            preluActivationWeights: o,
+                            leakyreluAlpha: h,
+                            activation: c
+                        })
+                    }
+                },
+                CL = "return abs(x);";
+            var TL = {
+                    kernelName: et,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            x: r
+                        } = t;
+                        if (n.shouldExecuteOnCPU([r]) && "complex64" !== r.dtype) {
+                            const e = n.texData.get(r.dataId),
+                                t = gO(e.values);
+                            return n.makeTensorInfo(r.shape, r.dtype, t)
+                        }
+                        let a;
+                        return a = Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS") ? new zO(r.shape, CL) : new PO(r.shape, CL), n.runWebGLProgram(a, [r], r.dtype)
+                    }
+                },
+                RL = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return acos(x);\n"
+                }),
+                _L = {
+                    kernelName: tt,
+                    backendName: "webgl",
+                    kernelFunc: RL
+                },
+                FL = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  if (x < 1.0) return NAN;\nreturn log(x + sqrt(x * x - 1.0));"
+                }),
+                ML = {
+                    kernelName: nt,
+                    backendName: "webgl",
+                    kernelFunc: FL
+                },
+                DL = "return a + b;",
+                OL = oL({
+                    opSnippet: DL,
+                    packedOpSnippet: DL,
+                    supportsComplex: !0,
+                    cpuKernelImpl: VD
+                }),
+                LL = {
+                    kernelName: rt,
+                    backendName: "webgl",
+                    kernelFunc: OL
+                };
+            var PL = {
+                kernelName: at,
+                backendName: "webgl",
+                kernelFunc: function e(t) {
+                    const {
+                        inputs: n,
+                        backend: r
+                    } = t, a = n;
+                    if (1 === a.length) return XO({
+                        inputs: {
+                            x: a[0]
+                        },
+                        backend: r
+                    });
+                    if (a.length > Ye().get("WEBGL_MAX_TEXTURES_IN_SHADER")) {
+                        const t = Math.floor(a.length / 2),
+                            n = e({
+                                inputs: a.slice(0, t),
+                                backend: r
+                            }),
+                            s = e({
+                                inputs: a.slice(t),
+                                backend: r
+                            });
+                        return e({
+                            inputs: [n, s],
+                            backend: r
+                        })
+                    }
+                    const s = a.map(e => e.dtype).reduce((e, t) => cs(e, t)),
+                        i = a.map(e => e.shape),
+                        o = Ye().getBool("WEBGL_PACK") ? new class {
+                            constructor(e, t) {
+                                this.outputShape = [], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = e, this.variableNames = t.map((e, t) => "T" + t);
+                                const n = [];
+                                this.variableNames.forEach(e => {
+                                    n.push(`vec4 v${e} = get${e}AtOutCoords();`)
+                                });
+                                const r = this.variableNames.map(e => "v" + e).join(" + ");
+                                this.userCode = `\n      void main() {\n        ${n.join("\n        ")}\n\n        vec4 result = ${r};\n        setOutput(result);\n      }\n    `
+                            }
+                        }(a[0].shape, i) : new class {
+                            constructor(e, t) {
+                                this.outputShape = [], this.outputShape = e, this.variableNames = t.map((e, t) => "T" + t);
+                                const n = [];
+                                this.variableNames.forEach(e => {
+                                    n.push(`float v${e} = get${e}AtOutCoords();`)
+                                });
+                                const r = this.variableNames.map(e => "v" + e).join(" + ");
+                                this.userCode = `\n      void main() {\n        ${n.join("\n        ")}\n\n        float result = ${r};\n        setOutput(result);\n      }\n    `
+                            }
+                        }(a[0].shape, i);
+                    return r.runWebGLProgram(o, a, s)
+                }
+            };
+            var $L = {
+                kernelName: st,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r, o = a.shape.length, u = va.parseAxisParam(s, a.shape);
+                    let l = u;
+                    const c = jd.getAxesPermutation(l, o);
+                    let h = a;
+                    null != c && (h = IL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: c
+                        }
+                    }), l = jd.getInnerMostAxes(l.length, o)), jd.assertAxesAreInnerMostDims("all", l, o);
+                    const [p, d] = jd.computeOutAndReduceShapes(h.shape, l), f = gL({
+                        inputs: {
+                            x: h
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [-1, va.sizeFromShape(d)]
+                        }
+                    }), m = xL(f, f.dtype, "all", n);
+                    let g;
+                    if (i) {
+                        g = gL({
+                            inputs: {
+                                x: m
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: jd.expandShapeToKeepDim(p, u)
+                            }
+                        })
+                    } else g = gL({
+                        inputs: {
+                            x: m
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: p
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(m), null != c && n.disposeIntermediateTensorInfo(h), g
+                }
+            };
+            var BL = {
+                kernelName: it,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        keepDims: i
+                    } = r, o = a.shape.length, u = va.parseAxisParam(s, a.shape);
+                    let l = u;
+                    const c = jd.getAxesPermutation(l, o);
+                    let h = a;
+                    null != c && (h = IL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: c
+                        }
+                    }), l = jd.getInnerMostAxes(l.length, o)), jd.assertAxesAreInnerMostDims("any", l, o);
+                    const [p, d] = jd.computeOutAndReduceShapes(h.shape, l), f = gL({
+                        inputs: {
+                            x: h
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [-1, va.sizeFromShape(d)]
+                        }
+                    }), m = xL(f, f.dtype, "any", n);
+                    let g;
+                    if (i) {
+                        g = gL({
+                            inputs: {
+                                x: m
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: jd.expandShapeToKeepDim(p, u)
+                            }
+                        })
+                    } else g = gL({
+                        inputs: {
+                            x: m
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: p
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(m), null != c && n.disposeIntermediateTensorInfo(h), g
+                }
+            };
+
+            function zL(e, t, n, r = null) {
+                const a = null != r ? r.shape : t.shape,
+                    s = a[a.length - 1],
+                    i = jd.computeOptimalWindowSize(s),
+                    o = new class {
+                        constructor(e, t, n, r) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, va.assert(e.length > 2, () => `Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);
+                            const a = e[e.length - 1],
+                                s = Math.ceil(a / t);
+                            this.outputShape = e.slice(0, -1), s > 1 && this.outputShape.push(s), r || this.variableNames.push("bestIndicesA");
+                            const i = this.outputShape,
+                                o = i.length,
+                                u = dD(o),
+                                l = FO("coords", o);
+                            let c, h;
+                            if (1 === s) {
+                                h = o + 1;
+                                const e = dD(h);
+                                c = `\n        ${e} sourceLocR = ${e}(${l.join()}, 0);\n        ++${l[o-1]};\n        ${e} sourceLocG = ${e}(${l.join()}, 0);\n        ++${l[o-2]};\n        ${e} sourceLocA = ${e}(${l.join()}, 0);\n        --${l[o-1]};\n        ${e} sourceLocB = ${e}(${l.join()}, 0);\n        --${l[o-2]};`
+                            } else h = o, c = `\n        ${u} sourceLocR = coords;\n        ++${l[o-1]};\n        ${u} sourceLocG = coords;\n        ++${l[o-2]};\n        ${u} sourceLocA = coords;\n        --${l[o-1]};\n        ${u} sourceLocB = coords;\n        --${l[o-2]};`;
+                            const p = ["x", "y", "z", "w", "u", "v"].slice(0, h),
+                                d = "." + p[h - 1],
+                                f = p.map(e => "int " + e),
+                                m = FO("sourceLocR", h - 1).concat("inIdx.r"),
+                                g = FO("sourceLocG", h - 1).concat("inIdx.g"),
+                                y = FO("sourceLocB", h - 1).concat("inIdx.b"),
+                                b = FO("sourceLocA", h - 1).concat("inIdx.a"),
+                                v = "max" === n ? "greaterThan" : "lessThan",
+                                x = r ? "" : `\n          inIdx = round(vec4(getBestIndicesAChannel(${m.join()}),\n                             getBestIndicesAChannel(${g.join()}),\n                             getBestIndicesAChannel(${y.join()}),\n                             getBestIndicesAChannel(${b.join()})));`,
+                                w = `vec4(\n            getAChannel(${m.join()}),\n            hasNextCol ? getAChannel(${g.join()}) : 0.,\n            hasNextRow ? getAChannel(${y.join()}) : 0.,\n            hasNextRow && hasNextCol ? getAChannel(${b.join()}) : 0.)`,
+                                k = r ? "" : `\n      float getBestIndicesAChannel(${f.join()}) {\n        return getChannel(getBestIndicesA(${p.join()}),\n                                          vec2(${p.slice(-2).join()}));\n      }`;
+                            this.userCode = `\n      float getAChannel(${f.join()}) {\n        return getChannel(getA(${p.join()}),\n                               vec2(${p.slice(-2).join()}));\n      }\n      ${k}\n      void main() {\n        ${u} coords = getOutputCoords();\n        bool hasNextCol = ${l[o-1]} < ${i[o-1]-1};\n        bool hasNextRow = ${l[o-2]} < ${i[o-2]-1};\n        ${c}\n        ivec4 srcIdx = ivec4(sourceLocR${d}, sourceLocG${d},\n          sourceLocB${d}, sourceLocA${d}) * ${t};\n        ivec4 inIdx = srcIdx;\n        vec4 bestIndex = vec4(inIdx);\n        vec4 bestValue = ${w};\n\n        for (int i = 0; i < ${t}; i++) {\n          inIdx = srcIdx;\n          ${x}\n          vec4 candidate = ${w};\n          bvec4 nan = isnan(candidate);\n          bvec4 replace = bvec4(\n            vec4(${v}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));\n\n          bestValue = vec4(replace.x  ? candidate.x : bestValue.x,\n                           replace.y  ? candidate.y : bestValue.y,\n                           replace.z  ? candidate.z : bestValue.z,\n                           replace.w  ? candidate.w : bestValue.w);\n          bestIndex = mix(bestIndex, vec4(inIdx), vec4(replace));\n          srcIdx++;\n        }\n        setOutput(bestIndex);\n      }\n    `
+                        }
+                    }(a, i, n, null == r),
+                    u = null == r ? [t] : [t, r],
+                    l = e.runWebGLProgram(o, u, "int32");
+                if (l.shape.length === t.shape.length) {
+                    const r = zL(e, t, n, l);
+                    return e.disposeIntermediateTensorInfo(l), r
+                }
+                return l
+            }
+
+            function UL(e, t, n, r) {
+                const a = [n];
+                if (jd.assertAxesAreInnerMostDims("arg" + r.charAt(0).toUpperCase() + r.slice(1), a, t.shape.length), !Ye().getBool("WEBGL_PACK_REDUCE") || t.shape.length <= 2) {
+                    const n = [],
+                        s = e.texData.get(t.dataId);
+                    let i = t;
+                    null !== s && s.isPacked && (i = e.unpackTensor(t), n.push(i));
+                    const [o, u] = jd.computeOutAndReduceShapes(i.shape, a), l = va.sizeFromShape(u), c = gL({
+                        inputs: {
+                            x: i
+                        },
+                        backend: e,
+                        attrs: {
+                            shape: [-1, l]
+                        }
+                    });
+                    n.push(c);
+                    const h = function e(t, n, r, a = null) {
+                        let s = n.shape[0],
+                            i = n.shape[1];
+                        null != a && (s = a.shape[0], i = a.shape[1]);
+                        const o = jd.computeOptimalWindowSize(i),
+                            u = {
+                                windowSize: o,
+                                inSize: i,
+                                batchSize: s,
+                                outSize: Math.ceil(i / o)
+                            },
+                            l = new class {
+                                constructor(e, t, n) {
+                                    this.variableNames = ["A"];
+                                    const {
+                                        windowSize: r,
+                                        batchSize: a,
+                                        outSize: s
+                                    } = e;
+                                    n || this.variableNames.push("bestIndicesA"), this.outputShape = [a, s];
+                                    const i = "max" === t ? ">" : "<",
+                                        o = n ? "inOffset + i;" : "round(getBestIndicesA(batch, inOffset + i));";
+                                    this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${r};\n\n        int bestIndex = inOffset;\n        float bestValue = getA(batch, bestIndex);\n\n        for (int i = 0; i < ${r}; i++) {\n          int inIdx = ${o};\n          float candidate = getA(batch, inIdx);\n          if (candidate ${i} bestValue) {\n            bestValue = candidate;\n            bestIndex = inIdx;\n          }\n        }\n        setOutput(float(bestIndex));\n      }\n    `
+                                }
+                            }(u, r, null == a),
+                            c = [n];
+                        null != a && c.push(a);
+                        const h = t.runWebGLProgram(l, c, "int32");
+                        if (1 === h.shape[1]) return h;
+                        const p = e(t, n, r, h);
+                        return t.disposeIntermediateTensorInfo(h), p
+                    }(e, c, r);
+                    n.push(h);
+                    const p = gL({
+                        inputs: {
+                            x: h
+                        },
+                        backend: e,
+                        attrs: {
+                            shape: o
+                        }
+                    });
+                    return n.forEach(t => e.disposeIntermediateTensorInfo(t)), p
+                }
+                return zL(e, t, r)
+            }
+            var WL = {
+                kernelName: ot,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s
+                    } = r;
+                    let i = va.parseAxisParam(s, a.shape);
+                    const o = jd.getAxesPermutation(i, a.shape.length);
+                    let u = a;
+                    const l = [];
+                    null != o && (u = IL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: o
+                        }
+                    }), l.push(u), i = jd.getInnerMostAxes(i.length, u.shape.length)), jd.assertAxesAreInnerMostDims("argMax", [i[0]], u.shape.length);
+                    const c = UL(n, u, i[0], "max");
+                    return l.forEach(e => n.disposeIntermediateTensorInfo(e)), c
+                }
+            };
+            var VL = {
+                    kernelName: ut,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            axis: s
+                        } = r;
+                        let i = va.parseAxisParam(s, a.shape);
+                        const o = jd.getAxesPermutation(i, a.shape.length);
+                        let u = a;
+                        const l = [];
+                        null != o && (u = IL({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: o
+                            }
+                        }), l.push(u), i = jd.getInnerMostAxes(i.length, u.shape.length)), jd.assertAxesAreInnerMostDims("argMin", [i[0]], u.shape.length);
+                        const c = UL(n, u, i[0], "min");
+                        return l.forEach(e => n.disposeIntermediateTensorInfo(e)), c
+                    }
+                },
+                jL = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return asin(x);\n"
+                }),
+                GL = {
+                    kernelName: lt,
+                    backendName: "webgl",
+                    kernelFunc: jL
+                },
+                HL = iL({
+                    opSnippet: "if (isnan(x)) return x;return log(x + sqrt(x * x + 1.0));"
+                }),
+                qL = {
+                    kernelName: ct,
+                    backendName: "webgl",
+                    kernelFunc: HL
+                },
+                KL = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return atan(x);\n"
+                }),
+                YL = {
+                    kernelName: ht,
+                    backendName: "webgl",
+                    kernelFunc: KL
+                },
+                XL = oL({
+                    opSnippet: "\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n\n  return atan(a, b);\n",
+                    packedOpSnippet: "\n  vec4 result = atan(a, b);\n  vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n"
+                }),
+                QL = {
+                    kernelName: dt,
+                    backendName: "webgl",
+                    kernelFunc: XL
+                },
+                ZL = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  if ((x < -1.0) || (x > 1.0)) return NAN;\nreturn (log(1.0 + x) - log(1.0 - x)) / 2.0;"
+                }),
+                JL = {
+                    kernelName: pt,
+                    backendName: "webgl",
+                    kernelFunc: ZL
+                },
+                eP = class {
+                    constructor(e, t, n, r = !1, a = !1) {
+                        if (this.variableNames = ["x"], "avg" === t && n) throw new Error("Cannot compute positions for average pool.");
+                        const s = e.filterWidth,
+                            i = e.strideHeight,
+                            o = e.strideWidth,
+                            u = e.dilationHeight,
+                            l = e.dilationWidth,
+                            c = e.effectiveFilterHeight,
+                            h = e.effectiveFilterWidth,
+                            p = e.padInfo.top,
+                            d = e.padInfo.left;
+                        this.outputShape = e.outShape;
+                        const f = "avg" === t,
+                            m = `((batch  * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,
+                            g = `(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`;
+                        let y = "0.0";
+                        if (f || (y = "-1.0 / 1e-20"), n) {
+                            const t = ">=";
+                            return void(this.userCode = `\n        const ivec2 strides = ivec2(${i}, ${o});\n        const ivec2 pads = ivec2(${p}, ${d});\n\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int batch = coords[0];\n          int d = coords[3];\n\n          ivec2 xRCCorner = coords.yz * strides - pads;\n          int xRCorner = xRCCorner.x;\n          int xCCorner = xRCCorner.y;\n\n          // max/min x(?, ?, d) to get y(yR, yC, d).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n          float avgValue = 0.0;\n\n          for (int wR = 0; wR < ${c};\n              wR += ${u}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${h};\n                wC += ${l}) {\n              int xC = xCCorner + wC;\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              float value = getX(batch, xR, xC, d);\n\n              // If a min / max value has already been found, use it. If not,\n              // use the current value.\n              float currMinMaxValue = mix(\n                  value, minMaxValue, minMaxValueFound);\n              if (value ${t} currMinMaxValue) {\n                minMaxValue = value;\n                minMaxValueFound = 1.0;\n                minMaxPosition = ${r?a?m:g:`wR * ${h} + wC`};\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `)
+                        }
+                        let b = `${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+                        "avg" === t && (b = "avgValue / count");
+                        const v = 4 * Math.floor(s / 4),
+                            x = s % 4,
+                            w = `\n      if (${f}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = max(values, minMaxValue);\n      }\n    `;
+                        this.userCode = `\n      const ivec2 strides = ivec2(${i}, ${o});\n      const ivec2 pads = ivec2(${p}, ${d});\n      const float initializationValue = ${y};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xR, int xC, int d) {\n        if (xC < 0 || xC >= ${e.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xR, xC, d);\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d = coords[3];\n\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // max/min x(?, ?, d) to get y(yR, yC, d).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${y});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wR = 0; wR < ${c};\n            wR += ${u}) {\n          int xR = xRCorner + wR;\n\n          if (xR < 0 || xR >= ${e.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${v}; wC += 4) {\n            int xC = xCCorner + wC * ${l};\n\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${l}, d),\n              getValue(batch, xR, xC + 2 * ${l}, d),\n              getValue(batch, xR, xC + 3 * ${l}, d)\n            );\n\n            ${w}\n          }\n\n          int xC = xCCorner + ${v};\n          if (${1===x}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              initializationValue,\n              initializationValue,\n              initializationValue\n            );\n\n            ${w}\n          } else if (${2===x}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${l}, d),\n              initializationValue,\n              initializationValue\n            );\n\n            ${w}\n          } else if (${3===x}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${l}, d),\n              getValue(batch, xR, xC + 2 * ${l}, d),\n              initializationValue\n            );\n\n            ${w}\n          }\n        }\n        setOutput(${b});\n      }\n    `
+                    }
+                },
+                tP = class {
+                    constructor(e, t, n, r = !1, a = !1) {
+                        if (this.variableNames = ["x"], "avg" === t && n) throw new Error("Cannot compute positions for average pool.");
+                        const s = e.filterWidth,
+                            i = e.strideDepth,
+                            o = e.strideHeight,
+                            u = e.strideWidth,
+                            l = e.dilationDepth,
+                            c = e.dilationHeight,
+                            h = e.dilationWidth,
+                            p = e.effectiveFilterDepth,
+                            d = e.effectiveFilterHeight,
+                            f = e.effectiveFilterWidth,
+                            m = e.padInfo.front,
+                            g = e.padInfo.top,
+                            y = e.padInfo.left;
+                        this.outputShape = e.outShape;
+                        const b = "avg" === t;
+                        let v = "0.0";
+                        if (b || (v = "-1.0 / 1e-20"), n) {
+                            const t = ">=";
+                            return void(this.userCode = `\n        const ivec3 strides =\n            ivec3(${i}, ${o}, ${u});\n        const ivec3 pads = ivec3(${m}, ${g}, ${y});\n\n        void main() {\n          ivec5 coords = getOutputCoords();\n          int batch = coords.x;\n          int ch = coords.u;\n\n          ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n          int xDCorner = xCorner.x;\n          int xRCorner = xCorner.y;\n          int xCCorner = xCorner.z;\n\n          // max/min x(?, ?, ?, ch) to get y(yD, yR, yC, ch).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n\n          for (int wD = 0; wD < ${p};\n              wD += ${l}) {\n            int xD = xDCorner + wD;\n\n            if (xD < 0 || xD >= ${e.inDepth}) {\n              continue;\n            }\n\n            for (int wR = 0; wR < ${d};\n                wR += ${c}) {\n              int xR = xRCorner + wR;\n\n              if (xR < 0 || xR >= ${e.inHeight}) {\n                continue;\n              }\n\n              for (int wC = 0; wC < ${f};\n                  wC += ${h}) {\n                int xC = xCCorner + wC;\n\n                if (xC < 0 || xC >= ${e.inWidth}) {\n                  continue;\n                }\n\n                float value = getX(batch, xD, xR, xC, ch);\n\n                // If a min / max value has already been found, use it. If not,\n                // use the current value.\n                float currMinMaxValue = mix(\n                    value, minMaxValue, minMaxValueFound);\n                if (value ${t} currMinMaxValue) {\n                  minMaxValue = value;\n                  minMaxValueFound = 1.0;\n                  minMaxPosition = ${r?a?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${d} * ${f} +\n                      wR * ${f} + wC`};\n                }\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `)
+                        }
+                        let x = `${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+                        "avg" === t && (x = "avgValue / count");
+                        const w = 4 * Math.floor(s / 4),
+                            k = s % 4,
+                            S = `\n      if (${b}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = max(values, minMaxValue);\n      }\n    `;
+                        this.userCode = `\n      const ivec3 strides =\n        ivec3(${i}, ${o}, ${u});\n      const ivec3 pads = ivec3(${m}, ${g}, ${y});\n      const float initializationValue = ${v};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xD, int xR, int xC, int ch) {\n        if (xC < 0 || xC >= ${e.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xD, xR, xC, ch);\n      }\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xDCorner = xCorner.x;\n        int xRCorner = xCorner.y;\n        int xCCorner = xCorner.z;\n\n        // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${v});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wD = 0; wD < ${p};\n            wD += ${l}) {\n          int xD = xDCorner + wD;\n\n          if (xD < 0 || xD >= ${e.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${d};\n            wR += ${c}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${w}; wC += 4) {\n              int xC = xCCorner + wC * ${h};\n\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${h}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${h}, ch),\n                getValue(batch, xD, xR, xC + 3 * ${h}, ch)\n              );\n\n              ${S}\n            }\n\n            int xC = xCCorner + ${w};\n            if (${1===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                initializationValue,\n                initializationValue,\n                initializationValue\n              );\n\n              ${S}\n            } else if (${2===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${h}, ch),\n                initializationValue,\n                initializationValue\n              );\n\n              ${S}\n            } else if (${3===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${h}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${h}, ch),\n                initializationValue\n              );\n\n              ${S}\n            }\n          }\n          setOutput(${x});\n        }\n      }\n    `
+                    }
+                };
+            var nP = {
+                kernelName: ft,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t;
+                    KM(a, "avgPool");
+                    const {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u
+                    } = r;
+                    va.assert(jd.eitherStridesOrDilationsAreOne(i, 1), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);
+                    const l = jd.computePool2DInfo(a.shape, s, i, 1, o, u);
+                    if (1 === l.filterWidth && 1 === l.filterHeight && va.arraysEqual(l.inShape, l.outShape)) return XO({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    const c = new eP(l, "avg", !1);
+                    return n.runWebGLProgram(c, [a], "float32")
+                }
+            };
+            var rP = {
+                kernelName: gt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u,
+                        dataFormat: l
+                    } = r, c = jd.computePool3DInfo(a.shape, s, i, [1, 1, 1], o, u, l), h = new tP(c, "avg", !1);
+                    return n.runWebGLProgram(h, [a], "float32")
+                }
+            };
+            var aP = {
+                kernelName: yt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s
+                    } = t, i = s, {
+                        filterSize: o,
+                        strides: u,
+                        pad: l,
+                        dimRoundingMode: c
+                    } = r, h = jd.computePool3DInfo(i.shape, o, u, [1, 1, 1], l, c), p = new class {
+                        constructor(e) {
+                            this.variableNames = ["dy"], this.outputShape = e.inShape;
+                            const t = e.filterDepth,
+                                n = e.filterHeight,
+                                r = e.filterWidth,
+                                a = e.strideDepth,
+                                s = e.strideHeight,
+                                i = e.strideWidth,
+                                o = e.dilationDepth,
+                                u = e.dilationHeight,
+                                l = e.dilationWidth,
+                                c = e.effectiveFilterDepth,
+                                h = e.effectiveFilterHeight,
+                                p = e.effectiveFilterWidth,
+                                d = c - 1 - e.padInfo.front,
+                                f = h - 1 - e.padInfo.top,
+                                m = p - 1 - e.padInfo.left,
+                                g = 1 / (t * n * r);
+                            this.userCode = `\n      const ivec3 pads = ivec3(${d}, ${f}, ${m});\n      const float avgMultiplier = float(${g});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, d) with pos mask(:, :, :, ch) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${c};\n            wD += ${o}) {\n          float dyD = float(dyDCorner + wD) / ${a}.0;\n\n          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${h};\n              wR += ${u}) {\n            float dyR = float(dyRCorner + wR) / ${s}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${p};\n                wC += ${l}) {\n              float dyC = float(dyCCorner + wC) / ${i}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n\n              dotProd += dyValue * avgMultiplier;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(h);
+                    return n.runWebGLProgram(p, [a], i.dtype)
+                }
+            };
+            var sP = {
+                kernelName: mt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s
+                    } = t, i = s;
+                    KM([a, s], "avgPoolGrad");
+                    const {
+                        filterSize: o,
+                        strides: u,
+                        pad: l
+                    } = r, c = jd.computePool2DInfo(i.shape, o, u, 1, l), h = new class {
+                        constructor(e) {
+                            this.variableNames = ["dy"], this.outputShape = e.inShape;
+                            const t = e.filterHeight,
+                                n = e.filterWidth,
+                                r = e.strideHeight,
+                                a = e.strideWidth,
+                                s = e.dilationHeight,
+                                i = e.dilationWidth,
+                                o = e.effectiveFilterHeight,
+                                u = e.effectiveFilterWidth,
+                                l = o - 1 - e.padInfo.top,
+                                c = u - 1 - e.padInfo.left,
+                                h = 1 / (t * n);
+                            this.userCode = `\n      const ivec2 pads = ivec2(${l}, ${c});\n      const float avgMultiplier = float(${h});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${o};\n            wR += ${s}) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${u};\n            wC+= ${i}) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n\n            dotProd += dyValue * avgMultiplier;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(c);
+                    return n.runWebGLProgram(h, [a], i.dtype)
+                }
+            };
+            var iP = {
+                    kernelName: bt,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            a: a,
+                            b: s
+                        } = t, {
+                            transposeA: i,
+                            transposeB: o
+                        } = r;
+                        return EL({
+                            a: a,
+                            b: s,
+                            transposeA: i,
+                            transposeB: o,
+                            backend: n
+                        })
+                    }
+                },
+                oP = {
+                    kernelName: ln,
+                    backendName: "webgl",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t,
+                        attrs: n
+                    }) => {
+                        const {
+                            x: r,
+                            mean: a,
+                            variance: s,
+                            offset: i,
+                            scale: o
+                        } = e;
+                        va.assert(a.shape.length === s.shape.length, () => "Batch normalization gradient requires mean and variance to have equal ranks."), va.assert(null == i || a.shape.length === i.shape.length, () => "Batch normalization gradient requires mean and offset to have equal ranks."), va.assert(null == o || a.shape.length === o.shape.length, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+                        let {
+                            varianceEpsilon: u
+                        } = n;
+                        null == u && (u = .001);
+                        const l = [r, a, s];
+                        let c = null;
+                        null != i && (c = i.shape, l.push(i));
+                        let h = null;
+                        null != o && (h = o.shape, l.push(o));
+                        const p = Ye().getBool("WEBGL_PACK_NORMALIZATION") ? new class {
+                            constructor(e, t, n, r, a, s) {
+                                this.packedInputs = !0, this.packedOutput = !0, this.variableNames = ["x", "mean", "variance"], jd.assertAndGetBroadcastShape(e, t), jd.assertAndGetBroadcastShape(e, n);
+                                let i = "vec4(0.0)";
+                                null != r && (jd.assertAndGetBroadcastShape(e, r), this.variableNames.push("offset"), i = "getOffsetAtOutCoords()");
+                                let o = "vec4(1.0)";
+                                null != a && (jd.assertAndGetBroadcastShape(e, a), this.variableNames.push("scale"), o = "getScaleAtOutCoords()"), this.outputShape = e, this.userCode = `\n      void main() {\n        vec4 offset = ${i};\n        vec4 scale = ${o};\n\n        vec4 x = getXAtOutCoords();\n        vec4 mean = getMeanAtOutCoords();\n        vec4 variance = getVarianceAtOutCoords();\n\n        vec4 inv = scale * inversesqrt(variance + vec4(${s}));\n\n        setOutput((x - mean) * inv + offset);\n      }\n    `
+                            }
+                        }(r.shape, a.shape, s.shape, c, h, u) : new class {
+                            constructor(e, t, n, r, a, s) {
+                                this.outputShape = [], this.variableNames = ["x", "mean", "variance"], jd.assertAndGetBroadcastShape(e, t), jd.assertAndGetBroadcastShape(e, n);
+                                let i = "0.0";
+                                null != r && (jd.assertAndGetBroadcastShape(e, r), this.variableNames.push("offset"), i = "getOffsetAtOutCoords()");
+                                let o = "1.0";
+                                null != a && (jd.assertAndGetBroadcastShape(e, a), this.variableNames.push("scale"), o = "getScaleAtOutCoords()"), this.outputShape = e, this.userCode = `\n      void main() {\n        float x = getXAtOutCoords();\n        float mean = getMeanAtOutCoords();\n        float variance = getVarianceAtOutCoords();\n        float offset = ${i};\n        float scale = ${o};\n        float inv = scale * inversesqrt(variance + float(${s}));\n        setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));\n      }\n    `
+                            }
+                        }(r.shape, a.shape, s.shape, c, h, u);
+                        return t.runWebGLProgram(p, l, l[0].dtype)
+                    }
+                },
+                uP = ["x", "y", "z", "w", "u", "v"];
+
+            function lP(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    begin: s,
+                    size: i
+                } = r, [o, u] = fo.parseSliceParams(a, s, i);
+                if (fo.assertParamsValid(a, o, u), 0 === va.sizeFromShape(u)) return n.makeTensorInfo(u, a.dtype, []);
+                if (n.shouldExecuteOnCPU([a]) || "string" === a.dtype) {
+                    const e = n.texData.get(a.dataId),
+                        t = yO(e.values, o, u, a.shape, a.dtype);
+                    return n.makeTensorInfo(u, a.dtype, t)
+                }
+                const {
+                    isPacked: l
+                } = n.texData.get(a.dataId), c = fo.isSliceContinous(a.shape, o, u);
+                if (l || !c) {
+                    const e = Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new class {
+                            constructor(e) {
+                                this.variableNames = ["source"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = e, this.rank = e.length, this.customUniforms = [{
+                                    name: "start",
+                                    arrayIndex: this.rank,
+                                    type: "int"
+                                }];
+                                const t = dD(this.rank),
+                                    n = FO("coords", this.rank),
+                                    r = FO("sourceLoc", this.rank),
+                                    a = 1 === this.rank ? "sourceLoc" : `vec2(${r.slice(-2).join()})`,
+                                    s = `getChannel(getSource(${r.join()}), ${a})`,
+                                    i = `\n      result.x = ${s};\n      if (++${n[this.rank-1]} < ${e[this.rank-1]}) {\n        ++${r[this.rank-1]};\n        result.y = ${s};\n        --${r[this.rank-1]};\n      }\n    `,
+                                    o = 1 === this.rank ? "" : `\n      --${n[this.rank-1]};\n      if (++${n[this.rank-2]} < ${e[this.rank-2]}) {\n        ++${r[this.rank-2]};\n        result.z = ${s};\n        if (++${n[this.rank-1]} < ${e[this.rank-1]}) {\n          ++${r[this.rank-1]};\n          result.w = ${s};\n        }\n      }\n    `,
+                                    u = this.rank <= 4 ? `sourceLoc = coords +\n            ${t}(${e.map((e,t)=>`start[${t}]`).join()});` : e.map((e, t) => `${r[t]} = ${n[t]} + start[${t}];`).join("\n");
+                                this.userCode = `\n      void main() {\n        ${t} coords = getOutputCoords();\n        ${t} sourceLoc;\n        ${u}\n        vec4 result = vec4(0.);\n        ${i}\n        ${o}\n        setOutput(result);\n      }\n    `
+                            }
+                        }(u) : new class {
+                            constructor(e) {
+                                this.variableNames = ["source"], this.outputShape = e, this.rank = e.length;
+                                const t = dD(this.rank);
+                                this.customUniforms = [{
+                                    name: "start",
+                                    arrayIndex: this.rank,
+                                    type: "int"
+                                }];
+                                const n = function (e) {
+                                    if (1 === e) return "sourceLoc";
+                                    if (e <= 6) return uP.slice(0, e).map(e => "sourceLoc." + e).join(",");
+                                    throw Error(`Slicing for rank ${e} is not yet supported`)
+                                }(this.rank);
+                                let r;
+                                r = `\n        ${t} sourceLoc;\n        ${t} coords = getOutputCoords();\n        ${e.map((e,t)=>`sourceLoc.${uP[t]} = start[${t}] + coords.${uP[t]};`).join("\n")}\n      `, this.userCode = `\n      void main() {\n        ${r}\n        setOutput(getSource(${n}));\n      }\n    `
+                            }
+                        }(u),
+                        t = [o];
+                    return n.runWebGLProgram(e, [a], a.dtype, t)
+                }
+                return n.uploadToGPU(a.dataId),
+                    function (e, t, n, r) {
+                        const a = r.texData.get(e.dataId),
+                            s = r.makeTensorInfo(n, e.dtype),
+                            i = r.texData.get(s.dataId);
+                        Object.assign(i, a), i.refCount = 1, i.shape = n, i.dtype = e.dtype;
+                        let o = fo.computeFlatOffset(t, va.computeStrides(e.shape));
+                        a.slice && (o += a.slice.flatOffset), i.slice = {
+                            flatOffset: o,
+                            origDataId: a.slice && a.slice.origDataId || e.dataId
+                        };
+                        const u = r.dataRefCount.get(i.slice.origDataId) || 1;
+                        return r.dataRefCount.set(i.slice.origDataId, u + 1), s
+                    }(a, o, u, n)
+            }
+            var cP = {
+                    kernelName: wr,
+                    backendName: "webgl",
+                    kernelFunc: lP
+                },
+                hP = {
+                    kernelName: vt,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            blockShape: s,
+                            crops: i
+                        } = r;
+                        va.assert(a.shape.length <= 4, () => "batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");
+                        const o = s.reduce((e, t) => e * t),
+                            u = jd.getReshaped(a.shape, s, o),
+                            l = jd.getPermuted(u.length, s.length),
+                            c = jd.getReshapedPermuted(a.shape, s, o),
+                            h = jd.getSliceBeginCoords(i, s.length),
+                            p = jd.getSliceSize(c, i, s.length),
+                            d = [],
+                            f = gL({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: u
+                                }
+                            }),
+                            m = IL({
+                                inputs: {
+                                    x: f
+                                },
+                                backend: n,
+                                attrs: {
+                                    perm: l
+                                }
+                            }),
+                            g = gL({
+                                inputs: {
+                                    x: m
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: c
+                                }
+                            }),
+                            y = lP({
+                                inputs: {
+                                    x: g
+                                },
+                                backend: n,
+                                attrs: {
+                                    begin: h,
+                                    size: p
+                                }
+                            });
+                        return d.push(f), d.push(m), d.push(g), d.forEach(e => n.disposeIntermediateTensorInfo(e)), y
+                    }
+                };
+            var pP = {
+                kernelName: xt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        weights: s
+                    } = t, {
+                        size: i
+                    } = r, o = n.readSync(a.dataId), u = n.readSync(s.dataId), l = jD(o, u, s.dtype, s.shape, i);
+                    return n.makeTensorInfo([i], s.dtype, l)
+                }
+            };
+            var dP = {
+                    kernelName: kt,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            s0: r,
+                            s1: a
+                        } = t, s = n.readSync(r.dataId), i = n.readSync(a.dataId), o = jd.assertAndGetBroadcastShape(Array.from(s), Array.from(i));
+                        return n.makeTensorInfo([o.length], "int32", Int32Array.from(o))
+                    }
+                },
+                fP = oL({
+                    opSnippet: "return float(a != b);",
+                    cpuKernelImpl: hO,
+                    dtype: "bool"
+                }),
+                mP = {
+                    kernelName: qn,
+                    backendName: "webgl",
+                    kernelFunc: fP
+                };
+
+            function gP(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    input: r
+                } = t;
+                return XO({
+                    inputs: {
+                        x: n.texData.get(r.dataId).complexTensorInfos.real
+                    },
+                    backend: n
+                })
+            }
+            var yP = {
+                kernelName: ir,
+                backendName: "webgl",
+                kernelFunc: gP
+            };
+            var bP = {
+                    kernelName: St,
+                    backendName: "webgl",
+                    kernelFunc: function e(t) {
+                        const {
+                            inputs: n,
+                            backend: r,
+                            attrs: a
+                        } = t, {
+                            x: s
+                        } = n, {
+                            dtype: i
+                        } = a;
+                        if ("complex64" === i) {
+                            if ("complex64" === s.dtype) return XO({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: r
+                            });
+                            const t = Pc(s.shape),
+                                n = e({
+                                    inputs: {
+                                        x: s
+                                    },
+                                    backend: r,
+                                    attrs: {
+                                        dtype: "float32"
+                                    }
+                                }),
+                                a = ZO({
+                                    inputs: {
+                                        real: n,
+                                        imag: t
+                                    },
+                                    backend: r
+                                });
+                            return t.dispose(), r.disposeIntermediateTensorInfo(n), a
+                        }
+                        if ("complex64" === s.dtype) {
+                            const t = gP({
+                                    inputs: {
+                                        input: s
+                                    },
+                                    backend: r
+                                }),
+                                n = e({
+                                    inputs: {
+                                        x: t
+                                    },
+                                    backend: r,
+                                    attrs: {
+                                        dtype: i
+                                    }
+                                });
+                            return r.disposeIntermediateTensorInfo(t), n
+                        }
+                        if (!va.hasEncodingLoss(s.dtype, i)) {
+                            const e = XO({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: r
+                            });
+                            return {
+                                dataId: e.dataId,
+                                shape: e.shape,
+                                dtype: i
+                            }
+                        }
+                        if ("int32" === i) return function (e, t) {
+                            const n = new PO(e.shape, "return float(int(x));"),
+                                r = t.runWebGLProgram(n, [e], "int32");
+                            return {
+                                dataId: r.dataId,
+                                shape: r.shape,
+                                dtype: r.dtype
+                            }
+                        }(s, r);
+                        if ("bool" === i) {
+                            const e = r.makeTensorInfo([], "bool", va.getTypedArrayFromDType("bool", 1)),
+                                t = fP({
+                                    inputs: {
+                                        a: s,
+                                        b: e
+                                    },
+                                    backend: r
+                                });
+                            return r.disposeIntermediateTensorInfo(e), t
+                        }
+                        throw new Error(`Error in Cast: failed to cast ${s.dtype} to ${i}`)
+                    }
+                },
+                vP = "return ceil(x);",
+                xP = iL({
+                    opSnippet: vP,
+                    packedOpSnippet: vP,
+                    cpuKernelImpl: HD
+                }),
+                wP = {
+                    kernelName: It,
+                    backendName: "webgl",
+                    kernelFunc: xP
+                };
+            var kP = {
+                kernelName: At,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        clipValueMin: s,
+                        clipValueMax: i
+                    } = r;
+                    let o;
+                    o = Ye().getBool("WEBGL_PACK_CLIP") ? new class {
+                        constructor(e) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.customUniforms = [{
+                                name: "minVal",
+                                type: "float"
+                            }, {
+                                name: "maxVal",
+                                type: "float"
+                            }], this.outputShape = e, this.userCode = "\n      void main() {\n        vec4 value = getAAtOutCoords();\n\n        if (any(isnan(value))) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, vec4(minVal), vec4(maxVal)));\n      }\n    "
+                        }
+                    }(a.shape) : new class {
+                        constructor(e) {
+                            this.variableNames = ["A"], this.customUniforms = [{
+                                name: "minVal",
+                                type: "float"
+                            }, {
+                                name: "maxVal",
+                                type: "float"
+                            }], this.outputShape = e, this.userCode = "\n\n      void main() {\n        float value = getAAtOutCoords();\n        if (isnan(value)) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, minVal, maxVal));\n      }\n    "
+                        }
+                    }(a.shape);
+                    const u = [
+                        [s],
+                        [i]
+                    ];
+                    return n.runWebGLProgram(o, [a], a.dtype, u)
+                }
+            };
+
+            function SP(e, t) {
+                return {
+                    dataId: t.dataId,
+                    dtype: t.dtype,
+                    shape: e.shape
+                }
+            }
+            var IP = {
+                kernelName: Nt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        x: r
+                    } = t, a = n.texData.get(r.dataId), s = new class {
+                        constructor(e) {
+                            this.variableNames = ["real", "imag"], this.outputShape = e, this.userCode = "\n      void main() {\n        float re = abs(getRealAtOutCoords());\n        float im = abs(getImagAtOutCoords());\n        float mx = max(re, im);\n\n        // sadly the length function in glsl is not underflow-safe\n        // (at least not on Intel GPUs). So the safe solution is\n        // to ensure underflow-safety in all cases.\n        setOutput(\n          mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))\n        );\n      }\n    "
+                        }
+                    }(r.shape), i = [SP(r, a.complexTensorInfos.real), SP(r, a.complexTensorInfos.imag)];
+                    return n.runWebGLProgram(s, i, i[0].dtype)
+                }
+            };
+
+            function AP(e, t, n) {
+                const r = e.indexOf(t);
+                return e.map((e, t) => t === r ? `${e} - ${n}` : e).join()
+            }
+
+            function EP(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    input: r
+                } = t;
+                return XO({
+                    inputs: {
+                        x: n.texData.get(r.dataId).complexTensorInfos.imag
+                    },
+                    backend: n
+                })
+            }
+            var NP = {
+                kernelName: gn,
+                backendName: "webgl",
+                kernelFunc: EP
+            };
+
+            function CP(e, t, n) {
+                const r = e[0].dtype;
+                if ("complex64" === r) {
+                    const r = e.map(e => gP({
+                            inputs: {
+                                input: e
+                            },
+                            backend: n
+                        })),
+                        a = e.map(e => EP({
+                            inputs: {
+                                input: e
+                            },
+                            backend: n
+                        })),
+                        s = CP(r, t, n),
+                        i = CP(a, t, n),
+                        o = ZO({
+                            inputs: {
+                                real: s,
+                                imag: i
+                            },
+                            backend: n
+                        });
+                    return r.forEach(e => n.disposeIntermediateTensorInfo(e)), a.forEach(e => n.disposeIntermediateTensorInfo(e)), n.disposeIntermediateTensorInfo(s), n.disposeIntermediateTensorInfo(i), o
+                }
+                let a = n.shouldExecuteOnCPU(e);
+                if ("string" === r && (a = !0), a) {
+                    const a = e.map(e => {
+                            const r = va.sizeFromShape(e.shape.slice(t));
+                            return gL({
+                                inputs: {
+                                    x: e
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: [-1, r]
+                                }
+                            })
+                        }),
+                        s = a.map(e => ({
+                            vals: n.readSync(e.dataId),
+                            shape: e.shape
+                        })),
+                        i = jd.computeOutShape(a.map(e => e.shape), 1),
+                        o = 1 === a[0].shape[0],
+                        u = qD(s, i, r, o),
+                        l = jd.computeOutShape(e.map(e => e.shape), t),
+                        c = n.makeTensorInfo(l, r, u);
+                    return a.forEach(e => n.disposeIntermediateTensorInfo(e)), c
+                }
+                if (e.length > Ye().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER")) {
+                    const r = Math.floor(e.length / 2),
+                        a = CP(e.slice(0, r), t, n),
+                        s = CP(e.slice(r), t, n),
+                        i = CP([a, s], t, n);
+                    return n.disposeIntermediateTensorInfo(a), n.disposeIntermediateTensorInfo(s), i
+                }
+                if (Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS") && e[0].shape.length > 1) {
+                    const a = new class {
+                        constructor(e, t) {
+                            this.packedInputs = !0, this.packedOutput = !0, this.outputShape = [], this.outputShape = jd.computeOutShape(e, t);
+                            const n = this.outputShape,
+                                r = n.length,
+                                a = dD(r),
+                                s = FO("coords", r),
+                                i = ["x", "y", "z", "w", "u", "v"].slice(0, r);
+                            this.variableNames = e.map((e, t) => "T" + t);
+                            const o = new Array(e.length - 1);
+                            o[0] = e[0][t];
+                            for (let n = 1; n < o.length; n++) o[n] = o[n - 1] + e[n][t];
+                            const u = i[t],
+                                l = i.slice(-2),
+                                c = i.join();
+                            let h = `if (${u} < ${o[0]}) {\n        return getChannel(\n            getT0(${c}), vec2(${l.join()}));\n        }`;
+                            for (let e = 1; e < o.length; e++) {
+                                const t = o[e - 1];
+                                h += `\n        if (${u} < ${o[e]}  && ${u} >= ${o[e-1]}) {\n          return getChannel(\n            getT${e}(${AP(i,u,t)}),\n            vec2(${AP(l,u,t)}));\n        }`
+                            }
+                            const p = o.length,
+                                d = o[o.length - 1];
+                            h += `\n        return getChannel(\n          getT${p}(${AP(i,u,d)}),\n          vec2(${AP(l,u,d)}));`, this.userCode = `\n      float getValue(${i.map(e=>"int "+e)}) {\n        ${h}\n      }\n\n      void main() {\n        ${a} coords = getOutputCoords();\n        vec4 result = vec4(getValue(${s}), 0., 0., 0.);\n\n        ${s[r-1]} = ${s[r-1]} + 1;\n        if (${s[r-1]} < ${n[r-1]}) {\n          result.g = getValue(${s});\n        }\n\n        ${s[r-2]} = ${s[r-2]} + 1;\n        if (${s[r-2]} < ${n[r-2]}) {\n          result.a = getValue(${s});\n        }\n\n        ${s[r-1]} = ${s[r-1]} - 1;\n        if (${s[r-2]} < ${n[r-2]} &&\n            ${s[r-1]} < ${n[r-1]}) {\n          result.b = getValue(${s});\n        }\n        setOutput(result);\n      }\n    `
+                        }
+                    }(e.map(e => e.shape), t);
+                    return n.runWebGLProgram(a, e, r)
+                }
+                const {
+                    tensors2D: s,
+                    outShape: i
+                } = function (e, t, n) {
+                    const r = jd.computeOutShape(e.map(e => e.shape), t);
+                    return {
+                        tensors2D: e.map(e => gL({
+                            inputs: {
+                                x: e
+                            },
+                            attrs: {
+                                shape: [-1, va.sizeFromShape(e.shape.slice(t))]
+                            },
+                            backend: n
+                        })),
+                        outShape: r
+                    }
+                }(e, t, n), o = new class {
+                    constructor(e) {
+                        this.outputShape = [], this.outputShape = jd.computeOutShape(e, 1), this.variableNames = e.map((e, t) => "T" + t);
+                        const t = new Array(e.length - 1);
+                        t[0] = e[0][1];
+                        for (let n = 1; n < t.length; n++) t[n] = t[n - 1] + e[n][1];
+                        const n = [`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];
+                        for (let e = 1; e < t.length; e++) {
+                            const r = t[e - 1];
+                            n.push(`else if (yC < ${t[e]}) setOutput(getT${e}(yR, yC-${r}));`)
+                        }
+                        const r = t.length,
+                            a = t[t.length - 1];
+                        n.push(`else setOutput(getT${r}(yR, yC-${a}));`), this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int yR = coords.x;\n        int yC = coords.y;\n\n        ${n.join("\n        ")}\n      }\n    `
+                    }
+                }(s.map(e => e.shape)), u = n.runWebGLProgram(o, s, r);
+                s.forEach(e => n.disposeIntermediateTensorInfo(e));
+                const l = gL({
+                    inputs: {
+                        x: u
+                    },
+                    attrs: {
+                        shape: i
+                    },
+                    backend: n
+                });
+                return n.disposeIntermediateTensorInfo(u), l
+            }
+
+            function TP(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    axis: a
+                } = r, s = va.parseAxisParam(a, t[0].shape)[0], i = jd.computeOutShape(t.map(e => e.shape), s);
+                if (0 === va.sizeFromShape(i)) return n.makeTensorInfo(i, t[0].dtype, []);
+                const o = t.filter(e => va.sizeFromShape(e.shape) > 0);
+                if (1 === o.length) return XO({
+                    inputs: {
+                        x: o[0]
+                    },
+                    backend: n
+                });
+                const u = o.map(e => e.shape);
+                return jd.assertParamsConsistent(u, s), CP(o, s, n)
+            }
+            var RP = {
+                    kernelName: Ct,
+                    backendName: "webgl",
+                    kernelFunc: TP
+                },
+                _P = class {
+                    constructor(e, t = !1, n = null, r = !1, a = !1) {
+                        this.variableNames = ["x", "W"], this.outputShape = e.outShape;
+                        const s = e.padInfo.top,
+                            i = e.padInfo.left,
+                            o = e.strideHeight,
+                            u = e.strideWidth,
+                            l = e.dilationHeight,
+                            c = e.dilationWidth,
+                            h = e.filterHeight,
+                            p = e.filterWidth,
+                            d = 4 * Math.floor(e.inChannels / 4),
+                            f = e.inChannels % 4,
+                            m = "channelsLast" === e.dataFormat,
+                            g = m ? 1 : 2,
+                            y = m ? 2 : 3,
+                            b = m ? 3 : 1;
+                        let v = "",
+                            x = "";
+                        n && (v = r ? `float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }` : a ? `float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }` : `\n          float activation(float x) {\n            ${n}\n          }\n        `, x = "result = activation(result);");
+                        const w = t ? "result += getBiasAtOutCoords();" : "";
+                        t && this.variableNames.push("bias"), r && this.variableNames.push("preluActivationWeights"), a && this.variableNames.push("leakyreluAlpha"), this.userCode = `\n      ${v}\n\n      const ivec2 strides = ivec2(${o}, ${u});\n      const ivec2 pads = ivec2(${s}, ${i});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d2 = coords[${b}];\n\n        ivec2 xRCCorner =\n            ivec2(coords[${g}], coords[${y}]) * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${h}; wR++) {\n          int xR = xRCorner + wR * ${l};\n\n          if (xR < 0 || xR >= ${e.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${p}; wC++) {\n            int xC = xCCorner + wC * ${c};\n\n            if (xC < 0 || xC >= ${e.inWidth}) {\n              continue;\n            }\n\n            for (int d1 = 0; d1 < ${d}; d1 += 4) {\n              vec4 wValues = vec4(\n                getW(wR, wC, d1, d2),\n                getW(wR, wC, d1 + 1, d2),\n                getW(wR, wC, d1 + 2, d2),\n                getW(wR, wC, d1 + 3, d2)\n              );\n\n              if (${m}) {\n                vec4 xValues = vec4(\n                  getX(batch, xR, xC, d1),\n                  getX(batch, xR, xC, d1 + 1),\n                  getX(batch, xR, xC, d1 + 2),\n                  getX(batch, xR, xC, d1 + 3)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec4 xValues = vec4(\n                  getX(batch, d1, xR, xC),\n                  getX(batch, d1 + 1, xR, xC),\n                  getX(batch, d1 + 2, xR, xC),\n                  getX(batch, d1 + 3, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n\n            if (${1===f}) {\n\n              if (${m}) {\n                dotProd +=\n                    getX(batch, xR, xC, ${d}) *\n                    getW(wR, wC, ${d}, d2);\n              } else {\n                dotProd +=\n                    getX(batch, ${d}, xR, xC) *\n                    getW(wR, wC, ${d}, d2);\n              }\n\n            } else if (${2===f}) {\n              vec2 wValues = vec2(\n                getW(wR, wC, ${d}, d2),\n                getW(wR, wC, ${d} + 1, d2)\n              );\n\n              if (${m}) {\n                vec2 xValues = vec2(\n                  getX(batch, xR, xC, ${d}),\n                  getX(batch, xR, xC, ${d} + 1)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec2 xValues = vec2(\n                  getX(batch, ${d}, xR, xC),\n                  getX(batch, ${d} + 1, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            } else if (${3===f}) {\n              vec3 wValues = vec3(\n                getW(wR, wC, ${d}, d2),\n                getW(wR, wC, ${d} + 1, d2),\n                getW(wR, wC, ${d} + 2, d2)\n              );\n\n              if (${m}) {\n                vec3 xValues = vec3(\n                  getX(batch, xR, xC, ${d}),\n                  getX(batch, xR, xC, ${d} + 1),\n                  getX(batch, xR, xC, ${d} + 2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec3 xValues = vec3(\n                  getX(batch, ${d}, xR, xC),\n                  getX(batch, ${d} + 1, xR, xC),\n                  getX(batch, ${d} + 2, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            }\n          }\n        }\n\n        float result = dotProd;\n        ${w}\n        ${x}\n        setOutput(result);\n      }\n    `
+                    }
+                };
+
+            function FP({
+                x: e,
+                filter: t,
+                convInfo: n,
+                backend: r,
+                bias: a = null,
+                preluActivationWeights: s = null,
+                leakyreluAlpha: i = 0,
+                activation: o = null
+            }) {
+                const u = e.shape,
+                    l = r.texData.get(e.dataId),
+                    c = n.inChannels,
+                    h = u[0] * u[1] * u[2],
+                    p = n.outChannels,
+                    d = "channelsLast" === n.dataFormat;
+                let f;
+                const m = [];
+                if (!((1 === h || 1 === p) && c > 1e3) && l.isPacked && d && null != l.texture && u[2] % 2 != 0 && va.arraysEqual(l.shape.slice(-3), u.slice(-3))) {
+                    const c = u[0] * u[1] * (u[2] + 1),
+                        h = {
+                            dataId: e.dataId,
+                            shape: [1, c, n.inChannels],
+                            dtype: e.dtype
+                        },
+                        p = l.shape;
+                    l.shape = l.shape.slice(), l.shape[l.shape.length - 2]++, va.assert(LM(l.shape, h.shape), () => `packed reshape ${l.shape} to ${h.shape} isn't free`);
+                    const d = gL({
+                        inputs: {
+                            x: t
+                        },
+                        backend: r,
+                        attrs: {
+                            shape: [1, n.inChannels, n.outChannels]
+                        }
+                    });
+                    m.push(d);
+                    const g = EL({
+                            a: h,
+                            b: d,
+                            backend: r,
+                            transposeA: !1,
+                            transposeB: !1,
+                            bias: a,
+                            activation: o,
+                            preluActivationWeights: s,
+                            leakyreluAlpha: i
+                        }),
+                        y = r.texData.get(g.dataId);
+                    va.assert(y.isPacked, () => "batchMatMul result is expected to be packed"), l.shape = p, y.shape = n.outShape, f = XO({
+                        inputs: {
+                            x: g
+                        },
+                        backend: r
+                    }), f.shape = n.outShape, m.push(g)
+                } else {
+                    const l = gL({
+                            inputs: {
+                                x: e
+                            },
+                            backend: r,
+                            attrs: {
+                                shape: [1, d ? u[0] * u[1] * u[2] : u[0] * u[2] * u[3], n.inChannels]
+                            }
+                        }),
+                        c = gL({
+                            inputs: {
+                                x: t
+                            },
+                            backend: r,
+                            attrs: {
+                                shape: [1, n.inChannels, n.outChannels]
+                            }
+                        }),
+                        h = EL({
+                            a: l,
+                            b: c,
+                            transposeA: !1,
+                            transposeB: !1,
+                            backend: r,
+                            bias: a,
+                            activation: o,
+                            preluActivationWeights: s,
+                            leakyreluAlpha: i
+                        });
+                    f = gL({
+                        inputs: {
+                            x: h
+                        },
+                        backend: r,
+                        attrs: {
+                            shape: n.outShape
+                        }
+                    }), m.push(l), m.push(c), m.push(h)
+                }
+                for (const e of m) r.disposeIntermediateTensorInfo(e);
+                return f
+            }
+
+            function MP({
+                x: e,
+                filter: t,
+                convInfo: n,
+                backend: r,
+                bias: a = null,
+                preluActivationWeights: s = null,
+                leakyreluAlpha: i = 0,
+                activation: o = null
+            }) {
+                const {
+                    filterWidth: u,
+                    filterHeight: l,
+                    inChannels: c,
+                    outWidth: h,
+                    outHeight: p,
+                    dataFormat: d
+                } = n, f = "channelsLast" === d, m = u * l * c, g = p * h, y = [m, g], b = [], v = gL({
+                    inputs: {
+                        x: e
+                    },
+                    backend: r,
+                    attrs: {
+                        shape: e.shape.slice(1)
+                    }
+                }), x = gL({
+                    inputs: {
+                        x: t
+                    },
+                    backend: r,
+                    attrs: {
+                        shape: [1, m, va.sizeFromShape(t.shape) / m]
+                    }
+                });
+                b.push(v), b.push(x);
+                const w = new class {
+                        constructor(e, t) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.customUniforms = [{
+                                name: "inputShape",
+                                type: "ivec3"
+                            }, {
+                                name: "pad",
+                                type: "ivec2"
+                            }, {
+                                name: "stride",
+                                type: "ivec2"
+                            }, {
+                                name: "dilation",
+                                type: "ivec2"
+                            }, {
+                                name: "inChannels",
+                                type: "int"
+                            }, {
+                                name: "itemsPerBlockRow",
+                                type: "int"
+                            }, {
+                                name: "outWidth",
+                                type: "int"
+                            }], this.outputShape = e, this.enableShapeUniforms = bD(this.outputShape.length);
+                            const {
+                                dataFormat: n
+                            } = t, r = XM(), a = "channelsLast" === n, s = a ? 0 : 1, i = a ? 1 : 2, o = this.enableShapeUniforms ? "if(blockIndex < outShape[1] && pos < outShape[0]) {" : `if(blockIndex < ${e[1]} && pos < ${e[0]}) {`;
+                            let u = "";
+                            for (let e = 0; e <= 1; e++)
+                                for (let t = 0; t <= 1; t++) u += `\n          blockIndex = rc.y + ${t};\n          pos = rc.x + ${e};\n\n          ${o}\n            offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];\n            d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);\n\n            if(d0 < inputShape[${s}] && d0 >= 0) {\n              // Use custom imod instead mod. On Intel GPU, mod may generate\n              // unexpected value.\n              // https://github.com/tensorflow/tfjs/issues/5447\n              offsetX = imod(blockIndex, outWidth) * stride[1] - pad[1];\n              d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /\n                  inChannels);\n\n              if(d1 < inputShape[${i}] && d1 >= 0) {\n\n                ch = imod(pos, inChannels);\n\n                if (${a}) {\n                  innerDims = vec2(d1, ch);\n                  result[${2*e+t}] = getChannel(\n                    getA(d0, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                } else {\n                  innerDims = vec2(d0, d1);\n                  result[${2*e+t}] = getChannel(\n                    getA(ch, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                }\n              }\n            }\n          }\n        `;
+                            this.userCode = `\n      void main() {\n        ivec2 rc = getOutputCoords();\n\n        vec4 result = vec4(0);\n\n        int blockIndex, pos, offsetY, d0, offsetX, d1, ch;\n        vec2 innerDims;\n\n        ${u}\n\n        ${r.output} = result;\n      }\n    `
+                        }
+                    }(y, n),
+                    k = [v.shape, [n.padInfo.top, n.padInfo.left],
+                        [n.strideHeight, n.strideWidth],
+                        [n.dilationHeight, n.dilationWidth],
+                        [n.inChannels],
+                        [n.filterWidth * n.inChannels],
+                        [n.outWidth]
+                    ],
+                    S = r.runWebGLProgram(w, [v], "float32", k),
+                    I = gL({
+                        inputs: {
+                            x: S
+                        },
+                        backend: r,
+                        attrs: {
+                            shape: [1, y[0], y[1]]
+                        }
+                    });
+                b.push(S), b.push(I);
+                const A = null != a,
+                    E = null != s,
+                    N = "leakyrelu" === o,
+                    C = o ? uL(o, !0) : null,
+                    T = new lL(I.shape, x.shape, [1, g, n.outChannels], !0, !1, A, C, E, N),
+                    R = [I, x];
+                if (a && R.push(a), E && R.push(s), N) {
+                    const e = r.makeTensorInfo([], "float32", va.createScalarValue(i, "float32"));
+                    R.push(e), b.push(e)
+                }
+                const _ = r.runWebGLProgram(T, R, "float32"),
+                    F = gL({
+                        inputs: {
+                            x: _
+                        },
+                        backend: r,
+                        attrs: {
+                            shape: f ? [1, p, h, n.outChannels] : [1, n.outChannels, p, h]
+                        }
+                    });
+                b.push(_);
+                for (const e of b) r.disposeIntermediateTensorInfo(e);
+                return F
+            }
+            var DP = {
+                kernelName: Tt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dataFormat: u,
+                        dilations: l,
+                        dimRoundingMode: c
+                    } = r, h = jd.convertConv2DDataFormat(u), p = jd.computeConv2DInfo(a.shape, s.shape, i, l, o, c, !1, h);
+                    let d;
+                    if (1 !== p.filterHeight || 1 !== p.filterWidth || 1 !== p.dilationHeight || 1 !== p.dilationWidth || 1 !== p.strideHeight || 1 !== p.strideWidth || "SAME" !== p.padInfo.type && "VALID" !== p.padInfo.type)
+                        if (Ye().getBool("WEBGL_CONV_IM2COL") && 1 === a.shape[0]) d = MP({
+                            x: a,
+                            filter: s,
+                            convInfo: p,
+                            backend: n
+                        });
+                        else {
+                            const e = new _P(p);
+                            d = n.runWebGLProgram(e, [a, s], "float32")
+                        }
+                    else d = FP({
+                        x: a,
+                        filter: s,
+                        convInfo: p,
+                        backend: n
+                    });
+                    const f = gL({
+                        inputs: {
+                            x: d
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: p.outShape
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(d), f
+                }
+            };
+            var OP = {
+                kernelName: Rt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        dy: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dataFormat: u,
+                        dimRoundingMode: l,
+                        filterShape: c
+                    } = r, h = jd.convertConv2DDataFormat(u), p = jd.computeConv2DInfo(a.shape, c, i, 1, o, l, !1, h), d = new class {
+                        constructor(e) {
+                            this.variableNames = ["x", "dy"], this.outputShape = e.filterShape;
+                            const t = e.strideHeight,
+                                n = e.strideWidth,
+                                r = e.padInfo.top,
+                                a = e.padInfo.left,
+                                s = "channelsLast" === e.dataFormat;
+                            this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int d2 = coords.w;\n\n        // Convolve x(?, ?, d1) with dy(:, :, d2) to get dw(wR, wC, d1, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yR = 0; yR < ${e.outHeight}; yR++) {\n            int xR = wR + yR * ${t} - ${r};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${e.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${a};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              if (${s}) {\n                float dyValue = getDy(b, yR, yC, d2);\n                float xValue = getX(b, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              } else {\n                float dyValue = getDy(b, d2, yR, yC);\n                float xValue = getX(b, d1, xR, xC);\n                dotProd += (xValue * dyValue);\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(p);
+                    return n.runWebGLProgram(d, [a, s], "float32")
+                }
+            };
+            var LP = {
+                kernelName: _t,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        filter: s
+                    } = t, {
+                        inputShape: i,
+                        strides: o,
+                        pad: u,
+                        dataFormat: l,
+                        dimRoundingMode: c
+                    } = r, h = jd.convertConv2DDataFormat(l), p = jd.computeConv2DInfo(i, s.shape, o, 1, u, c, !1, h), d = new class {
+                        constructor(e) {
+                            this.variableNames = ["dy", "W"], this.outputShape = e.inShape;
+                            const t = e.filterHeight,
+                                n = e.filterWidth,
+                                r = e.strideHeight,
+                                a = e.strideWidth,
+                                s = "channelsLast" === e.dataFormat,
+                                i = t - 1 - e.padInfo.top,
+                                o = n - 1 - e.padInfo.left,
+                                u = s ? 1 : 2,
+                                l = s ? 2 : 3,
+                                c = s ? 3 : 1;
+                            this.userCode = `\n      const ivec2 pads = ivec2(${i}, ${o});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[${c}];\n\n        ivec2 dyCorner = ivec2(coords[${u}], coords[${l}]) - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            for (int d2 = 0; d2 < ${e.outChannels}; d2++) {\n\n              if (${s}) {\n                float xValue = getDy(batch, idyR, idyC, d2);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              } else {\n                float xValue = getDy(batch, d2, idyR, idyC);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(p);
+                    return n.runWebGLProgram(d, [a, s], "float32")
+                }
+            };
+            var PP = {
+                kernelName: Ft,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dilations: u
+                    } = r, l = jd.computeConv3DInfo(a.shape, s.shape, i, u, o), c = new class {
+                        constructor(e) {
+                            this.variableNames = ["x", "W"], this.outputShape = e.outShape;
+                            const t = e.padInfo.front,
+                                n = e.padInfo.top,
+                                r = e.padInfo.left,
+                                a = e.strideDepth,
+                                s = e.strideHeight,
+                                i = e.strideWidth,
+                                o = e.dilationDepth,
+                                u = e.dilationHeight,
+                                l = e.dilationWidth,
+                                c = e.filterDepth,
+                                h = e.filterHeight,
+                                p = e.filterWidth,
+                                d = 4 * Math.floor(e.inChannels / 4),
+                                f = e.inChannels % 4;
+                            this.userCode = `\n      const ivec3 strides = ivec3(${a}, ${s}, ${i});\n      const ivec3 pads = ivec3(${t}, ${n}, ${r});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d2 = coords.u;\n\n        ivec3 xFRCCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xFCorner = xFRCCorner.x;\n        int xRCorner = xFRCCorner.y;\n        int xCCorner = xFRCCorner.z;\n\n        // Convolve x(?, ?, ?, d1) with w(:, :, :, d1, d2) to get\n        // y(yF, yR, yC, d2). ? = to be determined. : = across all\n        // values in that axis.\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${c}; wF++) {\n          int xF = xFCorner + wF * ${o};\n\n          if (xF < 0 || xF >= ${e.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${h}; wR++) {\n            int xR = xRCorner + wR * ${u};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${p}; wC++) {\n              int xC = xCCorner + wC * ${l};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              for (int d1 = 0; d1 < ${d}; d1 += 4) {\n                vec4 xValues = vec4(\n                  getX(batch, xF, xR, xC, d1),\n                  getX(batch, xF, xR, xC, d1 + 1),\n                  getX(batch, xF, xR, xC, d1 + 2),\n                  getX(batch, xF, xR, xC, d1 + 3)\n                );\n                vec4 wValues = vec4(\n                  getW(wF, wR, wC, d1, d2),\n                  getW(wF, wR, wC, d1 + 1, d2),\n                  getW(wF, wR, wC, d1 + 2, d2),\n                  getW(wF, wR, wC, d1 + 3, d2)\n                );\n\n                dotProd += dot(xValues, wValues);\n              }\n\n              if (${1===f}) {\n                dotProd +=\n                  getX(batch, xF, xR, xC, ${d}) *\n                  getW(wF, wR, wC, ${d}, d2);\n              } else if (${2===f}) {\n                vec2 xValues = vec2(\n                  getX(batch, xF, xR, xC, ${d}),\n                  getX(batch, xF, xR, xC, ${d} + 1)\n                );\n                vec2 wValues = vec2(\n                  getW(wF, wR, wC, ${d}, d2),\n                  getW(wF, wR, wC, ${d} + 1, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else if (${3===f}) {\n                vec3 xValues = vec3(\n                  getX(batch, xF, xR, xC, ${d}),\n                  getX(batch, xF, xR, xC, ${d} + 1),\n                  getX(batch, xF, xR, xC, ${d} + 2)\n                );\n                vec3 wValues = vec3(\n                  getW(wF, wR, wC, ${d}, d2),\n                  getW(wF, wR, wC, ${d} + 1, d2),\n                  getW(wF, wR, wC, ${d} + 2, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(l);
+                    return n.runWebGLProgram(c, [a, s], "float32")
+                }
+            };
+            var $P = {
+                kernelName: Mt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        dy: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        filterShape: u
+                    } = r, l = jd.computeConv3DInfo(a.shape, u, i, 1, o), c = new class {
+                        constructor(e) {
+                            this.variableNames = ["x", "dy"], this.outputShape = e.filterShape;
+                            const t = e.strideDepth,
+                                n = e.strideHeight,
+                                r = e.strideWidth,
+                                a = e.padInfo.front,
+                                s = e.padInfo.top,
+                                i = e.padInfo.left;
+                            this.userCode = `\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int wF = coords.x;\n        int wR = coords.y;\n        int wC = coords.z;\n        int d1 = coords.w;\n        int d2 = coords.u;\n\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yF = 0; yF < ${e.outDepth}; yF++) {\n            int xF = wF + yF * ${t} - ${a};\n\n            if (xF < 0 || xF >= ${e.inDepth}) {\n              continue;\n            }\n\n            for (int yR = 0; yR < ${e.outHeight}; yR++) {\n              int xR = wR + yR * ${n} - ${s};\n\n              if (xR < 0 || xR >= ${e.inHeight}) {\n                continue;\n              }\n\n              for (int yC = 0; yC < ${e.outWidth}; yC++) {\n                int xC = wC + yC * ${r} - ${i};\n\n                if (xC < 0 || xC >= ${e.inWidth}) {\n                  continue;\n                }\n\n                float dyValue = getDy(b, yF, yR, yC, d2);\n                float xValue = getX(b, xF, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(l);
+                    return n.runWebGLProgram(c, [a, s], "float32")
+                }
+            };
+            var BP = {
+                    kernelName: Dt,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            dy: a,
+                            filter: s
+                        } = t, {
+                            pad: i,
+                            strides: o,
+                            inputShape: u
+                        } = r, l = jd.computeConv3DInfo(u, s.shape, o, 1, i), c = new class {
+                            constructor(e) {
+                                this.variableNames = ["dy", "W"], this.outputShape = e.inShape;
+                                const t = e.filterDepth,
+                                    n = e.filterHeight,
+                                    r = e.filterWidth,
+                                    a = e.strideDepth,
+                                    s = e.strideHeight,
+                                    i = e.strideWidth,
+                                    o = t - 1 - e.padInfo.front,
+                                    u = n - 1 - e.padInfo.top,
+                                    l = r - 1 - e.padInfo.left;
+                                this.userCode = `\n      const ivec3 pads = ivec3(${o}, ${u}, ${l});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.u;\n\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyFCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${t}; wF++) {\n          float dyF = float(dyFCorner + wF) / ${a}.0;\n\n          if (dyF < 0.0 || dyF >= ${e.outDepth}.0 || fract(dyF) > 0.0) {\n            continue;\n          }\n          int idyF = int(dyF);\n\n          int wFPerm = ${t} - 1 - wF;\n\n          for (int wR = 0; wR < ${n}; wR++) {\n            float dyR = float(dyRCorner + wR) / ${s}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n              fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            int wRPerm = ${n} - 1 - wR;\n\n            for (int wC = 0; wC < ${r}; wC++) {\n              float dyC = float(dyCCorner + wC) / ${i}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              int wCPerm = ${r} - 1 - wC;\n\n              for (int d2 = 0; d2 < ${e.outChannels}; d2++) {\n                float xValue = getDy(batch, idyF, idyR, idyC, d2);\n                float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                            }
+                        }(l);
+                        return n.runWebGLProgram(c, [a, s], "float32")
+                    }
+                },
+                zP = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return cos(x);\n"
+                }),
+                UP = {
+                    kernelName: Ot,
+                    backendName: "webgl",
+                    kernelFunc: zP
+                },
+                WP = iL({
+                    opSnippet: "\n  float e2x = exp(-x);\n  return (e2x + 1.0 / e2x) / 2.0;\n"
+                }),
+                VP = {
+                    kernelName: Lt,
+                    backendName: "webgl",
+                    kernelFunc: WP
+                },
+                jP = {
+                    kernelName: $t,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            image: a,
+                            boxes: s,
+                            boxInd: i
+                        } = t, {
+                            cropSize: o,
+                            method: u,
+                            extrapolationValue: l
+                        } = r, c = new class {
+                            constructor(e, t, n, r, a) {
+                                this.variableNames = ["Image", "Boxes", "BoxInd"], this.outputShape = [];
+                                const [s, i, o, u] = e, [l] = t, [c, h] = n;
+                                this.outputShape = [l, c, h, u];
+                                const p = "bilinear" === r ? 1 : 0,
+                                    [d, f] = [i - 1 + ".0", o - 1 + ".0"],
+                                    [m, g, y] = c > 1 ? ["" + (i - 1) / (c - 1), "(y2-y1) * height_ratio", `y1*${d} + float(y)*(height_scale)`] : ["0.0", "0.0", "0.5 * (y1+y2) * " + d],
+                                    [b, v, x] = h > 1 ? ["" + (o - 1) / (h - 1), "(x2-x1) * width_ratio", `x1*${f} + float(x)*(width_scale)`] : ["0.0", "0.0", "0.5 * (x1+x2) * " + f];
+                                this.userCode = `\n      const float height_ratio = float(${m});\n      const float width_ratio = float(${b});\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int y = coords[1];\n        int x = coords[2];\n        int d = coords[3];\n\n        // get box vals\n        float y1 = getBoxes(b,0);\n        float x1 = getBoxes(b,1);\n        float y2 = getBoxes(b,2);\n        float x2 = getBoxes(b,3);\n\n        // get image in batch index\n        int bInd = round(getBoxInd(b));\n        if(bInd < 0 || bInd >= ${s}) {\n          return;\n        }\n\n        float height_scale = ${g};\n        float width_scale = ${v};\n\n        float in_y = ${y};\n        if( in_y < 0.0 || in_y > ${d} ) {\n          setOutput(float(${a}));\n          return;\n        }\n        float in_x = ${x};\n        if( in_x < 0.0 || in_x > ${f} ) {\n          setOutput(float(${a}));\n          return;\n        }\n\n        vec2 sourceFracIndexCR = vec2(in_x,in_y);\n        if(${p} == 1) {\n          // Compute the four integer indices.\n          ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);\n          ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));\n\n          float topLeft = getImage(b, sourceFloorCR.y, sourceFloorCR.x, d);\n          float bottomLeft = getImage(b, sourceCeilCR.y, sourceFloorCR.x, d);\n          float topRight = getImage(b, sourceFloorCR.y, sourceCeilCR.x, d);\n          float bottomRight = getImage(b, sourceCeilCR.y, sourceCeilCR.x, d);\n\n          vec2 fracCR = sourceFracIndexCR - vec2(sourceFloorCR);\n\n          float top = topLeft + (topRight - topLeft) * fracCR.x;\n          float bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;\n          float newValue = top + (bottom - top) * fracCR.y;\n          setOutput(newValue);\n        } else {\n          // Compute the coordinators of nearest neighbor point.\n          ivec2 sourceNearestCR = ivec2(floor(\n            sourceFracIndexCR + vec2(0.5,0.5)));\n          float newValue = getImage(b, sourceNearestCR.y, sourceNearestCR.x, d);\n          setOutput(newValue);\n        }\n      }\n    `
+                            }
+                        }(a.shape, s.shape, o, u, l);
+                        return n.runWebGLProgram(c, [a, s, i], "float32")
+                    }
+                },
+                GP = class {
+                    constructor(e, t, n) {
+                        this.variableNames = ["x"], this.customUniforms = [{
+                            name: "index",
+                            type: "float"
+                        }], this.outputShape = e;
+                        const r = e.length,
+                            a = t ? "0.0" : `getX(${HP(r,"coords")})`,
+                            s = e[e.length - 1];
+                        let i = "",
+                            o = "";
+                        t ? (i = n ? "end != " + (s - 1) : "end != 0", o = n ? "end + 1" : "end - 1") : (i = n ? "end + pow2 < " + s : "end >= pow2", o = n ? "end + pow2" : "end - pow2"), this.userCode = `\n      void main() {\n        ${dD(r)} coords = getOutputCoords();\n        int end = ${qP(r,"coords")};\n        float val = ${a};\n        int pow2 = int(pow(2.0, index));\n        if (${i}) {\n          int idx = ${o};\n          ${qP(r,"coords")} = idx;\n          val += getX(${HP(r,"coords")});\n        }\n        setOutput(val);\n      }\n    `
+                    }
+                };
+
+            function HP(e, t) {
+                if (1 === e) return "" + t;
+                if (2 === e) return `${t}.x, ${t}.y`;
+                if (3 === e) return `${t}.x, ${t}.y, ${t}.z`;
+                if (4 === e) return `${t}.x, ${t}.y, ${t}.z, ${t}.w`;
+                throw Error(`Cumulative sum for rank ${e} is not yet supported`)
+            }
+
+            function qP(e, t) {
+                if (1 === e) return "" + t;
+                if (2 === e) return t + ".y";
+                if (3 === e) return t + ".z";
+                if (4 === e) return t + ".w";
+                throw Error(`Cumulative sum for rank ${e} is not yet supported`)
+            }
+            var KP = {
+                kernelName: Pt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        exclusive: i,
+                        reverse: o
+                    } = r, u = a.shape.length, l = jd.getAxesPermutation([s], u);
+                    let c = a;
+                    null != l && (c = IL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: l
+                        }
+                    }));
+                    const h = jd.getInnerMostAxes(1, u)[0];
+                    if (h !== u - 1) throw new Error(`WebGL cumsum shader expects an inner-most axis=${a.shape.length-1} but got axis=${s}`);
+                    const p = c.shape[h];
+                    let d = XO({
+                        inputs: {
+                            x: c
+                        },
+                        backend: n
+                    });
+                    for (let e = 0; e <= Math.ceil(Math.log2(p)) - 1; e++) {
+                        const t = new GP(c.shape, !1, o),
+                            r = [
+                                [e]
+                            ],
+                            a = d;
+                        d = n.runWebGLProgram(t, [d], d.dtype, r), n.disposeIntermediateTensorInfo(a)
+                    }
+                    if (i) {
+                        const e = new GP(c.shape, i, o),
+                            t = d;
+                        d = n.runWebGLProgram(e, [d], d.dtype), n.disposeIntermediateTensorInfo(t)
+                    }
+                    if (null != l) {
+                        const e = IL({
+                            inputs: {
+                                x: d
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: jd.getUndoAxesPermutation(l)
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(c), e
+                    }
+                    return d
+                }
+            };
+            var YP = {
+                kernelName: Bt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        weights: s
+                    } = t, {
+                        size: i,
+                        binaryOutput: o
+                    } = r;
+                    if (1 === a.shape.length) {
+                        const e = n.readSync(a.dataId),
+                            t = n.readSync(s.dataId),
+                            r = jD(e, t, s.dtype, s.shape, i);
+                        return n.makeTensorInfo([i], s.dtype, r)
+                    }
+                    if (2 === a.shape.length) {
+                        const e = n.bufferSync(a),
+                            t = n.bufferSync(s),
+                            r = GD(e, t, i, o);
+                        return n.makeTensorInfo(r.shape, s.dtype, r.values)
+                    }
+                    throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)
+                }
+            };
+            var XP = {
+                    kernelName: zt,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            blockSize: s,
+                            dataFormat: i
+                        } = r, o = a.shape[0], u = ("NHWC" === i ? a.shape[1] : a.shape[2]) * s, l = ("NHWC" === i ? a.shape[2] : a.shape[3]) * s, c = ("NHWC" === i ? a.shape[3] : a.shape[1]) / (s * s), h = "NHWC" === i ? [o, u, l, c] : [o, c, u, l], p = new class {
+                            constructor(e, t, n) {
+                                this.variableNames = ["x"], this.outputShape = [], this.outputShape = e, this.blockSize = t, this.dataFormat = n, this.userCode = `\n    void main() {\n      ivec4 coords = getOutputCoords();\n      int b = coords[0];\n      int h = ${this.getHeightCoordString()};\n      int w = ${this.getWidthCoordString()};\n      int d = ${this.getDepthCoordString()};\n\n      int in_h = h / ${t};\n      int offset_h = imod(h, ${t});\n      int in_w = w / ${t};\n      int offset_w = imod(w, ${t});\n      int offset_d = (offset_h * ${t} + offset_w) *\n        ${this.getOutputDepthSize()};\n      int in_d = d + offset_d;\n\n      float result = ${this.getInputSamplingString()};\n      setOutput(result);\n    }\n  `
+                            }
+                            getHeightCoordString() {
+                                return "NHWC" === this.dataFormat ? "coords[1]" : "coords[2]"
+                            }
+                            getWidthCoordString() {
+                                return "NHWC" === this.dataFormat ? "coords[2]" : "coords[3]"
+                            }
+                            getDepthCoordString() {
+                                return "NHWC" === this.dataFormat ? "coords[3]" : "coords[1]"
+                            }
+                            getOutputDepthSize() {
+                                return "NHWC" === this.dataFormat ? this.outputShape[3] : this.outputShape[1]
+                            }
+                            getInputSamplingString() {
+                                return "NHWC" === this.dataFormat ? "getX(b, in_h, in_w, in_d)" : "getX(b, in_d, in_h, in_w)"
+                            }
+                        }(h, s, i);
+                        return n.runWebGLProgram(p, [a], a.dtype)
+                    }
+                },
+                QP = class {
+                    constructor(e, t = !1, n = null, r = !1, a = !1) {
+                        this.variableNames = ["x", "W"], this.customUniforms = [{
+                            name: "pads",
+                            type: "ivec2"
+                        }, {
+                            name: "strides",
+                            type: "ivec2"
+                        }, {
+                            name: "dilations",
+                            type: "ivec2"
+                        }, {
+                            name: "inDims",
+                            type: "ivec2"
+                        }], this.outputShape = e.outShape, this.enableShapeUniforms = bD(this.outputShape.length);
+                        const s = e.filterHeight,
+                            i = e.filterWidth,
+                            o = e.outChannels / e.inChannels;
+                        let u = "",
+                            l = "";
+                        n && (u = r ? `float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }` : a ? `float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }` : `\n          float activation(float x) {\n            ${n}\n          }\n        `, l = "result = activation(result);");
+                        const c = t ? "result += getBiasAtOutCoords();" : "";
+                        t && this.variableNames.push("bias"), r && this.variableNames.push("preluActivationWeights"), a && this.variableNames.push("leakyreluAlpha"), this.userCode = `\n      ${u}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${o};\n        int q = d2 - d1 * ${o};\n\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.\n        for (int wR = 0; wR < ${s}; wR++) {\n          int xR = xRCorner + wR * dilations[0];\n\n          if (xR < 0 || xR >= inDims[0]) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${i}; wC++) {\n            int xC = xCCorner + wC * dilations[1];\n\n            if (xC < 0 || xC >= inDims[1]) {\n              continue;\n            }\n\n            float xVal = getX(batch, xR, xC, d1);\n            float wVal = getW(wR, wC, d1, q);\n            dotProd += xVal * wVal;\n          }\n        }\n\n        float result = dotProd;\n        ${c}\n        ${l}\n        setOutput(result);\n      }\n    `
+                    }
+                },
+                ZP = class {
+                    constructor(e, t = !1, n = null, r = !1, a = !1) {
+                        this.variableNames = ["x", "W"], this.packedInputs = !0, this.packedOutput = !0, this.customUniforms = [{
+                            name: "pads",
+                            type: "ivec2"
+                        }, {
+                            name: "strides",
+                            type: "ivec2"
+                        }, {
+                            name: "dilations",
+                            type: "ivec2"
+                        }, {
+                            name: "inDims",
+                            type: "ivec2"
+                        }], this.outputShape = e.outShape, this.enableShapeUniforms = bD(this.outputShape.length);
+                        const s = e.outChannels / e.inChannels,
+                            i = e.padInfo.left,
+                            o = e.strideWidth,
+                            u = e.dilationWidth,
+                            l = e.filterHeight,
+                            c = e.filterWidth,
+                            h = c;
+                        let p = "\n      int xR; int xC; int xCOffset;\n      vec4 wTexel; vec4 previous; vec4 final;";
+                        for (let e = 0; e < c; e++) p += `\n          vec4 xTexelC${2*e};\n          int xTexelC${2*e}Ready;\n          vec4 xTexelC${2*e+1};\n          int xTexelC${2*e+1}Ready;\n          vec4 xC${e};`;
+                        p += `\n    for (int r = 0; r < ${l}; r++) {\n      `;
+                        for (let e = 0; e < c; e++) p += `\n          xTexelC${2*e} = vec4(0.0);\n          xTexelC${2*e}Ready = 0;\n          xTexelC${2*e+1} = vec4(0.0);\n          xTexelC${2*e+1}Ready = 0;\n          xC${e} = vec4(0.0);`;
+                        p += "\n        xR = xRCorner + r * dilations[0];\n        if (xR >=0 && xR < inDims[0]) {\n      ";
+                        for (let e = 0; e < (h + 1) / 2; e++) {
+                            const t = 2 * e;
+                            if (p += `\n          xC = xCCorner + ${t*u};\n          `, 1 === o) {
+                                if (t < c && (i % 2 == 1 ? (p += `\n                xCOffset = xC + 1;\n                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xCOffset, d1);\n\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n              `, p += 1 === u && t > 0 ? `\n                xC${t} = vec4(xTexelC${t-2}.zw, xTexelC${t}.xy);\n                ` : `\n                  xCOffset = xC + 1 - 2;\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                    previous = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      previous.zw = vec2(0.0);\n                    }\n\n                    xC${t} = vec4(previous.zw, xTexelC${t}.xy);\n                  } else {\n                    xC${t} = vec4(0.0, 0.0, xTexelC${t}.xy);\n                  }\n                  `) : p += `\n                if (xC >= 0 && xC < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                xC${t} = xTexelC${t};\n                `, t + 1 < c)) {
+                                    const e = i % 2 == 0 ? va.nearestLargerEven(u) : u;
+                                    u % 2 == 0 && i % 2 == 1 || u % 2 != 0 && i % 2 != 1 ? (p += `\n                  xCOffset = xC + imod(pads[1], 2) + ${e};\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                    xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      xTexelC${t+1}.zw = vec2(0.0);\n                    }\n                    xTexelC${t+1}Ready = 1;\n                  }\n                  `, u > 1 && (p += `\n                    xCOffset -= 2;\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                      xTexelC${t} = getX(batch, xR, xCOffset, d1);\n                      xTexelC${t}Ready = 1;\n                    }\n                    `), p += `\n                  xC${t+1} = vec4(xTexelC${t}.zw, xTexelC${t+1}.xy);\n                  `) : p += 1 === e ? `\n                    xC${t+1} = xTexelC${t};\n                    ` : `\n                    xCOffset = xC + ${e};\n\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                      xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n                      if (xCOffset + 1 >= inDims[1]) {\n                        xTexelC${t+1}.zw = vec2(0.0);\n                      }\n                      xTexelC${t+1}Ready = 1;\n                    }\n\n                    xC${t+1} = xTexelC${t+1};\n                    `
+                                }
+                            } else t < c && (i % 2 == 1 ? (p += `\n                xCOffset = xC + 1 - strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xCOffset, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${t+1}Ready == 0) {\n                  xTexelC${t+1} = getX(batch, xR, xC + 1, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xC + 2 >= inDims[1]) {\n                    xTexelC${t+1}.zw = vec2(0.0);\n                  }\n                  xTexelC${t+1}Ready = 1;\n                }\n\n                xC${t} = vec4(xTexelC${t}.zw, xTexelC${t+1}.zw);\n              `, t + 1 < c && (p += `\n                  final = vec4(0.0);\n                  xCOffset = xC + 1 + strides[1];\n                  if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                    final = getX(batch, xR, xCOffset, d1);\n                  }\n                  xC${t+1} = vec4(xTexelC${t+1}.xy, final.xy);\n                `)) : (p += `\n                if(xC >= 0 && xC < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                xCOffset = xC + strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                  xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t+1}.zw = vec2(0.);\n                  }\n                  xTexelC${t+1}Ready = 1;\n                }\n\n                xC${t} = vec4(\n                  xTexelC${t}.xy, xTexelC${t+1}.xy);\n              `, t + 1 < c && (p += `\n                  xC${t+1} = vec4(xTexelC${t}.zw, xTexelC${t+1}.zw);\n                `)));
+                            t < c && (p += `\n            wTexel = getW(r, ${t}, d1, q);\n            dotProd += xC${t} * vec4(wTexel.xz, wTexel.xz);\n          `, t + 1 < c && (p += `\n              wTexel = getW(r, ${t+1}, d1, q);\n              dotProd += xC${t+1} * vec4(wTexel.xz, wTexel.xz);\n            `))
+                        }
+                        p += "\n    }\n  ", p += "\n      }\n    ";
+                        let d = "",
+                            f = "";
+                        n && (d = r ? `vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }` : a ? `vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }` : `vec4 activation(vec4 x) {\n          ${n}\n        }`, f = "result = activation(result);");
+                        const m = t ? "result += getBiasAtOutCoords();" : "";
+                        t && this.variableNames.push("bias"), r && this.variableNames.push("preluActivationWeights"), a && this.variableNames.push("leakyreluAlpha"), this.userCode = `\n      ${d}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${s};\n        int q = d2 - d1 * ${s};\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n        vec4 dotProd = vec4(0.000000000000001);\n\n        ${p}\n\n        vec4 result = dotProd - vec4(0.000000000000001);\n        ${m}\n        ${f}\n        setOutput(result);\n      }\n    `
+                    }
+                };
+            var JP = {
+                kernelName: Ut,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dilations: u,
+                        dimRoundingMode: l
+                    } = r;
+                    let c = u;
+                    null == c && (c = [1, 1]), va.assert(jd.eitherStridesOrDilationsAreOne(i, c), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${c}'`);
+                    const h = jd.computeConv2DInfo(a.shape, s.shape, i, c, o, l, !0);
+                    let p;
+                    p = Ye().getBool("WEBGL_PACK_DEPTHWISECONV") && h.strideWidth <= 2 && h.outChannels / h.inChannels == 1 ? new ZP(h) : new QP(h);
+                    const d = [
+                        [h.padInfo.top, h.padInfo.left],
+                        [h.strideHeight, h.strideWidth],
+                        [h.dilationHeight, h.dilationWidth],
+                        [h.inHeight, h.inWidth]
+                    ];
+                    return n.runWebGLProgram(p, [a, s], "float32", d)
+                }
+            };
+            var e$ = {
+                kernelName: Wt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        dy: s
+                    } = t, {
+                        strides: i,
+                        dilations: o,
+                        pad: u,
+                        dimRoundingMode: l,
+                        filterShape: c
+                    } = r, h = jd.computeConv2DInfo(a.shape, c, i, o, u, l, !0), p = new class {
+                        constructor(e) {
+                            this.variableNames = ["x", "dy"], this.outputShape = e.filterShape;
+                            const t = e.strideHeight,
+                                n = e.strideWidth,
+                                r = e.padInfo.top,
+                                a = e.padInfo.left,
+                                s = e.outChannels / e.inChannels;
+                            this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int dm = coords.w;\n        int d2 = d1 * ${s} + dm;\n\n        float dotProd = 0.0;\n\n        // TO DO: Vec4 over the batch size\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yR = 0; yR < ${e.outHeight}; yR++) {\n            int xR = wR + yR * ${t} - ${r};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${e.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${a};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              float dyValue = getDy(b, yR, yC, d2);\n              float xValue = getX(b, xR, xC, d1);\n              dotProd += (xValue * dyValue);\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(h);
+                    return n.runWebGLProgram(p, [a, s], "float32")
+                }
+            };
+            var t$ = {
+                kernelName: Vt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        dilations: o,
+                        pad: u,
+                        dimRoundingMode: l,
+                        inputShape: c
+                    } = r, h = jd.computeConv2DInfo(c, s.shape, i, o, u, l, !0), p = new class {
+                        constructor(e) {
+                            this.variableNames = ["dy", "W"], this.outputShape = e.inShape;
+                            const t = e.filterHeight,
+                                n = e.filterWidth,
+                                r = e.strideHeight,
+                                a = e.strideWidth,
+                                s = t - 1 - e.padInfo.top,
+                                i = n - 1 - e.padInfo.left,
+                                o = e.outChannels / e.inChannels;
+                            this.userCode = `\n      const ivec2 pads = ivec2(${s}, ${i});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[3];\n        ivec2 dyCorner = coords.yz - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        float dotProd = 0.0;\n\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            // TO DO: Vec4 over the channelMul\n            for (int dm = 0; dm < ${o}; dm++) {\n              int d2 = d1 * ${o} + dm;\n              float xValue = getDy(batch, idyR, idyC, d2);\n              float wValue = getW(wRPerm, wCPerm, d1, dm);\n              dotProd += xValue * wValue;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(h);
+                    return n.runWebGLProgram(p, [a, s], "float32")
+                }
+            };
+            var n$ = {
+                kernelName: jt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        x: r
+                    } = t, a = [...r.shape, ...r.shape], s = va.sizeFromShape(r.shape), i = gL({
+                        inputs: {
+                            x: r
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [s]
+                        }
+                    }), o = new class {
+                        constructor(e) {
+                            this.variableNames = ["X"], this.outputShape = [e, e], this.userCode = "\n      void main() {\n          ivec2 coords = getOutputCoords();\n          float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;\n          setOutput(val);\n      }\n    "
+                        }
+                    }(s), u = n.runWebGLProgram(o, [i], i.dtype), l = gL({
+                        inputs: {
+                            x: u
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: a
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(i), n.disposeIntermediateTensorInfo(u), l
+                }
+            };
+            var r$ = {
+                kernelName: Gt,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s
+                    } = t, {
+                        strides: i,
+                        pad: o,
+                        dilations: u
+                    } = r, l = jd.computeDilation2DInfo(a.shape, s.shape, i, o, "NHWC", u);
+                    let c;
+                    const h = new class {
+                        constructor(e) {
+                            this.variableNames = ["x", "W"], this.outputShape = e.outShape;
+                            const {
+                                inHeight: t,
+                                inWidth: n,
+                                padInfo: r,
+                                strideHeight: a,
+                                strideWidth: s,
+                                filterHeight: i,
+                                filterWidth: o,
+                                dilationHeight: u,
+                                dilationWidth: l
+                            } = e, {
+                                top: c,
+                                left: h
+                            } = r;
+                            this.userCode = `\n      const ivec2 strides = ivec2(${a}, ${s});\n      const ivec2 pads = ivec2(${c}, ${h});\n      const float neg_infinity = -3.4e38;\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.w;\n        ivec2 outTopLeftCorner =\n            coords.yz * strides - pads;\n        int hBeg = outTopLeftCorner.x;\n        int wBeg = outTopLeftCorner.y;\n\n        float curVal = neg_infinity;\n        for (int h = 0; h < ${i}; h++) {\n          int hIn = hBeg + h * ${u};\n\n          if (hIn >= 0 && hIn < ${t}) {\n            for (int w = 0; w < ${o}; w++) {\n              int wIn = wBeg + w * ${l};\n\n              if (wIn >= 0 && wIn < ${n}) {\n                float xVal = getX(batch, hIn, wIn, d1);\n                float wVal = getW(h, w, d1);\n\n                float val = xVal + wVal;\n                if (val > curVal) {\n                  curVal = val;\n                }\n              }\n            }\n          }\n        }\n\n        float result = curVal;\n        setOutput(result);\n      }\n    `
+                        }
+                    }(l);
+                    c = n.runWebGLProgram(h, [a, s], "float32");
+                    const p = gL({
+                        inputs: {
+                            x: c
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: l.outShape
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(c), p
+                }
+            };
+            var a$ = {
+                    kernelName: Yt,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            equation: a
+                        } = r, s = t, {
+                            allDims: i,
+                            summedDims: o,
+                            idDims: u
+                        } = jd.decodeEinsumEquation(a, s.length);
+                        jd.checkEinsumDimSizes(i.length, u, s);
+                        const {
+                            path: l,
+                            steps: c
+                        } = jd.getEinsumComputePath(o, u), h = c.length;
+                        let p = null,
+                            d = i.length;
+                        const f = [];
+                        for (let e = 0; e < h; ++e) {
+                            for (const t of c[e]) {
+                                const {
+                                    permutationIndices: e,
+                                    expandDims: r
+                                } = jd.getEinsumPermutation(d, u[t]);
+                                let a;
+                                jd.isIdentityPermutation(e) ? a = s[t] : (a = IL({
+                                    inputs: {
+                                        x: s[t]
+                                    },
+                                    backend: n,
+                                    attrs: {
+                                        perm: e
+                                    }
+                                }), f.push(a));
+                                const i = a.shape.slice();
+                                for (let e = 0; e < r.length; ++e) i.splice(r[e], 0, 1);
+                                va.arraysEqual(a.shape, i) || (a = gL({
+                                    inputs: {
+                                        x: a
+                                    },
+                                    backend: n,
+                                    attrs: {
+                                        shape: i
+                                    }
+                                }), f.push(a)), null === p ? p = a : (p = fL({
+                                    inputs: {
+                                        a: a,
+                                        b: p
+                                    },
+                                    backend: n
+                                }), f.push(p))
+                            }
+                            e < h - 1 && (l[e] >= 0 && (p = kL({
+                                inputs: {
+                                    x: p
+                                },
+                                backend: n,
+                                attrs: {
+                                    axis: l[e] - (i.length - d),
+                                    keepDims: !1
+                                }
+                            }), f.push(p)), d--)
+                        }
+                        for (const e of f) e !== p && n.disposeIntermediateTensorInfo(e);
+                        return p
+                    }
+                },
+                s$ = iL({
+                    opSnippet: "return (x >= 0.0) ? x : (exp(x) - 1.0);",
+                    packedOpSnippet: "\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n"
+                }),
+                i$ = {
+                    kernelName: Xt,
+                    backendName: "webgl",
+                    kernelFunc: s$
+                },
+                o$ = {
+                    kernelName: Qt,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            dy: r,
+                            y: a
+                        } = t, s = Ye().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new YO("\n  vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));\n  return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));\n", r.shape, a.shape) : new KO("return (b >= 1.0) ? a : a * (b + 1.0);", r.shape, a.shape);
+                        return n.runWebGLProgram(s, [r, a], r.dtype)
+                    }
+                },
+                u$ = oL({
+                    opSnippet: "return float(a == b);",
+                    packedOpSnippet: "\n  return vec4(equal(a, b));\n",
+                    dtype: "bool",
+                    cpuKernelImpl: KD
+                }),
+                l$ = {
+                    kernelName: Jt,
+                    backendName: "webgl",
+                    kernelFunc: u$
+                },
+                c$ = iL({
+                    opSnippet: `\n  // Error function is calculated approximately with elementary function.\n  // See "Handbook of Mathematical Functions with Formulas,\n  // Graphs, and Mathematical Tables", Abramowitz and Stegun.\n  float p = ${jd.ERF_P};\n  float a1 = ${jd.ERF_A1};\n  float a2 = ${jd.ERF_A2};\n  float a3 = ${jd.ERF_A3};\n  float a4 = ${jd.ERF_A4};\n  float a5 = ${jd.ERF_A5};\n\n  float sign = sign(x);\n  x = abs(x);\n  float t = 1.0 / (1.0 + p * x);\n  return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));\n`
+                }),
+                h$ = {
+                    kernelName: Zt,
+                    backendName: "webgl",
+                    kernelFunc: c$
+                },
+                p$ = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return exp(x);\n",
+                    packedOpSnippet: "\n  vec4 result = exp(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n",
+                    cpuKernelImpl: YD,
+                    dtype: "float32"
+                }),
+                d$ = {
+                    kernelName: en,
+                    backendName: "webgl",
+                    kernelFunc: p$
+                };
+
+            function f$(e) {
+                const {
+                    inputs: t,
+                    attrs: n,
+                    backend: r
+                } = e, {
+                    dim: a
+                } = n, {
+                    input: s
+                } = t, i = s.shape.length, o = s.shape.slice();
+                let u = a;
+                return a < 0 && (va.assert(-(i + 1) <= a, () => `Axis must be in the interval [${-(i+1)}, ${i}]`), u = i + a + 1), o.splice(u, 0, 1), gL({
+                    inputs: {
+                        x: s
+                    },
+                    backend: r,
+                    attrs: {
+                        shape: o
+                    }
+                })
+            }
+            var m$ = {
+                    kernelName: tn,
+                    backendName: "webgl",
+                    kernelFunc: f$
+                },
+                g$ = "return exp(x) - 1.0;",
+                y$ = iL({
+                    opSnippet: g$,
+                    packedOpSnippet: g$,
+                    cpuKernelImpl: XD
+                }),
+                b$ = {
+                    kernelName: nn,
+                    backendName: "webgl",
+                    kernelFunc: y$
+                },
+                v$ = class {
+                    constructor(e, t, n) {
+                        this.variableNames = ["real", "imag"];
+                        const r = t[1];
+                        this.outputShape = t;
+                        const a = n ? "2.0 * " + Math.PI : "-2.0 * " + Math.PI,
+                            s = n ? r + ".0" : "1.0";
+                        let i;
+                        if ("real" === e) i = "return real * expR - imag * expI;";
+                        else {
+                            if ("imag" !== e) throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);
+                            i = "return real * expI + imag * expR;"
+                        }
+                        this.userCode = `\n      const float exponentMultiplier = ${a};\n\n      float unaryOpComplex(float real, float expR, float imag, float expI) {\n        ${i}\n      }\n\n      float mulMatDFT(int batch, int index) {\n        float indexRatio = float(index) / float(${r});\n        float exponentMultiplierTimesIndexRatio =\n            exponentMultiplier * indexRatio;\n\n        float result = 0.0;\n\n        for (int i = 0; i < ${r}; i++) {\n          // x = (-2|2 * PI / N) * index * i;\n          float x = exponentMultiplierTimesIndexRatio * float(i);\n          float expR = cos(x);\n          float expI = sin(x);\n          float real = getReal(batch, i);\n          float imag = getImag(batch, i);\n\n          result +=\n              unaryOpComplex(real, expR, imag, expI) / ${s};\n        }\n\n        return result;\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        setOutput(mulMatDFT(coords[0], coords[1]));\n      }\n    `
+                    }
+                };
+
+            function x$(e, t, n) {
+                const r = n.texData.get(e.dataId),
+                    a = va.sizeFromShape(e.shape),
+                    s = e.shape[e.shape.length - 1],
+                    i = gL({
+                        inputs: {
+                            x: e
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [a / s, s]
+                        }
+                    }),
+                    o = i.shape,
+                    u = new v$("real", o, t),
+                    l = new v$("imag", o, t),
+                    c = [{
+                        dataId: r.complexTensorInfos.real.dataId,
+                        dtype: r.complexTensorInfos.real.dtype,
+                        shape: o
+                    }, {
+                        dataId: r.complexTensorInfos.imag.dataId,
+                        dtype: r.complexTensorInfos.imag.dtype,
+                        shape: o
+                    }],
+                    h = n.runWebGLProgram(u, c, "float32"),
+                    p = n.runWebGLProgram(l, c, "float32"),
+                    d = ZO({
+                        inputs: {
+                            real: h,
+                            imag: p
+                        },
+                        backend: n
+                    });
+                n.disposeIntermediateTensorInfo(h), n.disposeIntermediateTensorInfo(p);
+                const f = gL({
+                    inputs: {
+                        x: d
+                    },
+                    backend: n,
+                    attrs: {
+                        shape: e.shape
+                    }
+                });
+                return n.disposeIntermediateTensorInfo(i), n.disposeIntermediateTensorInfo(d), f
+            }
+            var w$ = {
+                kernelName: rn,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        input: r
+                    } = t;
+                    return x$(r, !1, n)
+                }
+            };
+
+            function k$(e) {
+                const {
+                    backend: t,
+                    attrs: n
+                } = e, {
+                    shape: r,
+                    value: a
+                } = n;
+                let {
+                    dtype: s
+                } = n;
+                if (s = s || va.inferDtype(a), "string" === s) {
+                    const e = va.getArrayFromDType(s, va.sizeFromShape(r));
+                    return e.fill(a), t.makeTensorInfo(r, s, e)
+                } {
+                    const e = new class {
+                            constructor(e, t) {
+                                this.outputShape = [], this.customUniforms = [{
+                                    name: "value",
+                                    type: "float"
+                                }], this.variableNames = ["x"], this.outputShape = e, this.userCode = "\n      void main() {\n        // Input can be obtained from uniform value.\n        setOutput(value);\n      }\n    "
+                            }
+                        }(r, a),
+                        n = [
+                            [a]
+                        ];
+                    return t.runWebGLProgram(e, [], s, n)
+                }
+            }
+            var S$, I$ = {
+                    kernelName: an,
+                    backendName: "webgl",
+                    kernelFunc: k$
+                },
+                A$ = {
+                    kernelName: sn,
+                    backendName: "webgl",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t
+                    }) => {
+                        const {
+                            image: n
+                        } = e, r = t, a = new class {
+                            constructor(e) {
+                                this.variableNames = ["Image"], this.outputShape = [];
+                                const t = e[2];
+                                this.outputShape = e, this.userCode = `\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n\n          int coordX = ${t} - x - 1;\n          float outputValue;\n          if(coordX >= 0 && coordX < ${t}) {\n            outputValue = getImage(coords[0], coords[1], coordX, coords[3]);\n          } else {\n            outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `
+                            }
+                        }(n.shape);
+                        return r.runWebGLProgram(a, [n], n.dtype)
+                    }
+                },
+                E$ = "return floor(x);",
+                N$ = iL({
+                    opSnippet: E$,
+                    packedOpSnippet: E$,
+                    cpuKernelImpl: QD
+                }),
+                C$ = {
+                    kernelName: on,
+                    backendName: "webgl",
+                    kernelFunc: N$
+                },
+                T$ = oL({
+                    opSnippet: "\n  float s = sign(a) * sign(b);\n  int ia = round(a);\n  int ib = round(b);\n  if (ib != 0) {\n    // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n    return float(idiv(ia, ib, s));\n  } else {\n    return NAN;\n  }\n",
+                    packedOpSnippet: "\n  ivec4 ia = round(a);\n  ivec4 ib = round(b);\n  bvec4 cond = notEqual(ib, ivec4(0));\n  ivec4 result = ivec4(0);\n  vec4 s = sign(a) * sign(b);\n\n  // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n  if (cond[0]) {\n    result[0] = idiv(ia[0], ib[0], s[0]);\n  }\n  if (cond[1]) {\n    result[1] = idiv(ia[1], ib[1], s[1]);\n  }\n  if (cond[2]) {\n    result[2] = idiv(ia[2], ib[2], s[2]);\n  }\n  if (cond[3]) {\n    result[3] = idiv(ia[3], ib[3], s[3]);\n  }\n  return vec4(result);\n",
+                    dtype: "int32"
+                }),
+                R$ = {
+                    kernelName: un,
+                    backendName: "webgl",
+                    kernelFunc: T$
+                },
+                _$ = {
+                    kernelName: ta,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e;
+                        let {
+                            pixels: a
+                        } = t;
+                        const {
+                            numChannels: s
+                        } = r, i = "undefined" != typeof HTMLVideoElement && a instanceof HTMLVideoElement, o = "undefined" != typeof HTMLImageElement && a instanceof HTMLImageElement, [u, l] = i ? [a.videoWidth, a.videoHeight] : [a.width, a.height], c = [l, u], h = [l, u, s];
+                        (o || i) && (null == S$ && (S$ = document.createElement("canvas").getContext("2d")), S$.canvas.width = u, S$.canvas.height = l, S$.drawImage(a, 0, 0, u, l), a = S$.canvas);
+                        const p = n.makeTensorInfo(c, "int32");
+                        n.texData.get(p.dataId).usage = WF.PIXELS, n.gpgpu.uploadPixelDataToTexture(n.getTexture(p.dataId), a);
+                        const d = Ye().getBool("WEBGL_PACK") ? new class {
+                                constructor(e) {
+                                    this.variableNames = ["A"], this.packedInputs = !1, this.packedOutput = !0;
+                                    const t = XM(),
+                                        [n, r] = e;
+                                    this.outputShape = e, this.userCode = `\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n\n        vec4 result = vec4(0.);\n\n        for(int row=0; row<=1; row++) {\n          for(int col=0; col<=1; col++) {\n            texC = coords[1] + row;\n            depth = coords[2] + col;\n\n            vec2 uv = (vec2(texC, texR) + halfCR) /\n                       vec2(${r}.0, ${n}.0);\n            vec4 values = ${t.texture2D}(A, uv);\n            float value;\n            if (depth == 0) {\n              value = values.r;\n            } else if (depth == 1) {\n              value = values.g;\n            } else if (depth == 2) {\n              value = values.b;\n            } else if (depth == 3) {\n              value = values.a;\n            }\n\n            result[row * 2 + col] = floor(value * 255.0 + 0.5);\n          }\n        }\n\n        ${t.output} = result;\n      }\n    `
+                                }
+                            }(h) : new class {
+                                constructor(e) {
+                                    this.variableNames = ["A"];
+                                    const t = XM(),
+                                        [n, r] = e;
+                                    this.outputShape = e, this.userCode = `\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}.0, ${n}.0);\n\n        vec4 values = ${t.texture2D}(A, uv);\n        float value;\n        if (depth == 0) {\n          value = values.r;\n        } else if (depth == 1) {\n          value = values.g;\n        } else if (depth == 2) {\n          value = values.b;\n        } else if (depth == 3) {\n          value = values.a;\n        }\n\n        setOutput(floor(value * 255.0 + 0.5));\n      }\n    `
+                                }
+                            }(h),
+                            f = n.runWebGLProgram(d, [p], "int32");
+                        return n.disposeData(p.dataId), f
+                    }
+                };
+            var F$ = {
+                kernelName: aa,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t, {
+                        strides: u,
+                        pad: l,
+                        dataFormat: c,
+                        dilations: h,
+                        dimRoundingMode: p,
+                        activation: d,
+                        leakyreluAlpha: f
+                    } = r, m = jd.convertConv2DDataFormat(c), g = jd.computeConv2DInfo(a.shape, s.shape, u, h, l, p, !1, m);
+                    let y;
+                    const b = [];
+                    if (1 !== g.filterHeight || 1 !== g.filterWidth || 1 !== g.dilationHeight || 1 !== g.dilationWidth || 1 !== g.strideHeight || 1 !== g.strideWidth || "SAME" !== g.padInfo.type && "VALID" !== g.padInfo.type)
+                        if (Ye().getBool("WEBGL_CONV_IM2COL") && 1 === a.shape[0]) y = MP({
+                            x: a,
+                            filter: s,
+                            convInfo: g,
+                            backend: n,
+                            bias: i,
+                            activation: d,
+                            preluActivationWeights: o,
+                            leakyreluAlpha: f
+                        });
+                        else {
+                            const e = null != i,
+                                t = null != o,
+                                r = "leakyrelu" === d,
+                                u = d ? uL(d, !1) : null,
+                                l = new _P(g, e, u, t, r),
+                                c = [a, s];
+                            if (i && c.push(i), o && c.push(o), r) {
+                                const e = n.makeTensorInfo([], "float32", va.createScalarValue(f, "float32"));
+                                c.push(e), b.push(e)
+                            }
+                            y = n.runWebGLProgram(l, c, "float32")
+                        }
+                    else y = FP({
+                        x: a,
+                        filter: s,
+                        convInfo: g,
+                        backend: n,
+                        bias: i,
+                        activation: d,
+                        preluActivationWeights: o,
+                        leakyreluAlpha: f
+                    });
+                    const v = gL({
+                        inputs: {
+                            x: y
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: g.outShape
+                        }
+                    });
+                    return b.push(y), b.forEach(e => n.disposeIntermediateTensorInfo(e)), v
+                }
+            };
+            var M$ = {
+                kernelName: sa,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        filter: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t, {
+                        strides: u,
+                        pad: l,
+                        dilations: c,
+                        dimRoundingMode: h,
+                        activation: p,
+                        leakyreluAlpha: d
+                    } = r, f = [];
+                    let m = c;
+                    null == m && (m = [1, 1]), va.assert(jd.eitherStridesOrDilationsAreOne(u, m), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${u} and dilations '${m}'`);
+                    const g = jd.computeConv2DInfo(a.shape, s.shape, u, m, l, h, !0),
+                        y = Ye().getBool("WEBGL_PACK_DEPTHWISECONV") && g.strideWidth <= 2 && g.outChannels / g.inChannels == 1,
+                        b = p ? uL(p, y) : null,
+                        v = [a, s],
+                        x = null != i,
+                        w = null != o,
+                        k = "leakyrelu" === p;
+                    if (x && v.push(i), w && v.push(o), k) {
+                        const e = n.makeTensorInfo([], "float32", va.createScalarValue(d, "float32"));
+                        v.push(e), f.push(e)
+                    }
+                    let S;
+                    S = y ? new ZP(g, x, b, w, k) : new QP(g, x, b, w, k);
+                    const I = [
+                            [g.padInfo.top, g.padInfo.left],
+                            [g.strideHeight, g.strideWidth],
+                            [g.dilationHeight, g.dilationWidth],
+                            [g.inHeight, g.inWidth]
+                        ],
+                        A = n.runWebGLProgram(S, v, "float32", I);
+                    return f.forEach(e => n.disposeIntermediateTensorInfo(e)), A
+                }
+            };
+            var D$ = {
+                kernelName: hn,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        params: r,
+                        indices: a
+                    } = t, s = a.shape, i = s[s.length - 1], o = va.sizeFromShape(r.shape), [u, l, c, h] = jd.prepareAndValidate(r, a), p = gL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [l, i]
+                        }
+                    }), d = gL({
+                        inputs: {
+                            x: r
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [va.sizeFromShape(r.shape) / c, c]
+                        }
+                    });
+                    if (n.shouldExecuteOnCPU([r, a]) || "string" === r.dtype) {
+                        const e = n.readSync(a.dataId),
+                            t = n.bufferSync(r),
+                            s = ZD(e, t, r.dtype, l, i, c, h, r.shape, o);
+                        return n.makeTensorInfo(u, r.dtype, s.values)
+                    }
+                    const f = new class {
+                            constructor(e, t, n) {
+                                this.sliceDim = e, this.strides = t, this.variableNames = ["x", "indices"], this.outputShape = n;
+                                const r = dD(t.length),
+                                    a = dD(n.length),
+                                    s = this.sliceDim > 1 ? "strides[j]" : "strides";
+                                this.userCode = `\n        ${r} strides = ${r}(${this.strides});\n         void main() {\n          ${a} coords = getOutputCoords();\n          int flattenIndex = 0;\n          for (int j = 0; j < ${this.sliceDim}; j++) {\n            int index = round(getIndices(coords[0], j));\n            flattenIndex += index * ${s};\n          }\n          setOutput(getX(flattenIndex, coords[1]));\n        }\n      `
+                            }
+                        }(i, h, [l, c]),
+                        m = n.runWebGLProgram(f, [d, p], d.dtype),
+                        g = gL({
+                            inputs: {
+                                x: m
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: u
+                            }
+                        });
+                    return n.disposeIntermediateTensorInfo(p), n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(m), g
+                }
+            };
+
+            function O$(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a,
+                    indices: s
+                } = t, {
+                    axis: i,
+                    batchDims: o
+                } = r, u = va.parseAxisParam(i, a.shape)[0];
+                if (Ye().get("DEBUG")) {
+                    const e = n.readSync(s.dataId),
+                        t = a.shape[u];
+                    for (let n = 0; n < e.length; ++n) {
+                        const r = e[n];
+                        va.assert(r <= t - 1 && r >= 0, () => `GatherV2: the index value ${r} is not in [0, ${t-1}]`)
+                    }
+                }
+                const l = jd.segment_util.collectGatherOpShapeInfo(a, s, u, o),
+                    c = va.sizeFromShape(s.shape),
+                    h = [],
+                    p = gL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [l.batchSize, l.outerSize, l.dimSize, l.sliceSize]
+                        }
+                    }),
+                    d = gL({
+                        inputs: {
+                            x: s
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: [l.batchSize, c / l.batchSize]
+                        }
+                    });
+                h.push(p), h.push(d);
+                const f = [l.batchSize, l.outerSize, c / l.batchSize, l.sliceSize];
+                if (n.shouldExecuteOnCPU([a, s]) || "string" === a.dtype) {
+                    const e = n.bufferSync(d),
+                        t = n.bufferSync(p),
+                        r = JD(t, e, f);
+                    return h.forEach(e => n.disposeIntermediateTensorInfo(e)), n.makeTensorInfo(l.outputShape, r.dtype, r.values)
+                }
+                const m = new class {
+                        constructor(e, t) {
+                            this.variableNames = ["A", "indices"], this.outputShape = t, this.rank = t.length;
+                            const n = dD(this.rank),
+                                r = function (e, t) {
+                                    const n = ["resRC.x", "resRC.y", "resRC.z", "resRC.w"],
+                                        r = [];
+                                    for (let t = 0; t < e.length; t++) 2 === t ? r.push("index") : r.push("" + n[t]);
+                                    return r.join()
+                                }(e);
+                            this.userCode = `\n      void main() {\n        ${n} resRC = getOutputCoords();\n        int index = int(getIndices(resRC.x, resRC.z));\n        float inBounds = (index >= 0) && (index < ${e[2]}) ? 1.0 : 0.0;\n        setOutput(inBounds * getA(${r}));\n      }\n    `
+                        }
+                    }(p.shape, f),
+                    g = n.runWebGLProgram(m, [p, d], p.dtype);
+                h.push(g);
+                const y = gL({
+                    inputs: {
+                        x: g
+                    },
+                    backend: n,
+                    attrs: {
+                        shape: l.outputShape
+                    }
+                });
+                return h.forEach(e => n.disposeIntermediateTensorInfo(e)), y
+            }
+            var L$ = {
+                    kernelName: cn,
+                    backendName: "webgl",
+                    kernelFunc: O$
+                },
+                P$ = oL({
+                    opSnippet: "return float(a > b);",
+                    packedOpSnippet: "\n  return vec4(greaterThan(a, b));\n",
+                    cpuKernelImpl: eO,
+                    dtype: "bool"
+                }),
+                $$ = {
+                    kernelName: pn,
+                    backendName: "webgl",
+                    kernelFunc: P$
+                },
+                B$ = oL({
+                    opSnippet: "return float(a >= b);",
+                    packedOpSnippet: "\n  return vec4(greaterThanEqual(a, b));\n",
+                    dtype: "bool",
+                    cpuKernelImpl: tO
+                }),
+                z$ = {
+                    kernelName: dn,
+                    backendName: "webgl",
+                    kernelFunc: B$
+                };
+            var U$ = {
+                    kernelName: mn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            input: r
+                        } = t;
+                        return x$(r, !0, n)
+                    }
+                },
+                W$ = iL({
+                    opSnippet: "return float(!isnan(x) && !isinf(x));",
+                    dtype: "bool"
+                }),
+                V$ = {
+                    kernelName: yn,
+                    backendName: "webgl",
+                    kernelFunc: W$
+                },
+                j$ = iL({
+                    opSnippet: "return float(isinf(x));",
+                    dtype: "bool"
+                }),
+                G$ = {
+                    kernelName: bn,
+                    backendName: "webgl",
+                    kernelFunc: j$
+                },
+                H$ = iL({
+                    opSnippet: "return float(isnan(x));",
+                    dtype: "bool"
+                }),
+                q$ = {
+                    kernelName: vn,
+                    backendName: "webgl",
+                    kernelFunc: H$
+                },
+                K$ = oL({
+                    opSnippet: "return float(a < b);",
+                    packedOpSnippet: "\n  return vec4(lessThan(a, b));\n",
+                    cpuKernelImpl: nO,
+                    dtype: "bool"
+                }),
+                Y$ = {
+                    kernelName: wn,
+                    backendName: "webgl",
+                    kernelFunc: K$
+                },
+                X$ = oL({
+                    opSnippet: "return float(a <= b);",
+                    packedOpSnippet: "\n  return vec4(lessThanEqual(a, b));\n",
+                    cpuKernelImpl: rO,
+                    dtype: "bool"
+                }),
+                Q$ = {
+                    kernelName: kn,
+                    backendName: "webgl",
+                    kernelFunc: X$
+                };
+            var Z$ = {
+                    kernelName: Sn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            attrs: n
+                        } = e, {
+                            start: r,
+                            stop: a,
+                            num: s
+                        } = n, i = aO(r, a, s);
+                        return t.makeTensorInfo([i.length], "float32", i)
+                    }
+                },
+                J$ = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return x < 0.0 ? 0./0. : log(x);\n",
+                    packedOpSnippet: "\n  vec4 result = log(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);\n  result.g = isNaN.g ? x.g : (x.g < 0.0 ? 0./0. : result.g);\n  result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);\n  result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);\n  return result;\n",
+                    cpuKernelImpl: sO
+                }),
+                eB = {
+                    kernelName: In,
+                    backendName: "webgl",
+                    kernelFunc: J$
+                },
+                tB = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return log(1.0 + x);\n"
+                }),
+                nB = {
+                    kernelName: An,
+                    backendName: "webgl",
+                    kernelFunc: tB
+                },
+                rB = oL({
+                    opSnippet: "return float(a >= 1.0 && b >= 1.0);",
+                    packedOpSnippet: "\n  return vec4(\n    vec4(greaterThanEqual(a, vec4(1.0))) *\n    vec4(greaterThanEqual(b, vec4(1.0))));\n",
+                    dtype: "bool"
+                }),
+                aB = {
+                    kernelName: En,
+                    backendName: "webgl",
+                    kernelFunc: rB
+                },
+                sB = iL({
+                    opSnippet: "return float(!(x >= 1.0));"
+                }),
+                iB = {
+                    kernelName: Nn,
+                    backendName: "webgl",
+                    kernelFunc: sB
+                },
+                oB = oL({
+                    opSnippet: "return float(a >= 1.0 || b >= 1.0);",
+                    packedOpSnippet: "\n  return min(\n    vec4(greaterThanEqual(a, vec4(1.0))) +\n    vec4(greaterThanEqual(b, vec4(1.0))),\n    vec4(1.0));\n",
+                    dtype: "bool"
+                }),
+                uB = {
+                    kernelName: Cn,
+                    backendName: "webgl",
+                    kernelFunc: oB
+                },
+                lB = {
+                    kernelName: Rn,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            depthRadius: s,
+                            bias: i,
+                            alpha: o,
+                            beta: u
+                        } = r, l = Ye().getBool("WEBGL_PACK_NORMALIZATION") ? new class {
+                            constructor(e, t, n, r, a) {
+                                this.variableNames = ["x"], this.outputShape = [], this.packedInputs = !0, this.packedOutput = !0;
+                                const s = t,
+                                    i = e[3] - 1;
+                                let o;
+                                this.outputShape = e;
+                                const u = `float(${n}) + float(${r}) * sum`;
+                                o = .5 === a ? `inversesqrt(${u})` : 1 === a ? `1.0/(${u})` : `exp(log(${u}) * float(-${a}));`, this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords.x;\n        int r = coords.y;\n        int c = coords.z;\n        int d = coords.w;\n\n        bool hasNextCol = d < ${this.outputShape[3]};\n        bool hasNextRow = c < ${this.outputShape[2]};\n\n        vec4 sum = vec4(0.);\n        vec4 xFragAtOutputCoords = getX(b, r, c, d);\n\n        vec4 xAtOutputCoords = vec4(\n          getChannel(xFragAtOutputCoords, vec2(c, d)),\n          hasNextCol ?\n            getChannel(xFragAtOutputCoords, vec2(c, d + 1)) : 0.0,\n          hasNextRow ?\n            getChannel(xFragAtOutputCoords , vec2(c + 1, d)) : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0\n        );\n\n        int firstChannel = d - ${s};\n        vec2 cache = vec2(0.);\n        if(firstChannel >= 0){\n          vec4 firstChannelFrag = getX(b, r, c, firstChannel);\n          cache.x = getChannel(firstChannelFrag, vec2(c, firstChannel));\n            if(hasNextRow){\n              cache.y = getChannel(firstChannelFrag, vec2(c + 1, firstChannel));\n            }\n        }\n\n        ivec2 depth = ivec2(d, d + 1);\n        for (int j = - ${s}; j <= ${s}; j++) {\n          ivec2 idx = depth + j;\n          bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));\n          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${i}));\n\n          bool depthInRange = aboveLowerBound.x && belowUpperBound.x;\n          bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;\n\n          if(depthInRange || depthPlusOneInRange){\n            vec4 z = vec4(0.);\n            vec4 xFragAtCurrentDepth;\n            z.xz = cache.xy;\n            if(depthPlusOneInRange && hasNextCol){\n              xFragAtCurrentDepth = idx.y != d ?\n                getX(b, r, c, idx.y) : xFragAtOutputCoords;\n              z.y = getChannel(xFragAtCurrentDepth, vec2(c, idx.y));\n              if(hasNextRow){\n                z.w = getChannel(xFragAtCurrentDepth, vec2(c + 1, idx.y));\n              }\n            }\n            cache.xy = z.yw;\n            sum += z * z;\n          }\n        }\n        vec4 result = xAtOutputCoords * ${o};\n        setOutput(result);\n      }\n    `
+                            }
+                        }(a.shape, s, i, o, u) : new class {
+                            constructor(e, t, n, r, a) {
+                                this.variableNames = ["x"], this.outputShape = [];
+                                const s = t,
+                                    i = e[3] - 1;
+                                let o;
+                                this.outputShape = e;
+                                const u = `float(${n}) + float(${r}) * sum`;
+                                o = .5 === a ? `inversesqrt(${u})` : 1 === a ? `1.0/(${u})` : `exp(log(${u}) * float(-${a}));`, this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n        int d = coords[3];\n        float x = getX(b, r, c, d);\n        float sum = 0.0;\n        for (int j = -${s}; j <= ${s}; j++) {\n          int idx = d + j;\n          if (idx >= 0 && idx <=  ${i}) {\n            float z = getX(b, r, c, idx);\n            sum += z * z;\n          }\n        }\n        float val = x * ${o};\n        setOutput(val);\n      }\n    `
+                            }
+                        }(a.shape, s, i, o, u);
+                        return n.runWebGLProgram(l, [a], a.dtype)
+                    }
+                },
+                cB = {
+                    kernelName: _n,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a,
+                            y: s,
+                            dy: i
+                        } = t, {
+                            depthRadius: o,
+                            bias: u,
+                            alpha: l,
+                            beta: c
+                        } = r, h = new class {
+                            constructor(e, t, n, r, a) {
+                                this.variableNames = ["inputImage", "outputImage", "dy"], this.outputShape = [], this.outputShape = e, this.depth = e[3], this.depthRadius = t, this.bias = n, this.alpha = r, this.beta = a, this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n\n        float result = 0.0;\n        for (int d = 0; d < ${this.depth}; ++d) {\n          int depthBegin = int(max(0.0, float(d - ${t})));\n          int depthEnd = int(min(float(${this.depth}),\n              float(d + ${t} + 1)));\n\n          const int MIN_DEPTH_BEGIN = 0;\n          const int MAX_DEPTH_END = ${this.depth};\n\n          float norm = 0.0;\n          for (int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k) {\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd) {\n              norm += getInputImage(b, r, c, k) * getInputImage(b, r, c, k);\n            }\n            else {\n              break;\n            }\n          }\n\n          norm = float(${r}) * norm + float(${n});\n\n          for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd){\n              float dyi = -2.0 * float(${r})\n                * float(${a})\n                * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)\n                / norm;\n              if (k == d) {\n                dyi += pow(norm, -1.0 * ${a});\n              }\n              if (k == coords[3]) {\n                dyi *= getDy(b, r, c, d);\n                result += dyi;\n              }\n            }\n            else {\n              break;\n            }\n          }\n      }\n      setOutput(result);\n      }\n    `
+                            }
+                        }(a.shape, o, u, l, c);
+                        return n.runWebGLProgram(h, [a, s, i], a.dtype)
+                    }
+                };
+
+            function hB(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    reductionIndices: s,
+                    keepDims: i
+                } = r, o = a.shape.length, u = va.parseAxisParam(s, a.shape);
+                let l = u;
+                const c = jd.getAxesPermutation(l, o),
+                    h = null != c,
+                    p = n.shouldExecuteOnCPU([a]);
+                let d = a;
+                if (h) {
+                    if (p) {
+                        const e = n.texData.get(d.dataId).values,
+                            t = new Array(o);
+                        for (let e = 0; e < t.length; e++) t[e] = a.shape[c[e]];
+                        const r = TO(e, a.shape, a.dtype, c, t);
+                        d = n.makeTensorInfo(t, a.dtype);
+                        n.texData.get(d.dataId).values = r
+                    } else d = wL(a, c, n);
+                    l = jd.getInnerMostAxes(l.length, o)
+                }
+                jd.assertAxesAreInnerMostDims("max", l, o);
+                const [f, m] = jd.computeOutAndReduceShapes(d.shape, l);
+                let g, y = f;
+                if (i && (y = jd.expandShapeToKeepDim(f, u)), p) {
+                    const e = n.texData.get(d.dataId).values,
+                        t = iO(e, va.sizeFromShape(m), y, a.dtype);
+                    g = n.makeTensorInfo(y, a.dtype);
+                    n.texData.get(g.dataId).values = t
+                } else g = function (e, t, n, r) {
+                    const a = va.sizeFromShape(t),
+                        s = gL({
+                            inputs: {
+                                x: e
+                            },
+                            attrs: {
+                                shape: [va.sizeFromShape(e.shape) / a, a]
+                            },
+                            backend: r
+                        }),
+                        i = xL(s, e.dtype, "max", r),
+                        o = gL({
+                            inputs: {
+                                x: i
+                            },
+                            attrs: {
+                                shape: n
+                            },
+                            backend: r
+                        });
+                    return r.disposeIntermediateTensorInfo(s), r.disposeIntermediateTensorInfo(i), o
+                }(d, m, y, n);
+                return h && n.disposeIntermediateTensorInfo(d), g
+            }
+            var pB = {
+                    kernelName: Fn,
+                    backendName: "webgl",
+                    kernelFunc: hB
+                },
+                dB = oL({
+                    opSnippet: "\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n\n  return max(a, b);\n",
+                    packedOpSnippet: "\n  vec4 result = vec4(max(a, b));\n  vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n",
+                    cpuKernelImpl: oO
+                }),
+                fB = {
+                    kernelName: Mn,
+                    backendName: "webgl",
+                    kernelFunc: dB
+                };
+            var mB = {
+                kernelName: Dn,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t;
+                    KM(a, "maxPool");
+                    const {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dimRoundingMode: u
+                    } = r;
+                    va.assert(jd.eitherStridesOrDilationsAreOne(i, 1), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);
+                    const l = jd.computePool2DInfo(a.shape, s, i, 1, o, u);
+                    if (1 === l.filterWidth && 1 === l.filterHeight && va.arraysEqual(l.inShape, l.outShape)) return XO({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    const c = new eP(l, "max", !1);
+                    return n.runWebGLProgram(c, [a], a.dtype)
+                }
+            };
+            var gB = {
+                kernelName: Ln,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        filterSize: s,
+                        strides: i,
+                        pad: o,
+                        dataFormat: u,
+                        dimRoundingMode: l
+                    } = r, c = jd.computePool3DInfo(a.shape, s, i, [1, 1, 1], o, l, u), h = new tP(c, "max", !1);
+                    return n.runWebGLProgram(h, [a], a.dtype)
+                }
+            };
+            var yB = {
+                kernelName: Pn,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s
+                    } = t, i = s, {
+                        filterSize: o,
+                        strides: u,
+                        pad: l,
+                        dimRoundingMode: c
+                    } = r, h = jd.computePool3DInfo(i.shape, o, u, [1, 1, 1], l, c), p = new tP(h, "max", !0), d = n.runWebGLProgram(p, [i], i.dtype), f = new class {
+                        constructor(e) {
+                            this.variableNames = ["dy", "maxPos"], this.outputShape = e.inShape;
+                            const t = e.strideDepth,
+                                n = e.strideHeight,
+                                r = e.strideWidth,
+                                a = e.dilationDepth,
+                                s = e.dilationHeight,
+                                i = e.dilationWidth,
+                                o = e.effectiveFilterDepth,
+                                u = e.effectiveFilterHeight,
+                                l = e.effectiveFilterWidth,
+                                c = o - 1 - e.padInfo.front,
+                                h = u - 1 - e.padInfo.top,
+                                p = l - 1 - e.padInfo.left,
+                                d = o * u * l - 1;
+                            this.userCode = `\n      const ivec3 pads = ivec3(${c}, ${h}, ${p});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, ch) with pos mask(:, :, :, d) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${o};\n           wD += ${a}) {\n          float dyD = float(dyDCorner + wD) / ${t}.0;\n\n          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${u};\n              wR += ${s}) {\n            float dyR = float(dyRCorner + wR) / ${n}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${l};\n                wC += ${i}) {\n              float dyC = float(dyCCorner + wC) / ${r}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n              int maxPosValue = ${d} -\n                  int(getMaxPos(batch, idyD, idyR, idyC, ch));\n\n              // Get the current value, check it against the value from the\n              // position matrix.\n              int curPosValue =\n                  wD * ${u} * ${l} +\n                  wR * ${l} + wC;\n              float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n              dotProd += dyValue * mask;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(h), m = n.runWebGLProgram(f, [a, d], i.dtype);
+                    return n.disposeIntermediateTensorInfo(d), m
+                }
+            };
+            var bB = {
+                kernelName: On,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        dy: a,
+                        input: s,
+                        output: i
+                    } = t, o = s;
+                    KM([s, i], "maxPoolGrad");
+                    const {
+                        filterSize: u,
+                        strides: l,
+                        pad: c,
+                        dimRoundingMode: h
+                    } = r, p = jd.computePool2DInfo(o.shape, u, l, 1, c, h), d = new eP(p, "max", !0), f = n.runWebGLProgram(d, [o], o.dtype), m = new class {
+                        constructor(e) {
+                            this.variableNames = ["dy", "maxPos"], this.outputShape = e.inShape;
+                            const t = e.strideHeight,
+                                n = e.strideWidth,
+                                r = e.dilationHeight,
+                                a = e.effectiveFilterHeight,
+                                s = e.effectiveFilterWidth,
+                                i = a - 1 - e.padInfo.top,
+                                o = s - 1 - e.padInfo.left,
+                                u = a * s - 1;
+                            this.userCode = `\n      const ivec2 pads = ivec2(${i}, ${o});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${a};\n          wR += ${r}) {\n          float dyR = float(dyRCorner + wR) / ${t}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${s}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${n}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n            int maxPosValue = ${u} - int(getMaxPos(b, idyR, idyC, d));\n\n            // Get the current value, check it against the value from the\n            // position matrix.\n            int curPosValue = wR * ${s} + wC;\n            float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n            dotProd += dyValue * mask;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `
+                        }
+                    }(p), g = n.runWebGLProgram(m, [a, f], o.dtype);
+                    return n.disposeIntermediateTensorInfo(f), g
+                }
+            };
+            var vB = {
+                kernelName: $n,
+                backendName: "webgl",
+                kernelFunc: ({
+                    inputs: e,
+                    attrs: t,
+                    backend: n
+                }) => {
+                    const {
+                        x: r
+                    } = e, {
+                        filterSize: a,
+                        strides: s,
+                        pad: i,
+                        includeBatchInIndex: o
+                    } = t, u = n;
+                    va.assert(4 === r.shape.length, () => `Error in maxPool: input must be rank 4 but got rank ${r.shape.length}.`);
+                    const l = [1, 1];
+                    va.assert(jd.eitherStridesOrDilationsAreOne(s, l), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${l}'`);
+                    const c = jd.computePool2DInfo(r.shape, a, s, l, i),
+                        [h, p] = function (e, t, n, r) {
+                            let a = new eP(n, "max", !1);
+                            const s = r.runWebGLProgram(a, [e], "float32");
+                            return a = new eP(n, "max", !0, !0, t), [s, r.runWebGLProgram(a, [e], "float32")]
+                        }(r, o, c, u);
+                    return [h, p]
+                }
+            };
+            var xB = {
+                kernelName: Bn,
+                backendName: "webgl",
+                kernelFunc: ({
+                    inputs: e,
+                    attrs: t,
+                    backend: n
+                }) => {
+                    const {
+                        x: r
+                    } = e, {
+                        keepDims: a,
+                        axis: s
+                    } = t, i = n, o = r.shape.length, u = va.parseAxisParam(s, r.shape);
+                    let l = u;
+                    const c = jd.getAxesPermutation(l, o),
+                        h = null != c,
+                        p = i.shouldExecuteOnCPU([r]),
+                        d = [];
+                    let f = r;
+                    if (h) {
+                        if (p) {
+                            const e = i.texData.get(f.dataId).values,
+                                t = new Array(o);
+                            for (let e = 0; e < t.length; e++) t[e] = r.shape[c[e]];
+                            const n = TO(e, r.shape, r.dtype, c, t);
+                            f = i.makeTensorInfo(t, r.dtype);
+                            i.texData.get(f.dataId).values = n
+                        } else f = wL(r, c, i);
+                        d.push(f), l = jd.getInnerMostAxes(l.length, o)
+                    }
+                    jd.assertAxesAreInnerMostDims("sum", l, o);
+                    const [m, g] = jd.computeOutAndReduceShapes(f.shape, l);
+                    let y = m;
+                    a && (y = jd.expandShapeToKeepDim(m, u));
+                    const b = function (e, t, n, r) {
+                        const a = va.sizeFromShape(t),
+                            s = gL({
+                                inputs: {
+                                    x: e
+                                },
+                                attrs: {
+                                    shape: [va.sizeFromShape(e.shape) / a, a]
+                                },
+                                backend: r
+                            }),
+                            i = xL(s, "float32", "mean", r),
+                            o = gL({
+                                inputs: {
+                                    x: i
+                                },
+                                attrs: {
+                                    shape: n
+                                },
+                                backend: r
+                            });
+                        return r.disposeIntermediateTensorInfo(s), r.disposeIntermediateTensorInfo(i), o
+                    }(f, g, y, i);
+                    for (const e of d) i.disposeIntermediateTensorInfo(e);
+                    return b
+                }
+            };
+            var wB = {
+                    kernelName: zn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            axis: s,
+                            keepDims: i
+                        } = r, o = a.shape.length, u = va.parseAxisParam(s, a.shape);
+                        let l = u;
+                        const c = jd.getAxesPermutation(l, o);
+                        let h = a;
+                        null != c && (h = IL({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: c
+                            }
+                        }), l = jd.getInnerMostAxes(l.length, a.shape.length)), jd.assertAxesAreInnerMostDims("min", l, o);
+                        const [p, d] = jd.computeOutAndReduceShapes(h.shape, l), f = gL({
+                            inputs: {
+                                x: h
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [-1, va.sizeFromShape(d)]
+                            }
+                        }), m = xL(f, f.dtype, "min", n);
+                        let g;
+                        if (i) {
+                            g = gL({
+                                inputs: {
+                                    x: m
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: jd.expandShapeToKeepDim(p, u)
+                                }
+                            })
+                        } else g = gL({
+                            inputs: {
+                                x: m
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: p
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(m), null != c && n.disposeIntermediateTensorInfo(h), g
+                    }
+                },
+                kB = oL({
+                    opSnippet: "\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n\n  return min(a, b);\n",
+                    packedOpSnippet: "\n  vec4 result = vec4(min(a, b));\n  vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n",
+                    cpuKernelImpl: uO
+                }),
+                SB = {
+                    kernelName: Un,
+                    backendName: "webgl",
+                    kernelFunc: kB
+                },
+                IB = {
+                    kernelName: Wn,
+                    backendName: "webgl",
+                    kernelFunc: ({
+                        inputs: e,
+                        backend: t,
+                        attrs: n
+                    }) => {
+                        const {
+                            x: r
+                        } = e, {
+                            paddings: a,
+                            mode: s
+                        } = n, i = Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new class {
+                            constructor(e, t, n) {
+                                this.variableNames = ["x"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = t.map((t, n) => t[0] + e[n] + t[1]);
+                                const r = e.length,
+                                    a = dD(r),
+                                    s = t.map(e => e[0]).join(","),
+                                    i = t.map((t, n) => t[0] + e[n]).join(","),
+                                    o = FO("rc", r),
+                                    u = FO("source", r),
+                                    l = `${o[r-1]} < ${this.outputShape[r-1]}`,
+                                    c = 1 === r ? "source" : `vec2(${u.slice(-2).join()})`,
+                                    h = "reflect" === n ? 0 : 1;
+                                let p = "";
+                                if (1 === r) {
+                                    const e = `\n        ${a} source = rc;\n        if (source < start) {\n          source = start * 2 - source - ${h};\n        } else if (source >= end) {\n          source = (end - 1) * 2 - source + ${h};\n        }\n        source -= start;\n      `;
+                                    p = `\n        ${a} rc = outputLoc;\n        ${e}\n        result[0] = getChannel(getX(${u.join()}), ${c});\n        ${o[r-1]} += 1;\n        if(${l}) {\n          ${e}\n          result[1] = getChannel(getX(${u.join()}), ${c});\n        }\n      `
+                                } else {
+                                    const e = `\n        ${a} source = rc;\n        ${a} lt = ${a}(lessThan(source, start));\n        ${a} gte = ${a}(greaterThanEqual(source, end));\n        ${a} orig = 1 - (lt + gte);\n        source = orig * source +\n                lt * (start * 2 - source - ${h}) +\n                gte * ((end - 1) * 2 - source + ${h});\n        source -= start;\n      `;
+                                    p = `\n        ${a} rc = outputLoc;\n        ${e}\n        result[0] = getChannel(getX(${u.join()}), ${c});\n        ${o[r-1]} += 1;\n        if(${l}) {\n          ${e}\n          result[1] = getChannel(getX(${u.join()}), ${c});\n        }\n        rc = outputLoc;\n        ${o[r-2]} += 1;\n        if(${o[r-2]} < ${this.outputShape[r-2]}) {\n          ${e}\n          result[2] = getChannel(getX(${u.join()}), ${c});\n          ${o[r-1]} += 1;\n          if(${l}) {\n            ${e}\n            result[3] = getChannel(getX(${u.join()}), ${c});\n          }\n        }\n      `
+                                }
+                                this.userCode = `\n      const ${a} start = ${a}(${s});\n      const ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${p}\n        setOutput(result);\n      }\n    `
+                            }
+                        }(r.shape, a, s) : new class {
+                            constructor(e, t, n) {
+                                this.variableNames = ["x"], this.outputShape = t.map((t, n) => t[0] + e[n] + t[1]);
+                                const r = e.length,
+                                    a = dD(r),
+                                    s = t.map(e => e[0]).join(","),
+                                    i = t.map((t, n) => t[0] + e[n]).join(","),
+                                    o = ["coords[0]", "coords[1]", "coords[2]", "coords[3]"].slice(0, r),
+                                    u = "reflect" === n ? 0 : 1;
+                                this.userCode = 1 !== r ? `\n      ${a} start = ${a}(${s});\n      ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outC = getOutputCoords();\n        for (int i = 0; i < ${r}; i++) {\n          if (outC[i] < start[i]) {\n            outC[i] = start[i] * 2 - outC[i] - ${u};\n          } else if(outC[i] >= end[i]) {\n            outC[i] = (end[i] - 1) * 2 - outC[i] + ${u};\n          }\n        }\n        ${a} coords = outC - start;\n        setOutput(getX(${o}));\n      }\n    ` : `\n        int start = ${s};\n        int end = ${i};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start) {\n            outC = start * 2 - outC - ${u};\n          } else if(outC >= end) {\n            outC = (end - 1) * 2 - outC + ${u};\n          }\n          setOutput(getX(outC - start));\n        }\n      `
+                            }
+                        }(r.shape, a, s);
+                        return t.runWebGLProgram(i, [r], r.dtype)
+                    }
+                },
+                AB = oL({
+                    opSnippet: "if (b == 0.0) return NAN;\n  return mod(a, b);",
+                    packedOpSnippet: "\n  vec4 result = mod(a, b);\n  vec4 isNaN = vec4(equal(b, vec4(0.0)));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n"
+                }),
+                EB = {
+                    kernelName: Vn,
+                    backendName: "webgl",
+                    kernelFunc: AB
+                },
+                NB = oL({
+                    opSnippet: "\nif (a == b) {\n  return 1.0;\n};\nreturn a / b;",
+                    packedOpSnippet: "\n  // vec4 one = vec4(equal(a, b));\n  // return one + (vec4(1.0) - one) * a / b;\n  vec4 result = a / b;\n  if(a.x == b.x) {\n    result.x = 1.;\n  }\n  if(a.y == b.y) {\n    result.y = 1.;\n  }\n  if(a.z == b.z) {\n    result.z = 1.;\n  }\n  if(a.w == b.w) {\n    result.w = 1.;\n  }\n\n  return result;\n",
+                    checkOutOfBounds: !0
+                }),
+                CB = {
+                    kernelName: Kt,
+                    backendName: "webgl",
+                    kernelFunc: NB
+                },
+                TB = "return a - b;",
+                RB = oL({
+                    opSnippet: TB,
+                    packedOpSnippet: TB,
+                    supportsComplex: !0,
+                    cpuKernelImpl: EO
+                }),
+                _B = {
+                    kernelName: Vr,
+                    backendName: "webgl",
+                    kernelFunc: RB
+                };
+
+            function FB(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    logits: a
+                } = t, {
+                    dim: s
+                } = r, i = va.parseAxisParam([s], a.shape), o = hB({
+                    inputs: {
+                        x: a
+                    },
+                    backend: n,
+                    attrs: {
+                        reductionIndices: i,
+                        keepDims: !1
+                    }
+                }), u = jd.expandShapeToKeepDim(o.shape, i), l = gL({
+                    inputs: {
+                        x: o
+                    },
+                    backend: n,
+                    attrs: {
+                        shape: u
+                    }
+                }), c = RB({
+                    inputs: {
+                        a: a,
+                        b: l
+                    },
+                    backend: n
+                }), h = p$({
+                    inputs: {
+                        x: c
+                    },
+                    backend: n
+                }), p = kL({
+                    inputs: {
+                        x: h
+                    },
+                    backend: n,
+                    attrs: {
+                        axis: i,
+                        keepDims: !1
+                    }
+                }), d = gL({
+                    inputs: {
+                        x: p
+                    },
+                    backend: n,
+                    attrs: {
+                        shape: u
+                    }
+                }), f = NB({
+                    inputs: {
+                        a: h,
+                        b: d
+                    },
+                    backend: n
+                });
+                return n.disposeIntermediateTensorInfo(o), n.disposeIntermediateTensorInfo(l), n.disposeIntermediateTensorInfo(c), n.disposeIntermediateTensorInfo(h), n.disposeIntermediateTensorInfo(p), n.disposeIntermediateTensorInfo(d), f
+            }
+            var MB = {
+                kernelName: _r,
+                backendName: "webgl",
+                kernelFunc: FB
+            };
+            var DB = {
+                kernelName: jn,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        logits: a
+                    } = t, {
+                        numSamples: s,
+                        seed: i,
+                        normalized: o
+                    } = r, u = o ? a : FB({
+                        inputs: {
+                            logits: a
+                        },
+                        backend: n,
+                        attrs: {
+                            dim: a.shape.length - 1
+                        }
+                    }), l = u.shape[0], c = u.shape[1], h = new class {
+                        constructor(e, t, n) {
+                            this.variableNames = ["probs"], this.customUniforms = [{
+                                name: "seed",
+                                type: "float"
+                            }], this.outputShape = [e, n], this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n\n        float r = random(seed);\n        float cdf = 0.0;\n\n        for (int i = 0; i < ${t-1}; i++) {\n          cdf += getProbs(batch, i);\n\n          if (r < cdf) {\n            setOutput(float(i));\n            return;\n          }\n        }\n\n        // If no other event happened, last event happened.\n        setOutput(float(${t-1}));\n      }\n    `
+                        }
+                    }(l, c, s), p = [
+                        [i]
+                    ], d = n.runWebGLProgram(h, [u], "int32", p);
+                    return o || n.disposeIntermediateTensorInfo(u), d
+                }
+            };
+            var OB = {
+                    kernelName: Hn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            x: r
+                        } = t;
+                        if (n.shouldExecuteOnCPU([r])) {
+                            const e = n.texData.get(r.dataId),
+                                [t, a] = cO(e.values, r.shape, r.dtype);
+                            return n.makeTensorInfo(a, r.dtype, t)
+                        }
+                        let a;
+                        return a = Ye().getBool("WEBGL_PACK_UNARY_OPERATIONS") ? new zO(r.shape, "\n  vec4 result = -x;\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n") : new PO(r.shape, "if (isnan(x)) return x;\n  return -x;\n"), n.runWebGLProgram(a, [r], r.dtype)
+                    }
+                },
+                LB = jf.nonMaxSuppressionV3Impl;
+            var PB = {
+                    kernelName: Kn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        jd.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            boxes: a,
+                            scores: s
+                        } = t, {
+                            maxOutputSize: i,
+                            iouThreshold: o,
+                            scoreThreshold: u
+                        } = r, l = n.readSync(a.dataId), c = n.readSync(s.dataId), {
+                            selectedIndices: h
+                        } = LB(l, c, i, o, u);
+                        return n.makeTensorInfo([h.length], "int32", new Int32Array(h))
+                    }
+                },
+                $B = jf.nonMaxSuppressionV4Impl;
+            var BB = {
+                    kernelName: Yn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        jd.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            boxes: a,
+                            scores: s
+                        } = t, {
+                            maxOutputSize: i,
+                            iouThreshold: o,
+                            scoreThreshold: u,
+                            padToMaxOutputSize: l
+                        } = r, c = n.readSync(a.dataId), h = n.readSync(s.dataId), {
+                            selectedIndices: p,
+                            validOutputs: d
+                        } = $B(c, h, i, o, u, l);
+                        return [n.makeTensorInfo([p.length], "int32", new Int32Array(p)), n.makeTensorInfo([], "int32", new Int32Array([d]))]
+                    }
+                },
+                zB = jf.nonMaxSuppressionV5Impl;
+            var UB = {
+                    kernelName: Xn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        jd.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            boxes: a,
+                            scores: s
+                        } = t, {
+                            maxOutputSize: i,
+                            iouThreshold: o,
+                            scoreThreshold: u,
+                            softNmsSigma: l
+                        } = r, c = n.readSync(a.dataId), h = n.readSync(s.dataId), p = i, d = o, f = u, m = l, {
+                            selectedIndices: g,
+                            selectedScores: y
+                        } = zB(c, h, p, d, f, m);
+                        return [n.makeTensorInfo([g.length], "int32", new Int32Array(g)), n.makeTensorInfo([y.length], "float32", new Float32Array(y))]
+                    }
+                },
+                WB = {
+                    kernelName: Zn,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            indices: a
+                        } = t, {
+                            depth: s,
+                            onValue: i,
+                            offValue: o
+                        } = r, u = va.sizeFromShape(a.shape), l = new class {
+                            constructor(e, t, n, r) {
+                                this.variableNames = ["indices"], this.outputShape = [e, t], this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int index = round(getIndices(coords.x));\n        setOutput(mix(float(${r}), float(${n}),\n                      float(index == coords.y)));\n      }\n    `
+                            }
+                        }(u, s, i, o), c = gL({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [u]
+                            }
+                        }), h = n.runWebGLProgram(l, [c], a.dtype);
+                        n.disposeIntermediateTensorInfo(c);
+                        const p = gL({
+                            inputs: {
+                                x: h
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [...a.shape, s]
+                            }
+                        });
+                        return n.disposeIntermediateTensorInfo(h), p
+                    }
+                };
+
+            function VB(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, {
+                    x: r
+                } = t;
+                if ("complex64" === r.dtype) {
+                    const e = gP({
+                            inputs: {
+                                input: r
+                            },
+                            backend: n
+                        }),
+                        t = VB({
+                            inputs: {
+                                x: e
+                            },
+                            backend: n
+                        }),
+                        a = EP({
+                            inputs: {
+                                input: r
+                            },
+                            backend: n
+                        }),
+                        s = VB({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n
+                        }),
+                        i = ZO({
+                            inputs: {
+                                real: t,
+                                imag: s
+                            },
+                            backend: n
+                        });
+                    return n.disposeIntermediateTensorInfo(e), n.disposeIntermediateTensorInfo(t), n.disposeIntermediateTensorInfo(a), n.disposeIntermediateTensorInfo(s), i
+                }
+                return k$({
+                    attrs: {
+                        shape: r.shape,
+                        dtype: r.dtype,
+                        value: "string" === r.dtype ? "" : 0
+                    },
+                    backend: n
+                })
+            }
+            var jB = {
+                kernelName: Jr,
+                backendName: "webgl",
+                kernelFunc: VB
+            };
+            var GB = {
+                kernelName: Qn,
+                backendName: "webgl",
+                kernelFunc: function e(t) {
+                    const {
+                        inputs: n,
+                        backend: r
+                    } = t, {
+                        x: a
+                    } = n;
+                    if ("string" === a.dtype) throw new Error("onesLike is not supported under string dtype");
+                    if ("complex64" === a.dtype) {
+                        const t = gP({
+                                inputs: {
+                                    input: a
+                                },
+                                backend: r
+                            }),
+                            n = e({
+                                inputs: {
+                                    x: t
+                                },
+                                backend: r
+                            }),
+                            s = EP({
+                                inputs: {
+                                    input: a
+                                },
+                                backend: r
+                            }),
+                            i = VB({
+                                inputs: {
+                                    x: s
+                                },
+                                backend: r
+                            }),
+                            o = ZO({
+                                inputs: {
+                                    real: n,
+                                    imag: i
+                                },
+                                backend: r
+                            });
+                        return r.disposeIntermediateTensorInfo(t), r.disposeIntermediateTensorInfo(n), r.disposeIntermediateTensorInfo(s), r.disposeIntermediateTensorInfo(i), o
+                    }
+                    return k$({
+                        attrs: {
+                            shape: a.shape,
+                            dtype: a.dtype,
+                            value: 1
+                        },
+                        backend: r
+                    })
+                }
+            };
+            var HB = {
+                    kernelName: Jn,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            axis: a
+                        } = r;
+                        if (1 === t.length) return f$({
+                            inputs: {
+                                input: t[0]
+                            },
+                            backend: n,
+                            attrs: {
+                                dim: a
+                            }
+                        });
+                        const s = t[0].shape,
+                            i = t[0].dtype;
+                        t.forEach(e => {
+                            va.assertShapesMatch(s, e.shape, "All tensors passed to stack must have matching shapes"), va.assert(i === e.dtype, () => "All tensors passed to stack must have matching dtypes")
+                        });
+                        const o = [],
+                            u = TP({
+                                inputs: t.map(e => {
+                                    const t = f$({
+                                        inputs: {
+                                            input: e
+                                        },
+                                        backend: n,
+                                        attrs: {
+                                            dim: a
+                                        }
+                                    });
+                                    return o.push(t), t
+                                }),
+                                backend: n,
+                                attrs: {
+                                    axis: a
+                                }
+                            });
+                        return o.forEach(e => n.disposeIntermediateTensorInfo(e)), u
+                    }
+                },
+                qB = e => {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        paddings: s,
+                        constantValue: i
+                    } = r;
+                    if (0 === va.sizeFromShape(a.shape)) {
+                        return k$({
+                            backend: n,
+                            attrs: {
+                                shape: s.map((e, t) => e[0] + a.shape[t] + e[1]),
+                                value: i,
+                                dtype: a.dtype
+                            }
+                        })
+                    }
+                    const o = Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new class {
+                            constructor(e, t, n) {
+                                this.variableNames = ["x"], this.packedInputs = !0, this.packedOutput = !0, this.customUniforms = [{
+                                    name: "value",
+                                    type: "float"
+                                }], this.outputShape = t.map((t, n) => t[0] + e[n] + t[1]);
+                                const r = e.length,
+                                    a = dD(r),
+                                    s = t.map(e => e[0]).join(","),
+                                    i = t.map((t, n) => t[0] + e[n]).join(","),
+                                    o = FO("rc", r),
+                                    u = FO("source", r),
+                                    l = `${o[r-1]} < ${this.outputShape[r-1]}`,
+                                    c = 1 === r ? "source" : `vec2(${u.slice(-2).join()})`,
+                                    h = [a + " rc = outputLoc;", `${o[r-1]} += 1;\n       if(${l}) {\n      `, 1 === r ? "" : `}\n       rc = outputLoc;\n       ${o[r-2]} += 1;\n       if(${o[r-2]} < ${this.outputShape[r-2]}) {`, 1 === r ? "" : `  ${o[r-1]} += 1;\n         if(${l}) {`],
+                                    p = 1 === r ? "rc < start || rc >= end" : "any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))";
+                                let d = "";
+                                for (let e = 0, t = 1 === r ? 2 : 4; e < t; e++) d += `\n        ${h[e]}\n        if (${p}) {\n          result[${e}] = float(value);\n        } else {\n          ${a} source = rc - start;\n          result[${e}] = getChannel(getX(${u.join()}), ${c});\n        }\n      `;
+                                d += 1 === r ? "} " : "}}", this.userCode = `\n      const ${a} start = ${a}(${s});\n      const ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${d}\n        setOutput(result);\n      }\n    `
+                            }
+                        }(a.shape, s, i) : new class {
+                            constructor(e, t, n) {
+                                this.variableNames = ["x"], this.customUniforms = [{
+                                    name: "value",
+                                    type: "float"
+                                }], this.outputShape = t.map((t, n) => t[0] + e[n] + t[1]);
+                                const r = e.length,
+                                    a = dD(r),
+                                    s = t.map(e => e[0]).join(","),
+                                    i = t.map((t, n) => t[0] + e[n]).join(","),
+                                    o = ["coords[0]", "coords[1]", "coords[2]", "coords[3]"].slice(0, r);
+                                this.userCode = 1 !== r ? `\n      ${a} start = ${a}(${s});\n      ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outC = getOutputCoords();\n        if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {\n          setOutput(value);\n        } else {\n          ${a} coords = outC - start;\n          setOutput(getX(${o}));\n        }\n      }\n    ` : `\n        int start = ${s};\n        int end = ${i};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start || outC >= end) {\n            setOutput(value);\n          } else {\n            setOutput(getX(outC - start));\n          }\n        }\n      `
+                            }
+                        }(a.shape, s, i),
+                        u = [
+                            [i]
+                        ];
+                    return n.runWebGLProgram(o, [a], a.dtype, u)
+                },
+                KB = {
+                    kernelName: er,
+                    backendName: "webgl",
+                    kernelFunc: qB
+                },
+                YB = oL({
+                    opSnippet: "\n  if(a < 0.0 && floor(b) < b){\n    return NAN;\n  }\n  if (b == 0.0) {\n    return 1.0;\n  }\n  return (round(mod(b, 2.0)) != 1) ?\n      pow(abs(a), b) : sign(a) * pow(abs(a), b);\n",
+                    packedOpSnippet: "\n  // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.\n  vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));\n  vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);\n  vec4 result = multiplier * pow(abs(a), b);\n\n  // Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS\n  bvec4 isExpZero = equal(b, vec4(0.0));\n  result.r = isExpZero.r ? 1.0 : result.r;\n  result.g = isExpZero.g ? 1.0 : result.g;\n  result.b = isExpZero.b ? 1.0 : result.b;\n  result.a = isExpZero.a ? 1.0 : result.a;\n\n  vec4 isNaN = vec4(lessThan(a, vec4(0.0))) * vec4(lessThan(floor(b), b));\n  \n  result.r = isNaN.r > 0. ? NAN : result.r;\n  result.g = isNaN.g > 0. ? NAN : result.g;\n  result.b = isNaN.b > 0. ? NAN : result.b;\n  result.a = isNaN.a > 0. ? NAN : result.a;\n\n  return result;\n"
+                }),
+                XB = {
+                    kernelName: nr,
+                    backendName: "webgl",
+                    kernelFunc: YB
+                };
+            var QB = {
+                    kernelName: ar,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            axis: s,
+                            keepDims: i
+                        } = r, o = a.shape.length, u = [], l = va.parseAxisParam(s, a.shape);
+                        let c = l;
+                        const h = jd.getAxesPermutation(c, o);
+                        let p, d = a;
+                        if (null != h && (d = IL({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    perm: h
+                                }
+                            }), c = jd.getInnerMostAxes(c.length, o), u.push(d)), jd.assertAxesAreInnerMostDims("prod", c, o), n.shouldExecuteOnCPU([d])) {
+                            const e = n.texData.get(d.dataId).values,
+                                {
+                                    outVals: t,
+                                    outShape: r,
+                                    outDtype: a
+                                } = pO(d.shape, d.dtype, e, c);
+                            p = n.makeTensorInfo(r, a, t)
+                        } else {
+                            const [e, t] = jd.computeOutAndReduceShapes(d.shape, c), r = va.sizeFromShape(t), s = gL({
+                                inputs: {
+                                    x: d
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: [-1, r]
+                                }
+                            }), i = xL(s, hs(a.dtype), "prod", n);
+                            p = gL({
+                                inputs: {
+                                    x: i
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: e
+                                }
+                            }), u.push(s), u.push(i)
+                        }
+                        if (i) {
+                            u.push(p);
+                            const e = jd.expandShapeToKeepDim(p.shape, l);
+                            p = gL({
+                                inputs: {
+                                    x: p
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: e
+                                }
+                            })
+                        }
+                        return u.forEach(e => n.disposeIntermediateTensorInfo(e)), p
+                    }
+                },
+                ZB = e => {
+                    const {
+                        backend: t,
+                        attrs: n
+                    } = e, {
+                        start: r,
+                        stop: a,
+                        step: s,
+                        dtype: i
+                    } = n, o = dO(r, a, s, i);
+                    return t.makeTensorInfo([o.length], i, o)
+                },
+                JB = {
+                    kernelName: sr,
+                    backendName: "webgl",
+                    kernelFunc: ZB
+                },
+                ez = iL({
+                    opSnippet: "return 1.0 / x;"
+                }),
+                tz = {
+                    kernelName: or,
+                    backendName: "webgl",
+                    kernelFunc: ez
+                },
+                nz = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : x;\n",
+                    packedOpSnippet: "\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"
+                }),
+                rz = {
+                    kernelName: ur,
+                    backendName: "webgl",
+                    kernelFunc: nz
+                },
+                az = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n",
+                    packedOpSnippet: "\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"
+                }),
+                sz = {
+                    kernelName: fr,
+                    backendName: "webgl",
+                    kernelFunc: az
+                };
+            var iz = {
+                kernelName: pr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a
+                    } = t, {
+                        alignCorners: s,
+                        halfPixelCenters: i,
+                        size: o
+                    } = r, [u, l] = o, c = Ye().getBool("WEBGL_PACK_IMAGE_OPERATIONS") ? new class {
+                        constructor(e, t, n, r, a) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = [];
+                            const [s, i, o, u] = e;
+                            this.outputShape = [s, t, n, u];
+                            const l = [r && t > 1 ? i - 1 : i, r && n > 1 ? o - 1 : o],
+                                c = [r && t > 1 ? t - 1 : t, r && n > 1 ? n - 1 : n];
+                            let h;
+                            h = a ? "(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)" : "vec3(yRC) * effectiveInputOverOutputRatioRC", this.userCode = `\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]},\n          ${l[1]/c[1]});\n      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,\n                                     ${o}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${h};\n\n        // Compute the four integer indices.\n        ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));\n        ivec3 sourceCeilRC = ivec3(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${u-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        // In parallel, construct four corners for all four components in\n        // packed 2x2 cell.\n        vec4 topLeft = vec4(\n          getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 bottomLeft = vec4(\n          getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 topRight = vec4(\n          getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec4 bottomRight = vec4(\n          getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec3 fracRC = sourceFracIndexRC - vec3(sourceFloorRC);\n\n        vec4 top = mix(topLeft, topRight, fracRC.yyzz);\n        vec4 bottom = mix(bottomLeft, bottomRight, fracRC.yyzz);\n        vec4 newValue = mix(top, bottom, fracRC.x);\n\n        setOutput(newValue);\n      }\n    `
+                        }
+                    }(a.shape, u, l, s, i) : new class {
+                        constructor(e, t, n, r, a) {
+                            this.variableNames = ["A"], this.outputShape = [];
+                            const [s, i, o, u] = e;
+                            this.outputShape = [s, t, n, u];
+                            const l = [r && t > 1 ? i - 1 : i, r && n > 1 ? o - 1 : o],
+                                c = [r && t > 1 ? t - 1 : t, r && n > 1 ? n - 1 : n];
+                            let h;
+                            h = a ? "(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)" : "vec2(yRC) * effectiveInputOverOutputRatioRC", this.userCode = `\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]});\n      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${h};\n\n        // Compute the four integer indices.\n        ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));\n        ivec2 sourceCeilRC = ivec2(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        float topLeft = getA(b, sourceFloorRC.x, sourceFloorRC.y, d);\n        float bottomLeft = getA(b, sourceCeilRC.x, sourceFloorRC.y, d);\n        float topRight = getA(b, sourceFloorRC.x, sourceCeilRC.y, d);\n        float bottomRight = getA(b, sourceCeilRC.x, sourceCeilRC.y, d);\n\n        vec2 fracRC = sourceFracIndexRC - vec2(sourceFloorRC);\n\n        float top = topLeft + (topRight - topLeft) * fracRC.y;\n        float bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;\n        float newValue = top + (bottom - top) * fracRC.x;\n\n        setOutput(newValue);\n      }\n    `
+                        }
+                    }(a.shape, u, l, s, i);
+                    return n.runWebGLProgram(c, [a], "float32")
+                }
+            };
+            var oz = {
+                kernelName: dr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a,
+                        dy: s
+                    } = t, {
+                        alignCorners: i
+                    } = r, o = new class {
+                        constructor(e, t, n) {
+                            this.variableNames = ["dy"], this.outputShape = [], this.outputShape = t;
+                            const [, r, a] = t, [, s, i] = e, o = [n && s > 1 ? r - 1 : r, n && i > 1 ? a - 1 : a], u = [n && s > 1 ? s - 1 : s, n && i > 1 ? i - 1 : i], l = o[0] / u[0], c = o[1] / u[1], h = 1 / l, p = 1 / c, d = 2 * Math.ceil(h) + 2, f = 2 * Math.ceil(p) + 2;
+                            this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${l});\n        const float widthScale = float(${c});\n\n        const float invHeightScale = float(${h});\n        const float invWidthScale = float(${p});\n\n        const int winHeight = int(${d});\n        const int winWidth = int(${f});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(startRLerp - float(winHeight / 2));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(startCLerp - float(winWidth / 2));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${s}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${i}) {\n              continue;\n            }\n\n            float dxR = float(dyR) * heightScale;\n            int topDxRIndex = int(floor(dxR));\n            int bottomDxRIndex = int(min(ceil(dxR), ${r-1}.0));\n            float dxRLerp = dxR - float(topDxRIndex);\n            float inverseDxRLerp = 1.0 - dxRLerp;\n\n            float dxC = float(dyC) * widthScale;\n            int leftDxCIndex = int(floor(dxC));\n            int rightDxCIndex = int(min(ceil(dxC), ${a-1}.0));\n            float dxCLerp = dxC - float(leftDxCIndex);\n            float inverseDxCLerp = 1.0 - dxCLerp;\n\n            if (r == topDxRIndex && c == leftDxCIndex) {\n              // topLeft\n              accumulator +=\n                getDy(b, dyR, dyC, d) * inverseDxRLerp * inverseDxCLerp;\n            }\n\n            if (r == topDxRIndex && c == rightDxCIndex) {\n              // topRight\n              accumulator += getDy(b, dyR, dyC, d) * inverseDxRLerp * dxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == leftDxCIndex) {\n              // bottomLeft\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * inverseDxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == rightDxCIndex) {\n              // bottomRight\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * dxCLerp;\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `
+                        }
+                    }(s.shape, a.shape, i);
+                    return n.runWebGLProgram(o, [s], s.dtype)
+                }
+            };
+            var uz = {
+                kernelName: cr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a
+                    } = t, {
+                        alignCorners: s,
+                        halfPixelCenters: i,
+                        size: o
+                    } = r, [u, l] = o, c = Ye().getBool("WEBGL_PACK_IMAGE_OPERATIONS") ? new class {
+                        constructor(e, t, n, r, a) {
+                            this.variableNames = ["A"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = [];
+                            const [s, i, o, u] = e;
+                            this.outputShape = [s, t, n, u];
+                            const l = [r && t > 1 ? i - 1 : i, r && n > 1 ? o - 1 : o],
+                                c = [r && t > 1 ? t - 1 : t, r && n > 1 ? n - 1 : n],
+                                h = r ? "0.5" : "0.0";
+                            let p;
+                            p = a ? "max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))" : "vec3(yRC) * effectiveInputOverOutputRatioRC", this.userCode = `\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]},\n          ${l[1]/c[1]});\n      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,\n                                     ${o}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${p};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec3 sourceNearestRC = ivec3(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${h})));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${u-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        vec4 newValue = vec4(\n          getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),\n          hasNextCol ? getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d + 1) : 0.0);\n\n        setOutput(newValue);\n      }\n    `
+                        }
+                    }(a.shape, u, l, s, i) : new class {
+                        constructor(e, t, n, r, a) {
+                            this.variableNames = ["A"], this.outputShape = [];
+                            const [s, i, o, u] = e;
+                            this.outputShape = [s, t, n, u];
+                            const l = [r && t > 1 ? i - 1 : i, r && n > 1 ? o - 1 : o],
+                                c = [r && t > 1 ? t - 1 : t, r && n > 1 ? n - 1 : n],
+                                h = r ? "0.5" : "0.0";
+                            let p;
+                            p = a ? "max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))" : "vec2(yRC) * effectiveInputOverOutputRatioRC", this.userCode = `\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${l[0]/c[0]},\n          ${l[1]/c[1]});\n      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${p};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec2 sourceNearestRC = ivec2(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${h})));\n        float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);\n\n        setOutput(newValue);\n      }\n    `
+                        }
+                    }(a.shape, u, l, s, i);
+                    return n.runWebGLProgram(c, [a], a.dtype)
+                }
+            };
+            var lz = {
+                kernelName: hr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        images: a,
+                        dy: s
+                    } = t, {
+                        alignCorners: i
+                    } = r, o = new class {
+                        constructor(e, t, n) {
+                            this.variableNames = ["dy"], this.outputShape = [], this.outputShape = t;
+                            const [, r, a] = t, [, s, i] = e, o = [n && s > 1 ? r - 1 : r, n && i > 1 ? a - 1 : a], u = [n && s > 1 ? s - 1 : s, n && i > 1 ? i - 1 : i], l = o[0] / u[0], c = o[1] / u[1], h = 1 / l, p = 1 / c, d = 2 * Math.ceil(h) + 2, f = 2 * Math.ceil(p) + 2;
+                            this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${l});\n        const float widthScale = float(${c});\n\n        const float invHeightScale = float(${h});\n        const float invWidthScale = float(${p});\n\n        const int winHeight = int(${d});\n        const int winWidth = int(${f});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(floor(startRLerp - float(winHeight / 2)));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(floor(startCLerp - float(winWidth / 2)));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${s}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${i}) {\n              continue;\n            }\n\n            float sourceFracRow =\n              float(${o[0]}) *\n                (float(dyR) / float(${u[0]}));\n\n            float sourceFracCol =\n                float(${o[1]}) *\n                  (float(dyC) / float(${u[1]}));\n\n            int sourceNearestRow = int(min(\n                float(int(${r}) - 1),\n                ${n} ? float(round(sourceFracRow)) :\n                                  float(floor(sourceFracRow))));\n\n            int sourceNearestCol = int(min(\n                float(int(${a}) - 1),\n                ${n} ? float(round(sourceFracCol)) :\n                                  float(floor(sourceFracCol))));\n\n            if (r == sourceNearestRow && c == sourceNearestCol) {\n              accumulator += getDy(b, dyR, dyC, d);\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `
+                        }
+                    }(s.shape, a.shape, i);
+                    return n.runWebGLProgram(o, [s], s.dtype)
+                }
+            };
+            var cz = {
+                    kernelName: mr,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            dims: s
+                        } = r, i = a.shape.length, o = va.parseAxisParam(s, a.shape);
+                        if (0 === i) return XO({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n
+                        });
+                        const u = Ye().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new class {
+                            constructor(e, t) {
+                                this.variableNames = ["x"], this.packedInputs = !0, this.packedOutput = !0;
+                                const n = e.length;
+                                if (n > 4) throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);
+                                this.outputShape = e;
+                                const r = FO("rc", n),
+                                    a = `${r[n-1]} + 1 < ${this.outputShape[n-1]}`,
+                                    s = `${r[n-2]} + 1 < ${this.outputShape[n-2]}`,
+                                    i = dD(n);
+                                var o;
+
+                                function u(n) {
+                                    const r = e.map((r, a) => function (n, r) {
+                                        return -1 !== t.indexOf(n) && 1 !== e[n] ? `${e[n]} - ${r[n]} - 1` : "" + r[n]
+                                    }(a, n));
+                                    return `getChannel(getX(${r.join(",")}), vec2(${r.slice(-2).join(",")}))`
+                                }
+                                this.userCode = 1 === n ? `\n        void main(){\n          int rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = getChannel(getX(${e[0]} - rc - 1),\n            ${e[0]} - rc - 1);\n          if(${a}){\n              result.g = getChannel(getX(${e[0]} - (rc  + 1) - 1),\n                ${e[0]} - (rc  + 1) - 1);\n          }\n          setOutput(result);\n        }\n      ` : `\n        void main() {\n          ${i} rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = ${o=r.slice(),u(o)};\n          if(${a}){\n            result.g = ${function(e){return e[n-1]="("+e[n-1]+" + 1)",u(e)}(r.slice())};\n          }\n          if(${s}) {\n            result.b = ${function(e){return e[n-2]="("+e[n-2]+" + 1)",u(e)}(r.slice())};\n            if(${a}) {\n              result.a = ${function(e){return e[n-1]="("+e[n-1]+" + 1)",e[n-2]="("+e[n-2]+" + 1)",u(e)}(r.slice())};\n            }\n          }\n          setOutput(result);\n        }\n    `
+                            }
+                        }(a.shape, o) : new class {
+                            constructor(e, t) {
+                                this.variableNames = ["x"];
+                                const n = e.length;
+                                if (n > 4) throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);
+                                if (this.outputShape = e, 1 === n) return void(this.userCode = `\n        void main() {\n          int coord = getOutputCoords();\n          setOutput(getX(${e[0]} - coord - 1));\n        }\n      `);
+                                const r = e.map((n, r) => (n => -1 !== t.indexOf(n) && 1 !== e[n] ? `${e[n]} - coords[${n}] - 1` : `coords[${n}]`)(r)).join(","),
+                                    a = dD(n);
+                                this.userCode = `\n      void main() {\n        ${a} coords = getOutputCoords();\n        setOutput(getX(${r}));\n      }\n    `
+                            }
+                        }(a.shape, o);
+                        return n.runWebGLProgram(u, [a], a.dtype)
+                    }
+                },
+                hz = {
+                    kernelName: na,
+                    backendName: "webgl",
+                    kernelFunc: ({
+                        inputs: e,
+                        attrs: t,
+                        backend: n
+                    }) => {
+                        const {
+                            image: r
+                        } = e, {
+                            radians: a,
+                            fillValue: s,
+                            center: i
+                        } = t, o = n, u = new class {
+                            constructor(e, t) {
+                                this.variableNames = ["Image"], this.outputShape = [], this.customUniforms = [{
+                                    name: "params",
+                                    type: "vec4"
+                                }];
+                                const n = e[1],
+                                    r = e[2];
+                                this.outputShape = e;
+                                let a = "";
+                                a = "number" == typeof t ? `float outputValue = ${t.toFixed(2)};` : `\n        vec3 fill = vec3(${t.join(",")});\n        float outputValue = fill[coords[3]];`, this.userCode = `\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n          int y = coords[1];\n          float coordXFloat = (float(x) - params[0]) * params[3] -\n            (float(y) - params[1]) * params[2];\n          float coordYFloat = (float(x) - params[0]) * params[2] +\n            (float(y) - params[1]) * params[3];\n          int coordX = int(round(coordXFloat + params[0]));\n          int coordY = int(round(coordYFloat + params[1]));\n          ${a}\n          if(coordX >= 0 && coordX < ${r} && coordY >= 0 && coordY < ${n}) {\n            outputValue = getImage(coords[0], coordY, coordX, coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `
+                            }
+                        }(r.shape, s), [l, c] = jd.getImageCenter(i, r.shape[1], r.shape[2]), h = [
+                            [l, c, Math.sin(a), Math.cos(a)]
+                        ];
+                        return o.runWebGLProgram(u, [r], r.dtype, h)
+                    }
+                },
+                pz = iL({
+                    opSnippet: "\n  // OpenGL ES does not support round function.\n  // The algorithm is based on banker's rounding.\n  float base = floor(x);\n  if ((x - base) < 0.5) {\n    return floor(x);\n  } else if ((x - base) > 0.5) {\n    return ceil(x);\n  } else {\n    if (mod(base, 2.0) == 0.0) {\n      return base;\n    } else {\n      return base + 1.0;\n    }\n  }\n"
+                }),
+                dz = {
+                    kernelName: gr,
+                    backendName: "webgl",
+                    kernelFunc: pz
+                },
+                fz = iL({
+                    opSnippet: "return inversesqrt(x);",
+                    cpuKernelImpl: fO
+                }),
+                mz = {
+                    kernelName: yr,
+                    backendName: "webgl",
+                    kernelFunc: fz
+                },
+                gz = class {
+                    constructor(e, t, n, r, a, s, i = !0) {
+                        this.variableNames = ["updates", "indices", "defaultValue"], this.outputShape = s;
+                        const o = dD(a.length),
+                            u = dD(s.length);
+                        let l = "";
+                        1 === n ? l = "i" : 2 === n && (l = "i, j");
+                        const c = `getIndices(${l})`;
+                        let h = "";
+                        1 === r ? h = "i" : 2 === r && (h = "i, coords[1]");
+                        const p = `getUpdates(${h})`,
+                            d = t > 1 ? "strides[j]" : "strides";
+                        this.userCode = `\n        ${o} strides = ${o}(${a});\n\n        void main() {\n          ${u} coords = getOutputCoords();\n          float sum = 0.0;\n          bool found = false;\n          for (int i = 0; i < ${e}; i++) {\n            int flattenedIndex = 0;\n            for (int j = 0; j < ${t}; j++) {\n              int index = round(${c});\n              flattenedIndex += index * ${d};\n            }\n            if (flattenedIndex == coords[0]) {\n              sum += ${p};\n              found = true;\n            }\n          }\n          setOutput(mix(getDefaultValue(), sum, float(found)));\n        }\n      `
+                    }
+                };
+            var yz = {
+                kernelName: br,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        indices: a,
+                        updates: s
+                    } = t, {
+                        shape: i
+                    } = r, {
+                        sliceRank: o,
+                        numUpdates: u,
+                        sliceSize: l,
+                        strides: c,
+                        outputSize: h
+                    } = jd.calculateShapes(s, a, i), p = [h / l, l];
+                    if (0 === h) return n.makeTensorInfo(i, a.dtype);
+                    const d = gL({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [u, o]
+                            }
+                        }),
+                        f = gL({
+                            inputs: {
+                                x: s
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [u, l]
+                            }
+                        }),
+                        m = n.makeTensorInfo([], "float32", new Float32Array([0])),
+                        g = new gz(u, o, d.shape.length, f.shape.length, c, p),
+                        y = n.runWebGLProgram(g, [f, d, m], f.dtype),
+                        b = gL({
+                            inputs: {
+                                x: y
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: i
+                            }
+                        });
+                    return n.disposeIntermediateTensorInfo(d), n.disposeIntermediateTensorInfo(f), n.disposeIntermediateTensorInfo(y), n.disposeIntermediateTensorInfo(m), b
+                }
+            };
+            var bz = {
+                    kernelName: vr,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            condition: r,
+                            t: a,
+                            e: s
+                        } = t, i = new class {
+                            constructor(e, t, n) {
+                                let r, a;
+                                if (this.variableNames = ["c", "a", "b"], this.outputShape = t, n > 4) throw Error(`Where for rank ${n} is not yet supported`);
+                                if (1 === n) a = "resRC", r = "resRC";
+                                else {
+                                    const n = ["resRC.x", "resRC.y", "resRC.z", "resRC.w"],
+                                        s = [],
+                                        i = [];
+                                    for (let r = 0; r < t.length; r++) i.push("" + n[r]), r < e && s.push("" + n[r]);
+                                    r = s.join(), a = i.join()
+                                }
+                                const s = dD(n);
+                                this.userCode = `\n      void main() {\n        ${s} resRC = getOutputCoords();\n        float cVal = getC(${r});\n        if (cVal >= 1.0) {\n          setOutput(getA(${a}));\n        } else {\n          setOutput(getB(${a}));\n        }\n      }\n    `
+                            }
+                        }(r.shape.length, a.shape, a.shape.length);
+                        return n.runWebGLProgram(i, [r, a, s], cs(a.dtype, s.dtype))
+                    }
+                },
+                vz = iL({
+                    opSnippet: `\n  // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.\n  // see: https://arxiv.org/abs/1706.02515\n  float scaleAlpha = ${jd.SELU_SCALEALPHA};\n  float scale = ${jd.SELU_SCALE};\n  return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);\n`
+                }),
+                xz = {
+                    kernelName: xr,
+                    backendName: "webgl",
+                    kernelFunc: vz
+                },
+                wz = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return 1.0 / (1.0 + exp(-1.0 * x));\n",
+                    packedOpSnippet: "\n  vec4 result = 1.0 / (1.0 + exp(-1.0 * x));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n",
+                    cpuKernelImpl: mO
+                }),
+                kz = {
+                    kernelName: Ar,
+                    backendName: "webgl",
+                    kernelFunc: wz
+                },
+                Sz = iL({
+                    opSnippet: "\n  if (isnan(x)) { return 0.0; }\n  return sign(x);\n"
+                }),
+                Iz = {
+                    kernelName: Ir,
+                    backendName: "webgl",
+                    kernelFunc: Sz
+                },
+                Az = iL({
+                    opSnippet: "if (isnan(x)) return x;\n  return sin(x);\n"
+                }),
+                Ez = {
+                    kernelName: kr,
+                    backendName: "webgl",
+                    kernelFunc: Az
+                },
+                Nz = iL({
+                    opSnippet: "\n  float e2x = exp(x);\n  return (e2x - 1.0 / e2x) / 2.0;\n"
+                }),
+                Cz = {
+                    kernelName: Sr,
+                    backendName: "webgl",
+                    kernelFunc: Nz
+                },
+                Tz = iL({
+                    opSnippet: "\n  float epsilon = 1.1920928955078125e-7;\n  float threshold = log(epsilon) + 2.0;\n\n  bool too_large = x > -threshold;\n  bool too_small = x < threshold;\n\n  float result;\n  float exp_x = exp(x);\n\n  if (too_large){\n    result = x;\n  }\n  else if (too_small){\n    result = exp_x;\n  }\n  else{\n    result = log(exp_x + 1.0);\n  }\n  return result;\n"
+                }),
+                Rz = {
+                    kernelName: Er,
+                    backendName: "webgl",
+                    kernelFunc: Tz
+                },
+                _z = {
+                    kernelName: Tr,
+                    backendName: "webgl",
+                    kernelFunc: e => {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            blockShape: s,
+                            paddings: i
+                        } = r;
+                        va.assert(a.shape.length <= 4, () => "spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");
+                        const o = s.reduce((e, t) => e * t),
+                            u = [
+                                [0, 0]
+                            ];
+                        u.push(...i);
+                        for (let e = 1 + s.length; e < a.shape.length; ++e) u.push([0, 0]);
+                        const l = [],
+                            c = qB({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    paddings: u,
+                                    constantValue: 0
+                                }
+                            }),
+                            h = jd.getReshaped(c.shape, s, o, !1),
+                            p = jd.getPermuted(h.length, s.length, !1),
+                            d = jd.getReshapedPermuted(c.shape, s, o, !1),
+                            f = gL({
+                                inputs: {
+                                    x: c
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: h
+                                }
+                            }),
+                            m = IL({
+                                inputs: {
+                                    x: f
+                                },
+                                backend: n,
+                                attrs: {
+                                    perm: p
+                                }
+                            }),
+                            g = gL({
+                                inputs: {
+                                    x: m
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: d
+                                }
+                            });
+                        return l.push(c), l.push(f), l.push(m), l.forEach(e => n.disposeIntermediateTensorInfo(e)), g
+                    }
+                };
+            var Fz = {
+                kernelName: Fr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        indices: r,
+                        values: a,
+                        denseShape: s,
+                        defaultValue: i
+                    } = t;
+                    if (1 !== s.shape.length) throw new Error("Dense shape must be a vector, saw:\n         " + s.shape);
+                    if (2 !== r.shape.length) throw new Error("Indices must be a matrix, saw:\n         " + r.shape);
+                    if (1 !== a.shape.length) throw new Error("Values must be a vector, saw:\n         " + a.shape);
+                    if (0 !== i.shape.length) throw new Error("Default value must be a scalar, saw:\n        " + i.shape);
+                    const o = n.readSync(r.dataId),
+                        u = n.readSync(a.dataId),
+                        l = n.readSync(s.dataId),
+                        c = n.readSync(i.dataId)[0],
+                        [h, p, d, f, m] = bO(o, r.shape, r.dtype, u, a.dtype, l, c);
+                    return [n.makeTensorInfo(p, r.dtype, h), n.makeTensorInfo([p[0]], a.dtype, d), n.makeTensorInfo([f.length], "bool", new Uint8Array(f.map(e => Number(e)))), n.makeTensorInfo([m.length], r.dtype, new Int32Array(m))]
+                }
+            };
+            var Mz = {
+                kernelName: Mr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        inputIndices: r,
+                        inputShape: a,
+                        newShape: s
+                    } = t;
+                    if (2 !== r.shape.length) throw new Error("Input indices should be a matrix but received shape " + r.shape);
+                    if (1 !== a.shape.length) throw new Error("Input shape should be a vector but received shape " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Target shape should be a vector but received shape " + s.shape);
+                    const i = Array.from(n.readSync(a.dataId)),
+                        o = n.readSync(r.dataId),
+                        u = Array.from(n.readSync(s.dataId)),
+                        [l, c, h] = vO(o, r.shape, r.dtype, i, u);
+                    return [n.makeTensorInfo(c, r.dtype, l), n.makeTensorInfo([h.length], s.dtype, new Int32Array(h))]
+                }
+            };
+            var Dz = {
+                kernelName: Dr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        data: r,
+                        indices: a,
+                        segmentIds: s
+                    } = t;
+                    if (r.shape.length < 1) throw new Error("Data should be at least 1 dimensional but received scalar");
+                    if (1 !== a.shape.length) throw new Error("Indices should be a vector but received shape\n              " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Segment ids should be a vector but received shape\n              " + s.shape);
+                    const i = n.readSync(r.dataId),
+                        o = n.readSync(a.dataId),
+                        u = n.readSync(s.dataId),
+                        [l, c] = xO(i, r.shape, r.dtype, o, u, !0);
+                    return n.makeTensorInfo(c, r.dtype, l)
+                }
+            };
+            var Oz = {
+                kernelName: Or,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        data: r,
+                        indices: a,
+                        segmentIds: s
+                    } = t;
+                    if (r.shape.length < 1) throw new Error("Data should be at least 1 dimensional but received scalar");
+                    if (1 !== a.shape.length) throw new Error("Indices should be a vector but received shape\n             " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Segment ids should be a vector but received shape\n             " + s.shape);
+                    const i = n.readSync(r.dataId),
+                        o = n.readSync(a.dataId),
+                        u = n.readSync(s.dataId),
+                        [l, c] = xO(i, r.shape, r.dtype, o, u);
+                    return n.makeTensorInfo(c, r.dtype, l)
+                }
+            };
+            var Lz = {
+                kernelName: Lr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        sparseIndices: a,
+                        sparseValues: s,
+                        defaultValue: i
+                    } = t, {
+                        outputShape: o
+                    } = r, {
+                        sliceRank: u,
+                        numUpdates: l,
+                        strides: c,
+                        outputSize: h
+                    } = jd.calculateShapes(s, a, o), p = new gz(l, u, a.shape.length, s.shape.length, c, [h, 1], !1), d = n.runWebGLProgram(p, [s, a, i], s.dtype), f = gL({
+                        inputs: {
+                            x: d
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: o
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(d), f
+                }
+            };
+            var Pz = {
+                    kernelName: Rr,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            numOrSizeSplits: s,
+                            axis: i
+                        } = r, o = va.parseAxisParam(i, a.shape)[0], u = jd.prepareSplitSize(a, s, o), l = a.shape.length, c = new Array(l).fill(0), h = a.shape.slice();
+                        return u.map(e => {
+                            const t = [...h];
+                            t[o] = e;
+                            const r = lP({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    begin: c,
+                                    size: t
+                                }
+                            });
+                            return c[o] += e, r
+                        })
+                    }
+                },
+                $z = "return sqrt(x);",
+                Bz = iL({
+                    opSnippet: $z,
+                    packedOpSnippet: $z,
+                    cpuKernelImpl: wO
+                }),
+                zz = {
+                    kernelName: Nr,
+                    backendName: "webgl",
+                    kernelFunc: Bz
+                },
+                Uz = iL({
+                    opSnippet: "return x * x;"
+                }),
+                Wz = {
+                    kernelName: $r,
+                    backendName: "webgl",
+                    kernelFunc: Uz
+                },
+                Vz = oL({
+                    opSnippet: "return (a - b) * (a - b);",
+                    packedOpSnippet: "return (a - b) * (a - b);"
+                }),
+                jz = {
+                    kernelName: Pr,
+                    backendName: "webgl",
+                    kernelFunc: Vz
+                };
+            var Gz = {
+                kernelName: ea,
+                backendName: "webgl",
+                kernelFunc: function ({
+                    inputs: e,
+                    attrs: t,
+                    backend: n
+                }) {
+                    const {
+                        x: r
+                    } = e, a = `if (isnan(x)) return x;\n    return x > 0.0 ? 1.0 : float(${t.alpha});\n  `, s = new PO(r.shape, a);
+                    return n.runWebGLProgram(s, [r], r.dtype)
+                }
+            };
+            var Hz = {
+                kernelName: Br,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        begin: s,
+                        end: i,
+                        strides: o,
+                        beginMask: u,
+                        endMask: l,
+                        ellipsisMask: c,
+                        newAxisMask: h,
+                        shrinkAxisMask: p
+                    } = r, {
+                        finalShapeSparse: d,
+                        finalShape: f,
+                        isIdentity: m,
+                        sliceDim0: g,
+                        isSimpleSlice: y,
+                        begin: b,
+                        end: v,
+                        strides: x
+                    } = fo.sliceInfo(a.shape, s, i, o, u, l, c, h, p);
+                    let w;
+                    if (m) w = gL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: f
+                        }
+                    });
+                    else if (g || y) {
+                        va.assert(a.shape.length >= 1, () => "Input must have rank at least 1, got: " + a.shape.length);
+                        const e = fo.computeOutShape(b, v, x),
+                            t = lP({
+                                inputs: {
+                                    x: a
+                                },
+                                backend: n,
+                                attrs: {
+                                    begin: b,
+                                    size: e
+                                }
+                            });
+                        w = gL({
+                            inputs: {
+                                x: t
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: f
+                            }
+                        }), n.disposeIntermediateTensorInfo(t)
+                    } else {
+                        if (n.shouldExecuteOnCPU([a])) {
+                            const e = n.readSync(a.dataId),
+                                t = Ii(a.shape, a.dtype, e),
+                                r = kO(d, t, x, b);
+                            w = n.makeTensorInfo(f, a.dtype, r.values)
+                        } else {
+                            const e = new class {
+                                constructor(e, t, n) {
+                                    this.variableNames = ["x"], this.outputShape = n;
+                                    const r = n.length,
+                                        a = dD(n.length),
+                                        s = dD(n.length);
+                                    let i = "";
+                                    if (1 === r) i = "coords * strides + begin";
+                                    else {
+                                        let e = 0;
+                                        i = n.map((t, r) => (e++, 1 === n.length ? `coords * strides[${r}] + begin[${r}]` : `coords[${e-1}] * strides[${r}] + begin[${r}]`)).join(",")
+                                    }
+                                    this.userCode = `\n      ${a} begin = ${a}(${e});\n      ${a} strides = ${a}(${t});\n\n      void main() {\n        ${s} coords = getOutputCoords();\n        setOutput(getX(${i}));\n      }\n    `
+                                }
+                            }(b, x, d);
+                            w = n.runWebGLProgram(e, [a], a.dtype)
+                        }
+                    }
+                    const k = gL({
+                        inputs: {
+                            x: w
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: f
+                        }
+                    });
+                    return n.disposeIntermediateTensorInfo(w), k
+                }
+            };
+            var qz = {
+                kernelName: zr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        separator: a,
+                        nGramWidths: s,
+                        leftPad: i,
+                        rightPad: o,
+                        padWidth: u,
+                        preserveShortSequences: l
+                    } = r, {
+                        data: c,
+                        dataSplits: h
+                    } = t, p = n.readSync(c.dataId), d = n.readSync(h.dataId), [f, m] = SO(p, d, a, s, i, o, u, l);
+                    return [n.makeTensorInfo([f.length], "string", f), n.makeTensorInfo(h.shape, "int32", m)]
+                }
+            };
+            var Kz = {
+                kernelName: Ur,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        skipEmpty: a
+                    } = r, {
+                        input: s,
+                        delimiter: i
+                    } = t;
+                    if ("string" !== s.dtype) throw new Error("Input must be of datatype string");
+                    if (1 !== s.shape.length) throw new Error("Input must be a vector, got shape: " + s.shape);
+                    if (0 !== i.shape.length) throw new Error("Delimiter must be a scalar, got shape: " + i.shape);
+                    const o = n.readSync(s.dataId),
+                        u = n.readSync(i.dataId)[0],
+                        [l, c, h] = IO(o, u, a),
+                        p = c.length;
+                    return [n.makeTensorInfo([p, 2], "int32", l), n.makeTensorInfo([p], "string", c), n.makeTensorInfo([2], "int32", new Int32Array(h))]
+                }
+            };
+            var Yz = {
+                    kernelName: Wr,
+                    backendName: "webgl",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            numBuckets: a
+                        } = r, {
+                            input: s
+                        } = t;
+                        if ("string" !== s.dtype) throw new Error("Input must be of datatype string");
+                        if (a <= 0) throw new Error("Number of buckets must be at least 1");
+                        const i = n.readSync(s.dataId),
+                            o = AO(i, a);
+                        return n.makeTensorInfo(s.shape, "int32", o)
+                    }
+                },
+                Xz = iL({
+                    opSnippet: "return tan(x);"
+                }),
+                Qz = {
+                    kernelName: jr,
+                    backendName: "webgl",
+                    kernelFunc: Xz
+                },
+                Zz = iL({
+                    opSnippet: "\n  float e2x = exp(-2.0 * abs(x));\n  return sign(x) * (1.0 - e2x) / (1.0 + e2x);\n"
+                }),
+                Jz = {
+                    kernelName: Gr,
+                    backendName: "webgl",
+                    kernelFunc: Zz
+                };
+
+            function eU(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, {
+                    x: a
+                } = t, {
+                    reps: s
+                } = r;
+                if ("string" === a.dtype || a.shape.length > 5) {
+                    const e = n.readSync(a.dataId),
+                        t = "string" === a.dtype ? e.map(e => va.decodeString(e)) : e,
+                        r = Ii(a.shape, a.dtype, t),
+                        i = NO(r, s);
+                    return n.makeTensorInfo(i.shape, i.dtype, i.values)
+                }
+                const i = new class {
+                    constructor(e, t) {
+                        this.variableNames = ["A"];
+                        const n = new Array(e.length);
+                        for (let r = 0; r < n.length; r++) n[r] = e[r] * t[r];
+                        this.outputShape = n, this.rank = n.length;
+                        const r = dD(this.rank),
+                            a = function (e) {
+                                const t = e.length;
+                                if (t > 5) throw Error(`Tile for rank ${t} is not yet supported`);
+                                if (1 === t) return `imod(resRC, ${e[0]})`;
+                                const n = ["resRC.x", "resRC.y", "resRC.z", "resRC.w", "resRC.u"],
+                                    r = [];
+                                for (let t = 0; t < e.length; t++) r.push(`imod(${n[t]}, ${e[t]})`);
+                                return r.join()
+                            }(e);
+                        this.userCode = `\n      void main() {\n        ${r} resRC = getOutputCoords();\n        setOutput(getA(${a}));\n      }\n    `
+                    }
+                }(a.shape, s);
+                return n.runWebGLProgram(i, [a], a.dtype)
+            }
+            var tU = {
+                    kernelName: Hr,
+                    backendName: "webgl",
+                    kernelFunc: eU
+                },
+                nU = class {
+                    constructor(e) {
+                        this.variableNames = ["x", "indices"], this.customUniforms = [{
+                            name: "n",
+                            type: "int"
+                        }, {
+                            name: "firstPass",
+                            type: "int"
+                        }, {
+                            name: "k",
+                            type: "int"
+                        }], this.outputShape = e, this.userCode = "\n    void main() {\n         // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // The output size is half of the previous size.\n         // If the previous sequence is | | | | _ _ _ _  | | | |  _ _ _ _ (k=4),\n         // we only need to output the indices at positions |, the indices at\n         // positions _ can be thrown away, see Figure5(b) After Phase 2\n         // (Merge phase) in the Bitonic Top K paper referenced above.\n         // For example, the paper shows we only need to output the orange bars.\n         // The output sequence should look like this | | | | | | | |.\n         // Because the sequence is halved, to map the output index back\n         // to the previous sequence to find the corresponding value,\n         // we need to double the index. When we double the index,\n         // we basically interpolate a position, so 2i looks like\n         // | _ | _ | _ | _ | _ | _ | _. We move the | to the first k position\n         // of each 2k positions by - elemIdx % k. E.g. for output at\n         // index 4,5,6,7, we want to get the corresponding element at\n         // original index 8,9,10,11, for output at index 8,9,10,11,\n         // we want to get the corresponding element at original index\n         // 16,17,18,19, so on and so forth.\n\n         int i = elemIdx < k ? elemIdx : (elemIdx * 2 - imod(elemIdx, k));\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + k : int(getIndices(batch, i + k));\n\n         float x0 = getX(batch, i0);\n         float x1 = i1 < n ? getX(batch, i1) : x0;\n\n         setOutput(x0 >= x1 ? float(i0) : float(i1));\n       }\n     "
+                    }
+                };
+
+            function rU(e, t) {
+                null !== t && e.disposeIntermediateTensorInfo(t)
+            }
+
+            function aU(e) {
+                let t = 1;
+                for (; t < e;) t *= 2;
+                return t
+            }
+            var sU = {
+                kernelName: qr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        k: s,
+                        sorted: i
+                    } = r, o = Ye().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"), u = Ye().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"), l = a.shape, c = l[l.length - 1];
+                    if (n.shouldExecuteOnCPU([a]) || c < o || s > u) {
+                        const e = n.readSync(a.dataId),
+                            [t, r] = CO(e, l, a.dtype, s, i);
+                        return [n.makeTensorInfo(t.shape, t.dtype, t.values), n.makeTensorInfo(r.shape, r.dtype, r.values)]
+                    }
+                    if (0 === s) return l[l.length - 1] = 0, [n.makeTensorInfo(l, a.dtype, []), n.makeTensorInfo(l, "int32", [])];
+                    if (1 === c) return [a, k$({
+                        attrs: {
+                            shape: l,
+                            dtype: "int32",
+                            value: 0
+                        },
+                        backend: n
+                    })];
+                    const h = n.texData.get(a.dataId),
+                        p = null !== h && h.isPacked,
+                        d = p ? n.unpackTensor(a) : a,
+                        f = va.sizeFromShape(l) / c,
+                        m = gL({
+                            inputs: {
+                                x: d
+                            },
+                            attrs: {
+                                shape: [f, c]
+                            },
+                            backend: n
+                        });
+                    p && rU(n, d);
+                    const g = aU(s),
+                        y = aU(c);
+                    let b = null;
+                    const v = () => null === b ? [m, m] : [m, b],
+                        x = (e, t, r) => {
+                            const a = v(),
+                                s = new class {
+                                    constructor(e) {
+                                        this.variableNames = ["x", "indices"], this.customUniforms = [{
+                                            name: "n",
+                                            type: "int"
+                                        }, {
+                                            name: "firstPass",
+                                            type: "int"
+                                        }, {
+                                            name: "negativeInf",
+                                            type: "float"
+                                        }, {
+                                            name: "dir",
+                                            type: "int"
+                                        }, {
+                                            name: "inc",
+                                            type: "int"
+                                        }], this.outputShape = e, this.userCode = "\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // We compare elements pair-wise within a group of size 2 * inc.\n         // The comparing rule for each group alternates between ascending\n         // and descending. Within each group, we compare each pair at\n         // positions i and i+inc. To decide whether an element at position i\n         // is x0 or x1, we mod it by 2 * inc, if the result is smaller than\n         // inc, it is in the first half of the group, we denote it as x0,\n         // otherwise we denote it as x1.\n         // For example, as shown in the Bitonic top K paper referenced above,\n         // Figure5(a) shows that element[1] is in the\n         // second half of the group when group size is 2, but it is in the\n         // first half of the group when group size is 4.\n\n         bool isFirstInPair = imod(elemIdx, 2 * inc) < inc;\n         int i = isFirstInPair ? elemIdx : elemIdx - inc;\n\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + inc : int(getIndices(batch, i + inc));\n         float x0 = i0 < n ? getX(batch, i0) : negativeInf;\n         float x1 = i1 < n ? getX(batch, i1) : negativeInf;\n\n         // Denotes which direction indices are in (ascending or descending).\n         bool reverse = imod(elemIdx, 2 * dir) >= dir;\n         bool isGreater = x0 > x1 || (x0 == x1 && i1 > i0);\n         if (reverse == isGreater) { // Elements in opposite order of direction\n           int iTemp = i0;\n           i0 = i1;\n           i1 = iTemp;\n         }\n         if (isFirstInPair) {\n            setOutput(float(i0));\n         } else {\n            setOutput(float(i1));\n         }\n       }\n     "
+                                    }
+                                }(r),
+                                i = [
+                                    [c],
+                                    [null === b ? 1 : 0],
+                                    [Number.NEGATIVE_INFINITY],
+                                    [e],
+                                    [t]
+                                ],
+                                o = b;
+                            b = n.runWebGLProgram(s, a, "int32", i), rU(n, o)
+                        };
+                    for (let e = 1; e < g; e *= 2) {
+                        const t = 2 * e;
+                        for (let n = e; n >= 1; n /= 2) x(t, n, [f, y])
+                    }
+                    for (let e = y; e > g; e /= 2) {
+                        const t = v(),
+                            r = new nU([f, e / 2]),
+                            a = [
+                                [c],
+                                [null === b ? 1 : 0],
+                                [g]
+                            ],
+                            s = b;
+                        b = n.runWebGLProgram(r, t, "int32", a), rU(n, s);
+                        const i = g / 2,
+                            o = 2 * i;
+                        for (let e = i; e >= 1; e /= 2) x(o, e, b.shape)
+                    }
+                    let w = b;
+                    b = lP({
+                        inputs: {
+                            x: b
+                        },
+                        backend: n,
+                        attrs: {
+                            begin: 0,
+                            size: [f, s]
+                        }
+                    }), rU(n, w);
+                    let k = O$({
+                        inputs: {
+                            x: m,
+                            indices: b
+                        },
+                        backend: n,
+                        attrs: {
+                            axis: 1,
+                            batchDims: 1
+                        }
+                    });
+                    rU(n, m);
+                    const S = l.slice(0, -1);
+                    S.push(s), w = b, b = gL({
+                        inputs: {
+                            x: b
+                        },
+                        attrs: {
+                            shape: S
+                        },
+                        backend: n
+                    }), rU(n, w);
+                    const I = k;
+                    return k = gL({
+                        inputs: {
+                            x: k
+                        },
+                        attrs: {
+                            shape: S
+                        },
+                        backend: n
+                    }), rU(n, I), [k, b]
+                }
+            };
+            var iU = {
+                kernelName: Kr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        image: a,
+                        transforms: s
+                    } = t, {
+                        interpolation: i,
+                        fillMode: o,
+                        fillValue: u,
+                        outputShape: l
+                    } = r, [c, h, p, d] = a.shape, [f, m] = null != l ? l : [h, p], g = [c, f, m, d], y = new class {
+                        constructor(e, t, n, r, a, s) {
+                            this.variableNames = ["Image", "Transforms"], this.outputShape = s;
+                            const i = "nearest" === n ? 1 : 2;
+                            let o;
+                            switch (r) {
+                                case "constant":
+                                    o = 1;
+                                    break;
+                                case "reflect":
+                                    o = 2;
+                                    break;
+                                case "wrap":
+                                    o = 3;
+                                    break;
+                                case "nearest":
+                                    o = 4;
+                                    break;
+                                default:
+                                    o = 1
+                            }
+                            this.userCode = `\n            float mapCoord(float outCoord, float len) {\n              float inCoord = outCoord;\n              if(${o} == 2) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    if (inCoord < sz2) {\n                      inCoord = sz2 * float(int(float(-inCoord / sz2))) +\n                      inCoord;\n                    }\n                    inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1.0;\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    inCoord -= sz2 * float(int(float(inCoord / sz2)));\n                    if (inCoord >= len) {\n                      inCoord = sz2 - inCoord - 1.0;\n                    }\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${o} == 3) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord += len * (float(int(float(-inCoord / sz))) + 1.0);\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord -= len * float(int(float(inCoord / sz)));\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${o} == 4) {\n                return clamp(outCoord, 0.0, len - 1.0);\n              } else {\n                return outCoord;\n              }\n            }\n\n            float readWithFillValue(int batch, int coordY, int coordX,\n              int channel) {\n              float outputValue;\n              if (0 <= coordY && coordY < ${e} && 0 <= coordX && coordX < ${t}) {\n                  outputValue = getImage(batch, coordY, coordX, channel);\n              } else {\n                outputValue = float(${a});\n              }\n              return outputValue;\n            }\n\n            void main() {\n              ivec4 coords = getOutputCoords();\n              float outputValue;\n              int batch = coords[0];\n              int x = coords[2];\n              int y = coords[1];\n              int channel = coords[3];\n              float xf = float(x);\n              float yf = float(y);\n              float a1 = getTransforms(batch, 0);\n              float a2 = getTransforms(batch, 1);\n              float a3 = getTransforms(batch, 2);\n              float b1 = getTransforms(batch, 3);\n              float b2 = getTransforms(batch, 4);\n              float b3 = getTransforms(batch, 5);\n              float c1 = getTransforms(batch, 6);\n              float c2 = getTransforms(batch, 7);\n              float projection = c1 * xf + c2 * yf + 1.0;\n              if (projection == 0.0) {\n                outputValue = float(${a});\n              } else {\n                float inX = (a1 * xf + a2 * yf + a3) / projection;\n                float inY = (b1 * xf + b2 * yf + b3) / projection;\n                float mapX = mapCoord(inX, float(${t}));\n                float mapY = mapCoord(inY, float(${e}));\n\n                if (${i} == 1) {\n                  int coordY = int(round(mapY));\n                  int coordX = int(round(mapX));\n                  outputValue = readWithFillValue(batch, coordY, coordX,\n                    channel);\n                } else {\n                  float yFloor = floor(mapY);\n                  float xFloor = floor(mapX);\n                  float yCeil = yFloor + 1.0;\n                  float xCeil = xFloor + 1.0;\n                  float valueYFloor = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yFloor), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yFloor), int(xCeil), channel);\n                  float valueYCeil = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yCeil), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yCeil), int(xCeil), channel);\n                  outputValue = (yCeil - mapY) * valueYFloor +\n                  (mapY - yFloor) * valueYCeil;\n                }\n              }\n              setOutput(outputValue);\n            }\n        `
+                        }
+                    }(h, p, i, o, u, g);
+                    return n.runWebGLProgram(y, [a, s], "float32")
+                }
+            };
+            var oU = {
+                kernelName: Xr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, {
+                        axis: a
+                    } = n, {
+                        x: s
+                    } = t;
+                    KM(s, "unique"), console.warn("WARNING: ", "UI might be locked temporarily as data is being downloaded");
+                    const i = r.readSync(s.dataId),
+                        {
+                            outputValues: o,
+                            outputShape: u,
+                            indices: l
+                        } = RO(i, a, s.shape, s.dtype);
+                    return [r.makeTensorInfo(u, s.dtype, o), r.makeTensorInfo([l.length], "int32", l)]
+                }
+            };
+            var uU = {
+                kernelName: Qr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        value: a
+                    } = t;
+                    let {
+                        axis: s
+                    } = r;
+                    s < 0 && (s += a.shape.length);
+                    const i = a,
+                        o = i.shape.length,
+                        u = a.shape[s],
+                        l = new Array(o - 1);
+                    let c = 0;
+                    for (let e = 0; e < o; e++) e !== s && (l[c++] = i.shape[e]);
+                    const h = [],
+                        p = new Array(o).fill(0),
+                        d = i.shape.slice();
+                    d[s] = 1;
+                    const f = new Array(u);
+                    for (let e = 0; e < f.length; e++) {
+                        p[s] = e;
+                        const t = lP({
+                                inputs: {
+                                    x: i
+                                },
+                                backend: n,
+                                attrs: {
+                                    begin: p,
+                                    size: d
+                                }
+                            }),
+                            r = gL({
+                                inputs: {
+                                    x: t
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: l
+                                }
+                            });
+                        f[e] = r, h.push(t)
+                    }
+                    return h.forEach(e => n.disposeIntermediateTensorInfo(e)), f
+                }
+            };
+            var lU, cU, hU, pU, dU, fU = [NL, TL, _L, ML, LL, PL, $L, BL, WL, VL, GL, qL, YL, QL, JL, nP, rP, aP, sP, iP, oP, hP, pP, dP, bP, wP, kP, JO, IP, RP, DP, OP, LP, PP, $P, BP, UP, VP, jP, KP, YP, XP, JP, e$, t$, n$, r$, a$, i$, o$, l$, h$, d$, m$, b$, w$, I$, A$, C$, R$, _$, F$, M$, D$, L$, $$, z$, QO, U$, NP, V$, G$, q$, nL, Y$, Q$, Z$, eB, nB, aB, iB, uB, lB, cB, pB, fB, mB, gB, yB, bB, vB, xB, wB, SB, IB, EB, DB, mL, OB, PB, BB, UB, mP, WB, GB, HB, KB, XB, sL, QB, JB, yP, CB, tz, rz, sz, yL, iz, oz, uz, lz, cz, hz, dz, mz, yz, bz, xz, kz, Iz, Ez, Cz, cP, MB, Rz, _z, Fz, Mz, Dz, Oz, Lz, Pz, zz, Wz, jz, Gz, Hz, qz, Kz, Yz, _B, SL, Qz, Jz, tU, sU, iU, AL, oU, uU, {
+                kernelName: Zr,
+                backendName: "webgl",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a,
+                        segmentIds: s
+                    } = t, {
+                        numSegments: i
+                    } = r, o = a.shape.length, u = [];
+                    let l = 0;
+                    const c = jd.getAxesPermutation([l], o);
+                    let h = a;
+                    null != c && (h = IL({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: c
+                        }
+                    }), u.push(h), l = jd.getInnerMostAxes(1, o)[0]);
+                    const p = jd.segment_util.computeOutShape(h.shape, l, i),
+                        d = va.sizeFromShape([h.shape[l]]),
+                        f = gL({
+                            inputs: {
+                                x: h
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: [-1, d]
+                            }
+                        });
+                    u.push(f);
+                    const m = hs(a.dtype),
+                        g = (e, t, r, a, s) => {
+                            const i = e.shape[0],
+                                o = e.shape[1],
+                                l = jd.segment_util.segOpComputeOptimalWindowSize(o, s),
+                                c = {
+                                    windowSize: l,
+                                    inSize: o,
+                                    batchSize: i,
+                                    numSegments: s
+                                },
+                                h = new class {
+                                    constructor(e, t) {
+                                        this.variableNames = ["x", "segmentIds"];
+                                        const n = e.windowSize,
+                                            r = e.batchSize,
+                                            a = e.inSize,
+                                            s = e.numSegments,
+                                            i = s * Math.ceil(a / n);
+                                        this.outputShape = [r, i];
+                                        const o = 4 * Math.floor(n / 4),
+                                            u = n % 4,
+                                            l = "\n        sumValue += dot(values, segFilter);\n    ";
+                                        let c = "";
+                                        a % n > 0 && (c = `\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return initializationValue;\n        }\n      `);
+                                        let h = "";
+                                        a % n > 0 && (h = `\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return -1.0;\n        }\n      `), this.userCode = `\n      const float initializationValue = 0.0;\n\n      float getValue(int batch, int inIdx) {\n        ${c}\n        return getX(batch, inIdx);\n      }\n\n      float getSegmentIdAtIndex(int inIdx) {\n        ${h}\n        return getSegmentIds(inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = int(floor(float(outIdx) / float(\n          ${s})) * float(${n}));\n        int currentSeg = int(mod(float(outIdx), float(${s})));\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${o}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0\n          );\n\n          ${l}\n        }\n\n        int inIdx = inOffset + ${o};\n        if (${1===u}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          int inIdxSeg = int(getSegmentIdAtIndex(inIdx));\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            0,\n            0,\n            0\n          );\n\n          ${l}\n        } else if (${2===u}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n              0,\n              0\n          );\n\n          ${l}\n        } else if (${3===u}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            0\n          );\n\n          ${l}\n        }\n        setOutput(sumValue);\n      }\n    `
+                                    }
+                                }(c, t),
+                                p = n.compileAndRun(h, [e, r], a);
+                            if (u.push(p), p.shape[1] === s) return p;
+                            const d = ZB({
+                                    backend: n,
+                                    attrs: {
+                                        start: 0,
+                                        stop: s,
+                                        step: 1,
+                                        dtype: "float32"
+                                    }
+                                }),
+                                f = eU({
+                                    inputs: {
+                                        x: d
+                                    },
+                                    backend: n,
+                                    attrs: {
+                                        reps: [o / l]
+                                    }
+                                });
+                            u.push(d), u.push(f);
+                            return g(p, t, f, a, s)
+                        },
+                        y = gL({
+                            inputs: {
+                                x: g(f, "unsortedSegmentSum", s, m, i)
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: p
+                            }
+                        });
+                    let b = y;
+                    if (null != c) {
+                        u.push(y);
+                        const e = jd.getUndoAxesPermutation(c);
+                        b = IL({
+                            inputs: {
+                                x: b
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: e
+                            }
+                        })
+                    }
+                    return u.forEach(e => n.disposeIntermediateTensorInfo(e)), b
+                }
+            }, jB];
+            for (const e of fU) da(e);
+            (cU = lU || (lU = {}))[cU.float32 = 0] = "float32", cU[cU.int32 = 1] = "int32", cU[cU.bool = 2] = "bool", cU[cU.string = 3] = "string", cU[cU.complex64 = 4] = "complex64", (pU = hU || (hU = {}))[pU.linear = 0] = "linear", pU[pU.relu = 1] = "relu", pU[pU.relu6 = 2] = "relu6", pU[pU.prelu = 3] = "prelu", pU[pU.leakyrelu = 4] = "leakyrelu", pU[pU.sigmoid = 5] = "sigmoid", pU[pU.elu = 6] = "elu";
+            var mU = {
+                kernelName: ra,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    dU = e.wasm.cwrap(ra, null, ["number", "array", "number", "number", "array", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        a: a,
+                        b: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t;
+                    if ("float32" !== a.dtype || "float32" !== s.dtype) throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");
+                    const {
+                        transposeA: u,
+                        transposeB: l,
+                        activation: c,
+                        leakyreluAlpha: h
+                    } = r, p = n.dataIdMap.get(a.dataId).id, d = n.dataIdMap.get(s.dataId).id;
+                    let f = 0;
+                    if (null != i) {
+                        const e = n.dataIdMap.get(i.dataId);
+                        if (1 !== e.shape.length) throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${e.shape.length}.`);
+                        f = e.id
+                    }
+                    const m = null == o ? 0 : n.dataIdMap.get(o.dataId).id,
+                        g = hU[c];
+                    if (null == g) throw new Error(c + " activation not yet supported for FusedConv2D in the wasm backend.");
+                    const y = u ? a.shape[2] : a.shape[1],
+                        b = l ? s.shape[1] : s.shape[2],
+                        v = Yi.assertAndGetBroadcastShape(a.shape.slice(0, -2), s.shape.slice(0, -2)),
+                        x = n.makeOutput([...v, y, b], a.dtype),
+                        w = n.dataIdMap.get(x.dataId).id,
+                        k = new Uint8Array(new Int32Array(a.shape).buffer),
+                        S = new Uint8Array(new Int32Array(s.shape).buffer);
+                    return dU(p, k, a.shape.length, d, S, s.shape.length, u, l, g, f, m, h || 0, w), x
+                }
+            };
+
+            function gU(e, t) {
+                let n;
+                return {
+                    kernelName: e,
+                    backendName: "wasm",
+                    setupFunc: function (t) {
+                        n = t.wasm.cwrap(e, null, ["number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: r,
+                            inputs: {
+                                x: a
+                            }
+                        } = e, s = r.dataIdMap.get(a.dataId).id, i = r.makeOutput(a.shape, t || a.dtype), o = r.dataIdMap.get(i.dataId).id;
+                        return 0 === va.sizeFromShape(i.shape) || n(s, lU[a.dtype], o), i
+                    }
+                }
+            }
+            var yU = gU(et);
+
+            function bU(e, t, n) {
+                let r;
+                return {
+                    kernelName: e,
+                    backendName: "wasm",
+                    setupFunc: function (t) {
+                        r = t.wasm.cwrap(e, null, ["number", "array", "number", "number", "array", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: a
+                        } = e, {
+                            a: s,
+                            b: i
+                        } = a, o = t.dataIdMap.get(s.dataId).id, u = t.dataIdMap.get(i.dataId).id, l = null != n ? n : s.dtype, c = jd.assertAndGetBroadcastShape(s.shape, i.shape), h = t.makeOutput(c, l);
+                        if (0 === va.sizeFromShape(c)) return h;
+                        const p = new Uint8Array(new Int32Array(s.shape).buffer),
+                            d = new Uint8Array(new Int32Array(i.shape).buffer),
+                            f = t.dataIdMap.get(h.dataId).id;
+                        return r(o, p, s.shape.length, u, d, i.shape.length, lU[s.dtype], f), h
+                    }
+                }
+            }
+            var vU, xU = bU(rt);
+            var wU = {
+                kernelName: at,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    vU = e.wasm.cwrap(at, null, ["array", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, r = n.makeOutput(t[0].shape, t[0].dtype);
+                    if (0 === va.sizeFromShape(r.shape)) return r;
+                    const a = t.map(e => n.dataIdMap.get(e.dataId).id),
+                        s = new Uint8Array(new Int32Array(a).buffer),
+                        i = n.dataIdMap.get(r.dataId).id;
+                    return vU(s, a.length, lU[r.dtype], i), r
+                }
+            };
+
+            function kU(e) {
+                const {
+                    inputs: {
+                        x: t
+                    },
+                    backend: n
+                } = e, r = n.makeOutput(t.shape, t.dtype), a = n.typedArrayFromHeap(t);
+                return n.typedArrayFromHeap(r).set(a), r
+            }
+            var SU, IU = {
+                kernelName: fn,
+                backendName: "wasm",
+                kernelFunc: kU
+            };
+
+            function AU(e) {
+                const {
+                    inputs: t,
+                    backend: n,
+                    attrs: r
+                } = e, [a, s] = function (e, t) {
+                    const n = [],
+                        r = [];
+                    for (let a = 0; a < e.length; ++a) 1 !== e[a] && n.push(e[a]), 1 !== e[t[a]] && r.push(t[a]);
+                    for (let e = 0; e < r.length; ++e) {
+                        let t = -1;
+                        for (let n = 0; n < r.length; ++n) r[n] >= e && (-1 === t || r[t] > r[n]) && (t = n);
+                        r[t] = e
+                    }
+                    return [n, r]
+                }(t.x.shape, r.perm);
+                let i = !0;
+                for (let e = 0; e < s.length; e++) s[e] !== e && (i = !1);
+                const o = function (e, t) {
+                        const n = new Array(e.length);
+                        for (let r = 0; r < n.length; r++) n[r] = e[t[r]];
+                        return n
+                    }(t.x.shape, r.perm),
+                    u = {
+                        dataId: t.x.dataId,
+                        shape: a,
+                        dtype: t.x.dtype
+                    };
+                if (i) {
+                    const e = kU({
+                        inputs: t,
+                        backend: n
+                    });
+                    return e.shape = o, e
+                }
+                const l = n.makeOutput(o, u.dtype),
+                    c = n.dataIdMap.get(u.dataId).id,
+                    h = n.dataIdMap.get(l.dataId).id,
+                    p = new Uint8Array(new Int32Array(s).buffer),
+                    d = new Uint8Array(new Int32Array(u.shape).buffer);
+                return SU(c, d, u.shape.length, lU[u.dtype], h, p, s.length), l
+            }
+            var EU, NU = {
+                kernelName: Yr,
+                backendName: "wasm",
+                kernelFunc: AU,
+                setupFunc: function (e) {
+                    SU = e.wasm.cwrap(Yr, null, ["number", "array", "number", "number", "number", "array", "number"])
+                }
+            };
+
+            function CU(e, t, n) {
+                const r = e.shape,
+                    a = e.shape.length,
+                    s = va.parseAxisParam(t, r);
+                let i = s;
+                const o = jd.getAxesPermutation(i, a);
+                let u = null,
+                    l = !1;
+                if (null != o) {
+                    const t = new Array(a);
+                    for (let e = 0; e < t.length; e++) t[e] = r[o[e]];
+                    i = jd.getInnerMostAxes(i.length, a), u = AU({
+                        inputs: {
+                            x: e
+                        },
+                        attrs: {
+                            perm: o
+                        },
+                        backend: n
+                    });
+                    const s = n.dataIdMap.get(e.dataId).id;
+                    n.dataIdMap.get(u.dataId).id !== s && (l = !0)
+                }
+                return {
+                    transposed: u,
+                    originalAxes: s,
+                    axes: i,
+                    inputWasTransposed: l
+                }
+            }
+            var TU, RU = {
+                kernelName: st,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    EU = e.wasm.cwrap(st, null, ["number, number, number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        axis: a,
+                        keepDims: s
+                    } = r, {
+                        x: i
+                    } = n;
+                    let o = t.dataIdMap.get(i.dataId).id,
+                        u = i;
+                    const {
+                        transposed: l,
+                        axes: c,
+                        originalAxes: h,
+                        inputWasTransposed: p
+                    } = CU(i, a, t);
+                    if (p) {
+                        u = l, o = t.dataIdMap.get(l.dataId).id
+                    }
+                    const d = u.shape.length;
+                    jd.assertAxesAreInnerMostDims("all", c, d);
+                    const [f, m] = jd.computeOutAndReduceShapes(u.shape, c), g = va.sizeFromShape(m), y = t.makeOutput(f, i.dtype);
+                    if (0 !== va.sizeFromShape(u.shape)) {
+                        const e = t.dataIdMap.get(y.dataId).id;
+                        EU(o, g, e)
+                    }
+                    if (p && t.disposeData(l.dataId), s) {
+                        const e = jd.expandShapeToKeepDim(y.shape, h);
+                        y.shape = e
+                    }
+                    return y
+                }
+            };
+            var _U, FU = {
+                kernelName: it,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    TU = e.wasm.cwrap(it, null, ["number, number, number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        axis: a,
+                        keepDims: s
+                    } = r, {
+                        x: i
+                    } = n;
+                    let o = t.dataIdMap.get(i.dataId).id,
+                        u = i;
+                    const {
+                        transposed: l,
+                        axes: c,
+                        originalAxes: h,
+                        inputWasTransposed: p
+                    } = CU(i, a, t);
+                    if (p) {
+                        u = l, o = t.dataIdMap.get(l.dataId).id
+                    }
+                    const d = u.shape.length;
+                    jd.assertAxesAreInnerMostDims("any", c, d);
+                    const [f, m] = jd.computeOutAndReduceShapes(u.shape, c), g = va.sizeFromShape(m), y = t.makeOutput(f, i.dtype);
+                    if (0 !== va.sizeFromShape(u.shape)) {
+                        const e = t.dataIdMap.get(y.dataId).id;
+                        TU(o, g, e)
+                    }
+                    if (p && t.disposeData(l.dataId), s) {
+                        const e = jd.expandShapeToKeepDim(y.shape, h);
+                        y.shape = e
+                    }
+                    return y
+                }
+            };
+            var MU, DU = {
+                kernelName: ot,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        axis: a
+                    } = r, {
+                        x: s
+                    } = n, i = t.dataIdMap.get(s.dataId).id;
+                    let o = i,
+                        u = s;
+                    const {
+                        transposed: l,
+                        axes: c,
+                        inputWasTransposed: h
+                    } = CU(s, a, t);
+                    if (h) {
+                        const e = t.dataIdMap.get(l.dataId).id;
+                        e !== i && (u = l, o = e)
+                    }
+                    const p = u.shape.slice(0, -1),
+                        d = t.makeOutput(p, "int32"),
+                        f = t.dataIdMap.get(d.dataId).id,
+                        m = va.sizeFromShape(d.shape),
+                        g = u.shape[c[0]];
+                    return _U(o, lU[u.dtype], m, g, f), h && t.disposeData(l.dataId), d
+                },
+                setupFunc: function (e) {
+                    _U = e.wasm.cwrap(ot, null, ["number", "number", "number", "number", "number"])
+                }
+            };
+            var OU = {
+                kernelName: ft,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    MU = e.wasm.cwrap(ft, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, a = t.x, s = r.dataIdMap.get(a.dataId).id, {
+                        filterSize: i,
+                        strides: o,
+                        pad: u,
+                        dimRoundingMode: l
+                    } = n, c = jd.computePool2DInfo(a.shape, i, o, 1, u, l), h = c.filterHeight, p = c.filterWidth, d = c.padInfo.top, f = c.padInfo.right, m = c.padInfo.bottom, g = c.padInfo.left, y = c.strideHeight, b = c.strideWidth, v = c.inChannels;
+                    if ("channelsLast" !== c.dataFormat) throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);
+                    if (1 !== c.dilationWidth || 1 !== c.dilationHeight) throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${c.dilationHeight}, ${c.dilationWidth}].`);
+                    const x = r.makeOutput(c.outShape, "float32"),
+                        w = r.dataIdMap.get(x.dataId).id;
+                    return MU(s, a.shape[0], a.shape[1], a.shape[2], h, p, d, f, m, g, y, b, v, w), x
+                }
+            };
+
+            function LU(e) {
+                const {
+                    inputs: t,
+                    attrs: n
+                } = e, {
+                    x: r
+                } = t, {
+                    shape: a
+                } = n, s = va.sizeFromShape(r.shape), i = va.inferFromImplicitShape(a, s);
+                return va.assert(s === va.sizeFromShape(i), () => `new shape: ${i}, old shape: ${r.shape}. New shape and old shape must have the same number of elements.`), e.backend.incRef(r.dataId), {
+                    dataId: r.dataId,
+                    shape: i,
+                    dtype: r.dtype
+                }
+            }
+            var PU, $U = {
+                kernelName: lr,
+                backendName: "wasm",
+                kernelFunc: LU
+            };
+            var BU = {
+                kernelName: bt,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    PU = e.wasm.cwrap(bt, null, ["number", "array", "number", "number", "array", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        a: a,
+                        b: s
+                    } = t, {
+                        transposeA: i,
+                        transposeB: o
+                    } = r;
+                    if ("float32" !== a.dtype || "float32" !== s.dtype) throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");
+                    const u = a.shape.length,
+                        l = s.shape.length,
+                        c = i ? a.shape[u - 2] : a.shape[u - 1],
+                        h = o ? s.shape[l - 1] : s.shape[l - 2],
+                        p = i ? a.shape[u - 1] : a.shape[u - 2],
+                        d = o ? s.shape[l - 2] : s.shape[l - 1],
+                        f = a.shape.slice(0, -2),
+                        m = s.shape.slice(0, -2),
+                        g = va.sizeFromShape(f),
+                        y = va.sizeFromShape(m),
+                        b = Yi.assertAndGetBroadcastShape(a.shape.slice(0, -2), s.shape.slice(0, -2)).concat([p, d]);
+                    va.assert(c === h, () => `Error in matMul: inner shapes (${c}) and (${h}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);
+                    const v = o ? [y, d, h] : [y, h, d],
+                        x = LU({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: i ? [g, c, p] : [g, p, c]
+                            }
+                        }),
+                        w = LU({
+                            inputs: {
+                                x: s
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: v
+                            }
+                        }),
+                        k = n.dataIdMap.get(x.dataId).id,
+                        S = n.dataIdMap.get(w.dataId).id,
+                        I = i ? x.shape[2] : x.shape[1],
+                        A = o ? w.shape[1] : w.shape[2],
+                        E = Math.max(g, y),
+                        N = n.makeOutput([E, I, A], x.dtype),
+                        C = n.dataIdMap.get(N.dataId).id,
+                        T = new Uint8Array(new Int32Array(x.shape).buffer),
+                        R = new Uint8Array(new Int32Array(w.shape).buffer);
+                    return PU(k, T, x.shape.length, S, R, w.shape.length, i, o, C), n.disposeData(x.dataId), n.disposeData(w.dataId), N.shape = b, N
+                }
+            };
+
+            function zU(e) {
+                const {
+                    inputs: {
+                        x: t
+                    },
+                    attrs: {
+                        begin: n,
+                        size: r
+                    },
+                    backend: a
+                } = e, [s, i] = fo.parseSliceParams(t, n, r), o = fo.isSliceContinous(t.shape, s, i), u = a.readSync(t.dataId), l = a.makeOutput(i, t.dtype), c = va.computeStrides(t.shape), h = a.dataIdMap.get(l.dataId);
+                if (o) {
+                    const e = fo.computeFlatOffset(s, c);
+                    if ("string" === t.dtype) h.stringBytes = u.slice(e, e + va.sizeFromShape(i));
+                    else {
+                        a.typedArrayFromHeap(l).set(u.subarray(e, e + va.sizeFromShape(i)))
+                    }
+                    return l
+                }
+                if ("string" === t.dtype) {
+                    const e = vC(u, s, i, t.shape, t.dtype);
+                    return h.stringBytes = e, l
+                }
+                const p = a.typedArrayFromHeap(l),
+                    d = t.shape.length;
+                if (2 === d) ! function (e, t, n, r, a) {
+                    let s = 0;
+                    const i = r[0],
+                        o = r[1],
+                        u = i + a[0];
+                    for (let r = i; r < u; r++) {
+                        const i = r * t + o;
+                        n.set(e.subarray(i, i + a[1]), s), s += a[1]
+                    }
+                }(u, c[0], p, s, i);
+                else if (3 === d) ! function (e, t, n, r, a, s) {
+                    let i = 0;
+                    const o = a[0],
+                        u = a[1],
+                        l = a[2],
+                        c = o + s[0],
+                        h = u + s[1];
+                    for (let a = o; a < c; a++)
+                        for (let o = u; o < h; o++) {
+                            const u = a * t + o * n + l;
+                            r.set(e.subarray(u, u + s[2]), i), i += s[2]
+                        }
+                }(u, c[0], c[1], p, s, i);
+                else if (4 === d) ! function (e, t, n, r, a, s, i) {
+                    let o = 0;
+                    const u = s[0],
+                        l = s[1],
+                        c = s[2],
+                        h = u + i[0],
+                        p = l + i[1],
+                        d = c + i[2],
+                        f = s[3];
+                    for (let s = u; s < h; s++)
+                        for (let u = l; u < p; u++)
+                            for (let l = c; l < d; l++) {
+                                const c = s * t + u * n + l * r + f;
+                                a.set(e.subarray(c, c + i[3]), o), o += i[3]
+                            }
+                }(u, c[0], c[1], c[2], p, s, i);
+                else {
+                    const e = vC(u, s, i, t.shape, t.dtype);
+                    p.set(e)
+                }
+                return l
+            }
+            var UU = {
+                kernelName: wr,
+                backendName: "wasm",
+                kernelFunc: zU
+            };
+            var WU = {
+                kernelName: vt,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        blockShape: s,
+                        crops: i
+                    } = r, o = s.reduce((e, t) => e * t), u = jd.getReshaped(a.shape, s, o), l = jd.getPermuted(u.length, s.length), c = jd.getReshapedPermuted(a.shape, s, o), h = jd.getSliceBeginCoords(i, s.length), p = jd.getSliceSize(c, i, s.length), d = LU({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n,
+                        attrs: {
+                            shape: u
+                        }
+                    }), f = AU({
+                        inputs: {
+                            x: d
+                        },
+                        backend: n,
+                        attrs: {
+                            perm: l
+                        }
+                    }), m = zU({
+                        inputs: {
+                            x: LU({
+                                inputs: {
+                                    x: f
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: c
+                                }
+                            })
+                        },
+                        backend: n,
+                        attrs: {
+                            begin: h,
+                            size: p
+                        }
+                    });
+                    return n.disposeData(d.dataId), n.disposeData(f.dataId), n.disposeData(d.dataId), m
+                }
+            };
+
+            function VU(e) {
+                const {
+                    inputs: {
+                        x: t
+                    },
+                    attrs: {
+                        dtype: n
+                    },
+                    backend: r
+                } = e, a = r.makeOutput(t.shape, n), s = r.typedArrayFromHeap(t);
+                return r.typedArrayFromHeap(a).set(s), a
+            }
+            var jU, GU = {
+                    kernelName: St,
+                    backendName: "wasm",
+                    kernelFunc: VU
+                },
+                HU = gU(It);
+            var qU = {
+                kernelName: At,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    jU = e.wasm.cwrap(At, null, ["number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        clipValueMin: s,
+                        clipValueMax: i
+                    } = r, o = n.dataIdMap.get(a.dataId).id, u = n.makeOutput(a.shape, a.dtype), l = n.dataIdMap.get(u.dataId).id;
+                    return jU(o, s, i, l), u
+                }
+            };
+
+            function KU(e) {
+                const {
+                    inputs: t,
+                    backend: n
+                } = e, r = va.parseAxisParam(e.attrs.axis, t[0].shape)[0];
+                let a = jd.computeOutShape(t.map(e => e.shape), r);
+                const s = t.filter(e => va.sizeFromShape(e.shape) > 0);
+                if (1 === s.length) return kU({
+                    inputs: {
+                        x: s[0]
+                    },
+                    backend: n
+                });
+                const i = n.makeOutput(a, t[0].dtype);
+                if (0 === va.sizeFromShape(a)) return i;
+                const o = s.map(e => e.shape);
+                if (jd.assertParamsConsistent(o, r), "string" === s[0].dtype) {
+                    const e = s.map(e => {
+                            const t = va.sizeFromShape(e.shape.slice(r));
+                            return LU({
+                                inputs: {
+                                    x: e
+                                },
+                                backend: n,
+                                attrs: {
+                                    shape: [-1, t]
+                                }
+                            })
+                        }),
+                        o = e.map(e => ({
+                            vals: n.readSync(e.dataId),
+                            shape: e.shape
+                        }));
+                    a = jd.computeOutShape(e.map(e => e.shape), 1);
+                    const u = 1 === e[0].shape[0],
+                        l = fN(o, a, t[0].dtype, u),
+                        c = jd.computeOutShape(s.map(e => e.shape), r);
+                    i.shape = c;
+                    return n.dataIdMap.get(i.dataId).stringBytes = jd.fromStringArrayToUint8(l), e.forEach(e => n.disposeData(e.dataId)), i
+                }
+                const u = va.sizeFromShape(s[0].shape.slice(0, r));
+                let l = 0;
+                const c = s.map(e => {
+                        const t = va.sizeFromShape(e.shape.slice(r));
+                        return l += t, t
+                    }),
+                    h = s.map(e => n.typedArrayFromHeap(e)),
+                    p = n.typedArrayFromHeap(i);
+                for (let e = 0; e < u; e++) {
+                    let t = e * l;
+                    for (let n = 0; n < h.length; n++) {
+                        const r = c[n],
+                            a = e * r,
+                            s = h[n].subarray(a, a + r);
+                        p.set(s, t), t += r
+                    }
+                }
+                return i
+            }
+            var YU, XU = {
+                kernelName: Ct,
+                backendName: "wasm",
+                kernelFunc: KU
+            };
+            var QU, ZU = {
+                kernelName: Tt,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    YU = e.wasm.cwrap(Tt, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, {
+                        x: a,
+                        filter: s
+                    } = t, i = r.dataIdMap.get(a.dataId).id, o = r.dataIdMap.get(s.dataId).id, {
+                        strides: u,
+                        dilations: l,
+                        pad: c,
+                        dimRoundingMode: h,
+                        dataFormat: p
+                    } = n, d = jd.convertConv2DDataFormat(p), f = jd.computeConv2DInfo(a.shape, s.shape, u, l, c, h, !1, d), m = f.filterHeight, g = f.filterWidth, y = f.padInfo.top, b = f.padInfo.right, v = f.padInfo.bottom, x = f.padInfo.left, w = f.dilationHeight, k = f.dilationWidth, S = f.strideHeight, I = f.strideWidth, A = f.inChannels, E = f.outChannels, N = "SAME" === f.padInfo.type ? 1 : 0;
+                    if ("channelsLast" !== f.dataFormat) throw new Error(`wasm backend Conv2D does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);
+                    const C = r.makeOutput(f.outShape, "float32"),
+                        T = r.dataIdMap.get(C.dataId).id;
+                    return YU(i, a.shape[0], a.shape[1], a.shape[2], o, m, g, y, b, v, x, N, w, k, S, I, A, E, T), C
+                }
+            };
+            var JU, eW, tW, nW = {
+                    kernelName: _t,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        QU = e.wasm.cwrap(_t, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            dy: a,
+                            filter: s
+                        } = n, {
+                            strides: i,
+                            pad: o,
+                            dataFormat: u,
+                            dimRoundingMode: l,
+                            inputShape: c
+                        } = r, h = jd.convertConv2DDataFormat(u), p = jd.computeConv2DInfo(c, s.shape, i, 1, o, l, !1, h), {
+                            batchSize: d,
+                            filterHeight: f,
+                            filterWidth: m,
+                            inChannels: g,
+                            inHeight: y,
+                            inWidth: b,
+                            outChannels: v,
+                            outHeight: x,
+                            outWidth: w,
+                            strideHeight: k,
+                            strideWidth: S
+                        } = p, I = f - 1 - p.padInfo.top, A = m - 1 - p.padInfo.left, E = "channelsLast" === p.dataFormat, N = va.computeStrides(p.inShape), C = va.computeStrides(a.shape), [T, R, _] = va.computeStrides(s.shape), F = N[0], M = E ? N[1] : N[2], D = E ? N[2] : 1, O = E ? 1 : N[1], L = C[0], P = E ? C[1] : C[2], $ = E ? C[2] : 1, B = E ? 1 : C[1], z = t.makeOutput(p.inShape, "float32"), U = t.dataIdMap.get(z.dataId).id, W = t.dataIdMap.get(a.dataId).id, V = t.dataIdMap.get(s.dataId).id;
+                        return QU(W, V, d, f, m, y, b, g, x, w, v, k, S, I, A, T, R, _, F, M, D, O, L, P, $, B, U), z
+                    }
+                },
+                rW = gU(Ot),
+                aW = gU(Lt);
+            (eW = JU || (JU = {}))[eW.bilinear = 0] = "bilinear", eW[eW.nearest = 1] = "nearest";
+            var sW, iW = {
+                kernelName: $t,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    tW = e.wasm.cwrap($t, null, ["number", "number", "number", "number", "array", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        method: a,
+                        extrapolationValue: s,
+                        cropSize: i
+                    } = r, {
+                        image: o,
+                        boxes: u,
+                        boxInd: l
+                    } = n, c = u.shape[0], [h, p] = i, d = [c, h, p, o.shape[3]];
+                    let f, m = t.dataIdMap.get(o.dataId);
+                    "float32" !== o.dtype && (f = VU({
+                        backend: t,
+                        inputs: {
+                            x: o
+                        },
+                        attrs: {
+                            dtype: "float32"
+                        }
+                    }), m = t.dataIdMap.get(f.dataId));
+                    const g = m.id,
+                        y = t.dataIdMap.get(u.dataId).id,
+                        b = t.dataIdMap.get(l.dataId).id,
+                        v = t.makeOutput(d, "float32"),
+                        x = t.dataIdMap.get(v.dataId).id,
+                        w = new Uint8Array(new Int32Array(o.shape).buffer);
+                    return tW(g, y, b, c, w, h, p, JU[a], s, x), null != f && t.disposeData(f.dataId), v
+                }
+            };
+            var oW, uW = {
+                kernelName: Pt,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    sW = e.wasm.cwrap(Pt, null, ["number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        axis: s,
+                        exclusive: i,
+                        reverse: o
+                    } = r, u = a.shape.length;
+                    va.assert("float32" === a.dtype || "int32" === a.dtype, () => `cumsum does not support ${a.dtype} tensors in the WASM backend`);
+                    const l = jd.getAxesPermutation([s], u);
+                    let c = a;
+                    null !== l && (c = AU({
+                        inputs: {
+                            x: a
+                        },
+                        attrs: {
+                            perm: l
+                        },
+                        backend: n
+                    }));
+                    const h = jd.getInnerMostAxes(1, u)[0];
+                    jd.assertAxesAreInnerMostDims("cumsum", [h], u);
+                    const p = n.makeOutput(c.shape, c.dtype),
+                        d = c.shape[h],
+                        f = n.dataIdMap.get(c.dataId).id,
+                        m = n.dataIdMap.get(p.dataId).id;
+                    sW(f, i ? 1 : 0, o ? 1 : 0, d, m, lU[a.dtype]);
+                    let g = p;
+                    if (null !== l) {
+                        g = AU({
+                            inputs: {
+                                x: p
+                            },
+                            attrs: {
+                                perm: jd.getUndoAxesPermutation(l)
+                            },
+                            backend: n
+                        }), n.disposeData(c.dataId), n.disposeData(p.dataId)
+                    }
+                    return g
+                }
+            };
+            var lW, cW = {
+                kernelName: zt,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    oW = e.wasm.cwrap(zt, null, ["number", "number", "number", "array", "number", "array", "array", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = n, {
+                        blockSize: s,
+                        dataFormat: i
+                    } = r, o = a.shape[0], u = ("NHWC" === i ? a.shape[1] : a.shape[2]) * s, l = ("NHWC" === i ? a.shape[2] : a.shape[3]) * s, c = ("NHWC" === i ? a.shape[3] : a.shape[1]) / (s * s), h = "NHWC" === i ? [o, u, l, c] : [o, c, u, l], p = t.makeOutput(h, "float32"), d = t.dataIdMap.get(a.dataId).id, f = new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer), m = new Uint8Array(new Int32Array(h).buffer), g = new Uint8Array(new Int32Array(va.computeStrides(h)).buffer), y = t.dataIdMap.get(p.dataId).id;
+                    return oW(d, s, "NHWC" === i ? 1 : 0, f, a.shape.length - 1, m, g, h.length, y), p
+                }
+            };
+            var hW = {
+                    kernelName: Ut,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        lW = e.wasm.cwrap(Ut, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            attrs: n,
+                            backend: r
+                        } = e, {
+                            x: a,
+                            filter: s
+                        } = t, i = r.dataIdMap.get(a.dataId).id, o = r.dataIdMap.get(s.dataId).id, {
+                            strides: u,
+                            dilations: l,
+                            pad: c,
+                            dimRoundingMode: h
+                        } = n, p = null == l ? [1, 1] : l, d = jd.computeConv2DInfo(a.shape, s.shape, u, p, c, h, !0), f = d.filterHeight, m = d.filterWidth, g = d.padInfo.top, y = d.padInfo.right, b = d.padInfo.bottom, v = d.padInfo.left, x = d.dilationHeight, w = d.dilationWidth, k = d.strideHeight, S = d.strideWidth, I = d.inChannels, A = d.outChannels, E = "SAME" === d.padInfo.type ? 1 : 0;
+                        if ("channelsLast" !== d.dataFormat) throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${d.dataFormat}'. Please use 'channelsLast'.`);
+                        const N = r.makeOutput(d.outShape, "float32"),
+                            C = r.dataIdMap.get(N.dataId).id;
+                        return lW(i, a.shape[0], a.shape[1], a.shape[2], o, f, m, g, y, b, v, E, x, w, k, S, I, A, C), N
+                    }
+                },
+                pW = gU(Xt),
+                dW = bU(Jt, 0, "bool"),
+                fW = gU(en, "float32");
+
+            function mW(e) {
+                const {
+                    inputs: t,
+                    attrs: n,
+                    backend: r
+                } = e, {
+                    input: a
+                } = t, {
+                    dim: s
+                } = n, i = a.shape.length, o = a.shape.slice();
+                let u = s;
+                return s < 0 && (va.assert(-(i + 1) <= s, () => `Axis must be in the interval [${-(i+1)}, ${i}]`), u = i + s + 1), o.splice(u, 0, 1), LU({
+                    inputs: {
+                        x: a
+                    },
+                    backend: r,
+                    attrs: {
+                        shape: o
+                    }
+                })
+            }
+            var gW = {
+                kernelName: tn,
+                backendName: "wasm",
+                kernelFunc: mW
+            };
+
+            function yW(e) {
+                const {
+                    attrs: {
+                        shape: t,
+                        value: n,
+                        dtype: r
+                    },
+                    backend: a
+                } = e, s = a.makeOutput(t, r);
+                return a.typedArrayFromHeap(s).fill(n), s
+            }
+            var bW, vW = {
+                kernelName: an,
+                backendName: "wasm",
+                kernelFunc: yW
+            };
+            var xW, wW = {
+                    kernelName: sn,
+                    backendName: "wasm",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n
+                        } = e, {
+                            image: r
+                        } = t, a = n.makeOutput(r.shape, r.dtype), s = n.dataIdMap.get(r.dataId).id, i = n.dataIdMap.get(a.dataId).id, [o, u, l, c] = r.shape;
+                        return bW(s, o, u, l, c, i), a
+                    },
+                    setupFunc: function (e) {
+                        bW = e.wasm.cwrap(sn, null, ["number", "number", "number", "number", "number", "number"])
+                    }
+                },
+                kW = gU(on),
+                SW = bU(un);
+            var IW, AW = {
+                kernelName: ln,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    xW = e.wasm.cwrap(ln, null, ["number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        varianceEpsilon: a
+                    } = r, {
+                        x: s,
+                        mean: i,
+                        variance: o,
+                        offset: u,
+                        scale: l
+                    } = n, c = t.dataIdMap.get(s.dataId).id, h = t.dataIdMap.get(i.dataId).id, p = t.dataIdMap.get(o.dataId).id, d = null != u ? t.dataIdMap.get(u.dataId).id : 0, f = null != l ? t.dataIdMap.get(l.dataId).id : 0, m = t.makeOutput(s.shape, s.dtype);
+                    if (0 === va.sizeFromShape(s.shape)) return m;
+                    const g = t.dataIdMap.get(m.dataId).id;
+                    return xW(c, h, p, d, f, a, g), m
+                }
+            };
+            var EW, NW = {
+                kernelName: aa,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    IW = e.wasm.cwrap(aa, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, {
+                        x: a,
+                        filter: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t, {
+                        strides: u,
+                        pad: l,
+                        dilations: c,
+                        dataFormat: h,
+                        dimRoundingMode: p,
+                        activation: d,
+                        leakyreluAlpha: f
+                    } = n, m = jd.computeConv2DInfo(a.shape, s.shape, u, c, l, p), g = hU[d];
+                    if (null == g) throw new Error(d + " activation not yet supported for FusedConv2D in the wasm backend.");
+                    const y = r.dataIdMap.get(a.dataId).id,
+                        b = r.dataIdMap.get(s.dataId).id,
+                        v = m.outChannels;
+                    let x = 0;
+                    if (null != i) {
+                        const e = r.dataIdMap.get(i.dataId);
+                        if (1 !== e.shape.length) throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${e.shape.length}.`);
+                        if (e.shape[0] !== v) throw new Error(`FusedConv2D bias shape (${e.shape}) does not match the number of output channels (${v})`);
+                        x = e.id
+                    }
+                    const w = m.filterHeight,
+                        k = m.filterWidth,
+                        S = m.padInfo.top,
+                        I = m.padInfo.right,
+                        A = m.padInfo.bottom,
+                        E = m.padInfo.left,
+                        N = m.dilationHeight,
+                        C = m.dilationWidth,
+                        T = m.strideHeight,
+                        R = m.strideWidth,
+                        _ = m.inChannels,
+                        F = "SAME" === m.padInfo.type ? 1 : 0,
+                        M = m.batchSize,
+                        D = m.inHeight,
+                        O = m.inWidth;
+                    if ("NHWC" !== h) throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${h}'. Please use 'NHWC'.`);
+                    const L = r.makeOutput(m.outShape, "float32"),
+                        P = r.dataIdMap.get(L.dataId).id,
+                        $ = null == o ? 0 : r.dataIdMap.get(o.dataId).id;
+                    return IW(y, M, D, O, b, w, k, x, S, I, A, E, F, N, C, T, R, _, v, g, $, f || 0, P), L
+                }
+            };
+            var CW, TW = {
+                kernelName: sa,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    EW = e.wasm.cwrap(sa, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, {
+                        x: a,
+                        filter: s,
+                        bias: i,
+                        preluActivationWeights: o
+                    } = t, {
+                        strides: u,
+                        pad: l,
+                        dilations: c,
+                        dataFormat: h,
+                        dimRoundingMode: p,
+                        activation: d,
+                        leakyreluAlpha: f
+                    } = n, m = jd.computeConv2DInfo(a.shape, s.shape, u, c, l, p, !0), g = hU[d];
+                    if (null == g) throw new Error(d + " activation not yet supported for FusedDepthwiseConv2D in the wasm backend.");
+                    const y = r.dataIdMap.get(a.dataId).id,
+                        b = r.dataIdMap.get(s.dataId).id,
+                        v = m.outChannels;
+                    let x = 0;
+                    if (null != i) {
+                        const e = r.dataIdMap.get(i.dataId);
+                        if (1 !== e.shape.length) throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${e.shape.length}.`);
+                        if (e.shape[0] !== v) throw new Error(`FusedDepthwiseConv2D bias shape (${e.shape}) does not match the number of output channels (${v})`);
+                        x = e.id
+                    }
+                    const w = m.filterHeight,
+                        k = m.filterWidth,
+                        S = m.padInfo.top,
+                        I = m.padInfo.right,
+                        A = m.padInfo.bottom,
+                        E = m.padInfo.left,
+                        N = m.dilationHeight,
+                        C = m.dilationWidth,
+                        T = m.strideHeight,
+                        R = m.strideWidth,
+                        _ = m.inChannels,
+                        F = "SAME" === m.padInfo.type ? 1 : 0,
+                        M = m.batchSize,
+                        D = m.inHeight,
+                        O = m.inWidth;
+                    if ("NHWC" !== h) throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${h}'. Please use 'NHWC'.`);
+                    const L = r.makeOutput(m.outShape, "float32"),
+                        P = r.dataIdMap.get(L.dataId).id,
+                        $ = null == o ? 0 : r.dataIdMap.get(o.dataId).id;
+                    return EW(y, M, D, O, b, w, k, x, S, I, A, E, F, N, C, T, R, _, v, g, $, f || 0, P), L
+                }
+            };
+            var RW, _W = {
+                kernelName: hn,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    CW = e.wasm.cwrap(hn, null, ["number", "number", "number", "number", "number", "number", "array", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n
+                    } = e, {
+                        params: r,
+                        indices: a
+                    } = n, [s, i, o, u] = oo.prepareAndValidate(r, a), l = t.makeOutput(s, r.dtype);
+                    if (0 === i) return l;
+                    const c = a.shape,
+                        h = c[c.length - 1],
+                        p = t.dataIdMap.get(r.dataId).id,
+                        d = t.dataIdMap.get(a.dataId).id,
+                        f = new Uint8Array(new Int32Array(u).buffer),
+                        m = t.dataIdMap.get(l.dataId).id;
+                    return CW(p, lU[r.dtype], d, i, h, o, f, m), l
+                }
+            };
+            var FW, MW = {
+                    kernelName: cn,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        RW = e.wasm.cwrap("Gather", null, ["number", "number", "array", "number", "number", "number", "array", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            x: a,
+                            indices: s
+                        } = n, {
+                            axis: i,
+                            batchDims: o
+                        } = r, u = va.parseAxisParam(i, a.shape)[0], l = t.readSync(s.dataId), c = a.shape[u];
+                        for (let e = 0; e < l.length; ++e) {
+                            const t = l[e];
+                            va.assert(t <= c - 1 && t >= 0, () => `GatherV2: the index value ${t} is not in [0, ${c-1}]`)
+                        }
+                        const h = jd.segment_util.collectGatherOpShapeInfo(a, s, u, o),
+                            p = LU({
+                                inputs: {
+                                    x: a
+                                },
+                                attrs: {
+                                    shape: [h.batchSize, h.outerSize, h.dimSize, h.sliceSize]
+                                },
+                                backend: t
+                            }),
+                            d = va.sizeFromShape(s.shape),
+                            f = LU({
+                                inputs: {
+                                    x: s
+                                },
+                                attrs: {
+                                    shape: [h.batchSize, d / h.batchSize]
+                                },
+                                backend: t
+                            }),
+                            m = [h.batchSize, h.outerSize, d / h.batchSize, h.sliceSize],
+                            g = t.makeOutput(m, a.dtype);
+                        if (0 === va.sizeFromShape(a.shape)) return g;
+                        const y = p.shape.length - 1,
+                            b = t.dataIdMap.get(p.dataId).id,
+                            v = t.dataIdMap.get(f.dataId).id,
+                            x = t.dataIdMap.get(g.dataId).id,
+                            w = new Uint8Array(new Int32Array(va.computeStrides(p.shape)).buffer),
+                            k = new Uint8Array(new Int32Array(va.computeStrides(m)).buffer);
+                        return RW(b, lU[a.dtype], w, y, v, h.batchSize, k, x), t.disposeData(p.dataId), t.disposeData(f.dataId), g.shape = h.outputShape, g
+                    }
+                },
+                DW = bU(pn, 0, "bool"),
+                OW = bU(dn, 0, "bool");
+            var LW, PW = {
+                    kernelName: xn,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        FW = e.wasm.cwrap(xn, null, ["number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: {
+                                x: t
+                            },
+                            attrs: {
+                                alpha: n
+                            },
+                            backend: r
+                        } = e, a = r.dataIdMap.get(t.dataId).id, s = r.makeOutput(t.shape, "float32");
+                        if (0 !== va.sizeFromShape(t.shape)) {
+                            const e = r.dataIdMap.get(s.dataId).id;
+                            FW(a, lU[t.dtype], n, e)
+                        }
+                        return s
+                    }
+                },
+                $W = bU(wn, 0, "bool"),
+                BW = bU(kn, 0, "bool"),
+                zW = gU(In),
+                UW = bU(En, 0, "bool");
+            var WW, VW = {
+                    kernelName: Fn,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        LW = e.wasm.cwrap(Fn, null, ["number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            reductionIndices: a,
+                            keepDims: s
+                        } = r, {
+                            x: i
+                        } = n;
+                        let o = t.dataIdMap.get(i.dataId).id,
+                            u = i;
+                        const {
+                            transposed: l,
+                            axes: c,
+                            originalAxes: h,
+                            inputWasTransposed: p
+                        } = CU(i, a, t);
+                        if (p) {
+                            u = l, o = t.dataIdMap.get(l.dataId).id
+                        }
+                        const d = u.shape.length;
+                        jd.assertAxesAreInnerMostDims("max", c, d);
+                        const [f, m] = jd.computeOutAndReduceShapes(u.shape, c), g = va.sizeFromShape(m), y = t.makeOutput(f, i.dtype);
+                        if (0 !== va.sizeFromShape(u.shape)) {
+                            const e = t.dataIdMap.get(y.dataId).id;
+                            LW(o, lU[i.dtype], g, e)
+                        }
+                        if (p && t.disposeData(l.dataId), s) {
+                            const e = jd.expandShapeToKeepDim(y.shape, h);
+                            y.shape = e
+                        }
+                        return y
+                    }
+                },
+                jW = bU(Mn);
+            var GW, HW = {
+                kernelName: Dn,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    WW = e.wasm.cwrap(Dn, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        attrs: n,
+                        backend: r
+                    } = e, a = t.x, s = r.dataIdMap.get(a.dataId).id;
+                    va.assert("float32" === a.dtype, () => `Error in MaxPool: only float32 input is supported. Got ${a.dtype}.`);
+                    const {
+                        filterSize: i,
+                        strides: o,
+                        pad: u,
+                        dimRoundingMode: l
+                    } = n, c = jd.computePool2DInfo(a.shape, i, o, 1, u, l), h = c.filterHeight, p = c.filterWidth, d = c.padInfo.top, f = c.padInfo.right, m = c.padInfo.bottom, g = c.padInfo.left, y = c.dilationHeight, b = c.dilationWidth, v = c.strideHeight, x = c.strideWidth, w = c.inChannels, k = c.outChannels;
+                    if ("channelsLast" !== c.dataFormat) throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);
+                    const S = r.makeOutput(c.outShape, "float32"),
+                        I = r.dataIdMap.get(S.dataId).id;
+                    return WW(s, a.shape[0], a.shape[1], a.shape[2], h, p, d, f, m, g, y, b, v, x, w, k, I), S
+                }
+            };
+            var qW, KW = {
+                kernelName: Bn,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    GW = e.wasm.cwrap(Bn, null, ["number, number, number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        axis: a,
+                        keepDims: s
+                    } = r, {
+                        x: i
+                    } = n, o = t.dataIdMap.get(i.dataId).id;
+                    let u = o,
+                        l = i;
+                    const {
+                        transposed: c,
+                        axes: h,
+                        originalAxes: p,
+                        inputWasTransposed: d
+                    } = CU(i, a, t);
+                    let f = h;
+                    if (d) {
+                        const e = t.dataIdMap.get(c.dataId).id;
+                        e !== o && (l = c, u = e, f = jd.getInnerMostAxes(f.length, l.shape.length))
+                    }
+                    jd.assertAxesAreInnerMostDims("mean", f, l.shape.length);
+                    const [m, g] = jd.computeOutAndReduceShapes(l.shape, f), y = va.sizeFromShape(g);
+                    let b = l;
+                    "float32" !== l.dtype && (b = VU({
+                        backend: t,
+                        inputs: {
+                            x: l
+                        },
+                        attrs: {
+                            dtype: "float32"
+                        }
+                    }), u = t.dataIdMap.get(b.dataId).id);
+                    const v = t.makeOutput(m, "float32");
+                    if (0 !== va.sizeFromShape(l.shape)) {
+                        const e = t.dataIdMap.get(v.dataId).id;
+                        GW(u, y, e)
+                    }
+                    if (d && t.disposeData(c.dataId), s) {
+                        const e = jd.expandShapeToKeepDim(v.shape, p);
+                        v.shape = e
+                    }
+                    return "float32" !== l.dtype && t.disposeData(b.dataId), v
+                }
+            };
+            var YW, XW, QW, ZW = {
+                    kernelName: zn,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        qW = e.wasm.cwrap(zn, null, ["number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            axis: a,
+                            keepDims: s
+                        } = r, {
+                            x: i
+                        } = n, o = t.dataIdMap.get(i.dataId).id;
+                        let u = o,
+                            l = i;
+                        const {
+                            transposed: c,
+                            axes: h,
+                            originalAxes: p,
+                            inputWasTransposed: d
+                        } = CU(i, a, t);
+                        if (d) {
+                            const e = t.dataIdMap.get(c.dataId).id;
+                            e !== o && (l = c, u = e)
+                        }
+                        const f = l.shape.length;
+                        jd.assertAxesAreInnerMostDims("min", h, f);
+                        const [m, g] = jd.computeOutAndReduceShapes(l.shape, h), y = va.sizeFromShape(g), b = t.makeOutput(m, l.dtype);
+                        if (0 !== va.sizeFromShape(l.shape)) {
+                            const e = t.dataIdMap.get(b.dataId).id;
+                            qW(u, lU[i.dtype], y, e)
+                        }
+                        if (d && t.disposeData(c.dataId), s) {
+                            const e = jd.expandShapeToKeepDim(b.shape, p);
+                            b.shape = e
+                        }
+                        return b
+                    }
+                },
+                JW = bU(Un);
+            (XW = YW || (YW = {}))[XW.reflect = 0] = "reflect", XW[XW.symmetric = 1] = "symmetric";
+            var eV, tV = {
+                    kernelName: Wn,
+                    backendName: "wasm",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: {
+                                x: t
+                            },
+                            backend: n,
+                            attrs: {
+                                paddings: r,
+                                mode: a
+                            }
+                        } = e, s = r.map((e, n) => e[0] + t.shape[n] + e[1]), i = n.dataIdMap.get(t.dataId).id, o = n.makeOutput(s, t.dtype), u = n.dataIdMap.get(o.dataId).id, l = new Uint8Array(new Int32Array(t.shape).buffer), c = r.map(e => e[0]), h = r.map(e => e[1]), p = new Uint8Array(new Int32Array(c).buffer), d = new Uint8Array(new Int32Array(h).buffer);
+                        return QW(i, l, t.shape.length, lU[t.dtype], p, d, YW[a], u), o
+                    },
+                    setupFunc: function (e) {
+                        QW = e.wasm.cwrap(Wn, null, ["number", "array", "number", "number", "array", "array", "number", "number"])
+                    }
+                },
+                nV = bU(Gn),
+                rV = gU(Hn);
+
+            function aV(e, t) {
+                const n = new Int32Array(e.wasm.HEAPU8.buffer, t, 4),
+                    r = n[0],
+                    a = n[1],
+                    s = n[2],
+                    i = n[3];
+                return e.wasm._free(t), {
+                    pSelectedIndices: r,
+                    selectedSize: a,
+                    pSelectedScores: s,
+                    pValidOutputs: i
+                }
+            }
+            var sV, iV = {
+                kernelName: Kn,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    eV = e.wasm.cwrap(Kn, "number", ["number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        iouThreshold: a,
+                        maxOutputSize: s,
+                        scoreThreshold: i
+                    } = r, {
+                        boxes: o,
+                        scores: u
+                    } = n, l = t.dataIdMap.get(o.dataId).id, c = t.dataIdMap.get(u.dataId).id, h = eV(l, c, s, a, i), {
+                        pSelectedIndices: p,
+                        selectedSize: d,
+                        pSelectedScores: f,
+                        pValidOutputs: m
+                    } = aV(t, h);
+                    return t.wasm._free(f), t.wasm._free(m), t.makeOutput([d], "int32", p)
+                }
+            };
+            var oV, uV = {
+                kernelName: Yn,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    sV = e.wasm.cwrap(Yn, "number", ["number", "number", "number", "number", "number", "bool"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        iouThreshold: a,
+                        maxOutputSize: s,
+                        scoreThreshold: i,
+                        padToMaxOutputSize: o
+                    } = r, {
+                        boxes: u,
+                        scores: l
+                    } = n, c = t.dataIdMap.get(u.dataId).id, h = t.dataIdMap.get(l.dataId).id, p = sV(c, h, s, a, i, o), {
+                        pSelectedIndices: d,
+                        selectedSize: f,
+                        pSelectedScores: m,
+                        pValidOutputs: g
+                    } = aV(t, p);
+                    return t.wasm._free(m), [t.makeOutput([f], "int32", d), t.makeOutput([], "int32", g)]
+                }
+            };
+            var lV, cV = {
+                    kernelName: Xn,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        oV = e.wasm.cwrap(Xn, "number", ["number", "number", "number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            iouThreshold: a,
+                            maxOutputSize: s,
+                            scoreThreshold: i,
+                            softNmsSigma: o
+                        } = r, {
+                            boxes: u,
+                            scores: l
+                        } = n, c = t.dataIdMap.get(u.dataId).id, h = t.dataIdMap.get(l.dataId).id, p = oV(c, h, s, a, i, o), {
+                            pSelectedIndices: d,
+                            selectedSize: f,
+                            pSelectedScores: m,
+                            pValidOutputs: g
+                        } = aV(t, p);
+                        return t.wasm._free(g), [t.makeOutput([f], "int32", d), t.makeOutput([f], "float32", m)]
+                    }
+                },
+                hV = bU(qn, 0, "bool");
+            var pV = {
+                kernelName: Zn,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    lV = e.wasm.cwrap(Zn, null, ["number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        indices: a
+                    } = t, {
+                        depth: s,
+                        onValue: i,
+                        offValue: o
+                    } = r, u = n.makeOutput([...a.shape, s], "int32"), l = n.dataIdMap.get(u.dataId).id, c = n.dataIdMap.get(a.dataId).id;
+                    return lV(c, s, i, o, l), u
+                }
+            };
+            var dV = {
+                kernelName: Qn,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: {
+                            x: t
+                        },
+                        backend: n
+                    } = e, r = n.makeOutput(t.shape, t.dtype);
+                    return n.typedArrayFromHeap(r).fill(1), r
+                }
+            };
+            var fV, mV = {
+                kernelName: Jn,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        axis: a
+                    } = r;
+                    if (1 === t.length) return mW({
+                        inputs: {
+                            input: t[0]
+                        },
+                        backend: n,
+                        attrs: {
+                            dim: a
+                        }
+                    });
+                    const s = t[0].shape,
+                        i = t[0].dtype;
+                    t.forEach(e => {
+                        va.assertShapesMatch(s, e.shape, "All tensors passed to stack must have matching shapes"), va.assert(i === e.dtype, () => "All tensors passed to stack must have matching dtypes")
+                    });
+                    const o = [],
+                        u = KU({
+                            inputs: t.map(e => {
+                                const t = mW({
+                                    inputs: {
+                                        input: e
+                                    },
+                                    backend: n,
+                                    attrs: {
+                                        dim: a
+                                    }
+                                });
+                                return o.push(t), t
+                            }),
+                            backend: n,
+                            attrs: {
+                                axis: a
+                            }
+                        });
+                    return o.forEach(e => n.disposeData(e.dataId)), u
+                }
+            };
+            var gV, yV = {
+                    kernelName: er,
+                    backendName: "wasm",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: {
+                                x: t
+                            },
+                            backend: n,
+                            attrs: {
+                                paddings: r,
+                                constantValue: a
+                            }
+                        } = e, s = r.map((e, n) => e[0] + t.shape[n] + e[1]);
+                        if (0 === va.sizeFromShape(t.shape)) return yW({
+                            backend: n,
+                            attrs: {
+                                shape: s,
+                                value: a,
+                                dtype: t.dtype
+                            }
+                        });
+                        const i = n.dataIdMap.get(t.dataId).id,
+                            o = n.makeOutput(s, t.dtype),
+                            u = n.dataIdMap.get(o.dataId).id,
+                            l = new Uint8Array(new Int32Array(t.shape).buffer),
+                            c = r.map(e => e[0]),
+                            h = r.map(e => e[1]),
+                            p = new Uint8Array(new Int32Array(c).buffer),
+                            d = new Uint8Array(new Int32Array(h).buffer);
+                        return fV(i, l, t.shape.length, lU[t.dtype], p, d, a, u), o
+                    },
+                    setupFunc: function (e) {
+                        fV = e.wasm.cwrap(er, null, ["number", "array", "number", "number", "array", "array", "number", "number"])
+                    }
+                },
+                bV = bU(nr);
+            var vV, xV = {
+                kernelName: rr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    gV = e.wasm.cwrap(rr, null, ["number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        x: r,
+                        alpha: a
+                    } = t, s = n.dataIdMap.get(r.dataId).id, i = n.dataIdMap.get(a.dataId).id;
+                    let o = s;
+                    const u = r;
+                    let l = u;
+                    "float32" !== u.dtype && (l = VU({
+                        backend: n,
+                        inputs: {
+                            x: r
+                        },
+                        attrs: {
+                            dtype: "float32"
+                        }
+                    }), o = n.dataIdMap.get(l.dataId).id);
+                    const c = n.makeOutput(r.shape, "float32"),
+                        h = n.dataIdMap.get(c.dataId).id;
+                    return gV(o, i, h), "float32" !== u.dtype && n.disposeData(l.dataId), c
+                }
+            };
+            var wV, kV = {
+                    kernelName: ar,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        vV = e.wasm.cwrap(ar, null, ["number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            axis: a,
+                            keepDims: s
+                        } = r, {
+                            x: i
+                        } = n, o = t.dataIdMap.get(i.dataId).id;
+                        let u = o,
+                            l = i;
+                        const {
+                            transposed: c,
+                            axes: h,
+                            originalAxes: p,
+                            inputWasTransposed: d
+                        } = CU(i, a, t);
+                        let f = h;
+                        if (d) {
+                            const e = t.dataIdMap.get(c.dataId).id;
+                            e !== o && (l = c, u = e, f = jd.getInnerMostAxes(f.length, l.shape.length))
+                        }
+                        jd.assertAxesAreInnerMostDims("prod", f, l.shape.length);
+                        const [m, g] = jd.computeOutAndReduceShapes(l.shape, f), y = va.sizeFromShape(g), b = t.makeOutput(m, l.dtype);
+                        if (0 !== va.sizeFromShape(l.shape)) {
+                            const e = t.dataIdMap.get(b.dataId).id;
+                            vV(u, y, lU[b.dtype], e)
+                        }
+                        if (d && t.disposeData(c.dataId), s) {
+                            const e = jd.expandShapeToKeepDim(b.shape, p);
+                            b.shape = e
+                        }
+                        return b
+                    }
+                },
+                SV = {
+                    kernelName: sr,
+                    backendName: "wasm",
+                    kernelFunc: e => {
+                        const {
+                            backend: t,
+                            attrs: n
+                        } = e, {
+                            start: r,
+                            stop: a,
+                            step: s,
+                            dtype: i
+                        } = n, o = pC(r, a, s, i), u = t.makeOutput([o.length], i);
+                        return t.typedArrayFromHeap(u).set(o), u
+                    }
+                },
+                IV = bU(Kt),
+                AV = gU(ur),
+                EV = gU(fr);
+            var NV, CV = {
+                kernelName: pr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    wV = e.wasm.cwrap(pr, null, ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        images: a
+                    } = n, {
+                        alignCorners: s,
+                        halfPixelCenters: i,
+                        size: o
+                    } = r, [u, l] = o, [c, h, p, d] = a.shape, f = [c, u, l, d];
+                    let m, g = t.dataIdMap.get(a.dataId);
+                    "float32" !== g.dtype && (m = VU({
+                        backend: t,
+                        inputs: {
+                            x: a
+                        },
+                        attrs: {
+                            dtype: "float32"
+                        }
+                    }), g = t.dataIdMap.get(m.dataId));
+                    const y = g.id,
+                        b = t.makeOutput(f, "float32");
+                    if (0 === va.sizeFromShape(a.shape)) return b;
+                    const v = t.dataIdMap.get(b.dataId).id;
+                    return wV(y, c, h, p, d, u, l, s ? 1 : 0, i ? 1 : 0, v), null != m && t.disposeData(m.dataId), b
+                }
+            };
+            var TV, RV = {
+                kernelName: mr,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        dims: s
+                    } = r, i = va.parseAxisParam(s, a.shape);
+                    if (0 === a.shape.length) return kU({
+                        inputs: {
+                            x: a
+                        },
+                        backend: n
+                    });
+                    const o = n.makeOutput(a.shape, a.dtype),
+                        u = n.dataIdMap.get(a.dataId).id,
+                        l = n.dataIdMap.get(o.dataId).id,
+                        c = new Uint8Array(new Int32Array(i).buffer),
+                        h = new Uint8Array(new Int32Array(a.shape).buffer);
+                    NV(u, c, i.length, h, a.shape.length, l);
+                    const p = LU({
+                        inputs: {
+                            x: o
+                        },
+                        attrs: {
+                            shape: a.shape
+                        },
+                        backend: n
+                    });
+                    return n.disposeData(o.dataId), p
+                },
+                setupFunc: function (e) {
+                    NV = e.wasm.cwrap(mr, null, ["number", "array", "number", "array", "number", "number"])
+                }
+            };
+            var _V, FV = {
+                    kernelName: na,
+                    backendName: "wasm",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            backend: n,
+                            attrs: r
+                        } = e, {
+                            image: a
+                        } = t, {
+                            radians: s,
+                            fillValue: i,
+                            center: o
+                        } = r, u = n.makeOutput(a.shape, a.dtype), l = n.dataIdMap.get(a.dataId).id, c = n.dataIdMap.get(u.dataId).id, [h, p, d, f] = a.shape, [m, g] = jd.getImageCenter(o, p, d), y = "number" == typeof i ? [i, i, i, 0 === i ? 0 : 255] : [...i, 255], b = new Uint8Array(new Int32Array(y).buffer);
+                        return TV(l, h, p, d, f, s, m, g, b, y.length, c), u
+                    },
+                    setupFunc: function (e) {
+                        TV = e.wasm.cwrap(na, null, ["number", "number", "number", "number", "number", "number", "number", "number", "array", "number", "number"])
+                    }
+                },
+                MV = gU(gr),
+                DV = gU(yr);
+            var OV, LV = {
+                kernelName: br,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    _V = e.wasm.cwrap(br, null, ["number", "number", "number", "number", "number", "number", "array", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        indices: a,
+                        updates: s
+                    } = n, {
+                        shape: i
+                    } = r, o = t.makeOutput(i, s.dtype);
+                    if (0 === va.sizeFromShape(i)) return o;
+                    const {
+                        sliceRank: u,
+                        numUpdates: l,
+                        sliceSize: c,
+                        strides: h,
+                        outputSize: p
+                    } = lo.calculateShapes(s, a, i), d = t.dataIdMap.get(a.dataId).id, f = t.dataIdMap.get(s.dataId).id, m = new Uint8Array(new Int32Array(h).buffer), g = t.dataIdMap.get(o.dataId).id;
+                    return _V(d, f, lU[s.dtype], u, l, c, m, p, g), o
+                }
+            };
+            var PV, $V = {
+                kernelName: vr,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n
+                    } = e, {
+                        condition: r,
+                        t: a,
+                        e: s
+                    } = t, i = n.dataIdMap.get(r.dataId).id, o = n.dataIdMap.get(a.dataId).id, u = n.dataIdMap.get(s.dataId).id, l = n.makeOutput(a.shape, a.dtype), c = n.dataIdMap.get(l.dataId).id, h = r.shape.length, p = a.shape.length, d = 0 === h || h > 1 || 1 === p ? 1 : va.sizeFromShape(a.shape.slice(1));
+                    return OV(i, o, u, d, c), l
+                },
+                setupFunc: function (e) {
+                    OV = e.wasm.cwrap("SelectV2", null, ["number", "number", "number", "number", "number"])
+                }
+            };
+            var BV, zV = {
+                    kernelName: "Sigmoid",
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        PV = e.wasm.cwrap(Ar, null, ["number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: {
+                                x: n
+                            }
+                        } = e, r = t.dataIdMap.get(n.dataId).id, a = t.makeOutput(n.shape, n.dtype), s = t.dataIdMap.get(a.dataId).id;
+                        return 0 === va.sizeFromShape(a.shape) || PV(r, s), a
+                    }
+                },
+                UV = gU(kr);
+            var WV = {
+                kernelName: _r,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    BV = e.wasm.cwrap(_r, null, ["number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: {
+                            logits: n
+                        },
+                        attrs: {
+                            dim: r
+                        }
+                    } = e, a = t.dataIdMap.get(n.dataId).id, s = t.makeOutput(n.shape, n.dtype), i = t.dataIdMap.get(s.dataId).id, o = n.shape[r], u = va.sizeFromShape(n.shape) / o;
+                    return 0 === va.sizeFromShape(s.shape) || BV(a, i, o, u), s
+                }
+            };
+            var VV, jV = {
+                kernelName: Tr,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, {
+                        blockShape: s,
+                        paddings: i
+                    } = r, o = va.sizeFromShape(s), u = [
+                        [0, 0]
+                    ];
+                    u.push(...i);
+                    for (let e = 1 + s.length; e < a.shape.length; ++e) u.push([0, 0]);
+                    const l = yV.kernelFunc({
+                            inputs: {
+                                x: a
+                            },
+                            backend: n,
+                            attrs: {
+                                paddings: u,
+                                constantValue: 0
+                            }
+                        }),
+                        c = jd.getReshaped(l.shape, s, o, !1),
+                        h = jd.getPermuted(c.length, s.length, !1),
+                        p = jd.getReshapedPermuted(l.shape, s, o, !1),
+                        d = LU({
+                            inputs: {
+                                x: l
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: c
+                            }
+                        }),
+                        f = AU({
+                            inputs: {
+                                x: d
+                            },
+                            backend: n,
+                            attrs: {
+                                perm: h
+                            }
+                        }),
+                        m = LU({
+                            inputs: {
+                                x: f
+                            },
+                            backend: n,
+                            attrs: {
+                                shape: p
+                            }
+                        });
+                    return n.disposeData(l.dataId), n.disposeData(d.dataId), n.disposeData(f.dataId), m
+                }
+            };
+            var GV, HV = {
+                kernelName: Fr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    VV = e.wasm.cwrap("SparseFillEmptyRows", "number", ["number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n
+                    } = e, {
+                        indices: r,
+                        values: a,
+                        denseShape: s,
+                        defaultValue: i
+                    } = n, o = r.shape[0], u = r.shape[1], l = t.readSync(s.dataId)[0], c = [o + l, u], h = t.dataIdMap.get(r.dataId).id, p = t.dataIdMap.get(a.dataId).id, d = t.dataIdMap.get(i.dataId).id, f = t.makeOutput(c, r.dtype), m = t.dataIdMap.get(f.dataId).id, g = t.makeOutput(c.slice(0, 1), a.dtype), y = t.dataIdMap.get(g.dataId).id, b = t.makeOutput([l], "bool"), v = t.dataIdMap.get(b.dataId).id, x = t.makeOutput([o], r.dtype), w = t.dataIdMap.get(x.dataId).id, k = t.makeOutput([4], "int32"), S = t.dataIdMap.get(k.dataId).id, I = VV(h, p, lU[a.dtype], o, l, u, d, m, y, v, w, S), A = t.readSync(k.dataId);
+                    let E;
+                    switch (A[0]) {
+                        case 1:
+                            E = jd.getSparseFillEmptyRowsIndicesDenseShapeMismatch(A[1]);
+                            break;
+                        case 2:
+                            E = jd.getSparseFillEmptyRowsNegativeIndexErrorMessage(A[1], A[2]);
+                            break;
+                        case 3:
+                            E = jd.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(A[1], A[2], A[3]);
+                            break;
+                        default:
+                            E = ""
+                    }
+                    if (t.disposeData(k.dataId), E) throw t.disposeData(f.dataId), t.disposeData(g.dataId), t.disposeData(b.dataId), t.disposeData(x.dataId), new Error(E);
+                    let N = f,
+                        C = g;
+                    return I !== c[0] && (N = zU({
+                        inputs: {
+                            x: f
+                        },
+                        attrs: {
+                            begin: 0,
+                            size: [I, u]
+                        },
+                        backend: t
+                    }), C = zU({
+                        inputs: {
+                            x: g
+                        },
+                        attrs: {
+                            begin: 0,
+                            size: I
+                        },
+                        backend: t
+                    }), t.disposeData(f.dataId), t.disposeData(g.dataId)), [N, C, b, x]
+                }
+            };
+            var qV, KV = {
+                kernelName: Mr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    GV = e.wasm.cwrap(Mr, null, ["number", "number", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n
+                    } = e, {
+                        inputIndices: r,
+                        inputShape: a,
+                        newShape: s
+                    } = n;
+                    if (2 !== r.shape.length) throw new Error("Input indices should be a matrix but received shape\n        " + r.shape);
+                    if (1 !== a.shape.length) throw new Error("Input shape should be a vector but received shape\n        " + a.shape);
+                    if (1 !== s.shape.length) throw new Error("Target shape should be a vector but received shape " + s.shape);
+                    const i = t.dataIdMap.get(r.dataId).id,
+                        o = t.dataIdMap.get(a.dataId).id,
+                        u = t.dataIdMap.get(s.dataId).id,
+                        l = r.shape[0],
+                        c = va.sizeFromShape(s.shape),
+                        h = t.makeOutput([l, c], r.dtype),
+                        p = t.dataIdMap.get(h.dataId).id,
+                        d = t.makeOutput([c], s.dtype),
+                        f = t.dataIdMap.get(d.dataId).id,
+                        m = t.makeOutput([3], "int32"),
+                        g = t.dataIdMap.get(m.dataId).id;
+                    GV(i, o, u, l, p, f, g);
+                    const y = t.readSync(m.dataId);
+                    let b;
+                    switch (y[0]) {
+                        case 0:
+                            b = jd.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(y[1], y[2]);
+                            break;
+                        case 1:
+                            b = jd.getSparseReshapeNegativeOutputDimErrorMessage(y[1], y[2]);
+                            break;
+                        case 2:
+                            b = jd.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();
+                            break;
+                        case 3: {
+                            const e = Array.from(t.readSync(a.dataId)),
+                                n = Array.from(t.readSync(d.dataId));
+                            b = jd.getSparseReshapeInputOutputMultipleErrorMessage(e, n);
+                            break
+                        }
+                        case 4: {
+                            const e = Array.from(t.readSync(a.dataId)),
+                                n = Array.from(t.readSync(d.dataId));
+                            b = jd.getSparseReshapeInputOutputMismatchErrorMessage(e, n);
+                            break
+                        }
+                        default:
+                            b = ""
+                    }
+                    if (t.disposeData(m.dataId), b) throw t.disposeData(h.dataId), t.disposeData(d.dataId), new Error(b);
+                    return [h, d]
+                }
+            };
+
+            function YV(e) {
+                qV = e.wasm.cwrap("SparseSegmentReduction", null, ["number", "number", "number", "number", "number", "number", "number", "number", "number"])
+            }
+
+            function XV(e, t) {
+                const {
+                    backend: n,
+                    inputs: r
+                } = e, {
+                    data: a,
+                    indices: s,
+                    segmentIds: i
+                } = r, o = s.shape[0], u = n.readSync(i.dataId, o - 1, o)[0], l = o > 0 ? u + 1 : 0;
+                if (l < 0) throw new Error(jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+                const c = a.shape.slice();
+                c[0] = l;
+                const h = n.dataIdMap.get(a.dataId).id,
+                    p = n.dataIdMap.get(s.dataId).id,
+                    d = n.dataIdMap.get(i.dataId).id,
+                    f = n.makeOutput(c, a.dtype),
+                    m = n.dataIdMap.get(f.dataId).id,
+                    g = n.makeOutput([4], "int32"),
+                    y = n.dataIdMap.get(g.dataId).id;
+                qV(h, lU[a.dtype], a.shape[0], p, d, m, y, t, 0);
+                const b = n.readSync(g.dataId);
+                let v;
+                switch (b[0]) {
+                    case 0:
+                        v = jd.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();
+                        break;
+                    case 1:
+                        v = jd.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();
+                        break;
+                    case 2:
+                        v = jd.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(b[1], b[2]);
+                        break;
+                    case 3:
+                        v = jd.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(b[1], b[2], b[3]);
+                        break;
+                    default:
+                        v = ""
+                }
+                if (n.disposeData(g.dataId), v) throw n.disposeData(f.dataId), new Error(v);
+                return f
+            }
+            var QV = {
+                kernelName: Dr,
+                backendName: "wasm",
+                setupFunc: YV,
+                kernelFunc: function (e) {
+                    return XV(e, !0)
+                }
+            };
+            var ZV = {
+                kernelName: Or,
+                backendName: "wasm",
+                setupFunc: YV,
+                kernelFunc: function (e) {
+                    return XV(e, !1)
+                }
+            };
+            var JV, ej = {
+                    kernelName: Rr,
+                    backendName: "wasm",
+                    kernelFunc: function (e) {
+                        const {
+                            inputs: t,
+                            attrs: n,
+                            backend: r
+                        } = e, {
+                            x: a
+                        } = t, {
+                            numOrSizeSplits: s,
+                            axis: i
+                        } = n, o = va.parseAxisParam(i, a.shape)[0], u = jd.prepareSplitSize(a, s, o), l = new Array(a.shape.length).fill(0), c = a.shape.slice();
+                        return u.map(e => {
+                            const t = [...c];
+                            t[o] = e;
+                            const n = zU({
+                                inputs: {
+                                    x: a
+                                },
+                                attrs: {
+                                    begin: l,
+                                    size: t
+                                },
+                                backend: r
+                            });
+                            return l[o] += e, n
+                        })
+                    }
+                },
+                tj = gU(Nr),
+                nj = gU($r),
+                rj = bU(Pr);
+            var aj, sj = {
+                kernelName: ea,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    JV = e.wasm.cwrap(ea, null, ["number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        alpha: a
+                    } = r, {
+                        x: s
+                    } = n, i = t.dataIdMap.get(s.dataId).id, o = t.makeOutput(s.shape, s.dtype), u = t.dataIdMap.get(o.dataId).id;
+                    return JV(i, a, lU[s.dtype], u), o
+                }
+            };
+            var ij, oj = {
+                    kernelName: Br,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        aj = e.wasm.cwrap(Br, null, ["number", "array", "number", "array", "array", "array", "array", "array", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            x: a
+                        } = n, {
+                            begin: s,
+                            end: i,
+                            strides: o,
+                            beginMask: u,
+                            endMask: l,
+                            ellipsisMask: c,
+                            newAxisMask: h,
+                            shrinkAxisMask: p
+                        } = r, {
+                            finalShapeSparse: d,
+                            finalShape: f,
+                            isIdentity: m,
+                            sliceDim0: g,
+                            isSimpleSlice: y,
+                            begin: b,
+                            end: v,
+                            strides: x
+                        } = fo.sliceInfo(a.shape, s, i, o, u, l, c, h, p);
+                        let w;
+                        if (m) w = LU({
+                            inputs: {
+                                x: a
+                            },
+                            backend: t,
+                            attrs: {
+                                shape: f
+                            }
+                        });
+                        else if (g || y) {
+                            va.assert(a.shape.length >= 1, () => "Input must have rank at least 1, got: " + a.shape.length);
+                            const e = fo.computeOutShape(b, v, x),
+                                n = zU({
+                                    inputs: {
+                                        x: a
+                                    },
+                                    backend: t,
+                                    attrs: {
+                                        begin: b,
+                                        size: e
+                                    }
+                                });
+                            w = LU({
+                                inputs: {
+                                    x: n
+                                },
+                                backend: t,
+                                attrs: {
+                                    shape: f
+                                }
+                            }), t.disposeData(n.dataId)
+                        } else {
+                            const e = t.makeOutput(d, "float32"),
+                                n = t.dataIdMap.get(a.dataId).id,
+                                r = new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer),
+                                s = new Uint8Array(new Int32Array(b).buffer),
+                                i = new Uint8Array(new Int32Array(v).buffer),
+                                o = new Uint8Array(new Int32Array(x).buffer),
+                                u = new Uint8Array(new Int32Array(d).buffer),
+                                l = new Uint8Array(new Int32Array(va.computeStrides(d)).buffer),
+                                c = t.dataIdMap.get(e.dataId).id;
+                            aj(n, r, a.shape.length, s, i, o, u, l, d.length, c), w = LU({
+                                inputs: {
+                                    x: e
+                                },
+                                backend: t,
+                                attrs: {
+                                    shape: f
+                                }
+                            }), t.disposeData(e.dataId)
+                        }
+                        return w
+                    }
+                },
+                uj = bU(Vr);
+            var lj, cj, hj = {
+                    kernelName: Cr,
+                    backendName: "wasm",
+                    setupFunc: function (e) {
+                        ij = e.wasm.cwrap(Cr, null, ["number", "number", "number", "number"])
+                    },
+                    kernelFunc: function (e) {
+                        const {
+                            backend: t,
+                            inputs: n,
+                            attrs: r
+                        } = e, {
+                            axis: a,
+                            keepDims: s
+                        } = r, {
+                            x: i
+                        } = n, o = t.dataIdMap.get(i.dataId).id;
+                        let u = o,
+                            l = i;
+                        const {
+                            transposed: c,
+                            axes: h,
+                            originalAxes: p,
+                            inputWasTransposed: d
+                        } = CU(i, a, t);
+                        let f = h;
+                        if (d) {
+                            const e = t.dataIdMap.get(c.dataId).id;
+                            e !== o && (l = c, u = e, f = jd.getInnerMostAxes(f.length, l.shape.length))
+                        }
+                        jd.assertAxesAreInnerMostDims("sum", f, l.shape.length);
+                        const [m, g] = jd.computeOutAndReduceShapes(l.shape, f), y = va.sizeFromShape(g), b = t.makeOutput(m, l.dtype);
+                        if (0 !== va.sizeFromShape(l.shape)) {
+                            const e = t.dataIdMap.get(b.dataId).id;
+                            ij(u, y, lU[b.dtype], e)
+                        }
+                        if (d && t.disposeData(c.dataId), s) {
+                            const e = jd.expandShapeToKeepDim(b.shape, p);
+                            b.shape = e
+                        }
+                        return b
+                    }
+                },
+                pj = gU(jr),
+                dj = gU(Gr);
+            var fj;
+            var mj = [mU, yU, xU, wU, RU, FU, DU, OU, BU, WU, GU, HU, qU, XU, ZU, nW, rW, aW, iW, uW, cW, hW, pW, dW, fW, gW, vW, wW, kW, SW, AW, NW, TW, _W, MW, DW, OW, IU, PW, $W, BW, zW, UW, VW, jW, HW, KW, ZW, JW, tV, nV, rV, iV, uV, cV, hV, pV, dV, mV, yV, bV, xV, kV, SV, IV, AV, EV, $U, CV, RV, FV, MV, DV, LV, $V, zV, UV, UU, WV, jV, HV, KV, QV, ZV, ej, tj, nj, rj, sj, oj, uj, hj, pj, dj, {
+                kernelName: Hr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    lj = e.wasm.cwrap(Hr, null, ["number", "array", "number", "array", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        x: a
+                    } = t, s = n.dataIdMap.get(a.dataId).id, {
+                        reps: i
+                    } = r, o = new Array(a.shape.length);
+                    for (let e = 0; e < o.length; e++) o[e] = a.shape[e] * i[e];
+                    const u = new Uint8Array(new Int32Array(a.shape).buffer),
+                        l = new Uint8Array(new Int32Array(o).buffer),
+                        c = n.makeOutput(o, a.dtype),
+                        h = n.dataIdMap.get(c.dataId).id;
+                    return lj(s, u, a.shape.length, l, o.length, lU[c.dtype], h), c
+                }
+            }, {
+                kernelName: qr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    cj = e.wasm.cwrap(qr, null, ["number", "array", "number", "number", "number", "bool", "number", "number"])
+                },
+                kernelFunc: ({
+                    inputs: e,
+                    backend: t,
+                    attrs: n
+                }) => {
+                    const {
+                        x: r
+                    } = e, {
+                        k: a,
+                        sorted: s
+                    } = n, i = t.dataIdMap.get(r.dataId).id, o = new Uint8Array(new Int32Array(r.shape).buffer), u = r.shape.slice();
+                    u[u.length - 1] = a;
+                    const l = t.makeOutput(u, r.dtype),
+                        c = t.dataIdMap.get(l.dataId).id,
+                        h = t.makeOutput(u, "int32"),
+                        p = t.dataIdMap.get(h.dataId).id;
+                    return cj(i, o, r.shape.length, lU[r.dtype], a, s, c, p), [l, h]
+                }
+            }, {
+                kernelName: Kr,
+                backendName: "wasm",
+                setupFunc: function (e) {
+                    fj = e.wasm.cwrap(Kr, null, ["number", "number", "bool", "number", "number", "number", "number", "number", "number", "array", "number", "number", "number", "number", "number"])
+                },
+                kernelFunc: function (e) {
+                    const {
+                        backend: t,
+                        inputs: n,
+                        attrs: r
+                    } = e, {
+                        image: a,
+                        transforms: s
+                    } = n, {
+                        interpolation: i,
+                        fillMode: o,
+                        fillValue: u,
+                        outputShape: l
+                    } = r, [c, h, p, d] = a.shape, [f, m] = null != l ? l : [h, p], g = [c, f, m, d], y = new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer), b = t.makeOutput(g, a.dtype), v = t.dataIdMap.get(b.dataId).id, x = t.dataIdMap.get(a.dataId).id, w = t.dataIdMap.get(s.dataId).id, k = "nearest" === i ? 1 : 2;
+                    let S;
+                    switch (o) {
+                        case "constant":
+                            S = 1;
+                            break;
+                        case "reflect":
+                            S = 2;
+                            break;
+                        case "wrap":
+                            S = 3;
+                            break;
+                        case "nearest":
+                            S = 4;
+                            break;
+                        default:
+                            S = 1
+                    }
+                    return fj(x, w, s.shape[0] > 1, c, f, m, d, p, h, y, a.shape.length - 1, k, S, u, v), b
+                }
+            }, NU, {
+                kernelName: Qr,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: t,
+                        backend: n,
+                        attrs: r
+                    } = e, {
+                        value: a
+                    } = t;
+                    let {
+                        axis: s
+                    } = r;
+                    s < 0 && (s += a.shape.length);
+                    const i = a.shape[s],
+                        o = a.shape.length,
+                        u = new Array(o - 1);
+                    let l = 0;
+                    for (let e = 0; e < o; e++) e !== s && (u[l++] = a.shape[e]);
+                    const c = new Array(i),
+                        h = new Array(o).fill(0),
+                        p = a.shape.slice();
+                    p[s] = 1;
+                    for (let e = 0; e < c.length; e++) h[s] = e, c[e] = zU({
+                        inputs: {
+                            x: a
+                        },
+                        attrs: {
+                            begin: h,
+                            size: p
+                        },
+                        backend: n
+                    });
+                    return c.map(({
+                        dataId: e,
+                        dtype: t
+                    }) => ({
+                        dataId: e,
+                        dtype: t,
+                        shape: u
+                    }))
+                }
+            }, {
+                kernelName: Jr,
+                backendName: "wasm",
+                kernelFunc: function (e) {
+                    const {
+                        inputs: {
+                            x: t
+                        },
+                        backend: n
+                    } = e, r = n.makeOutput(t.shape, t.dtype);
+                    return n.typedArrayFromHeap(r).fill(0), r
+                }
+            }];
+            for (const e of mj) da(e);
+            var gj = Ye();
+            gj.registerFlag("WASM_HAS_SIMD_SUPPORT", async () => WebAssembly.validate(new Uint8Array([0, 97, 115, 109, 1, 0, 0, 0, 1, 4, 1, 96, 0, 0, 3, 2, 1, 0, 10, 9, 1, 7, 0, 65, 0, 253, 15, 26, 11]))), gj.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT", async () => {
+                if (gj.get("IS_NODE")) return !1;
+                try {
+                    return (new MessageChannel).port1.postMessage(new SharedArrayBuffer(1)), WebAssembly.validate(new Uint8Array([0, 97, 115, 109, 1, 0, 0, 0, 1, 4, 1, 96, 0, 0, 3, 2, 1, 0, 5, 4, 1, 3, 1, 1, 10, 11, 1, 9, 0, 65, 0, 254, 16, 2, 0, 26, 11]))
+                } catch (e) {
+                    return !1
+                }
+            });
+            var yj = k(Y()),
+                bj = k(X()),
+                vj = class extends Z {
+                    constructor(e) {
+                        super(), this.wasm = e, this.dataIdNextNumber = 1, this.wasm.tfjs.initWithThreadsCount(Rj), _j = this.wasm.tfjs.getThreadsCount(), this.dataIdMap = new Q(this, eu())
+                    }
+                    write(e, t, n) {
+                        const r = {
+                            id: this.dataIdNextNumber++
+                        };
+                        return this.move(r, e, t, n, 1), r
+                    }
+                    numDataIds() {
+                        return this.dataIdMap.numDataIds()
+                    }
+                    async time(e) {
+                        const t = va.now();
+                        e();
+                        return {
+                            kernelMs: va.now() - t
+                        }
+                    }
+                    move(e, t, n, r, a) {
+                        const s = this.dataIdNextNumber++;
+                        if ("string" === r) {
+                            const i = t;
+                            return void this.dataIdMap.set(e, {
+                                id: s,
+                                stringBytes: i,
+                                shape: n,
+                                dtype: r,
+                                memoryOffset: null,
+                                refCount: a
+                            })
+                        }
+                        const i = va.sizeFromShape(n),
+                            o = i * va.bytesPerElement(r),
+                            u = this.wasm._malloc(o);
+                        this.dataIdMap.set(e, {
+                            id: s,
+                            memoryOffset: u,
+                            shape: n,
+                            dtype: r,
+                            refCount: a
+                        }), this.wasm.tfjs.registerTensor(s, i, u), null != t && this.wasm.HEAPU8.set(new Uint8Array(t.buffer, t.byteOffset, o), u)
+                    }
+                    async read(e) {
+                        return this.readSync(e)
+                    }
+                    readSync(e, t, n) {
+                        const {
+                            memoryOffset: r,
+                            dtype: a,
+                            shape: s,
+                            stringBytes: i
+                        } = this.dataIdMap.get(e);
+                        if ("string" === a) return null != t && 0 !== t || !(null == n || n >= i.length) ? i.slice(t, n) : i;
+                        t = t || 0, n = n || va.sizeFromShape(s);
+                        const o = va.bytesPerElement(a);
+                        return function (e, t) {
+                            switch (t) {
+                                case "float32":
+                                    return new Float32Array(e);
+                                case "int32":
+                                    return new Int32Array(e);
+                                case "bool":
+                                    return new Uint8Array(e);
+                                default:
+                                    throw new Error("Unknown dtype " + t)
+                            }
+                        }(this.wasm.HEAPU8.slice(r + t * o, r + n * o).buffer, a)
+                    }
+                    disposeData(e, t = !1) {
+                        if (this.dataIdMap.has(e)) {
+                            const n = this.dataIdMap.get(e);
+                            if (n.refCount--, !t && n.refCount > 0) return !1;
+                            this.wasm._free(n.memoryOffset), this.wasm.tfjs.disposeData(n.id), this.dataIdMap.delete(e)
+                        }
+                        return !0
+                    }
+                    refCount(e) {
+                        if (this.dataIdMap.has(e)) {
+                            return this.dataIdMap.get(e).refCount
+                        }
+                        return 0
+                    }
+                    incRef(e) {
+                        const t = this.dataIdMap.get(e);
+                        null != t && t.refCount++
+                    }
+                    floatPrecision() {
+                        return 32
+                    }
+                    getMemoryOffset(e) {
+                        return this.dataIdMap.get(e).memoryOffset
+                    }
+                    dispose() {
+                        this.wasm.tfjs.dispose(), "PThread" in this.wasm && this.wasm.PThread.terminateAllThreads(), this.wasm = null
+                    }
+                    memory() {
+                        return {
+                            unreliable: !1
+                        }
+                    }
+                    makeOutput(e, t, n) {
+                        let r;
+                        if (null == n) r = this.write(null, e, t);
+                        else {
+                            const a = this.dataIdNextNumber++;
+                            r = {
+                                id: a
+                            }, this.dataIdMap.set(r, {
+                                id: a,
+                                memoryOffset: n,
+                                shape: e,
+                                dtype: t,
+                                refCount: 1
+                            });
+                            const s = va.sizeFromShape(e);
+                            this.wasm.tfjs.registerTensor(a, s, n)
+                        }
+                        return {
+                            dataId: r,
+                            shape: e,
+                            dtype: t
+                        }
+                    }
+                    typedArrayFromHeap({
+                        shape: e,
+                        dtype: t,
+                        dataId: n
+                    }) {
+                        const r = this.wasm.HEAPU8.buffer,
+                            {
+                                memoryOffset: a
+                            } = this.dataIdMap.get(n),
+                            s = va.sizeFromShape(e);
+                        switch (t) {
+                            case "float32":
+                                return new Float32Array(r, a, s);
+                            case "int32":
+                                return new Int32Array(r, a, s);
+                            case "bool":
+                                return new Uint8Array(r, a, s);
+                            default:
+                                throw new Error("Unknown dtype " + t)
+                        }
+                    }
+                };
+
+            function xj(e, t, n) {
+                if (null != Sj) return Sj;
+                let r = "tfjs-backend-wasm.wasm";
+                return e && t ? r = "tfjs-backend-wasm-threaded-simd.wasm" : e && (r = "tfjs-backend-wasm-simd.wasm"), null != Aj && null != Aj[r] ? Aj[r] : n + r
+            }
+            async function wj() {
+                const [e, t] = await Promise.all([Ye().getAsync("WASM_HAS_SIMD_SUPPORT"), Ye().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);
+                return new Promise((n, r) => {
+                    const a = {};
+                    var s;
+                    a.locateFile = (n, r) => {
+                        if (n.endsWith(".worker.js")) {
+                            const e = new Blob(['var Module={};function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;this.alert=threadAlert;Module["instantiateWasm"]=function(info,receiveInstance){var instance=new WebAssembly.Instance(Module["wasmModule"],info);Module["wasmModule"]=null;receiveInstance(instance);return instance.exports};function moduleLoaded(){}this.onmessage=function(e){try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob==="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance;moduleLoaded()})}else if(e.data.cmd==="objectTransfer"){Module["PThread"].receiveObjectTransfer(e.data)}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.threadInfoStruct,0,0);var max=e.data.stackBase;var top=e.data.stackBase+e.data.stackSize;Module["establishStackSpace"](top,max);Module["_emscripten_tls_init"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].setThreadStatus(Module["_pthread_self"](),1);try{var result=Module["invokeEntryPoint"](e.data.start_routine,e.data.arg);if(!Module["getNoExitRuntime"]())Module["PThread"].threadExit(result)}catch(ex){if(ex==="Canceled!"){Module["PThread"].threadCancel()}else if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["getNoExitRuntime"]()){}else{Module["PThread"].threadExit(ex.status)}}else{Module["PThread"].threadExit(-2);throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["PThread"].threadCancel()}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processThreadQueue"){if(Module["_pthread_self"]()){Module["_emscripten_current_thread_process_queued_calls"]()}}else{err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){err("worker.js onmessage() captured an uncaught exception: "+ex);if(ex&&ex.stack)err(ex.stack);throw ex}};if(typeof process==="object"&&typeof process.versions==="object"&&typeof process.versions.node==="string"){self={location:{href:__filename}};var onmessage=this.onmessage;var nodeWorkerThreads=require("worker_threads");global.Worker=nodeWorkerThreads.Worker;var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",function(data){onmessage({data:data})});var nodeFS=require("fs");var nodeRead=function(filename){return nodeFS.readFileSync(filename,"utf8")};function globalEval(x){global.require=require;global.Module=Module;eval.call(null,x)}importScripts=function(f){globalEval(nodeRead(f))};postMessage=function(msg){parentPort.postMessage(msg)};if(typeof performance==="undefined"){performance={now:function(){return Date.now()}}}}'], {
+                                type: "application/javascript"
+                            });
+                            return URL.createObjectURL(e)
+                        }
+                        return n.endsWith(".wasm") ? xj(e, t, null != Ij ? Ij : r) : r + n
+                    }, Nj && (a.instantiateWasm = (s = xj(e, t, null != Ij ? Ij : ""), (e, t) => (va.fetch(s, {
+                        credentials: "same-origin"
+                    }).then(n => {
+                        n.ok || e.env.a(`failed to load wasm binary file at '${s}'`), n.arrayBuffer().then(n => {
+                            WebAssembly.instantiate(n, e).then(e => {
+                                t(e.instance, e.module)
+                            })
+                        })
+                    }), {})));
+                    let i, o = !1;
+                    a.onAbort = () => {
+                        if (o) return;
+                        if (Ej) return;
+                        Ej = !0;
+                        r({
+                            message: "Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"
+                        })
+                    }, t && e && null == Sj ? (a.mainScriptUrlOrBlob = new Blob(["var WasmBackendModuleThreadedSimd = " + yj.default.toString()], {
+                        type: "text/javascript"
+                    }), i = (0, yj.default)(a)) : i = (0, bj.default)(a), i.then(e => {
+                        o = !0, Ej = !1;
+                        e.tfjs = {
+                            init: e.cwrap("init", null, []),
+                            initWithThreadsCount: e.cwrap("init_with_threads_count", null, ["number"]),
+                            getThreadsCount: e.cwrap("get_threads_count", "number", []),
+                            registerTensor: e.cwrap("register_tensor", null, ["number", "number", "number"]),
+                            disposeData: e.cwrap("dispose_data", null, ["number"]),
+                            dispose: e.cwrap("dispose", null, [])
+                        }, n({
+                            wasm: e
+                        })
+                    })
+                })
+            }
+            var kj = ["tfjs-backend-wasm.wasm", "tfjs-backend-wasm-simd.wasm", "tfjs-backend-wasm-threaded-simd.wasm"],
+                Sj = null,
+                Ij = null,
+                Aj = {},
+                Ej = !1,
+                Nj = !1;
+
+            function Cj(e, t = !1) {
+                if (Zo("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."), Ej) throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");
+                Sj = e, Nj = t
+            }
+
+            function Tj(e, t = !1) {
+                if (Ej) throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");
+                if ("string" == typeof e) Ij = e;
+                else {
+                    Aj = e;
+                    const t = kj.filter(e => null == Aj[e]);
+                    if (t.length > 0) throw new Error(`There were no entries found for the following binaries: ${t.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)
+                }
+                Nj = t
+            }
+            var Rj = -1,
+                _j = -1;
+
+            function Fj(e) {
+                Rj = e
+            }
+
+            function Mj() {
+                if (-1 === _j) throw new Error("WASM backend not initialized.");
+                return _j
+            }
+            var Dj = "3.13.0";
+            du("wasm", async () => {
+                const {
+                    wasm: e
+                } = await wj();
+                return new vj(e)
+            }, 2);
+            var Oj = {
+                tfjs: "3.13.0",
+                "tfjs-core": "3.13.0",
+                "tfjs-data": "3.13.0",
+                "tfjs-layers": "3.13.0",
+                "tfjs-converter": "3.13.0",
+                "tfjs-backend-cpu": "3.13.0",
+                "tfjs-backend-webgl": "3.13.0",
+                "tfjs-backend-wasm": "3.13.0"
+            };
+
+            function Lj(e, t, n = !1) {
+                if (e.beginPath(), t.slice(1).forEach(({
+                        x: n,
+                        y: r
+                    }, a) => {
+                        const s = t[a];
+                        e.moveTo(s.x, s.y), e.lineTo(n, r)
+                    }), n) {
+                    const n = t[t.length - 1],
+                        r = t[0];
+                    if (!n || !r) return;
+                    e.moveTo(n.x, n.y), e.lineTo(r.x, r.y)
+                }
+                e.stroke()
+            }
+            p({}, {
+                AnchorPosition: () => kG,
+                DrawBox: () => EG,
+                DrawBoxOptions: () => AG,
+                DrawFaceLandmarks: () => hH,
+                DrawFaceLandmarksOptions: () => cH,
+                DrawTextField: () => IG,
+                DrawTextFieldOptions: () => SG,
+                drawContour: () => Lj,
+                drawDetections: () => NG,
+                drawFaceExpressions: () => oH,
+                drawFaceLandmarks: () => pH
+            });
+            p({}, {
+                computeReshapedDimensions: () => qj,
+                getCenterPoint: () => Kj,
+                isDimensions: () => Hj,
+                isEven: () => jj,
+                isFloat: () => Vj,
+                isTensor: () => $j,
+                isTensor1D: () => Bj,
+                isTensor2D: () => zj,
+                isTensor3D: () => Uj,
+                isTensor4D: () => Wj,
+                isValidNumber: () => Xj,
+                isValidProbablitiy: () => Qj,
+                range: () => Yj,
+                round: () => Gj
+            });
+            var Pj = class {
+                constructor(e, t) {
+                    if (!Xj(e) || !Xj(t)) throw new Error("Dimensions.constructor - expected width and height to be valid numbers, instead have " + JSON.stringify({
+                        width: e,
+                        height: t
+                    }));
+                    this._width = e, this._height = t
+                }
+                get width() {
+                    return this._width
+                }
+                get height() {
+                    return this._height
+                }
+                reverse() {
+                    return new Pj(1 / this.width, 1 / this.height)
+                }
+            };
+
+            function $j(e, t) {
+                return e instanceof Ya && e.shape.length === t
+            }
+
+            function Bj(e) {
+                return $j(e, 1)
+            }
+
+            function zj(e) {
+                return $j(e, 2)
+            }
+
+            function Uj(e) {
+                return $j(e, 3)
+            }
+
+            function Wj(e) {
+                return $j(e, 4)
+            }
+
+            function Vj(e) {
+                return e % 1 != 0
+            }
+
+            function jj(e) {
+                return e % 2 == 0
+            }
+
+            function Gj(e, t = 2) {
+                const n = 10 ** t;
+                return Math.floor(e * n) / n
+            }
+
+            function Hj(e) {
+                return e && e.width && e.height
+            }
+
+            function qj({
+                width: e,
+                height: t
+            }, n) {
+                const r = n / Math.max(t, e);
+                return new Pj(Math.round(e * r), Math.round(t * r))
+            }
+
+            function Kj(e) {
+                return e.reduce((e, t) => e.add(t), new Zj(0, 0)).div(new Zj(e.length, e.length))
+            }
+
+            function Yj(e, t, n) {
+                return Array(e).fill(0).map((e, r) => t + r * n)
+            }
+
+            function Xj(e) {
+                return !!e && e !== 1 / 0 && e !== -1 / 0 && !Number.isNaN(e) || 0 === e
+            }
+
+            function Qj(e) {
+                return Xj(e) && e >= 0 && e <= 1
+            }
+            var Zj = class {
+                    constructor(e, t) {
+                        this._x = e, this._y = t
+                    }
+                    get x() {
+                        return this._x
+                    }
+                    get y() {
+                        return this._y
+                    }
+                    add(e) {
+                        return new Zj(this.x + e.x, this.y + e.y)
+                    }
+                    sub(e) {
+                        return new Zj(this.x - e.x, this.y - e.y)
+                    }
+                    mul(e) {
+                        return new Zj(this.x * e.x, this.y * e.y)
+                    }
+                    div(e) {
+                        return new Zj(this.x / e.x, this.y / e.y)
+                    }
+                    abs() {
+                        return new Zj(Math.abs(this.x), Math.abs(this.y))
+                    }
+                    magnitude() {
+                        return Math.sqrt(this.x ** 2 + this.y ** 2)
+                    }
+                    floor() {
+                        return new Zj(Math.floor(this.x), Math.floor(this.y))
+                    }
+                },
+                Jj = class {
+                    static isRect(e) {
+                        return !!e && [e.x, e.y, e.width, e.height].every(Xj)
+                    }
+                    static assertIsValidBox(e, t, n = !1) {
+                        if (!Jj.isRect(e)) throw new Error(`${t} - invalid box: ${JSON.stringify(e)}, expected object with properties x, y, width, height`);
+                        if (!n && (e.width < 0 || e.height < 0)) throw new Error(`${t} - width (${e.width}) and height (${e.height}) must be positive numbers`)
+                    }
+                    constructor(e, t = !0) {
+                        const n = e || {},
+                            r = [n.left, n.top, n.right, n.bottom].every(Xj),
+                            a = [n.x, n.y, n.width, n.height].every(Xj);
+                        if (!a && !r) throw new Error("Box.constructor - expected box to be IBoundingBox | IRect, instead have " + JSON.stringify(n));
+                        const [s, i, o, u] = a ? [n.x, n.y, n.width, n.height] : [n.left, n.top, n.right - n.left, n.bottom - n.top];
+                        Jj.assertIsValidBox({
+                            x: s,
+                            y: i,
+                            width: o,
+                            height: u
+                        }, "Box.constructor", t), this._x = s, this._y = i, this._width = o, this._height = u
+                    }
+                    get x() {
+                        return this._x
+                    }
+                    get y() {
+                        return this._y
+                    }
+                    get width() {
+                        return this._width
+                    }
+                    get height() {
+                        return this._height
+                    }
+                    get left() {
+                        return this.x
+                    }
+                    get top() {
+                        return this.y
+                    }
+                    get right() {
+                        return this.x + this.width
+                    }
+                    get bottom() {
+                        return this.y + this.height
+                    }
+                    get area() {
+                        return this.width * this.height
+                    }
+                    get topLeft() {
+                        return new Zj(this.left, this.top)
+                    }
+                    get topRight() {
+                        return new Zj(this.right, this.top)
+                    }
+                    get bottomLeft() {
+                        return new Zj(this.left, this.bottom)
+                    }
+                    get bottomRight() {
+                        return new Zj(this.right, this.bottom)
+                    }
+                    round() {
+                        const [e, t, n, r] = [this.x, this.y, this.width, this.height].map(e => Math.round(e));
+                        return new Jj({
+                            x: e,
+                            y: t,
+                            width: n,
+                            height: r
+                        })
+                    }
+                    floor() {
+                        const [e, t, n, r] = [this.x, this.y, this.width, this.height].map(e => Math.floor(e));
+                        return new Jj({
+                            x: e,
+                            y: t,
+                            width: n,
+                            height: r
+                        })
+                    }
+                    toSquare() {
+                        let {
+                            x: e,
+                            y: t,
+                            width: n,
+                            height: r
+                        } = this;
+                        const a = Math.abs(n - r);
+                        return n < r && (e -= a / 2, n += a), r < n && (t -= a / 2, r += a), new Jj({
+                            x: e,
+                            y: t,
+                            width: n,
+                            height: r
+                        })
+                    }
+                    rescale(e) {
+                        const t = Hj(e) ? e.width : e,
+                            n = Hj(e) ? e.height : e;
+                        return new Jj({
+                            x: this.x * t,
+                            y: this.y * n,
+                            width: this.width * t,
+                            height: this.height * n
+                        })
+                    }
+                    pad(e, t) {
+                        const [n, r, a, s] = [this.x - e / 2, this.y - t / 2, this.width + e, this.height + t];
+                        return new Jj({
+                            x: n,
+                            y: r,
+                            width: a,
+                            height: s
+                        })
+                    }
+                    clipAtImageBorders(e, t) {
+                        const {
+                            x: n,
+                            y: r,
+                            right: a,
+                            bottom: s
+                        } = this, i = Math.max(n, 0), o = Math.max(r, 0), u = a - i, l = s - o, c = Math.min(u, e - i), h = Math.min(l, t - o);
+                        return new Jj({
+                            x: i,
+                            y: o,
+                            width: c,
+                            height: h
+                        }).floor()
+                    }
+                    shift(e, t) {
+                        const {
+                            width: n,
+                            height: r
+                        } = this, a = this.x + e, s = this.y + t;
+                        return new Jj({
+                            x: a,
+                            y: s,
+                            width: n,
+                            height: r
+                        })
+                    }
+                    padAtBorders(e, t) {
+                        const n = this.width + 1,
+                            r = this.height + 1;
+                        let a = n,
+                            s = r,
+                            i = this.left,
+                            o = this.top,
+                            u = this.right,
+                            l = this.bottom;
+                        return u > t && (a = -u + t + n, u = t), l > e && (s = -l + e + r, l = e), i < 1 && (s = 2 - i, i = 1), o < 1 && (s = 2 - o, o = 1), {
+                            dy: 1,
+                            edy: s,
+                            dx: 1,
+                            edx: a,
+                            y: o,
+                            ey: l,
+                            x: i,
+                            ex: u,
+                            w: n,
+                            h: r
+                        }
+                    }
+                    calibrate(e) {
+                        return new Jj({
+                            left: this.left + e.left * this.width,
+                            top: this.top + e.top * this.height,
+                            right: this.right + e.right * this.width,
+                            bottom: this.bottom + e.bottom * this.height
+                        }).toSquare().round()
+                    }
+                },
+                eG = class extends Jj {
+                    constructor(e, t, n, r, a = !1) {
+                        super({
+                            left: e,
+                            top: t,
+                            right: n,
+                            bottom: r
+                        }, a)
+                    }
+                },
+                tG = class {
+                    constructor(e, t, n, r, a) {
+                        this._imageDims = new Pj(a.width, a.height), this._score = e, this._classScore = t, this._className = n, this._box = new Jj(r).rescale(this._imageDims)
+                    }
+                    get score() {
+                        return this._score
+                    }
+                    get classScore() {
+                        return this._classScore
+                    }
+                    get className() {
+                        return this._className
+                    }
+                    get box() {
+                        return this._box
+                    }
+                    get imageDims() {
+                        return this._imageDims
+                    }
+                    get imageWidth() {
+                        return this.imageDims.width
+                    }
+                    get imageHeight() {
+                        return this.imageDims.height
+                    }
+                    get relativeBox() {
+                        return new Jj(this._box).rescale(this.imageDims.reverse())
+                    }
+                    forSize(e, t) {
+                        return new tG(this.score, this.classScore, this.className, this.relativeBox, {
+                            width: e,
+                            height: t
+                        })
+                    }
+                },
+                nG = class extends tG {
+                    constructor(e, t, n) {
+                        super(e, e, "", t, n)
+                    }
+                    forSize(e, t) {
+                        const {
+                            score: n,
+                            relativeBox: r,
+                            imageDims: a
+                        } = super.forSize(e, t);
+                        return new nG(n, r, a)
+                    }
+                };
+
+            function rG(e, t, n = !0) {
+                const r = Math.max(0, Math.min(e.right, t.right) - Math.max(e.left, t.left)) * Math.max(0, Math.min(e.bottom, t.bottom) - Math.max(e.top, t.top));
+                return n ? r / (e.area + t.area - r) : r / Math.min(e.area, t.area)
+            }
+
+            function aG(e, t) {
+                return ru(() => {
+                    const [n, r, a] = t, s = Vl([...e.shape.slice(0, 3), 1], n, "float32"), i = Vl([...e.shape.slice(0, 3), 1], r, "float32"), o = Vl([...e.shape.slice(0, 3), 1], a, "float32"), u = Xu([s, i, o], 3);
+                    return gc(e, u)
+                })
+            }
+
+            function sG(e) {
+                return 1 / (1 + Math.exp(-e))
+            }
+            var iG, oG = class extends Jj {
+                    constructor(e, t, n, r, a = !1) {
+                        super({
+                            x: e,
+                            y: t,
+                            width: n,
+                            height: r
+                        }, a)
+                    }
+                },
+                uG = class {
+                    constructor(e, t, n = new Zj(0, 0)) {
+                        const {
+                            width: r,
+                            height: a
+                        } = t;
+                        this._imgDims = new Pj(r, a), this._shift = n, this._positions = e.map(e => e.mul(new Zj(r, a)).add(n))
+                    }
+                    get shift() {
+                        return new Zj(this._shift.x, this._shift.y)
+                    }
+                    get imageWidth() {
+                        return this._imgDims.width
+                    }
+                    get imageHeight() {
+                        return this._imgDims.height
+                    }
+                    get positions() {
+                        return this._positions
+                    }
+                    get relativePositions() {
+                        return this._positions.map(e => e.sub(this._shift).div(new Zj(this.imageWidth, this.imageHeight)))
+                    }
+                    forSize(e, t) {
+                        return new this.constructor(this.relativePositions, {
+                            width: e,
+                            height: t
+                        })
+                    }
+                    shiftBy(e, t) {
+                        return new this.constructor(this.relativePositions, this._imgDims, new Zj(e, t))
+                    }
+                    shiftByPoint(e) {
+                        return this.shiftBy(e.x, e.y)
+                    }
+                    align(e, t = {}) {
+                        if (e) {
+                            const n = e instanceof nG ? e.box.floor() : new Jj(e);
+                            return this.shiftBy(n.x, n.y).align(null, t)
+                        }
+                        const {
+                            useDlibAlignment: n,
+                            minBoxPadding: r
+                        } = {
+                            useDlibAlignment: !1,
+                            minBoxPadding: .2,
+                            ...t
+                        };
+                        return n ? this.alignDlib() : this.alignMinBbox(r)
+                    }
+                    alignDlib() {
+                        const e = this.getRefPointsForAlignment(),
+                            [t, n, r] = e,
+                            a = e => r.sub(e).magnitude(),
+                            s = (a(t) + a(n)) / 2,
+                            i = Math.floor(s / .45),
+                            o = Kj(e),
+                            u = Math.floor(Math.max(0, o.x - .5 * i)),
+                            l = Math.floor(Math.max(0, o.y - .43 * i));
+                        return new oG(u, l, Math.min(i, this.imageWidth + u), Math.min(i, this.imageHeight + l))
+                    }
+                    alignMinBbox(e) {
+                        const t = function (e) {
+                            const t = e.map(e => e.x),
+                                n = e.map(e => e.y),
+                                r = t.reduce((e, t) => t < e ? t : e, 1 / 0),
+                                a = n.reduce((e, t) => t < e ? t : e, 1 / 0),
+                                s = t.reduce((e, t) => e < t ? t : e, 0),
+                                i = n.reduce((e, t) => e < t ? t : e, 0);
+                            return new eG(r, a, s, i)
+                        }(this.positions);
+                        return t.pad(t.width * e, t.height * e)
+                    }
+                    getRefPointsForAlignment() {
+                        throw new Error("getRefPointsForAlignment not implemented by base class")
+                    }
+                },
+                lG = class extends uG {
+                    getJawOutline() {
+                        return this.positions.slice(0, 17)
+                    }
+                    getLeftEyeBrow() {
+                        return this.positions.slice(17, 22)
+                    }
+                    getRightEyeBrow() {
+                        return this.positions.slice(22, 27)
+                    }
+                    getNose() {
+                        return this.positions.slice(27, 36)
+                    }
+                    getLeftEye() {
+                        return this.positions.slice(36, 42)
+                    }
+                    getRightEye() {
+                        return this.positions.slice(42, 48)
+                    }
+                    getMouth() {
+                        return this.positions.slice(48, 68)
+                    }
+                    getRefPointsForAlignment() {
+                        return [this.getLeftEye(), this.getRightEye(), this.getMouth()].map(Kj)
+                    }
+                };
+
+            function cG(e) {
+                return e.detection instanceof nG
+            }
+
+            function hG(e, t) {
+                return {
+                    ...e,
+                    ...{
+                        detection: t
+                    }
+                }
+            }
+
+            function pG() {
+                const e = window.fetch;
+                if (!e) throw new Error("fetch - missing fetch implementation for browser environment");
+                return {
+                    Canvas: HTMLCanvasElement,
+                    CanvasRenderingContext2D: CanvasRenderingContext2D,
+                    Image: HTMLImageElement,
+                    ImageData: ImageData,
+                    Video: HTMLVideoElement,
+                    createCanvasElement: () => document.createElement("canvas"),
+                    createImageElement: () => document.createElement("img"),
+                    createVideoElement: () => document.createElement("video"),
+                    fetch: e,
+                    readFile: () => {
+                        throw new Error("readFile - filesystem not available for browser environment")
+                    }
+                }
+            }
+
+            function dG() {
+                return "object" == typeof s && void 0 !== r && null != r.versions && null != r.versions.node
+            }
+
+            function fG(e) {
+                let t = "";
+                if (!e && dG()) try {
+                    e = h("fs")
+                } catch (e) {
+                    t = e.toString()
+                }
+                return {
+                    readFile: e ? t => new Promise((n, r) => {
+                        e.readFile(t, (e, t) => e ? r(e) : n(t))
+                    }) : () => {
+                        throw new Error("readFile - failed to require fs in nodejs environment with error: " + t)
+                    }
+                }
+            }
+
+            function mG() {
+                const e = s.Canvas || s.HTMLCanvasElement,
+                    t = s.Image || s.HTMLImageElement,
+                    n = s.Video || s.HTMLVideoElement,
+                    r = s.fetch,
+                    a = fG();
+                return {
+                    Canvas: e || class {},
+                    CanvasRenderingContext2D: s.CanvasRenderingContext2D || class {},
+                    Image: t || class {},
+                    ImageData: s.ImageData || class {},
+                    Video: s.HTMLVideoElement || class {},
+                    createCanvasElement: () => {
+                        if (e) return new e;
+                        throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")
+                    },
+                    createImageElement: () => {
+                        if (t) return new t;
+                        throw new Error("createImageElement - missing Image implementation for nodejs environment")
+                    },
+                    createVideoElement: () => {
+                        if (n) return new n;
+                        throw new Error("createVideoElement - missing Video implementation for nodejs environment")
+                    },
+                    fetch: r,
+                    ...a
+                }
+            }
+
+            function gG() {
+                return "object" == typeof window && "undefined" != typeof document && "undefined" != typeof HTMLImageElement && "undefined" != typeof HTMLCanvasElement && "undefined" != typeof HTMLVideoElement && "undefined" != typeof ImageData && "undefined" != typeof CanvasRenderingContext2D
+            }
+
+            function yG(e) {
+                iG = e
+            }
+
+            function bG() {
+                return gG() ? yG(pG()) : dG() ? yG(mG()) : null
+            }
+            var vG = {
+                getEnv: function () {
+                    if (!iG) throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");
+                    return iG
+                },
+                setEnv: yG,
+                initialize: bG,
+                createBrowserEnv: pG,
+                createFileSystem: fG,
+                createNodejsEnv: mG,
+                monkeyPatch: function (e) {
+                    if (iG || bG(), !iG) throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");
+                    const {
+                        Canvas: t = iG.Canvas,
+                        Image: n = iG.Image
+                    } = e;
+                    iG.Canvas = t, iG.Image = n, iG.createCanvasElement = e.createCanvasElement || (() => new t), iG.createImageElement = e.createImageElement || (() => new n), iG.ImageData = e.ImageData || iG.ImageData, iG.Video = e.Video || iG.Video, iG.fetch = e.fetch || iG.fetch, iG.readFile = e.readFile || iG.readFile
+                },
+                isBrowser: gG,
+                isNodejs: dG
+            };
+
+            function xG(e) {
+                return vG.isNodejs() || "string" != typeof e ? e : document.getElementById(e)
+            }
+
+            function wG(e) {
+                const {
+                    Canvas: t,
+                    CanvasRenderingContext2D: n
+                } = vG.getEnv();
+                if (e instanceof n) return e;
+                const r = xG(e);
+                if (!(r instanceof t)) throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");
+                const a = r.getContext("2d");
+                if (!a) throw new Error("resolveContext2d - canvas 2d context is null");
+                return a
+            }
+            bG();
+            var kG = (e => (e.TOP_LEFT = "TOP_LEFT", e.TOP_RIGHT = "TOP_RIGHT", e.BOTTOM_LEFT = "BOTTOM_LEFT", e.BOTTOM_RIGHT = "BOTTOM_RIGHT", e))(kG || {}),
+                SG = class {
+                    constructor(e = {}) {
+                        const {
+                            anchorPosition: t,
+                            backgroundColor: n,
+                            fontColor: r,
+                            fontSize: a,
+                            fontStyle: s,
+                            padding: i
+                        } = e;
+                        this.anchorPosition = t || "TOP_LEFT", this.backgroundColor = n || "rgba(0, 0, 0, 0.5)", this.fontColor = r || "rgba(255, 255, 255, 1)", this.fontSize = a || 14, this.fontStyle = s || "Georgia", this.padding = i || 4
+                    }
+                },
+                IG = class {
+                    constructor(e, t, n = {}) {
+                        this.text = "string" == typeof e ? [e] : e instanceof IG ? e.text : e, this.anchor = t, this.options = new SG(n)
+                    }
+                    measureWidth(e) {
+                        const {
+                            padding: t
+                        } = this.options;
+                        return this.text.map(t => e.measureText(t).width).reduce((e, t) => e < t ? t : e, 0) + 2 * t
+                    }
+                    measureHeight() {
+                        const {
+                            fontSize: e,
+                            padding: t
+                        } = this.options;
+                        return this.text.length * e + 2 * t
+                    }
+                    getUpperLeft(e, t) {
+                        const {
+                            anchorPosition: n
+                        } = this.options, r = "BOTTOM_RIGHT" === n || "TOP_RIGHT" === n, a = "BOTTOM_LEFT" === n || "BOTTOM_RIGHT" === n, s = this.measureWidth(e), i = this.measureHeight(), o = r ? this.anchor.x - s : this.anchor.x, u = a ? this.anchor.y - i : this.anchor.y;
+                        if (t) {
+                            const {
+                                width: e,
+                                height: n
+                            } = t;
+                            return {
+                                x: Math.max(Math.min(o, e - s), 0),
+                                y: Math.max(Math.min(u, n - i), 0)
+                            }
+                        }
+                        return {
+                            x: o,
+                            y: u
+                        }
+                    }
+                    draw(e) {
+                        const t = xG(e),
+                            n = wG(t),
+                            {
+                                backgroundColor: r,
+                                fontColor: a,
+                                fontSize: s,
+                                fontStyle: i,
+                                padding: o
+                            } = this.options;
+                        n.font = `${s}px ${i}`;
+                        const u = this.measureWidth(n),
+                            l = this.measureHeight();
+                        n.fillStyle = r;
+                        const c = this.getUpperLeft(n, t);
+                        n.fillRect(c.x, c.y, u, l), n.fillStyle = a, this.text.forEach((e, t) => {
+                            const r = o + c.x,
+                                a = o + c.y + (t + 1) * s;
+                            n.fillText(e, r, a)
+                        })
+                    }
+                },
+                AG = class {
+                    constructor(e = {}) {
+                        const {
+                            boxColor: t,
+                            lineWidth: n,
+                            label: r,
+                            drawLabelOptions: a
+                        } = e;
+                        this.boxColor = t || "rgba(0, 0, 255, 1)", this.lineWidth = n || 2, this.label = r;
+                        const s = {
+                            anchorPosition: "BOTTOM_LEFT",
+                            backgroundColor: this.boxColor
+                        };
+                        this.drawLabelOptions = new SG({
+                            ...s,
+                            ...a
+                        })
+                    }
+                },
+                EG = class {
+                    constructor(e, t = {}) {
+                        this.box = new Jj(e), this.options = new AG(t)
+                    }
+                    draw(e) {
+                        const t = wG(e),
+                            {
+                                boxColor: n,
+                                lineWidth: r
+                            } = this.options,
+                            {
+                                x: a,
+                                y: s,
+                                width: i,
+                                height: o
+                            } = this.box;
+                        t.strokeStyle = n, t.lineWidth = r, t.strokeRect(a, s, i, o);
+                        const {
+                            label: u
+                        } = this.options;
+                        u && new IG([u], {
+                            x: a - r / 2,
+                            y: s
+                        }, this.options.drawLabelOptions).draw(e)
+                    }
+                };
+
+            function NG(e, t) {
+                (Array.isArray(t) ? t : [t]).forEach(t => {
+                    const n = t instanceof nG ? t.score : cG(t) ? t.detection.score : void 0,
+                        r = t instanceof nG ? t.box : cG(t) ? t.detection.box : new Jj(t),
+                        a = n ? "" + Gj(n) : void 0;
+                    new EG(r, {
+                        label: a
+                    }).draw(e)
+                })
+            }
+
+            function CG(e) {
+                const {
+                    Image: t,
+                    Video: n
+                } = vG.getEnv();
+                return e instanceof t && e.complete || e instanceof n && e.readyState >= 3
+            }
+
+            function TG(e) {
+                const {
+                    Image: t,
+                    Video: n
+                } = vG.getEnv();
+                return e instanceof t ? new Pj(e.naturalWidth, e.naturalHeight) : e instanceof n ? new Pj(e.videoWidth, e.videoHeight) : new Pj(e.width, e.height)
+            }
+
+            function RG({
+                width: e,
+                height: t
+            }) {
+                const {
+                    createCanvasElement: n
+                } = vG.getEnv(), r = n();
+                return r.width = e, r.height = t, r
+            }
+
+            function _G(e, t) {
+                const {
+                    ImageData: n
+                } = vG.getEnv();
+                if (!(e instanceof n || CG(e))) throw new Error("createCanvasFromMedia - media has not finished loading yet");
+                const {
+                    width: r,
+                    height: a
+                } = t || TG(e), s = RG({
+                    width: r,
+                    height: a
+                });
+                return e instanceof n ? wG(s).putImageData(e, 0, 0) : wG(s).drawImage(e, 0, 0, r, a), s
+            }
+
+            function FG(e) {
+                const {
+                    Image: t,
+                    Canvas: n,
+                    Video: r
+                } = vG.getEnv();
+                return e instanceof t || e instanceof n || e instanceof r
+            }
+            var MG = class {
+                constructor(e, t = !1) {
+                    if (this._imageTensors = [], this._canvases = [], this._treatAsBatchInput = !1, this._inputDimensions = [], this._inputSize = 0, !Array.isArray(e)) throw new Error("NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have " + e);
+                    this._treatAsBatchInput = t, this._batchSize = e.length, e.forEach((e, t) => {
+                        if (Uj(e)) return this._imageTensors[t] = e, void(this._inputDimensions[t] = e.shape);
+                        if (Wj(e)) {
+                            const n = e.shape[0];
+                            if (1 !== n) throw new Error(`NetInput - tf.Tensor4D with batchSize ${n} passed, but not supported in input array`);
+                            return this._imageTensors[t] = e, void(this._inputDimensions[t] = e.shape.slice(1))
+                        }
+                        const n = e instanceof vG.getEnv().Canvas ? e : _G(e);
+                        this._canvases[t] = n, this._inputDimensions[t] = [n.height, n.width, 3]
+                    })
+                }
+                get imageTensors() {
+                    return this._imageTensors
+                }
+                get canvases() {
+                    return this._canvases
+                }
+                get isBatchInput() {
+                    return this.batchSize > 1 || this._treatAsBatchInput
+                }
+                get batchSize() {
+                    return this._batchSize
+                }
+                get inputDimensions() {
+                    return this._inputDimensions
+                }
+                get inputSize() {
+                    return this._inputSize
+                }
+                get reshapedInputDimensions() {
+                    return Yj(this.batchSize, 0, 1).map((e, t) => this.getReshapedInputDimensions(t))
+                }
+                getInput(e) {
+                    return this.canvases[e] || this.imageTensors[e]
+                }
+                getInputDimensions(e) {
+                    return this._inputDimensions[e]
+                }
+                getInputHeight(e) {
+                    return this._inputDimensions[e][0]
+                }
+                getInputWidth(e) {
+                    return this._inputDimensions[e][1]
+                }
+                getReshapedInputDimensions(e) {
+                    if ("number" != typeof this.inputSize) throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");
+                    return qj({
+                        width: this.getInputWidth(e),
+                        height: this.getInputHeight(e)
+                    }, this.inputSize)
+                }
+                toBatchTensor(e, t = !0) {
+                    return this._inputSize = e, ru(() => {
+                        const n = Yj(this.batchSize, 0, 1).map(n => {
+                            const r = this.getInput(n);
+                            if (r instanceof Ya) {
+                                let n = Wj(r) ? r : Bl(r);
+                                return n = function (e, t = !1) {
+                                    return ru(() => {
+                                        const [n, r] = e.shape.slice(1);
+                                        if (n === r) return e;
+                                        const a = Math.abs(n - r),
+                                            s = Math.round(a * (t ? .5 : 1)),
+                                            i = n > r ? 2 : 1,
+                                            o = t => {
+                                                const n = e.shape.slice();
+                                                return n[i] = t, Vl(n, 0, "float32")
+                                            },
+                                            u = o(s),
+                                            l = a - u.shape[i],
+                                            c = [t && l ? o(l) : null, e, u].filter(e => !!e).map(e => Ai(e, "float32"));
+                                        return Xu(c, i)
+                                    })
+                                }(n, t), n.shape[1] === e && n.shape[2] === e || (n = Nd.resizeBilinear(n, [e, e], !1, !1)), n.as3D(e, e, 3)
+                            }
+                            if (r instanceof vG.getEnv().Canvas) return eo.fromPixels(function (e, t, n = !1) {
+                                const {
+                                    Image: r,
+                                    Canvas: a
+                                } = vG.getEnv();
+                                if (!(e instanceof r || e instanceof a)) throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");
+                                if (t <= 0) return RG({
+                                    width: 1,
+                                    height: 1
+                                });
+                                const s = TG(e),
+                                    i = t / Math.max(s.height, s.width),
+                                    o = i * s.width,
+                                    u = i * s.height,
+                                    l = RG({
+                                        width: t,
+                                        height: t
+                                    }),
+                                    c = e instanceof a ? e : _G(e),
+                                    h = Math.abs(o - u) / 2,
+                                    p = n && o < u ? h : 0,
+                                    d = n && u < o ? h : 0;
+                                return c.width > 0 && c.height > 0 && wG(l).drawImage(c, p, d, o, u), l
+                            }(r, e, t));
+                            throw new Error(`toBatchTensor - at batchIdx ${n}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${r}`)
+                        });
+                        return Gh(n.map(e => Ai(e, "float32"))).as4D(this.batchSize, e, e, 3)
+                    })
+                }
+            };
+            async function DG(e) {
+                if (e instanceof MG) return e;
+                const t = Array.isArray(e) ? e : [e];
+                if (!t.length) throw new Error("toNetInput - empty array passed as input");
+                const n = t => Array.isArray(e) ? ` at input index ${t}:` : "",
+                    r = t.map(xG);
+                return r.forEach((e, r) => {
+                    if (!FG(e) && !Uj(e) && !Wj(e)) {
+                        if ("string" == typeof t[r]) throw new Error(`toNetInput -${n(r)} string passed, but could not resolve HTMLElement for element id ${t[r]}`);
+                        throw new Error(`toNetInput -${n(r)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`)
+                    }
+                    if (Wj(e)) {
+                        const t = e.shape[0];
+                        if (1 !== t) throw new Error(`toNetInput -${n(r)} tf.Tensor4D with batchSize ${t} passed, but not supported in input array`)
+                    }
+                }), await Promise.all(r.map(e => {
+                    return FG(e) && (t = e, new Promise((e, n) => {
+                        function r(e) {
+                            e.currentTarget && (e.currentTarget.removeEventListener("load", a), e.currentTarget.removeEventListener("error", r), n(e))
+                        }
+
+                        function a(t) {
+                            t.currentTarget && (t.currentTarget.removeEventListener("load", a), t.currentTarget.removeEventListener("error", r), e(t))
+                        }(t instanceof vG.getEnv().Canvas || CG(t)) && e(null), t.addEventListener("load", a), t.addEventListener("error", r)
+                    }));
+                    var t
+                })), new MG(r, Array.isArray(e))
+            }
+            async function OG(e, t) {
+                const {
+                    Canvas: n
+                } = vG.getEnv();
+                let r = e;
+                if (!(e instanceof n)) {
+                    const t = await DG(e);
+                    if (t.batchSize > 1) throw new Error("extractFaces - batchSize > 1 not supported");
+                    const a = t.getInput(0);
+                    r = a instanceof n ? a : await async function (e, t) {
+                        const n = t || vG.getEnv().createCanvasElement(),
+                            [r, a, s] = e.shape.slice(Wj(e) ? 1 : 0),
+                            i = ru(() => e.as3D(r, a, s).toInt());
+                        return await eo.toPixels(i, n), i.dispose(), n
+                    }(a)
+                }
+                const a = wG(r);
+                return t.map(e => e instanceof nG ? e.forSize(r.width, r.height).box.floor() : e).map(e => e.clipAtImageBorders(r.width, r.height)).map(({
+                    x: e,
+                    y: t,
+                    width: n,
+                    height: r
+                }) => {
+                    const s = RG({
+                        width: n,
+                        height: r
+                    });
+                    return n > 0 && r > 0 && wG(s).putImageData(a.getImageData(e, t, n, r), 0, 0), s
+                })
+            }
+            async function LG(e, t) {
+                if (!Uj(e) && !Wj(e)) throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");
+                if (Wj(e) && e.shape[0] > 1) throw new Error("extractFaceTensors - batchSize > 1 not supported");
+                return ru(() => {
+                    const [n, r, a] = e.shape.slice(Wj(e) ? 1 : 0);
+                    return t.map(e => e instanceof nG ? e.forSize(r, n).box : e).map(e => e.clipAtImageBorders(r, n)).map(({
+                        x: t,
+                        y: s,
+                        width: i,
+                        height: o
+                    }) => Dh(e.as3D(n, r, a), [s, t, 0], [o, i, a]))
+                })
+            }
+            async function PG(e) {
+                return (await async function (e, t) {
+                    const {
+                        fetch: n
+                    } = vG.getEnv(), r = await n(e, t);
+                    if (!(r.status < 400)) throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);
+                    return r
+                }(e)).json()
+            }
+
+            function $G(e, t) {
+                const n = t + "-weights_manifest.json";
+                if (!e) return {
+                    modelBaseUri: "",
+                    manifestUri: n
+                };
+                if ("/" === e) return {
+                    modelBaseUri: "/",
+                    manifestUri: "/" + n
+                };
+                const r = e.startsWith("http://") ? "http://" : e.startsWith("https://") ? "https://" : "",
+                    a = (e = e.replace(r, "")).split("/").filter(e => e),
+                    s = e.endsWith(".json") ? a[a.length - 1] : n;
+                let i = r + (e.endsWith(".json") ? a.slice(0, a.length - 1) : a).join("/");
+                return i = e.startsWith("/") ? "/" + i : i, {
+                    modelBaseUri: i,
+                    manifestUri: "/" === i ? "/" + s : `${i}/${s}`
+                }
+            }
+            var BG = class {
+                constructor(e) {
+                    this._params = void 0, this._paramMappings = [], this._name = e
+                }
+                get params() {
+                    return this._params
+                }
+                get paramMappings() {
+                    return this._paramMappings
+                }
+                get isLoaded() {
+                    return !!this.params
+                }
+                getParamFromPath(e) {
+                    const {
+                        obj: t,
+                        objProp: n
+                    } = this.traversePropertyPath(e);
+                    return t[n]
+                }
+                reassignParamFromPath(e, t) {
+                    const {
+                        obj: n,
+                        objProp: r
+                    } = this.traversePropertyPath(e);
+                    n[r].dispose(), n[r] = t
+                }
+                getParamList() {
+                    return this._paramMappings.map(({
+                        paramPath: e
+                    }) => ({
+                        path: e,
+                        tensor: this.getParamFromPath(e)
+                    }))
+                }
+                getTrainableParams() {
+                    return this.getParamList().filter(e => e.tensor instanceof Qa)
+                }
+                getFrozenParams() {
+                    return this.getParamList().filter(e => !(e.tensor instanceof Qa))
+                }
+                variable() {
+                    this.getFrozenParams().forEach(({
+                        path: e,
+                        tensor: t
+                    }) => {
+                        this.reassignParamFromPath(e, t.variable())
+                    })
+                }
+                freeze() {
+                    this.getTrainableParams().forEach(({
+                        path: e,
+                        tensor: t
+                    }) => {
+                        const n = Ls(t.dataSync());
+                        t.dispose(), this.reassignParamFromPath(e, n)
+                    })
+                }
+                dispose(e = !0) {
+                    this.getParamList().forEach(t => {
+                        if (e && t.tensor.isDisposed) throw new Error("param tensor has already been disposed for path " + t.path);
+                        t.tensor.dispose()
+                    }), this._params = void 0
+                }
+                serializeParams() {
+                    return new Float32Array(this.getParamList().map(({
+                        tensor: e
+                    }) => Array.from(e.dataSync())).reduce((e, t) => e.concat(t)))
+                }
+                async load(e) {
+                    e instanceof Float32Array ? this.extractWeights(e) : await this.loadFromUri(e)
+                }
+                async loadFromUri(e) {
+                    if (e && "string" != typeof e) throw new Error(this._name + ".loadFromUri - expected model uri");
+                    const t = await async function (e, t) {
+                        const {
+                            manifestUri: n,
+                            modelBaseUri: r
+                        } = $G(e, t), a = await PG(n);
+                        return Ci.loadWeights(a, r)
+                    }(e, this.getDefaultModelName());
+                    this.loadFromWeightMap(t)
+                }
+                async loadFromDisk(e) {
+                    if (e && "string" != typeof e) throw new Error(this._name + ".loadFromDisk - expected model file path");
+                    const {
+                        readFile: t
+                    } = vG.getEnv(), {
+                        manifestUri: n,
+                        modelBaseUri: r
+                    } = $G(e, this.getDefaultModelName()), a = Ci.weightsLoaderFactory(e => Promise.all(e.map(e => t(e).then(e => e.buffer)))), s = JSON.parse((await t(n)).toString()), i = await a(s, r);
+                    this.loadFromWeightMap(i)
+                }
+                loadFromWeightMap(e) {
+                    const {
+                        paramMappings: t,
+                        params: n
+                    } = this.extractParamsFromWeightMap(e);
+                    this._paramMappings = t, this._params = n
+                }
+                extractWeights(e) {
+                    const {
+                        paramMappings: t,
+                        params: n
+                    } = this.extractParams(e);
+                    this._paramMappings = t, this._params = n
+                }
+                traversePropertyPath(e) {
+                    if (!this.params) throw new Error("traversePropertyPath - model has no loaded params");
+                    const t = e.split("/").reduce((t, n) => {
+                            if (!t.nextObj.hasOwnProperty(n)) throw new Error(`traversePropertyPath - object does not have property ${n}, for path ${e}`);
+                            return {
+                                obj: t.nextObj,
+                                objProp: n,
+                                nextObj: t.nextObj[n]
+                            }
+                        }, {
+                            nextObj: this.params
+                        }),
+                        {
+                            obj: n,
+                            objProp: r
+                        } = t;
+                    if (!(n && r && n[r] instanceof Ya)) throw new Error("traversePropertyPath - parameter is not a tensor, for path " + e);
+                    return {
+                        obj: n,
+                        objProp: r
+                    }
+                }
+            };
+
+            function zG(e, t, n) {
+                return ru(() => {
+                    let r = Nh(e, t.depthwise_filter, t.pointwise_filter, n, "same");
+                    return r = gu(r, t.bias), r
+                })
+            }
+
+            function UG(e, t, n = !1) {
+                return ru(() => {
+                    const r = gh(n ? gu(ml(e, t.conv0.filters, [2, 2], "same"), t.conv0.bias) : zG(e, t.conv0, [2, 2])),
+                        a = zG(r, t.conv1, [1, 1]),
+                        s = zG(gh(gu(r, a)), t.conv2, [1, 1]);
+                    return gh(gu(r, gu(a, s)))
+                })
+            }
+
+            function WG(e, t, n = !1, r = !0) {
+                return ru(() => {
+                    const a = gh(n ? gu(ml(e, t.conv0.filters, r ? [2, 2] : [1, 1], "same"), t.conv0.bias) : zG(e, t.conv0, r ? [2, 2] : [1, 1])),
+                        s = zG(a, t.conv1, [1, 1]),
+                        i = zG(gh(gu(a, s)), t.conv2, [1, 1]),
+                        o = zG(gh(gu(a, gu(s, i))), t.conv3, [1, 1]);
+                    return gh(gu(a, gu(s, gu(i, o))))
+                })
+            }
+
+            function VG(e, t, n = "same", r = !1) {
+                return ru(() => {
+                    const a = gu(ml(e, t.filters, [1, 1], n), t.bias);
+                    return r ? gh(a) : a
+                })
+            }
+
+            function jG(e, t) {
+                Object.keys(e).forEach(n => {
+                    t.some(e => e.originalPath === n) || e[n].dispose()
+                })
+            }
+
+            function GG(e, t) {
+                return (n, r, a, s) => {
+                    const i = Qh(e(n * r * a * a), [a, a, n, r]),
+                        o = Yh(e(r));
+                    return t.push({
+                        paramPath: s + "/filters"
+                    }, {
+                        paramPath: s + "/bias"
+                    }), {
+                        filters: i,
+                        bias: o
+                    }
+                }
+            }
+
+            function HG(e, t) {
+                return (n, r, a) => {
+                    const s = Xh(e(n * r), [n, r]),
+                        i = Yh(e(r));
+                    return t.push({
+                        paramPath: a + "/weights"
+                    }, {
+                        paramPath: a + "/bias"
+                    }), {
+                        weights: s,
+                        bias: i
+                    }
+                }
+            }
+            var qG = class {
+                constructor(e, t, n) {
+                    this.depthwise_filter = e, this.pointwise_filter = t, this.bias = n
+                }
+            };
+
+            function KG(e, t) {
+                return (n, r, a) => {
+                    const s = Qh(e(9 * n), [3, 3, n, 1]),
+                        i = Qh(e(n * r), [1, 1, n, r]),
+                        o = Yh(e(r));
+                    return t.push({
+                        paramPath: a + "/depthwise_filter"
+                    }, {
+                        paramPath: a + "/pointwise_filter"
+                    }, {
+                        paramPath: a + "/bias"
+                    }), new qG(s, i, o)
+                }
+            }
+
+            function YG(e) {
+                return t => {
+                    const n = e(t + "/depthwise_filter", 4),
+                        r = e(t + "/pointwise_filter", 4),
+                        a = e(t + "/bias", 1);
+                    return new qG(n, r, a)
+                }
+            }
+
+            function XG(e, t) {
+                return (n, r, a) => {
+                    const s = e[n];
+                    if (!$j(s, r)) throw new Error(`expected weightMap[${n}] to be a Tensor${r}D, instead have ${s}`);
+                    return t.push({
+                        originalPath: n,
+                        paramPath: a || n
+                    }), s
+                }
+            }
+
+            function QG(e) {
+                let t = e;
+                return {
+                    extractWeights: function (e) {
+                        const n = t.slice(0, e);
+                        return t = t.slice(e), n
+                    },
+                    getRemainingWeights: function () {
+                        return t
+                    }
+                }
+            }
+
+            function ZG(e, t) {
+                const n = GG(e, t),
+                    r = KG(e, t);
+
+                function a(e, t, a, s = !1) {
+                    return {
+                        conv0: s ? n(e, t, 3, a + "/conv0") : r(e, t, a + "/conv0"),
+                        conv1: r(t, t, a + "/conv1"),
+                        conv2: r(t, t, a + "/conv2")
+                    }
+                }
+                return {
+                    extractDenseBlock3Params: a,
+                    extractDenseBlock4Params: function (e, t, n, s = !1) {
+                        const {
+                            conv0: i,
+                            conv1: o,
+                            conv2: u
+                        } = a(e, t, n, s);
+                        return {
+                            conv0: i,
+                            conv1: o,
+                            conv2: u,
+                            conv3: r(t, t, n + "/conv3")
+                        }
+                    }
+                }
+            }
+
+            function JG(e) {
+                return t => ({
+                    filters: e(t + "/filters", 4),
+                    bias: e(t + "/bias", 1)
+                })
+            }
+
+            function eH(e, t) {
+                const n = XG(e, t),
+                    r = JG(n),
+                    a = YG(n);
+                return {
+                    extractDenseBlock3Params: function (e, t = !1) {
+                        return {
+                            conv0: t ? r(e + "/conv0") : a(e + "/conv0"),
+                            conv1: a(e + "/conv1"),
+                            conv2: a(e + "/conv2")
+                        }
+                    },
+                    extractDenseBlock4Params: function (e, t = !1) {
+                        return {
+                            conv0: t ? r(e + "/conv0") : a(e + "/conv0"),
+                            conv1: a(e + "/conv1"),
+                            conv2: a(e + "/conv2"),
+                            conv3: a(e + "/conv3")
+                        }
+                    }
+                }
+            }
+            var tH = class extends BG {
+                constructor() {
+                    super("FaceFeatureExtractor")
+                }
+                forwardInput(e) {
+                    const {
+                        params: t
+                    } = this;
+                    if (!t) throw new Error("FaceFeatureExtractor - load model before inference");
+                    return ru(() => {
+                        let n = WG(aG(Ai(e.toBatchTensor(112, !0), "float32"), [122.782, 117.001, 104.298]).div(255), t.dense0, !0);
+                        return n = WG(n, t.dense1), n = WG(n, t.dense2), n = WG(n, t.dense3), n = Ku(n, [7, 7], [2, 2], "valid"), n
+                    })
+                }
+                async forward(e) {
+                    return this.forwardInput(await DG(e))
+                }
+                getDefaultModelName() {
+                    return "face_feature_extractor_model"
+                }
+                extractParamsFromWeightMap(e) {
+                    return function (e) {
+                        const t = [],
+                            {
+                                extractDenseBlock4Params: n
+                            } = eH(e, t),
+                            r = {
+                                dense0: n("dense0", !0),
+                                dense1: n("dense1"),
+                                dense2: n("dense2"),
+                                dense3: n("dense3")
+                            };
+                        return jG(e, t), {
+                            params: r,
+                            paramMappings: t
+                        }
+                    }(e)
+                }
+                extractParams(e) {
+                    return function (e) {
+                        const t = [],
+                            {
+                                extractWeights: n,
+                                getRemainingWeights: r
+                            } = QG(e),
+                            {
+                                extractDenseBlock4Params: a
+                            } = ZG(n, t),
+                            s = a(3, 32, "dense0", !0),
+                            i = a(32, 64, "dense1"),
+                            o = a(64, 128, "dense2"),
+                            u = a(128, 256, "dense3");
+                        if (0 !== r().length) throw new Error("weights remaing after extract: " + r().length);
+                        return {
+                            paramMappings: t,
+                            params: {
+                                dense0: s,
+                                dense1: i,
+                                dense2: o,
+                                dense3: u
+                            }
+                        }
+                    }(e)
+                }
+            };
+
+            function nH(e, t) {
+                return ru(() => gu(Gi(e, t.weights), t.bias))
+            }
+
+            function rH(e) {
+                const t = {},
+                    n = {};
+                return Object.keys(e).forEach(r => {
+                    (r.startsWith("fc") ? n : t)[r] = e[r]
+                }), {
+                    featureExtractorMap: t,
+                    classifierMap: n
+                }
+            }
+            var aH = class extends BG {
+                    constructor(e, t) {
+                        super(e), this._faceFeatureExtractor = t
+                    }
+                    get faceFeatureExtractor() {
+                        return this._faceFeatureExtractor
+                    }
+                    runNet(e) {
+                        const {
+                            params: t
+                        } = this;
+                        if (!t) throw new Error(this._name + " - load model before inference");
+                        return ru(() => {
+                            const n = e instanceof MG ? this.faceFeatureExtractor.forwardInput(e) : e;
+                            return nH(n.as2D(n.shape[0], -1), t.fc)
+                        })
+                    }
+                    dispose(e = !0) {
+                        this.faceFeatureExtractor.dispose(e), super.dispose(e)
+                    }
+                    loadClassifierParams(e) {
+                        const {
+                            params: t,
+                            paramMappings: n
+                        } = this.extractClassifierParams(e);
+                        this._params = t, this._paramMappings = n
+                    }
+                    extractClassifierParams(e) {
+                        return function (e, t, n) {
+                            const r = [],
+                                {
+                                    extractWeights: a,
+                                    getRemainingWeights: s
+                                } = QG(e),
+                                i = HG(a, r)(t, n, "fc");
+                            if (0 !== s().length) throw new Error("weights remaing after extract: " + s().length);
+                            return {
+                                paramMappings: r,
+                                params: {
+                                    fc: i
+                                }
+                            }
+                        }(e, this.getClassifierChannelsIn(), this.getClassifierChannelsOut())
+                    }
+                    extractParamsFromWeightMap(e) {
+                        const {
+                            featureExtractorMap: t,
+                            classifierMap: n
+                        } = rH(e);
+                        return this.faceFeatureExtractor.loadFromWeightMap(t),
+                            function (e) {
+                                const t = [],
+                                    n = XG(e, t),
+                                    r = {
+                                        fc: (a = "fc", {
+                                            weights: n(a + "/weights", 2),
+                                            bias: n(a + "/bias", 1)
+                                        })
+                                    };
+                                var a;
+                                return jG(e, t), {
+                                    params: r,
+                                    paramMappings: t
+                                }
+                            }(n)
+                    }
+                    extractParams(e) {
+                        const t = this.getClassifierChannelsIn(),
+                            n = this.getClassifierChannelsOut(),
+                            r = n * t + n,
+                            a = e.slice(0, e.length - r),
+                            s = e.slice(e.length - r);
+                        return this.faceFeatureExtractor.extractWeights(a), this.extractClassifierParams(s)
+                    }
+                },
+                sH = ["neutral", "happy", "sad", "angry", "fearful", "disgusted", "surprised"],
+                iH = class {
+                    constructor(e) {
+                        if (this.neutral = 0, this.happy = 0, this.sad = 0, this.angry = 0, this.fearful = 0, this.disgusted = 0, this.surprised = 0, 7 !== e.length) throw new Error("FaceExpressions.constructor - expected probabilities.length to be 7, have: " + e.length);
+                        sH.forEach((t, n) => {
+                            this[t] = e[n]
+                        })
+                    }
+                    asSortedArray() {
+                        return sH.map(e => ({
+                            expression: e,
+                            probability: this[e]
+                        })).sort((e, t) => t.probability - e.probability)
+                    }
+                };
+
+            function oH(e, t, n = .1, r) {
+                (Array.isArray(t) ? t : [t]).forEach(t => {
+                    const a = t instanceof iH ? t : t.expressions instanceof iH ? t.expressions : void 0;
+                    if (!a) throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");
+                    const s = a.asSortedArray().filter(e => e.probability > n),
+                        i = cG(t) ? t.detection.box.bottomLeft : r || new Zj(0, 0);
+                    new IG(s.map(e => `${e.expression} (${Gj(e.probability)})`), i).draw(e)
+                })
+            }
+
+            function uH(e) {
+                return cG(e) && e.landmarks instanceof uG && e.unshiftedLandmarks instanceof uG && e.alignedRect instanceof nG
+            }
+
+            function lH(e, t) {
+                const {
+                    box: n
+                } = e.detection, r = t.shiftBy(n.x, n.y), a = r.align(), {
+                    imageDims: s
+                } = e.detection, i = {
+                    landmarks: r,
+                    unshiftedLandmarks: t,
+                    alignedRect: new nG(e.detection.score, a.rescale(s.reverse()), s),
+                    angle: function (e) {
+                        const t = (e, t, n, r) => Math.atan2(r - t, n - e) % Math.PI,
+                            n = {
+                                roll: void 0,
+                                pitch: void 0,
+                                yaw: void 0
+                            };
+                        if (!e || !e._positions || 68 !== e._positions.length) return n;
+                        const r = e._positions;
+                        n.roll = -t(r[36]._x, r[36]._y, r[45]._x, r[45]._y), n.pitch = t(0, Math.abs(r[0]._x - r[30]._x) / r[30]._x, Math.PI, Math.abs(r[16]._x - r[30]._x) / r[30]._x);
+                        const a = r.reduce((e, t) => e < t._y ? e : t._y, 1 / 0),
+                            s = r.reduce((e, t) => e > t._y ? e : t._y, -1 / 0);
+                        return n.yaw = Math.PI * (e._imgDims._height / (s - a) / 1.4 - 1), n
+                    }(t)
+                };
+                return {
+                    ...e,
+                    ...i
+                }
+            }
+            var cH = class {
+                    constructor(e = {}) {
+                        const {
+                            drawLines: t = !0,
+                            drawPoints: n = !0,
+                            lineWidth: r,
+                            lineColor: a,
+                            pointSize: s,
+                            pointColor: i
+                        } = e;
+                        this.drawLines = t, this.drawPoints = n, this.lineWidth = r || 1, this.pointSize = s || 2, this.lineColor = a || "rgba(0, 255, 255, 1)", this.pointColor = i || "rgba(255, 0, 255, 1)"
+                    }
+                },
+                hH = class {
+                    constructor(e, t = {}) {
+                        this.faceLandmarks = e, this.options = new cH(t)
+                    }
+                    draw(e) {
+                        const t = wG(e),
+                            {
+                                drawLines: n,
+                                drawPoints: r,
+                                lineWidth: a,
+                                lineColor: s,
+                                pointSize: i,
+                                pointColor: o
+                            } = this.options;
+                        if (n && this.faceLandmarks instanceof lG && (t.strokeStyle = s, t.lineWidth = a, Lj(t, this.faceLandmarks.getJawOutline()), Lj(t, this.faceLandmarks.getLeftEyeBrow()), Lj(t, this.faceLandmarks.getRightEyeBrow()), Lj(t, this.faceLandmarks.getNose()), Lj(t, this.faceLandmarks.getLeftEye(), !0), Lj(t, this.faceLandmarks.getRightEye(), !0), Lj(t, this.faceLandmarks.getMouth(), !0)), r) {
+                            t.strokeStyle = o, t.fillStyle = o;
+                            const e = e => {
+                                t.beginPath(), t.arc(e.x, e.y, i, 0, 2 * Math.PI), t.fill()
+                            };
+                            this.faceLandmarks.positions.forEach(e)
+                        }
+                    }
+                };
+
+            function pH(e, t) {
+                (Array.isArray(t) ? t : [t]).forEach(t => {
+                    const n = t instanceof uG ? t : uH(t) ? t.landmarks : void 0;
+                    if (!n) throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");
+                    new hH(n).draw(e)
+                })
+            }
+
+            function dH(e, t) {
+                const n = [],
+                    {
+                        extractWeights: r,
+                        getRemainingWeights: a
+                    } = QG(e),
+                    {
+                        extractConvParams: s,
+                        extractSeparableConvParams: i,
+                        extractReductionBlockParams: o,
+                        extractMainBlockParams: u
+                    } = function (e, t) {
+                        const n = GG(e, t),
+                            r = KG(e, t);
+                        return {
+                            extractConvParams: n,
+                            extractSeparableConvParams: r,
+                            extractReductionBlockParams: function (e, t, a) {
+                                return {
+                                    separable_conv0: r(e, t, a + "/separable_conv0"),
+                                    separable_conv1: r(t, t, a + "/separable_conv1"),
+                                    expansion_conv: n(e, t, 1, a + "/expansion_conv")
+                                }
+                            },
+                            extractMainBlockParams: function (e, t) {
+                                return {
+                                    separable_conv0: r(e, e, t + "/separable_conv0"),
+                                    separable_conv1: r(e, e, t + "/separable_conv1"),
+                                    separable_conv2: r(e, e, t + "/separable_conv2")
+                                }
+                            }
+                        }
+                    }(r, n),
+                    l = {
+                        conv_in: s(3, 32, 3, "entry_flow/conv_in"),
+                        reduction_block_0: o(32, 64, "entry_flow/reduction_block_0"),
+                        reduction_block_1: o(64, 128, "entry_flow/reduction_block_1")
+                    },
+                    c = {};
+                Yj(t, 0, 1).forEach(e => {
+                    c["main_block_" + e] = u(128, "middle_flow/main_block_" + e)
+                });
+                const h = {
+                    reduction_block: o(128, 256, "exit_flow/reduction_block"),
+                    separable_conv: i(256, 512, "exit_flow/separable_conv")
+                };
+                if (0 !== a().length) throw new Error("weights remaing after extract: " + a().length);
+                return {
+                    paramMappings: n,
+                    params: {
+                        entry_flow: l,
+                        middle_flow: c,
+                        exit_flow: h
+                    }
+                }
+            }
+
+            function fH(e, t) {
+                const n = [],
+                    {
+                        extractConvParams: r,
+                        extractSeparableConvParams: a,
+                        extractReductionBlockParams: s,
+                        extractMainBlockParams: i
+                    } = function (e, t) {
+                        const n = XG(e, t),
+                            r = JG(n),
+                            a = YG(n);
+                        return {
+                            extractConvParams: r,
+                            extractSeparableConvParams: a,
+                            extractReductionBlockParams: function (e) {
+                                return {
+                                    separable_conv0: a(e + "/separable_conv0"),
+                                    separable_conv1: a(e + "/separable_conv1"),
+                                    expansion_conv: r(e + "/expansion_conv")
+                                }
+                            },
+                            extractMainBlockParams: function (e) {
+                                return {
+                                    separable_conv0: a(e + "/separable_conv0"),
+                                    separable_conv1: a(e + "/separable_conv1"),
+                                    separable_conv2: a(e + "/separable_conv2")
+                                }
+                            }
+                        }
+                    }(e, n),
+                    o = {
+                        conv_in: r("entry_flow/conv_in"),
+                        reduction_block_0: s("entry_flow/reduction_block_0"),
+                        reduction_block_1: s("entry_flow/reduction_block_1")
+                    },
+                    u = {};
+                Yj(t, 0, 1).forEach(e => {
+                    u["main_block_" + e] = i("middle_flow/main_block_" + e)
+                });
+                const l = {
+                    reduction_block: s("exit_flow/reduction_block"),
+                    separable_conv: a("exit_flow/separable_conv")
+                };
+                return jG(e, n), {
+                    params: {
+                        entry_flow: o,
+                        middle_flow: u,
+                        exit_flow: l
+                    },
+                    paramMappings: n
+                }
+            }
+
+            function mH(e, t, n) {
+                return gu(ml(e, t.filters, n, "same"), t.bias)
+            }
+
+            function gH(e, t, n = !0) {
+                let r = n ? gh(e) : e;
+                return r = zG(r, t.separable_conv0, [1, 1]), r = zG(gh(r), t.separable_conv1, [1, 1]), r = Fc(r, [3, 3], [2, 2], "same"), r = gu(r, mH(e, t.expansion_conv, [2, 2])), r
+            }
+            var yH = class extends BG {
+                constructor(e) {
+                    super("TinyXception"), this._numMainBlocks = e
+                }
+                forwardInput(e) {
+                    const {
+                        params: t
+                    } = this;
+                    if (!t) throw new Error("TinyXception - load model before inference");
+                    return ru(() => {
+                        const n = aG(Ai(e.toBatchTensor(112, !0), "float32"), [122.782, 117.001, 104.298]).div(255);
+                        let r = gh(mH(n, t.entry_flow.conv_in, [2, 2]));
+                        return r = gH(r, t.entry_flow.reduction_block_0, !1), r = gH(r, t.entry_flow.reduction_block_1), Yj(this._numMainBlocks, 0, 1).forEach(e => {
+                            r = function (e, t) {
+                                let n = zG(gh(e), t.separable_conv0, [1, 1]);
+                                return n = zG(gh(n), t.separable_conv1, [1, 1]), n = zG(gh(n), t.separable_conv2, [1, 1]), n = gu(n, e), n
+                            }(r, t.middle_flow["main_block_" + e])
+                        }), r = gH(r, t.exit_flow.reduction_block), r = gh(zG(r, t.exit_flow.separable_conv, [1, 1])), r
+                    })
+                }
+                async forward(e) {
+                    return this.forwardInput(await DG(e))
+                }
+                getDefaultModelName() {
+                    return "tiny_xception_model"
+                }
+                extractParamsFromWeightMap(e) {
+                    return fH(e, this._numMainBlocks)
+                }
+                extractParams(e) {
+                    return dH(e, this._numMainBlocks)
+                }
+            };
+            var bH = class extends aH {
+                postProcess(e, t, n) {
+                    const r = n.map(({
+                            width: e,
+                            height: n
+                        }) => {
+                            const r = t / Math.max(n, e);
+                            return {
+                                width: e * r,
+                                height: n * r
+                            }
+                        }),
+                        a = r.length;
+                    return ru(() => {
+                        const n = (e, t) => Gh([Vl([68], e, "float32"), Vl([68], t, "float32")], 1).as2D(1, 136).as1D(),
+                            s = (e, t) => {
+                                const {
+                                    width: n,
+                                    height: a
+                                } = r[e];
+                                return t(n, a) ? Math.abs(n - a) / 2 : 0
+                            };
+                        return e.mul(Vl([a, 136], t, "float32")).sub(Gh(Array.from(Array(a), (e, t) => n((e => s(e, (e, t) => e < t))(t), (e => s(e, (e, t) => t < e))(t))))).div(Gh(Array.from(Array(a), (e, t) => n(r[t].width, r[t].height))))
+                    })
+                }
+                forwardInput(e) {
+                    return ru(() => {
+                        const t = this.runNet(e);
+                        return this.postProcess(t, e.inputSize, e.inputDimensions.map(([e, t]) => ({
+                            height: e,
+                            width: t
+                        })))
+                    })
+                }
+                async forward(e) {
+                    return this.forwardInput(await DG(e))
+                }
+                async detectLandmarks(e) {
+                    const t = await DG(e),
+                        n = ru(() => ap(this.forwardInput(t))),
+                        r = await Promise.all(n.map(async (e, n) => {
+                            const r = Array.from(e.dataSync()),
+                                a = r.filter((e, t) => jj(t)),
+                                s = r.filter((e, t) => !jj(t));
+                            return new lG(Array(68).fill(0).map((e, t) => new Zj(a[t], s[t])), {
+                                height: t.getInputHeight(n),
+                                width: t.getInputWidth(n)
+                            })
+                        }));
+                    return n.forEach(e => e.dispose()), t.isBatchInput ? r : r[0]
+                }
+                getClassifierChannelsOut() {
+                    return 136
+                }
+            };
+            var vH = class extends BG {
+                constructor() {
+                    super("TinyFaceFeatureExtractor")
+                }
+                forwardInput(e) {
+                    const {
+                        params: t
+                    } = this;
+                    if (!t) throw new Error("TinyFaceFeatureExtractor - load model before inference");
+                    return ru(() => {
+                        let n = UG(aG(Ai(e.toBatchTensor(112, !0), "float32"), [122.782, 117.001, 104.298]).div(255), t.dense0, !0);
+                        return n = UG(n, t.dense1), n = UG(n, t.dense2), n = Ku(n, [14, 14], [2, 2], "valid"), n
+                    })
+                }
+                async forward(e) {
+                    return this.forwardInput(await DG(e))
+                }
+                getDefaultModelName() {
+                    return "face_feature_extractor_tiny_model"
+                }
+                extractParamsFromWeightMap(e) {
+                    return function (e) {
+                        const t = [],
+                            {
+                                extractDenseBlock3Params: n
+                            } = eH(e, t),
+                            r = {
+                                dense0: n("dense0", !0),
+                                dense1: n("dense1"),
+                                dense2: n("dense2")
+                            };
+                        return jG(e, t), {
+                            params: r,
+                            paramMappings: t
+                        }
+                    }(e)
+                }
+                extractParams(e) {
+                    return function (e) {
+                        const t = [],
+                            {
+                                extractWeights: n,
+                                getRemainingWeights: r
+                            } = QG(e),
+                            {
+                                extractDenseBlock3Params: a
+                            } = ZG(n, t),
+                            s = a(3, 32, "dense0", !0),
+                            i = a(32, 64, "dense1"),
+                            o = a(64, 128, "dense2");
+                        if (0 !== r().length) throw new Error("weights remaing after extract: " + r().length);
+                        return {
+                            paramMappings: t,
+                            params: {
+                                dense0: s,
+                                dense1: i,
+                                dense2: o
+                            }
+                        }
+                    }(e)
+                }
+            };
+
+            function xH(e, t, n, r, a = "same") {
+                const {
+                    filters: s,
+                    bias: i
+                } = t.conv;
+                let o = ml(e, s, n, a);
+                return o = gu(o, i), o = function (e, t) {
+                    return gu(vu(e, t.weights), t.biases)
+                }(o, t.scale), r ? gh(o) : o
+            }
+
+            function wH(e, t) {
+                return xH(e, t, [1, 1], !1)
+            }
+
+            function kH(e, t) {
+                return xH(e, t, [2, 2], !0, "valid")
+            }
+
+            function SH(e, t) {
+                function n(n, r, a, s) {
+                    const i = function (t, n, r) {
+                            const a = e(t),
+                                s = a.length / (n * r * r);
+                            if (Vj(s)) throw new Error(`depth has to be an integer: ${s}, weights.length: ${a.length}, numFilters: ${n}, filterSize: ${r}`);
+                            return ru(() => qi(Qh(a, [n, s, r, r]), [2, 3, 1, 0]))
+                        }(n, r, a),
+                        o = Yh(e(r));
+                    return t.push({
+                        paramPath: s + "/filters"
+                    }, {
+                        paramPath: s + "/bias"
+                    }), {
+                        filters: i,
+                        bias: o
+                    }
+                }
+
+                function r(r, a, s, i) {
+                    return {
+                        conv: n(r, a, s, i + "/conv"),
+                        scale: function (n, r) {
+                            const a = Yh(e(n)),
+                                s = Yh(e(n));
+                            return t.push({
+                                paramPath: r + "/weights"
+                            }, {
+                                paramPath: r + "/biases"
+                            }), {
+                                weights: a,
+                                biases: s
+                            }
+                        }(a, i + "/scale")
+                    }
+                }
+                return {
+                    extractConvLayerParams: r,
+                    extractResidualLayerParams: function (e, t, n, a, s = !1) {
+                        return {
+                            conv1: r((s ? .5 : 1) * e, t, n, a + "/conv1"),
+                            conv2: r(e, t, n, a + "/conv2")
+                        }
+                    }
+                }
+            }
+
+            function IH(e, t) {
+                const n = XG(e, t);
+
+                function r(e) {
+                    return {
+                        conv: {
+                            filters: n(e + "/conv/filters", 4),
+                            bias: n(e + "/conv/bias", 1)
+                        },
+                        scale: function (e) {
+                            return {
+                                weights: n(e + "/scale/weights", 1),
+                                biases: n(e + "/scale/biases", 1)
+                            }
+                        }(e)
+                    }
+                }
+                return {
+                    extractConvLayerParams: r,
+                    extractResidualLayerParams: function (e) {
+                        return {
+                            conv1: r(e + "/conv1"),
+                            conv2: r(e + "/conv2")
+                        }
+                    }
+                }
+            }
+
+            function AH(e, t) {
+                let n = function (e, t) {
+                    return xH(e, t, [1, 1], !0)
+                }(e, t.conv1);
+                return n = wH(n, t.conv2), n = gu(n, e), n = gh(n), n
+            }
+
+            function EH(e, t) {
+                let n = kH(e, t.conv1);
+                n = wH(n, t.conv2);
+                let r = Ku(e, 2, 2, "valid");
+                const a = Pc(r.shape),
+                    s = r.shape[3] !== n.shape[3];
+                if (r.shape[1] !== n.shape[1] || r.shape[2] !== n.shape[2]) {
+                    const e = [...n.shape];
+                    e[1] = 1;
+                    const t = Pc(e);
+                    n = Xu([n, t], 1);
+                    const r = [...n.shape];
+                    r[2] = 1;
+                    const a = Pc(r);
+                    n = Xu([n, a], 2)
+                }
+                return r = s ? Xu([r, a], 3) : r, n = gu(r, n), n = gh(n), n
+            }
+
+            function NH(e, t) {
+                function n(n, r, a, s, i) {
+                    const o = Qh(e(n * r * a * a), [a, a, n, r]),
+                        u = Yh(e(r));
+                    return t.push({
+                        paramPath: s + "/filters"
+                    }, {
+                        paramPath: `${s}/${i?"batch_norm_offset":"bias"}`
+                    }), {
+                        filters: o,
+                        bias: u
+                    }
+                }
+
+                function r(e, t, r, a) {
+                    const {
+                        filters: s,
+                        bias: i
+                    } = n(e, t, r, a, !0);
+                    return {
+                        filters: s,
+                        batch_norm_offset: i
+                    }
+                }
+
+                function a(n, a, s) {
+                    return {
+                        depthwise_conv: function (n, r) {
+                            const a = Qh(e(9 * n), [3, 3, n, 1]),
+                                s = Yh(e(n)),
+                                i = Yh(e(n)),
+                                o = Yh(e(n)),
+                                u = Yh(e(n));
+                            return t.push({
+                                paramPath: r + "/filters"
+                            }, {
+                                paramPath: r + "/batch_norm_scale"
+                            }, {
+                                paramPath: r + "/batch_norm_offset"
+                            }, {
+                                paramPath: r + "/batch_norm_mean"
+                            }, {
+                                paramPath: r + "/batch_norm_variance"
+                            }), {
+                                filters: a,
+                                batch_norm_scale: s,
+                                batch_norm_offset: i,
+                                batch_norm_mean: o,
+                                batch_norm_variance: u
+                            }
+                        }(n, s + "/depthwise_conv"),
+                        pointwise_conv: r(n, a, 1, s + "/pointwise_conv")
+                    }
+                }
+                return {
+                    extractMobilenetV1Params: function () {
+                        return {
+                            conv_0: r(3, 32, 3, "mobilenetv1/conv_0"),
+                            conv_1: a(32, 64, "mobilenetv1/conv_1"),
+                            conv_2: a(64, 128, "mobilenetv1/conv_2"),
+                            conv_3: a(128, 128, "mobilenetv1/conv_3"),
+                            conv_4: a(128, 256, "mobilenetv1/conv_4"),
+                            conv_5: a(256, 256, "mobilenetv1/conv_5"),
+                            conv_6: a(256, 512, "mobilenetv1/conv_6"),
+                            conv_7: a(512, 512, "mobilenetv1/conv_7"),
+                            conv_8: a(512, 512, "mobilenetv1/conv_8"),
+                            conv_9: a(512, 512, "mobilenetv1/conv_9"),
+                            conv_10: a(512, 512, "mobilenetv1/conv_10"),
+                            conv_11: a(512, 512, "mobilenetv1/conv_11"),
+                            conv_12: a(512, 1024, "mobilenetv1/conv_12"),
+                            conv_13: a(1024, 1024, "mobilenetv1/conv_13")
+                        }
+                    },
+                    extractPredictionLayerParams: function () {
+                        return {
+                            conv_0: r(1024, 256, 1, "prediction_layer/conv_0"),
+                            conv_1: r(256, 512, 3, "prediction_layer/conv_1"),
+                            conv_2: r(512, 128, 1, "prediction_layer/conv_2"),
+                            conv_3: r(128, 256, 3, "prediction_layer/conv_3"),
+                            conv_4: r(256, 128, 1, "prediction_layer/conv_4"),
+                            conv_5: r(128, 256, 3, "prediction_layer/conv_5"),
+                            conv_6: r(256, 64, 1, "prediction_layer/conv_6"),
+                            conv_7: r(64, 128, 3, "prediction_layer/conv_7"),
+                            box_predictor_0: {
+                                box_encoding_predictor: n(512, 12, 1, "prediction_layer/box_predictor_0/box_encoding_predictor"),
+                                class_predictor: n(512, 9, 1, "prediction_layer/box_predictor_0/class_predictor")
+                            },
+                            box_predictor_1: {
+                                box_encoding_predictor: n(1024, 24, 1, "prediction_layer/box_predictor_1/box_encoding_predictor"),
+                                class_predictor: n(1024, 18, 1, "prediction_layer/box_predictor_1/class_predictor")
+                            },
+                            box_predictor_2: {
+                                box_encoding_predictor: n(512, 24, 1, "prediction_layer/box_predictor_2/box_encoding_predictor"),
+                                class_predictor: n(512, 18, 1, "prediction_layer/box_predictor_2/class_predictor")
+                            },
+                            box_predictor_3: {
+                                box_encoding_predictor: n(256, 24, 1, "prediction_layer/box_predictor_3/box_encoding_predictor"),
+                                class_predictor: n(256, 18, 1, "prediction_layer/box_predictor_3/class_predictor")
+                            },
+                            box_predictor_4: {
+                                box_encoding_predictor: n(256, 24, 1, "prediction_layer/box_predictor_4/box_encoding_predictor"),
+                                class_predictor: n(256, 18, 1, "prediction_layer/box_predictor_4/class_predictor")
+                            },
+                            box_predictor_5: {
+                                box_encoding_predictor: n(128, 24, 1, "prediction_layer/box_predictor_5/box_encoding_predictor"),
+                                class_predictor: n(128, 18, 1, "prediction_layer/box_predictor_5/class_predictor")
+                            }
+                        }
+                    }
+                }
+            }
+
+            function CH(e) {
+                const t = [],
+                    {
+                        extractMobilenetV1Params: n,
+                        extractPredictionLayerParams: r
+                    } = function (e, t) {
+                        const n = XG(e, t);
+
+                        function r(e, t, r) {
+                            return {
+                                filters: n(`${e}/Conv2d_${t}_pointwise/weights`, 4, r + "/filters"),
+                                batch_norm_offset: n(`${e}/Conv2d_${t}_pointwise/convolution_bn_offset`, 1, r + "/batch_norm_offset")
+                            }
+                        }
+
+                        function a(e) {
+                            const t = "mobilenetv1/conv_" + e,
+                                a = `MobilenetV1/Conv2d_${e}_depthwise`,
+                                s = t + "/depthwise_conv",
+                                i = t + "/pointwise_conv";
+                            return {
+                                depthwise_conv: {
+                                    filters: n(a + "/depthwise_weights", 4, s + "/filters"),
+                                    batch_norm_scale: n(a + "/BatchNorm/gamma", 1, s + "/batch_norm_scale"),
+                                    batch_norm_offset: n(a + "/BatchNorm/beta", 1, s + "/batch_norm_offset"),
+                                    batch_norm_mean: n(a + "/BatchNorm/moving_mean", 1, s + "/batch_norm_mean"),
+                                    batch_norm_variance: n(a + "/BatchNorm/moving_variance", 1, s + "/batch_norm_variance")
+                                },
+                                pointwise_conv: r("MobilenetV1", e, i)
+                            }
+                        }
+
+                        function s(e, t) {
+                            return {
+                                filters: n(e + "/weights", 4, t + "/filters"),
+                                bias: n(e + "/biases", 1, t + "/bias")
+                            }
+                        }
+
+                        function i(e) {
+                            return {
+                                box_encoding_predictor: s(`Prediction/BoxPredictor_${e}/BoxEncodingPredictor`, `prediction_layer/box_predictor_${e}/box_encoding_predictor`),
+                                class_predictor: s(`Prediction/BoxPredictor_${e}/ClassPredictor`, `prediction_layer/box_predictor_${e}/class_predictor`)
+                            }
+                        }
+                        return {
+                            extractMobilenetV1Params: function () {
+                                return {
+                                    conv_0: r("MobilenetV1", 0, "mobilenetv1/conv_0"),
+                                    conv_1: a(1),
+                                    conv_2: a(2),
+                                    conv_3: a(3),
+                                    conv_4: a(4),
+                                    conv_5: a(5),
+                                    conv_6: a(6),
+                                    conv_7: a(7),
+                                    conv_8: a(8),
+                                    conv_9: a(9),
+                                    conv_10: a(10),
+                                    conv_11: a(11),
+                                    conv_12: a(12),
+                                    conv_13: a(13)
+                                }
+                            },
+                            extractPredictionLayerParams: function () {
+                                return {
+                                    conv_0: r("Prediction", 0, "prediction_layer/conv_0"),
+                                    conv_1: r("Prediction", 1, "prediction_layer/conv_1"),
+                                    conv_2: r("Prediction", 2, "prediction_layer/conv_2"),
+                                    conv_3: r("Prediction", 3, "prediction_layer/conv_3"),
+                                    conv_4: r("Prediction", 4, "prediction_layer/conv_4"),
+                                    conv_5: r("Prediction", 5, "prediction_layer/conv_5"),
+                                    conv_6: r("Prediction", 6, "prediction_layer/conv_6"),
+                                    conv_7: r("Prediction", 7, "prediction_layer/conv_7"),
+                                    box_predictor_0: i(0),
+                                    box_predictor_1: i(1),
+                                    box_predictor_2: i(2),
+                                    box_predictor_3: i(3),
+                                    box_predictor_4: i(4),
+                                    box_predictor_5: i(5)
+                                }
+                            }
+                        }
+                    }(e, t),
+                    a = e["Output/extra_dim"];
+                if (t.push({
+                        originalPath: "Output/extra_dim",
+                        paramPath: "output_layer/extra_dim"
+                    }), !Uj(a)) throw new Error("expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have " + a);
+                const s = {
+                    mobilenetv1: n(),
+                    prediction_layer: r(),
+                    output_layer: {
+                        extra_dim: a
+                    }
+                };
+                return jG(e, t), {
+                    params: s,
+                    paramMappings: t
+                }
+            }
+
+            function TH(e, t, n) {
+                return ru(() => {
+                    let r = ml(e, t.filters, n, "same");
+                    return r = gu(r, t.batch_norm_offset), cl(r, 0, 6)
+                })
+            }
+
+            function RH(e, t) {
+                return ru(() => {
+                    let n, r = TH(e, t.conv_0, [2, 2]);
+                    if ([t.conv_1, t.conv_2, t.conv_3, t.conv_4, t.conv_5, t.conv_6, t.conv_7, t.conv_8, t.conv_9, t.conv_10, t.conv_11, t.conv_12, t.conv_13].forEach((e, t) => {
+                            const a = t + 1,
+                                s = function (e) {
+                                    return [2, 4, 6, 12].some(t => t === e) ? [2, 2] : [1, 1]
+                                }(a);
+                            r = function (e, t, n) {
+                                return ru(() => {
+                                    let r = Nl(e, t.filters, n, "same");
+                                    return r = nl(r, t.batch_norm_mean, t.batch_norm_variance, t.batch_norm_offset, t.batch_norm_scale, .0010000000474974513), cl(r, 0, 6)
+                                })
+                            }(r, e.depthwise_conv, s), r = TH(r, e.pointwise_conv, [1, 1]), 11 === a && (n = r)
+                        }), null === n) throw new Error("mobileNetV1 - output of conv layer 11 is null");
+                    return {
+                        out: r,
+                        conv11: n
+                    }
+                })
+            }
+
+            function _H(e, t, n) {
+                const r = e.arraySync(),
+                    a = Math.min(r[t][0], r[t][2]),
+                    s = Math.min(r[t][1], r[t][3]),
+                    i = Math.max(r[t][0], r[t][2]),
+                    o = Math.max(r[t][1], r[t][3]),
+                    u = Math.min(r[n][0], r[n][2]),
+                    l = Math.min(r[n][1], r[n][3]),
+                    c = Math.max(r[n][0], r[n][2]),
+                    h = Math.max(r[n][1], r[n][3]),
+                    p = (i - a) * (o - s),
+                    d = (c - u) * (h - l);
+                if (p <= 0 || d <= 0) return 0;
+                const f = Math.max(a, u),
+                    m = Math.max(s, l),
+                    g = Math.min(i, c),
+                    y = Math.min(o, h),
+                    b = Math.max(g - f, 0) * Math.max(y - m, 0);
+                return b / (p + d - b)
+            }
+
+            function FH(e, t) {
+                const {
+                    sizes: n,
+                    centers: r
+                } = function (e) {
+                    const t = ap(qi(e, [1, 0])),
+                        n = [gc(t[2], t[0]), gc(t[3], t[1])];
+                    return {
+                        sizes: n,
+                        centers: [gu(t[0], bu(n[0], 2)), gu(t[1], bu(n[1], 2))]
+                    }
+                }(e), a = ap(qi(t, [1, 0])), s = bu(vu($l(bu(a[2], 5)), n[0]), 2), i = gu(vu(bu(a[0], 10), n[0]), r[0]), o = bu(vu($l(bu(a[3], 5)), n[1]), 2), u = gu(vu(bu(a[1], 10), n[1]), r[1]);
+                return qi(Gh([gc(i, s), gc(u, o), gu(i, s), gu(u, o)]), [1, 0])
+            }
+
+            function MH(e, t) {
+                return ru(() => {
+                    const n = e.shape[0];
+                    return {
+                        boxPredictionEncoding: qu(VG(e, t.box_encoding_predictor), [n, -1, 1, 4]),
+                        classPrediction: qu(VG(e, t.class_predictor), [n, -1, 3])
+                    }
+                })
+            }
+            var DH = class {
+                    constructor({
+                        minConfidence: e,
+                        maxResults: t
+                    } = {}) {
+                        if (this._name = "SsdMobilenetv1Options", this._minConfidence = e || .5, this._maxResults = t || 100, "number" != typeof this._minConfidence || this._minConfidence <= 0 || this._minConfidence >= 1) throw new Error(this._name + " - expected minConfidence to be a number between 0 and 1");
+                        if ("number" != typeof this._maxResults) throw new Error(this._name + " - expected maxResults to be a number")
+                    }
+                    get minConfidence() {
+                        return this._minConfidence
+                    }
+                    get maxResults() {
+                        return this._maxResults
+                    }
+                },
+                OH = [new Zj(.738768, .874946), new Zj(2.42204, 2.65704), new Zj(4.30971, 7.04493), new Zj(10.246, 4.59428), new Zj(12.6868, 11.8741)],
+                LH = [new Zj(1.603231, 2.094468), new Zj(6.041143, 7.080126), new Zj(2.882459, 3.518061), new Zj(4.266906, 5.178857), new Zj(9.041765, 10.66308)],
+                PH = [117.001, 114.697, 97.404],
+                $H = e => "number" == typeof e;
+
+            function BH(e) {
+                return ru(() => {
+                    const t = vu(e, Ah(.10000000149011612));
+                    return gu(gh(gc(e, t)), t)
+                })
+            }
+
+            function zH(e, t) {
+                return ru(() => {
+                    let n = Qc(e, [
+                        [0, 0],
+                        [1, 1],
+                        [1, 1],
+                        [0, 0]
+                    ]);
+                    return n = ml(n, t.conv.filters, [1, 1], "valid"), n = gc(n, t.bn.sub), n = vu(n, t.bn.truediv), n = gu(n, t.conv.bias), BH(n)
+                })
+            }
+
+            function UH(e, t) {
+                return ru(() => {
+                    let n = Qc(e, [
+                        [0, 0],
+                        [1, 1],
+                        [1, 1],
+                        [0, 0]
+                    ]);
+                    return n = Nh(n, t.depthwise_filter, t.pointwise_filter, [1, 1], "valid"), n = gu(n, t.bias), BH(n)
+                })
+            }
+
+            function WH(e, t) {
+                const n = GG(e, t);
+                const r = KG(e, t);
+                return {
+                    extractConvParams: n,
+                    extractConvWithBatchNormParams: function (r, a, s) {
+                        return {
+                            conv: n(r, a, 3, s + "/conv"),
+                            bn: function (n, r) {
+                                const a = Yh(e(n)),
+                                    s = Yh(e(n));
+                                return t.push({
+                                    paramPath: r + "/sub"
+                                }, {
+                                    paramPath: r + "/truediv"
+                                }), {
+                                    sub: a,
+                                    truediv: s
+                                }
+                            }(a, s + "/bn")
+                        }
+                    },
+                    extractSeparableConvParams: r
+                }
+            }
+
+            function VH(e, t) {
+                const n = XG(e, t);
+
+                function r(e) {
+                    return {
+                        filters: n(e + "/filters", 4),
+                        bias: n(e + "/bias", 1)
+                    }
+                }
+                return {
+                    extractConvParams: r,
+                    extractConvWithBatchNormParams: function (e) {
+                        return {
+                            conv: r(e + "/conv"),
+                            bn: function (e) {
+                                return {
+                                    sub: n(e + "/sub", 1),
+                                    truediv: n(e + "/truediv", 1)
+                                }
+                            }(e + "/bn")
+                        }
+                    },
+                    extractSeparableConvParams: YG(n)
+                }
+            }
+            var jH = class {
+                    constructor({
+                        inputSize: e,
+                        scoreThreshold: t
+                    } = {}) {
+                        if (this._name = "TinyYolov2Options", this._inputSize = e || 416, this._scoreThreshold = t || .5, "number" != typeof this._inputSize || this._inputSize % 32 != 0) throw new Error(this._name + " - expected inputSize to be a number divisible by 32");
+                        if ("number" != typeof this._scoreThreshold || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) throw new Error(this._name + " - expected scoreThreshold to be a number between 0 and 1")
+                    }
+                    get inputSize() {
+                        return this._inputSize
+                    }
+                    get scoreThreshold() {
+                        return this._scoreThreshold
+                    }
+                },
+                GH = class extends BG {
+                    constructor(e) {
+                        super("TinyYolov2"),
+                            function (e) {
+                                if (!e) throw new Error("invalid config: " + e);
+                                if ("boolean" != typeof e.withSeparableConvs) throw new Error("config.withSeparableConvs has to be a boolean, have: " + e.withSeparableConvs);
+                                if (!$H(e.iouThreshold) || e.iouThreshold < 0 || e.iouThreshold > 1) throw new Error("config.iouThreshold has to be a number between [0, 1], have: " + e.iouThreshold);
+                                if (!Array.isArray(e.classes) || !e.classes.length || !e.classes.every(e => "string" == typeof e)) throw new Error("config.classes has to be an array class names: string[], have: " + JSON.stringify(e.classes));
+                                if (!Array.isArray(e.anchors) || !e.anchors.length || !e.anchors.map(e => e || {}).every(e => $H(e.x) && $H(e.y))) throw new Error("config.anchors has to be an array of { x: number, y: number }, have: " + JSON.stringify(e.anchors));
+                                if (e.meanRgb && (!Array.isArray(e.meanRgb) || 3 !== e.meanRgb.length || !e.meanRgb.every($H))) throw new Error("config.meanRgb has to be an array of shape [number, number, number], have: " + JSON.stringify(e.meanRgb))
+                            }(e), this._config = e
+                    }
+                    get config() {
+                        return this._config
+                    }
+                    get withClassScores() {
+                        return this.config.withClassScores || this.config.classes.length > 1
+                    }
+                    get boxEncodingSize() {
+                        return 5 + (this.withClassScores ? this.config.classes.length : 0)
+                    }
+                    runTinyYolov2(e, t) {
+                        let n = zH(e, t.conv0);
+                        return n = Fc(n, [2, 2], [2, 2], "same"), n = zH(n, t.conv1), n = Fc(n, [2, 2], [2, 2], "same"), n = zH(n, t.conv2), n = Fc(n, [2, 2], [2, 2], "same"), n = zH(n, t.conv3), n = Fc(n, [2, 2], [2, 2], "same"), n = zH(n, t.conv4), n = Fc(n, [2, 2], [2, 2], "same"), n = zH(n, t.conv5), n = Fc(n, [2, 2], [1, 1], "same"), n = zH(n, t.conv6), n = zH(n, t.conv7), VG(n, t.conv8, "valid", !1)
+                    }
+                    runMobilenet(e, t) {
+                        let n = this.config.isFirstLayerConv2d ? BH(VG(e, t.conv0, "valid", !1)) : UH(e, t.conv0);
+                        return n = Fc(n, [2, 2], [2, 2], "same"), n = UH(n, t.conv1), n = Fc(n, [2, 2], [2, 2], "same"), n = UH(n, t.conv2), n = Fc(n, [2, 2], [2, 2], "same"), n = UH(n, t.conv3), n = Fc(n, [2, 2], [2, 2], "same"), n = UH(n, t.conv4), n = Fc(n, [2, 2], [2, 2], "same"), n = UH(n, t.conv5), n = Fc(n, [2, 2], [1, 1], "same"), n = t.conv6 ? UH(n, t.conv6) : n, n = t.conv7 ? UH(n, t.conv7) : n, VG(n, t.conv8, "valid", !1)
+                    }
+                    forwardInput(e, t) {
+                        const {
+                            params: n
+                        } = this;
+                        if (!n) throw new Error("TinyYolov2 - load model before inference");
+                        return ru(() => {
+                            let r = Ai(e.toBatchTensor(t, !1), "float32");
+                            return r = this.config.meanRgb ? aG(r, this.config.meanRgb) : r, r = r.div(255), this.config.withSeparableConvs ? this.runMobilenet(r, n) : this.runTinyYolov2(r, n)
+                        })
+                    }
+                    async forward(e, t) {
+                        return this.forwardInput(await DG(e), t)
+                    }
+                    async detect(e, t = {}) {
+                        const {
+                            inputSize: n,
+                            scoreThreshold: r
+                        } = new jH(t), a = await DG(e), s = await this.forwardInput(a, n), i = ru(() => ap(s)[0].expandDims()), o = {
+                            width: a.getInputWidth(0),
+                            height: a.getInputHeight(0)
+                        }, u = await this.extractBoxes(i, a.getReshapedInputDimensions(0), r);
+                        s.dispose(), i.dispose();
+                        const l = u.map(e => e.box),
+                            c = u.map(e => e.score),
+                            h = u.map(e => e.classScore),
+                            p = u.map(e => this.config.classes[e.label]);
+                        return function (e, t, n, r = !0) {
+                            let a = t.map((e, t) => ({
+                                score: e,
+                                boxIndex: t
+                            })).sort((e, t) => e.score - t.score).map(e => e.boxIndex);
+                            const s = [];
+                            for (; a.length > 0;) {
+                                const t = a.pop();
+                                s.push(t);
+                                const i = a,
+                                    o = [];
+                                for (let n = 0; n < i.length; n++) {
+                                    const a = i[n],
+                                        s = e[t],
+                                        u = e[a];
+                                    o.push(rG(s, u, r))
+                                }
+                                a = a.filter((e, t) => o[t] <= n)
+                            }
+                            return s
+                        }(l.map(e => e.rescale(n)), c, this.config.iouThreshold, !0).map(e => new tG(c[e], h[e], p[e], l[e], o))
+                    }
+                    getDefaultModelName() {
+                        return ""
+                    }
+                    extractParamsFromWeightMap(e) {
+                        return function (e, t) {
+                            const n = [],
+                                {
+                                    extractConvParams: r,
+                                    extractConvWithBatchNormParams: a,
+                                    extractSeparableConvParams: s
+                                } = VH(e, n);
+                            let i;
+                            if (t.withSeparableConvs) {
+                                const e = t.filterSizes && t.filterSizes.length || 9;
+                                i = {
+                                    conv0: t.isFirstLayerConv2d ? r("conv0") : s("conv0"),
+                                    conv1: s("conv1"),
+                                    conv2: s("conv2"),
+                                    conv3: s("conv3"),
+                                    conv4: s("conv4"),
+                                    conv5: s("conv5"),
+                                    conv6: e > 7 ? s("conv6") : void 0,
+                                    conv7: e > 8 ? s("conv7") : void 0,
+                                    conv8: r("conv8")
+                                }
+                            } else i = {
+                                conv0: a("conv0"),
+                                conv1: a("conv1"),
+                                conv2: a("conv2"),
+                                conv3: a("conv3"),
+                                conv4: a("conv4"),
+                                conv5: a("conv5"),
+                                conv6: a("conv6"),
+                                conv7: a("conv7"),
+                                conv8: r("conv8")
+                            };
+                            return jG(e, n), {
+                                params: i,
+                                paramMappings: n
+                            }
+                        }(e, this.config)
+                    }
+                    extractParams(e) {
+                        const t = this.config.filterSizes || GH.DEFAULT_FILTER_SIZES,
+                            n = t ? t.length : void 0;
+                        if (7 !== n && 8 !== n && 9 !== n) throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${n} filterSizes in config`);
+                        return function (e, t, n, r) {
+                            const {
+                                extractWeights: a,
+                                getRemainingWeights: s
+                            } = QG(e), i = [], {
+                                extractConvParams: o,
+                                extractConvWithBatchNormParams: u,
+                                extractSeparableConvParams: l
+                            } = WH(a, i);
+                            let c;
+                            if (t.withSeparableConvs) {
+                                const [e, a, s, i, u, h, p, d, f] = r;
+                                c = {
+                                    conv0: t.isFirstLayerConv2d ? o(e, a, 3, "conv0") : l(e, a, "conv0"),
+                                    conv1: l(a, s, "conv1"),
+                                    conv2: l(s, i, "conv2"),
+                                    conv3: l(i, u, "conv3"),
+                                    conv4: l(u, h, "conv4"),
+                                    conv5: l(h, p, "conv5"),
+                                    conv6: d ? l(p, d, "conv6") : void 0,
+                                    conv7: f ? l(d, f, "conv7") : void 0,
+                                    conv8: o(f || d || p, 5 * n, 1, "conv8")
+                                }
+                            } else {
+                                const [e, t, a, s, i, l, h, p, d] = r;
+                                c = {
+                                    conv0: u(e, t, "conv0"),
+                                    conv1: u(t, a, "conv1"),
+                                    conv2: u(a, s, "conv2"),
+                                    conv3: u(s, i, "conv3"),
+                                    conv4: u(i, l, "conv4"),
+                                    conv5: u(l, h, "conv5"),
+                                    conv6: u(h, p, "conv6"),
+                                    conv7: u(p, d, "conv7"),
+                                    conv8: o(d, 5 * n, 1, "conv8")
+                                }
+                            }
+                            if (0 !== s().length) throw new Error("weights remaing after extract: " + s().length);
+                            return {
+                                params: c,
+                                paramMappings: i
+                            }
+                        }(e, this.config, this.boxEncodingSize, t)
+                    }
+                    async extractBoxes(e, t, n) {
+                        const {
+                            width: r,
+                            height: a
+                        } = t, s = Math.max(r, a), i = s / r, o = s / a, u = e.shape[1], l = this.config.anchors.length, [c, h, p] = ru(() => {
+                            const t = e.reshape([u, u, l, this.boxEncodingSize]);
+                            return [t.slice([0, 0, 0, 0], [u, u, l, 4]), t.slice([0, 0, 0, 4], [u, u, l, 1]), this.withClassScores ? Lh(t.slice([0, 0, 0, 5], [u, u, l, this.config.classes.length]), 3) : Ah(0)]
+                        }), d = [], f = await h.array(), m = await c.array();
+                        for (let e = 0; e < u; e++)
+                            for (let t = 0; t < u; t++)
+                                for (let r = 0; r < l; r++) {
+                                    const a = sG(f[e][t][r][0]);
+                                    if (!n || a > n) {
+                                        const n = (t + sG(m[e][t][r][0])) / u * i,
+                                            s = (e + sG(m[e][t][r][1])) / u * o,
+                                            l = Math.exp(m[e][t][r][2]) * this.config.anchors[r].x / u * i,
+                                            c = Math.exp(m[e][t][r][3]) * this.config.anchors[r].y / u * o,
+                                            h = n - l / 2,
+                                            f = s - c / 2,
+                                            g = {
+                                                row: e,
+                                                col: t,
+                                                anchor: r
+                                            },
+                                            {
+                                                classScore: y,
+                                                label: b
+                                            } = this.withClassScores ? await this.extractPredictedClass(p, g) : {
+                                                classScore: 1,
+                                                label: 0
+                                            };
+                                        d.push({
+                                            box: new eG(h, f, h + l, f + c),
+                                            score: a,
+                                            classScore: a * y,
+                                            label: b,
+                                            ...g
+                                        })
+                                    }
+                                }
+                        return c.dispose(), h.dispose(), p.dispose(), d
+                    }
+                    async extractPredictedClass(e, t) {
+                        const {
+                            row: n,
+                            col: r,
+                            anchor: a
+                        } = t, s = await e.array();
+                        return Array(this.config.classes.length).fill(0).map((e, t) => s[n][r][a][t]).map((e, t) => ({
+                            classScore: e,
+                            label: t
+                        })).reduce((e, t) => e.classScore > t.classScore ? e : t)
+                    }
+                },
+                HH = GH;
+            HH.DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];
+            var qH = class extends jH {
+                    constructor() {
+                        super(...arguments), this._name = "TinyFaceDetectorOptions"
+                    }
+                },
+                KH = class {
+                    async then(e) {
+                        return e(await this.run())
+                    }
+                    async run() {
+                        throw new Error("ComposableTask - run is not implemented")
+                    }
+                };
+            async function YH(e, t, n, r, a = (({
+                alignedRect: e
+            }) => e)) {
+                const s = e.map(e => uH(e) ? a(e) : e.detection),
+                    i = r || (t instanceof Ya ? await LG(t, s) : await OG(t, s)),
+                    o = await n(i);
+                return i.forEach(e => e instanceof Ya && e.dispose()), o
+            }
+            var XH = [new Zj(1.603231, 2.094468), new Zj(6.041143, 7.080126), new Zj(2.882459, 3.518061), new Zj(4.266906, 5.178857), new Zj(9.041765, 10.66308)],
+                QH = [117.001, 114.697, 97.404],
+                ZH = {
+                    ssdMobilenetv1: new class extends BG {
+                        constructor() {
+                            super("SsdMobilenetv1")
+                        }
+                        forwardInput(e) {
+                            const {
+                                params: t
+                            } = this;
+                            if (!t) throw new Error("SsdMobilenetv1 - load model before inference");
+                            return ru(() => {
+                                const n = Ai(e.toBatchTensor(512, !1), "float32"),
+                                    r = RH(gc(bu(n, 127.5), 1), t.mobilenetv1),
+                                    {
+                                        boxPredictions: a,
+                                        classPredictions: s
+                                    } = function (e, t, n) {
+                                        return ru(() => {
+                                            const r = TH(e, n.conv_0, [1, 1]),
+                                                a = TH(r, n.conv_1, [2, 2]),
+                                                s = TH(a, n.conv_2, [1, 1]),
+                                                i = TH(s, n.conv_3, [2, 2]),
+                                                o = TH(i, n.conv_4, [1, 1]),
+                                                u = TH(o, n.conv_5, [2, 2]),
+                                                l = TH(u, n.conv_6, [1, 1]),
+                                                c = TH(l, n.conv_7, [2, 2]),
+                                                h = MH(t, n.box_predictor_0),
+                                                p = MH(e, n.box_predictor_1),
+                                                d = MH(a, n.box_predictor_2),
+                                                f = MH(i, n.box_predictor_3),
+                                                m = MH(u, n.box_predictor_4),
+                                                g = MH(c, n.box_predictor_5);
+                                            return {
+                                                boxPredictions: Xu([h.boxPredictionEncoding, p.boxPredictionEncoding, d.boxPredictionEncoding, f.boxPredictionEncoding, m.boxPredictionEncoding, g.boxPredictionEncoding], 1),
+                                                classPredictions: Xu([h.classPrediction, p.classPrediction, d.classPrediction, f.classPrediction, m.classPrediction, g.classPrediction], 1)
+                                            }
+                                        })
+                                    }(r.out, r.conv11, t.prediction_layer);
+                                return function (e, t, n) {
+                                    return ru(() => {
+                                        const r = e.shape[0];
+                                        let a = FH(qu(Ul(n.extra_dim, [r, 1, 1]), [-1, 4]), qu(e, [-1, 4]));
+                                        a = qu(a, [r, a.shape[0] / r, 4]);
+                                        const s = Qu(Zu(t, [0, 0, 1], [-1, -1, -1]));
+                                        let i = Zu(s, [0, 0, 0], [-1, -1, 1]);
+                                        i = qu(i, [r, i.shape[1]]);
+                                        return {
+                                            boxes: ap(a),
+                                            scores: ap(i)
+                                        }
+                                    })
+                                }(a, s, t.output_layer)
+                            })
+                        }
+                        async forward(e) {
+                            return this.forwardInput(await DG(e))
+                        }
+                        async locateFaces(e, t = {}) {
+                            const {
+                                maxResults: n,
+                                minConfidence: r
+                            } = new DH(t), a = await DG(e), {
+                                boxes: s,
+                                scores: i
+                            } = this.forwardInput(a), o = s[0], u = i[0];
+                            for (let e = 1; e < s.length; e++) s[e].dispose(), i[e].dispose();
+                            const l = Array.from(u.dataSync()),
+                                c = function (e, t, n, r, a) {
+                                    const s = e.shape[0],
+                                        i = Math.min(n, s),
+                                        o = t.map((e, t) => ({
+                                            score: e,
+                                            boxIndex: t
+                                        })).filter(e => e.score > a).sort((e, t) => t.score - e.score),
+                                        u = [];
+                                    return o.forEach(t => {
+                                        if (u.length >= i) return;
+                                        const n = t.score;
+                                        for (let n = u.length - 1; n >= 0; --n) {
+                                            const s = _H(e, t.boxIndex, u[n]);
+                                            if (0 !== s && (t.score *= s <= r ? 1 : 0, t.score <= a)) break
+                                        }
+                                        n === t.score && u.push(t.boxIndex)
+                                    }), u
+                                }(o, l, n, .5, r),
+                                h = a.getReshapedInputDimensions(0),
+                                p = a.inputSize,
+                                d = p / h.width,
+                                f = p / h.height,
+                                m = o.arraySync(),
+                                g = c.map(e => {
+                                    const [t, n] = [Math.max(0, m[e][0]), Math.min(1, m[e][2])].map(e => e * f), [r, s] = [Math.max(0, m[e][1]), Math.min(1, m[e][3])].map(e => e * d);
+                                    return new nG(l[e], new oG(r, t, s - r, n - t), {
+                                        height: a.getInputHeight(0),
+                                        width: a.getInputWidth(0)
+                                    })
+                                });
+                            return o.dispose(), u.dispose(), g
+                        }
+                        getDefaultModelName() {
+                            return "ssd_mobilenetv1_model"
+                        }
+                        extractParamsFromWeightMap(e) {
+                            return CH(e)
+                        }
+                        extractParams(e) {
+                            return function (e) {
+                                const t = [],
+                                    {
+                                        extractWeights: n,
+                                        getRemainingWeights: r
+                                    } = QG(e),
+                                    {
+                                        extractMobilenetV1Params: a,
+                                        extractPredictionLayerParams: s
+                                    } = NH(n, t),
+                                    i = a(),
+                                    o = s(),
+                                    u = {
+                                        extra_dim: to(n(20472), [1, 5118, 4])
+                                    };
+                                if (t.push({
+                                        paramPath: "output_layer/extra_dim"
+                                    }), 0 !== r().length) throw new Error("weights remaing after extract: " + r().length);
+                                return {
+                                    params: {
+                                        mobilenetv1: i,
+                                        prediction_layer: o,
+                                        output_layer: u
+                                    },
+                                    paramMappings: t
+                                }
+                            }(e)
+                        }
+                    },
+                    tinyFaceDetector: new class extends HH {
+                        constructor() {
+                            super({
+                                withSeparableConvs: !0,
+                                iouThreshold: .4,
+                                classes: ["face"],
+                                anchors: XH,
+                                meanRgb: QH,
+                                isFirstLayerConv2d: !0,
+                                filterSizes: [3, 16, 32, 64, 128, 256, 512]
+                            })
+                        }
+                        get anchors() {
+                            return this.config.anchors
+                        }
+                        async locateFaces(e, t) {
+                            return (await this.detect(e, t)).map(e => new nG(e.score, e.relativeBox, {
+                                width: e.imageWidth,
+                                height: e.imageHeight
+                            }))
+                        }
+                        getDefaultModelName() {
+                            return "tiny_face_detector_model"
+                        }
+                        extractParamsFromWeightMap(e) {
+                            return super.extractParamsFromWeightMap(e)
+                        }
+                    },
+                    tinyYolov2: new class extends HH {
+                        constructor(e = !0) {
+                            super({
+                                withSeparableConvs: e,
+                                iouThreshold: .4,
+                                classes: ["face"],
+                                ...e ? {
+                                    anchors: LH,
+                                    meanRgb: PH
+                                } : {
+                                    anchors: OH,
+                                    withClassScores: !0
+                                }
+                            })
+                        }
+                        get withSeparableConvs() {
+                            return this.config.withSeparableConvs
+                        }
+                        get anchors() {
+                            return this.config.anchors
+                        }
+                        async locateFaces(e, t) {
+                            return (await this.detect(e, t)).map(e => new nG(e.score, e.relativeBox, {
+                                width: e.imageWidth,
+                                height: e.imageHeight
+                            }))
+                        }
+                        getDefaultModelName() {
+                            return this.withSeparableConvs ? "tiny_yolov2_separable_conv_model" : "tiny_yolov2_model"
+                        }
+                        extractParamsFromWeightMap(e) {
+                            return super.extractParamsFromWeightMap(e)
+                        }
+                    },
+                    faceLandmark68Net: new class extends bH {
+                        constructor(e = new tH) {
+                            super("FaceLandmark68Net", e)
+                        }
+                        getDefaultModelName() {
+                            return "face_landmark_68_model"
+                        }
+                        getClassifierChannelsIn() {
+                            return 256
+                        }
+                    },
+                    faceLandmark68TinyNet: new class extends bH {
+                        constructor(e = new vH) {
+                            super("FaceLandmark68TinyNet", e)
+                        }
+                        getDefaultModelName() {
+                            return "face_landmark_68_tiny_model"
+                        }
+                        getClassifierChannelsIn() {
+                            return 128
+                        }
+                    },
+                    faceRecognitionNet: new class extends BG {
+                        constructor() {
+                            super("FaceRecognitionNet")
+                        }
+                        forwardInput(e) {
+                            const {
+                                params: t
+                            } = this;
+                            if (!t) throw new Error("FaceRecognitionNet - load model before inference");
+                            return ru(() => {
+                                let n = kH(aG(Ai(e.toBatchTensor(150, !0), "float32"), [122.782, 117.001, 104.298]).div(255), t.conv32_down);
+                                n = Fc(n, 3, 2, "valid"), n = AH(n, t.conv32_1), n = AH(n, t.conv32_2), n = AH(n, t.conv32_3), n = EH(n, t.conv64_down), n = AH(n, t.conv64_1), n = AH(n, t.conv64_2), n = AH(n, t.conv64_3), n = EH(n, t.conv128_down), n = AH(n, t.conv128_1), n = AH(n, t.conv128_2), n = EH(n, t.conv256_down), n = AH(n, t.conv256_1), n = AH(n, t.conv256_2), n = EH(n, t.conv256_down_out);
+                                const r = n.mean([1, 2]);
+                                return Gi(r, t.fc)
+                            })
+                        }
+                        async forward(e) {
+                            return this.forwardInput(await DG(e))
+                        }
+                        async computeFaceDescriptor(e) {
+                            var t;
+                            if (null == (t = null == e ? void 0 : e.shape) ? void 0 : t.some(e => e <= 0)) return new Float32Array(128);
+                            const n = await DG(e),
+                                r = ru(() => ap(this.forwardInput(n))),
+                                a = await Promise.all(r.map(e => e.data()));
+                            return r.forEach(e => e.dispose()), n.isBatchInput ? a : a[0]
+                        }
+                        getDefaultModelName() {
+                            return "face_recognition_model"
+                        }
+                        extractParamsFromWeightMap(e) {
+                            return function (e) {
+                                const t = [],
+                                    {
+                                        extractConvLayerParams: n,
+                                        extractResidualLayerParams: r
+                                    } = IH(e, t),
+                                    a = n("conv32_down"),
+                                    s = r("conv32_1"),
+                                    i = r("conv32_2"),
+                                    o = r("conv32_3"),
+                                    u = r("conv64_down"),
+                                    l = r("conv64_1"),
+                                    c = r("conv64_2"),
+                                    h = r("conv64_3"),
+                                    p = r("conv128_down"),
+                                    d = r("conv128_1"),
+                                    f = r("conv128_2"),
+                                    m = r("conv256_down"),
+                                    g = r("conv256_1"),
+                                    y = r("conv256_2"),
+                                    b = r("conv256_down_out"),
+                                    {
+                                        fc: v
+                                    } = e;
+                                if (t.push({
+                                        originalPath: "fc",
+                                        paramPath: "fc"
+                                    }), !zj(v)) throw new Error("expected weightMap[fc] to be a Tensor2D, instead have " + v);
+                                const x = {
+                                    conv32_down: a,
+                                    conv32_1: s,
+                                    conv32_2: i,
+                                    conv32_3: o,
+                                    conv64_down: u,
+                                    conv64_1: l,
+                                    conv64_2: c,
+                                    conv64_3: h,
+                                    conv128_down: p,
+                                    conv128_1: d,
+                                    conv128_2: f,
+                                    conv256_down: m,
+                                    conv256_1: g,
+                                    conv256_2: y,
+                                    conv256_down_out: b,
+                                    fc: v
+                                };
+                                return jG(e, t), {
+                                    params: x,
+                                    paramMappings: t
+                                }
+                            }(e)
+                        }
+                        extractParams(e) {
+                            return function (e) {
+                                const {
+                                    extractWeights: t,
+                                    getRemainingWeights: n
+                                } = QG(e), r = [], {
+                                    extractConvLayerParams: a,
+                                    extractResidualLayerParams: s
+                                } = SH(t, r), i = a(4704, 32, 7, "conv32_down"), o = s(9216, 32, 3, "conv32_1"), u = s(9216, 32, 3, "conv32_2"), l = s(9216, 32, 3, "conv32_3"), c = s(36864, 64, 3, "conv64_down", !0), h = s(36864, 64, 3, "conv64_1"), p = s(36864, 64, 3, "conv64_2"), d = s(36864, 64, 3, "conv64_3"), f = s(147456, 128, 3, "conv128_down", !0), m = s(147456, 128, 3, "conv128_1"), g = s(147456, 128, 3, "conv128_2"), y = s(589824, 256, 3, "conv256_down", !0), b = s(589824, 256, 3, "conv256_1"), v = s(589824, 256, 3, "conv256_2"), x = s(589824, 256, 3, "conv256_down_out"), w = ru(() => qi(Xh(t(32768), [128, 256]), [1, 0]));
+                                if (r.push({
+                                        paramPath: "fc"
+                                    }), 0 !== n().length) throw new Error("weights remaing after extract: " + n().length);
+                                return {
+                                    params: {
+                                        conv32_down: i,
+                                        conv32_1: o,
+                                        conv32_2: u,
+                                        conv32_3: l,
+                                        conv64_down: c,
+                                        conv64_1: h,
+                                        conv64_2: p,
+                                        conv64_3: d,
+                                        conv128_down: f,
+                                        conv128_1: m,
+                                        conv128_2: g,
+                                        conv256_down: y,
+                                        conv256_1: b,
+                                        conv256_2: v,
+                                        conv256_down_out: x,
+                                        fc: w
+                                    },
+                                    paramMappings: r
+                                }
+                            }(e)
+                        }
+                    },
+                    faceExpressionNet: new class extends aH {
+                        constructor(e = new tH) {
+                            super("FaceExpressionNet", e)
+                        }
+                        forwardInput(e) {
+                            return ru(() => Lh(this.runNet(e)))
+                        }
+                        async forward(e) {
+                            return this.forwardInput(await DG(e))
+                        }
+                        async predictExpressions(e) {
+                            const t = await DG(e),
+                                n = await this.forwardInput(t),
+                                r = await Promise.all(ap(n).map(async e => {
+                                    const t = e.dataSync();
+                                    return e.dispose(), t
+                                }));
+                            n.dispose();
+                            const a = r.map(e => new iH(e));
+                            return t.isBatchInput ? a : a[0]
+                        }
+                        getDefaultModelName() {
+                            return "face_expression_model"
+                        }
+                        getClassifierChannelsIn() {
+                            return 256
+                        }
+                        getClassifierChannelsOut() {
+                            return 7
+                        }
+                    },
+                    ageGenderNet: new class extends BG {
+                        constructor(e = new yH(2)) {
+                            super("AgeGenderNet"), this._faceFeatureExtractor = e
+                        }
+                        get faceFeatureExtractor() {
+                            return this._faceFeatureExtractor
+                        }
+                        runNet(e) {
+                            const {
+                                params: t
+                            } = this;
+                            if (!t) throw new Error(this._name + " - load model before inference");
+                            return ru(() => {
+                                const n = e instanceof MG ? this.faceFeatureExtractor.forwardInput(e) : e,
+                                    r = Ku(n, [7, 7], [2, 2], "valid").as2D(n.shape[0], -1);
+                                return {
+                                    age: nH(r, t.fc.age).as1D(),
+                                    gender: nH(r, t.fc.gender)
+                                }
+                            })
+                        }
+                        forwardInput(e) {
+                            return ru(() => {
+                                const {
+                                    age: t,
+                                    gender: n
+                                } = this.runNet(e);
+                                return {
+                                    age: t,
+                                    gender: Lh(n)
+                                }
+                            })
+                        }
+                        async forward(e) {
+                            return this.forwardInput(await DG(e))
+                        }
+                        async predictAgeAndGender(e) {
+                            const t = await DG(e),
+                                n = await this.forwardInput(t),
+                                r = ap(n.age),
+                                a = ap(n.gender),
+                                s = r.map((e, t) => ({
+                                    ageTensor: e,
+                                    genderTensor: a[t]
+                                })),
+                                i = await Promise.all(s.map(async ({
+                                    ageTensor: e,
+                                    genderTensor: t
+                                }) => {
+                                    const n = e.dataSync()[0],
+                                        r = t.dataSync()[0],
+                                        a = r > .5,
+                                        s = a ? "male" : "female",
+                                        i = a ? r : 1 - r;
+                                    return e.dispose(), t.dispose(), {
+                                        age: n,
+                                        gender: s,
+                                        genderProbability: i
+                                    }
+                                }));
+                            return n.age.dispose(), n.gender.dispose(), t.isBatchInput ? i : i[0]
+                        }
+                        getDefaultModelName() {
+                            return "age_gender_model"
+                        }
+                        dispose(e = !0) {
+                            this.faceFeatureExtractor.dispose(e), super.dispose(e)
+                        }
+                        loadClassifierParams(e) {
+                            const {
+                                params: t,
+                                paramMappings: n
+                            } = this.extractClassifierParams(e);
+                            this._params = t, this._paramMappings = n
+                        }
+                        extractClassifierParams(e) {
+                            return function (e) {
+                                const t = [],
+                                    {
+                                        extractWeights: n,
+                                        getRemainingWeights: r
+                                    } = QG(e),
+                                    a = HG(n, t),
+                                    s = a(512, 1, "fc/age"),
+                                    i = a(512, 2, "fc/gender");
+                                if (0 !== r().length) throw new Error("weights remaing after extract: " + r().length);
+                                return {
+                                    paramMappings: t,
+                                    params: {
+                                        fc: {
+                                            age: s,
+                                            gender: i
+                                        }
+                                    }
+                                }
+                            }(e)
+                        }
+                        extractParamsFromWeightMap(e) {
+                            const {
+                                featureExtractorMap: t,
+                                classifierMap: n
+                            } = rH(e);
+                            return this.faceFeatureExtractor.loadFromWeightMap(t),
+                                function (e) {
+                                    const t = [],
+                                        n = XG(e, t);
+
+                                    function r(e) {
+                                        return {
+                                            weights: n(e + "/weights", 2),
+                                            bias: n(e + "/bias", 1)
+                                        }
+                                    }
+                                    const a = {
+                                        fc: {
+                                            age: r("fc/age"),
+                                            gender: r("fc/gender")
+                                        }
+                                    };
+                                    return jG(e, t), {
+                                        params: a,
+                                        paramMappings: t
+                                    }
+                                }(n)
+                        }
+                        extractParams(e) {
+                            const t = e.slice(0, e.length - 1539),
+                                n = e.slice(e.length - 1539);
+                            return this.faceFeatureExtractor.extractWeights(t), this.extractClassifierParams(n)
+                        }
+                    }
+                },
+                JH = class extends KH {
+                    constructor(e, t, n) {
+                        super(), this.parentTask = e, this.input = t, this.extractedFaces = n
+                    }
+                },
+                eq = class extends JH {
+                    async run() {
+                        const e = await this.parentTask,
+                            t = await YH(e, this.input, async e => Promise.all(e.map(e => ZH.faceExpressionNet.predictExpressions(e))), this.extractedFaces);
+                        return e.map((e, n) => {
+                            return r = e, a = t[n], {
+                                ...r,
+                                expressions: a
+                            };
+                            var r, a
+                        })
+                    }
+                    withAgeAndGender() {
+                        return new rq(this, this.input)
+                    }
+                },
+                tq = class extends eq {
+                    withAgeAndGender() {
+                        return new aq(this, this.input)
+                    }
+                    withFaceDescriptors() {
+                        return new iq(this, this.input)
+                    }
+                },
+                nq = class extends KH {
+                    constructor(e, t, n) {
+                        super(), this.parentTask = e, this.input = t, this.extractedFaces = n
+                    }
+                },
+                rq = class extends nq {
+                    async run() {
+                        const e = await this.parentTask,
+                            t = await YH(e, this.input, async e => Promise.all(e.map(e => ZH.ageGenderNet.predictAgeAndGender(e))), this.extractedFaces);
+                        return e.map((e, n) => {
+                            const {
+                                age: r,
+                                gender: a,
+                                genderProbability: s
+                            } = t[n];
+                            return function (e, t) {
+                                return {
+                                    ...e,
+                                    ...{
+                                        age: t
+                                    }
+                                }
+                            }(function (e, t, n) {
+                                return {
+                                    ...e,
+                                    ...{
+                                        gender: t,
+                                        genderProbability: n
+                                    }
+                                }
+                            }(e, a, s), r)
+                        })
+                    }
+                    withFaceExpressions() {
+                        return new eq(this, this.input)
+                    }
+                },
+                aq = class extends rq {
+                    withFaceExpressions() {
+                        return new tq(this, this.input)
+                    }
+                    withFaceDescriptors() {
+                        return new iq(this, this.input)
+                    }
+                },
+                sq = class extends KH {
+                    constructor(e, t) {
+                        super(), this.parentTask = e, this.input = t
+                    }
+                },
+                iq = class extends sq {
+                    async run() {
+                        const e = await this.parentTask;
+                        return (await YH(e, this.input, e => Promise.all(e.map(e => ZH.faceRecognitionNet.computeFaceDescriptor(e))), null, e => e.landmarks.align(null, {
+                            useDlibAlignment: !0
+                        }))).map((t, n) => function (e, t) {
+                            return {
+                                ...e,
+                                ...{
+                                    descriptor: t
+                                }
+                            }
+                        }(e[n], t))
+                    }
+                    withFaceExpressions() {
+                        return new tq(this, this.input)
+                    }
+                    withAgeAndGender() {
+                        return new aq(this, this.input)
+                    }
+                },
+                oq = class extends KH {
+                    constructor(e, t, n) {
+                        super(), this.parentTask = e, this.input = t, this.useTinyLandmarkNet = n
+                    }
+                    get landmarkNet() {
+                        return this.useTinyLandmarkNet ? ZH.faceLandmark68TinyNet : ZH.faceLandmark68Net
+                    }
+                },
+                uq = class extends oq {
+                    async run() {
+                        const e = await this.parentTask,
+                            t = e.map(e => e.detection),
+                            n = this.input instanceof Ya ? await LG(this.input, t) : await OG(this.input, t),
+                            r = await Promise.all(n.map(e => this.landmarkNet.detectLandmarks(e)));
+                        return n.forEach(e => e instanceof Ya && e.dispose()), e.map((e, t) => lH(e, r[t]))
+                    }
+                    withFaceExpressions() {
+                        return new tq(this, this.input)
+                    }
+                    withAgeAndGender() {
+                        return new aq(this, this.input)
+                    }
+                    withFaceDescriptors() {
+                        return new iq(this, this.input)
+                    }
+                },
+                lq = class extends KH {
+                    constructor(e, t = new DH) {
+                        super(), this.input = e, this.options = t
+                    }
+                },
+                cq = class extends lq {
+                    async run() {
+                        const {
+                            input: e,
+                            options: t
+                        } = this;
+                        let n;
+                        if (t instanceof qH) n = ZH.tinyFaceDetector.locateFaces(e, t);
+                        else if (t instanceof DH) n = ZH.ssdMobilenetv1.locateFaces(e, t);
+                        else {
+                            if (!(t instanceof jH)) throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");
+                            n = ZH.tinyYolov2.locateFaces(e, t)
+                        }
+                        return n
+                    }
+                    runAndExtendWithFaceDetections() {
+                        return new Promise((e, t) => {
+                            this.run().then(t => e(t.map(e => hG({}, e)))).catch(e => t(e))
+                        })
+                    }
+                    withFaceLandmarks(e = !1) {
+                        return new uq(this.runAndExtendWithFaceDetections(), this.input, e)
+                    }
+                    withFaceExpressions() {
+                        return new eq(this.runAndExtendWithFaceDetections(), this.input)
+                    }
+                    withAgeAndGender() {
+                        return new rq(this.runAndExtendWithFaceDetections(), this.input)
+                    }
+                };
+            class hq {
+                constructor() {
+                    this.FACE_ANGLE_TOO_LARGE = !1, this.FACE_CLOSE_TO_BORDER = !1, this.FACE_NOT_FOUND = !1, this.FACE_TOO_SMALL = !1, this.PROBABILITY_TOO_SMALL = !1, this.TOO_MANY_FACES = !1
+                }
+            }
+
+            function pq(e, t) {
+                if (!(e instanceof t)) throw new TypeError("Cannot call a class as a function")
+            }
+
+            function dq(e, t) {
+                for (var n = 0; n < t.length; n++) {
+                    var r = t[n];
+                    r.enumerable = r.enumerable || !1, r.configurable = !0, "value" in r && (r.writable = !0), Object.defineProperty(e, r.key, r)
+                }
+            }
+            /**
+             * KalmanFilter
+             * @class
+             * @author Wouter Bulten
+             * @see {@link http://github.com/wouterbulten/kalmanjs}
+             * @version Version: 1.0.0-beta
+             * @copyright Copyright 2015-2018 Wouter Bulten
+             * @license MIT License
+             * @preserve
+             */
+            var fq = function () {
+                function e() {
+                    var t = arguments.length > 0 && void 0 !== arguments[0] ? arguments[0] : {},
+                        n = t.R,
+                        r = void 0 === n ? 1 : n,
+                        a = t.Q,
+                        s = void 0 === a ? 1 : a,
+                        i = t.A,
+                        o = void 0 === i ? 1 : i,
+                        u = t.B,
+                        l = void 0 === u ? 0 : u,
+                        c = t.C,
+                        h = void 0 === c ? 1 : c;
+                    pq(this, e), this.R = r, this.Q = s, this.A = o, this.C = h, this.B = l, this.cov = NaN, this.x = NaN
+                }
+                var t, n, r;
+                return t = e, (n = [{
+                    key: "filter",
+                    value: function (e) {
+                        var t = arguments.length > 1 && void 0 !== arguments[1] ? arguments[1] : 0;
+                        if (isNaN(this.x)) this.x = 1 / this.C * e, this.cov = 1 / this.C * this.Q * (1 / this.C);
+                        else {
+                            var n = this.predict(t),
+                                r = this.uncertainty(),
+                                a = r * this.C * (1 / (this.C * r * this.C + this.Q));
+                            this.x = n + a * (e - this.C * n), this.cov = r - a * this.C * r
+                        }
+                        return this.x
+                    }
+                }, {
+                    key: "predict",
+                    value: function () {
+                        var e = arguments.length > 0 && void 0 !== arguments[0] ? arguments[0] : 0;
+                        return this.A * this.x + this.B * e
+                    }
+                }, {
+                    key: "uncertainty",
+                    value: function () {
+                        return this.A * this.cov * this.A + this.R
+                    }
+                }, {
+                    key: "lastMeasurement",
+                    value: function () {
+                        return this.x
+                    }
+                }, {
+                    key: "setMeasurementNoise",
+                    value: function (e) {
+                        this.Q = e
+                    }
+                }, {
+                    key: "setProcessNoise",
+                    value: function (e) {
+                        this.R = e
+                    }
+                }]) && dq(t.prototype, n), r && dq(t, r), e
+            }();
+            const mq = window.cv,
+                gq = {
+                    R: 1,
+                    Q: 50
+                },
+                yq = (e, t) => bq[e].filter(t),
+                bq = {
+                    nosetipx: new fq(gq),
+                    nosetipy: new fq(gq),
+                    frontx: new fq(gq),
+                    fronty: new fq(gq),
+                    topx: new fq(gq),
+                    topy: new fq(gq),
+                    sidex: new fq(gq),
+                    sidey: new fq(gq)
+                };
+            class vq {
+                constructor(e, t) {
+                    this.selectedOpenCvMethod = "0", this.noseTip = e[30], this.bottomNose = e[33], this.lefteyeleftcorner = e[36], this.lefteyerightcorner = e[39], this.righteyerightcorner = e[45], this.righteyeleftcorner = e[42], this.leftmouth = e[48], this.rightmouth = e[54], this.leftnostril = e[31], this.rightnostril = e[35], this.focal_length = t.width, this.center = [t.width / 2, t.height / 2]
+                }
+                estimateHeadPose() {
+                    const e = mq.matFromArray(10, 2, mq.CV_64F, [this.noseTip._x, this.noseTip._y, this.bottomNose._x, this.bottomNose._y, this.leftnostril._x, this.leftnostril._y, this.rightnostril._x, this.rightnostril._y, this.lefteyeleftcorner._x, this.lefteyeleftcorner._y, this.lefteyerightcorner._x, this.lefteyerightcorner._y, this.righteyerightcorner._x, this.righteyerightcorner._y, this.righteyeleftcorner._x, this.righteyeleftcorner._y, this.leftmouth._x, this.leftmouth._y, this.rightmouth._x, this.rightmouth._y]),
+                        t = mq.matFromArray(10, 3, mq.CV_64F, [0, 0, 0, 0, -60, -78, -67, -58, -100, 67, -58, -100, -262, 168, -240, -115, 170, -210, 262, 168, -240, 115, 170, -210, -148, -192, -181, 148, -192, -181]),
+                        n = mq.matFromArray(3, 3, mq.CV_64F, [this.focal_length, 0, this.center[0], 0, this.focal_length, this.center[1], 0, 0, 1]),
+                        r = mq.Mat.zeros(4, 1, mq.CV_64F),
+                        a = new mq.Mat,
+                        s = new mq.Mat,
+                        i = new mq.Mat;
+                    mq.solvePnPRansac(t, e, n, r, a, s, !1, 100, 8, .99, i, parseInt(this.selectedOpenCvMethod, 10)), mq.solvePnPRefineVVS(t, e, n, r, a, s);
+                    const o = mq.matFromArray(3, 1, mq.CV_64F, [0, 0, 1e3]),
+                        u = new mq.Mat,
+                        l = new mq.Mat;
+                    mq.projectPoints(o, a, s, n, r, u, l);
+                    let c = u.data64F[0],
+                        h = u.data64F[1];
+                    const p = mq.matFromArray(3, 1, mq.CV_64F, [0, 1e3, 0]),
+                        d = new mq.Mat,
+                        f = new mq.Mat;
+                    mq.projectPoints(p, a, s, n, r, d, f);
+                    let m = d.data64F[0],
+                        g = d.data64F[1];
+                    const y = mq.matFromArray(3, 1, mq.CV_64F, [1e3, 0, 0]),
+                        b = new mq.Mat,
+                        v = new mq.Mat;
+                    mq.projectPoints(y, a, s, n, r, b, v);
+                    let x = b.data64F[0],
+                        w = b.data64F[1];
+                    return this.drawingData = {
+                        noseTip: this.noseTip,
+                        frontx: c,
+                        fronty: h,
+                        topx: m,
+                        topy: g,
+                        sidex: x,
+                        sidey: w
+                    }, a.data64F.map(e => e / Math.PI * 180)
+                }
+                draw(e) {
+                    const {
+                        noseTip: t,
+                        frontx: n,
+                        fronty: r,
+                        topx: a,
+                        topy: s,
+                        sidex: i,
+                        sidey: o
+                    } = this.drawingData;
+                    try {
+                        const u = {
+                                _x: yq("nosetipx", t.x),
+                                _y: yq("nosetipy", t.y)
+                            },
+                            l = yq("frontx", n),
+                            c = yq("fronty", r),
+                            h = yq("topx", a),
+                            p = yq("topy", s),
+                            d = yq("sidex", i),
+                            f = yq("sidey", o);
+                        e.strokeStyle = "white", e.fillStyle = "white", e.beginPath(), e.arc(l, c, 2, 0, 2 * Math.PI), e.fill(), e.arc(u._x, u._y, 2, 0, 2 * Math.PI), e.fill(), e.lineWidth = 2, e.beginPath(), e.strokeStyle = "red", e.fillStyle = "red", e.moveTo(l, c), e.lineTo(u._x, u._y), e.stroke(), e.beginPath(), e.strokeStyle = "green", e.fillStyle = "green", e.moveTo(h, p), e.lineTo(u._x, u._y), e.stroke(), e.beginPath(), e.strokeStyle = "blue", e.fillStyle = "blue", e.moveTo(d, f), e.lineTo(u._x, u._y), e.stroke()
+                    } catch (e) {
+                        console.error("errrrr"), console.error(e)
+                    }
+                }
+            }
+            const xq = () => {
+                    let e = {
+                        isMobile: !1,
+                        isAndroid: !1,
+                        isLinux: !1,
+                        isMac: !1,
+                        isWin: !1,
+                        isChrome: !1,
+                        isFirefox: !1,
+                        isSafari: !1,
+                        isIos: !1
+                    };
+                    if (e.isMobile = /Android|webOS|iPhone|iPad|iPod|BlackBerry/i.test(navigator.userAgent), e.isAndroid = /Android/i.test(navigator.userAgent), e.isIos = /iPhone|iPad|iPod/i.test(navigator.userAgent), e.isLinux = /linux/i.test(navigator.platform), e.isMac = /mac/i.test(navigator.platform), e.isWin = /win/i.test(navigator.platform), e.isChrome = /chrome/i.test(navigator.userAgent), e.isFirefox = /firefox/i.test(navigator.userAgent), e.isSafari = !e.isChrome && /safari/i.test(navigator.userAgent), e.isMobile = /Android|webOS|iPhone|iPad|iPod|BlackBerry/i.test(navigator.userAgent), e.isAndroid = /Android/i.test(navigator.userAgent), e.isIos = /iPhone|iPad|iPod/i.test(navigator.userAgent), !e.isIos) {
+                        navigator.userAgent.match(/Mac/) && navigator.maxTouchPoints && navigator.maxTouchPoints > 2 && (e.isIos = !0, e.isMobile = !0)
+                    }
+                    return e
+                },
+                wq = window.cv;
+            class kq {
+                constructor(e, t, n) {
+                    this.probabilityThreshold = 50, this.counterSuccessfulResults = 0, this.MAX_FPS_DETECTION = 200, this.rendering = {
+                        now: 0,
+                        elapsed: 0,
+                        then: 0
+                    }, this.MAX_NUMBER_FACES = 1, this.MAX_ANGLE_TURN_HEAD = 30, this.MIN_FACE_SIZE_FOR_MOBILE = 224, this.MIN_FACE_SIZE_FOR_DESKTOP = 350, this.MIN_OCCUPIED_SPACE_FOR_DESKTOP = 15, this.MIN_OCCUPIED_SPACE_FOR_MOBILE = 15, this.X_OFFSET_FROM_FRAME = 15, this.Y_OFFSET_FROM_FRAME = 5, this.DEFAULT_NUMBER_SUCCESSFUL_RESULTS_FOR_AUTO_CAPTURING = 5, this.videoRatio = 1, this.showDetectedFace = !1, this.headTurnCounter = 0, this.counterSuccessfulHeadTurns = 0, this.stream = e, this.isMobile = t, this.modelPath = n
+                }
+                setProbabilityThreshold(e) {
+                    this.probabilityThreshold = e
+                }
+                static getInstance(e, t, n) {
+                    return kq.instance || (kq.instance = new kq(e, t, n)), kq.instance
+                }
+                updateHtmlElements(e) {
+                    this.videoElement = e
+                }
+                static async initDetector(e) {
+                    await d.setBackend(this.BACKEND), await d.enableProdMode(), await d.ENV.set("DEBUG", !1), await d.ready(), await this.loadingModels(e), wq || console.error("open cv not found"), l.loadedModels({
+                        tinyFaceDetectorLoaded: ZH.tinyFaceDetector.isLoaded,
+                        faceLandmark68TinyNetLoaded: ZH.faceLandmark68TinyNet.isLoaded
+                    });
+                    const t = performance.now();
+                    await this.warmingUpModels(), l.detectorInitialized(performance.now() - t)
+                }
+                static async loadingModels(e) {
+                    this.modelsLoaded() || (await ZH.tinyFaceDetector.load(e), await ZH.faceLandmark68TinyNet.load(e))
+                }
+                static modelsLoaded() {
+                    return ZH.tinyFaceDetector.isLoaded && ZH.faceLandmark68TinyNet.isLoaded
+                }
+                static async warmingUpModels() {
+                    const e = new Image;
+                    e.src = "data:image/jpg;base64,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", await this.detectAllFaces(e)
+                }
+                static async detectAllFaces(e) {
+                    d.engine().startScope();
+                    let t = [];
+                    try {
+                        let n = e;
+                        if (this.isIos && e instanceof HTMLVideoElement) {
+                            const t = document.createElement("canvas"),
+                                r = 480,
+                                a = 640;
+                            t.width = r, t.height = a;
+                            t.getContext("2d").drawImage(e, 0, 0, r, a), n = t
+                        }
+                        t = await
+
+                        function (e, t = new DH) {
+                            return new cq(e, t)
+                        }(n, this.tinyFaceDetectorOptions).withFaceLandmarks(!0)
+                    } catch (e) {} finally {
+                        d.engine().endScope()
+                    }
+                    return t
+                }
+                async startDetector() {
+                    if (!this.videoElement) throw Error("Video element not found");
+                    if (0 === this.videoElement.clientWidth) throw Error("Video element too small");
+                    this.rendering.then = Date.now(), this.counterSuccessfulResults = 0, this.updateForIos(), this.videoRatio = function (e) {
+                        const t = e.clientWidth / e.videoWidth,
+                            n = e.clientHeight / e.videoHeight;
+                        return Math.min(t, n)
+                    }(this.videoElement), kq.modelsLoaded() || await kq.initDetector(this.modelPath), this.requestAnimationFrame = requestAnimationFrame(() => this.detection())
+                }
+                stopDetector() {
+                    this.requestAnimationFrame && (cancelAnimationFrame(this.requestAnimationFrame), this.requestAnimationFrame = void 0, this.innerCanvas && this.innerCanvas.remove(), this.innerCanvas = void 0)
+                }
+                updateForIos() {
+                    xq().isIos && (this.DEFAULT_NUMBER_SUCCESSFUL_RESULTS_FOR_AUTO_CAPTURING = 3, this.MAX_FPS_DETECTION = 500, kq.isIos = !0)
+                }
+                async detection() {
+                    if (this.requestAnimationFrame = requestAnimationFrame(() => this.detection()), this.rendering.now = Date.now(), this.rendering.elapsed = this.rendering.now - this.rendering.then, this.rendering.elapsed > this.MAX_FPS_DETECTION) {
+                        this.rendering.then = this.rendering.now - this.rendering.elapsed % this.MAX_FPS_DETECTION;
+                        const e = await kq.detectAllFaces(this.videoElement);
+                        (this.showDetectedFace || kq.debug) && this.drawFaces(e);
+                        const t = await this.checkDetectedFaceOnError(e);
+                        Object.values(t).every(e => !e) ? (this.counterSuccessfulResults++, this.autoCapturing()) : (this.counterSuccessfulResults = 0, this.stream.returnErrors(t))
+                    }
+                }
+                async checkDetectedFaceOnError(e) {
+                    const t = function e(t, n) {
+                        const {
+                            width: r,
+                            height: a
+                        } = new Pj(n.width, n.height);
+                        if (r <= 0 || a <= 0) throw new Error("resizeResults - invalid dimensions: " + JSON.stringify({
+                            width: r,
+                            height: a
+                        }));
+                        if (Array.isArray(t)) return t.map(t => e(t, {
+                            width: r,
+                            height: a
+                        }));
+                        if (uH(t)) {
+                            const e = t.detection.forSize(r, a),
+                                n = t.unshiftedLandmarks.forSize(e.box.width, e.box.height);
+                            return lH(hG(t, e), n)
+                        }
+                        return cG(t) ? hG(t, t.detection.forSize(r, a)) : t instanceof uG || t instanceof nG ? t.forSize(r, a) : t
+                    }
+                    /**
+                     * @license
+                     * Copyright 2017 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2018 Google LLC
+                     *
+                     * Use of this source code is governed by an MIT-style
+                     * license that can be found in the LICENSE file or at
+                     * https://opensource.org/licenses/MIT.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2018 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     *
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2018 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2019 Google LLC
+                     *
+                     * Use of this source code is governed by an MIT-style
+                     * license that can be found in the LICENSE file or at
+                     * https://opensource.org/licenses/MIT.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2019 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     *
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2019 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2020 Google Inc. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2020 Google LLC
+                     *
+                     * Use of this source code is governed by an MIT-style
+                     * license that can be found in the LICENSE file or at
+                     * https://opensource.org/licenses/MIT.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2020 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2020 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the License);
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an AS IS BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2021 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2021 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * https://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2022 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /**
+                     * @license
+                     * Copyright 2018 Google LLC. All Rights Reserved.
+                     * Licensed under the Apache License, Version 2.0 (the "License");
+                     * you may not use this file except in compliance with the License.
+                     * You may obtain a copy of the License at
+                     *
+                     * http://www.apache.org/licenses/LICENSE-2.0
+                     *
+                     * Unless required by applicable law or agreed to in writing, software
+                     * distributed under the License is distributed on an "AS IS" BASIS,
+                     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+                     * See the License for the specific language governing permissions and
+                     * limitations under the License.
+                     * =============================================================================
+                     */
+                    /** @license See the LICENSE file. */
+                    (e, {
+                        width: this.videoElement.videoWidth,
+                        height: this.videoElement.videoHeight
+                    }), n = new hq, r = this.checkNumberFaces(t);
+                    return n.FACE_NOT_FOUND = !r && 0 === t.length, n.TOO_MANY_FACES = !r && t.length > this.MAX_NUMBER_FACES, r && (wq && (n.FACE_ANGLE_TOO_LARGE = !this.checkHeadRotation(t)), n.PROBABILITY_TOO_SMALL = !this.checkProbability(t), n.FACE_TOO_SMALL = !this.checkFaceSize(t), n.FACE_CLOSE_TO_BORDER = !this.checkFaceIndent(t)), n
+                }
+                autoCapturing() {
+                    kq.debug && console.info(this.counterSuccessfulResults), this.counterSuccessfulResults >= this.DEFAULT_NUMBER_SUCCESSFUL_RESULTS_FOR_AUTO_CAPTURING && (this.stream.autoCapturing(), this.counterSuccessfulResults = 0)
+                }
+                checkNumberFaces(e) {
+                    return e.length === this.MAX_NUMBER_FACES
+                }
+                checkProbability(e) {
+                    return e[0].detection.score > +this.probabilityThreshold / 100
+                }
+                checkFaceSize(e) {
+                    const t = e[0].detection.box,
+                        {
+                            width: n,
+                            height: r,
+                            area: a
+                        } = t,
+                        {
+                            imageHeight: s,
+                            imageWidth: i
+                        } = e[0].detection,
+                        o = 100 / (s * i / a);
+                    return Math.min(n / this.videoRatio, r / this.videoRatio) > (this.isMobile ? this.MIN_FACE_SIZE_FOR_MOBILE : this.MIN_FACE_SIZE_FOR_DESKTOP) && o > (this.isMobile ? this.MIN_OCCUPIED_SPACE_FOR_MOBILE : this.MIN_OCCUPIED_SPACE_FOR_DESKTOP)
+                }
+                checkFaceIndent(e) {
+                    const t = e[0].detection.box,
+                        {
+                            imageHeight: n,
+                            imageWidth: r
+                        } = e[0].detection,
+                        {
+                            top: a,
+                            left: s,
+                            bottom: i,
+                            right: o
+                        } = t,
+                        u = s,
+                        l = r - o,
+                        c = n - i;
+                    return !(a < this.Y_OFFSET_FROM_FRAME || u < this.X_OFFSET_FROM_FRAME || l < this.X_OFFSET_FROM_FRAME || c < this.Y_OFFSET_FROM_FRAME)
+                }
+                checkHeadRotation(e) {
+                    const t = document.createElement("canvas"),
+                        n = function (e, t, n = !1) {
+                            const {
+                                width: r,
+                                height: a
+                            } = n ? TG(t) : t;
+                            return e.width = r, e.height = a, {
+                                width: r,
+                                height: a
+                            }
+                        }(t, this.videoElement, !0);
+                    t.remove();
+                    const r = new vq(e[0].landmarks.positions, n);
+                    let a = [0, 0, 0];
+                    try {
+                        a = r.estimateHeadPose()
+                    } catch (e) {}
+                    return a[2] > this.MAX_ANGLE_TURN_HEAD || -1 * a[2] > this.MAX_ANGLE_TURN_HEAD ? this.headTurnCounter++ : (this.counterSuccessfulHeadTurns++, this.counterSuccessfulHeadTurns >= 5 && (this.headTurnCounter = 0, this.counterSuccessfulHeadTurns = 0)), this.headTurnCounter <= 1
+                }
+                drawFaces(e) {
+                    this.innerCanvas || this.createCanvasForDrawingMask(this.videoElement);
+                    const t = this.innerCanvas;
+                    if (!t) return console.error("no canvas for drawing");
+                    const n = t.getContext("2d");
+                    if (n) {
+                        n.clearRect(0, 0, t.width, t.height), n.font = 'small-caps 20px "Segoe UI"', n.fillStyle = "white";
+                        for (const t of e) {
+                            n.lineWidth = 3, n.strokeStyle = "deepskyblue", n.fillStyle = "deepskyblue", n.globalAlpha = .6, n.beginPath(), n.rect(t.detection.box.x, t.detection.box.y, t.detection.box.width, t.detection.box.height), n.stroke(), n.globalAlpha = 1, n.globalAlpha = .8, n.fillStyle = "lightblue";
+                            const e = 2;
+                            for (let r = 0; r < t.landmarks.positions.length; r++) n.beginPath(), n.arc(t.landmarks.positions[r].x, t.landmarks.positions[r].y, e, 0, 2 * Math.PI), n.fill()
+                        }
+                    }
+                }
+                createCanvasForDrawingMask(e) {
+                    this.innerCanvas = document.createElement("canvas"), this.innerCanvas.className = "canvas-on-video", this.innerCanvas.width = e.videoWidth, this.innerCanvas.height = e.videoHeight, this.videoElement.after(this.innerCanvas), this.innerCanvas.offsetWidth && this.innerCanvas.setAttribute("style", `margin-left:${-this.innerCanvas.offsetWidth/2}px; margin-top: ${-this.innerCanvas.offsetHeight/2}px;`)
+                }
+            }
+            kq.tinyFaceDetectorOptions = new qH({
+                inputSize: 192
+            }), kq.BACKEND = "webgl", kq.debug = !1;
+            var Sq = function (e, t, n) {
+                return e(n = {
+                    path: t,
+                    exports: {},
+                    require: function (e, t) {
+                        return function () {
+                            throw new Error("Dynamic requires are not currently supported by @rollup/plugin-commonjs")
+                        }()
+                    }
+                }, n.exports), n.exports
+            }((function (e, t) {
+                ! function () {
+                    var n = {};
+
+                    function r(e) {
+                        return h(">" + d("B", e.length), e)
+                    }
+
+                    function a(e) {
+                        return h(">" + d("H", e.length), e)
+                    }
+
+                    function s(e) {
+                        return h(">" + d("L", e.length), e)
+                    }
+
+                    function i(e, t, n) {
+                        var i, o, u, l, c = "",
+                            p = "";
+                        if ("Byte" == t)(i = e.length) <= 4 ? p = r(e) + d("\0", 4 - i) : (p = h(">L", [n]), c = r(e));
+                        else if ("Short" == t)(i = e.length) <= 2 ? p = a(e) + d("\0\0", 2 - i) : (p = h(">L", [n]), c = a(e));
+                        else if ("Long" == t)(i = e.length) <= 1 ? p = s(e) : (p = h(">L", [n]), c = s(e));
+                        else if ("Ascii" == t)(i = (o = e + "\0").length) > 4 ? (p = h(">L", [n]), c = o) : p = o + d("\0", 4 - i);
+                        else if ("Rational" == t) {
+                            if ("number" == typeof e[0]) i = 1, u = e[0], l = e[1], o = h(">L", [u]) + h(">L", [l]);
+                            else {
+                                i = e.length, o = "";
+                                for (var f = 0; f < i; f++) u = e[f][0], l = e[f][1], o += h(">L", [u]) + h(">L", [l])
+                            }
+                            p = h(">L", [n]), c = o
+                        } else if ("SRational" == t) {
+                            if ("number" == typeof e[0]) i = 1, u = e[0], l = e[1], o = h(">l", [u]) + h(">l", [l]);
+                            else {
+                                i = e.length, o = "";
+                                for (f = 0; f < i; f++) u = e[f][0], l = e[f][1], o += h(">l", [u]) + h(">l", [l])
+                            }
+                            p = h(">L", [n]), c = o
+                        } else "Undefined" == t && ((i = e.length) > 4 ? (p = h(">L", [n]), c = e) : p = e + d("\0", 4 - i));
+                        return [h(">L", [i]), p, c]
+                    }
+
+                    function o(e, t, n) {
+                        var r, a = Object.keys(e).length,
+                            s = h(">H", [a]);
+                        r = ["0th", "1st"].indexOf(t) > -1 ? 2 + 12 * a + 4 : 2 + 12 * a;
+                        var o = "",
+                            u = "";
+                        for (var l in e)
+                            if ("string" == typeof l && (l = parseInt(l)), !("0th" == t && [34665, 34853].indexOf(l) > -1 || "Exif" == t && 40965 == l || "1st" == t && [513, 514].indexOf(l) > -1)) {
+                                var c = e[l],
+                                    p = h(">H", [l]),
+                                    d = g[t][l].type,
+                                    f = h(">H", [m[d]]);
+                                "number" == typeof c && (c = [c]);
+                                var y = i(c, d, 8 + r + n + u.length);
+                                o += p + f + y[0] + y[1], u += y[2]
+                            } return [s + o, u]
+                    }
+
+                    function u(e) {
+                        var t;
+                        if ("ÿØ" == e.slice(0, 2)) t = function (e) {
+                            for (var t, n = 0; n < e.length; n++)
+                                if ("ÿá" == (t = e[n]).slice(0, 2) && "Exif\0\0" == t.slice(4, 10)) return t;
+                            return null
+                        }(f(e)), this.tiftag = t ? t.slice(10) : null;
+                        else if (["II", "MM"].indexOf(e.slice(0, 2)) > -1) this.tiftag = e;
+                        else {
+                            if ("Exif" != e.slice(0, 4)) throw new Error("Given file is neither JPEG nor TIFF.");
+                            this.tiftag = e.slice(6)
+                        }
+                    }
+                    if (n.version = "1.0.4", n.remove = function (e) {
+                            var t = !1;
+                            if ("ÿØ" == e.slice(0, 2));
+                            else {
+                                if ("data:image/jpeg;base64," != e.slice(0, 23) && "data:image/jpg;base64," != e.slice(0, 22)) throw new Error("Given data is not jpeg.");
+                                e = c(e.split(",")[1]), t = !0
+                            }
+                            var n = f(e).filter((function (e) {
+                                return !("ÿá" == e.slice(0, 2) && "Exif\0\0" == e.slice(4, 10))
+                            })).join("");
+                            return t && (n = "data:image/jpeg;base64," + l(n)), n
+                        }, n.insert = function (e, t) {
+                            var n = !1;
+                            if ("Exif\0\0" != e.slice(0, 6)) throw new Error("Given data is not exif.");
+                            if ("ÿØ" == t.slice(0, 2));
+                            else {
+                                if ("data:image/jpeg;base64," != t.slice(0, 23) && "data:image/jpg;base64," != t.slice(0, 22)) throw new Error("Given data is not jpeg.");
+                                t = c(t.split(",")[1]), n = !0
+                            }
+                            var r = "ÿá" + h(">H", [e.length + 2]) + e,
+                                a = function (e, t) {
+                                    var n = !1,
+                                        r = [];
+                                    e.forEach((function (a, s) {
+                                        "ÿá" == a.slice(0, 2) && "Exif\0\0" == a.slice(4, 10) && (n ? r.unshift(s) : (e[s] = t, n = !0))
+                                    })), r.forEach((function (t) {
+                                        e.splice(t, 1)
+                                    })), !n && t && (e = [e[0], t].concat(e.slice(1)));
+                                    return e.join("")
+                                }(f(t), r);
+                            return n && (a = "data:image/jpeg;base64," + l(a)), a
+                        }, n.load = function (e) {
+                            var t;
+                            if ("string" != typeof e) throw new Error("'load' gots invalid type argument.");
+                            if ("ÿØ" == e.slice(0, 2)) t = e;
+                            else if ("data:image/jpeg;base64," == e.slice(0, 23) || "data:image/jpg;base64," == e.slice(0, 22)) t = c(e.split(",")[1]);
+                            else {
+                                if ("Exif" != e.slice(0, 4)) throw new Error("'load' gots invalid file data.");
+                                t = e.slice(6)
+                            }
+                            var n = {
+                                    "0th": {},
+                                    Exif: {},
+                                    GPS: {},
+                                    Interop: {},
+                                    "1st": {},
+                                    thumbnail: null
+                                },
+                                r = new u(t);
+                            if (null === r.tiftag) return n;
+                            "II" == r.tiftag.slice(0, 2) ? r.endian_mark = "<" : r.endian_mark = ">";
+                            var a = p(r.endian_mark + "L", r.tiftag.slice(4, 8))[0];
+                            n["0th"] = r.get_ifd(a, "0th");
+                            var s = n["0th"].first_ifd_pointer;
+                            if (delete n["0th"].first_ifd_pointer, 34665 in n["0th"] && (a = n["0th"][34665], n.Exif = r.get_ifd(a, "Exif")), 34853 in n["0th"] && (a = n["0th"][34853], n.GPS = r.get_ifd(a, "GPS")), 40965 in n.Exif && (a = n.Exif[40965], n.Interop = r.get_ifd(a, "Interop")), "\0\0\0\0" != s && (a = p(r.endian_mark + "L", s)[0], n["1st"] = r.get_ifd(a, "1st"), 513 in n["1st"] && 514 in n["1st"])) {
+                                var i = n["1st"][513] + n["1st"][514],
+                                    o = r.tiftag.slice(n["1st"][513], i);
+                                n.thumbnail = o
+                            }
+                            return n
+                        }, n.dump = function (e) {
+                            var t, r, a, s, i, u, l = (t = e, JSON.parse(JSON.stringify(t))),
+                                c = !1,
+                                p = !1,
+                                d = !1,
+                                g = !1;
+                            r = "0th" in l ? l["0th"] : {}, "Exif" in l && Object.keys(l.Exif).length || "Interop" in l && Object.keys(l.Interop).length ? (r[34665] = 1, c = !0, a = l.Exif, "Interop" in l && Object.keys(l.Interop).length ? (a[40965] = 1, d = !0, s = l.Interop) : Object.keys(a).indexOf(n.ExifIFD.InteroperabilityTag.toString()) > -1 && delete a[40965]) : Object.keys(r).indexOf(n.ImageIFD.ExifTag.toString()) > -1 && delete r[34665], "GPS" in l && Object.keys(l.GPS).length ? (r[n.ImageIFD.GPSTag] = 1, p = !0, i = l.GPS) : Object.keys(r).indexOf(n.ImageIFD.GPSTag.toString()) > -1 && delete r[n.ImageIFD.GPSTag], "1st" in l && "thumbnail" in l && null != l.thumbnail && (g = !0, l["1st"][513] = 1, l["1st"][514] = 1, u = l["1st"]);
+                            var y, b, v, x = o(r, "0th", 0),
+                                w = x[0].length + 12 * c + 12 * p + 4 + x[1].length,
+                                k = "",
+                                S = 0,
+                                I = "",
+                                A = 0,
+                                E = "",
+                                N = 0,
+                                C = "";
+                            (c && (S = (y = o(a, "Exif", w))[0].length + 12 * d + y[1].length), p && (A = (I = o(i, "GPS", w + S).join("")).length), d) && (N = (E = o(s, "Interop", w + S + A).join("")).length);
+                            if (g && (b = o(u, "1st", w + S + A + N), (v = function (e) {
+                                    var t = f(e);
+                                    for (;
+                                        "ÿà" <= t[1].slice(0, 2) && t[1].slice(0, 2) <= "ÿï";) t = [t[0]].concat(t.slice(2));
+                                    return t.join("")
+                                }(l.thumbnail)).length > 64e3)) throw new Error("Given thumbnail is too large. max 64kB");
+                            var T = "",
+                                R = "",
+                                _ = "",
+                                F = "\0\0\0\0";
+                            if (c) {
+                                var M = h(">L", [D = 8 + w]);
+                                T = h(">H", [34665]) + h(">H", [m.Long]) + h(">L", [1]) + M
+                            }
+                            if (p) {
+                                M = h(">L", [D = 8 + w + S]);
+                                R = h(">H", [34853]) + h(">H", [m.Long]) + h(">L", [1]) + M
+                            }
+                            if (d) {
+                                M = h(">L", [D = 8 + w + S + A]);
+                                _ = h(">H", [40965]) + h(">H", [m.Long]) + h(">L", [1]) + M
+                            }
+                            if (g) {
+                                var D;
+                                F = h(">L", [D = 8 + w + S + A + N]);
+                                var O = "
\0\0\0\0
" + h(">L", [D + b[0].length + 24 + 4 + b[1].length]),
+                                    L = "\0\0\0\0
" + h(">L", [v.length]);
+                                C = b[0] + O + L + "\0\0\0\0" + b[1] + v
+                            }
+                            var P = x[0] + T + R + F + x[1];
+                            return c && (k = y[0] + _ + y[1]), "Exif\0\0MM\0*\0\0\0\b" + P + k + I + E + C
+                        }, u.prototype = {
+                            get_ifd: function (e, t) {
+                                var n, r = {},
+                                    a = p(this.endian_mark + "H", this.tiftag.slice(e, e + 2))[0],
+                                    s = e + 2;
+                                n = ["0th", "1st"].indexOf(t) > -1 ? "Image" : t;
+                                for (var i = 0; i < a; i++) {
+                                    e = s + 12 * i;
+                                    var o = p(this.endian_mark + "H", this.tiftag.slice(e, e + 2))[0],
+                                        u = [p(this.endian_mark + "H", this.tiftag.slice(e + 2, e + 4))[0], p(this.endian_mark + "L", this.tiftag.slice(e + 4, e + 8))[0], this.tiftag.slice(e + 8, e + 12)];
+                                    o in g[n] && (r[o] = this.convert_value(u))
+                                }
+                                return "0th" == t && (e = s + 12 * a, r.first_ifd_pointer = this.tiftag.slice(e, e + 4)), r
+                            },
+                            convert_value: function (e) {
+                                var t, n = null,
+                                    r = e[0],
+                                    a = e[1],
+                                    s = e[2];
+                                if (1 == r) a > 4 ? (t = p(this.endian_mark + "L", s)[0], n = p(this.endian_mark + d("B", a), this.tiftag.slice(t, t + a))) : n = p(this.endian_mark + d("B", a), s.slice(0, a));
+                                else if (2 == r) a > 4 ? (t = p(this.endian_mark + "L", s)[0], n = this.tiftag.slice(t, t + a - 1)) : n = s.slice(0, a - 1);
+                                else if (3 == r) a > 2 ? (t = p(this.endian_mark + "L", s)[0], n = p(this.endian_mark + d("H", a), this.tiftag.slice(t, t + 2 * a))) : n = p(this.endian_mark + d("H", a), s.slice(0, 2 * a));
+                                else if (4 == r) a > 1 ? (t = p(this.endian_mark + "L", s)[0], n = p(this.endian_mark + d("L", a), this.tiftag.slice(t, t + 4 * a))) : n = p(this.endian_mark + d("L", a), s);
+                                else if (5 == r)
+                                    if (t = p(this.endian_mark + "L", s)[0], a > 1) {
+                                        n = [];
+                                        for (var i = 0; i < a; i++) n.push([p(this.endian_mark + "L", this.tiftag.slice(t + 8 * i, t + 4 + 8 * i))[0], p(this.endian_mark + "L", this.tiftag.slice(t + 4 + 8 * i, t + 8 + 8 * i))[0]])
+                                    } else n = [p(this.endian_mark + "L", this.tiftag.slice(t, t + 4))[0], p(this.endian_mark + "L", this.tiftag.slice(t + 4, t + 8))[0]];
+                                else if (7 == r) a > 4 ? (t = p(this.endian_mark + "L", s)[0], n = this.tiftag.slice(t, t + a)) : n = s.slice(0, a);
+                                else if (9 == r) a > 1 ? (t = p(this.endian_mark + "L", s)[0], n = p(this.endian_mark + d("l", a), this.tiftag.slice(t, t + 4 * a))) : n = p(this.endian_mark + d("l", a), s);
+                                else {
+                                    if (10 != r) throw new Error("Exif might be wrong. Got incorrect value type to decode. type:" + r);
+                                    if (t = p(this.endian_mark + "L", s)[0], a > 1) {
+                                        n = [];
+                                        for (i = 0; i < a; i++) n.push([p(this.endian_mark + "l", this.tiftag.slice(t + 8 * i, t + 4 + 8 * i))[0], p(this.endian_mark + "l", this.tiftag.slice(t + 4 + 8 * i, t + 8 + 8 * i))[0]])
+                                    } else n = [p(this.endian_mark + "l", this.tiftag.slice(t, t + 4))[0], p(this.endian_mark + "l", this.tiftag.slice(t + 4, t + 8))[0]]
+                                }
+                                return n instanceof Array && 1 == n.length ? n[0] : n
+                            }
+                        }, "undefined" != typeof window && "function" == typeof window.btoa) var l = window.btoa;
+                    if (void 0 === l) l = function (e) {
+                        for (var t, n, r, a, s, i, o, u = "", l = 0, c = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/="; l < e.length;) a = (t = e.charCodeAt(l++)) >> 2, s = (3 & t) << 4 | (n = e.charCodeAt(l++)) >> 4, i = (15 & n) << 2 | (r = e.charCodeAt(l++)) >> 6, o = 63 & r, isNaN(n) ? i = o = 64 : isNaN(r) && (o = 64), u = u + c.charAt(a) + c.charAt(s) + c.charAt(i) + c.charAt(o);
+                        return u
+                    };
+                    if ("undefined" != typeof window && "function" == typeof window.atob) var c = window.atob;
+                    if (void 0 === c) c = function (e) {
+                        var t, n, r, a, s, i, o = "",
+                            u = 0,
+                            l = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";
+                        for (e = e.replace(/[^A-Za-z0-9\+\/\=]/g, ""); u < e.length;) t = l.indexOf(e.charAt(u++)) << 2 | (a = l.indexOf(e.charAt(u++))) >> 4, n = (15 & a) << 4 | (s = l.indexOf(e.charAt(u++))) >> 2, r = (3 & s) << 6 | (i = l.indexOf(e.charAt(u++))), o += String.fromCharCode(t), 64 != s && (o += String.fromCharCode(n)), 64 != i && (o += String.fromCharCode(r));
+                        return o
+                    };
+
+                    function h(e, t) {
+                        if (!(t instanceof Array)) throw new Error("'pack' error. Got invalid type argument.");
+                        if (e.length - 1 != t.length) throw new Error("'pack' error. " + (e.length - 1) + " marks, " + t.length + " elements.");
+                        var n;
+                        if ("<" == e[0]) n = !0;
+                        else {
+                            if (">" != e[0]) throw new Error("");
+                            n = !1
+                        }
+                        for (var r = "", a = 1, s = null, i = null, o = null; i = e[a];) {
+                            if ("b" == i.toLowerCase()) {
+                                if (s = t[a - 1], "b" == i && s < 0 && (s += 256), s > 255 || s < 0) throw new Error("'pack' error.");
+                                o = String.fromCharCode(s)
+                            } else if ("H" == i) {
+                                if ((s = t[a - 1]) > 65535 || s < 0) throw new Error("'pack' error.");
+                                o = String.fromCharCode(Math.floor(s % 65536 / 256)) + String.fromCharCode(s % 256), n && (o = o.split("").reverse().join(""))
+                            } else {
+                                if ("l" != i.toLowerCase()) throw new Error("'pack' error.");
+                                if (s = t[a - 1], "l" == i && s < 0 && (s += 4294967296), s > 4294967295 || s < 0) throw new Error("'pack' error.");
+                                o = String.fromCharCode(Math.floor(s / 16777216)) + String.fromCharCode(Math.floor(s % 16777216 / 65536)) + String.fromCharCode(Math.floor(s % 65536 / 256)) + String.fromCharCode(s % 256), n && (o = o.split("").reverse().join(""))
+                            }
+                            r += o, a += 1
+                        }
+                        return r
+                    }
+
+                    function p(e, t) {
+                        if ("string" != typeof t) throw new Error("'unpack' error. Got invalid type argument.");
+                        for (var n, r = 0, a = 1; a < e.length; a++)
+                            if ("b" == e[a].toLowerCase()) r += 1;
+                            else if ("h" == e[a].toLowerCase()) r += 2;
+                        else {
+                            if ("l" != e[a].toLowerCase()) throw new Error("'unpack' error. Got invalid mark.");
+                            r += 4
+                        }
+                        if (r != t.length) throw new Error("'unpack' error. Mismatch between symbol and string length. " + r + ":" + t.length);
+                        if ("<" == e[0]) n = !0;
+                        else {
+                            if (">" != e[0]) throw new Error("'unpack' error.");
+                            n = !1
+                        }
+                        for (var s = [], i = 0, o = 1, u = null, l = null, c = null, h = ""; l = e[o];) {
+                            if ("b" == l.toLowerCase()) c = 1, u = (h = t.slice(i, i + c)).charCodeAt(0), "b" == l && u >= 128 && (u -= 256);
+                            else if ("H" == l) c = 2, h = t.slice(i, i + c), n && (h = h.split("").reverse().join("")), u = 256 * h.charCodeAt(0) + h.charCodeAt(1);
+                            else {
+                                if ("l" != l.toLowerCase()) throw new Error("'unpack' error. " + l);
+                                c = 4, h = t.slice(i, i + c), n && (h = h.split("").reverse().join("")), u = 16777216 * h.charCodeAt(0) + 65536 * h.charCodeAt(1) + 256 * h.charCodeAt(2) + h.charCodeAt(3), "l" == l && u >= 2147483648 && (u -= 4294967296)
+                            }
+                            s.push(u), i += c, o += 1
+                        }
+                        return s
+                    }
+
+                    function d(e, t) {
+                        for (var n = "", r = 0; r < t; r++) n += e;
+                        return n
+                    }
+
+                    function f(e) {
+                        if ("ÿØ" != e.slice(0, 2)) throw new Error("Given data isn't JPEG.");
+                        for (var t = 2, n = ["ÿØ"];;) {
+                            if ("ÿÚ" == e.slice(t, t + 2)) {
+                                n.push(e.slice(t));
+                                break
+                            }
+                            var r = t + p(">H", e.slice(t + 2, t + 4))[0] + 2;
+                            if (n.push(e.slice(t, r)), (t = r) >= e.length) throw new Error("Wrong JPEG data.")
+                        }
+                        return n
+                    }
+                    var m = {
+                            Byte: 1,
+                            Ascii: 2,
+                            Short: 3,
+                            Long: 4,
+                            Rational: 5,
+                            Undefined: 7,
+                            SLong: 9,
+                            SRational: 10
+                        },
+                        g = {
+                            Image: {
+                                11: {
+                                    name: "ProcessingSoftware",
+                                    type: "Ascii"
+                                },
+                                254: {
+                                    name: "NewSubfileType",
+                                    type: "Long"
+                                },
+                                255: {
+                                    name: "SubfileType",
+                                    type: "Short"
+                                },
+                                256: {
+                                    name: "ImageWidth",
+                                    type: "Long"
+                                },
+                                257: {
+                                    name: "ImageLength",
+                                    type: "Long"
+                                },
+                                258: {
+                                    name: "BitsPerSample",
+                                    type: "Short"
+                                },
+                                259: {
+                                    name: "Compression",
+                                    type: "Short"
+                                },
+                                262: {
+                                    name: "PhotometricInterpretation",
+                                    type: "Short"
+                                },
+                                263: {
+                                    name: "Threshholding",
+                                    type: "Short"
+                                },
+                                264: {
+                                    name: "CellWidth",
+                                    type: "Short"
+                                },
+                                265: {
+                                    name: "CellLength",
+                                    type: "Short"
+                                },
+                                266: {
+                                    name: "FillOrder",
+                                    type: "Short"
+                                },
+                                269: {
+                                    name: "DocumentName",
+                                    type: "Ascii"
+                                },
+                                270: {
+                                    name: "ImageDescription",
+                                    type: "Ascii"
+                                },
+                                271: {
+                                    name: "Make",
+                                    type: "Ascii"
+                                },
+                                272: {
+                                    name: "Model",
+                                    type: "Ascii"
+                                },
+                                273: {
+                                    name: "StripOffsets",
+                                    type: "Long"
+                                },
+                                274: {
+                                    name: "Orientation",
+                                    type: "Short"
+                                },
+                                277: {
+                                    name: "SamplesPerPixel",
+                                    type: "Short"
+                                },
+                                278: {
+                                    name: "RowsPerStrip",
+                                    type: "Long"
+                                },
+                                279: {
+                                    name: "StripByteCounts",
+                                    type: "Long"
+                                },
+                                282: {
+                                    name: "XResolution",
+                                    type: "Rational"
+                                },
+                                283: {
+                                    name: "YResolution",
+                                    type: "Rational"
+                                },
+                                284: {
+                                    name: "PlanarConfiguration",
+                                    type: "Short"
+                                },
+                                290: {
+                                    name: "GrayResponseUnit",
+                                    type: "Short"
+                                },
+                                291: {
+                                    name: "GrayResponseCurve",
+                                    type: "Short"
+                                },
+                                292: {
+                                    name: "T4Options",
+                                    type: "Long"
+                                },
+                                293: {
+                                    name: "T6Options",
+                                    type: "Long"
+                                },
+                                296: {
+                                    name: "ResolutionUnit",
+                                    type: "Short"
+                                },
+                                301: {
+                                    name: "TransferFunction",
+                                    type: "Short"
+                                },
+                                305: {
+                                    name: "Software",
+                                    type: "Ascii"
+                                },
+                                306: {
+                                    name: "DateTime",
+                                    type: "Ascii"
+                                },
+                                315: {
+                                    name: "Artist",
+                                    type: "Ascii"
+                                },
+                                316: {
+                                    name: "HostComputer",
+                                    type: "Ascii"
+                                },
+                                317: {
+                                    name: "Predictor",
+                                    type: "Short"
+                                },
+                                318: {
+                                    name: "WhitePoint",
+                                    type: "Rational"
+                                },
+                                319: {
+                                    name: "PrimaryChromaticities",
+                                    type: "Rational"
+                                },
+                                320: {
+                                    name: "ColorMap",
+                                    type: "Short"
+                                },
+                                321: {
+                                    name: "HalftoneHints",
+                                    type: "Short"
+                                },
+                                322: {
+                                    name: "TileWidth",
+                                    type: "Short"
+                                },
+                                323: {
+                                    name: "TileLength",
+                                    type: "Short"
+                                },
+                                324: {
+                                    name: "TileOffsets",
+                                    type: "Short"
+                                },
+                                325: {
+                                    name: "TileByteCounts",
+                                    type: "Short"
+                                },
+                                330: {
+                                    name: "SubIFDs",
+                                    type: "Long"
+                                },
+                                332: {
+                                    name: "InkSet",
+                                    type: "Short"
+                                },
+                                333: {
+                                    name: "InkNames",
+                                    type: "Ascii"
+                                },
+                                334: {
+                                    name: "NumberOfInks",
+                                    type: "Short"
+                                },
+                                336: {
+                                    name: "DotRange",
+                                    type: "Byte"
+                                },
+                                337: {
+                                    name: "TargetPrinter",
+                                    type: "Ascii"
+                                },
+                                338: {
+                                    name: "ExtraSamples",
+                                    type: "Short"
+                                },
+                                339: {
+                                    name: "SampleFormat",
+                                    type: "Short"
+                                },
+                                340: {
+                                    name: "SMinSampleValue",
+                                    type: "Short"
+                                },
+                                341: {
+                                    name: "SMaxSampleValue",
+                                    type: "Short"
+                                },
+                                342: {
+                                    name: "TransferRange",
+                                    type: "Short"
+                                },
+                                343: {
+                                    name: "ClipPath",
+                                    type: "Byte"
+                                },
+                                344: {
+                                    name: "XClipPathUnits",
+                                    type: "Long"
+                                },
+                                345: {
+                                    name: "YClipPathUnits",
+                                    type: "Long"
+                                },
+                                346: {
+                                    name: "Indexed",
+                                    type: "Short"
+                                },
+                                347: {
+                                    name: "JPEGTables",
+                                    type: "Undefined"
+                                },
+                                351: {
+                                    name: "OPIProxy",
+                                    type: "Short"
+                                },
+                                512: {
+                                    name: "JPEGProc",
+                                    type: "Long"
+                                },
+                                513: {
+                                    name: "JPEGInterchangeFormat",
+                                    type: "Long"
+                                },
+                                514: {
+                                    name: "JPEGInterchangeFormatLength",
+                                    type: "Long"
+                                },
+                                515: {
+                                    name: "JPEGRestartInterval",
+                                    type: "Short"
+                                },
+                                517: {
+                                    name: "JPEGLosslessPredictors",
+                                    type: "Short"
+                                },
+                                518: {
+                                    name: "JPEGPointTransforms",
+                                    type: "Short"
+                                },
+                                519: {
+                                    name: "JPEGQTables",
+                                    type: "Long"
+                                },
+                                520: {
+                                    name: "JPEGDCTables",
+                                    type: "Long"
+                                },
+                                521: {
+                                    name: "JPEGACTables",
+                                    type: "Long"
+                                },
+                                529: {
+                                    name: "YCbCrCoefficients",
+                                    type: "Rational"
+                                },
+                                530: {
+                                    name: "YCbCrSubSampling",
+                                    type: "Short"
+                                },
+                                531: {
+                                    name: "YCbCrPositioning",
+                                    type: "Short"
+                                },
+                                532: {
+                                    name: "ReferenceBlackWhite",
+                                    type: "Rational"
+                                },
+                                700: {
+                                    name: "XMLPacket",
+                                    type: "Byte"
+                                },
+                                18246: {
+                                    name: "Rating",
+                                    type: "Short"
+                                },
+                                18249: {
+                                    name: "RatingPercent",
+                                    type: "Short"
+                                },
+                                32781: {
+                                    name: "ImageID",
+                                    type: "Ascii"
+                                },
+                                33421: {
+                                    name: "CFARepeatPatternDim",
+                                    type: "Short"
+                                },
+                                33422: {
+                                    name: "CFAPattern",
+                                    type: "Byte"
+                                },
+                                33423: {
+                                    name: "BatteryLevel",
+                                    type: "Rational"
+                                },
+                                33432: {
+                                    name: "Copyright",
+                                    type: "Ascii"
+                                },
+                                33434: {
+                                    name: "ExposureTime",
+                                    type: "Rational"
+                                },
+                                34377: {
+                                    name: "ImageResources",
+                                    type: "Byte"
+                                },
+                                34665: {
+                                    name: "ExifTag",
+                                    type: "Long"
+                                },
+                                34675: {
+                                    name: "InterColorProfile",
+                                    type: "Undefined"
+                                },
+                                34853: {
+                                    name: "GPSTag",
+                                    type: "Long"
+                                },
+                                34857: {
+                                    name: "Interlace",
+                                    type: "Short"
+                                },
+                                34858: {
+                                    name: "TimeZoneOffset",
+                                    type: "Long"
+                                },
+                                34859: {
+                                    name: "SelfTimerMode",
+                                    type: "Short"
+                                },
+                                37387: {
+                                    name: "FlashEnergy",
+                                    type: "Rational"
+                                },
+                                37388: {
+                                    name: "SpatialFrequencyResponse",
+                                    type: "Undefined"
+                                },
+                                37389: {
+                                    name: "Noise",
+                                    type: "Undefined"
+                                },
+                                37390: {
+                                    name: "FocalPlaneXResolution",
+                                    type: "Rational"
+                                },
+                                37391: {
+                                    name: "FocalPlaneYResolution",
+                                    type: "Rational"
+                                },
+                                37392: {
+                                    name: "FocalPlaneResolutionUnit",
+                                    type: "Short"
+                                },
+                                37393: {
+                                    name: "ImageNumber",
+                                    type: "Long"
+                                },
+                                37394: {
+                                    name: "SecurityClassification",
+                                    type: "Ascii"
+                                },
+                                37395: {
+                                    name: "ImageHistory",
+                                    type: "Ascii"
+                                },
+                                37397: {
+                                    name: "ExposureIndex",
+                                    type: "Rational"
+                                },
+                                37398: {
+                                    name: "TIFFEPStandardID",
+                                    type: "Byte"
+                                },
+                                37399: {
+                                    name: "SensingMethod",
+                                    type: "Short"
+                                },
+                                40091: {
+                                    name: "XPTitle",
+                                    type: "Byte"
+                                },
+                                40092: {
+                                    name: "XPComment",
+                                    type: "Byte"
+                                },
+                                40093: {
+                                    name: "XPAuthor",
+                                    type: "Byte"
+                                },
+                                40094: {
+                                    name: "XPKeywords",
+                                    type: "Byte"
+                                },
+                                40095: {
+                                    name: "XPSubject",
+                                    type: "Byte"
+                                },
+                                50341: {
+                                    name: "PrintImageMatching",
+                                    type: "Undefined"
+                                },
+                                50706: {
+                                    name: "DNGVersion",
+                                    type: "Byte"
+                                },
+                                50707: {
+                                    name: "DNGBackwardVersion",
+                                    type: "Byte"
+                                },
+                                50708: {
+                                    name: "UniqueCameraModel",
+                                    type: "Ascii"
+                                },
+                                50709: {
+                                    name: "LocalizedCameraModel",
+                                    type: "Byte"
+                                },
+                                50710: {
+                                    name: "CFAPlaneColor",
+                                    type: "Byte"
+                                },
+                                50711: {
+                                    name: "CFALayout",
+                                    type: "Short"
+                                },
+                                50712: {
+                                    name: "LinearizationTable",
+                                    type: "Short"
+                                },
+                                50713: {
+                                    name: "BlackLevelRepeatDim",
+                                    type: "Short"
+                                },
+                                50714: {
+                                    name: "BlackLevel",
+                                    type: "Rational"
+                                },
+                                50715: {
+                                    name: "BlackLevelDeltaH",
+                                    type: "SRational"
+                                },
+                                50716: {
+                                    name: "BlackLevelDeltaV",
+                                    type: "SRational"
+                                },
+                                50717: {
+                                    name: "WhiteLevel",
+                                    type: "Short"
+                                },
+                                50718: {
+                                    name: "DefaultScale",
+                                    type: "Rational"
+                                },
+                                50719: {
+                                    name: "DefaultCropOrigin",
+                                    type: "Short"
+                                },
+                                50720: {
+                                    name: "DefaultCropSize",
+                                    type: "Short"
+                                },
+                                50721: {
+                                    name: "ColorMatrix1",
+                                    type: "SRational"
+                                },
+                                50722: {
+                                    name: "ColorMatrix2",
+                                    type: "SRational"
+                                },
+                                50723: {
+                                    name: "CameraCalibration1",
+                                    type: "SRational"
+                                },
+                                50724: {
+                                    name: "CameraCalibration2",
+                                    type: "SRational"
+                                },
+                                50725: {
+                                    name: "ReductionMatrix1",
+                                    type: "SRational"
+                                },
+                                50726: {
+                                    name: "ReductionMatrix2",
+                                    type: "SRational"
+                                },
+                                50727: {
+                                    name: "AnalogBalance",
+                                    type: "Rational"
+                                },
+                                50728: {
+                                    name: "AsShotNeutral",
+                                    type: "Short"
+                                },
+                                50729: {
+                                    name: "AsShotWhiteXY",
+                                    type: "Rational"
+                                },
+                                50730: {
+                                    name: "BaselineExposure",
+                                    type: "SRational"
+                                },
+                                50731: {
+                                    name: "BaselineNoise",
+                                    type: "Rational"
+                                },
+                                50732: {
+                                    name: "BaselineSharpness",
+                                    type: "Rational"
+                                },
+                                50733: {
+                                    name: "BayerGreenSplit",
+                                    type: "Long"
+                                },
+                                50734: {
+                                    name: "LinearResponseLimit",
+                                    type: "Rational"
+                                },
+                                50735: {
+                                    name: "CameraSerialNumber",
+                                    type: "Ascii"
+                                },
+                                50736: {
+                                    name: "LensInfo",
+                                    type: "Rational"
+                                },
+                                50737: {
+                                    name: "ChromaBlurRadius",
+                                    type: "Rational"
+                                },
+                                50738: {
+                                    name: "AntiAliasStrength",
+                                    type: "Rational"
+                                },
+                                50739: {
+                                    name: "ShadowScale",
+                                    type: "SRational"
+                                },
+                                50740: {
+                                    name: "DNGPrivateData",
+                                    type: "Byte"
+                                },
+                                50741: {
+                                    name: "MakerNoteSafety",
+                                    type: "Short"
+                                },
+                                50778: {
+                                    name: "CalibrationIlluminant1",
+                                    type: "Short"
+                                },
+                                50779: {
+                                    name: "CalibrationIlluminant2",
+                                    type: "Short"
+                                },
+                                50780: {
+                                    name: "BestQualityScale",
+                                    type: "Rational"
+                                },
+                                50781: {
+                                    name: "RawDataUniqueID",
+                                    type: "Byte"
+                                },
+                                50827: {
+                                    name: "OriginalRawFileName",
+                                    type: "Byte"
+                                },
+                                50828: {
+                                    name: "OriginalRawFileData",
+                                    type: "Undefined"
+                                },
+                                50829: {
+                                    name: "ActiveArea",
+                                    type: "Short"
+                                },
+                                50830: {
+                                    name: "MaskedAreas",
+                                    type: "Short"
+                                },
+                                50831: {
+                                    name: "AsShotICCProfile",
+                                    type: "Undefined"
+                                },
+                                50832: {
+                                    name: "AsShotPreProfileMatrix",
+                                    type: "SRational"
+                                },
+                                50833: {
+                                    name: "CurrentICCProfile",
+                                    type: "Undefined"
+                                },
+                                50834: {
+                                    name: "CurrentPreProfileMatrix",
+                                    type: "SRational"
+                                },
+                                50879: {
+                                    name: "ColorimetricReference",
+                                    type: "Short"
+                                },
+                                50931: {
+                                    name: "CameraCalibrationSignature",
+                                    type: "Byte"
+                                },
+                                50932: {
+                                    name: "ProfileCalibrationSignature",
+                                    type: "Byte"
+                                },
+                                50934: {
+                                    name: "AsShotProfileName",
+                                    type: "Byte"
+                                },
+                                50935: {
+                                    name: "NoiseReductionApplied",
+                                    type: "Rational"
+                                },
+                                50936: {
+                                    name: "ProfileName",
+                                    type: "Byte"
+                                },
+                                50937: {
+                                    name: "ProfileHueSatMapDims",
+                                    type: "Long"
+                                },
+                                50938: {
+                                    name: "ProfileHueSatMapData1",
+                                    type: "Float"
+                                },
+                                50939: {
+                                    name: "ProfileHueSatMapData2",
+                                    type: "Float"
+                                },
+                                50940: {
+                                    name: "ProfileToneCurve",
+                                    type: "Float"
+                                },
+                                50941: {
+                                    name: "ProfileEmbedPolicy",
+                                    type: "Long"
+                                },
+                                50942: {
+                                    name: "ProfileCopyright",
+                                    type: "Byte"
+                                },
+                                50964: {
+                                    name: "ForwardMatrix1",
+                                    type: "SRational"
+                                },
+                                50965: {
+                                    name: "ForwardMatrix2",
+                                    type: "SRational"
+                                },
+                                50966: {
+                                    name: "PreviewApplicationName",
+                                    type: "Byte"
+                                },
+                                50967: {
+                                    name: "PreviewApplicationVersion",
+                                    type: "Byte"
+                                },
+                                50968: {
+                                    name: "PreviewSettingsName",
+                                    type: "Byte"
+                                },
+                                50969: {
+                                    name: "PreviewSettingsDigest",
+                                    type: "Byte"
+                                },
+                                50970: {
+                                    name: "PreviewColorSpace",
+                                    type: "Long"
+                                },
+                                50971: {
+                                    name: "PreviewDateTime",
+                                    type: "Ascii"
+                                },
+                                50972: {
+                                    name: "RawImageDigest",
+                                    type: "Undefined"
+                                },
+                                50973: {
+                                    name: "OriginalRawFileDigest",
+                                    type: "Undefined"
+                                },
+                                50974: {
+                                    name: "SubTileBlockSize",
+                                    type: "Long"
+                                },
+                                50975: {
+                                    name: "RowInterleaveFactor",
+                                    type: "Long"
+                                },
+                                50981: {
+                                    name: "ProfileLookTableDims",
+                                    type: "Long"
+                                },
+                                50982: {
+                                    name: "ProfileLookTableData",
+                                    type: "Float"
+                                },
+                                51008: {
+                                    name: "OpcodeList1",
+                                    type: "Undefined"
+                                },
+                                51009: {
+                                    name: "OpcodeList2",
+                                    type: "Undefined"
+                                },
+                                51022: {
+                                    name: "OpcodeList3",
+                                    type: "Undefined"
+                                }
+                            },
+                            Exif: {
+                                33434: {
+                                    name: "ExposureTime",
+                                    type: "Rational"
+                                },
+                                33437: {
+                                    name: "FNumber",
+                                    type: "Rational"
+                                },
+                                34850: {
+                                    name: "ExposureProgram",
+                                    type: "Short"
+                                },
+                                34852: {
+                                    name: "SpectralSensitivity",
+                                    type: "Ascii"
+                                },
+                                34855: {
+                                    name: "ISOSpeedRatings",
+                                    type: "Short"
+                                },
+                                34856: {
+                                    name: "OECF",
+                                    type: "Undefined"
+                                },
+                                34864: {
+                                    name: "SensitivityType",
+                                    type: "Short"
+                                },
+                                34865: {
+                                    name: "StandardOutputSensitivity",
+                                    type: "Long"
+                                },
+                                34866: {
+                                    name: "RecommendedExposureIndex",
+                                    type: "Long"
+                                },
+                                34867: {
+                                    name: "ISOSpeed",
+                                    type: "Long"
+                                },
+                                34868: {
+                                    name: "ISOSpeedLatitudeyyy",
+                                    type: "Long"
+                                },
+                                34869: {
+                                    name: "ISOSpeedLatitudezzz",
+                                    type: "Long"
+                                },
+                                36864: {
+                                    name: "ExifVersion",
+                                    type: "Undefined"
+                                },
+                                36867: {
+                                    name: "DateTimeOriginal",
+                                    type: "Ascii"
+                                },
+                                36868: {
+                                    name: "DateTimeDigitized",
+                                    type: "Ascii"
+                                },
+                                37121: {
+                                    name: "ComponentsConfiguration",
+                                    type: "Undefined"
+                                },
+                                37122: {
+                                    name: "CompressedBitsPerPixel",
+                                    type: "Rational"
+                                },
+                                37377: {
+                                    name: "ShutterSpeedValue",
+                                    type: "SRational"
+                                },
+                                37378: {
+                                    name: "ApertureValue",
+                                    type: "Rational"
+                                },
+                                37379: {
+                                    name: "BrightnessValue",
+                                    type: "SRational"
+                                },
+                                37380: {
+                                    name: "ExposureBiasValue",
+                                    type: "SRational"
+                                },
+                                37381: {
+                                    name: "MaxApertureValue",
+                                    type: "Rational"
+                                },
+                                37382: {
+                                    name: "SubjectDistance",
+                                    type: "Rational"
+                                },
+                                37383: {
+                                    name: "MeteringMode",
+                                    type: "Short"
+                                },
+                                37384: {
+                                    name: "LightSource",
+                                    type: "Short"
+                                },
+                                37385: {
+                                    name: "Flash",
+                                    type: "Short"
+                                },
+                                37386: {
+                                    name: "FocalLength",
+                                    type: "Rational"
+                                },
+                                37396: {
+                                    name: "SubjectArea",
+                                    type: "Short"
+                                },
+                                37500: {
+                                    name: "MakerNote",
+                                    type: "Undefined"
+                                },
+                                37510: {
+                                    name: "UserComment",
+                                    type: "Ascii"
+                                },
+                                37520: {
+                                    name: "SubSecTime",
+                                    type: "Ascii"
+                                },
+                                37521: {
+                                    name: "SubSecTimeOriginal",
+                                    type: "Ascii"
+                                },
+                                37522: {
+                                    name: "SubSecTimeDigitized",
+                                    type: "Ascii"
+                                },
+                                40960: {
+                                    name: "FlashpixVersion",
+                                    type: "Undefined"
+                                },
+                                40961: {
+                                    name: "ColorSpace",
+                                    type: "Short"
+                                },
+                                40962: {
+                                    name: "PixelXDimension",
+                                    type: "Long"
+                                },
+                                40963: {
+                                    name: "PixelYDimension",
+                                    type: "Long"
+                                },
+                                40964: {
+                                    name: "RelatedSoundFile",
+                                    type: "Ascii"
+                                },
+                                40965: {
+                                    name: "InteroperabilityTag",
+                                    type: "Long"
+                                },
+                                41483: {
+                                    name: "FlashEnergy",
+                                    type: "Rational"
+                                },
+                                41484: {
+                                    name: "SpatialFrequencyResponse",
+                                    type: "Undefined"
+                                },
+                                41486: {
+                                    name: "FocalPlaneXResolution",
+                                    type: "Rational"
+                                },
+                                41487: {
+                                    name: "FocalPlaneYResolution",
+                                    type: "Rational"
+                                },
+                                41488: {
+                                    name: "FocalPlaneResolutionUnit",
+                                    type: "Short"
+                                },
+                                41492: {
+                                    name: "SubjectLocation",
+                                    type: "Short"
+                                },
+                                41493: {
+                                    name: "ExposureIndex",
+                                    type: "Rational"
+                                },
+                                41495: {
+                                    name: "SensingMethod",
+                                    type: "Short"
+                                },
+                                41728: {
+                                    name: "FileSource",
+                                    type: "Undefined"
+                                },
+                                41729: {
+                                    name: "SceneType",
+                                    type: "Undefined"
+                                },
+                                41730: {
+                                    name: "CFAPattern",
+                                    type: "Undefined"
+                                },
+                                41985: {
+                                    name: "CustomRendered",
+                                    type: "Short"
+                                },
+                                41986: {
+                                    name: "ExposureMode",
+                                    type: "Short"
+                                },
+                                41987: {
+                                    name: "WhiteBalance",
+                                    type: "Short"
+                                },
+                                41988: {
+                                    name: "DigitalZoomRatio",
+                                    type: "Rational"
+                                },
+                                41989: {
+                                    name: "FocalLengthIn35mmFilm",
+                                    type: "Short"
+                                },
+                                41990: {
+                                    name: "SceneCaptureType",
+                                    type: "Short"
+                                },
+                                41991: {
+                                    name: "GainControl",
+                                    type: "Short"
+                                },
+                                41992: {
+                                    name: "Contrast",
+                                    type: "Short"
+                                },
+                                41993: {
+                                    name: "Saturation",
+                                    type: "Short"
+                                },
+                                41994: {
+                                    name: "Sharpness",
+                                    type: "Short"
+                                },
+                                41995: {
+                                    name: "DeviceSettingDescription",
+                                    type: "Undefined"
+                                },
+                                41996: {
+                                    name: "SubjectDistanceRange",
+                                    type: "Short"
+                                },
+                                42016: {
+                                    name: "ImageUniqueID",
+                                    type: "Ascii"
+                                },
+                                42032: {
+                                    name: "CameraOwnerName",
+                                    type: "Ascii"
+                                },
+                                42033: {
+                                    name: "BodySerialNumber",
+                                    type: "Ascii"
+                                },
+                                42034: {
+                                    name: "LensSpecification",
+                                    type: "Rational"
+                                },
+                                42035: {
+                                    name: "LensMake",
+                                    type: "Ascii"
+                                },
+                                42036: {
+                                    name: "LensModel",
+                                    type: "Ascii"
+                                },
+                                42037: {
+                                    name: "LensSerialNumber",
+                                    type: "Ascii"
+                                },
+                                42240: {
+                                    name: "Gamma",
+                                    type: "Rational"
+                                }
+                            },
+                            GPS: {
+                                0: {
+                                    name: "GPSVersionID",
+                                    type: "Byte"
+                                },
+                                1: {
+                                    name: "GPSLatitudeRef",
+                                    type: "Ascii"
+                                },
+                                2: {
+                                    name: "GPSLatitude",
+                                    type: "Rational"
+                                },
+                                3: {
+                                    name: "GPSLongitudeRef",
+                                    type: "Ascii"
+                                },
+                                4: {
+                                    name: "GPSLongitude",
+                                    type: "Rational"
+                                },
+                                5: {
+                                    name: "GPSAltitudeRef",
+                                    type: "Byte"
+                                },
+                                6: {
+                                    name: "GPSAltitude",
+                                    type: "Rational"
+                                },
+                                7: {
+                                    name: "GPSTimeStamp",
+                                    type: "Rational"
+                                },
+                                8: {
+                                    name: "GPSSatellites",
+                                    type: "Ascii"
+                                },
+                                9: {
+                                    name: "GPSStatus",
+                                    type: "Ascii"
+                                },
+                                10: {
+                                    name: "GPSMeasureMode",
+                                    type: "Ascii"
+                                },
+                                11: {
+                                    name: "GPSDOP",
+                                    type: "Rational"
+                                },
+                                12: {
+                                    name: "GPSSpeedRef",
+                                    type: "Ascii"
+                                },
+                                13: {
+                                    name: "GPSSpeed",
+                                    type: "Rational"
+                                },
+                                14: {
+                                    name: "GPSTrackRef",
+                                    type: "Ascii"
+                                },
+                                15: {
+                                    name: "GPSTrack",
+                                    type: "Rational"
+                                },
+                                16: {
+                                    name: "GPSImgDirectionRef",
+                                    type: "Ascii"
+                                },
+                                17: {
+                                    name: "GPSImgDirection",
+                                    type: "Rational"
+                                },
+                                18: {
+                                    name: "GPSMapDatum",
+                                    type: "Ascii"
+                                },
+                                19: {
+                                    name: "GPSDestLatitudeRef",
+                                    type: "Ascii"
+                                },
+                                20: {
+                                    name: "GPSDestLatitude",
+                                    type: "Rational"
+                                },
+                                21: {
+                                    name: "GPSDestLongitudeRef",
+                                    type: "Ascii"
+                                },
+                                22: {
+                                    name: "GPSDestLongitude",
+                                    type: "Rational"
+                                },
+                                23: {
+                                    name: "GPSDestBearingRef",
+                                    type: "Ascii"
+                                },
+                                24: {
+                                    name: "GPSDestBearing",
+                                    type: "Rational"
+                                },
+                                25: {
+                                    name: "GPSDestDistanceRef",
+                                    type: "Ascii"
+                                },
+                                26: {
+                                    name: "GPSDestDistance",
+                                    type: "Rational"
+                                },
+                                27: {
+                                    name: "GPSProcessingMethod",
+                                    type: "Undefined"
+                                },
+                                28: {
+                                    name: "GPSAreaInformation",
+                                    type: "Undefined"
+                                },
+                                29: {
+                                    name: "GPSDateStamp",
+                                    type: "Ascii"
+                                },
+                                30: {
+                                    name: "GPSDifferential",
+                                    type: "Short"
+                                },
+                                31: {
+                                    name: "GPSHPositioningError",
+                                    type: "Rational"
+                                }
+                            },
+                            Interop: {
+                                1: {
+                                    name: "InteroperabilityIndex",
+                                    type: "Ascii"
+                                }
+                            }
+                        };
+                    g["0th"] = g.Image, g["1st"] = g.Image, n.TAGS = g, n.ImageIFD = {
+                        ProcessingSoftware: 11,
+                        NewSubfileType: 254,
+                        SubfileType: 255,
+                        ImageWidth: 256,
+                        ImageLength: 257,
+                        BitsPerSample: 258,
+                        Compression: 259,
+                        PhotometricInterpretation: 262,
+                        Threshholding: 263,
+                        CellWidth: 264,
+                        CellLength: 265,
+                        FillOrder: 266,
+                        DocumentName: 269,
+                        ImageDescription: 270,
+                        Make: 271,
+                        Model: 272,
+                        StripOffsets: 273,
+                        Orientation: 274,
+                        SamplesPerPixel: 277,
+                        RowsPerStrip: 278,
+                        StripByteCounts: 279,
+                        XResolution: 282,
+                        YResolution: 283,
+                        PlanarConfiguration: 284,
+                        GrayResponseUnit: 290,
+                        GrayResponseCurve: 291,
+                        T4Options: 292,
+                        T6Options: 293,
+                        ResolutionUnit: 296,
+                        TransferFunction: 301,
+                        Software: 305,
+                        DateTime: 306,
+                        Artist: 315,
+                        HostComputer: 316,
+                        Predictor: 317,
+                        WhitePoint: 318,
+                        PrimaryChromaticities: 319,
+                        ColorMap: 320,
+                        HalftoneHints: 321,
+                        TileWidth: 322,
+                        TileLength: 323,
+                        TileOffsets: 324,
+                        TileByteCounts: 325,
+                        SubIFDs: 330,
+                        InkSet: 332,
+                        InkNames: 333,
+                        NumberOfInks: 334,
+                        DotRange: 336,
+                        TargetPrinter: 337,
+                        ExtraSamples: 338,
+                        SampleFormat: 339,
+                        SMinSampleValue: 340,
+                        SMaxSampleValue: 341,
+                        TransferRange: 342,
+                        ClipPath: 343,
+                        XClipPathUnits: 344,
+                        YClipPathUnits: 345,
+                        Indexed: 346,
+                        JPEGTables: 347,
+                        OPIProxy: 351,
+                        JPEGProc: 512,
+                        JPEGInterchangeFormat: 513,
+                        JPEGInterchangeFormatLength: 514,
+                        JPEGRestartInterval: 515,
+                        JPEGLosslessPredictors: 517,
+                        JPEGPointTransforms: 518,
+                        JPEGQTables: 519,
+                        JPEGDCTables: 520,
+                        JPEGACTables: 521,
+                        YCbCrCoefficients: 529,
+                        YCbCrSubSampling: 530,
+                        YCbCrPositioning: 531,
+                        ReferenceBlackWhite: 532,
+                        XMLPacket: 700,
+                        Rating: 18246,
+                        RatingPercent: 18249,
+                        ImageID: 32781,
+                        CFARepeatPatternDim: 33421,
+                        CFAPattern: 33422,
+                        BatteryLevel: 33423,
+                        Copyright: 33432,
+                        ExposureTime: 33434,
+                        ImageResources: 34377,
+                        ExifTag: 34665,
+                        InterColorProfile: 34675,
+                        GPSTag: 34853,
+                        Interlace: 34857,
+                        TimeZoneOffset: 34858,
+                        SelfTimerMode: 34859,
+                        FlashEnergy: 37387,
+                        SpatialFrequencyResponse: 37388,
+                        Noise: 37389,
+                        FocalPlaneXResolution: 37390,
+                        FocalPlaneYResolution: 37391,
+                        FocalPlaneResolutionUnit: 37392,
+                        ImageNumber: 37393,
+                        SecurityClassification: 37394,
+                        ImageHistory: 37395,
+                        ExposureIndex: 37397,
+                        TIFFEPStandardID: 37398,
+                        SensingMethod: 37399,
+                        XPTitle: 40091,
+                        XPComment: 40092,
+                        XPAuthor: 40093,
+                        XPKeywords: 40094,
+                        XPSubject: 40095,
+                        PrintImageMatching: 50341,
+                        DNGVersion: 50706,
+                        DNGBackwardVersion: 50707,
+                        UniqueCameraModel: 50708,
+                        LocalizedCameraModel: 50709,
+                        CFAPlaneColor: 50710,
+                        CFALayout: 50711,
+                        LinearizationTable: 50712,
+                        BlackLevelRepeatDim: 50713,
+                        BlackLevel: 50714,
+                        BlackLevelDeltaH: 50715,
+                        BlackLevelDeltaV: 50716,
+                        WhiteLevel: 50717,
+                        DefaultScale: 50718,
+                        DefaultCropOrigin: 50719,
+                        DefaultCropSize: 50720,
+                        ColorMatrix1: 50721,
+                        ColorMatrix2: 50722,
+                        CameraCalibration1: 50723,
+                        CameraCalibration2: 50724,
+                        ReductionMatrix1: 50725,
+                        ReductionMatrix2: 50726,
+                        AnalogBalance: 50727,
+                        AsShotNeutral: 50728,
+                        AsShotWhiteXY: 50729,
+                        BaselineExposure: 50730,
+                        BaselineNoise: 50731,
+                        BaselineSharpness: 50732,
+                        BayerGreenSplit: 50733,
+                        LinearResponseLimit: 50734,
+                        CameraSerialNumber: 50735,
+                        LensInfo: 50736,
+                        ChromaBlurRadius: 50737,
+                        AntiAliasStrength: 50738,
+                        ShadowScale: 50739,
+                        DNGPrivateData: 50740,
+                        MakerNoteSafety: 50741,
+                        CalibrationIlluminant1: 50778,
+                        CalibrationIlluminant2: 50779,
+                        BestQualityScale: 50780,
+                        RawDataUniqueID: 50781,
+                        OriginalRawFileName: 50827,
+                        OriginalRawFileData: 50828,
+                        ActiveArea: 50829,
+                        MaskedAreas: 50830,
+                        AsShotICCProfile: 50831,
+                        AsShotPreProfileMatrix: 50832,
+                        CurrentICCProfile: 50833,
+                        CurrentPreProfileMatrix: 50834,
+                        ColorimetricReference: 50879,
+                        CameraCalibrationSignature: 50931,
+                        ProfileCalibrationSignature: 50932,
+                        AsShotProfileName: 50934,
+                        NoiseReductionApplied: 50935,
+                        ProfileName: 50936,
+                        ProfileHueSatMapDims: 50937,
+                        ProfileHueSatMapData1: 50938,
+                        ProfileHueSatMapData2: 50939,
+                        ProfileToneCurve: 50940,
+                        ProfileEmbedPolicy: 50941,
+                        ProfileCopyright: 50942,
+                        ForwardMatrix1: 50964,
+                        ForwardMatrix2: 50965,
+                        PreviewApplicationName: 50966,
+                        PreviewApplicationVersion: 50967,
+                        PreviewSettingsName: 50968,
+                        PreviewSettingsDigest: 50969,
+                        PreviewColorSpace: 50970,
+                        PreviewDateTime: 50971,
+                        RawImageDigest: 50972,
+                        OriginalRawFileDigest: 50973,
+                        SubTileBlockSize: 50974,
+                        RowInterleaveFactor: 50975,
+                        ProfileLookTableDims: 50981,
+                        ProfileLookTableData: 50982,
+                        OpcodeList1: 51008,
+                        OpcodeList2: 51009,
+                        OpcodeList3: 51022,
+                        NoiseProfile: 51041
+                    }, n.ExifIFD = {
+                        ExposureTime: 33434,
+                        FNumber: 33437,
+                        ExposureProgram: 34850,
+                        SpectralSensitivity: 34852,
+                        ISOSpeedRatings: 34855,
+                        OECF: 34856,
+                        SensitivityType: 34864,
+                        StandardOutputSensitivity: 34865,
+                        RecommendedExposureIndex: 34866,
+                        ISOSpeed: 34867,
+                        ISOSpeedLatitudeyyy: 34868,
+                        ISOSpeedLatitudezzz: 34869,
+                        ExifVersion: 36864,
+                        DateTimeOriginal: 36867,
+                        DateTimeDigitized: 36868,
+                        ComponentsConfiguration: 37121,
+                        CompressedBitsPerPixel: 37122,
+                        ShutterSpeedValue: 37377,
+                        ApertureValue: 37378,
+                        BrightnessValue: 37379,
+                        ExposureBiasValue: 37380,
+                        MaxApertureValue: 37381,
+                        SubjectDistance: 37382,
+                        MeteringMode: 37383,
+                        LightSource: 37384,
+                        Flash: 37385,
+                        FocalLength: 37386,
+                        SubjectArea: 37396,
+                        MakerNote: 37500,
+                        UserComment: 37510,
+                        SubSecTime: 37520,
+                        SubSecTimeOriginal: 37521,
+                        SubSecTimeDigitized: 37522,
+                        FlashpixVersion: 40960,
+                        ColorSpace: 40961,
+                        PixelXDimension: 40962,
+                        PixelYDimension: 40963,
+                        RelatedSoundFile: 40964,
+                        InteroperabilityTag: 40965,
+                        FlashEnergy: 41483,
+                        SpatialFrequencyResponse: 41484,
+                        FocalPlaneXResolution: 41486,
+                        FocalPlaneYResolution: 41487,
+                        FocalPlaneResolutionUnit: 41488,
+                        SubjectLocation: 41492,
+                        ExposureIndex: 41493,
+                        SensingMethod: 41495,
+                        FileSource: 41728,
+                        SceneType: 41729,
+                        CFAPattern: 41730,
+                        CustomRendered: 41985,
+                        ExposureMode: 41986,
+                        WhiteBalance: 41987,
+                        DigitalZoomRatio: 41988,
+                        FocalLengthIn35mmFilm: 41989,
+                        SceneCaptureType: 41990,
+                        GainControl: 41991,
+                        Contrast: 41992,
+                        Saturation: 41993,
+                        Sharpness: 41994,
+                        DeviceSettingDescription: 41995,
+                        SubjectDistanceRange: 41996,
+                        ImageUniqueID: 42016,
+                        CameraOwnerName: 42032,
+                        BodySerialNumber: 42033,
+                        LensSpecification: 42034,
+                        LensMake: 42035,
+                        LensModel: 42036,
+                        LensSerialNumber: 42037,
+                        Gamma: 42240
+                    }, n.GPSIFD = {
+                        GPSVersionID: 0,
+                        GPSLatitudeRef: 1,
+                        GPSLatitude: 2,
+                        GPSLongitudeRef: 3,
+                        GPSLongitude: 4,
+                        GPSAltitudeRef: 5,
+                        GPSAltitude: 6,
+                        GPSTimeStamp: 7,
+                        GPSSatellites: 8,
+                        GPSStatus: 9,
+                        GPSMeasureMode: 10,
+                        GPSDOP: 11,
+                        GPSSpeedRef: 12,
+                        GPSSpeed: 13,
+                        GPSTrackRef: 14,
+                        GPSTrack: 15,
+                        GPSImgDirectionRef: 16,
+                        GPSImgDirection: 17,
+                        GPSMapDatum: 18,
+                        GPSDestLatitudeRef: 19,
+                        GPSDestLatitude: 20,
+                        GPSDestLongitudeRef: 21,
+                        GPSDestLongitude: 22,
+                        GPSDestBearingRef: 23,
+                        GPSDestBearing: 24,
+                        GPSDestDistanceRef: 25,
+                        GPSDestDistance: 26,
+                        GPSProcessingMethod: 27,
+                        GPSAreaInformation: 28,
+                        GPSDateStamp: 29,
+                        GPSDifferential: 30,
+                        GPSHPositioningError: 31
+                    }, n.InteropIFD = {
+                        InteroperabilityIndex: 1
+                    }, n.GPSHelper = {
+                        degToDmsRational: function (e) {
+                            var t = Math.abs(e),
+                                n = t % 1 * 60,
+                                r = n % 1 * 60;
+                            return [
+                                [Math.floor(t), 1],
+                                [Math.floor(n), 1],
+                                [Math.round(100 * r), 100]
+                            ]
+                        },
+                        dmsRationalToDeg: function (e, t) {
+                            var n = "S" === t || "W" === t ? -1 : 1;
+                            return (e[0][0] / e[0][1] + e[1][0] / e[1][1] / 60 + e[2][0] / e[2][1] / 3600) * n
+                        }
+                    }, e.exports && (t = e.exports = n), t.piexif = n
+                }()
+            }));
+            const Iq = async (e, t, n) => {
+                const r = await (e => new Promise((t, n) => {
+                        const r = new FileReader;
+                        r.onloadend = () => t(r.result), r.readAsDataURL(e)
+                    }))(e),
+                    a = Sq.load(r),
+                    s = (await navigator.mediaDevices.enumerateDevices()).filter(e => "videoinput" === e.kind).map(e => e.label),
+                    i = {
+                        width: n.width,
+                        height: n.height,
+                        usedCamera: t.label,
+                        camerasOnDevice: s
+                    },
+                    o = {
+                        "0th": Object.assign(Object.assign({}, a["0th"]), {
+                            [Sq.ImageIFD.Model]: i.usedCamera,
+                            [Sq.ImageIFD.ImageWidth]: i.width,
+                            [Sq.ImageIFD.ImageLength]: i.height
+                        }),
+                        Exif: Object.assign(Object.assign({}, a.Exif), {
+                            [Sq.ExifIFD.UserComment]: JSON.stringify(i)
+                        }),
+                        GPS: Object.assign({}, a.GPS),
+                        Interop: Object.assign({}, a.Interop),
+                        "1st": Object.assign({}, a["1st"]),
+                        thumbnail: a.thumbnail
+                    },
+                    u = Sq.dump(o),
+                    l = Sq.insert(u, r);
+                return await fetch(l).then(e => e.blob())
+            }, Aq = new Image;
+            var Eq;
+            Aq.src = "data:image/svg+xml;base64,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",
+                function (e) {
+                    e.JPEG = "image/jpeg", e.PNG = "image/png"
+                }(Eq || (Eq = {}));
+            class Nq {
+                constructor(e, t) {
+                    this.showMask = !1, this.videoSize = {
+                        width: 0,
+                        height: 0
+                    }, this.streamPaused = !1, this.pauseStream = () => {
+                        this.streamPaused = !0, this.videoElement.pause(), this.dropMask(), this.faceDetection && kq.getInstance().stopDetector()
+                    }, this.device = e, this.detector = kq.getInstance(this, e.isMobile, t)
+                }
+                setShowMask(e) {
+                    this.showMask = e
+                }
+                setFaceDetection(e) {
+                    this.faceDetection = e
+                }
+                setProbabilityThreshold(e) {
+                    kq.getInstance().setProbabilityThreshold(e)
+                }
+                setCallbackErrors(e) {
+                    this.callbackErrors = e
+                }
+                setCallbackAutoCapturing(e) {
+                    this.callbackAutoCapturing = e
+                }
+                setCallbackFaceDetectionErrors(e) {
+                    this.callbackFaceDetectionErrors = e
+                }
+                static getInstance(e, t) {
+                    return Nq.instance || (Nq.instance = new Nq(e, t)), Nq.instance
+                }
+                returnErrors(e) {
+                    this.callbackFaceDetectionErrors(e)
+                }
+                autoCapturing() {
+                    this.callbackAutoCapturing()
+                }
+                updateHtmlElements(e, t) {
+                    this.videoElement = e, this.canvasElement = t, this.detector.updateHtmlElements(e)
+                }
+                async startDetection() {
+                    return this.detector.startDetector()
+                }
+                static orientationChange() {
+                    Nq.instance && Nq.getInstance().drawMask()
+                }
+                startStream(e) {
+                    this.stream && this.stream.getTracks().forEach(e => e.stop()), this.stream = e, "srcObject" in this.videoElement ? this.videoElement.srcObject = e : this.videoElement.src = window.URL.createObjectURL(e), this.videoElement.play().then(() => {
+                        this.streamPaused = !1, l.cameraReadyEvent(), this.drawMask(), this.faceDetection && this.startDetection()
+                    })
+                }
+                async initStream(e) {
+                    this.startStream(e)
+                }
+                async resumeStream() {
+                    this.streamPaused = !1, await this.videoElement.play(), this.drawMask(), this.faceDetection && await kq.getInstance().startDetector()
+                }
+                updateCanvasSize(e) {
+                    this.videoSize = {
+                        width: this.videoElement.videoWidth,
+                        height: this.videoElement.videoHeight
+                    }, e.width = this.videoSize.width, e.height = this.videoSize.height
+                }
+                dropStream() {
+                    this.streamStopped() || (this.stream.getTracks().forEach(e => e.stop()), this.videoElement.srcObject = null), this.faceDetection && kq.getInstance().stopDetector()
+                }
+                streamStopped() {
+                    return !(this.stream && this.stream.getTracks && this.stream.getTracks().length > 0)
+                }
+                drawMask() {
+                    this.showMask && !this.streamPaused && setTimeout(() => {
+                        this.updateCanvasSize(this.canvasElement);
+                        const e = this.canvasElement,
+                            t = e.getContext("2d"),
+                            n = e.width / Aq.width,
+                            r = e.height / Aq.height,
+                            a = Math.min(n, r),
+                            s = e.width < e.height,
+                            i = !s && this.device.isMobile ? 50 : 100,
+                            o = s ? 200 : 100,
+                            u = (e.width - Aq.width * a) / 2,
+                            l = (e.height - Aq.height * a) / 2;
+                        t.drawImage(Aq, 0, 0, Aq.width, Aq.height, u + i, l + o, Aq.width * a - 2 * i, Aq.height * a - 2 * o)
+                    }, 150)
+                }
+                dropMask() {
+                    this.canvasElement.getContext("2d").clearRect(0, 0, 1e4, 1e4)
+                }
+                takePhoto() {
+                    return new Promise(async e => {
+                        const t = document.createElement("canvas");
+                        t.width = this.videoElement.videoWidth, t.height = this.videoElement.videoHeight, e([await this.getFrame(t)])
+                    })
+                }
+                getFrame(e) {
+                    return new Promise(t => {
+                        e.getContext("2d").drawImage(this.videoElement, 0, 0, e.width, e.height), e.toBlob(e => {
+                            if (e.type !== Eq.JPEG || this.device.isIos) t(e);
+                            else try {
+                                Iq(e, this.stream.getTracks()[0], this.videoSize).then(e => t(e))
+                            } catch (n) {
+                                t(e), this.callbackErrors(n, !1)
+                            }
+                        }, Eq.PNG, 1)
+                    })
+                }
+            }
+            window.addEventListener("resize", Nq.orientationChange, !1), window.addEventListener("orientationchange", Nq.orientationChange, !1);
+            let Cq = class {
+                constructor(e) {
+                    Object(u.f)(this, e), this.eventVideoStarted = Object(u.b)(this, "videoStarted", 7), this.eventCloseCamera = Object(u.b)(this, "closeCamera", 7), this.errorCameraEvent = Object(u.b)(this, "errorCamera", 7), this.eventMakePhoto = Object(u.b)(this, "makePhoto", 7), this.eventTakePhoto = Object(u.b)(this, "takePhoto", 7), this.callbackErrors = (e, t) => {
+                        t ? (this.errorCameraEvent.emit(e), this.eventCloseCamera.emit()) : this.errorCameraEvent.emit(e)
+                    }, this.callbackAutoCapturing = () => {
+                        this.device.isMobile ? this.eventMakePhoto.emit() : this.eventTakePhoto.emit()
+                    }, this.callbackFaceDetectionErrors = e => {
+                        l.detectionError(e)
+                    }
+                }
+                componentDidLoad() {
+                    this.startStream()
+                }
+                render() {
+                    const e = "camera " + (this.device.isMobile ? "cameraMobile" : "");
+                    return Object(u.d)("div", {
+                        class: e
+                    }, Object(u.d)("video", {
+                        loop: !0,
+                        autoplay: !0,
+                        playsinline: !0,
+                        muted: !0,
+                        class: "cameraVideo",
+                        "aria-hidden": "true",
+                        ref: e => this.cameraVideo = e
+                    }), Object(u.d)("canvas", {
+                        class: "cameraCanvas",
+                        ref: e => this.cameraCanvas = e
+                    }))
+                }
+                startStream() {
+                    Nq.instance || Nq.getInstance(this.device, this.modelPath);
+                    const e = Nq.getInstance();
+                    e.updateHtmlElements(this.cameraVideo, this.cameraCanvas), e.setShowMask(this.showMask), e.setFaceDetection(this.faceDetection), e.setProbabilityThreshold(this.probabilityThreshold), e.setCallbackErrors(this.callbackErrors), e.setCallbackAutoCapturing(this.callbackAutoCapturing), e.setCallbackFaceDetectionErrors(this.callbackFaceDetectionErrors)
+                }
+            };
+            Cq.style = ".camera{width:100%;height:100%;color:white;display:flex;align-items:center;justify-content:center;position:relative}.cameraCanvas{transform:scale(-1, 1);-webkit-transform:scale(-1, 1);max-width:100%;max-height:100%;z-index:2}.cameraVideo{transform:scale(-1, 1);-webkit-transform:scale(-1, 1);z-index:1;position:absolute;max-width:100%;max-height:100%}.cameraMobile{position:fixed;top:0;left:0;background:black}.canvas-on-video{max-width:100%;max-height:100%;position:absolute;left:50%;top:50%;z-index:2;transform:scale(-1, 1)}";
+            let Tq = class {
+                constructor(e) {
+                    Object(u.f)(this, e), this.updateState = () => {
+                        this._showMask = "true" === this.show_mask, this._disable_control_panel = "true" === this.disable_control_panel, this._face_detection = "true" === this.face_detection, this._stopAfterCapturing = "true" === this.stop_after_capturing, this._probabilityThreshold = this.probability_threshold ? +this.probability_threshold : 50, this.url_logo = this.url_logo ? this.url_logo : "data:image/png;charset=utf-8;base64,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", this._debug = "true" === this.debug, kq.debug = this._debug
+                    }, this.photoIsReady = e => {
+                        this._stopAfterCapturing && this.closeCamera(), l.captureEvent(e)
+                    }, this.enableButton = () => {
+                        this._resolutionOnPhoto = !0
+                    }, this.updateState(), this._cameraStatus = 0, this._resolutionOnPhoto = !1, this._device = xq(), this._mobile = this._device.isMobile, this._mobileMakePhoto = !1
+                }
+                videoStarted() {
+                    this._resolutionOnPhoto = !0
+                }
+                async retakePhoto() {
+                    this._mobileMakePhoto = !1, await Nq.getInstance().resumeStream()
+                }
+                takePhoto() {
+                    Nq.getInstance().takePhoto().then(e => {
+                        this.photoIsReady(e)
+                    })
+                }
+                async openCamera() {
+                    this._cameraStatus = 1;
+                    const e = this._device.isMac ? {
+                        audio: !1,
+                        video: {
+                            facingMode: "user",
+                            width: 1920,
+                            height: 1080
+                        }
+                    } : {
+                        audio: !1,
+                        video: {
+                            facingMode: "user",
+                            width: {
+                                ideal: 1920
+                            },
+                            height: {
+                                ideal: 1080
+                            }
+                        }
+                    };
+                    setTimeout(() => {
+                        navigator.mediaDevices.getUserMedia(e).then(e => {
+                            Nq.getInstance().initStream(e), this.videoStarted()
+                        }).catch(e => {
+                            this.closeCamera(), l.errorEvent(e)
+                        })
+                    }, 100)
+                }
+                async loadModels(e) {
+                    await kq.initDetector(e.detail.path)
+                }
+                errorCamera(e) {
+                    l.errorEvent(e.detail.message)
+                }
+                closeCamera() {
+                    Nq.instance && (this._cameraStatus = 0, l.closeEvent(), Nq.getInstance().dropStream(), this._mobileMakePhoto = !1, this._resolutionOnPhoto = !1)
+                }
+                makePhoto() {
+                    this._mobile ? (this._mobileMakePhoto = !0, Nq.getInstance().pauseStream()) : this.takePhoto()
+                }
+                componentWillLoad() {
+                    l.init(this.component), navigator.mediaDevices || l.errorEvent("This browser does not support webRTC")
+                }
+                componentDidUpdate() {
+                    this.updateState()
+                }
+                render() {
+                    let e;
+                    return e = this._cameraStatus ? Object(u.d)("camera-comp", {
+                        class: "block",
+                        showMask: this._showMask,
+                        device: this._device,
+                        faceDetection: this._face_detection,
+                        modelPath: this.model_path,
+                        probabilityThreshold: this._probabilityThreshold
+                    }) : Object(u.d)("img", {
+                        src: this.url_logo,
+                        class: "logo",
+                        style: this.logo_style ? JSON.parse(this.logo_style) : "",
+                        alt: "logo"
+                    }), Object(u.d)("div", {
+                        class: "cameraContainer",
+                        id: "cameraContainer",
+                        style: {
+                            backgroundColor: this.background_color
+                        }
+                    }, Object(u.d)("div", {
+                        class: "wrapperCamera",
+                        style: {
+                            height: this._disable_control_panel ? "100%" : "65%"
+                        }
+                    }, e), Object(u.d)("control-panel", {
+                        class: "block",
+                        mobile: this._mobile,
+                        disableControlPanel: this._disable_control_panel,
+                        cameraStatus: this._cameraStatus,
+                        resolutionOnPhoto: this._resolutionOnPhoto,
+                        mobileMakePhoto: this._mobileMakePhoto,
+                        faceDetection: this._face_detection
+                    }))
+                }
+                get component() {
+                    return Object(u.c)(this)
+                }
+            };
+            Tq.style = ".cameraContainer{width:100%;height:100%;z-index:10;position:relative;overflow:hidden}.logo{max-height:450px;max-width:450px}.wrapperCamera{display:flex;justify-content:center;align-items:center}.block{width:100%;height:100%}@media screen and (orientation: portrait){.history{width:100vw;font-size:20px}@media screen and (max-height: 667px){.logo{background-size:120px;max-height:120px;max-width:120px}}@media screen and (min-height: 668px) and (max-height: 812px){.logo{background-size:140px;max-height:140px;max-width:140px}}@media (max-height: 667px){.logo{margin:0 auto}}}@media screen and (max-height: 414px){.logo{background-size:140px;max-height:140px;max-width:140px}}";
+            const Rq = (e, t) => [...t];
+            let _q = class {
+                constructor(e) {
+                    Object(u.f)(this, e), this.eventOpenCamera = Object(u.b)(this, "openCamera", 7), this.eventCloseCamera = Object(u.b)(this, "closeCamera", 7), this.eventMakePhoto = Object(u.b)(this, "makePhoto", 7), this.eventRetakePhoto = Object(u.b)(this, "retakePhoto", 7), this.eventTakePhoto = Object(u.b)(this, "takePhoto", 7), this.makePhoto = () => {
+                        this.eventMakePhoto.emit()
+                    }, this.closeCamera = () => {
+                        this.eventCloseCamera.emit()
+                    }, this.openCamera = () => {
+                        this.eventOpenCamera.emit()
+                    }, this.takePhoto = () => {
+                        this.eventTakePhoto.emit()
+                    }, this.retakePhoto = async () => {
+                        this.eventRetakePhoto.emit()
+                    }
+                }
+                render() {
+                    //alert("being rendered")
+                    return Object(u.d)("div", {
+                        class: "controlPanel" + (this.mobile ? " controlMobile" : "")
+                    }, this.mobile ? Object(u.d)("div", null, this.cameraStatus ? Object(u.d)("div", null, this.mobileMakePhoto ? Object(u.d)("div", null, Object(u.d)("img", {
+                        onClick: this.takePhoto,
+                        "aria-label": window.acuant_passive_liveness_language["SUBMIT_ALT"],
+                        class: "shoot takePhoto",
+                        id: "acceptSelfieButton",
+                        tabindex: 1,
+                        src: "data:image/svg+xml;base64,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"
+                    }), Object(u.d)("div", {
+                        onClick: this.retakePhoto,
+                        "aria-label": window.acuant_passive_liveness_language["RETAKE_TEXT"],
+                        class: "retake",
+                        id: "retakeButton",
+                        tabindex: 2
+                    }, window.acuant_passive_liveness_language["RETAKE_TEXT"])) : Object(u.d)("button", {
+                        onClick: this.makePhoto,
+                        disabled: !this.resolutionOnPhoto,
+                        "aria-label": window.acuant_passive_liveness_language["INTRO_TEXT"],
+                        tabindex: -1,
+                        class: "shoot",
+                        id: "manualCaptureSelfieButton"
+                    }), Object(u.d)("div", {
+                        onClick: this.closeCamera,
+                        tabindex: 3,
+                        class: "close"
+                    }, window.acuant_passive_liveness_language["CLOSE_TEXT"])) : Object(u.d)("div", null, this.disableControlPanel ? Object(u.d)("div", null) : Object(u.d)("button", {
+                        onClick: this.openCamera
+                    }, "Turn on camera"))) : Object(u.d)(Rq, null, this.disableControlPanel ? Object(u.d)("div", null) : Object(u.d)(Rq, null, this.cameraStatus ? Object(u.d)(Rq, null, Object(u.d)("button", {
+                        onClick: this.makePhoto,
+                        disabled: !this.resolutionOnPhoto
+                    }, "Capture"), Object(u.d)("button", {
+                        onClick: this.closeCamera,
+                        class: "cameraOff"
+                    }, "Close camera")) : Object(u.d)("button", {
+                        onClick: this.openCamera
+                    }, "Open camera"))))
+                }
+            };
+            _q.style = '.controlPanel{height:35%;display:flex;justify-content:space-evenly;align-items:center;width:100%;max-width:450px;margin:auto}.buttons{width:100%;max-width:450px;display:flex;justify-content:space-evenly;align-items:center}button{width:164px;height:50px;border:none;margin-bottom:10px;-webkit-border-radius:10px;-moz-border-radius:10px;border-radius:10px;color:#fff;font-family:"Avenir", Helvetica, sans-serif;font-weight:600;font-size:16px;background-color:#4ca585;outline:none;cursor:pointer}button[disabled]{background-color:#aaa;cursor:not-allowed;pointer-events:none}.cameraOff{background-color:#667a87}.cameraOff:hover{background-color:#92a8b5}.shoot{width:50px;height:50px;border-radius:25px;border:2px solid white;outline:none;background-color:red;position:fixed;left:calc(50% - 25px);bottom:10px}.shoot[disabled]{background-color:#aaa;cursor:not-allowed;pointer-events:none}.close{position:fixed;right:6vw;top:2vh;text-shadow:0 0 4px black;color:white;font-size:20px;user-select:none}.retake{position:fixed;right:6vw;bottom:35px;text-shadow:0 0 4px black;color:white;font-size:20px;user-select:none}.takePhoto{width:46px;height:46px;bottom:20px;background:none}@media screen and (orientation:landscape) and (min-width: 360px){.shoot{bottom:calc(50% - 25px);left:auto;right:5vw}}'
+        }.call(this, "/index.js", n(77), "/", n(36), n(241).Buffer, n(245).setImmediate)
+}, function (e, t, n) {
+    "use strict";
+    (function (e) {
+        /*!
+         * The buffer module from node.js, for the browser.
+         *
+         * @author   Feross Aboukhadijeh <http://feross.org>
+         * @license  MIT
+         */
+        var r = n(242),
+            a = n(243),
+            s = n(244);
+
+        function i() {
+            return u.TYPED_ARRAY_SUPPORT ? 2147483647 : 1073741823
+        }
+
+        function o(e, t) {
+            if (i() < t) throw new RangeError("Invalid typed array length");
+            return u.TYPED_ARRAY_SUPPORT ? (e = new Uint8Array(t)).__proto__ = u.prototype : (null === e && (e = new u(t)), e.length = t), e
+        }
+
+        function u(e, t, n) {
+            if (!(u.TYPED_ARRAY_SUPPORT || this instanceof u)) return new u(e, t, n);
+            if ("number" == typeof e) {
+                if ("string" == typeof t) throw new Error("If encoding is specified then the first argument must be a string");
+                return h(this, e)
+            }
+            return l(this, e, t, n)
+        }
+
+        function l(e, t, n, r) {
+            if ("number" == typeof t) throw new TypeError('"value" argument must not be a number');
+            return "undefined" != typeof ArrayBuffer && t instanceof ArrayBuffer ? function (e, t, n, r) {
+                if (t.byteLength, n < 0 || t.byteLength < n) throw new RangeError("'offset' is out of bounds");
+                if (t.byteLength < n + (r || 0)) throw new RangeError("'length' is out of bounds");
+                t = void 0 === n && void 0 === r ? new Uint8Array(t) : void 0 === r ? new Uint8Array(t, n) : new Uint8Array(t, n, r);
+                u.TYPED_ARRAY_SUPPORT ? (e = t).__proto__ = u.prototype : e = p(e, t);
+                return e
+            }(e, t, n, r) : "string" == typeof t ? function (e, t, n) {
+                "string" == typeof n && "" !== n || (n = "utf8");
+                if (!u.isEncoding(n)) throw new TypeError('"encoding" must be a valid string encoding');
+                var r = 0 | f(t, n),
+                    a = (e = o(e, r)).write(t, n);
+                a !== r && (e = e.slice(0, a));
+                return e
+            }(e, t, n) : function (e, t) {
+                if (u.isBuffer(t)) {
+                    var n = 0 | d(t.length);
+                    return 0 === (e = o(e, n)).length || t.copy(e, 0, 0, n), e
+                }
+                if (t) {
+                    if ("undefined" != typeof ArrayBuffer && t.buffer instanceof ArrayBuffer || "length" in t) return "number" != typeof t.length || (r = t.length) != r ? o(e, 0) : p(e, t);
+                    if ("Buffer" === t.type && s(t.data)) return p(e, t.data)
+                }
+                var r;
+                throw new TypeError("First argument must be a string, Buffer, ArrayBuffer, Array, or array-like object.")
+            }(e, t)
+        }
+
+        function c(e) {
+            if ("number" != typeof e) throw new TypeError('"size" argument must be a number');
+            if (e < 0) throw new RangeError('"size" argument must not be negative')
+        }
+
+        function h(e, t) {
+            if (c(t), e = o(e, t < 0 ? 0 : 0 | d(t)), !u.TYPED_ARRAY_SUPPORT)
+                for (var n = 0; n < t; ++n) e[n] = 0;
+            return e
+        }
+
+        function p(e, t) {
+            var n = t.length < 0 ? 0 : 0 | d(t.length);
+            e = o(e, n);
+            for (var r = 0; r < n; r += 1) e[r] = 255 & t[r];
+            return e
+        }
+
+        function d(e) {
+            if (e >= i()) throw new RangeError("Attempt to allocate Buffer larger than maximum size: 0x" + i().toString(16) + " bytes");
+            return 0 | e
+        }
+
+        function f(e, t) {
+            if (u.isBuffer(e)) return e.length;
+            if ("undefined" != typeof ArrayBuffer && "function" == typeof ArrayBuffer.isView && (ArrayBuffer.isView(e) || e instanceof ArrayBuffer)) return e.byteLength;
+            "string" != typeof e && (e = "" + e);
+            var n = e.length;
+            if (0 === n) return 0;
+            for (var r = !1;;) switch (t) {
+                case "ascii":
+                case "latin1":
+                case "binary":
+                    return n;
+                case "utf8":
+                case "utf-8":
+                case void 0:
+                    return z(e).length;
+                case "ucs2":
+                case "ucs-2":
+                case "utf16le":
+                case "utf-16le":
+                    return 2 * n;
+                case "hex":
+                    return n >>> 1;
+                case "base64":
+                    return U(e).length;
+                default:
+                    if (r) return z(e).length;
+                    t = ("" + t).toLowerCase(), r = !0
+            }
+        }
+
+        function m(e, t, n) {
+            var r = !1;
+            if ((void 0 === t || t < 0) && (t = 0), t > this.length) return "";
+            if ((void 0 === n || n > this.length) && (n = this.length), n <= 0) return "";
+            if ((n >>>= 0) <= (t >>>= 0)) return "";
+            for (e || (e = "utf8");;) switch (e) {
+                case "hex":
+                    return T(this, t, n);
+                case "utf8":
+                case "utf-8":
+                    return E(this, t, n);
+                case "ascii":
+                    return N(this, t, n);
+                case "latin1":
+                case "binary":
+                    return C(this, t, n);
+                case "base64":
+                    return A(this, t, n);
+                case "ucs2":
+                case "ucs-2":
+                case "utf16le":
+                case "utf-16le":
+                    return R(this, t, n);
+                default:
+                    if (r) throw new TypeError("Unknown encoding: " + e);
+                    e = (e + "").toLowerCase(), r = !0
+            }
+        }
+
+        function g(e, t, n) {
+            var r = e[t];
+            e[t] = e[n], e[n] = r
+        }
+
+        function y(e, t, n, r, a) {
+            if (0 === e.length) return -1;
+            if ("string" == typeof n ? (r = n, n = 0) : n > 2147483647 ? n = 2147483647 : n < -2147483648 && (n = -2147483648), n = +n, isNaN(n) && (n = a ? 0 : e.length - 1), n < 0 && (n = e.length + n), n >= e.length) {
+                if (a) return -1;
+                n = e.length - 1
+            } else if (n < 0) {
+                if (!a) return -1;
+                n = 0
+            }
+            if ("string" == typeof t && (t = u.from(t, r)), u.isBuffer(t)) return 0 === t.length ? -1 : b(e, t, n, r, a);
+            if ("number" == typeof t) return t &= 255, u.TYPED_ARRAY_SUPPORT && "function" == typeof Uint8Array.prototype.indexOf ? a ? Uint8Array.prototype.indexOf.call(e, t, n) : Uint8Array.prototype.lastIndexOf.call(e, t, n) : b(e, [t], n, r, a);
+            throw new TypeError("val must be string, number or Buffer")
+        }
+
+        function b(e, t, n, r, a) {
+            var s, i = 1,
+                o = e.length,
+                u = t.length;
+            if (void 0 !== r && ("ucs2" === (r = String(r).toLowerCase()) || "ucs-2" === r || "utf16le" === r || "utf-16le" === r)) {
+                if (e.length < 2 || t.length < 2) return -1;
+                i = 2, o /= 2, u /= 2, n /= 2
+            }
+
+            function l(e, t) {
+                return 1 === i ? e[t] : e.readUInt16BE(t * i)
+            }
+            if (a) {
+                var c = -1;
+                for (s = n; s < o; s++)
+                    if (l(e, s) === l(t, -1 === c ? 0 : s - c)) {
+                        if (-1 === c && (c = s), s - c + 1 === u) return c * i
+                    } else -1 !== c && (s -= s - c), c = -1
+            } else
+                for (n + u > o && (n = o - u), s = n; s >= 0; s--) {
+                    for (var h = !0, p = 0; p < u; p++)
+                        if (l(e, s + p) !== l(t, p)) {
+                            h = !1;
+                            break
+                        } if (h) return s
+                }
+            return -1
+        }
+
+        function v(e, t, n, r) {
+            n = Number(n) || 0;
+            var a = e.length - n;
+            r ? (r = Number(r)) > a && (r = a) : r = a;
+            var s = t.length;
+            if (s % 2 != 0) throw new TypeError("Invalid hex string");
+            r > s / 2 && (r = s / 2);
+            for (var i = 0; i < r; ++i) {
+                var o = parseInt(t.substr(2 * i, 2), 16);
+                if (isNaN(o)) return i;
+                e[n + i] = o
+            }
+            return i
+        }
+
+        function x(e, t, n, r) {
+            return W(z(t, e.length - n), e, n, r)
+        }
+
+        function w(e, t, n, r) {
+            return W(function (e) {
+                for (var t = [], n = 0; n < e.length; ++n) t.push(255 & e.charCodeAt(n));
+                return t
+            }(t), e, n, r)
+        }
+
+        function k(e, t, n, r) {
+            return w(e, t, n, r)
+        }
+
+        function S(e, t, n, r) {
+            return W(U(t), e, n, r)
+        }
+
+        function I(e, t, n, r) {
+            return W(function (e, t) {
+                for (var n, r, a, s = [], i = 0; i < e.length && !((t -= 2) < 0); ++i) n = e.charCodeAt(i), r = n >> 8, a = n % 256, s.push(a), s.push(r);
+                return s
+            }(t, e.length - n), e, n, r)
+        }
+
+        function A(e, t, n) {
+            return 0 === t && n === e.length ? r.fromByteArray(e) : r.fromByteArray(e.slice(t, n))
+        }
+
+        function E(e, t, n) {
+            n = Math.min(e.length, n);
+            for (var r = [], a = t; a < n;) {
+                var s, i, o, u, l = e[a],
+                    c = null,
+                    h = l > 239 ? 4 : l > 223 ? 3 : l > 191 ? 2 : 1;
+                if (a + h <= n) switch (h) {
+                    case 1:
+                        l < 128 && (c = l);
+                        break;
+                    case 2:
+                        128 == (192 & (s = e[a + 1])) && (u = (31 & l) << 6 | 63 & s) > 127 && (c = u);
+                        break;
+                    case 3:
+                        s = e[a + 1], i = e[a + 2], 128 == (192 & s) && 128 == (192 & i) && (u = (15 & l) << 12 | (63 & s) << 6 | 63 & i) > 2047 && (u < 55296 || u > 57343) && (c = u);
+                        break;
+                    case 4:
+                        s = e[a + 1], i = e[a + 2], o = e[a + 3], 128 == (192 & s) && 128 == (192 & i) && 128 == (192 & o) && (u = (15 & l) << 18 | (63 & s) << 12 | (63 & i) << 6 | 63 & o) > 65535 && u < 1114112 && (c = u)
+                }
+                null === c ? (c = 65533, h = 1) : c > 65535 && (c -= 65536, r.push(c >>> 10 & 1023 | 55296), c = 56320 | 1023 & c), r.push(c), a += h
+            }
+            return function (e) {
+                var t = e.length;
+                if (t <= 4096) return String.fromCharCode.apply(String, e);
+                var n = "",
+                    r = 0;
+                for (; r < t;) n += String.fromCharCode.apply(String, e.slice(r, r += 4096));
+                return n
+            }(r)
+        }
+        t.Buffer = u, t.SlowBuffer = function (e) {
+            +e != e && (e = 0);
+            return u.alloc(+e)
+        }, t.INSPECT_MAX_BYTES = 50, u.TYPED_ARRAY_SUPPORT = void 0 !== e.TYPED_ARRAY_SUPPORT ? e.TYPED_ARRAY_SUPPORT : function () {
+            try {
+                var e = new Uint8Array(1);
+                return e.__proto__ = {
+                    __proto__: Uint8Array.prototype,
+                    foo: function () {
+                        return 42
+                    }
+                }, 42 === e.foo() && "function" == typeof e.subarray && 0 === e.subarray(1, 1).byteLength
+            } catch (e) {
+                return !1
+            }
+        }(), t.kMaxLength = i(), u.poolSize = 8192, u._augment = function (e) {
+            return e.__proto__ = u.prototype, e
+        }, u.from = function (e, t, n) {
+            return l(null, e, t, n)
+        }, u.TYPED_ARRAY_SUPPORT && (u.prototype.__proto__ = Uint8Array.prototype, u.__proto__ = Uint8Array, "undefined" != typeof Symbol && Symbol.species && u[Symbol.species] === u && Object.defineProperty(u, Symbol.species, {
+            value: null,
+            configurable: !0
+        })), u.alloc = function (e, t, n) {
+            return function (e, t, n, r) {
+                return c(t), t <= 0 ? o(e, t) : void 0 !== n ? "string" == typeof r ? o(e, t).fill(n, r) : o(e, t).fill(n) : o(e, t)
+            }(null, e, t, n)
+        }, u.allocUnsafe = function (e) {
+            return h(null, e)
+        }, u.allocUnsafeSlow = function (e) {
+            return h(null, e)
+        }, u.isBuffer = function (e) {
+            return !(null == e || !e._isBuffer)
+        }, u.compare = function (e, t) {
+            if (!u.isBuffer(e) || !u.isBuffer(t)) throw new TypeError("Arguments must be Buffers");
+            if (e === t) return 0;
+            for (var n = e.length, r = t.length, a = 0, s = Math.min(n, r); a < s; ++a)
+                if (e[a] !== t[a]) {
+                    n = e[a], r = t[a];
+                    break
+                } return n < r ? -1 : r < n ? 1 : 0
+        }, u.isEncoding = function (e) {
+            switch (String(e).toLowerCase()) {
+                case "hex":
+                case "utf8":
+                case "utf-8":
+                case "ascii":
+                case "latin1":
+                case "binary":
+                case "base64":
+                case "ucs2":
+                case "ucs-2":
+                case "utf16le":
+                case "utf-16le":
+                    return !0;
+                default:
+                    return !1
+            }
+        }, u.concat = function (e, t) {
+            if (!s(e)) throw new TypeError('"list" argument must be an Array of Buffers');
+            if (0 === e.length) return u.alloc(0);
+            var n;
+            if (void 0 === t)
+                for (t = 0, n = 0; n < e.length; ++n) t += e[n].length;
+            var r = u.allocUnsafe(t),
+                a = 0;
+            for (n = 0; n < e.length; ++n) {
+                var i = e[n];
+                if (!u.isBuffer(i)) throw new TypeError('"list" argument must be an Array of Buffers');
+                i.copy(r, a), a += i.length
+            }
+            return r
+        }, u.byteLength = f, u.prototype._isBuffer = !0, u.prototype.swap16 = function () {
+            var e = this.length;
+            if (e % 2 != 0) throw new RangeError("Buffer size must be a multiple of 16-bits");
+            for (var t = 0; t < e; t += 2) g(this, t, t + 1);
+            return this
+        }, u.prototype.swap32 = function () {
+            var e = this.length;
+            if (e % 4 != 0) throw new RangeError("Buffer size must be a multiple of 32-bits");
+            for (var t = 0; t < e; t += 4) g(this, t, t + 3), g(this, t + 1, t + 2);
+            return this
+        }, u.prototype.swap64 = function () {
+            var e = this.length;
+            if (e % 8 != 0) throw new RangeError("Buffer size must be a multiple of 64-bits");
+            for (var t = 0; t < e; t += 8) g(this, t, t + 7), g(this, t + 1, t + 6), g(this, t + 2, t + 5), g(this, t + 3, t + 4);
+            return this
+        }, u.prototype.toString = function () {
+            var e = 0 | this.length;
+            return 0 === e ? "" : 0 === arguments.length ? E(this, 0, e) : m.apply(this, arguments)
+        }, u.prototype.equals = function (e) {
+            if (!u.isBuffer(e)) throw new TypeError("Argument must be a Buffer");
+            return this === e || 0 === u.compare(this, e)
+        }, u.prototype.inspect = function () {
+            var e = "",
+                n = t.INSPECT_MAX_BYTES;
+            return this.length > 0 && (e = this.toString("hex", 0, n).match(/.{2}/g).join(" "), this.length > n && (e += " ... ")), "<Buffer " + e + ">"
+        }, u.prototype.compare = function (e, t, n, r, a) {
+            if (!u.isBuffer(e)) throw new TypeError("Argument must be a Buffer");
+            if (void 0 === t && (t = 0), void 0 === n && (n = e ? e.length : 0), void 0 === r && (r = 0), void 0 === a && (a = this.length), t < 0 || n > e.length || r < 0 || a > this.length) throw new RangeError("out of range index");
+            if (r >= a && t >= n) return 0;
+            if (r >= a) return -1;
+            if (t >= n) return 1;
+            if (this === e) return 0;
+            for (var s = (a >>>= 0) - (r >>>= 0), i = (n >>>= 0) - (t >>>= 0), o = Math.min(s, i), l = this.slice(r, a), c = e.slice(t, n), h = 0; h < o; ++h)
+                if (l[h] !== c[h]) {
+                    s = l[h], i = c[h];
+                    break
+                } return s < i ? -1 : i < s ? 1 : 0
+        }, u.prototype.includes = function (e, t, n) {
+            return -1 !== this.indexOf(e, t, n)
+        }, u.prototype.indexOf = function (e, t, n) {
+            return y(this, e, t, n, !0)
+        }, u.prototype.lastIndexOf = function (e, t, n) {
+            return y(this, e, t, n, !1)
+        }, u.prototype.write = function (e, t, n, r) {
+            if (void 0 === t) r = "utf8", n = this.length, t = 0;
+            else if (void 0 === n && "string" == typeof t) r = t, n = this.length, t = 0;
+            else {
+                if (!isFinite(t)) throw new Error("Buffer.write(string, encoding, offset[, length]) is no longer supported");
+                t |= 0, isFinite(n) ? (n |= 0, void 0 === r && (r = "utf8")) : (r = n, n = void 0)
+            }
+            var a = this.length - t;
+            if ((void 0 === n || n > a) && (n = a), e.length > 0 && (n < 0 || t < 0) || t > this.length) throw new RangeError("Attempt to write outside buffer bounds");
+            r || (r = "utf8");
+            for (var s = !1;;) switch (r) {
+                case "hex":
+                    return v(this, e, t, n);
+                case "utf8":
+                case "utf-8":
+                    return x(this, e, t, n);
+                case "ascii":
+                    return w(this, e, t, n);
+                case "latin1":
+                case "binary":
+                    return k(this, e, t, n);
+                case "base64":
+                    return S(this, e, t, n);
+                case "ucs2":
+                case "ucs-2":
+                case "utf16le":
+                case "utf-16le":
+                    return I(this, e, t, n);
+                default:
+                    if (s) throw new TypeError("Unknown encoding: " + r);
+                    r = ("" + r).toLowerCase(), s = !0
+            }
+        }, u.prototype.toJSON = function () {
+            return {
+                type: "Buffer",
+                data: Array.prototype.slice.call(this._arr || this, 0)
+            }
+        };
+
+        function N(e, t, n) {
+            var r = "";
+            n = Math.min(e.length, n);
+            for (var a = t; a < n; ++a) r += String.fromCharCode(127 & e[a]);
+            return r
+        }
+
+        function C(e, t, n) {
+            var r = "";
+            n = Math.min(e.length, n);
+            for (var a = t; a < n; ++a) r += String.fromCharCode(e[a]);
+            return r
+        }
+
+        function T(e, t, n) {
+            var r = e.length;
+            (!t || t < 0) && (t = 0), (!n || n < 0 || n > r) && (n = r);
+            for (var a = "", s = t; s < n; ++s) a += B(e[s]);
+            return a
+        }
+
+        function R(e, t, n) {
+            for (var r = e.slice(t, n), a = "", s = 0; s < r.length; s += 2) a += String.fromCharCode(r[s] + 256 * r[s + 1]);
+            return a
+        }
+
+        function _(e, t, n) {
+            if (e % 1 != 0 || e < 0) throw new RangeError("offset is not uint");
+            if (e + t > n) throw new RangeError("Trying to access beyond buffer length")
+        }
+
+        function F(e, t, n, r, a, s) {
+            if (!u.isBuffer(e)) throw new TypeError('"buffer" argument must be a Buffer instance');
+            if (t > a || t < s) throw new RangeError('"value" argument is out of bounds');
+            if (n + r > e.length) throw new RangeError("Index out of range")
+        }
+
+        function M(e, t, n, r) {
+            t < 0 && (t = 65535 + t + 1);
+            for (var a = 0, s = Math.min(e.length - n, 2); a < s; ++a) e[n + a] = (t & 255 << 8 * (r ? a : 1 - a)) >>> 8 * (r ? a : 1 - a)
+        }
+
+        function D(e, t, n, r) {
+            t < 0 && (t = 4294967295 + t + 1);
+            for (var a = 0, s = Math.min(e.length - n, 4); a < s; ++a) e[n + a] = t >>> 8 * (r ? a : 3 - a) & 255
+        }
+
+        function O(e, t, n, r, a, s) {
+            if (n + r > e.length) throw new RangeError("Index out of range");
+            if (n < 0) throw new RangeError("Index out of range")
+        }
+
+        function L(e, t, n, r, s) {
+            return s || O(e, 0, n, 4), a.write(e, t, n, r, 23, 4), n + 4
+        }
+
+        function P(e, t, n, r, s) {
+            return s || O(e, 0, n, 8), a.write(e, t, n, r, 52, 8), n + 8
+        }
+        u.prototype.slice = function (e, t) {
+            var n, r = this.length;
+            if ((e = ~~e) < 0 ? (e += r) < 0 && (e = 0) : e > r && (e = r), (t = void 0 === t ? r : ~~t) < 0 ? (t += r) < 0 && (t = 0) : t > r && (t = r), t < e && (t = e), u.TYPED_ARRAY_SUPPORT)(n = this.subarray(e, t)).__proto__ = u.prototype;
+            else {
+                var a = t - e;
+                n = new u(a, void 0);
+                for (var s = 0; s < a; ++s) n[s] = this[s + e]
+            }
+            return n
+        }, u.prototype.readUIntLE = function (e, t, n) {
+            e |= 0, t |= 0, n || _(e, t, this.length);
+            for (var r = this[e], a = 1, s = 0; ++s < t && (a *= 256);) r += this[e + s] * a;
+            return r
+        }, u.prototype.readUIntBE = function (e, t, n) {
+            e |= 0, t |= 0, n || _(e, t, this.length);
+            for (var r = this[e + --t], a = 1; t > 0 && (a *= 256);) r += this[e + --t] * a;
+            return r
+        }, u.prototype.readUInt8 = function (e, t) {
+            return t || _(e, 1, this.length), this[e]
+        }, u.prototype.readUInt16LE = function (e, t) {
+            return t || _(e, 2, this.length), this[e] | this[e + 1] << 8
+        }, u.prototype.readUInt16BE = function (e, t) {
+            return t || _(e, 2, this.length), this[e] << 8 | this[e + 1]
+        }, u.prototype.readUInt32LE = function (e, t) {
+            return t || _(e, 4, this.length), (this[e] | this[e + 1] << 8 | this[e + 2] << 16) + 16777216 * this[e + 3]
+        }, u.prototype.readUInt32BE = function (e, t) {
+            return t || _(e, 4, this.length), 16777216 * this[e] + (this[e + 1] << 16 | this[e + 2] << 8 | this[e + 3])
+        }, u.prototype.readIntLE = function (e, t, n) {
+            e |= 0, t |= 0, n || _(e, t, this.length);
+            for (var r = this[e], a = 1, s = 0; ++s < t && (a *= 256);) r += this[e + s] * a;
+            return r >= (a *= 128) && (r -= Math.pow(2, 8 * t)), r
+        }, u.prototype.readIntBE = function (e, t, n) {
+            e |= 0, t |= 0, n || _(e, t, this.length);
+            for (var r = t, a = 1, s = this[e + --r]; r > 0 && (a *= 256);) s += this[e + --r] * a;
+            return s >= (a *= 128) && (s -= Math.pow(2, 8 * t)), s
+        }, u.prototype.readInt8 = function (e, t) {
+            return t || _(e, 1, this.length), 128 & this[e] ? -1 * (255 - this[e] + 1) : this[e]
+        }, u.prototype.readInt16LE = function (e, t) {
+            t || _(e, 2, this.length);
+            var n = this[e] | this[e + 1] << 8;
+            return 32768 & n ? 4294901760 | n : n
+        }, u.prototype.readInt16BE = function (e, t) {
+            t || _(e, 2, this.length);
+            var n = this[e + 1] | this[e] << 8;
+            return 32768 & n ? 4294901760 | n : n
+        }, u.prototype.readInt32LE = function (e, t) {
+            return t || _(e, 4, this.length), this[e] | this[e + 1] << 8 | this[e + 2] << 16 | this[e + 3] << 24
+        }, u.prototype.readInt32BE = function (e, t) {
+            return t || _(e, 4, this.length), this[e] << 24 | this[e + 1] << 16 | this[e + 2] << 8 | this[e + 3]
+        }, u.prototype.readFloatLE = function (e, t) {
+            return t || _(e, 4, this.length), a.read(this, e, !0, 23, 4)
+        }, u.prototype.readFloatBE = function (e, t) {
+            return t || _(e, 4, this.length), a.read(this, e, !1, 23, 4)
+        }, u.prototype.readDoubleLE = function (e, t) {
+            return t || _(e, 8, this.length), a.read(this, e, !0, 52, 8)
+        }, u.prototype.readDoubleBE = function (e, t) {
+            return t || _(e, 8, this.length), a.read(this, e, !1, 52, 8)
+        }, u.prototype.writeUIntLE = function (e, t, n, r) {
+            (e = +e, t |= 0, n |= 0, r) || F(this, e, t, n, Math.pow(2, 8 * n) - 1, 0);
+            var a = 1,
+                s = 0;
+            for (this[t] = 255 & e; ++s < n && (a *= 256);) this[t + s] = e / a & 255;
+            return t + n
+        }, u.prototype.writeUIntBE = function (e, t, n, r) {
+            (e = +e, t |= 0, n |= 0, r) || F(this, e, t, n, Math.pow(2, 8 * n) - 1, 0);
+            var a = n - 1,
+                s = 1;
+            for (this[t + a] = 255 & e; --a >= 0 && (s *= 256);) this[t + a] = e / s & 255;
+            return t + n
+        }, u.prototype.writeUInt8 = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 1, 255, 0), u.TYPED_ARRAY_SUPPORT || (e = Math.floor(e)), this[t] = 255 & e, t + 1
+        }, u.prototype.writeUInt16LE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 2, 65535, 0), u.TYPED_ARRAY_SUPPORT ? (this[t] = 255 & e, this[t + 1] = e >>> 8) : M(this, e, t, !0), t + 2
+        }, u.prototype.writeUInt16BE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 2, 65535, 0), u.TYPED_ARRAY_SUPPORT ? (this[t] = e >>> 8, this[t + 1] = 255 & e) : M(this, e, t, !1), t + 2
+        }, u.prototype.writeUInt32LE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 4, 4294967295, 0), u.TYPED_ARRAY_SUPPORT ? (this[t + 3] = e >>> 24, this[t + 2] = e >>> 16, this[t + 1] = e >>> 8, this[t] = 255 & e) : D(this, e, t, !0), t + 4
+        }, u.prototype.writeUInt32BE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 4, 4294967295, 0), u.TYPED_ARRAY_SUPPORT ? (this[t] = e >>> 24, this[t + 1] = e >>> 16, this[t + 2] = e >>> 8, this[t + 3] = 255 & e) : D(this, e, t, !1), t + 4
+        }, u.prototype.writeIntLE = function (e, t, n, r) {
+            if (e = +e, t |= 0, !r) {
+                var a = Math.pow(2, 8 * n - 1);
+                F(this, e, t, n, a - 1, -a)
+            }
+            var s = 0,
+                i = 1,
+                o = 0;
+            for (this[t] = 255 & e; ++s < n && (i *= 256);) e < 0 && 0 === o && 0 !== this[t + s - 1] && (o = 1), this[t + s] = (e / i >> 0) - o & 255;
+            return t + n
+        }, u.prototype.writeIntBE = function (e, t, n, r) {
+            if (e = +e, t |= 0, !r) {
+                var a = Math.pow(2, 8 * n - 1);
+                F(this, e, t, n, a - 1, -a)
+            }
+            var s = n - 1,
+                i = 1,
+                o = 0;
+            for (this[t + s] = 255 & e; --s >= 0 && (i *= 256);) e < 0 && 0 === o && 0 !== this[t + s + 1] && (o = 1), this[t + s] = (e / i >> 0) - o & 255;
+            return t + n
+        }, u.prototype.writeInt8 = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 1, 127, -128), u.TYPED_ARRAY_SUPPORT || (e = Math.floor(e)), e < 0 && (e = 255 + e + 1), this[t] = 255 & e, t + 1
+        }, u.prototype.writeInt16LE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 2, 32767, -32768), u.TYPED_ARRAY_SUPPORT ? (this[t] = 255 & e, this[t + 1] = e >>> 8) : M(this, e, t, !0), t + 2
+        }, u.prototype.writeInt16BE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 2, 32767, -32768), u.TYPED_ARRAY_SUPPORT ? (this[t] = e >>> 8, this[t + 1] = 255 & e) : M(this, e, t, !1), t + 2
+        }, u.prototype.writeInt32LE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 4, 2147483647, -2147483648), u.TYPED_ARRAY_SUPPORT ? (this[t] = 255 & e, this[t + 1] = e >>> 8, this[t + 2] = e >>> 16, this[t + 3] = e >>> 24) : D(this, e, t, !0), t + 4
+        }, u.prototype.writeInt32BE = function (e, t, n) {
+            return e = +e, t |= 0, n || F(this, e, t, 4, 2147483647, -2147483648), e < 0 && (e = 4294967295 + e + 1), u.TYPED_ARRAY_SUPPORT ? (this[t] = e >>> 24, this[t + 1] = e >>> 16, this[t + 2] = e >>> 8, this[t + 3] = 255 & e) : D(this, e, t, !1), t + 4
+        }, u.prototype.writeFloatLE = function (e, t, n) {
+            return L(this, e, t, !0, n)
+        }, u.prototype.writeFloatBE = function (e, t, n) {
+            return L(this, e, t, !1, n)
+        }, u.prototype.writeDoubleLE = function (e, t, n) {
+            return P(this, e, t, !0, n)
+        }, u.prototype.writeDoubleBE = function (e, t, n) {
+            return P(this, e, t, !1, n)
+        }, u.prototype.copy = function (e, t, n, r) {
+            if (n || (n = 0), r || 0 === r || (r = this.length), t >= e.length && (t = e.length), t || (t = 0), r > 0 && r < n && (r = n), r === n) return 0;
+            if (0 === e.length || 0 === this.length) return 0;
+            if (t < 0) throw new RangeError("targetStart out of bounds");
+            if (n < 0 || n >= this.length) throw new RangeError("sourceStart out of bounds");
+            if (r < 0) throw new RangeError("sourceEnd out of bounds");
+            r > this.length && (r = this.length), e.length - t < r - n && (r = e.length - t + n);
+            var a, s = r - n;
+            if (this === e && n < t && t < r)
+                for (a = s - 1; a >= 0; --a) e[a + t] = this[a + n];
+            else if (s < 1e3 || !u.TYPED_ARRAY_SUPPORT)
+                for (a = 0; a < s; ++a) e[a + t] = this[a + n];
+            else Uint8Array.prototype.set.call(e, this.subarray(n, n + s), t);
+            return s
+        }, u.prototype.fill = function (e, t, n, r) {
+            if ("string" == typeof e) {
+                if ("string" == typeof t ? (r = t, t = 0, n = this.length) : "string" == typeof n && (r = n, n = this.length), 1 === e.length) {
+                    var a = e.charCodeAt(0);
+                    a < 256 && (e = a)
+                }
+                if (void 0 !== r && "string" != typeof r) throw new TypeError("encoding must be a string");
+                if ("string" == typeof r && !u.isEncoding(r)) throw new TypeError("Unknown encoding: " + r)
+            } else "number" == typeof e && (e &= 255);
+            if (t < 0 || this.length < t || this.length < n) throw new RangeError("Out of range index");
+            if (n <= t) return this;
+            var s;
+            if (t >>>= 0, n = void 0 === n ? this.length : n >>> 0, e || (e = 0), "number" == typeof e)
+                for (s = t; s < n; ++s) this[s] = e;
+            else {
+                var i = u.isBuffer(e) ? e : z(new u(e, r).toString()),
+                    o = i.length;
+                for (s = 0; s < n - t; ++s) this[s + t] = i[s % o]
+            }
+            return this
+        };
+        var $ = /[^+\/0-9A-Za-z-_]/g;
+
+        function B(e) {
+            return e < 16 ? "0" + e.toString(16) : e.toString(16)
+        }
+
+        function z(e, t) {
+            var n;
+            t = t || 1 / 0;
+            for (var r = e.length, a = null, s = [], i = 0; i < r; ++i) {
+                if ((n = e.charCodeAt(i)) > 55295 && n < 57344) {
+                    if (!a) {
+                        if (n > 56319) {
+                            (t -= 3) > -1 && s.push(239, 191, 189);
+                            continue
+                        }
+                        if (i + 1 === r) {
+                            (t -= 3) > -1 && s.push(239, 191, 189);
+                            continue
+                        }
+                        a = n;
+                        continue
+                    }
+                    if (n < 56320) {
+                        (t -= 3) > -1 && s.push(239, 191, 189), a = n;
+                        continue
+                    }
+                    n = 65536 + (a - 55296 << 10 | n - 56320)
+                } else a && (t -= 3) > -1 && s.push(239, 191, 189);
+                if (a = null, n < 128) {
+                    if ((t -= 1) < 0) break;
+                    s.push(n)
+                } else if (n < 2048) {
+                    if ((t -= 2) < 0) break;
+                    s.push(n >> 6 | 192, 63 & n | 128)
+                } else if (n < 65536) {
+                    if ((t -= 3) < 0) break;
+                    s.push(n >> 12 | 224, n >> 6 & 63 | 128, 63 & n | 128)
+                } else {
+                    if (!(n < 1114112)) throw new Error("Invalid code point");
+                    if ((t -= 4) < 0) break;
+                    s.push(n >> 18 | 240, n >> 12 & 63 | 128, n >> 6 & 63 | 128, 63 & n | 128)
+                }
+            }
+            return s
+        }
+
+        function U(e) {
+            return r.toByteArray(function (e) {
+                if ((e = function (e) {
+                        return e.trim ? e.trim() : e.replace(/^\s+|\s+$/g, "")
+                    }(e).replace($, "")).length < 2) return "";
+                for (; e.length % 4 != 0;) e += "=";
+                return e
+            }(e))
+        }
+
+        function W(e, t, n, r) {
+            for (var a = 0; a < r && !(a + n >= t.length || a >= e.length); ++a) t[a + n] = e[a];
+            return a
+        }
+    }).call(this, n(36))
+}, function (e, t, n) {
+    "use strict";
+    t.byteLength = function (e) {
+        var t = l(e),
+            n = t[0],
+            r = t[1];
+        return 3 * (n + r) / 4 - r
+    }, t.toByteArray = function (e) {
+        var t, n, r = l(e),
+            i = r[0],
+            o = r[1],
+            u = new s(function (e, t, n) {
+                return 3 * (t + n) / 4 - n
+            }(0, i, o)),
+            c = 0,
+            h = o > 0 ? i - 4 : i;
+        for (n = 0; n < h; n += 4) t = a[e.charCodeAt(n)] << 18 | a[e.charCodeAt(n + 1)] << 12 | a[e.charCodeAt(n + 2)] << 6 | a[e.charCodeAt(n + 3)], u[c++] = t >> 16 & 255, u[c++] = t >> 8 & 255, u[c++] = 255 & t;
+        2 === o && (t = a[e.charCodeAt(n)] << 2 | a[e.charCodeAt(n + 1)] >> 4, u[c++] = 255 & t);
+        1 === o && (t = a[e.charCodeAt(n)] << 10 | a[e.charCodeAt(n + 1)] << 4 | a[e.charCodeAt(n + 2)] >> 2, u[c++] = t >> 8 & 255, u[c++] = 255 & t);
+        return u
+    }, t.fromByteArray = function (e) {
+        for (var t, n = e.length, a = n % 3, s = [], i = 0, o = n - a; i < o; i += 16383) s.push(c(e, i, i + 16383 > o ? o : i + 16383));
+        1 === a ? (t = e[n - 1], s.push(r[t >> 2] + r[t << 4 & 63] + "==")) : 2 === a && (t = (e[n - 2] << 8) + e[n - 1], s.push(r[t >> 10] + r[t >> 4 & 63] + r[t << 2 & 63] + "="));
+        return s.join("")
+    };
+    for (var r = [], a = [], s = "undefined" != typeof Uint8Array ? Uint8Array : Array, i = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/", o = 0, u = i.length; o < u; ++o) r[o] = i[o], a[i.charCodeAt(o)] = o;
+
+    function l(e) {
+        var t = e.length;
+        if (t % 4 > 0) throw new Error("Invalid string. Length must be a multiple of 4");
+        var n = e.indexOf("=");
+        return -1 === n && (n = t), [n, n === t ? 0 : 4 - n % 4]
+    }
+
+    function c(e, t, n) {
+        for (var a, s, i = [], o = t; o < n; o += 3) a = (e[o] << 16 & 16711680) + (e[o + 1] << 8 & 65280) + (255 & e[o + 2]), i.push(r[(s = a) >> 18 & 63] + r[s >> 12 & 63] + r[s >> 6 & 63] + r[63 & s]);
+        return i.join("")
+    }
+    a["-".charCodeAt(0)] = 62, a["_".charCodeAt(0)] = 63
+}, function (e, t) {
+    /*! ieee754. BSD-3-Clause License. Feross Aboukhadijeh <https://feross.org/opensource> */
+    t.read = function (e, t, n, r, a) {
+        var s, i, o = 8 * a - r - 1,
+            u = (1 << o) - 1,
+            l = u >> 1,
+            c = -7,
+            h = n ? a - 1 : 0,
+            p = n ? -1 : 1,
+            d = e[t + h];
+        for (h += p, s = d & (1 << -c) - 1, d >>= -c, c += o; c > 0; s = 256 * s + e[t + h], h += p, c -= 8);
+        for (i = s & (1 << -c) - 1, s >>= -c, c += r; c > 0; i = 256 * i + e[t + h], h += p, c -= 8);
+        if (0 === s) s = 1 - l;
+        else {
+            if (s === u) return i ? NaN : 1 / 0 * (d ? -1 : 1);
+            i += Math.pow(2, r), s -= l
+        }
+        return (d ? -1 : 1) * i * Math.pow(2, s - r)
+    }, t.write = function (e, t, n, r, a, s) {
+        var i, o, u, l = 8 * s - a - 1,
+            c = (1 << l) - 1,
+            h = c >> 1,
+            p = 23 === a ? Math.pow(2, -24) - Math.pow(2, -77) : 0,
+            d = r ? 0 : s - 1,
+            f = r ? 1 : -1,
+            m = t < 0 || 0 === t && 1 / t < 0 ? 1 : 0;
+        for (t = Math.abs(t), isNaN(t) || t === 1 / 0 ? (o = isNaN(t) ? 1 : 0, i = c) : (i = Math.floor(Math.log(t) / Math.LN2), t * (u = Math.pow(2, -i)) < 1 && (i--, u *= 2), (t += i + h >= 1 ? p / u : p * Math.pow(2, 1 - h)) * u >= 2 && (i++, u /= 2), i + h >= c ? (o = 0, i = c) : i + h >= 1 ? (o = (t * u - 1) * Math.pow(2, a), i += h) : (o = t * Math.pow(2, h - 1) * Math.pow(2, a), i = 0)); a >= 8; e[n + d] = 255 & o, d += f, o /= 256, a -= 8);
+        for (i = i << a | o, l += a; l > 0; e[n + d] = 255 & i, d += f, i /= 256, l -= 8);
+        e[n + d - f] |= 128 * m
+    }
+}, function (e, t) {
+    var n = {}.toString;
+    e.exports = Array.isArray || function (e) {
+        return "[object Array]" == n.call(e)
+    }
+}, function (e, t, n) {
+    (function (e) {
+        var r = void 0 !== e && e || "undefined" != typeof self && self || window,
+            a = Function.prototype.apply;
+
+        function s(e, t) {
+            this._id = e, this._clearFn = t
+        }
+        t.setTimeout = function () {
+            return new s(a.call(setTimeout, r, arguments), clearTimeout)
+        }, t.setInterval = function () {
+            return new s(a.call(setInterval, r, arguments), clearInterval)
+        }, t.clearTimeout = t.clearInterval = function (e) {
+            e && e.close()
+        }, s.prototype.unref = s.prototype.ref = function () {}, s.prototype.close = function () {
+            this._clearFn.call(r, this._id)
+        }, t.enroll = function (e, t) {
+            clearTimeout(e._idleTimeoutId), e._idleTimeout = t
+        }, t.unenroll = function (e) {
+            clearTimeout(e._idleTimeoutId), e._idleTimeout = -1
+        }, t._unrefActive = t.active = function (e) {
+            clearTimeout(e._idleTimeoutId);
+            var t = e._idleTimeout;
+            t >= 0 && (e._idleTimeoutId = setTimeout((function () {
+                e._onTimeout && e._onTimeout()
+            }), t))
+        }, n(246), t.setImmediate = "undefined" != typeof self && self.setImmediate || void 0 !== e && e.setImmediate || this && this.setImmediate, t.clearImmediate = "undefined" != typeof self && self.clearImmediate || void 0 !== e && e.clearImmediate || this && this.clearImmediate
+    }).call(this, n(36))
+}, function (e, t, n) {
+    (function (e, t) {
+        ! function (e, n) {
+            "use strict";
+            if (!e.setImmediate) {
+                var r, a, s, i, o, u = 1,
+                    l = {},
+                    c = !1,
+                    h = e.document,
+                    p = Object.getPrototypeOf && Object.getPrototypeOf(e);
+                p = p && p.setTimeout ? p : e, "[object process]" === {}.toString.call(e.process) ? r = function (e) {
+                    t.nextTick((function () {
+                        f(e)
+                    }))
+                } : ! function () {
+                    if (e.postMessage && !e.importScripts) {
+                        var t = !0,
+                            n = e.onmessage;
+                        return e.onmessage = function () {
+                            t = !1
+                        }, e.postMessage("", "*"), e.onmessage = n, t
+                    }
+                }() ? e.MessageChannel ? ((s = new MessageChannel).port1.onmessage = function (e) {
+                    f(e.data)
+                }, r = function (e) {
+                    s.port2.postMessage(e)
+                }) : h && "onreadystatechange" in h.createElement("script") ? (a = h.documentElement, r = function (e) {
+                    var t = h.createElement("script");
+                    t.onreadystatechange = function () {
+                        f(e), t.onreadystatechange = null, a.removeChild(t), t = null
+                    }, a.appendChild(t)
+                }) : r = function (e) {
+                    setTimeout(f, 0, e)
+                } : (i = "setImmediate$" + Math.random() + "$", o = function (t) {
+                    t.source === e && "string" == typeof t.data && 0 === t.data.indexOf(i) && f(+t.data.slice(i.length))
+                }, e.addEventListener ? e.addEventListener("message", o, !1) : e.attachEvent("onmessage", o), r = function (t) {
+                    e.postMessage(i + t, "*")
+                }), p.setImmediate = function (e) {
+                    "function" != typeof e && (e = new Function("" + e));
+                    for (var t = new Array(arguments.length - 1), n = 0; n < t.length; n++) t[n] = arguments[n + 1];
+                    var a = {
+                        callback: e,
+                        args: t
+                    };
+                    return l[u] = a, r(u), u++
+                }, p.clearImmediate = d
+            }
+
+            function d(e) {
+                delete l[e]
+            }
+
+            function f(e) {
+                if (c) setTimeout(f, 0, e);
+                else {
+                    var t = l[e];
+                    if (t) {
+                        c = !0;
+                        try {
+                            ! function (e) {
+                                var t = e.callback,
+                                    n = e.args;
+                                switch (n.length) {
+                                    case 0:
+                                        t();
+                                        break;
+                                    case 1:
+                                        t(n[0]);
+                                        break;
+                                    case 2:
+                                        t(n[0], n[1]);
+                                        break;
+                                    case 3:
+                                        t(n[0], n[1], n[2]);
+                                        break;
+                                    default:
+                                        t.apply(void 0, n)
+                                }
+                            }(t)
+                        } finally {
+                            d(e), c = !1
+                        }
+                    }
+                }
+            }
+        }("undefined" == typeof self ? void 0 === e ? this : e : self)
+    }).call(this, n(36), n(77))
+}, function (e, t, n) {
+    "use strict";
+    n.r(t);
+    n(130), n(134), n(153), n(159), n(160), n(163), n(164), n(86), n(166), n(169), n(177), n(180), n(181), n(182), n(185), n(186), n(187), n(188), n(189), n(190), n(191), n(192), n(194), n(195), n(196), n(197), n(198), n(199), n(200), n(201), n(206), n(207), n(208), n(209), n(210), n(211), n(221), n(222), n(223), n(225), n(226), n(227), n(228), n(229), n(230), n(231), n(232), n(233), n(234), n(235), n(236), n(237);
+    var r = n(0);
+    const a = (e, t) => "undefined" == typeof window ? Promise.resolve() : Object(r.e)().then(() => Object(r.a)([
+        ["camera-comp_3", [
+            [1, "camera-component", {
+                    url_logo: [1],
+                    show_mask: [1],
+                    background_color: [1],
+                    disable_control_panel: [1],
+                    stop_after_capturing: [1],
+                    face_detection: [1],
+                    model_path: [1],
+                    probability_threshold: [1],
+                    debug: [1],
+                    logo_style: [1],
+                    _resolutionOnPhoto: [32],
+                    _cameraStatus: [32],
+                    _mobileMakePhoto: [32],
+                    _stopAfterCapturing: [32],
+                    _showMask: [32],
+                    _disable_control_panel: [32],
+                    _face_detection: [32],
+                    _probabilityThreshold: [32],
+                    _debug: [32],
+                    video: [32],
+                    canvas: [32]
+                },
+                [
+                    [0, "videoStarted", "videoStarted"],
+                    [0, "retakePhoto", "retakePhoto"],
+                    [0, "takePhoto", "takePhoto"],
+                    [0, "openCamera", "openCamera"],
+                    [0, "loadModels", "loadModels"],
+                    [0, "errorCamera", "errorCamera"],
+                    [0, "closeCamera", "closeCamera"],
+                    [0, "makePhoto", "makePhoto"]
+                ]
+            ],
+            [0, "camera-comp", {
+                showMask: [4, "show-mask"],
+                modelPath: [1, "model-path"],
+                device: [16],
+                probabilityThreshold: [2, "probability-threshold"],
+                faceDetection: [4, "face-detection"]
+            }],
+            [0, "control-panel", {
+                mobile: [4],
+                disableControlPanel: [4, "disable-control-panel"],
+                cameraStatus: [2, "camera-status"],
+                resolutionOnPhoto: [4, "resolution-on-photo"],
+                mobileMakePhoto: [4, "mobile-make-photo"],
+                faceDetection: [4, "face-detection"]
+            }]
+        ]]
+    ], t));
+
+    function s(e) {
+        return (s = "function" == typeof Symbol && "symbol" == typeof Symbol.iterator ? function (e) {
+            return typeof e
+        } : function (e) {
+            return e && "function" == typeof Symbol && e.constructor === Symbol && e !== Symbol.prototype ? "symbol" : typeof e
+        })(e)
+    }
+
+    function i() {
+        /*! regenerator-runtime -- Copyright (c) 2014-present, Facebook, Inc. -- license (MIT): https://github.com/facebook/regenerator/blob/main/LICENSE */
+        i = function () {
+            return e
+        };
+        var e = {},
+            t = Object.prototype,
+            n = t.hasOwnProperty,
+            r = Object.defineProperty || function (e, t, n) {
+                e[t] = n.value
+            },
+            a = "function" == typeof Symbol ? Symbol : {},
+            o = a.iterator || "@@iterator",
+            u = a.asyncIterator || "@@asyncIterator",
+            l = a.toStringTag || "@@toStringTag";
+
+        function c(e, t, n) {
+            return Object.defineProperty(e, t, {
+                value: n,
+                enumerable: !0,
+                configurable: !0,
+                writable: !0
+            }), e[t]
+        }
+        try {
+            c({}, "")
+        } catch (e) {
+            c = function (e, t, n) {
+                return e[t] = n
+            }
+        }
+
+        function h(e, t, n, a) {
+            var s = t && t.prototype instanceof f ? t : f,
+                i = Object.create(s.prototype),
+                o = new N(a || []);
+            return r(i, "_invoke", {
+                value: S(e, n, o)
+            }), i
+        }
+
+        function p(e, t, n) {
+            try {
+                return {
+                    type: "normal",
+                    arg: e.call(t, n)
+                }
+            } catch (e) {
+                return {
+                    type: "throw",
+                    arg: e
+                }
+            }
+        }
+        e.wrap = h;
+        var d = {};
+
+        function f() {}
+
+        function m() {}
+
+        function g() {}
+        var y = {};
+        c(y, o, (function () {
+            return this
+        }));
+        var b = Object.getPrototypeOf,
+            v = b && b(b(C([])));
+        v && v !== t && n.call(v, o) && (y = v);
+        var x = g.prototype = f.prototype = Object.create(y);
+
+        function w(e) {
+            ["next", "throw", "return"].forEach((function (t) {
+                c(e, t, (function (e) {
+                    return this._invoke(t, e)
+                }))
+            }))
+        }
+
+        function k(e, t) {
+            var a;
+            r(this, "_invoke", {
+                value: function (r, i) {
+                    function o() {
+                        return new t((function (a, o) {
+                            ! function r(a, i, o, u) {
+                                var l = p(e[a], e, i);
+                                if ("throw" !== l.type) {
+                                    var c = l.arg,
+                                        h = c.value;
+                                    return h && "object" == s(h) && n.call(h, "__await") ? t.resolve(h.__await).then((function (e) {
+                                        r("next", e, o, u)
+                                    }), (function (e) {
+                                        r("throw", e, o, u)
+                                    })) : t.resolve(h).then((function (e) {
+                                        c.value = e, o(c)
+                                    }), (function (e) {
+                                        return r("throw", e, o, u)
+                                    }))
+                                }
+                                u(l.arg)
+                            }(r, i, a, o)
+                        }))
+                    }
+                    return a = a ? a.then(o, o) : o()
+                }
+            })
+        }
+
+        function S(e, t, n) {
+            var r = "suspendedStart";
+            return function (a, s) {
+                if ("executing" === r) throw new Error("Generator is already running");
+                if ("completed" === r) {
+                    if ("throw" === a) throw s;
+                    return T()
+                }
+                for (n.method = a, n.arg = s;;) {
+                    var i = n.delegate;
+                    if (i) {
+                        var o = I(i, n);
+                        if (o) {
+                            if (o === d) continue;
+                            return o
+                        }
+                    }
+                    if ("next" === n.method) n.sent = n._sent = n.arg;
+                    else if ("throw" === n.method) {
+                        if ("suspendedStart" === r) throw r = "completed", n.arg;
+                        n.dispatchException(n.arg)
+                    } else "return" === n.method && n.abrupt("return", n.arg);
+                    r = "executing";
+                    var u = p(e, t, n);
+                    if ("normal" === u.type) {
+                        if (r = n.done ? "completed" : "suspendedYield", u.arg === d) continue;
+                        return {
+                            value: u.arg,
+                            done: n.done
+                        }
+                    }
+                    "throw" === u.type && (r = "completed", n.method = "throw", n.arg = u.arg)
+                }
+            }
+        }
+
+        function I(e, t) {
+            var n = t.method,
+                r = e.iterator[n];
+            if (void 0 === r) return t.delegate = null, "throw" === n && e.iterator.return && (t.method = "return", t.arg = void 0, I(e, t), "throw" === t.method) || "return" !== n && (t.method = "throw", t.arg = new TypeError("The iterator does not provide a '" + n + "' method")), d;
+            var a = p(r, e.iterator, t.arg);
+            if ("throw" === a.type) return t.method = "throw", t.arg = a.arg, t.delegate = null, d;
+            var s = a.arg;
+            return s ? s.done ? (t[e.resultName] = s.value, t.next = e.nextLoc, "return" !== t.method && (t.method = "next", t.arg = void 0), t.delegate = null, d) : s : (t.method = "throw", t.arg = new TypeError("iterator result is not an object"), t.delegate = null, d)
+        }
+
+        function A(e) {
+            var t = {
+                tryLoc: e[0]
+            };
+            1 in e && (t.catchLoc = e[1]), 2 in e && (t.finallyLoc = e[2], t.afterLoc = e[3]), this.tryEntries.push(t)
+        }
+
+        function E(e) {
+            var t = e.completion || {};
+            t.type = "normal", delete t.arg, e.completion = t
+        }
+
+        function N(e) {
+            this.tryEntries = [{
+                tryLoc: "root"
+            }], e.forEach(A, this), this.reset(!0)
+        }
+
+        function C(e) {
+            if (e) {
+                var t = e[o];
+                if (t) return t.call(e);
+                if ("function" == typeof e.next) return e;
+                if (!isNaN(e.length)) {
+                    var r = -1,
+                        a = function t() {
+                            for (; ++r < e.length;)
+                                if (n.call(e, r)) return t.value = e[r], t.done = !1, t;
+                            return t.value = void 0, t.done = !0, t
+                        };
+                    return a.next = a
+                }
+            }
+            return {
+                next: T
+            }
+        }
+
+        function T() {
+            return {
+                value: void 0,
+                done: !0
+            }
+        }
+        return m.prototype = g, r(x, "constructor", {
+            value: g,
+            configurable: !0
+        }), r(g, "constructor", {
+            value: m,
+            configurable: !0
+        }), m.displayName = c(g, l, "GeneratorFunction"), e.isGeneratorFunction = function (e) {
+            var t = "function" == typeof e && e.constructor;
+            return !!t && (t === m || "GeneratorFunction" === (t.displayName || t.name))
+        }, e.mark = function (e) {
+            return Object.setPrototypeOf ? Object.setPrototypeOf(e, g) : (e.__proto__ = g, c(e, l, "GeneratorFunction")), e.prototype = Object.create(x), e
+        }, e.awrap = function (e) {
+            return {
+                __await: e
+            }
+        }, w(k.prototype), c(k.prototype, u, (function () {
+            return this
+        })), e.AsyncIterator = k, e.async = function (t, n, r, a, s) {
+            void 0 === s && (s = Promise);
+            var i = new k(h(t, n, r, a), s);
+            return e.isGeneratorFunction(n) ? i : i.next().then((function (e) {
+                return e.done ? e.value : i.next()
+            }))
+        }, w(x), c(x, l, "Generator"), c(x, o, (function () {
+            return this
+        })), c(x, "toString", (function () {
+            return "[object Generator]"
+        })), e.keys = function (e) {
+            var t = Object(e),
+                n = [];
+            for (var r in t) n.push(r);
+            return n.reverse(),
+                function e() {
+                    for (; n.length;) {
+                        var r = n.pop();
+                        if (r in t) return e.value = r, e.done = !1, e
+                    }
+                    return e.done = !0, e
+                }
+        }, e.values = C, N.prototype = {
+            constructor: N,
+            reset: function (e) {
+                if (this.prev = 0, this.next = 0, this.sent = this._sent = void 0, this.done = !1, this.delegate = null, this.method = "next", this.arg = void 0, this.tryEntries.forEach(E), !e)
+                    for (var t in this) "t" === t.charAt(0) && n.call(this, t) && !isNaN(+t.slice(1)) && (this[t] = void 0)
+            },
+            stop: function () {
+                this.done = !0;
+                var e = this.tryEntries[0].completion;
+                if ("throw" === e.type) throw e.arg;
+                return this.rval
+            },
+            dispatchException: function (e) {
+                if (this.done) throw e;
+                var t = this;
+
+                function r(n, r) {
+                    return i.type = "throw", i.arg = e, t.next = n, r && (t.method = "next", t.arg = void 0), !!r
+                }
+                for (var a = this.tryEntries.length - 1; a >= 0; --a) {
+                    var s = this.tryEntries[a],
+                        i = s.completion;
+                    if ("root" === s.tryLoc) return r("end");
+                    if (s.tryLoc <= this.prev) {
+                        var o = n.call(s, "catchLoc"),
+                            u = n.call(s, "finallyLoc");
+                        if (o && u) {
+                            if (this.prev < s.catchLoc) return r(s.catchLoc, !0);
+                            if (this.prev < s.finallyLoc) return r(s.finallyLoc)
+                        } else if (o) {
+                            if (this.prev < s.catchLoc) return r(s.catchLoc, !0)
+                        } else {
+                            if (!u) throw new Error("try statement without catch or finally");
+                            if (this.prev < s.finallyLoc) return r(s.finallyLoc)
+                        }
+                    }
+                }
+            },
+            abrupt: function (e, t) {
+                for (var r = this.tryEntries.length - 1; r >= 0; --r) {
+                    var a = this.tryEntries[r];
+                    if (a.tryLoc <= this.prev && n.call(a, "finallyLoc") && this.prev < a.finallyLoc) {
+                        var s = a;
+                        break
+                    }
+                }
+                s && ("break" === e || "continue" === e) && s.tryLoc <= t && t <= s.finallyLoc && (s = null);
+                var i = s ? s.completion : {};
+                return i.type = e, i.arg = t, s ? (this.method = "next", this.next = s.finallyLoc, d) : this.complete(i)
+            },
+            complete: function (e, t) {
+                if ("throw" === e.type) throw e.arg;
+                return "break" === e.type || "continue" === e.type ? this.next = e.arg : "return" === e.type ? (this.rval = this.arg = e.arg, this.method = "return", this.next = "end") : "normal" === e.type && t && (this.next = t), d
+            },
+            finish: function (e) {
+                for (var t = this.tryEntries.length - 1; t >= 0; --t) {
+                    var n = this.tryEntries[t];
+                    if (n.finallyLoc === e) return this.complete(n.completion, n.afterLoc), E(n), d
+                }
+            },
+            catch: function (e) {
+                for (var t = this.tryEntries.length - 1; t >= 0; --t) {
+                    var n = this.tryEntries[t];
+                    if (n.tryLoc === e) {
+                        var r = n.completion;
+                        if ("throw" === r.type) {
+                            var a = r.arg;
+                            E(n)
+                        }
+                        return a
+                    }
+                }
+                throw new Error("illegal catch attempt")
+            },
+            delegateYield: function (e, t, n) {
+                return this.delegate = {
+                    iterator: C(e),
+                    resultName: t,
+                    nextLoc: n
+                }, "next" === this.method && (this.arg = void 0), d
+            }
+        }, e
+    }
+
+    function o(e, t, n, r, a, s, i) {
+        try {
+            var o = e[s](i),
+                u = o.value
+        } catch (e) {
+            return void n(e)
+        }
+        o.done ? t(u) : Promise.resolve(u).then(r, a)
+    }
+
+    function u(e) {
+        return function () {
+            var t = this,
+                n = arguments;
+            return new Promise((function (r, a) {
+                var s = e.apply(t, n);
+
+                function i(e) {
+                    o(s, r, a, i, u, "next", e)
+                }
+
+                function u(e) {
+                    o(s, r, a, i, u, "throw", e)
+                }
+                i(void 0)
+            }))
+        }
+    }! function () {
+        if ("undefined" != typeof window && void 0 !== window.Reflect && void 0 !== window.customElements) {
+            var e = HTMLElement;
+            window.HTMLElement = function () {
+                return Reflect.construct(e, [], this.constructor)
+            }, HTMLElement.prototype = e.prototype, HTMLElement.prototype.constructor = HTMLElement, Object.setPrototypeOf(HTMLElement, e)
+        }
+    }(),
+    function () {
+        var e = [];
+        if ("undefined" != typeof window) {
+            var t = window;
+            t.customElements && (!t.Element || t.Element.prototype.closest && t.Element.prototype.matches && t.Element.prototype.remove && t.Element.prototype.getRootNode) || e.push(Promise.resolve().then(n.t.bind(null, 248, 7)));
+            "function" == typeof Object.assign && Object.entries && Array.prototype.find && Array.prototype.includes && String.prototype.startsWith && String.prototype.endsWith && (!t.NodeList || t.NodeList.prototype.forEach) && t.fetch && function () {
+                try {
+                    var e = new URL("b", "http://a");
+                    return e.pathname = "c%20d", "http://a/c%20d" === e.href && e.searchParams
+                } catch (e) {
+                    return !1
+                }
+            }() && "undefined" != typeof WeakMap || e.push(Promise.resolve().then(n.t.bind(null, 249, 7)))
+        }
+        return Promise.all(e)
+    }().then((function () {
+        a(window)
+    })), window.AcuantPassiveLiveness = function () {
+        var e, t = null,
+            n = null,
+            r = null,
+            a = null,
+            s = null,
+            o = null,
+            l = null,
+            c = null,
+            h = null,
+            p = null,
+            d = null,
+            f = {
+                start: function (e,states) {
+                    d = states ? states : {
+                        FACE_NOT_FOUND: "FACE NOT FOUND",
+                        TOO_MANY_FACES: "TOO MANY FACES",
+                        FACE_ANGLE_TOO_LARGE: "FACE ANGLE TOO LARGE",
+                        PROBABILITY_TOO_SMALL: "PROBABILITY TOO SMALL",
+                        FACE_TOO_SMALL: "FACE TOO SMALL",
+                        FACE_CLOSE_TO_BORDER: "TOO CLOSE TO THE FRAME",        
+                        CLOSE_TEXT: "close",
+                        RETAKE_TEXT: "retake",
+                        INTRO_TEXT: "Camera is on, ready for selfie. If possible center your face in the camera and it will automatically take the picture. Otherwise select this button for manual capture.",
+                        SUBMIT_ALT: "Photo has been taken, select this button to submit this photo.",
+                        CAPTURE_ALT: "Take photo"
+                    };
+                    window.acuant_passive_liveness_language = d;
+                    var t = arguments.length > 1 && void 0 !== arguments[1] ? arguments[1] : null;
+                    m() ? (t && (d = t), x(), I(e), w()) : B(e.onCaptured);
+                },
+                startManualCapture: function (e) {
+                    return B(e)
+                },
+                end: function () {
+                    return k()
+                },
+                getLiveness: (e = u(i().mark((function e(t, n) {
+                    var r, a, s, o, u, l, c;
+                    return i().wrap((function (e) {
+                        for (;;) switch (e.prev = e.next) {
+                            case 0:
+                                return r = t.endpoint, a = t.token, s = t.subscriptionId, o = t.image, u = {
+                                    Settings: {
+                                        SubscriptionId: s,
+                                        AdditionalSettings: {
+                                            OS: "UNKNOWN"
+                                        }
+                                    },
+                                    Image: o
+                                }, e.prev = 2, e.next = 5, fetch(r + "/api/v1/liveness", {
+                                    method: "POST",
+                                    body: JSON.stringify(u),
+                                    headers: {
+                                        "Content-Type": "application/json",
+                                        Authorization: "Basic " + a,
+                                        Accept: "application/json"
+                                    }
+                                });
+                            case 5:
+                                return l = e.sent, e.next = 8, l.json();
+                            case 8:
+                                c = e.sent, n(c), e.next = 16;
+                                break;
+                            case 12:
+                                e.prev = 12, e.t0 = e.catch(2), console.log("Error while getting liveness: " + e.t0), n(null);
+                            case 16:
+                            case "end":
+                                return e.stop()
+                        }
+                    }), e, null, [
+                        [2, 12]
+                    ])
+                }))), function (t, n) {
+                    return e.apply(this, arguments)
+                })
+            },
+            m = function () {
+                return "mediaDevices" in navigator && g()
+            },
+            g = function () {
+                var e, t = !1;
+                return e = navigator.userAgent || navigator.vendor || window.opera, (/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino|android|playbook|silk/i.test(e) || /1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(e.substr(0, 4))) && (t = !0), (t || y()) && !v()
+            },
+            y = function () {
+                return /iPad|iPhone|iPod/.test(navigator.platform) || b()
+            },
+            b = function () {
+                return "MacIntel" === navigator.platform && navigator.maxTouchPoints > 1
+            },
+            v = function () {
+                return navigator.userAgent.toLowerCase().indexOf("firefox") > -1
+            },
+            x = function () {
+                if (!(n = document.getElementById("acuant-face-capture-camera"))) {
+                    var e = document.getElementById("acuant-face-capture-container");
+                    (n = document.createElement("camera-component")).id = "acuant-face-capture-camera", n.model_path = acuantConfig.path, n.url_logo = " ", n.logo_style = '{"display": "none" }', n.show_mask = !0, n.disable_control_panel = !0, n.stop_after_capturing = !0, n.face_detection = !0, n.background_color = "white", n.probability_threshold = 70, e.appendChild(n)
+                }
+            },
+            w = function () {
+                var e, t, r, a;
+                
+                // (-1 == (a = navigator.userAgent.toLowerCase()).indexOf("safari") || a.indexOf("chrome") > -1) && (e = window.navigator.standalone, t = window.navigator.userAgent.toLowerCase(), r = /mobile.*safari/.test(t), !/iphone|ipod|ipad/.test(t) || r || e) ? n.parentElement.requestFullscreen().then((function () {
+                //     S()
+                // })).catch((function () {
+                //     console.log("Error while requesting full screen. Launching camera."), S()
+                // })): 
+                S();
+            },
+            k = function () {
+                if (n) {
+                    var e = new Event("closeCamera");
+                    n.dispatchEvent(e)
+                }
+            },
+            S = function () {
+                var e = new Event("openCamera");
+                n.dispatchEvent(e)
+            },
+            I = function (e) {
+                D(e.onError), L(e.onOpened), P(e.onClosed), _(e.onDetection), C(e.onCaptured), E(e.onPhotoTaken), N(e.onPhotoRetake), A(e.onDetectorInitialized)
+            },
+            A = function (e) {
+                p = e, n.addEventListener("detectorInitialized", p)
+            },
+            E = function (e) {
+                r = e, n.addEventListener("makePhoto", r)
+            },
+            N = function (e) {
+                a = e, n.addEventListener("retakePhoto", a)
+            },
+            C = function (e) {
+                s = T(e), n.addEventListener("capture", s)
+            },
+            T = function (e) {
+                return function (t) {
+                    q();
+                    var n = t.detail.photos[0];
+                    R(n, (function (t) {
+                        document.fullscreenElement ? document.exitFullscreen().then((function () {
+                            U(t, e)
+                        })) : U(t, e)
+                    }))
+                }
+            },
+            R = function (e, t) {
+                var n = new FileReader;
+                n.onloadend = function () {
+                    return t(n.result)
+                }, n.readAsDataURL(e)
+            },
+            _ = function (e) {
+                o = F(e), n.addEventListener("detectionError", o)
+            },
+            F = function (e) {
+                return function (t) {
+                    var n = t.detail.error,
+                        r = M(n, !0);
+                    e(d[r])
+                }
+            },
+            M = function (e, t) {
+                return Object.keys(e).find((function (n) {
+                    return e[n] === t
+                }))
+            },
+            D = function (e) {
+                l = O(e), n.addEventListener("error", l)
+            },
+            O = function (e) {
+                return function (t) {
+                    q();
+                    var n = {
+                        code: 1,
+                        message: "Camera permission not granted"
+                    };
+                    "isTrusted" in t ? document.fullscreenElement ? document.exitFullscreen().then((function () {
+                        e(n)
+                    })) : e(n) : e({
+                        code: 2,
+                        message: "Unexpected error: " + t
+                    })
+                }
+            },
+            L = function (e) {
+                c = e, n.addEventListener("openCamera", c)
+            },
+            P = function (e) {
+                h = $(e), n.addEventListener("closeCamera", h)
+            },
+            $ = function (e) {
+                return function () {
+                    q(), document.fullscreenElement ? document.exitFullscreen().then((function () {
+                        e()
+                    })) : e()
+                }
+            },
+            B = function (e) {
+                t || ((t = document.createElement("input")).type = "file", t.capture = "user", t.accept = "image/*", t.onchange = function (t) {
+                    return z(t, e)
+                }), t.click()
+            },
+            z = function (e, t) {
+                var n = e.target,
+                    r = new FileReader;
+                r.onload = function (e) {
+                    return U(e.target.result, t)
+                }, r.readAsDataURL(n.files[0])
+            },
+            U = function (e, t) {
+                var n = document.createElement("img"),
+                    r = document.createElement("canvas");
+                n.onload = function () {
+                    var e = 1080,
+                        a = 720,
+                        s = n.width,
+                        i = n.height,
+                        o = r.getContext("2d");
+                    (s > i ? i : s) > a ? s < i ? a = (e = a) * (i / s) : e = a * (s / i) : (e = n.width, a = n.height);
+                    r.width = e, r.height = a, o.drawImage(n, 0, 0, e, a);
+                    var u = r.toDataURL("image/jpeg", .8);
+                    o.clearRect(0, 0, e, a);
+                    var l = j(u);
+                    W(l).then(t)
+                }, n.src = e
+            },
+            W = function () {
+                var e = u(i().mark((function e(t) {
+                    var n, r;
+                    return i().wrap((function (e) {
+                        for (;;) switch (e.prev = e.next) {
+                            case 0:
+                                if (n = G(t), !AcuantJavascriptWebSdk.singleWorkerModel) {
+                                    e.next = 11;
+                                    break
+                                }
+                                return e.next = 4, new Promise((function (e) {
+                                    AcuantJavascriptWebSdk.startImageWorker(e)
+                                }));
+                            case 4:
+                                return e.next = 6, V(n);
+                            case 6:
+                                return r = e.sent, AcuantJavascriptWebSdk.endImageWorker(), e.abrupt("return", r);
+                            case 11:
+                                return e.next = 13, V(n);
+                            case 13:
+                                return e.abrupt("return", e.sent);
+                            case 14:
+                            case "end":
+                                return e.stop()
+                        }
+                    }), e)
+                })));
+                return function (t) {
+                    return e.apply(this, arguments)
+                }
+            }();
+
+        function V(e) {
+            return new Promise((function (t) {
+                AcuantJavascriptWebSdk.sign(e, {
+                    onSuccess: function (e) {
+                        return t(H(e))
+                    },
+                    onFail: t
+                })
+            }))
+        }
+        var j = function (e) {
+                return AcuantJavascriptWebSdk.addMetadata(e, {
+                    dateTimeOriginal: (new Date).toUTCString()
+                })
+            },
+            G = function (e) {
+                for (var t = window.atob(e.split("base64,")[1]), n = t.length, r = new Uint8Array(new ArrayBuffer(n)), a = 0; a < n; a++) r[a] = t.charCodeAt(a);
+                return r
+            },
+            H = function (e) {
+                for (var t = "", n = new Uint8Array(e), r = n.byteLength, a = 0; a < r; a++) t += String.fromCharCode(n[a]);
+                return window.btoa(t)
+            },
+            q = function () {
+                n.removeEventListener("makePhoto", r), n.removeEventListener("retakePhoto", a), n.removeEventListener("capture", s), n.removeEventListener("detectionError", o), n.removeEventListener("error", l), n.removeEventListener("openCamera", c), n.removeEventListener("closeCamera", h), n.removeEventListener("detectorInitialized", p), r = null, a = null, s = null, o = null, l = null, c = null, h = null, p = null, n = null
+            };
+        return f
+    }()
+}, function (e, t) {
+    var n, r, a, s, i;
+    (function () {
+        "use strict";
+        var e = new Set("annotation-xml color-profile font-face font-face-src font-face-uri font-face-format font-face-name missing-glyph".split(" "));
+
+        function t(t) {
+            var n = e.has(t);
+            return t = /^[a-z][.0-9_a-z]*-[\-.0-9_a-z]*$/.test(t), !n && t
+        }
+
+        function n(e) {
+            var t = e.isConnected;
+            if (void 0 !== t) return t;
+            for (; e && !(e.__CE_isImportDocument || e instanceof Document);) e = e.parentNode || (window.ShadowRoot && e instanceof ShadowRoot ? e.host : void 0);
+            return !(!e || !(e.__CE_isImportDocument || e instanceof Document))
+        }
+
+        function r(e, t) {
+            for (; t && t !== e && !t.nextSibling;) t = t.parentNode;
+            return t && t !== e ? t.nextSibling : null
+        }
+
+        function a(e, t, n) {
+            n = void 0 === n ? new Set : n;
+            for (var s = e; s;) {
+                if (s.nodeType === Node.ELEMENT_NODE) {
+                    var i = s;
+                    t(i);
+                    var o = i.localName;
+                    if ("link" === o && "import" === i.getAttribute("rel")) {
+                        if ((s = i.import) instanceof Node && !n.has(s))
+                            for (n.add(s), s = s.firstChild; s; s = s.nextSibling) a(s, t, n);
+                        s = r(e, i);
+                        continue
+                    }
+                    if ("template" === o) {
+                        s = r(e, i);
+                        continue
+                    }
+                    if (i = i.__CE_shadowRoot)
+                        for (i = i.firstChild; i; i = i.nextSibling) a(i, t, n)
+                }
+                s = s.firstChild ? s.firstChild : r(e, s)
+            }
+        }
+
+        function s(e, t, n) {
+            e[t] = n
+        }
+
+        function i() {
+            this.a = new Map, this.g = new Map, this.c = [], this.f = [], this.b = !1
+        }
+
+        function o(e, t) {
+            e.b && a(t, (function (t) {
+                return u(e, t)
+            }))
+        }
+
+        function u(e, t) {
+            if (e.b && !t.__CE_patched) {
+                t.__CE_patched = !0;
+                for (var n = 0; n < e.c.length; n++) e.c[n](t);
+                for (n = 0; n < e.f.length; n++) e.f[n](t)
+            }
+        }
+
+        function l(e, t) {
+            var n = [];
+            for (a(t, (function (e) {
+                    return n.push(e)
+                })), t = 0; t < n.length; t++) {
+                var r = n[t];
+                1 === r.__CE_state ? e.connectedCallback(r) : p(e, r)
+            }
+        }
+
+        function c(e, t) {
+            var n = [];
+            for (a(t, (function (e) {
+                    return n.push(e)
+                })), t = 0; t < n.length; t++) {
+                var r = n[t];
+                1 === r.__CE_state && e.disconnectedCallback(r)
+            }
+        }
+
+        function h(e, t, n) {
+            var r = (n = void 0 === n ? {} : n).u || new Set,
+                s = n.i || function (t) {
+                    return p(e, t)
+                },
+                i = [];
+            if (a(t, (function (t) {
+                    if ("link" === t.localName && "import" === t.getAttribute("rel")) {
+                        var n = t.import;
+                        n instanceof Node && (n.__CE_isImportDocument = !0, n.__CE_hasRegistry = !0), n && "complete" === n.readyState ? n.__CE_documentLoadHandled = !0 : t.addEventListener("load", (function () {
+                            var n = t.import;
+                            if (!n.__CE_documentLoadHandled) {
+                                n.__CE_documentLoadHandled = !0;
+                                var a = new Set(r);
+                                a.delete(n), h(e, n, {
+                                    u: a,
+                                    i: s
+                                })
+                            }
+                        }))
+                    } else i.push(t)
+                }), r), e.b)
+                for (t = 0; t < i.length; t++) u(e, i[t]);
+            for (t = 0; t < i.length; t++) s(i[t])
+        }
+
+        function p(e, t) {
+            if (void 0 === t.__CE_state) {
+                var r = t.ownerDocument;
+                if ((r.defaultView || r.__CE_isImportDocument && r.__CE_hasRegistry) && (r = e.a.get(t.localName))) {
+                    r.constructionStack.push(t);
+                    var a = r.constructorFunction;
+                    try {
+                        try {
+                            if (new a !== t) throw Error("The custom element constructor did not produce the element being upgraded.")
+                        } finally {
+                            r.constructionStack.pop()
+                        }
+                    } catch (e) {
+                        throw t.__CE_state = 2, e
+                    }
+                    if (t.__CE_state = 1, t.__CE_definition = r, r.attributeChangedCallback)
+                        for (r = r.observedAttributes, a = 0; a < r.length; a++) {
+                            var s = r[a],
+                                i = t.getAttribute(s);
+                            null !== i && e.attributeChangedCallback(t, s, null, i, null)
+                        }
+                    n(t) && e.connectedCallback(t)
+                }
+            }
+        }
+
+        function d(e) {
+            var t = document;
+            this.c = e, this.a = t, this.b = void 0, h(this.c, this.a), "loading" === this.a.readyState && (this.b = new MutationObserver(this.f.bind(this)), this.b.observe(this.a, {
+                childList: !0,
+                subtree: !0
+            }))
+        }
+
+        function f(e) {
+            e.b && e.b.disconnect()
+        }
+
+        function m() {
+            var e = this;
+            this.b = this.a = void 0, this.c = new Promise((function (t) {
+                e.b = t, e.a && t(e.a)
+            }))
+        }
+
+        function g(e) {
+            if (e.a) throw Error("Already resolved.");
+            e.a = void 0, e.b && e.b(void 0)
+        }
+
+        function y(e) {
+            this.c = !1, this.a = e, this.j = new Map, this.f = function (e) {
+                return e()
+            }, this.b = !1, this.g = [], this.o = new d(e)
+        }
+        i.prototype.connectedCallback = function (e) {
+            var t = e.__CE_definition;
+            t.connectedCallback && t.connectedCallback.call(e)
+        }, i.prototype.disconnectedCallback = function (e) {
+            var t = e.__CE_definition;
+            t.disconnectedCallback && t.disconnectedCallback.call(e)
+        }, i.prototype.attributeChangedCallback = function (e, t, n, r, a) {
+            var s = e.__CE_definition;
+            s.attributeChangedCallback && -1 < s.observedAttributes.indexOf(t) && s.attributeChangedCallback.call(e, t, n, r, a)
+        }, d.prototype.f = function (e) {
+            var t = this.a.readyState;
+            for ("interactive" !== t && "complete" !== t || f(this), t = 0; t < e.length; t++)
+                for (var n = e[t].addedNodes, r = 0; r < n.length; r++) h(this.c, n[r])
+        }, y.prototype.l = function (e, n) {
+            var r = this;
+            if (!(n instanceof Function)) throw new TypeError("Custom element constructors must be functions.");
+            if (!t(e)) throw new SyntaxError("The element name '" + e + "' is not valid.");
+            if (this.a.a.get(e)) throw Error("A custom element with name '" + e + "' has already been defined.");
+            if (this.c) throw Error("A custom element is already being defined.");
+            this.c = !0;
+            try {
+                var a = function (e) {
+                        var t = s[e];
+                        if (void 0 !== t && !(t instanceof Function)) throw Error("The '" + e + "' callback must be a function.");
+                        return t
+                    },
+                    s = n.prototype;
+                if (!(s instanceof Object)) throw new TypeError("The custom element constructor's prototype is not an object.");
+                var i = a("connectedCallback"),
+                    o = a("disconnectedCallback"),
+                    u = a("adoptedCallback"),
+                    l = a("attributeChangedCallback"),
+                    c = n.observedAttributes || []
+            } catch (e) {
+                return
+            } finally {
+                this.c = !1
+            }
+            n = {
+                    localName: e,
+                    constructorFunction: n,
+                    connectedCallback: i,
+                    disconnectedCallback: o,
+                    adoptedCallback: u,
+                    attributeChangedCallback: l,
+                    observedAttributes: c,
+                    constructionStack: []
+                },
+                function (e, t, n) {
+                    e.a.set(t, n), e.g.set(n.constructorFunction, n)
+                }(this.a, e, n), this.g.push(n), this.b || (this.b = !0, this.f((function () {
+                    return function (e) {
+                        if (!1 !== e.b) {
+                            e.b = !1;
+                            for (var t = e.g, n = [], r = new Map, a = 0; a < t.length; a++) r.set(t[a].localName, []);
+                            for (h(e.a, document, {
+                                    i: function (t) {
+                                        if (void 0 === t.__CE_state) {
+                                            var a = t.localName,
+                                                s = r.get(a);
+                                            s ? s.push(t) : e.a.a.get(a) && n.push(t)
+                                        }
+                                    }
+                                }), a = 0; a < n.length; a++) p(e.a, n[a]);
+                            for (; 0 < t.length;) {
+                                var s = t.shift();
+                                a = s.localName, s = r.get(s.localName);
+                                for (var i = 0; i < s.length; i++) p(e.a, s[i]);
+                                (a = e.j.get(a)) && g(a)
+                            }
+                        }
+                    }(r)
+                })))
+        }, y.prototype.i = function (e) {
+            h(this.a, e)
+        }, y.prototype.get = function (e) {
+            if (e = this.a.a.get(e)) return e.constructorFunction
+        }, y.prototype.m = function (e) {
+            if (!t(e)) return Promise.reject(new SyntaxError("'" + e + "' is not a valid custom element name."));
+            var n = this.j.get(e);
+            return n || (n = new m, this.j.set(e, n), this.a.a.get(e) && !this.g.some((function (t) {
+                return t.localName === e
+            })) && g(n)), n.c
+        }, y.prototype.s = function (e) {
+            f(this.o);
+            var t = this.f;
+            this.f = function (n) {
+                return e((function () {
+                    return t(n)
+                }))
+            }
+        }, window.CustomElementRegistry = y, y.prototype.define = y.prototype.l, y.prototype.upgrade = y.prototype.i, y.prototype.get = y.prototype.get, y.prototype.whenDefined = y.prototype.m, y.prototype.polyfillWrapFlushCallback = y.prototype.s;
+        var b = window.Document.prototype.createElement,
+            v = window.Document.prototype.createElementNS,
+            x = window.Document.prototype.importNode,
+            w = window.Document.prototype.prepend,
+            k = window.Document.prototype.append,
+            S = window.DocumentFragment.prototype.prepend,
+            I = window.DocumentFragment.prototype.append,
+            A = window.Node.prototype.cloneNode,
+            E = window.Node.prototype.appendChild,
+            N = window.Node.prototype.insertBefore,
+            C = window.Node.prototype.removeChild,
+            T = window.Node.prototype.replaceChild,
+            R = Object.getOwnPropertyDescriptor(window.Node.prototype, "textContent"),
+            _ = window.Element.prototype.attachShadow,
+            F = Object.getOwnPropertyDescriptor(window.Element.prototype, "innerHTML"),
+            M = window.Element.prototype.getAttribute,
+            D = window.Element.prototype.setAttribute,
+            O = window.Element.prototype.removeAttribute,
+            L = window.Element.prototype.getAttributeNS,
+            P = window.Element.prototype.setAttributeNS,
+            $ = window.Element.prototype.removeAttributeNS,
+            B = window.Element.prototype.insertAdjacentElement,
+            z = window.Element.prototype.insertAdjacentHTML,
+            U = window.Element.prototype.prepend,
+            W = window.Element.prototype.append,
+            V = window.Element.prototype.before,
+            j = window.Element.prototype.after,
+            G = window.Element.prototype.replaceWith,
+            H = window.Element.prototype.remove,
+            q = window.HTMLElement,
+            K = Object.getOwnPropertyDescriptor(window.HTMLElement.prototype, "innerHTML"),
+            Y = window.HTMLElement.prototype.insertAdjacentElement,
+            X = window.HTMLElement.prototype.insertAdjacentHTML,
+            Q = new function () {};
+
+        function Z(e, t, r) {
+            function a(t) {
+                return function (r) {
+                    for (var a = [], s = 0; s < arguments.length; ++s) a[s] = arguments[s];
+                    s = [];
+                    for (var i = [], o = 0; o < a.length; o++) {
+                        var u = a[o];
+                        if (u instanceof Element && n(u) && i.push(u), u instanceof DocumentFragment)
+                            for (u = u.firstChild; u; u = u.nextSibling) s.push(u);
+                        else s.push(u)
+                    }
+                    for (t.apply(this, a), a = 0; a < i.length; a++) c(e, i[a]);
+                    if (n(this))
+                        for (a = 0; a < s.length; a++)(i = s[a]) instanceof Element && l(e, i)
+                }
+            }
+            void 0 !== r.h && (t.prepend = a(r.h)), void 0 !== r.append && (t.append = a(r.append))
+        }
+        var J, ee = window.customElements;
+        if (!ee || ee.forcePolyfill || "function" != typeof ee.define || "function" != typeof ee.get) {
+            var te = new i;
+            J = te, window.HTMLElement = function () {
+                    function e() {
+                        var e = this.constructor,
+                            t = J.g.get(e);
+                        if (!t) throw Error("The custom element being constructed was not registered with `customElements`.");
+                        var n = t.constructionStack;
+                        if (0 === n.length) return n = b.call(document, t.localName), Object.setPrototypeOf(n, e.prototype), n.__CE_state = 1, n.__CE_definition = t, u(J, n), n;
+                        var r = n[t = n.length - 1];
+                        if (r === Q) throw Error("The HTMLElement constructor was either called reentrantly for this constructor or called multiple times.");
+                        return n[t] = Q, Object.setPrototypeOf(r, e.prototype), u(J, r), r
+                    }
+                    return e.prototype = q.prototype, Object.defineProperty(e.prototype, "constructor", {
+                        writable: !0,
+                        configurable: !0,
+                        enumerable: !1,
+                        value: e
+                    }), e
+                }(),
+                function () {
+                    var e = te;
+                    s(Document.prototype, "createElement", (function (t) {
+                        if (this.__CE_hasRegistry) {
+                            var n = e.a.get(t);
+                            if (n) return new n.constructorFunction
+                        }
+                        return t = b.call(this, t), u(e, t), t
+                    })), s(Document.prototype, "importNode", (function (t, n) {
+                        return t = x.call(this, t, !!n), this.__CE_hasRegistry ? h(e, t) : o(e, t), t
+                    })), s(Document.prototype, "createElementNS", (function (t, n) {
+                        if (this.__CE_hasRegistry && (null === t || "http://www.w3.org/1999/xhtml" === t)) {
+                            var r = e.a.get(n);
+                            if (r) return new r.constructorFunction
+                        }
+                        return t = v.call(this, t, n), u(e, t), t
+                    })), Z(e, Document.prototype, {
+                        h: w,
+                        append: k
+                    })
+                }(), Z(te, DocumentFragment.prototype, {
+                    h: S,
+                    append: I
+                }),
+                function () {
+                    function e(e, r) {
+                        Object.defineProperty(e, "textContent", {
+                            enumerable: r.enumerable,
+                            configurable: !0,
+                            get: r.get,
+                            set: function (e) {
+                                if (this.nodeType === Node.TEXT_NODE) r.set.call(this, e);
+                                else {
+                                    var a = void 0;
+                                    if (this.firstChild) {
+                                        var s = this.childNodes,
+                                            i = s.length;
+                                        if (0 < i && n(this)) {
+                                            a = Array(i);
+                                            for (var o = 0; o < i; o++) a[o] = s[o]
+                                        }
+                                    }
+                                    if (r.set.call(this, e), a)
+                                        for (e = 0; e < a.length; e++) c(t, a[e])
+                                }
+                            }
+                        })
+                    }
+                    var t = te;
+                    s(Node.prototype, "insertBefore", (function (e, r) {
+                        if (e instanceof DocumentFragment) {
+                            var a = Array.prototype.slice.apply(e.childNodes);
+                            if (e = N.call(this, e, r), n(this))
+                                for (r = 0; r < a.length; r++) l(t, a[r]);
+                            return e
+                        }
+                        return a = n(e), r = N.call(this, e, r), a && c(t, e), n(this) && l(t, e), r
+                    })), s(Node.prototype, "appendChild", (function (e) {
+                        if (e instanceof DocumentFragment) {
+                            var r = Array.prototype.slice.apply(e.childNodes);
+                            if (e = E.call(this, e), n(this))
+                                for (var a = 0; a < r.length; a++) l(t, r[a]);
+                            return e
+                        }
+                        return r = n(e), a = E.call(this, e), r && c(t, e), n(this) && l(t, e), a
+                    })), s(Node.prototype, "cloneNode", (function (e) {
+                        return e = A.call(this, !!e), this.ownerDocument.__CE_hasRegistry ? h(t, e) : o(t, e), e
+                    })), s(Node.prototype, "removeChild", (function (e) {
+                        var r = n(e),
+                            a = C.call(this, e);
+                        return r && c(t, e), a
+                    })), s(Node.prototype, "replaceChild", (function (e, r) {
+                        if (e instanceof DocumentFragment) {
+                            var a = Array.prototype.slice.apply(e.childNodes);
+                            if (e = T.call(this, e, r), n(this))
+                                for (c(t, r), r = 0; r < a.length; r++) l(t, a[r]);
+                            return e
+                        }
+                        a = n(e);
+                        var s = T.call(this, e, r),
+                            i = n(this);
+                        return i && c(t, r), a && c(t, e), i && l(t, e), s
+                    })), R && R.get ? e(Node.prototype, R) : function (e, t) {
+                        e.b = !0, e.c.push(t)
+                    }(t, (function (t) {
+                        e(t, {
+                            enumerable: !0,
+                            configurable: !0,
+                            get: function () {
+                                for (var e = [], t = 0; t < this.childNodes.length; t++) {
+                                    var n = this.childNodes[t];
+                                    n.nodeType !== Node.COMMENT_NODE && e.push(n.textContent)
+                                }
+                                return e.join("")
+                            },
+                            set: function (e) {
+                                for (; this.firstChild;) C.call(this, this.firstChild);
+                                null != e && "" !== e && E.call(this, document.createTextNode(e))
+                            }
+                        })
+                    }))
+                }(),
+                function () {
+                    function e(e, t) {
+                        Object.defineProperty(e, "innerHTML", {
+                            enumerable: t.enumerable,
+                            configurable: !0,
+                            get: t.get,
+                            set: function (e) {
+                                var r = this,
+                                    s = void 0;
+                                if (n(this) && (s = [], a(this, (function (e) {
+                                        e !== r && s.push(e)
+                                    }))), t.set.call(this, e), s)
+                                    for (var u = 0; u < s.length; u++) {
+                                        var l = s[u];
+                                        1 === l.__CE_state && i.disconnectedCallback(l)
+                                    }
+                                return this.ownerDocument.__CE_hasRegistry ? h(i, this) : o(i, this), e
+                            }
+                        })
+                    }
+
+                    function t(e, t) {
+                        s(e, "insertAdjacentElement", (function (e, r) {
+                            var a = n(r);
+                            return e = t.call(this, e, r), a && c(i, r), n(e) && l(i, r), e
+                        }))
+                    }
+
+                    function r(e, t) {
+                        function n(e, t) {
+                            for (var n = []; e !== t; e = e.nextSibling) n.push(e);
+                            for (t = 0; t < n.length; t++) h(i, n[t])
+                        }
+                        s(e, "insertAdjacentHTML", (function (e, r) {
+                            if ("beforebegin" === (e = e.toLowerCase())) {
+                                var a = this.previousSibling;
+                                t.call(this, e, r), n(a || this.parentNode.firstChild, this)
+                            } else if ("afterbegin" === e) a = this.firstChild, t.call(this, e, r), n(this.firstChild, a);
+                            else if ("beforeend" === e) a = this.lastChild, t.call(this, e, r), n(a || this.firstChild, null);
+                            else {
+                                if ("afterend" !== e) throw new SyntaxError("The value provided (" + String(e) + ") is not one of 'beforebegin', 'afterbegin', 'beforeend', or 'afterend'.");
+                                a = this.nextSibling, t.call(this, e, r), n(this.nextSibling, a)
+                            }
+                        }))
+                    }
+                    var i = te;
+                    _ && s(Element.prototype, "attachShadow", (function (e) {
+                            e = _.call(this, e);
+                            var t = i;
+                            if (t.b && !e.__CE_patched) {
+                                e.__CE_patched = !0;
+                                for (var n = 0; n < t.c.length; n++) t.c[n](e)
+                            }
+                            return this.__CE_shadowRoot = e
+                        })), F && F.get ? e(Element.prototype, F) : K && K.get ? e(HTMLElement.prototype, K) : function (e, t) {
+                            e.b = !0, e.f.push(t)
+                        }(i, (function (t) {
+                            e(t, {
+                                enumerable: !0,
+                                configurable: !0,
+                                get: function () {
+                                    return A.call(this, !0).innerHTML
+                                },
+                                set: function (e) {
+                                    var t = "template" === this.localName,
+                                        n = t ? this.content : this,
+                                        r = v.call(document, this.namespaceURI, this.localName);
+                                    for (r.innerHTML = e; 0 < n.childNodes.length;) C.call(n, n.childNodes[0]);
+                                    for (e = t ? r.content : r; 0 < e.childNodes.length;) E.call(n, e.childNodes[0])
+                                }
+                            })
+                        })), s(Element.prototype, "setAttribute", (function (e, t) {
+                            if (1 !== this.__CE_state) return D.call(this, e, t);
+                            var n = M.call(this, e);
+                            D.call(this, e, t), t = M.call(this, e), i.attributeChangedCallback(this, e, n, t, null)
+                        })), s(Element.prototype, "setAttributeNS", (function (e, t, n) {
+                            if (1 !== this.__CE_state) return P.call(this, e, t, n);
+                            var r = L.call(this, e, t);
+                            P.call(this, e, t, n), n = L.call(this, e, t), i.attributeChangedCallback(this, t, r, n, e)
+                        })), s(Element.prototype, "removeAttribute", (function (e) {
+                            if (1 !== this.__CE_state) return O.call(this, e);
+                            var t = M.call(this, e);
+                            O.call(this, e), null !== t && i.attributeChangedCallback(this, e, t, null, null)
+                        })), s(Element.prototype, "removeAttributeNS", (function (e, t) {
+                            if (1 !== this.__CE_state) return $.call(this, e, t);
+                            var n = L.call(this, e, t);
+                            $.call(this, e, t);
+                            var r = L.call(this, e, t);
+                            n !== r && i.attributeChangedCallback(this, t, n, r, e)
+                        })), Y ? t(HTMLElement.prototype, Y) : B ? t(Element.prototype, B) : console.warn("Custom Elements: `Element#insertAdjacentElement` was not patched."), X ? r(HTMLElement.prototype, X) : z ? r(Element.prototype, z) : console.warn("Custom Elements: `Element#insertAdjacentHTML` was not patched."), Z(i, Element.prototype, {
+                            h: U,
+                            append: W
+                        }),
+                        function (e) {
+                            function t(t) {
+                                return function (r) {
+                                    for (var a = [], s = 0; s < arguments.length; ++s) a[s] = arguments[s];
+                                    s = [];
+                                    for (var i = [], o = 0; o < a.length; o++) {
+                                        var u = a[o];
+                                        if (u instanceof Element && n(u) && i.push(u), u instanceof DocumentFragment)
+                                            for (u = u.firstChild; u; u = u.nextSibling) s.push(u);
+                                        else s.push(u)
+                                    }
+                                    for (t.apply(this, a), a = 0; a < i.length; a++) c(e, i[a]);
+                                    if (n(this))
+                                        for (a = 0; a < s.length; a++)(i = s[a]) instanceof Element && l(e, i)
+                                }
+                            }
+                            var r = Element.prototype;
+                            void 0 !== V && (r.before = t(V)), void 0 !== V && (r.after = t(j)), void 0 !== G && s(r, "replaceWith", (function (t) {
+                                for (var r = [], a = 0; a < arguments.length; ++a) r[a] = arguments[a];
+                                a = [];
+                                for (var s = [], i = 0; i < r.length; i++) {
+                                    var o = r[i];
+                                    if (o instanceof Element && n(o) && s.push(o), o instanceof DocumentFragment)
+                                        for (o = o.firstChild; o; o = o.nextSibling) a.push(o);
+                                    else a.push(o)
+                                }
+                                for (i = n(this), G.apply(this, r), r = 0; r < s.length; r++) c(e, s[r]);
+                                if (i)
+                                    for (c(e, this), r = 0; r < a.length; r++)(s = a[r]) instanceof Element && l(e, s)
+                            })), void 0 !== H && s(r, "remove", (function () {
+                                var t = n(this);
+                                H.call(this), t && c(e, this)
+                            }))
+                        }(i)
+                }(), document.__CE_hasRegistry = !0;
+            var ne = new y(te);
+            Object.defineProperty(window, "customElements", {
+                configurable: !0,
+                enumerable: !0,
+                value: ne
+            })
+        }
+    }).call(self), "string" != typeof document.baseURI && Object.defineProperty(Document.prototype, "baseURI", {
+            enumerable: !0,
+            configurable: !0,
+            get: function () {
+                var e = document.querySelector("base");
+                return e && e.href ? e.href : document.URL
+            }
+        }), "function" != typeof window.CustomEvent && (window.CustomEvent = function (e, t) {
+            t = t || {
+                bubbles: !1,
+                cancelable: !1,
+                detail: void 0
+            };
+            var n = document.createEvent("CustomEvent");
+            return n.initCustomEvent(e, t.bubbles, t.cancelable, t.detail), n
+        }, window.CustomEvent.prototype = window.Event.prototype), n = Event.prototype, r = document, a = window, n.composedPath || (n.composedPath = function () {
+            if (this.path) return this.path;
+            var e = this.target;
+            for (this.path = []; null !== e.parentNode;) this.path.push(e), e = e.parentNode;
+            return this.path.push(r, a), this.path
+        }), "function" != typeof (
+            /*!
+            Element.closest and Element.matches
+            https://github.com/jonathantneal/closest
+            Creative Commons Zero v1.0 Universal
+            */
+            s = window.Element.prototype).matches && (s.matches = s.msMatchesSelector || s.mozMatchesSelector || s.webkitMatchesSelector || function (e) {
+            e = (this.document || this.ownerDocument).querySelectorAll(e);
+            for (var t = 0; e[t] && e[t] !== this;) ++t;
+            return !!e[t]
+        }), "function" != typeof s.closest && (s.closest = function (e) {
+            for (var t = this; t && 1 === t.nodeType;) {
+                if (t.matches(e)) return t;
+                t = t.parentNode
+            }
+            return null
+        }),
+        /*!
+        Element.getRootNode()
+        */
+        function (e) {
+            function t(e) {
+                return e && e.parentNode ? t(e.parentNode) : e
+            }
+            "function" != typeof e.getRootNode && (e.getRootNode = function (e) {
+                return e && e.composed ? function e(n) {
+                    return (n = t(n)) && 11 === n.nodeType ? e(n.host) : n
+                }(this) : t(this)
+            })
+        }(Element.prototype),
+        /*!
+        Element.isConnected()
+        */
+        function (e) {
+            "isConnected" in e || Object.defineProperty(e, "isConnected", {
+                configurable: !0,
+                enumerable: !0,
+                get: function () {
+                    var e = this.getRootNode({
+                        composed: !0
+                    });
+                    return e && 9 === e.nodeType
+                }
+            })
+        }(Element.prototype), [Element.prototype, CharacterData.prototype, DocumentType.prototype].forEach((function (e) {
+            e.hasOwnProperty("remove") || Object.defineProperty(e, "remove", {
+                configurable: !0,
+                enumerable: !0,
+                writable: !0,
+                value: function () {
+                    null !== this.parentNode && this.parentNode.removeChild(this)
+                }
+            })
+        })), "classList" in (i = Element.prototype) || Object.defineProperty(i, "classList", {
+            get: function () {
+                var e = this,
+                    t = (e.getAttribute("class") || "").replace(/^\s+|\s$/g, "").split(/\s+/g);
+
+                function n() {
+                    t.length > 0 ? e.setAttribute("class", t.join(" ")) : e.removeAttribute("class")
+                }
+                return "" === t[0] && t.splice(0, 1), t.toggle = function (e, r) {
+                    void 0 !== r ? r ? t.add(e) : t.remove(e) : -1 !== t.indexOf(e) ? t.splice(t.indexOf(e), 1) : t.push(e), n()
+                }, t.add = function () {
+                    for (var e = [].slice.call(arguments), r = 0, a = e.length; r < a; r++) - 1 === t.indexOf(e[r]) && t.push(e[r]);
+                    n()
+                }, t.remove = function () {
+                    for (var e = [].slice.call(arguments), r = 0, a = e.length; r < a; r++) - 1 !== t.indexOf(e[r]) && t.splice(t.indexOf(e[r]), 1);
+                    n()
+                }, t.item = function (e) {
+                    return t[e]
+                }, t.contains = function (e) {
+                    return -1 !== t.indexOf(e)
+                }, t.replace = function (e, r) {
+                    -1 !== t.indexOf(e) && t.splice(t.indexOf(e), 1, r), n()
+                }, t.value = e.getAttribute("class") || "", t
+            }
+        }),
+        /*!
+        DOMTokenList
+        */
+        function (e) {
+            try {
+                document.body.classList.add()
+            } catch (r) {
+                var t = e.add,
+                    n = e.remove;
+                e.add = function () {
+                    for (var e = 0; e < arguments.length; e++) t.call(this, arguments[e])
+                }, e.remove = function () {
+                    for (var e = 0; e < arguments.length; e++) n.call(this, arguments[e])
+                }
+            }
+        }(DOMTokenList.prototype)
+}, function (e, t, n) {
+    (function (e) {
+        ! function (t) {
+            "use strict";
+            ! function (e) {
+                var t = {};
+
+                function n(r) {
+                    if (t[r]) return t[r].exports;
+                    var a = t[r] = {
+                        i: r,
+                        l: !1,
+                        exports: {}
+                    };
+                    return e[r].call(a.exports, a, a.exports, n), a.l = !0, a.exports
+                }
+                n.m = e, n.c = t, n.d = function (e, t, r) {
+                    n.o(e, t) || Object.defineProperty(e, t, {
+                        enumerable: !0,
+                        get: r
+                    })
+                }, n.r = function (e) {
+                    "undefined" != typeof Symbol && Symbol.toStringTag && Object.defineProperty(e, Symbol.toStringTag, {
+                        value: "Module"
+                    }), Object.defineProperty(e, "__esModule", {
+                        value: !0
+                    })
+                }, n.t = function (e, t) {
+                    if (1 & t && (e = n(e)), 8 & t) return e;
+                    if (4 & t && "object" == typeof e && e && e.__esModule) return e;
+                    var r = Object.create(null);
+                    if (n.r(r), Object.defineProperty(r, "default", {
+                            enumerable: !0,
+                            value: e
+                        }), 2 & t && "string" != typeof e)
+                        for (var a in e) n.d(r, a, function (t) {
+                            return e[t]
+                        }.bind(null, a));
+                    return r
+                }, n.n = function (e) {
+                    var t = e && e.__esModule ? function () {
+                        return e.default
+                    } : function () {
+                        return e
+                    };
+                    return n.d(t, "a", t), t
+                }, n.o = function (e, t) {
+                    return Object.prototype.hasOwnProperty.call(e, t)
+                }, n.p = "", n(n.s = 0)
+            }([function (e, t, n) {
+                n(1), n(55), n(62), n(68), n(70), n(71), n(72), n(73), n(75), n(76), n(78), n(87), n(88), n(89), n(98), n(99), n(101), n(102), n(103), n(105), n(106), n(107), n(108), n(110), n(111), n(112), n(113), n(114), n(115), n(116), n(117), n(118), n(127), n(130), n(131), n(133), n(135), n(136), n(137), n(138), n(139), n(141), n(143), n(146), n(148), n(150), n(151), n(153), n(154), n(155), n(156), n(157), n(159), n(160), n(162), n(163), n(164), n(165), n(166), n(167), n(168), n(169), n(170), n(172), n(173), n(183), n(184), n(185), n(189), n(191), n(192), n(193), n(194), n(195), n(196), n(198), n(201), n(202), n(203), n(204), n(208), n(209), n(212), n(213), n(214), n(215), n(216), n(217), n(218), n(219), n(221), n(222), n(223), n(226), n(227), n(228), n(229), n(230), n(231), n(232), n(233), n(234), n(235), n(236), n(237), n(238), n(240), n(241), n(243), n(248), e.exports = n(246)
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(45),
+                    i = n(14),
+                    o = n(46),
+                    u = n(39),
+                    l = n(47),
+                    c = n(48),
+                    h = n(52),
+                    p = n(49),
+                    d = n(53),
+                    f = p("isConcatSpreadable"),
+                    m = d >= 51 || !a((function () {
+                        var e = [];
+                        return e[f] = !1, e.concat()[0] !== e
+                    })),
+                    g = h("concat"),
+                    y = function (e) {
+                        if (!i(e)) return !1;
+                        var t = e[f];
+                        return void 0 !== t ? !!t : s(e)
+                    };
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !m || !g
+                }, {
+                    concat: function (e) {
+                        var t, n, r, a, s, i = o(this),
+                            h = c(i, 0),
+                            p = 0;
+                        for (t = -1, r = arguments.length; t < r; t++)
+                            if (y(s = -1 === t ? i : arguments[t])) {
+                                if (p + (a = u(s.length)) > 9007199254740991) throw TypeError("Maximum allowed index exceeded");
+                                for (n = 0; n < a; n++, p++) n in s && l(h, p, s[n])
+                            } else {
+                                if (p >= 9007199254740991) throw TypeError("Maximum allowed index exceeded");
+                                l(h, p++, s)
+                            } return h.length = p, h
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(4).f,
+                    s = n(18),
+                    i = n(21),
+                    o = n(22),
+                    u = n(32),
+                    l = n(44);
+                e.exports = function (e, t) {
+                    var n, c, h, p, d, f = e.target,
+                        m = e.global,
+                        g = e.stat;
+                    if (n = m ? r : g ? r[f] || o(f, {}) : (r[f] || {}).prototype)
+                        for (c in t) {
+                            if (p = t[c], h = e.noTargetGet ? (d = a(n, c)) && d.value : n[c], !l(m ? c : f + (g ? "." : "#") + c, e.forced) && void 0 !== h) {
+                                if (typeof p == typeof h) continue;
+                                u(p, h)
+                            }(e.sham || h && h.sham) && s(p, "sham", !0), i(n, c, p, e)
+                        }
+                }
+            }, function (t, n) {
+                var r = function (e) {
+                    return e && e.Math == Math && e
+                };
+                t.exports = r("object" == typeof globalThis && globalThis) || r("object" == typeof window && window) || r("object" == typeof self && self) || r("object" == typeof e && e) || Function("return this")()
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(7),
+                    s = n(8),
+                    i = n(9),
+                    o = n(13),
+                    u = n(15),
+                    l = n(16),
+                    c = Object.getOwnPropertyDescriptor;
+                t.f = r ? c : function (e, t) {
+                    if (e = i(e), t = o(t, !0), l) try {
+                        return c(e, t)
+                    } catch (e) {}
+                    if (u(e, t)) return s(!a.f.call(e, t), e[t])
+                }
+            }, function (e, t, n) {
+                var r = n(6);
+                e.exports = !r((function () {
+                    return 7 != Object.defineProperty({}, 1, {
+                        get: function () {
+                            return 7
+                        }
+                    })[1]
+                }))
+            }, function (e, t) {
+                e.exports = function (e) {
+                    try {
+                        return !!e()
+                    } catch (e) {
+                        return !0
+                    }
+                }
+            }, function (e, t, n) {
+                var r = {}.propertyIsEnumerable,
+                    a = Object.getOwnPropertyDescriptor,
+                    s = a && !r.call({
+                        1: 2
+                    }, 1);
+                t.f = s ? function (e) {
+                    var t = a(this, e);
+                    return !!t && t.enumerable
+                } : r
+            }, function (e, t) {
+                e.exports = function (e, t) {
+                    return {
+                        enumerable: !(1 & e),
+                        configurable: !(2 & e),
+                        writable: !(4 & e),
+                        value: t
+                    }
+                }
+            }, function (e, t, n) {
+                var r = n(10),
+                    a = n(12);
+                e.exports = function (e) {
+                    return r(a(e))
+                }
+            }, function (e, t, n) {
+                var r = n(6),
+                    a = n(11),
+                    s = "".split;
+                e.exports = r((function () {
+                    return !Object("z").propertyIsEnumerable(0)
+                })) ? function (e) {
+                    return "String" == a(e) ? s.call(e, "") : Object(e)
+                } : Object
+            }, function (e, t) {
+                var n = {}.toString;
+                e.exports = function (e) {
+                    return n.call(e).slice(8, -1)
+                }
+            }, function (e, t) {
+                e.exports = function (e) {
+                    if (null == e) throw TypeError("Can't call method on " + e);
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(14);
+                e.exports = function (e, t) {
+                    if (!r(e)) return e;
+                    var n, a;
+                    if (t && "function" == typeof (n = e.toString) && !r(a = n.call(e))) return a;
+                    if ("function" == typeof (n = e.valueOf) && !r(a = n.call(e))) return a;
+                    if (!t && "function" == typeof (n = e.toString) && !r(a = n.call(e))) return a;
+                    throw TypeError("Can't convert object to primitive value")
+                }
+            }, function (e, t) {
+                e.exports = function (e) {
+                    return "object" == typeof e ? null !== e : "function" == typeof e
+                }
+            }, function (e, t) {
+                var n = {}.hasOwnProperty;
+                e.exports = function (e, t) {
+                    return n.call(e, t)
+                }
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(6),
+                    s = n(17);
+                e.exports = !r && !a((function () {
+                    return 7 != Object.defineProperty(s("div"), "a", {
+                        get: function () {
+                            return 7
+                        }
+                    }).a
+                }))
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(14),
+                    s = r.document,
+                    i = a(s) && a(s.createElement);
+                e.exports = function (e) {
+                    return i ? s.createElement(e) : {}
+                }
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(19),
+                    s = n(8);
+                e.exports = r ? function (e, t, n) {
+                    return a.f(e, t, s(1, n))
+                } : function (e, t, n) {
+                    return e[t] = n, e
+                }
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(16),
+                    s = n(20),
+                    i = n(13),
+                    o = Object.defineProperty;
+                t.f = r ? o : function (e, t, n) {
+                    if (s(e), t = i(t, !0), s(n), a) try {
+                        return o(e, t, n)
+                    } catch (e) {}
+                    if ("get" in n || "set" in n) throw TypeError("Accessors not supported");
+                    return "value" in n && (e[t] = n.value), e
+                }
+            }, function (e, t, n) {
+                var r = n(14);
+                e.exports = function (e) {
+                    if (!r(e)) throw TypeError(String(e) + " is not an object");
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(18),
+                    s = n(15),
+                    i = n(22),
+                    o = n(23),
+                    u = n(25),
+                    l = u.get,
+                    c = u.enforce,
+                    h = String(String).split("String");
+                (e.exports = function (e, t, n, o) {
+                    var u = !!o && !!o.unsafe,
+                        l = !!o && !!o.enumerable,
+                        p = !!o && !!o.noTargetGet;
+                    "function" == typeof n && ("string" != typeof t || s(n, "name") || a(n, "name", t), c(n).source = h.join("string" == typeof t ? t : "")), e !== r ? (u ? !p && e[t] && (l = !0) : delete e[t], l ? e[t] = n : a(e, t, n)) : l ? e[t] = n : i(t, n)
+                })(Function.prototype, "toString", (function () {
+                    return "function" == typeof this && l(this).source || o(this)
+                }))
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(18);
+                e.exports = function (e, t) {
+                    try {
+                        a(r, e, t)
+                    } catch (n) {
+                        r[e] = t
+                    }
+                    return t
+                }
+            }, function (e, t, n) {
+                var r = n(24),
+                    a = Function.toString;
+                "function" != typeof r.inspectSource && (r.inspectSource = function (e) {
+                    return a.call(e)
+                }), e.exports = r.inspectSource
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(22),
+                    s = r["__core-js_shared__"] || a("__core-js_shared__", {});
+                e.exports = s
+            }, function (e, t, n) {
+                var r, a, s, i = n(26),
+                    o = n(3),
+                    u = n(14),
+                    l = n(18),
+                    c = n(15),
+                    h = n(27),
+                    p = n(31),
+                    d = o.WeakMap;
+                if (i) {
+                    var f = new d,
+                        m = f.get,
+                        g = f.has,
+                        y = f.set;
+                    r = function (e, t) {
+                        return y.call(f, e, t), t
+                    }, a = function (e) {
+                        return m.call(f, e) || {}
+                    }, s = function (e) {
+                        return g.call(f, e)
+                    }
+                } else {
+                    var b = h("state");
+                    p[b] = !0, r = function (e, t) {
+                        return l(e, b, t), t
+                    }, a = function (e) {
+                        return c(e, b) ? e[b] : {}
+                    }, s = function (e) {
+                        return c(e, b)
+                    }
+                }
+                e.exports = {
+                    set: r,
+                    get: a,
+                    has: s,
+                    enforce: function (e) {
+                        return s(e) ? a(e) : r(e, {})
+                    },
+                    getterFor: function (e) {
+                        return function (t) {
+                            var n;
+                            if (!u(t) || (n = a(t)).type !== e) throw TypeError("Incompatible receiver, " + e + " required");
+                            return n
+                        }
+                    }
+                }
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(23),
+                    s = r.WeakMap;
+                e.exports = "function" == typeof s && /native code/.test(a(s))
+            }, function (e, t, n) {
+                var r = n(28),
+                    a = n(30),
+                    s = r("keys");
+                e.exports = function (e) {
+                    return s[e] || (s[e] = a(e))
+                }
+            }, function (e, t, n) {
+                var r = n(29),
+                    a = n(24);
+                (e.exports = function (e, t) {
+                    return a[e] || (a[e] = void 0 !== t ? t : {})
+                })("versions", []).push({
+                    version: "3.6.5",
+                    mode: r ? "pure" : "global",
+                    copyright: "© 2020 Denis Pushkarev (zloirock.ru)"
+                })
+            }, function (e, t) {
+                e.exports = !1
+            }, function (e, t) {
+                var n = 0,
+                    r = Math.random();
+                e.exports = function (e) {
+                    return "Symbol(" + String(void 0 === e ? "" : e) + ")_" + (++n + r).toString(36)
+                }
+            }, function (e, t) {
+                e.exports = {}
+            }, function (e, t, n) {
+                var r = n(15),
+                    a = n(33),
+                    s = n(4),
+                    i = n(19);
+                e.exports = function (e, t) {
+                    for (var n = a(t), o = i.f, u = s.f, l = 0; l < n.length; l++) {
+                        var c = n[l];
+                        r(e, c) || o(e, c, u(t, c))
+                    }
+                }
+            }, function (e, t, n) {
+                var r = n(34),
+                    a = n(36),
+                    s = n(43),
+                    i = n(20);
+                e.exports = r("Reflect", "ownKeys") || function (e) {
+                    var t = a.f(i(e)),
+                        n = s.f;
+                    return n ? t.concat(n(e)) : t
+                }
+            }, function (e, t, n) {
+                var r = n(35),
+                    a = n(3),
+                    s = function (e) {
+                        return "function" == typeof e ? e : void 0
+                    };
+                e.exports = function (e, t) {
+                    return arguments.length < 2 ? s(r[e]) || s(a[e]) : r[e] && r[e][t] || a[e] && a[e][t]
+                }
+            }, function (e, t, n) {
+                var r = n(3);
+                e.exports = r
+            }, function (e, t, n) {
+                var r = n(37),
+                    a = n(42).concat("length", "prototype");
+                t.f = Object.getOwnPropertyNames || function (e) {
+                    return r(e, a)
+                }
+            }, function (e, t, n) {
+                var r = n(15),
+                    a = n(9),
+                    s = n(38).indexOf,
+                    i = n(31);
+                e.exports = function (e, t) {
+                    var n, o = a(e),
+                        u = 0,
+                        l = [];
+                    for (n in o) !r(i, n) && r(o, n) && l.push(n);
+                    for (; t.length > u;) r(o, n = t[u++]) && (~s(l, n) || l.push(n));
+                    return l
+                }
+            }, function (e, t, n) {
+                var r = n(9),
+                    a = n(39),
+                    s = n(41),
+                    i = function (e) {
+                        return function (t, n, i) {
+                            var o, u = r(t),
+                                l = a(u.length),
+                                c = s(i, l);
+                            if (e && n != n) {
+                                for (; l > c;)
+                                    if ((o = u[c++]) != o) return !0
+                            } else
+                                for (; l > c; c++)
+                                    if ((e || c in u) && u[c] === n) return e || c || 0;
+                            return !e && -1
+                        }
+                    };
+                e.exports = {
+                    includes: i(!0),
+                    indexOf: i(!1)
+                }
+            }, function (e, t, n) {
+                var r = n(40),
+                    a = Math.min;
+                e.exports = function (e) {
+                    return e > 0 ? a(r(e), 9007199254740991) : 0
+                }
+            }, function (e, t) {
+                var n = Math.ceil,
+                    r = Math.floor;
+                e.exports = function (e) {
+                    return isNaN(e = +e) ? 0 : (e > 0 ? r : n)(e)
+                }
+            }, function (e, t, n) {
+                var r = n(40),
+                    a = Math.max,
+                    s = Math.min;
+                e.exports = function (e, t) {
+                    var n = r(e);
+                    return n < 0 ? a(n + t, 0) : s(n, t)
+                }
+            }, function (e, t) {
+                e.exports = ["constructor", "hasOwnProperty", "isPrototypeOf", "propertyIsEnumerable", "toLocaleString", "toString", "valueOf"]
+            }, function (e, t) {
+                t.f = Object.getOwnPropertySymbols
+            }, function (e, t, n) {
+                var r = n(6),
+                    a = /#|\.prototype\./,
+                    s = function (e, t) {
+                        var n = o[i(e)];
+                        return n == l || n != u && ("function" == typeof t ? r(t) : !!t)
+                    },
+                    i = s.normalize = function (e) {
+                        return String(e).replace(a, ".").toLowerCase()
+                    },
+                    o = s.data = {},
+                    u = s.NATIVE = "N",
+                    l = s.POLYFILL = "P";
+                e.exports = s
+            }, function (e, t, n) {
+                var r = n(11);
+                e.exports = Array.isArray || function (e) {
+                    return "Array" == r(e)
+                }
+            }, function (e, t, n) {
+                var r = n(12);
+                e.exports = function (e) {
+                    return Object(r(e))
+                }
+            }, function (e, t, n) {
+                var r = n(13),
+                    a = n(19),
+                    s = n(8);
+                e.exports = function (e, t, n) {
+                    var i = r(t);
+                    i in e ? a.f(e, i, s(0, n)) : e[i] = n
+                }
+            }, function (e, t, n) {
+                var r = n(14),
+                    a = n(45),
+                    s = n(49)("species");
+                e.exports = function (e, t) {
+                    var n;
+                    return a(e) && ("function" != typeof (n = e.constructor) || n !== Array && !a(n.prototype) ? r(n) && null === (n = n[s]) && (n = void 0) : n = void 0), new(void 0 === n ? Array : n)(0 === t ? 0 : t)
+                }
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(28),
+                    s = n(15),
+                    i = n(30),
+                    o = n(50),
+                    u = n(51),
+                    l = a("wks"),
+                    c = r.Symbol,
+                    h = u ? c : c && c.withoutSetter || i;
+                e.exports = function (e) {
+                    return s(l, e) || (o && s(c, e) ? l[e] = c[e] : l[e] = h("Symbol." + e)), l[e]
+                }
+            }, function (e, t, n) {
+                var r = n(6);
+                e.exports = !!Object.getOwnPropertySymbols && !r((function () {
+                    return !String(Symbol())
+                }))
+            }, function (e, t, n) {
+                var r = n(50);
+                e.exports = r && !Symbol.sham && "symbol" == typeof Symbol.iterator
+            }, function (e, t, n) {
+                var r = n(6),
+                    a = n(49),
+                    s = n(53),
+                    i = a("species");
+                e.exports = function (e) {
+                    return s >= 51 || !r((function () {
+                        var t = [];
+                        return (t.constructor = {})[i] = function () {
+                            return {
+                                foo: 1
+                            }
+                        }, 1 !== t[e](Boolean).foo
+                    }))
+                }
+            }, function (e, t, n) {
+                var r, a, s = n(3),
+                    i = n(54),
+                    o = s.process,
+                    u = o && o.versions,
+                    l = u && u.v8;
+                l ? a = (r = l.split("."))[0] + r[1] : i && (!(r = i.match(/Edge\/(\d+)/)) || r[1] >= 74) && (r = i.match(/Chrome\/(\d+)/)) && (a = r[1]), e.exports = a && +a
+            }, function (e, t, n) {
+                var r = n(34);
+                e.exports = r("navigator", "userAgent") || ""
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(56),
+                    s = n(57);
+                r({
+                    target: "Array",
+                    proto: !0
+                }, {
+                    copyWithin: a
+                }), s("copyWithin")
+            }, function (e, t, n) {
+                var r = n(46),
+                    a = n(41),
+                    s = n(39),
+                    i = Math.min;
+                e.exports = [].copyWithin || function (e, t) {
+                    var n = r(this),
+                        o = s(n.length),
+                        u = a(e, o),
+                        l = a(t, o),
+                        c = arguments.length > 2 ? arguments[2] : void 0,
+                        h = i((void 0 === c ? o : a(c, o)) - l, o - u),
+                        p = 1;
+                    for (l < u && u < l + h && (p = -1, l += h - 1, u += h - 1); h-- > 0;) l in n ? n[u] = n[l] : delete n[u], u += p, l += p;
+                    return n
+                }
+            }, function (e, t, n) {
+                var r = n(49),
+                    a = n(58),
+                    s = n(19),
+                    i = r("unscopables"),
+                    o = Array.prototype;
+                null == o[i] && s.f(o, i, {
+                    configurable: !0,
+                    value: a(null)
+                }), e.exports = function (e) {
+                    o[i][e] = !0
+                }
+            }, function (e, t, n) {
+                var r, a = n(20),
+                    s = n(59),
+                    i = n(42),
+                    o = n(31),
+                    u = n(61),
+                    l = n(17),
+                    c = n(27)("IE_PROTO"),
+                    h = function () {},
+                    p = function (e) {
+                        return "<script>" + e + "<\/script>"
+                    },
+                    d = function () {
+                        try {
+                            r = document.domain && new ActiveXObject("htmlfile")
+                        } catch (e) {}
+                        var e, t;
+                        d = r ? function (e) {
+                            e.write(p("")), e.close();
+                            var t = e.parentWindow.Object;
+                            return e = null, t
+                        }(r) : ((t = l("iframe")).style.display = "none", u.appendChild(t), t.src = String("javascript:"), (e = t.contentWindow.document).open(), e.write(p("document.F=Object")), e.close(), e.F);
+                        for (var n = i.length; n--;) delete d.prototype[i[n]];
+                        return d()
+                    };
+                o[c] = !0, e.exports = Object.create || function (e, t) {
+                    var n;
+                    return null !== e ? (h.prototype = a(e), n = new h, h.prototype = null, n[c] = e) : n = d(), void 0 === t ? n : s(n, t)
+                }
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(19),
+                    s = n(20),
+                    i = n(60);
+                e.exports = r ? Object.defineProperties : function (e, t) {
+                    s(e);
+                    for (var n, r = i(t), o = r.length, u = 0; o > u;) a.f(e, n = r[u++], t[n]);
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(37),
+                    a = n(42);
+                e.exports = Object.keys || function (e) {
+                    return r(e, a)
+                }
+            }, function (e, t, n) {
+                var r = n(34);
+                e.exports = r("document", "documentElement")
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(63).every,
+                    s = n(66),
+                    i = n(67),
+                    o = s("every"),
+                    u = i("every");
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !o || !u
+                }, {
+                    every: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(64),
+                    a = n(10),
+                    s = n(46),
+                    i = n(39),
+                    o = n(48),
+                    u = [].push,
+                    l = function (e) {
+                        var t = 1 == e,
+                            n = 2 == e,
+                            l = 3 == e,
+                            c = 4 == e,
+                            h = 6 == e,
+                            p = 5 == e || h;
+                        return function (d, f, m, g) {
+                            for (var y, b, v = s(d), x = a(v), w = r(f, m, 3), k = i(x.length), S = 0, I = g || o, A = t ? I(d, k) : n ? I(d, 0) : void 0; k > S; S++)
+                                if ((p || S in x) && (b = w(y = x[S], S, v), e))
+                                    if (t) A[S] = b;
+                                    else if (b) switch (e) {
+                                case 3:
+                                    return !0;
+                                case 5:
+                                    return y;
+                                case 6:
+                                    return S;
+                                case 2:
+                                    u.call(A, y)
+                            } else if (c) return !1;
+                            return h ? -1 : l || c ? c : A
+                        }
+                    };
+                e.exports = {
+                    forEach: l(0),
+                    map: l(1),
+                    filter: l(2),
+                    some: l(3),
+                    every: l(4),
+                    find: l(5),
+                    findIndex: l(6)
+                }
+            }, function (e, t, n) {
+                var r = n(65);
+                e.exports = function (e, t, n) {
+                    if (r(e), void 0 === t) return e;
+                    switch (n) {
+                        case 0:
+                            return function () {
+                                return e.call(t)
+                            };
+                        case 1:
+                            return function (n) {
+                                return e.call(t, n)
+                            };
+                        case 2:
+                            return function (n, r) {
+                                return e.call(t, n, r)
+                            };
+                        case 3:
+                            return function (n, r, a) {
+                                return e.call(t, n, r, a)
+                            }
+                    }
+                    return function () {
+                        return e.apply(t, arguments)
+                    }
+                }
+            }, function (e, t) {
+                e.exports = function (e) {
+                    if ("function" != typeof e) throw TypeError(String(e) + " is not a function");
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(6);
+                e.exports = function (e, t) {
+                    var n = [][e];
+                    return !!n && r((function () {
+                        n.call(null, t || function () {
+                            throw 1
+                        }, 1)
+                    }))
+                }
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(6),
+                    s = n(15),
+                    i = Object.defineProperty,
+                    o = {},
+                    u = function (e) {
+                        throw e
+                    };
+                e.exports = function (e, t) {
+                    if (s(o, e)) return o[e];
+                    t || (t = {});
+                    var n = [][e],
+                        l = !!s(t, "ACCESSORS") && t.ACCESSORS,
+                        c = s(t, 0) ? t[0] : u,
+                        h = s(t, 1) ? t[1] : void 0;
+                    return o[e] = !!n && !a((function () {
+                        if (l && !r) return !0;
+                        var e = {
+                            length: -1
+                        };
+                        l ? i(e, 1, {
+                            enumerable: !0,
+                            get: u
+                        }) : e[1] = 1, n.call(e, c, h)
+                    }))
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(69),
+                    s = n(57);
+                r({
+                    target: "Array",
+                    proto: !0
+                }, {
+                    fill: a
+                }), s("fill")
+            }, function (e, t, n) {
+                var r = n(46),
+                    a = n(41),
+                    s = n(39);
+                e.exports = function (e) {
+                    for (var t = r(this), n = s(t.length), i = arguments.length, o = a(i > 1 ? arguments[1] : void 0, n), u = i > 2 ? arguments[2] : void 0, l = void 0 === u ? n : a(u, n); l > o;) t[o++] = e;
+                    return t
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(63).filter,
+                    s = n(52),
+                    i = n(67),
+                    o = s("filter"),
+                    u = i("filter");
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !o || !u
+                }, {
+                    filter: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(63).find,
+                    s = n(57),
+                    i = n(67),
+                    o = !0,
+                    u = i("find");
+                "find" in [] && Array(1).find((function () {
+                    o = !1
+                })), r({
+                    target: "Array",
+                    proto: !0,
+                    forced: o || !u
+                }, {
+                    find: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                }), s("find")
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(63).findIndex,
+                    s = n(57),
+                    i = n(67),
+                    o = !0,
+                    u = i("findIndex");
+                "findIndex" in [] && Array(1).findIndex((function () {
+                    o = !1
+                })), r({
+                    target: "Array",
+                    proto: !0,
+                    forced: o || !u
+                }, {
+                    findIndex: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                }), s("findIndex")
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(74),
+                    s = n(46),
+                    i = n(39),
+                    o = n(40),
+                    u = n(48);
+                r({
+                    target: "Array",
+                    proto: !0
+                }, {
+                    flat: function () {
+                        var e = arguments.length ? arguments[0] : void 0,
+                            t = s(this),
+                            n = i(t.length),
+                            r = u(t, 0);
+                        return r.length = a(r, t, t, n, 0, void 0 === e ? 1 : o(e)), r
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(45),
+                    a = n(39),
+                    s = n(64),
+                    i = function (e, t, n, o, u, l, c, h) {
+                        for (var p, d = u, f = 0, m = !!c && s(c, h, 3); f < o;) {
+                            if (f in n) {
+                                if (p = m ? m(n[f], f, t) : n[f], l > 0 && r(p)) d = i(e, t, p, a(p.length), d, l - 1) - 1;
+                                else {
+                                    if (d >= 9007199254740991) throw TypeError("Exceed the acceptable array length");
+                                    e[d] = p
+                                }
+                                d++
+                            }
+                            f++
+                        }
+                        return d
+                    };
+                e.exports = i
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(74),
+                    s = n(46),
+                    i = n(39),
+                    o = n(65),
+                    u = n(48);
+                r({
+                    target: "Array",
+                    proto: !0
+                }, {
+                    flatMap: function (e) {
+                        var t, n = s(this),
+                            r = i(n.length);
+                        return o(e), (t = u(n, 0)).length = a(t, n, n, r, 0, 1, e, arguments.length > 1 ? arguments[1] : void 0), t
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(77);
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: [].forEach != a
+                }, {
+                    forEach: a
+                })
+            }, function (e, t, n) {
+                var r = n(63).forEach,
+                    a = n(66),
+                    s = n(67),
+                    i = a("forEach"),
+                    o = s("forEach");
+                e.exports = i && o ? [].forEach : function (e) {
+                    return r(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(79);
+                r({
+                    target: "Array",
+                    stat: !0,
+                    forced: !n(86)((function (e) {
+                        Array.from(e)
+                    }))
+                }, {
+                    from: a
+                })
+            }, function (e, t, n) {
+                var r = n(64),
+                    a = n(46),
+                    s = n(80),
+                    i = n(81),
+                    o = n(39),
+                    u = n(47),
+                    l = n(83);
+                e.exports = function (e) {
+                    var t, n, c, h, p, d, f = a(e),
+                        m = "function" == typeof this ? this : Array,
+                        g = arguments.length,
+                        y = g > 1 ? arguments[1] : void 0,
+                        b = void 0 !== y,
+                        v = l(f),
+                        x = 0;
+                    if (b && (y = r(y, g > 2 ? arguments[2] : void 0, 2)), null == v || m == Array && i(v))
+                        for (n = new m(t = o(f.length)); t > x; x++) d = b ? y(f[x], x) : f[x], u(n, x, d);
+                    else
+                        for (p = (h = v.call(f)).next, n = new m; !(c = p.call(h)).done; x++) d = b ? s(h, y, [c.value, x], !0) : c.value, u(n, x, d);
+                    return n.length = x, n
+                }
+            }, function (e, t, n) {
+                var r = n(20);
+                e.exports = function (e, t, n, a) {
+                    try {
+                        return a ? t(r(n)[0], n[1]) : t(n)
+                    } catch (t) {
+                        var s = e.return;
+                        throw void 0 !== s && r(s.call(e)), t
+                    }
+                }
+            }, function (e, t, n) {
+                var r = n(49),
+                    a = n(82),
+                    s = r("iterator"),
+                    i = Array.prototype;
+                e.exports = function (e) {
+                    return void 0 !== e && (a.Array === e || i[s] === e)
+                }
+            }, function (e, t) {
+                e.exports = {}
+            }, function (e, t, n) {
+                var r = n(84),
+                    a = n(82),
+                    s = n(49)("iterator");
+                e.exports = function (e) {
+                    if (null != e) return e[s] || e["@@iterator"] || a[r(e)]
+                }
+            }, function (e, t, n) {
+                var r = n(85),
+                    a = n(11),
+                    s = n(49)("toStringTag"),
+                    i = "Arguments" == a(function () {
+                        return arguments
+                    }());
+                e.exports = r ? a : function (e) {
+                    var t, n, r;
+                    return void 0 === e ? "Undefined" : null === e ? "Null" : "string" == typeof (n = function (e, t) {
+                        try {
+                            return e[t]
+                        } catch (e) {}
+                    }(t = Object(e), s)) ? n : i ? a(t) : "Object" == (r = a(t)) && "function" == typeof t.callee ? "Arguments" : r
+                }
+            }, function (e, t, n) {
+                var r = {};
+                r[n(49)("toStringTag")] = "z", e.exports = "[object z]" === String(r)
+            }, function (e, t, n) {
+                var r = n(49)("iterator"),
+                    a = !1;
+                try {
+                    var s = 0,
+                        i = {
+                            next: function () {
+                                return {
+                                    done: !!s++
+                                }
+                            },
+                            return: function () {
+                                a = !0
+                            }
+                        };
+                    i[r] = function () {
+                        return this
+                    }, Array.from(i, (function () {
+                        throw 2
+                    }))
+                } catch (e) {}
+                e.exports = function (e, t) {
+                    if (!t && !a) return !1;
+                    var n = !1;
+                    try {
+                        var s = {};
+                        s[r] = function () {
+                            return {
+                                next: function () {
+                                    return {
+                                        done: n = !0
+                                    }
+                                }
+                            }
+                        }, e(s)
+                    } catch (e) {}
+                    return n
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(38).includes,
+                    s = n(57);
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !n(67)("indexOf", {
+                        ACCESSORS: !0,
+                        1: 0
+                    })
+                }, {
+                    includes: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                }), s("includes")
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(38).indexOf,
+                    s = n(66),
+                    i = n(67),
+                    o = [].indexOf,
+                    u = !!o && 1 / [1].indexOf(1, -0) < 0,
+                    l = s("indexOf"),
+                    c = i("indexOf", {
+                        ACCESSORS: !0,
+                        1: 0
+                    });
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: u || !l || !c
+                }, {
+                    indexOf: function (e) {
+                        return u ? o.apply(this, arguments) || 0 : a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(9),
+                    a = n(57),
+                    s = n(82),
+                    i = n(25),
+                    o = n(90),
+                    u = i.set,
+                    l = i.getterFor("Array Iterator");
+                e.exports = o(Array, "Array", (function (e, t) {
+                    u(this, {
+                        type: "Array Iterator",
+                        target: r(e),
+                        index: 0,
+                        kind: t
+                    })
+                }), (function () {
+                    var e = l(this),
+                        t = e.target,
+                        n = e.kind,
+                        r = e.index++;
+                    return !t || r >= t.length ? (e.target = void 0, {
+                        value: void 0,
+                        done: !0
+                    }) : "keys" == n ? {
+                        value: r,
+                        done: !1
+                    } : "values" == n ? {
+                        value: t[r],
+                        done: !1
+                    } : {
+                        value: [r, t[r]],
+                        done: !1
+                    }
+                }), "values"), s.Arguments = s.Array, a("keys"), a("values"), a("entries")
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(91),
+                    s = n(93),
+                    i = n(96),
+                    o = n(95),
+                    u = n(18),
+                    l = n(21),
+                    c = n(49),
+                    h = n(29),
+                    p = n(82),
+                    d = n(92),
+                    f = d.IteratorPrototype,
+                    m = d.BUGGY_SAFARI_ITERATORS,
+                    g = c("iterator"),
+                    y = function () {
+                        return this
+                    };
+                e.exports = function (e, t, n, c, d, b, v) {
+                    a(n, t, c);
+                    var x, w, k, S = function (e) {
+                            if (e === d && C) return C;
+                            if (!m && e in E) return E[e];
+                            switch (e) {
+                                case "keys":
+                                case "values":
+                                case "entries":
+                                    return function () {
+                                        return new n(this, e)
+                                    }
+                            }
+                            return function () {
+                                return new n(this)
+                            }
+                        },
+                        I = t + " Iterator",
+                        A = !1,
+                        E = e.prototype,
+                        N = E[g] || E["@@iterator"] || d && E[d],
+                        C = !m && N || S(d),
+                        T = "Array" == t && E.entries || N;
+                    if (T && (x = s(T.call(new e)), f !== Object.prototype && x.next && (h || s(x) === f || (i ? i(x, f) : "function" != typeof x[g] && u(x, g, y)), o(x, I, !0, !0), h && (p[I] = y))), "values" == d && N && "values" !== N.name && (A = !0, C = function () {
+                            return N.call(this)
+                        }), h && !v || E[g] === C || u(E, g, C), p[t] = C, d)
+                        if (w = {
+                                values: S("values"),
+                                keys: b ? C : S("keys"),
+                                entries: S("entries")
+                            }, v)
+                            for (k in w)(m || A || !(k in E)) && l(E, k, w[k]);
+                        else r({
+                            target: t,
+                            proto: !0,
+                            forced: m || A
+                        }, w);
+                    return w
+                }
+            }, function (e, t, n) {
+                var r = n(92).IteratorPrototype,
+                    a = n(58),
+                    s = n(8),
+                    i = n(95),
+                    o = n(82),
+                    u = function () {
+                        return this
+                    };
+                e.exports = function (e, t, n) {
+                    var l = t + " Iterator";
+                    return e.prototype = a(r, {
+                        next: s(1, n)
+                    }), i(e, l, !1, !0), o[l] = u, e
+                }
+            }, function (e, t, n) {
+                var r, a, s, i = n(93),
+                    o = n(18),
+                    u = n(15),
+                    l = n(49),
+                    c = n(29),
+                    h = l("iterator"),
+                    p = !1;
+                [].keys && ("next" in (s = [].keys()) ? (a = i(i(s))) !== Object.prototype && (r = a) : p = !0), null == r && (r = {}), c || u(r, h) || o(r, h, (function () {
+                    return this
+                })), e.exports = {
+                    IteratorPrototype: r,
+                    BUGGY_SAFARI_ITERATORS: p
+                }
+            }, function (e, t, n) {
+                var r = n(15),
+                    a = n(46),
+                    s = n(27),
+                    i = n(94),
+                    o = s("IE_PROTO"),
+                    u = Object.prototype;
+                e.exports = i ? Object.getPrototypeOf : function (e) {
+                    return e = a(e), r(e, o) ? e[o] : "function" == typeof e.constructor && e instanceof e.constructor ? e.constructor.prototype : e instanceof Object ? u : null
+                }
+            }, function (e, t, n) {
+                var r = n(6);
+                e.exports = !r((function () {
+                    function e() {}
+                    return e.prototype.constructor = null, Object.getPrototypeOf(new e) !== e.prototype
+                }))
+            }, function (e, t, n) {
+                var r = n(19).f,
+                    a = n(15),
+                    s = n(49)("toStringTag");
+                e.exports = function (e, t, n) {
+                    e && !a(e = n ? e : e.prototype, s) && r(e, s, {
+                        configurable: !0,
+                        value: t
+                    })
+                }
+            }, function (e, t, n) {
+                var r = n(20),
+                    a = n(97);
+                e.exports = Object.setPrototypeOf || ("__proto__" in {} ? function () {
+                    var e, t = !1,
+                        n = {};
+                    try {
+                        (e = Object.getOwnPropertyDescriptor(Object.prototype, "__proto__").set).call(n, []), t = n instanceof Array
+                    } catch (e) {}
+                    return function (n, s) {
+                        return r(n), a(s), t ? e.call(n, s) : n.__proto__ = s, n
+                    }
+                }() : void 0)
+            }, function (e, t, n) {
+                var r = n(14);
+                e.exports = function (e) {
+                    if (!r(e) && null !== e) throw TypeError("Can't set " + String(e) + " as a prototype");
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(10),
+                    s = n(9),
+                    i = n(66),
+                    o = [].join,
+                    u = a != Object,
+                    l = i("join", ",");
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: u || !l
+                }, {
+                    join: function (e) {
+                        return o.call(s(this), void 0 === e ? "," : e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(100);
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: a !== [].lastIndexOf
+                }, {
+                    lastIndexOf: a
+                })
+            }, function (e, t, n) {
+                var r = n(9),
+                    a = n(40),
+                    s = n(39),
+                    i = n(66),
+                    o = n(67),
+                    u = Math.min,
+                    l = [].lastIndexOf,
+                    c = !!l && 1 / [1].lastIndexOf(1, -0) < 0,
+                    h = i("lastIndexOf"),
+                    p = o("indexOf", {
+                        ACCESSORS: !0,
+                        1: 0
+                    }),
+                    d = c || !h || !p;
+                e.exports = d ? function (e) {
+                    if (c) return l.apply(this, arguments) || 0;
+                    var t = r(this),
+                        n = s(t.length),
+                        i = n - 1;
+                    for (arguments.length > 1 && (i = u(i, a(arguments[1]))), i < 0 && (i = n + i); i >= 0; i--)
+                        if (i in t && t[i] === e) return i || 0;
+                    return -1
+                } : l
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(63).map,
+                    s = n(52),
+                    i = n(67),
+                    o = s("map"),
+                    u = i("map");
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !o || !u
+                }, {
+                    map: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(47);
+                r({
+                    target: "Array",
+                    stat: !0,
+                    forced: a((function () {
+                        function e() {}
+                        return !(Array.of.call(e) instanceof e)
+                    }))
+                }, {
+                    of: function () {
+                        for (var e = 0, t = arguments.length, n = new("function" == typeof this ? this : Array)(t); t > e;) s(n, e, arguments[e++]);
+                        return n.length = t, n
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(104).left,
+                    s = n(66),
+                    i = n(67),
+                    o = s("reduce"),
+                    u = i("reduce", {
+                        1: 0
+                    });
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !o || !u
+                }, {
+                    reduce: function (e) {
+                        return a(this, e, arguments.length, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(65),
+                    a = n(46),
+                    s = n(10),
+                    i = n(39),
+                    o = function (e) {
+                        return function (t, n, o, u) {
+                            r(n);
+                            var l = a(t),
+                                c = s(l),
+                                h = i(l.length),
+                                p = e ? h - 1 : 0,
+                                d = e ? -1 : 1;
+                            if (o < 2)
+                                for (;;) {
+                                    if (p in c) {
+                                        u = c[p], p += d;
+                                        break
+                                    }
+                                    if (p += d, e ? p < 0 : h <= p) throw TypeError("Reduce of empty array with no initial value")
+                                }
+                            for (; e ? p >= 0 : h > p; p += d) p in c && (u = n(u, c[p], p, l));
+                            return u
+                        }
+                    };
+                e.exports = {
+                    left: o(!1),
+                    right: o(!0)
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(104).right,
+                    s = n(66),
+                    i = n(67),
+                    o = s("reduceRight"),
+                    u = i("reduce", {
+                        1: 0
+                    });
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !o || !u
+                }, {
+                    reduceRight: function (e) {
+                        return a(this, e, arguments.length, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(14),
+                    s = n(45),
+                    i = n(41),
+                    o = n(39),
+                    u = n(9),
+                    l = n(47),
+                    c = n(49),
+                    h = n(52),
+                    p = n(67),
+                    d = h("slice"),
+                    f = p("slice", {
+                        ACCESSORS: !0,
+                        0: 0,
+                        1: 2
+                    }),
+                    m = c("species"),
+                    g = [].slice,
+                    y = Math.max;
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !d || !f
+                }, {
+                    slice: function (e, t) {
+                        var n, r, c, h = u(this),
+                            p = o(h.length),
+                            d = i(e, p),
+                            f = i(void 0 === t ? p : t, p);
+                        if (s(h) && ("function" != typeof (n = h.constructor) || n !== Array && !s(n.prototype) ? a(n) && null === (n = n[m]) && (n = void 0) : n = void 0, n === Array || void 0 === n)) return g.call(h, d, f);
+                        for (r = new(void 0 === n ? Array : n)(y(f - d, 0)), c = 0; d < f; d++, c++) d in h && l(r, c, h[d]);
+                        return r.length = c, r
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(63).some,
+                    s = n(66),
+                    i = n(67),
+                    o = s("some"),
+                    u = i("some");
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !o || !u
+                }, {
+                    some: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                n(109)("Array")
+            }, function (e, t, n) {
+                var r = n(34),
+                    a = n(19),
+                    s = n(49),
+                    i = n(5),
+                    o = s("species");
+                e.exports = function (e) {
+                    var t = r(e),
+                        n = a.f;
+                    i && t && !t[o] && n(t, o, {
+                        configurable: !0,
+                        get: function () {
+                            return this
+                        }
+                    })
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(41),
+                    s = n(40),
+                    i = n(39),
+                    o = n(46),
+                    u = n(48),
+                    l = n(47),
+                    c = n(52),
+                    h = n(67),
+                    p = c("splice"),
+                    d = h("splice", {
+                        ACCESSORS: !0,
+                        0: 0,
+                        1: 2
+                    }),
+                    f = Math.max,
+                    m = Math.min;
+                r({
+                    target: "Array",
+                    proto: !0,
+                    forced: !p || !d
+                }, {
+                    splice: function (e, t) {
+                        var n, r, c, h, p, d, g = o(this),
+                            y = i(g.length),
+                            b = a(e, y),
+                            v = arguments.length;
+                        if (0 === v ? n = r = 0 : 1 === v ? (n = 0, r = y - b) : (n = v - 2, r = m(f(s(t), 0), y - b)), y + n - r > 9007199254740991) throw TypeError("Maximum allowed length exceeded");
+                        for (c = u(g, r), h = 0; h < r; h++)(p = b + h) in g && l(c, h, g[p]);
+                        if (c.length = r, n < r) {
+                            for (h = b; h < y - r; h++) d = h + n, (p = h + r) in g ? g[d] = g[p] : delete g[d];
+                            for (h = y; h > y - r + n; h--) delete g[h - 1]
+                        } else if (n > r)
+                            for (h = y - r; h > b; h--) d = h + n - 1, (p = h + r - 1) in g ? g[d] = g[p] : delete g[d];
+                        for (h = 0; h < n; h++) g[h + b] = arguments[h + 2];
+                        return g.length = y - r + n, c
+                    }
+                })
+            }, function (e, t, n) {
+                n(57)("flat")
+            }, function (e, t, n) {
+                n(57)("flatMap")
+            }, function (e, t, n) {
+                var r = n(14),
+                    a = n(19),
+                    s = n(93),
+                    i = n(49)("hasInstance"),
+                    o = Function.prototype;
+                i in o || a.f(o, i, {
+                    value: function (e) {
+                        if ("function" != typeof this || !r(e)) return !1;
+                        if (!r(this.prototype)) return e instanceof this;
+                        for (; e = s(e);)
+                            if (this.prototype === e) return !0;
+                        return !1
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(19).f,
+                    s = Function.prototype,
+                    i = s.toString,
+                    o = /^\s*function ([^ (]*)/;
+                r && !("name" in s) && a(s, "name", {
+                    configurable: !0,
+                    get: function () {
+                        try {
+                            return i.call(this).match(o)[1]
+                        } catch (e) {
+                            return ""
+                        }
+                    }
+                })
+            }, function (e, t, n) {
+                n(2)({
+                    global: !0
+                }, {
+                    globalThis: n(3)
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(34),
+                    s = n(6),
+                    i = a("JSON", "stringify"),
+                    o = /[\uD800-\uDFFF]/g,
+                    u = /^[\uD800-\uDBFF]$/,
+                    l = /^[\uDC00-\uDFFF]$/,
+                    c = function (e, t, n) {
+                        var r = n.charAt(t - 1),
+                            a = n.charAt(t + 1);
+                        return u.test(e) && !l.test(a) || l.test(e) && !u.test(r) ? "\\u" + e.charCodeAt(0).toString(16) : e
+                    },
+                    h = s((function () {
+                        return '"\\udf06\\ud834"' !== i("\udf06\ud834") || '"\\udead"' !== i("\udead")
+                    }));
+                i && r({
+                    target: "JSON",
+                    stat: !0,
+                    forced: h
+                }, {
+                    stringify: function (e, t, n) {
+                        var r = i.apply(null, arguments);
+                        return "string" == typeof r ? r.replace(o, c) : r
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(3);
+                n(95)(r.JSON, "JSON", !0)
+            }, function (e, t, n) {
+                var r = n(119),
+                    a = n(125);
+                e.exports = r("Map", (function (e) {
+                    return function () {
+                        return e(this, arguments.length ? arguments[0] : void 0)
+                    }
+                }), a)
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(3),
+                    s = n(44),
+                    i = n(21),
+                    o = n(120),
+                    u = n(122),
+                    l = n(123),
+                    c = n(14),
+                    h = n(6),
+                    p = n(86),
+                    d = n(95),
+                    f = n(124);
+                e.exports = function (e, t, n) {
+                    var m = -1 !== e.indexOf("Map"),
+                        g = -1 !== e.indexOf("Weak"),
+                        y = m ? "set" : "add",
+                        b = a[e],
+                        v = b && b.prototype,
+                        x = b,
+                        w = {},
+                        k = function (e) {
+                            var t = v[e];
+                            i(v, e, "add" == e ? function (e) {
+                                return t.call(this, 0 === e ? 0 : e), this
+                            } : "delete" == e ? function (e) {
+                                return !(g && !c(e)) && t.call(this, 0 === e ? 0 : e)
+                            } : "get" == e ? function (e) {
+                                return g && !c(e) ? void 0 : t.call(this, 0 === e ? 0 : e)
+                            } : "has" == e ? function (e) {
+                                return !(g && !c(e)) && t.call(this, 0 === e ? 0 : e)
+                            } : function (e, n) {
+                                return t.call(this, 0 === e ? 0 : e, n), this
+                            })
+                        };
+                    if (s(e, "function" != typeof b || !(g || v.forEach && !h((function () {
+                            (new b).entries().next()
+                        }))))) x = n.getConstructor(t, e, m, y), o.REQUIRED = !0;
+                    else if (s(e, !0)) {
+                        var S = new x,
+                            I = S[y](g ? {} : -0, 1) != S,
+                            A = h((function () {
+                                S.has(1)
+                            })),
+                            E = p((function (e) {
+                                new b(e)
+                            })),
+                            N = !g && h((function () {
+                                for (var e = new b, t = 5; t--;) e[y](t, t);
+                                return !e.has(-0)
+                            }));
+                        E || ((x = t((function (t, n) {
+                            l(t, x, e);
+                            var r = f(new b, t, x);
+                            return null != n && u(n, r[y], r, m), r
+                        }))).prototype = v, v.constructor = x), (A || N) && (k("delete"), k("has"), m && k("get")), (N || I) && k(y), g && v.clear && delete v.clear
+                    }
+                    return w[e] = x, r({
+                        global: !0,
+                        forced: x != b
+                    }, w), d(x, e), g || n.setStrong(x, e, m), x
+                }
+            }, function (e, t, n) {
+                var r = n(31),
+                    a = n(14),
+                    s = n(15),
+                    i = n(19).f,
+                    o = n(30),
+                    u = n(121),
+                    l = o("meta"),
+                    c = 0,
+                    h = Object.isExtensible || function () {
+                        return !0
+                    },
+                    p = function (e) {
+                        i(e, l, {
+                            value: {
+                                objectID: "O" + ++c,
+                                weakData: {}
+                            }
+                        })
+                    },
+                    d = e.exports = {
+                        REQUIRED: !1,
+                        fastKey: function (e, t) {
+                            if (!a(e)) return "symbol" == typeof e ? e : ("string" == typeof e ? "S" : "P") + e;
+                            if (!s(e, l)) {
+                                if (!h(e)) return "F";
+                                if (!t) return "E";
+                                p(e)
+                            }
+                            return e[l].objectID
+                        },
+                        getWeakData: function (e, t) {
+                            if (!s(e, l)) {
+                                if (!h(e)) return !0;
+                                if (!t) return !1;
+                                p(e)
+                            }
+                            return e[l].weakData
+                        },
+                        onFreeze: function (e) {
+                            return u && d.REQUIRED && h(e) && !s(e, l) && p(e), e
+                        }
+                    };
+                r[l] = !0
+            }, function (e, t, n) {
+                var r = n(6);
+                e.exports = !r((function () {
+                    return Object.isExtensible(Object.preventExtensions({}))
+                }))
+            }, function (e, t, n) {
+                var r = n(20),
+                    a = n(81),
+                    s = n(39),
+                    i = n(64),
+                    o = n(83),
+                    u = n(80),
+                    l = function (e, t) {
+                        this.stopped = e, this.result = t
+                    };
+                (e.exports = function (e, t, n, c, h) {
+                    var p, d, f, m, g, y, b, v = i(t, n, c ? 2 : 1);
+                    if (h) p = e;
+                    else {
+                        if ("function" != typeof (d = o(e))) throw TypeError("Target is not iterable");
+                        if (a(d)) {
+                            for (f = 0, m = s(e.length); m > f; f++)
+                                if ((g = c ? v(r(b = e[f])[0], b[1]) : v(e[f])) && g instanceof l) return g;
+                            return new l(!1)
+                        }
+                        p = d.call(e)
+                    }
+                    for (y = p.next; !(b = y.call(p)).done;)
+                        if ("object" == typeof (g = u(p, v, b.value, c)) && g && g instanceof l) return g;
+                    return new l(!1)
+                }).stop = function (e) {
+                    return new l(!0, e)
+                }
+            }, function (e, t) {
+                e.exports = function (e, t, n) {
+                    if (!(e instanceof t)) throw TypeError("Incorrect " + (n ? n + " " : "") + "invocation");
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(14),
+                    a = n(96);
+                e.exports = function (e, t, n) {
+                    var s, i;
+                    return a && "function" == typeof (s = t.constructor) && s !== n && r(i = s.prototype) && i !== n.prototype && a(e, i), e
+                }
+            }, function (e, t, n) {
+                var r = n(19).f,
+                    a = n(58),
+                    s = n(126),
+                    i = n(64),
+                    o = n(123),
+                    u = n(122),
+                    l = n(90),
+                    c = n(109),
+                    h = n(5),
+                    p = n(120).fastKey,
+                    d = n(25),
+                    f = d.set,
+                    m = d.getterFor;
+                e.exports = {
+                    getConstructor: function (e, t, n, l) {
+                        var c = e((function (e, r) {
+                                o(e, c, t), f(e, {
+                                    type: t,
+                                    index: a(null),
+                                    first: void 0,
+                                    last: void 0,
+                                    size: 0
+                                }), h || (e.size = 0), null != r && u(r, e[l], e, n)
+                            })),
+                            d = m(t),
+                            g = function (e, t, n) {
+                                var r, a, s = d(e),
+                                    i = y(e, t);
+                                return i ? i.value = n : (s.last = i = {
+                                    index: a = p(t, !0),
+                                    key: t,
+                                    value: n,
+                                    previous: r = s.last,
+                                    next: void 0,
+                                    removed: !1
+                                }, s.first || (s.first = i), r && (r.next = i), h ? s.size++ : e.size++, "F" !== a && (s.index[a] = i)), e
+                            },
+                            y = function (e, t) {
+                                var n, r = d(e),
+                                    a = p(t);
+                                if ("F" !== a) return r.index[a];
+                                for (n = r.first; n; n = n.next)
+                                    if (n.key == t) return n
+                            };
+                        return s(c.prototype, {
+                            clear: function () {
+                                for (var e = d(this), t = e.index, n = e.first; n;) n.removed = !0, n.previous && (n.previous = n.previous.next = void 0), delete t[n.index], n = n.next;
+                                e.first = e.last = void 0, h ? e.size = 0 : this.size = 0
+                            },
+                            delete: function (e) {
+                                var t = d(this),
+                                    n = y(this, e);
+                                if (n) {
+                                    var r = n.next,
+                                        a = n.previous;
+                                    delete t.index[n.index], n.removed = !0, a && (a.next = r), r && (r.previous = a), t.first == n && (t.first = r), t.last == n && (t.last = a), h ? t.size-- : this.size--
+                                }
+                                return !!n
+                            },
+                            forEach: function (e) {
+                                for (var t, n = d(this), r = i(e, arguments.length > 1 ? arguments[1] : void 0, 3); t = t ? t.next : n.first;)
+                                    for (r(t.value, t.key, this); t && t.removed;) t = t.previous
+                            },
+                            has: function (e) {
+                                return !!y(this, e)
+                            }
+                        }), s(c.prototype, n ? {
+                            get: function (e) {
+                                var t = y(this, e);
+                                return t && t.value
+                            },
+                            set: function (e, t) {
+                                return g(this, 0 === e ? 0 : e, t)
+                            }
+                        } : {
+                            add: function (e) {
+                                return g(this, e = 0 === e ? 0 : e, e)
+                            }
+                        }), h && r(c.prototype, "size", {
+                            get: function () {
+                                return d(this).size
+                            }
+                        }), c
+                    },
+                    setStrong: function (e, t, n) {
+                        var r = t + " Iterator",
+                            a = m(t),
+                            s = m(r);
+                        l(e, t, (function (e, t) {
+                            f(this, {
+                                type: r,
+                                target: e,
+                                state: a(e),
+                                kind: t,
+                                last: void 0
+                            })
+                        }), (function () {
+                            for (var e = s(this), t = e.kind, n = e.last; n && n.removed;) n = n.previous;
+                            return e.target && (e.last = n = n ? n.next : e.state.first) ? "keys" == t ? {
+                                value: n.key,
+                                done: !1
+                            } : "values" == t ? {
+                                value: n.value,
+                                done: !1
+                            } : {
+                                value: [n.key, n.value],
+                                done: !1
+                            } : (e.target = void 0, {
+                                value: void 0,
+                                done: !0
+                            })
+                        }), n ? "entries" : "values", !n, !0), c(t)
+                    }
+                }
+            }, function (e, t, n) {
+                var r = n(21);
+                e.exports = function (e, t, n) {
+                    for (var a in t) r(e, a, t[a], n);
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(3),
+                    s = n(44),
+                    i = n(21),
+                    o = n(15),
+                    u = n(11),
+                    l = n(124),
+                    c = n(13),
+                    h = n(6),
+                    p = n(58),
+                    d = n(36).f,
+                    f = n(4).f,
+                    m = n(19).f,
+                    g = n(128).trim,
+                    y = a.Number,
+                    b = y.prototype,
+                    v = "Number" == u(p(b)),
+                    x = function (e) {
+                        var t, n, r, a, s, i, o, u, l = c(e, !1);
+                        if ("string" == typeof l && l.length > 2)
+                            if (43 === (t = (l = g(l)).charCodeAt(0)) || 45 === t) {
+                                if (88 === (n = l.charCodeAt(2)) || 120 === n) return NaN
+                            } else if (48 === t) {
+                            switch (l.charCodeAt(1)) {
+                                case 66:
+                                case 98:
+                                    r = 2, a = 49;
+                                    break;
+                                case 79:
+                                case 111:
+                                    r = 8, a = 55;
+                                    break;
+                                default:
+                                    return +l
+                            }
+                            for (i = (s = l.slice(2)).length, o = 0; o < i; o++)
+                                if ((u = s.charCodeAt(o)) < 48 || u > a) return NaN;
+                            return parseInt(s, r)
+                        }
+                        return +l
+                    };
+                if (s("Number", !y(" 0o1") || !y("0b1") || y("+0x1"))) {
+                    for (var w, k = function (e) {
+                            var t = arguments.length < 1 ? 0 : e,
+                                n = this;
+                            return n instanceof k && (v ? h((function () {
+                                b.valueOf.call(n)
+                            })) : "Number" != u(n)) ? l(new y(x(t)), n, k) : x(t)
+                        }, S = r ? d(y) : "MAX_VALUE,MIN_VALUE,NaN,NEGATIVE_INFINITY,POSITIVE_INFINITY,EPSILON,isFinite,isInteger,isNaN,isSafeInteger,MAX_SAFE_INTEGER,MIN_SAFE_INTEGER,parseFloat,parseInt,isInteger".split(","), I = 0; S.length > I; I++) o(y, w = S[I]) && !o(k, w) && m(k, w, f(y, w));
+                    k.prototype = b, b.constructor = k, i(a, "Number", k)
+                }
+            }, function (e, t, n) {
+                var r = n(12),
+                    a = "[" + n(129) + "]",
+                    s = RegExp("^" + a + a + "*"),
+                    i = RegExp(a + a + "*$"),
+                    o = function (e) {
+                        return function (t) {
+                            var n = String(r(t));
+                            return 1 & e && (n = n.replace(s, "")), 2 & e && (n = n.replace(i, "")), n
+                        }
+                    };
+                e.exports = {
+                    start: o(1),
+                    end: o(2),
+                    trim: o(3)
+                }
+            }, function (e, t) {
+                e.exports = "\t\n\v\f\r                　\u2028\u2029\ufeff"
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    EPSILON: Math.pow(2, -52)
+                })
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    isFinite: n(132)
+                })
+            }, function (e, t, n) {
+                var r = n(3).isFinite;
+                e.exports = Number.isFinite || function (e) {
+                    return "number" == typeof e && r(e)
+                }
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    isInteger: n(134)
+                })
+            }, function (e, t, n) {
+                var r = n(14),
+                    a = Math.floor;
+                e.exports = function (e) {
+                    return !r(e) && isFinite(e) && a(e) === e
+                }
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    isNaN: function (e) {
+                        return e != e
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(134),
+                    s = Math.abs;
+                r({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    isSafeInteger: function (e) {
+                        return a(e) && s(e) <= 9007199254740991
+                    }
+                })
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    MAX_SAFE_INTEGER: 9007199254740991
+                })
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Number",
+                    stat: !0
+                }, {
+                    MIN_SAFE_INTEGER: -9007199254740991
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(140);
+                r({
+                    target: "Number",
+                    stat: !0,
+                    forced: Number.parseFloat != a
+                }, {
+                    parseFloat: a
+                })
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(128).trim,
+                    s = n(129),
+                    i = r.parseFloat,
+                    o = 1 / i(s + "-0") != -1 / 0;
+                e.exports = o ? function (e) {
+                    var t = a(String(e)),
+                        n = i(t);
+                    return 0 === n && "-" == t.charAt(0) ? -0 : n
+                } : i
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(142);
+                r({
+                    target: "Number",
+                    stat: !0,
+                    forced: Number.parseInt != a
+                }, {
+                    parseInt: a
+                })
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(128).trim,
+                    s = n(129),
+                    i = r.parseInt,
+                    o = /^[+-]?0[Xx]/,
+                    u = 8 !== i(s + "08") || 22 !== i(s + "0x16");
+                e.exports = u ? function (e, t) {
+                    var n = a(String(e));
+                    return i(n, t >>> 0 || (o.test(n) ? 16 : 10))
+                } : i
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(40),
+                    s = n(144),
+                    i = n(145),
+                    o = n(6),
+                    u = 1..toFixed,
+                    l = Math.floor,
+                    c = function (e, t, n) {
+                        return 0 === t ? n : t % 2 == 1 ? c(e, t - 1, n * e) : c(e * e, t / 2, n)
+                    };
+                r({
+                    target: "Number",
+                    proto: !0,
+                    forced: u && ("0.000" !== 8e-5.toFixed(3) || "1" !== .9.toFixed(0) || "1.25" !== 1.255.toFixed(2) || "1000000000000000128" !== (0xde0b6b3a7640080).toFixed(0)) || !o((function () {
+                        u.call({})
+                    }))
+                }, {
+                    toFixed: function (e) {
+                        var t, n, r, o, u = s(this),
+                            h = a(e),
+                            p = [0, 0, 0, 0, 0, 0],
+                            d = "",
+                            f = "0",
+                            m = function (e, t) {
+                                for (var n = -1, r = t; ++n < 6;) r += e * p[n], p[n] = r % 1e7, r = l(r / 1e7)
+                            },
+                            g = function (e) {
+                                for (var t = 6, n = 0; --t >= 0;) n += p[t], p[t] = l(n / e), n = n % e * 1e7
+                            },
+                            y = function () {
+                                for (var e = 6, t = ""; --e >= 0;)
+                                    if ("" !== t || 0 === e || 0 !== p[e]) {
+                                        var n = String(p[e]);
+                                        t = "" === t ? n : t + i.call("0", 7 - n.length) + n
+                                    } return t
+                            };
+                        if (h < 0 || h > 20) throw RangeError("Incorrect fraction digits");
+                        if (u != u) return "NaN";
+                        if (u <= -1e21 || u >= 1e21) return String(u);
+                        if (u < 0 && (d = "-", u = -u), u > 1e-21)
+                            if (n = (t = function (e) {
+                                    for (var t = 0, n = e; n >= 4096;) t += 12, n /= 4096;
+                                    for (; n >= 2;) t += 1, n /= 2;
+                                    return t
+                                }(u * c(2, 69, 1)) - 69) < 0 ? u * c(2, -t, 1) : u / c(2, t, 1), n *= 4503599627370496, (t = 52 - t) > 0) {
+                                for (m(0, n), r = h; r >= 7;) m(1e7, 0), r -= 7;
+                                for (m(c(10, r, 1), 0), r = t - 1; r >= 23;) g(1 << 23), r -= 23;
+                                g(1 << r), m(1, 1), g(2), f = y()
+                            } else m(0, n), m(1 << -t, 0), f = y() + i.call("0", h);
+                        return h > 0 ? d + ((o = f.length) <= h ? "0." + i.call("0", h - o) + f : f.slice(0, o - h) + "." + f.slice(o - h)) : d + f
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(11);
+                e.exports = function (e) {
+                    if ("number" != typeof e && "Number" != r(e)) throw TypeError("Incorrect invocation");
+                    return +e
+                }
+            }, function (e, t, n) {
+                var r = n(40),
+                    a = n(12);
+                e.exports = "".repeat || function (e) {
+                    var t = String(a(this)),
+                        n = "",
+                        s = r(e);
+                    if (s < 0 || s == 1 / 0) throw RangeError("Wrong number of repetitions");
+                    for (; s > 0;
+                        (s >>>= 1) && (t += t)) 1 & s && (n += t);
+                    return n
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(147);
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: Object.assign !== a
+                }, {
+                    assign: a
+                })
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(6),
+                    s = n(60),
+                    i = n(43),
+                    o = n(7),
+                    u = n(46),
+                    l = n(10),
+                    c = Object.assign,
+                    h = Object.defineProperty;
+                e.exports = !c || a((function () {
+                    if (r && 1 !== c({
+                            b: 1
+                        }, c(h({}, "a", {
+                            enumerable: !0,
+                            get: function () {
+                                h(this, "b", {
+                                    value: 3,
+                                    enumerable: !1
+                                })
+                            }
+                        }), {
+                            b: 2
+                        })).b) return !0;
+                    var e = {},
+                        t = {},
+                        n = Symbol();
+                    return e[n] = 7, "abcdefghijklmnopqrst".split("").forEach((function (e) {
+                        t[e] = e
+                    })), 7 != c({}, e)[n] || "abcdefghijklmnopqrst" != s(c({}, t)).join("")
+                })) ? function (e, t) {
+                    for (var n = u(e), a = arguments.length, c = 1, h = i.f, p = o.f; a > c;)
+                        for (var d, f = l(arguments[c++]), m = h ? s(f).concat(h(f)) : s(f), g = m.length, y = 0; g > y;) d = m[y++], r && !p.call(f, d) || (n[d] = f[d]);
+                    return n
+                } : c
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(5),
+                    s = n(149),
+                    i = n(46),
+                    o = n(65),
+                    u = n(19);
+                a && r({
+                    target: "Object",
+                    proto: !0,
+                    forced: s
+                }, {
+                    __defineGetter__: function (e, t) {
+                        u.f(i(this), e, {
+                            get: o(t),
+                            enumerable: !0,
+                            configurable: !0
+                        })
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(29),
+                    a = n(3),
+                    s = n(6);
+                e.exports = r || !s((function () {
+                    var e = Math.random();
+                    __defineSetter__.call(null, e, (function () {})), delete a[e]
+                }))
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(5),
+                    s = n(149),
+                    i = n(46),
+                    o = n(65),
+                    u = n(19);
+                a && r({
+                    target: "Object",
+                    proto: !0,
+                    forced: s
+                }, {
+                    __defineSetter__: function (e, t) {
+                        u.f(i(this), e, {
+                            set: o(t),
+                            enumerable: !0,
+                            configurable: !0
+                        })
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(152).entries;
+                r({
+                    target: "Object",
+                    stat: !0
+                }, {
+                    entries: function (e) {
+                        return a(e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(60),
+                    s = n(9),
+                    i = n(7).f,
+                    o = function (e) {
+                        return function (t) {
+                            for (var n, o = s(t), u = a(o), l = u.length, c = 0, h = []; l > c;) n = u[c++], r && !i.call(o, n) || h.push(e ? [n, o[n]] : o[n]);
+                            return h
+                        }
+                    };
+                e.exports = {
+                    entries: o(!0),
+                    values: o(!1)
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(121),
+                    s = n(6),
+                    i = n(14),
+                    o = n(120).onFreeze,
+                    u = Object.freeze;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: s((function () {
+                        u(1)
+                    })),
+                    sham: !a
+                }, {
+                    freeze: function (e) {
+                        return u && i(e) ? u(o(e)) : e
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(122),
+                    s = n(47);
+                r({
+                    target: "Object",
+                    stat: !0
+                }, {
+                    fromEntries: function (e) {
+                        var t = {};
+                        return a(e, (function (e, n) {
+                            s(t, e, n)
+                        }), void 0, !0), t
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(9),
+                    i = n(4).f,
+                    o = n(5),
+                    u = a((function () {
+                        i(1)
+                    }));
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: !o || u,
+                    sham: !o
+                }, {
+                    getOwnPropertyDescriptor: function (e, t) {
+                        return i(s(e), t)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(5),
+                    s = n(33),
+                    i = n(9),
+                    o = n(4),
+                    u = n(47);
+                r({
+                    target: "Object",
+                    stat: !0,
+                    sham: !a
+                }, {
+                    getOwnPropertyDescriptors: function (e) {
+                        for (var t, n, r = i(e), a = o.f, l = s(r), c = {}, h = 0; l.length > h;) void 0 !== (n = a(r, t = l[h++])) && u(c, t, n);
+                        return c
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(158).f;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: a((function () {
+                        return !Object.getOwnPropertyNames(1)
+                    }))
+                }, {
+                    getOwnPropertyNames: s
+                })
+            }, function (e, t, n) {
+                var r = n(9),
+                    a = n(36).f,
+                    s = {}.toString,
+                    i = "object" == typeof window && window && Object.getOwnPropertyNames ? Object.getOwnPropertyNames(window) : [];
+                e.exports.f = function (e) {
+                    return i && "[object Window]" == s.call(e) ? function (e) {
+                        try {
+                            return a(e)
+                        } catch (e) {
+                            return i.slice()
+                        }
+                    }(e) : a(r(e))
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(46),
+                    i = n(93),
+                    o = n(94);
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: a((function () {
+                        i(1)
+                    })),
+                    sham: !o
+                }, {
+                    getPrototypeOf: function (e) {
+                        return i(s(e))
+                    }
+                })
+            }, function (e, t, n) {
+                n(2)({
+                    target: "Object",
+                    stat: !0
+                }, {
+                    is: n(161)
+                })
+            }, function (e, t) {
+                e.exports = Object.is || function (e, t) {
+                    return e === t ? 0 !== e || 1 / e == 1 / t : e != e && t != t
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(14),
+                    i = Object.isExtensible;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: a((function () {
+                        i(1)
+                    }))
+                }, {
+                    isExtensible: function (e) {
+                        return !!s(e) && (!i || i(e))
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(14),
+                    i = Object.isFrozen;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: a((function () {
+                        i(1)
+                    }))
+                }, {
+                    isFrozen: function (e) {
+                        return !s(e) || !!i && i(e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(6),
+                    s = n(14),
+                    i = Object.isSealed;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: a((function () {
+                        i(1)
+                    }))
+                }, {
+                    isSealed: function (e) {
+                        return !s(e) || !!i && i(e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(46),
+                    s = n(60);
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: n(6)((function () {
+                        s(1)
+                    }))
+                }, {
+                    keys: function (e) {
+                        return s(a(e))
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(5),
+                    s = n(149),
+                    i = n(46),
+                    o = n(13),
+                    u = n(93),
+                    l = n(4).f;
+                a && r({
+                    target: "Object",
+                    proto: !0,
+                    forced: s
+                }, {
+                    __lookupGetter__: function (e) {
+                        var t, n = i(this),
+                            r = o(e, !0);
+                        do {
+                            if (t = l(n, r)) return t.get
+                        } while (n = u(n))
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(5),
+                    s = n(149),
+                    i = n(46),
+                    o = n(13),
+                    u = n(93),
+                    l = n(4).f;
+                a && r({
+                    target: "Object",
+                    proto: !0,
+                    forced: s
+                }, {
+                    __lookupSetter__: function (e) {
+                        var t, n = i(this),
+                            r = o(e, !0);
+                        do {
+                            if (t = l(n, r)) return t.set
+                        } while (n = u(n))
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(14),
+                    s = n(120).onFreeze,
+                    i = n(121),
+                    o = n(6),
+                    u = Object.preventExtensions;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: o((function () {
+                        u(1)
+                    })),
+                    sham: !i
+                }, {
+                    preventExtensions: function (e) {
+                        return u && a(e) ? u(s(e)) : e
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(14),
+                    s = n(120).onFreeze,
+                    i = n(121),
+                    o = n(6),
+                    u = Object.seal;
+                r({
+                    target: "Object",
+                    stat: !0,
+                    forced: o((function () {
+                        u(1)
+                    })),
+                    sham: !i
+                }, {
+                    seal: function (e) {
+                        return u && a(e) ? u(s(e)) : e
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(85),
+                    a = n(21),
+                    s = n(171);
+                r || a(Object.prototype, "toString", s, {
+                    unsafe: !0
+                })
+            }, function (e, t, n) {
+                var r = n(85),
+                    a = n(84);
+                e.exports = r ? {}.toString : function () {
+                    return "[object " + a(this) + "]"
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(152).values;
+                r({
+                    target: "Object",
+                    stat: !0
+                }, {
+                    values: function (e) {
+                        return a(e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r, a, s, i, o = n(2),
+                    u = n(29),
+                    l = n(3),
+                    c = n(34),
+                    h = n(174),
+                    p = n(21),
+                    d = n(126),
+                    f = n(95),
+                    m = n(109),
+                    g = n(14),
+                    y = n(65),
+                    b = n(123),
+                    v = n(11),
+                    x = n(23),
+                    w = n(122),
+                    k = n(86),
+                    S = n(175),
+                    I = n(176).set,
+                    A = n(178),
+                    E = n(179),
+                    N = n(181),
+                    C = n(180),
+                    T = n(182),
+                    R = n(25),
+                    _ = n(44),
+                    F = n(49),
+                    M = n(53),
+                    D = F("species"),
+                    O = "Promise",
+                    L = R.get,
+                    P = R.set,
+                    $ = R.getterFor(O),
+                    B = h,
+                    z = l.TypeError,
+                    U = l.document,
+                    W = l.process,
+                    V = c("fetch"),
+                    j = C.f,
+                    G = j,
+                    H = "process" == v(W),
+                    q = !!(U && U.createEvent && l.dispatchEvent),
+                    K = _(O, (function () {
+                        if (x(B) === String(B)) {
+                            if (66 === M) return !0;
+                            if (!H && "function" != typeof PromiseRejectionEvent) return !0
+                        }
+                        if (u && !B.prototype.finally) return !0;
+                        if (M >= 51 && /native code/.test(B)) return !1;
+                        var e = B.resolve(1),
+                            t = function (e) {
+                                e((function () {}), (function () {}))
+                            };
+                        return (e.constructor = {})[D] = t, !(e.then((function () {})) instanceof t)
+                    })),
+                    Y = K || !k((function (e) {
+                        B.all(e).catch((function () {}))
+                    })),
+                    X = function (e) {
+                        var t;
+                        return !(!g(e) || "function" != typeof (t = e.then)) && t
+                    },
+                    Q = function (e, t, n) {
+                        if (!t.notified) {
+                            t.notified = !0;
+                            var r = t.reactions;
+                            A((function () {
+                                for (var a = t.value, s = 1 == t.state, i = 0; r.length > i;) {
+                                    var o, u, l, c = r[i++],
+                                        h = s ? c.ok : c.fail,
+                                        p = c.resolve,
+                                        d = c.reject,
+                                        f = c.domain;
+                                    try {
+                                        h ? (s || (2 === t.rejection && te(e, t), t.rejection = 1), !0 === h ? o = a : (f && f.enter(), o = h(a), f && (f.exit(), l = !0)), o === c.promise ? d(z("Promise-chain cycle")) : (u = X(o)) ? u.call(o, p, d) : p(o)) : d(a)
+                                    } catch (e) {
+                                        f && !l && f.exit(), d(e)
+                                    }
+                                }
+                                t.reactions = [], t.notified = !1, n && !t.rejection && J(e, t)
+                            }))
+                        }
+                    },
+                    Z = function (e, t, n) {
+                        var r, a;
+                        q ? ((r = U.createEvent("Event")).promise = t, r.reason = n, r.initEvent(e, !1, !0), l.dispatchEvent(r)) : r = {
+                            promise: t,
+                            reason: n
+                        }, (a = l["on" + e]) ? a(r) : "unhandledrejection" === e && N("Unhandled promise rejection", n)
+                    },
+                    J = function (e, t) {
+                        I.call(l, (function () {
+                            var n, r = t.value;
+                            if (ee(t) && (n = T((function () {
+                                    H ? W.emit("unhandledRejection", r, e) : Z("unhandledrejection", e, r)
+                                })), t.rejection = H || ee(t) ? 2 : 1, n.error)) throw n.value
+                        }))
+                    },
+                    ee = function (e) {
+                        return 1 !== e.rejection && !e.parent
+                    },
+                    te = function (e, t) {
+                        I.call(l, (function () {
+                            H ? W.emit("rejectionHandled", e) : Z("rejectionhandled", e, t.value)
+                        }))
+                    },
+                    ne = function (e, t, n, r) {
+                        return function (a) {
+                            e(t, n, a, r)
+                        }
+                    },
+                    re = function (e, t, n, r) {
+                        t.done || (t.done = !0, r && (t = r), t.value = n, t.state = 2, Q(e, t, !0))
+                    },
+                    ae = function (e, t, n, r) {
+                        if (!t.done) {
+                            t.done = !0, r && (t = r);
+                            try {
+                                if (e === n) throw z("Promise can't be resolved itself");
+                                var a = X(n);
+                                a ? A((function () {
+                                    var r = {
+                                        done: !1
+                                    };
+                                    try {
+                                        a.call(n, ne(ae, e, r, t), ne(re, e, r, t))
+                                    } catch (n) {
+                                        re(e, r, n, t)
+                                    }
+                                })) : (t.value = n, t.state = 1, Q(e, t, !1))
+                            } catch (n) {
+                                re(e, {
+                                    done: !1
+                                }, n, t)
+                            }
+                        }
+                    };
+                K && (B = function (e) {
+                    b(this, B, O), y(e), r.call(this);
+                    var t = L(this);
+                    try {
+                        e(ne(ae, this, t), ne(re, this, t))
+                    } catch (e) {
+                        re(this, t, e)
+                    }
+                }, (r = function (e) {
+                    P(this, {
+                        type: O,
+                        done: !1,
+                        notified: !1,
+                        parent: !1,
+                        reactions: [],
+                        rejection: !1,
+                        state: 0,
+                        value: void 0
+                    })
+                }).prototype = d(B.prototype, {
+                    then: function (e, t) {
+                        var n = $(this),
+                            r = j(S(this, B));
+                        return r.ok = "function" != typeof e || e, r.fail = "function" == typeof t && t, r.domain = H ? W.domain : void 0, n.parent = !0, n.reactions.push(r), 0 != n.state && Q(this, n, !1), r.promise
+                    },
+                    catch: function (e) {
+                        return this.then(void 0, e)
+                    }
+                }), a = function () {
+                    var e = new r,
+                        t = L(e);
+                    this.promise = e, this.resolve = ne(ae, e, t), this.reject = ne(re, e, t)
+                }, C.f = j = function (e) {
+                    return e === B || e === s ? new a(e) : G(e)
+                }, u || "function" != typeof h || (i = h.prototype.then, p(h.prototype, "then", (function (e, t) {
+                    var n = this;
+                    return new B((function (e, t) {
+                        i.call(n, e, t)
+                    })).then(e, t)
+                }), {
+                    unsafe: !0
+                }), "function" == typeof V && o({
+                    global: !0,
+                    enumerable: !0,
+                    forced: !0
+                }, {
+                    fetch: function (e) {
+                        return E(B, V.apply(l, arguments))
+                    }
+                }))), o({
+                    global: !0,
+                    wrap: !0,
+                    forced: K
+                }, {
+                    Promise: B
+                }), f(B, O, !1, !0), m(O), s = c(O), o({
+                    target: O,
+                    stat: !0,
+                    forced: K
+                }, {
+                    reject: function (e) {
+                        var t = j(this);
+                        return t.reject.call(void 0, e), t.promise
+                    }
+                }), o({
+                    target: O,
+                    stat: !0,
+                    forced: u || K
+                }, {
+                    resolve: function (e) {
+                        return E(u && this === s ? B : this, e)
+                    }
+                }), o({
+                    target: O,
+                    stat: !0,
+                    forced: Y
+                }, {
+                    all: function (e) {
+                        var t = this,
+                            n = j(t),
+                            r = n.resolve,
+                            a = n.reject,
+                            s = T((function () {
+                                var n = y(t.resolve),
+                                    s = [],
+                                    i = 0,
+                                    o = 1;
+                                w(e, (function (e) {
+                                    var u = i++,
+                                        l = !1;
+                                    s.push(void 0), o++, n.call(t, e).then((function (e) {
+                                        l || (l = !0, s[u] = e, --o || r(s))
+                                    }), a)
+                                })), --o || r(s)
+                            }));
+                        return s.error && a(s.value), n.promise
+                    },
+                    race: function (e) {
+                        var t = this,
+                            n = j(t),
+                            r = n.reject,
+                            a = T((function () {
+                                var a = y(t.resolve);
+                                w(e, (function (e) {
+                                    a.call(t, e).then(n.resolve, r)
+                                }))
+                            }));
+                        return a.error && r(a.value), n.promise
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(3);
+                e.exports = r.Promise
+            }, function (e, t, n) {
+                var r = n(20),
+                    a = n(65),
+                    s = n(49)("species");
+                e.exports = function (e, t) {
+                    var n, i = r(e).constructor;
+                    return void 0 === i || null == (n = r(i)[s]) ? t : a(n)
+                }
+            }, function (e, t, n) {
+                var r, a, s, i = n(3),
+                    o = n(6),
+                    u = n(11),
+                    l = n(64),
+                    c = n(61),
+                    h = n(17),
+                    p = n(177),
+                    d = i.location,
+                    f = i.setImmediate,
+                    m = i.clearImmediate,
+                    g = i.process,
+                    y = i.MessageChannel,
+                    b = i.Dispatch,
+                    v = 0,
+                    x = {},
+                    w = function (e) {
+                        if (x.hasOwnProperty(e)) {
+                            var t = x[e];
+                            delete x[e], t()
+                        }
+                    },
+                    k = function (e) {
+                        return function () {
+                            w(e)
+                        }
+                    },
+                    S = function (e) {
+                        w(e.data)
+                    },
+                    I = function (e) {
+                        i.postMessage(e + "", d.protocol + "//" + d.host)
+                    };
+                f && m || (f = function (e) {
+                    for (var t = [], n = 1; arguments.length > n;) t.push(arguments[n++]);
+                    return x[++v] = function () {
+                        ("function" == typeof e ? e : Function(e)).apply(void 0, t)
+                    }, r(v), v
+                }, m = function (e) {
+                    delete x[e]
+                }, "process" == u(g) ? r = function (e) {
+                    g.nextTick(k(e))
+                } : b && b.now ? r = function (e) {
+                    b.now(k(e))
+                } : y && !p ? (s = (a = new y).port2, a.port1.onmessage = S, r = l(s.postMessage, s, 1)) : !i.addEventListener || "function" != typeof postMessage || i.importScripts || o(I) || "file:" === d.protocol ? r = "onreadystatechange" in h("script") ? function (e) {
+                    c.appendChild(h("script")).onreadystatechange = function () {
+                        c.removeChild(this), w(e)
+                    }
+                } : function (e) {
+                    setTimeout(k(e), 0)
+                } : (r = I, i.addEventListener("message", S, !1))), e.exports = {
+                    set: f,
+                    clear: m
+                }
+            }, function (e, t, n) {
+                var r = n(54);
+                e.exports = /(iphone|ipod|ipad).*applewebkit/i.test(r)
+            }, function (e, t, n) {
+                var r, a, s, i, o, u, l, c, h = n(3),
+                    p = n(4).f,
+                    d = n(11),
+                    f = n(176).set,
+                    m = n(177),
+                    g = h.MutationObserver || h.WebKitMutationObserver,
+                    y = h.process,
+                    b = h.Promise,
+                    v = "process" == d(y),
+                    x = p(h, "queueMicrotask"),
+                    w = x && x.value;
+                w || (r = function () {
+                    var e, t;
+                    for (v && (e = y.domain) && e.exit(); a;) {
+                        t = a.fn, a = a.next;
+                        try {
+                            t()
+                        } catch (e) {
+                            throw a ? i() : s = void 0, e
+                        }
+                    }
+                    s = void 0, e && e.enter()
+                }, v ? i = function () {
+                    y.nextTick(r)
+                } : g && !m ? (o = !0, u = document.createTextNode(""), new g(r).observe(u, {
+                    characterData: !0
+                }), i = function () {
+                    u.data = o = !o
+                }) : b && b.resolve ? (l = b.resolve(void 0), c = l.then, i = function () {
+                    c.call(l, r)
+                }) : i = function () {
+                    f.call(h, r)
+                }), e.exports = w || function (e) {
+                    var t = {
+                        fn: e,
+                        next: void 0
+                    };
+                    s && (s.next = t), a || (a = t, i()), s = t
+                }
+            }, function (e, t, n) {
+                var r = n(20),
+                    a = n(14),
+                    s = n(180);
+                e.exports = function (e, t) {
+                    if (r(e), a(t) && t.constructor === e) return t;
+                    var n = s.f(e);
+                    return (0, n.resolve)(t), n.promise
+                }
+            }, function (e, t, n) {
+                var r = n(65),
+                    a = function (e) {
+                        var t, n;
+                        this.promise = new e((function (e, r) {
+                            if (void 0 !== t || void 0 !== n) throw TypeError("Bad Promise constructor");
+                            t = e, n = r
+                        })), this.resolve = r(t), this.reject = r(n)
+                    };
+                e.exports.f = function (e) {
+                    return new a(e)
+                }
+            }, function (e, t, n) {
+                var r = n(3);
+                e.exports = function (e, t) {
+                    var n = r.console;
+                    n && n.error && (1 === arguments.length ? n.error(e) : n.error(e, t))
+                }
+            }, function (e, t) {
+                e.exports = function (e) {
+                    try {
+                        return {
+                            error: !1,
+                            value: e()
+                        }
+                    } catch (e) {
+                        return {
+                            error: !0,
+                            value: e
+                        }
+                    }
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(65),
+                    s = n(180),
+                    i = n(182),
+                    o = n(122);
+                r({
+                    target: "Promise",
+                    stat: !0
+                }, {
+                    allSettled: function (e) {
+                        var t = this,
+                            n = s.f(t),
+                            r = n.resolve,
+                            u = n.reject,
+                            l = i((function () {
+                                var n = a(t.resolve),
+                                    s = [],
+                                    i = 0,
+                                    u = 1;
+                                o(e, (function (e) {
+                                    var a = i++,
+                                        o = !1;
+                                    s.push(void 0), u++, n.call(t, e).then((function (e) {
+                                        o || (o = !0, s[a] = {
+                                            status: "fulfilled",
+                                            value: e
+                                        }, --u || r(s))
+                                    }), (function (e) {
+                                        o || (o = !0, s[a] = {
+                                            status: "rejected",
+                                            reason: e
+                                        }, --u || r(s))
+                                    }))
+                                })), --u || r(s)
+                            }));
+                        return l.error && u(l.value), n.promise
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(29),
+                    s = n(174),
+                    i = n(6),
+                    o = n(34),
+                    u = n(175),
+                    l = n(179),
+                    c = n(21);
+                r({
+                    target: "Promise",
+                    proto: !0,
+                    real: !0,
+                    forced: !!s && i((function () {
+                        s.prototype.finally.call({
+                            then: function () {}
+                        }, (function () {}))
+                    }))
+                }, {
+                    finally: function (e) {
+                        var t = u(this, o("Promise")),
+                            n = "function" == typeof e;
+                        return this.then(n ? function (n) {
+                            return l(t, e()).then((function () {
+                                return n
+                            }))
+                        } : e, n ? function (n) {
+                            return l(t, e()).then((function () {
+                                throw n
+                            }))
+                        } : e)
+                    }
+                }), a || "function" != typeof s || s.prototype.finally || c(s.prototype, "finally", o("Promise").prototype.finally)
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(3),
+                    s = n(44),
+                    i = n(124),
+                    o = n(19).f,
+                    u = n(36).f,
+                    l = n(186),
+                    c = n(187),
+                    h = n(188),
+                    p = n(21),
+                    d = n(6),
+                    f = n(25).set,
+                    m = n(109),
+                    g = n(49)("match"),
+                    y = a.RegExp,
+                    b = y.prototype,
+                    v = /a/g,
+                    x = /a/g,
+                    w = new y(v) !== v,
+                    k = h.UNSUPPORTED_Y;
+                if (r && s("RegExp", !w || k || d((function () {
+                        return x[g] = !1, y(v) != v || y(x) == x || "/a/i" != y(v, "i")
+                    })))) {
+                    for (var S = function (e, t) {
+                            var n, r = this instanceof S,
+                                a = l(e),
+                                s = void 0 === t;
+                            if (!r && a && e.constructor === S && s) return e;
+                            w ? a && !s && (e = e.source) : e instanceof S && (s && (t = c.call(e)), e = e.source), k && (n = !!t && t.indexOf("y") > -1) && (t = t.replace(/y/g, ""));
+                            var o = i(w ? new y(e, t) : y(e, t), r ? this : b, S);
+                            return k && n && f(o, {
+                                sticky: n
+                            }), o
+                        }, I = function (e) {
+                            e in S || o(S, e, {
+                                configurable: !0,
+                                get: function () {
+                                    return y[e]
+                                },
+                                set: function (t) {
+                                    y[e] = t
+                                }
+                            })
+                        }, A = u(y), E = 0; A.length > E;) I(A[E++]);
+                    b.constructor = S, S.prototype = b, p(a, "RegExp", S)
+                }
+                m("RegExp")
+            }, function (e, t, n) {
+                var r = n(14),
+                    a = n(11),
+                    s = n(49)("match");
+                e.exports = function (e) {
+                    var t;
+                    return r(e) && (void 0 !== (t = e[s]) ? !!t : "RegExp" == a(e))
+                }
+            }, function (e, t, n) {
+                var r = n(20);
+                e.exports = function () {
+                    var e = r(this),
+                        t = "";
+                    return e.global && (t += "g"), e.ignoreCase && (t += "i"), e.multiline && (t += "m"), e.dotAll && (t += "s"), e.unicode && (t += "u"), e.sticky && (t += "y"), t
+                }
+            }, function (e, t, n) {
+                var r = n(6);
+
+                function a(e, t) {
+                    return RegExp(e, t)
+                }
+                t.UNSUPPORTED_Y = r((function () {
+                    var e = a("a", "y");
+                    return e.lastIndex = 2, null != e.exec("abcd")
+                })), t.BROKEN_CARET = r((function () {
+                    var e = a("^r", "gy");
+                    return e.lastIndex = 2, null != e.exec("str")
+                }))
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(190);
+                r({
+                    target: "RegExp",
+                    proto: !0,
+                    forced: /./.exec !== a
+                }, {
+                    exec: a
+                })
+            }, function (e, t, n) {
+                var r, a, s = n(187),
+                    i = n(188),
+                    o = RegExp.prototype.exec,
+                    u = String.prototype.replace,
+                    l = o,
+                    c = (r = /a/, a = /b*/g, o.call(r, "a"), o.call(a, "a"), 0 !== r.lastIndex || 0 !== a.lastIndex),
+                    h = i.UNSUPPORTED_Y || i.BROKEN_CARET,
+                    p = void 0 !== /()??/.exec("")[1];
+                (c || p || h) && (l = function (e) {
+                    var t, n, r, a, i = this,
+                        l = h && i.sticky,
+                        d = s.call(i),
+                        f = i.source,
+                        m = 0,
+                        g = e;
+                    return l && (-1 === (d = d.replace("y", "")).indexOf("g") && (d += "g"), g = String(e).slice(i.lastIndex), i.lastIndex > 0 && (!i.multiline || i.multiline && "\n" !== e[i.lastIndex - 1]) && (f = "(?: " + f + ")", g = " " + g, m++), n = new RegExp("^(?:" + f + ")", d)), p && (n = new RegExp("^" + f + "$(?!\\s)", d)), c && (t = i.lastIndex), r = o.call(l ? n : i, g), l ? r ? (r.input = r.input.slice(m), r[0] = r[0].slice(m), r.index = i.lastIndex, i.lastIndex += r[0].length) : i.lastIndex = 0 : c && r && (i.lastIndex = i.global ? r.index + r[0].length : t), p && r && r.length > 1 && u.call(r[0], n, (function () {
+                        for (a = 1; a < arguments.length - 2; a++) void 0 === arguments[a] && (r[a] = void 0)
+                    })), r
+                }), e.exports = l
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(19),
+                    s = n(187),
+                    i = n(188).UNSUPPORTED_Y;
+                r && ("g" != /./g.flags || i) && a.f(RegExp.prototype, "flags", {
+                    configurable: !0,
+                    get: s
+                })
+            }, function (e, t, n) {
+                var r = n(5),
+                    a = n(188).UNSUPPORTED_Y,
+                    s = n(19).f,
+                    i = n(25).get,
+                    o = RegExp.prototype;
+                r && a && s(RegExp.prototype, "sticky", {
+                    configurable: !0,
+                    get: function () {
+                        if (this !== o) {
+                            if (this instanceof RegExp) return !!i(this).sticky;
+                            throw TypeError("Incompatible receiver, RegExp required")
+                        }
+                    }
+                })
+            }, function (e, t, n) {
+                n(189);
+                var r, a, s = n(2),
+                    i = n(14),
+                    o = (r = !1, (a = /[ac]/).exec = function () {
+                        return r = !0, /./.exec.apply(this, arguments)
+                    }, !0 === a.test("abc") && r),
+                    u = /./.test;
+                s({
+                    target: "RegExp",
+                    proto: !0,
+                    forced: !o
+                }, {
+                    test: function (e) {
+                        if ("function" != typeof this.exec) return u.call(this, e);
+                        var t = this.exec(e);
+                        if (null !== t && !i(t)) throw new Error("RegExp exec method returned something other than an Object or null");
+                        return !!t
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(21),
+                    a = n(20),
+                    s = n(6),
+                    i = n(187),
+                    o = RegExp.prototype,
+                    u = o.toString,
+                    l = s((function () {
+                        return "/a/b" != u.call({
+                            source: "a",
+                            flags: "b"
+                        })
+                    })),
+                    c = "toString" != u.name;
+                (l || c) && r(RegExp.prototype, "toString", (function () {
+                    var e = a(this),
+                        t = String(e.source),
+                        n = e.flags;
+                    return "/" + t + "/" + String(void 0 === n && e instanceof RegExp && !("flags" in o) ? i.call(e) : n)
+                }), {
+                    unsafe: !0
+                })
+            }, function (e, t, n) {
+                var r = n(119),
+                    a = n(125);
+                e.exports = r("Set", (function (e) {
+                    return function () {
+                        return e(this, arguments.length ? arguments[0] : void 0)
+                    }
+                }), a)
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(197).codeAt;
+                r({
+                    target: "String",
+                    proto: !0
+                }, {
+                    codePointAt: function (e) {
+                        return a(this, e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(40),
+                    a = n(12),
+                    s = function (e) {
+                        return function (t, n) {
+                            var s, i, o = String(a(t)),
+                                u = r(n),
+                                l = o.length;
+                            return u < 0 || u >= l ? e ? "" : void 0 : (s = o.charCodeAt(u)) < 55296 || s > 56319 || u + 1 === l || (i = o.charCodeAt(u + 1)) < 56320 || i > 57343 ? e ? o.charAt(u) : s : e ? o.slice(u, u + 2) : i - 56320 + (s - 55296 << 10) + 65536
+                        }
+                    };
+                e.exports = {
+                    codeAt: s(!1),
+                    charAt: s(!0)
+                }
+            }, function (e, t, n) {
+                var r, a = n(2),
+                    s = n(4).f,
+                    i = n(39),
+                    o = n(199),
+                    u = n(12),
+                    l = n(200),
+                    c = n(29),
+                    h = "".endsWith,
+                    p = Math.min,
+                    d = l("endsWith");
+                a({
+                    target: "String",
+                    proto: !0,
+                    forced: !(!c && !d && (r = s(String.prototype, "endsWith"), r && !r.writable) || d)
+                }, {
+                    endsWith: function (e) {
+                        var t = String(u(this));
+                        o(e);
+                        var n = arguments.length > 1 ? arguments[1] : void 0,
+                            r = i(t.length),
+                            a = void 0 === n ? r : p(i(n), r),
+                            s = String(e);
+                        return h ? h.call(t, s, a) : t.slice(a - s.length, a) === s
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(186);
+                e.exports = function (e) {
+                    if (r(e)) throw TypeError("The method doesn't accept regular expressions");
+                    return e
+                }
+            }, function (e, t, n) {
+                var r = n(49)("match");
+                e.exports = function (e) {
+                    var t = /./;
+                    try {
+                        "/./" [e](t)
+                    } catch (n) {
+                        try {
+                            return t[r] = !1, "/./" [e](t)
+                        } catch (e) {}
+                    }
+                    return !1
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(41),
+                    s = String.fromCharCode,
+                    i = String.fromCodePoint;
+                r({
+                    target: "String",
+                    stat: !0,
+                    forced: !!i && 1 != i.length
+                }, {
+                    fromCodePoint: function (e) {
+                        for (var t, n = [], r = arguments.length, i = 0; r > i;) {
+                            if (t = +arguments[i++], a(t, 1114111) !== t) throw RangeError(t + " is not a valid code point");
+                            n.push(t < 65536 ? s(t) : s(55296 + ((t -= 65536) >> 10), t % 1024 + 56320))
+                        }
+                        return n.join("")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(199),
+                    s = n(12);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: !n(200)("includes")
+                }, {
+                    includes: function (e) {
+                        return !!~String(s(this)).indexOf(a(e), arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(197).charAt,
+                    a = n(25),
+                    s = n(90),
+                    i = a.set,
+                    o = a.getterFor("String Iterator");
+                s(String, "String", (function (e) {
+                    i(this, {
+                        type: "String Iterator",
+                        string: String(e),
+                        index: 0
+                    })
+                }), (function () {
+                    var e, t = o(this),
+                        n = t.string,
+                        a = t.index;
+                    return a >= n.length ? {
+                        value: void 0,
+                        done: !0
+                    } : (e = r(n, a), t.index += e.length, {
+                        value: e,
+                        done: !1
+                    })
+                }))
+            }, function (e, t, n) {
+                var r = n(205),
+                    a = n(20),
+                    s = n(39),
+                    i = n(12),
+                    o = n(206),
+                    u = n(207);
+                r("match", 1, (function (e, t, n) {
+                    return [function (t) {
+                        var n = i(this),
+                            r = null == t ? void 0 : t[e];
+                        return void 0 !== r ? r.call(t, n) : new RegExp(t)[e](String(n))
+                    }, function (e) {
+                        var r = n(t, e, this);
+                        if (r.done) return r.value;
+                        var i = a(e),
+                            l = String(this);
+                        if (!i.global) return u(i, l);
+                        var c = i.unicode;
+                        i.lastIndex = 0;
+                        for (var h, p = [], d = 0; null !== (h = u(i, l));) {
+                            var f = String(h[0]);
+                            p[d] = f, "" === f && (i.lastIndex = o(l, s(i.lastIndex), c)), d++
+                        }
+                        return 0 === d ? null : p
+                    }]
+                }))
+            }, function (e, t, n) {
+                n(189);
+                var r = n(21),
+                    a = n(6),
+                    s = n(49),
+                    i = n(190),
+                    o = n(18),
+                    u = s("species"),
+                    l = !a((function () {
+                        var e = /./;
+                        return e.exec = function () {
+                            var e = [];
+                            return e.groups = {
+                                a: "7"
+                            }, e
+                        }, "7" !== "".replace(e, "$<a>")
+                    })),
+                    c = "$0" === "a".replace(/./, "$0"),
+                    h = s("replace"),
+                    p = !!/./ [h] && "" === /./ [h]("a", "$0"),
+                    d = !a((function () {
+                        var e = /(?:)/,
+                            t = e.exec;
+                        e.exec = function () {
+                            return t.apply(this, arguments)
+                        };
+                        var n = "ab".split(e);
+                        return 2 !== n.length || "a" !== n[0] || "b" !== n[1]
+                    }));
+                e.exports = function (e, t, n, h) {
+                    var f = s(e),
+                        m = !a((function () {
+                            var t = {};
+                            return t[f] = function () {
+                                return 7
+                            }, 7 != "" [e](t)
+                        })),
+                        g = m && !a((function () {
+                            var t = !1,
+                                n = /a/;
+                            return "split" === e && ((n = {}).constructor = {}, n.constructor[u] = function () {
+                                return n
+                            }, n.flags = "", n[f] = /./ [f]), n.exec = function () {
+                                return t = !0, null
+                            }, n[f](""), !t
+                        }));
+                    if (!m || !g || "replace" === e && (!l || !c || p) || "split" === e && !d) {
+                        var y = /./ [f],
+                            b = n(f, "" [e], (function (e, t, n, r, a) {
+                                return t.exec === i ? m && !a ? {
+                                    done: !0,
+                                    value: y.call(t, n, r)
+                                } : {
+                                    done: !0,
+                                    value: e.call(n, t, r)
+                                } : {
+                                    done: !1
+                                }
+                            }), {
+                                REPLACE_KEEPS_$0: c,
+                                REGEXP_REPLACE_SUBSTITUTES_UNDEFINED_CAPTURE: p
+                            }),
+                            v = b[0],
+                            x = b[1];
+                        r(String.prototype, e, v), r(RegExp.prototype, f, 2 == t ? function (e, t) {
+                            return x.call(e, this, t)
+                        } : function (e) {
+                            return x.call(e, this)
+                        })
+                    }
+                    h && o(RegExp.prototype[f], "sham", !0)
+                }
+            }, function (e, t, n) {
+                var r = n(197).charAt;
+                e.exports = function (e, t, n) {
+                    return t + (n ? r(e, t).length : 1)
+                }
+            }, function (e, t, n) {
+                var r = n(11),
+                    a = n(190);
+                e.exports = function (e, t) {
+                    var n = e.exec;
+                    if ("function" == typeof n) {
+                        var s = n.call(e, t);
+                        if ("object" != typeof s) throw TypeError("RegExp exec method returned something other than an Object or null");
+                        return s
+                    }
+                    if ("RegExp" !== r(e)) throw TypeError("RegExp#exec called on incompatible receiver");
+                    return a.call(e, t)
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(91),
+                    s = n(12),
+                    i = n(39),
+                    o = n(65),
+                    u = n(20),
+                    l = n(11),
+                    c = n(186),
+                    h = n(187),
+                    p = n(18),
+                    d = n(6),
+                    f = n(49),
+                    m = n(175),
+                    g = n(206),
+                    y = n(25),
+                    b = n(29),
+                    v = f("matchAll"),
+                    x = y.set,
+                    w = y.getterFor("RegExp String Iterator"),
+                    k = RegExp.prototype,
+                    S = k.exec,
+                    I = "".matchAll,
+                    A = !!I && !d((function () {
+                        "a".matchAll(/./)
+                    })),
+                    E = a((function (e, t, n, r) {
+                        x(this, {
+                            type: "RegExp String Iterator",
+                            regexp: e,
+                            string: t,
+                            global: n,
+                            unicode: r,
+                            done: !1
+                        })
+                    }), "RegExp String", (function () {
+                        var e = w(this);
+                        if (e.done) return {
+                            value: void 0,
+                            done: !0
+                        };
+                        var t = e.regexp,
+                            n = e.string,
+                            r = function (e, t) {
+                                var n, r = e.exec;
+                                if ("function" == typeof r) {
+                                    if ("object" != typeof (n = r.call(e, t))) throw TypeError("Incorrect exec result");
+                                    return n
+                                }
+                                return S.call(e, t)
+                            }(t, n);
+                        return null === r ? {
+                            value: void 0,
+                            done: e.done = !0
+                        } : e.global ? ("" == String(r[0]) && (t.lastIndex = g(n, i(t.lastIndex), e.unicode)), {
+                            value: r,
+                            done: !1
+                        }) : (e.done = !0, {
+                            value: r,
+                            done: !1
+                        })
+                    })),
+                    N = function (e) {
+                        var t, n, r, a, s, o, l = u(this),
+                            c = String(e);
+                        return t = m(l, RegExp), void 0 === (n = l.flags) && l instanceof RegExp && !("flags" in k) && (n = h.call(l)), r = void 0 === n ? "" : String(n), a = new t(t === RegExp ? l.source : l, r), s = !!~r.indexOf("g"), o = !!~r.indexOf("u"), a.lastIndex = i(l.lastIndex), new E(a, c, s, o)
+                    };
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: A
+                }, {
+                    matchAll: function (e) {
+                        var t, n, r, a = s(this);
+                        if (null != e) {
+                            if (c(e) && !~String(s("flags" in k ? e.flags : h.call(e))).indexOf("g")) throw TypeError("`.matchAll` does not allow non-global regexes");
+                            if (A) return I.apply(a, arguments);
+                            if (void 0 === (n = e[v]) && b && "RegExp" == l(e) && (n = N), null != n) return o(n).call(e, a)
+                        } else if (A) return I.apply(a, arguments);
+                        return t = String(a), r = new RegExp(e, "g"), b ? N.call(r, t) : r[v](t)
+                    }
+                }), b || v in k || p(k, v, N)
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(210).end;
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(211)
+                }, {
+                    padEnd: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(39),
+                    a = n(145),
+                    s = n(12),
+                    i = Math.ceil,
+                    o = function (e) {
+                        return function (t, n, o) {
+                            var u, l, c = String(s(t)),
+                                h = c.length,
+                                p = void 0 === o ? " " : String(o),
+                                d = r(n);
+                            return d <= h || "" == p ? c : (u = d - h, (l = a.call(p, i(u / p.length))).length > u && (l = l.slice(0, u)), e ? c + l : l + c)
+                        }
+                    };
+                e.exports = {
+                    start: o(!1),
+                    end: o(!0)
+                }
+            }, function (e, t, n) {
+                var r = n(54);
+                e.exports = /Version\/10\.\d+(\.\d+)?( Mobile\/\w+)? Safari\//.test(r)
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(210).start;
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(211)
+                }, {
+                    padStart: function (e) {
+                        return a(this, e, arguments.length > 1 ? arguments[1] : void 0)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(9),
+                    s = n(39);
+                r({
+                    target: "String",
+                    stat: !0
+                }, {
+                    raw: function (e) {
+                        for (var t = a(e.raw), n = s(t.length), r = arguments.length, i = [], o = 0; n > o;) i.push(String(t[o++])), o < r && i.push(String(arguments[o]));
+                        return i.join("")
+                    }
+                })
+            }, function (e, t, n) {
+                n(2)({
+                    target: "String",
+                    proto: !0
+                }, {
+                    repeat: n(145)
+                })
+            }, function (e, t, n) {
+                var r = n(205),
+                    a = n(20),
+                    s = n(46),
+                    i = n(39),
+                    o = n(40),
+                    u = n(12),
+                    l = n(206),
+                    c = n(207),
+                    h = Math.max,
+                    p = Math.min,
+                    d = Math.floor,
+                    f = /\$([$&'`]|\d\d?|<[^>]*>)/g,
+                    m = /\$([$&'`]|\d\d?)/g;
+                r("replace", 2, (function (e, t, n, r) {
+                    var g = r.REGEXP_REPLACE_SUBSTITUTES_UNDEFINED_CAPTURE,
+                        y = r.REPLACE_KEEPS_$0,
+                        b = g ? "$" : "$0";
+                    return [function (n, r) {
+                        var a = u(this),
+                            s = null == n ? void 0 : n[e];
+                        return void 0 !== s ? s.call(n, a, r) : t.call(String(a), n, r)
+                    }, function (e, r) {
+                        if (!g && y || "string" == typeof r && -1 === r.indexOf(b)) {
+                            var s = n(t, e, this, r);
+                            if (s.done) return s.value
+                        }
+                        var u = a(e),
+                            d = String(this),
+                            f = "function" == typeof r;
+                        f || (r = String(r));
+                        var m = u.global;
+                        if (m) {
+                            var x = u.unicode;
+                            u.lastIndex = 0
+                        }
+                        for (var w = [];;) {
+                            var k = c(u, d);
+                            if (null === k) break;
+                            if (w.push(k), !m) break;
+                            "" === String(k[0]) && (u.lastIndex = l(d, i(u.lastIndex), x))
+                        }
+                        for (var S, I = "", A = 0, E = 0; E < w.length; E++) {
+                            k = w[E];
+                            for (var N = String(k[0]), C = h(p(o(k.index), d.length), 0), T = [], R = 1; R < k.length; R++) T.push(void 0 === (S = k[R]) ? S : String(S));
+                            var _ = k.groups;
+                            if (f) {
+                                var F = [N].concat(T, C, d);
+                                void 0 !== _ && F.push(_);
+                                var M = String(r.apply(void 0, F))
+                            } else M = v(N, d, C, T, _, r);
+                            C >= A && (I += d.slice(A, C) + M, A = C + N.length)
+                        }
+                        return I + d.slice(A)
+                    }];
+
+                    function v(e, n, r, a, i, o) {
+                        var u = r + e.length,
+                            l = a.length,
+                            c = m;
+                        return void 0 !== i && (i = s(i), c = f), t.call(o, c, (function (t, s) {
+                            var o;
+                            switch (s.charAt(0)) {
+                                case "$":
+                                    return "$";
+                                case "&":
+                                    return e;
+                                case "`":
+                                    return n.slice(0, r);
+                                case "'":
+                                    return n.slice(u);
+                                case "<":
+                                    o = i[s.slice(1, -1)];
+                                    break;
+                                default:
+                                    var c = +s;
+                                    if (0 === c) return t;
+                                    if (c > l) {
+                                        var h = d(c / 10);
+                                        return 0 === h ? t : h <= l ? void 0 === a[h - 1] ? s.charAt(1) : a[h - 1] + s.charAt(1) : t
+                                    }
+                                    o = a[c - 1]
+                            }
+                            return void 0 === o ? "" : o
+                        }))
+                    }
+                }))
+            }, function (e, t, n) {
+                var r = n(205),
+                    a = n(20),
+                    s = n(12),
+                    i = n(161),
+                    o = n(207);
+                r("search", 1, (function (e, t, n) {
+                    return [function (t) {
+                        var n = s(this),
+                            r = null == t ? void 0 : t[e];
+                        return void 0 !== r ? r.call(t, n) : new RegExp(t)[e](String(n))
+                    }, function (e) {
+                        var r = n(t, e, this);
+                        if (r.done) return r.value;
+                        var s = a(e),
+                            u = String(this),
+                            l = s.lastIndex;
+                        i(l, 0) || (s.lastIndex = 0);
+                        var c = o(s, u);
+                        return i(s.lastIndex, l) || (s.lastIndex = l), null === c ? -1 : c.index
+                    }]
+                }))
+            }, function (e, t, n) {
+                var r = n(205),
+                    a = n(186),
+                    s = n(20),
+                    i = n(12),
+                    o = n(175),
+                    u = n(206),
+                    l = n(39),
+                    c = n(207),
+                    h = n(190),
+                    p = n(6),
+                    d = [].push,
+                    f = Math.min,
+                    m = !p((function () {
+                        return !RegExp(4294967295, "y")
+                    }));
+                r("split", 2, (function (e, t, n) {
+                    var r;
+                    return r = "c" == "abbc".split(/(b)*/)[1] || 4 != "test".split(/(?:)/, -1).length || 2 != "ab".split(/(?:ab)*/).length || 4 != ".".split(/(.?)(.?)/).length || ".".split(/()()/).length > 1 || "".split(/.?/).length ? function (e, n) {
+                        var r = String(i(this)),
+                            s = void 0 === n ? 4294967295 : n >>> 0;
+                        if (0 === s) return [];
+                        if (void 0 === e) return [r];
+                        if (!a(e)) return t.call(r, e, s);
+                        for (var o, u, l, c = [], p = (e.ignoreCase ? "i" : "") + (e.multiline ? "m" : "") + (e.unicode ? "u" : "") + (e.sticky ? "y" : ""), f = 0, m = new RegExp(e.source, p + "g");
+                            (o = h.call(m, r)) && !((u = m.lastIndex) > f && (c.push(r.slice(f, o.index)), o.length > 1 && o.index < r.length && d.apply(c, o.slice(1)), l = o[0].length, f = u, c.length >= s));) m.lastIndex === o.index && m.lastIndex++;
+                        return f === r.length ? !l && m.test("") || c.push("") : c.push(r.slice(f)), c.length > s ? c.slice(0, s) : c
+                    } : "0".split(void 0, 0).length ? function (e, n) {
+                        return void 0 === e && 0 === n ? [] : t.call(this, e, n)
+                    } : t, [function (t, n) {
+                        var a = i(this),
+                            s = null == t ? void 0 : t[e];
+                        return void 0 !== s ? s.call(t, a, n) : r.call(String(a), t, n)
+                    }, function (e, a) {
+                        var i = n(r, e, this, a, r !== t);
+                        if (i.done) return i.value;
+                        var h = s(e),
+                            p = String(this),
+                            d = o(h, RegExp),
+                            g = h.unicode,
+                            y = (h.ignoreCase ? "i" : "") + (h.multiline ? "m" : "") + (h.unicode ? "u" : "") + (m ? "y" : "g"),
+                            b = new d(m ? h : "^(?:" + h.source + ")", y),
+                            v = void 0 === a ? 4294967295 : a >>> 0;
+                        if (0 === v) return [];
+                        if (0 === p.length) return null === c(b, p) ? [p] : [];
+                        for (var x = 0, w = 0, k = []; w < p.length;) {
+                            b.lastIndex = m ? w : 0;
+                            var S, I = c(b, m ? p : p.slice(w));
+                            if (null === I || (S = f(l(b.lastIndex + (m ? 0 : w)), p.length)) === x) w = u(p, w, g);
+                            else {
+                                if (k.push(p.slice(x, w)), k.length === v) return k;
+                                for (var A = 1; A <= I.length - 1; A++)
+                                    if (k.push(I[A]), k.length === v) return k;
+                                w = x = S
+                            }
+                        }
+                        return k.push(p.slice(x)), k
+                    }]
+                }), !m)
+            }, function (e, t, n) {
+                var r, a = n(2),
+                    s = n(4).f,
+                    i = n(39),
+                    o = n(199),
+                    u = n(12),
+                    l = n(200),
+                    c = n(29),
+                    h = "".startsWith,
+                    p = Math.min,
+                    d = l("startsWith");
+                a({
+                    target: "String",
+                    proto: !0,
+                    forced: !(!c && !d && (r = s(String.prototype, "startsWith"), r && !r.writable) || d)
+                }, {
+                    startsWith: function (e) {
+                        var t = String(u(this));
+                        o(e);
+                        var n = i(p(arguments.length > 1 ? arguments[1] : void 0, t.length)),
+                            r = String(e);
+                        return h ? h.call(t, r, n) : t.slice(n, n + r.length) === r
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(128).trim;
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(220)("trim")
+                }, {
+                    trim: function () {
+                        return a(this)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(6),
+                    a = n(129);
+                e.exports = function (e) {
+                    return r((function () {
+                        return !!a[e]() || "​…᠎" != "​…᠎" [e]() || a[e].name !== e
+                    }))
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(128).end,
+                    s = n(220)("trimEnd"),
+                    i = s ? function () {
+                        return a(this)
+                    } : "".trimEnd;
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: s
+                }, {
+                    trimEnd: i,
+                    trimRight: i
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(128).start,
+                    s = n(220)("trimStart"),
+                    i = s ? function () {
+                        return a(this)
+                    } : "".trimStart;
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: s
+                }, {
+                    trimStart: i,
+                    trimLeft: i
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("anchor")
+                }, {
+                    anchor: function (e) {
+                        return a(this, "a", "name", e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(12),
+                    a = /"/g;
+                e.exports = function (e, t, n, s) {
+                    var i = String(r(e)),
+                        o = "<" + t;
+                    return "" !== n && (o += " " + n + '="' + String(s).replace(a, "&quot;") + '"'), o + ">" + i + "</" + t + ">"
+                }
+            }, function (e, t, n) {
+                var r = n(6);
+                e.exports = function (e) {
+                    return r((function () {
+                        var t = "" [e]('"');
+                        return t !== t.toLowerCase() || t.split('"').length > 3
+                    }))
+                }
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("big")
+                }, {
+                    big: function () {
+                        return a(this, "big", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("blink")
+                }, {
+                    blink: function () {
+                        return a(this, "blink", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("bold")
+                }, {
+                    bold: function () {
+                        return a(this, "b", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("fixed")
+                }, {
+                    fixed: function () {
+                        return a(this, "tt", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("fontcolor")
+                }, {
+                    fontcolor: function (e) {
+                        return a(this, "font", "color", e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("fontsize")
+                }, {
+                    fontsize: function (e) {
+                        return a(this, "font", "size", e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("italics")
+                }, {
+                    italics: function () {
+                        return a(this, "i", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("link")
+                }, {
+                    link: function (e) {
+                        return a(this, "a", "href", e)
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("small")
+                }, {
+                    small: function () {
+                        return a(this, "small", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("strike")
+                }, {
+                    strike: function () {
+                        return a(this, "strike", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("sub")
+                }, {
+                    sub: function () {
+                        return a(this, "sub", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r = n(2),
+                    a = n(224);
+                r({
+                    target: "String",
+                    proto: !0,
+                    forced: n(225)("sup")
+                }, {
+                    sup: function () {
+                        return a(this, "sup", "", "")
+                    }
+                })
+            }, function (e, t, n) {
+                var r, a = n(3),
+                    s = n(126),
+                    i = n(120),
+                    o = n(119),
+                    u = n(239),
+                    l = n(14),
+                    c = n(25).enforce,
+                    h = n(26),
+                    p = !a.ActiveXObject && "ActiveXObject" in a,
+                    d = Object.isExtensible,
+                    f = function (e) {
+                        return function () {
+                            return e(this, arguments.length ? arguments[0] : void 0)
+                        }
+                    },
+                    m = e.exports = o("WeakMap", f, u);
+                if (h && p) {
+                    r = u.getConstructor(f, "WeakMap", !0), i.REQUIRED = !0;
+                    var g = m.prototype,
+                        y = g.delete,
+                        b = g.has,
+                        v = g.get,
+                        x = g.set;
+                    s(g, {
+                        delete: function (e) {
+                            if (l(e) && !d(e)) {
+                                var t = c(this);
+                                return t.frozen || (t.frozen = new r), y.call(this, e) || t.frozen.delete(e)
+                            }
+                            return y.call(this, e)
+                        },
+                        has: function (e) {
+                            if (l(e) && !d(e)) {
+                                var t = c(this);
+                                return t.frozen || (t.frozen = new r), b.call(this, e) || t.frozen.has(e)
+                            }
+                            return b.call(this, e)
+                        },
+                        get: function (e) {
+                            if (l(e) && !d(e)) {
+                                var t = c(this);
+                                return t.frozen || (t.frozen = new r), b.call(this, e) ? v.call(this, e) : t.frozen.get(e)
+                            }
+                            return v.call(this, e)
+                        },
+                        set: function (e, t) {
+                            if (l(e) && !d(e)) {
+                                var n = c(this);
+                                n.frozen || (n.frozen = new r), b.call(this, e) ? x.call(this, e, t) : n.frozen.set(e, t)
+                            } else x.call(this, e, t);
+                            return this
+                        }
+                    })
+                }
+            }, function (e, t, n) {
+                var r = n(126),
+                    a = n(120).getWeakData,
+                    s = n(20),
+                    i = n(14),
+                    o = n(123),
+                    u = n(122),
+                    l = n(63),
+                    c = n(15),
+                    h = n(25),
+                    p = h.set,
+                    d = h.getterFor,
+                    f = l.find,
+                    m = l.findIndex,
+                    g = 0,
+                    y = function (e) {
+                        return e.frozen || (e.frozen = new b)
+                    },
+                    b = function () {
+                        this.entries = []
+                    },
+                    v = function (e, t) {
+                        return f(e.entries, (function (e) {
+                            return e[0] === t
+                        }))
+                    };
+                b.prototype = {
+                    get: function (e) {
+                        var t = v(this, e);
+                        if (t) return t[1]
+                    },
+                    has: function (e) {
+                        return !!v(this, e)
+                    },
+                    set: function (e, t) {
+                        var n = v(this, e);
+                        n ? n[1] = t : this.entries.push([e, t])
+                    },
+                    delete: function (e) {
+                        var t = m(this.entries, (function (t) {
+                            return t[0] === e
+                        }));
+                        return ~t && this.entries.splice(t, 1), !!~t
+                    }
+                }, e.exports = {
+                    getConstructor: function (e, t, n, l) {
+                        var h = e((function (e, r) {
+                                o(e, h, t), p(e, {
+                                    type: t,
+                                    id: g++,
+                                    frozen: void 0
+                                }), null != r && u(r, e[l], e, n)
+                            })),
+                            f = d(t),
+                            m = function (e, t, n) {
+                                var r = f(e),
+                                    i = a(s(t), !0);
+                                return !0 === i ? y(r).set(t, n) : i[r.id] = n, e
+                            };
+                        return r(h.prototype, {
+                            delete: function (e) {
+                                var t = f(this);
+                                if (!i(e)) return !1;
+                                var n = a(e);
+                                return !0 === n ? y(t).delete(e) : n && c(n, t.id) && delete n[t.id]
+                            },
+                            has: function (e) {
+                                var t = f(this);
+                                if (!i(e)) return !1;
+                                var n = a(e);
+                                return !0 === n ? y(t).has(e) : n && c(n, t.id)
+                            }
+                        }), r(h.prototype, n ? {
+                            get: function (e) {
+                                var t = f(this);
+                                if (i(e)) {
+                                    var n = a(e);
+                                    return !0 === n ? y(t).get(e) : n ? n[t.id] : void 0
+                                }
+                            },
+                            set: function (e, t) {
+                                return m(this, e, t)
+                            }
+                        } : {
+                            add: function (e) {
+                                return m(this, e, !0)
+                            }
+                        }), h
+                    }
+                }
+            }, function (e, t, n) {
+                n(119)("WeakSet", (function (e) {
+                    return function () {
+                        return e(this, arguments.length ? arguments[0] : void 0)
+                    }
+                }), n(239))
+            }, function (e, t, n) {
+                var r = n(3),
+                    a = n(242),
+                    s = n(77),
+                    i = n(18);
+                for (var o in a) {
+                    var u = r[o],
+                        l = u && u.prototype;
+                    if (l && l.forEach !== s) try {
+                        i(l, "forEach", s)
+                    } catch (e) {
+                        l.forEach = s
+                    }
+                }
+            }, function (e, t) {
+                e.exports = {
+                    CSSRuleList: 0,
+                    CSSStyleDeclaration: 0,
+                    CSSValueList: 0,
+                    ClientRectList: 0,
+                    DOMRectList: 0,
+                    DOMStringList: 0,
+                    DOMTokenList: 1,
+                    DataTransferItemList: 0,
+                    FileList: 0,
+                    HTMLAllCollection: 0,
+                    HTMLCollection: 0,
+                    HTMLFormElement: 0,
+                    HTMLSelectElement: 0,
+                    MediaList: 0,
+                    MimeTypeArray: 0,
+                    NamedNodeMap: 0,
+                    NodeList: 1,
+                    PaintRequestList: 0,
+                    Plugin: 0,
+                    PluginArray: 0,
+                    SVGLengthList: 0,
+                    SVGNumberList: 0,
+                    SVGPathSegList: 0,
+                    SVGPointList: 0,
+                    SVGStringList: 0,
+                    SVGTransformList: 0,
+                    SourceBufferList: 0,
+                    StyleSheetList: 0,
+                    TextTrackCueList: 0,
+                    TextTrackList: 0,
+                    TouchList: 0
+                }
+            }, function (e, t, n) {
+                n(203);
+                var r, a = n(2),
+                    s = n(5),
+                    i = n(244),
+                    o = n(3),
+                    u = n(59),
+                    l = n(21),
+                    c = n(123),
+                    h = n(15),
+                    p = n(147),
+                    d = n(79),
+                    f = n(197).codeAt,
+                    m = n(245),
+                    g = n(95),
+                    y = n(246),
+                    b = n(25),
+                    v = o.URL,
+                    x = y.URLSearchParams,
+                    w = y.getState,
+                    k = b.set,
+                    S = b.getterFor("URL"),
+                    I = Math.floor,
+                    A = Math.pow,
+                    E = /[A-Za-z]/,
+                    N = /[\d+-.A-Za-z]/,
+                    C = /\d/,
+                    T = /^(0x|0X)/,
+                    R = /^[0-7]+$/,
+                    _ = /^\d+$/,
+                    F = /^[\dA-Fa-f]+$/,
+                    M = /[\u0000\u0009\u000A\u000D #%/:?@[\\]]/,
+                    D = /[\u0000\u0009\u000A\u000D #/:?@[\\]]/,
+                    O = /^[\u0000-\u001F ]+|[\u0000-\u001F ]+$/g,
+                    L = /[\u0009\u000A\u000D]/g,
+                    P = function (e, t) {
+                        var n, r, a;
+                        if ("[" == t.charAt(0)) {
+                            if ("]" != t.charAt(t.length - 1)) return "Invalid host";
+                            if (!(n = B(t.slice(1, -1)))) return "Invalid host";
+                            e.host = n
+                        } else if (q(e)) {
+                            if (t = m(t), M.test(t)) return "Invalid host";
+                            if (null === (n = $(t))) return "Invalid host";
+                            e.host = n
+                        } else {
+                            if (D.test(t)) return "Invalid host";
+                            for (n = "", r = d(t), a = 0; a < r.length; a++) n += G(r[a], U);
+                            e.host = n
+                        }
+                    },
+                    $ = function (e) {
+                        var t, n, r, a, s, i, o, u = e.split(".");
+                        if (u.length && "" == u[u.length - 1] && u.pop(), (t = u.length) > 4) return e;
+                        for (n = [], r = 0; r < t; r++) {
+                            if ("" == (a = u[r])) return e;
+                            if (s = 10, a.length > 1 && "0" == a.charAt(0) && (s = T.test(a) ? 16 : 8, a = a.slice(8 == s ? 1 : 2)), "" === a) i = 0;
+                            else {
+                                if (!(10 == s ? _ : 8 == s ? R : F).test(a)) return e;
+                                i = parseInt(a, s)
+                            }
+                            n.push(i)
+                        }
+                        for (r = 0; r < t; r++)
+                            if (i = n[r], r == t - 1) {
+                                if (i >= A(256, 5 - t)) return null
+                            } else if (i > 255) return null;
+                        for (o = n.pop(), r = 0; r < n.length; r++) o += n[r] * A(256, 3 - r);
+                        return o
+                    },
+                    B = function (e) {
+                        var t, n, r, a, s, i, o, u = [0, 0, 0, 0, 0, 0, 0, 0],
+                            l = 0,
+                            c = null,
+                            h = 0,
+                            p = function () {
+                                return e.charAt(h)
+                            };
+                        if (":" == p()) {
+                            if (":" != e.charAt(1)) return;
+                            h += 2, c = ++l
+                        }
+                        for (; p();) {
+                            if (8 == l) return;
+                            if (":" != p()) {
+                                for (t = n = 0; n < 4 && F.test(p());) t = 16 * t + parseInt(p(), 16), h++, n++;
+                                if ("." == p()) {
+                                    if (0 == n) return;
+                                    if (h -= n, l > 6) return;
+                                    for (r = 0; p();) {
+                                        if (a = null, r > 0) {
+                                            if (!("." == p() && r < 4)) return;
+                                            h++
+                                        }
+                                        if (!C.test(p())) return;
+                                        for (; C.test(p());) {
+                                            if (s = parseInt(p(), 10), null === a) a = s;
+                                            else {
+                                                if (0 == a) return;
+                                                a = 10 * a + s
+                                            }
+                                            if (a > 255) return;
+                                            h++
+                                        }
+                                        u[l] = 256 * u[l] + a, 2 != ++r && 4 != r || l++
+                                    }
+                                    if (4 != r) return;
+                                    break
+                                }
+                                if (":" == p()) {
+                                    if (h++, !p()) return
+                                } else if (p()) return;
+                                u[l++] = t
+                            } else {
+                                if (null !== c) return;
+                                h++, c = ++l
+                            }
+                        }
+                        if (null !== c)
+                            for (i = l - c, l = 7; 0 != l && i > 0;) o = u[l], u[l--] = u[c + i - 1], u[c + --i] = o;
+                        else if (8 != l) return;
+                        return u
+                    },
+                    z = function (e) {
+                        var t, n, r, a;
+                        if ("number" == typeof e) {
+                            for (t = [], n = 0; n < 4; n++) t.unshift(e % 256), e = I(e / 256);
+                            return t.join(".")
+                        }
+                        if ("object" == typeof e) {
+                            for (t = "", r = function (e) {
+                                    for (var t = null, n = 1, r = null, a = 0, s = 0; s < 8; s++) 0 !== e[s] ? (a > n && (t = r, n = a), r = null, a = 0) : (null === r && (r = s), ++a);
+                                    return a > n && (t = r, n = a), t
+                                }(e), n = 0; n < 8; n++) a && 0 === e[n] || (a && (a = !1), r === n ? (t += n ? ":" : "::", a = !0) : (t += e[n].toString(16), n < 7 && (t += ":")));
+                            return "[" + t + "]"
+                        }
+                        return e
+                    },
+                    U = {},
+                    W = p({}, U, {
+                        " ": 1,
+                        '"': 1,
+                        "<": 1,
+                        ">": 1,
+                        "`": 1
+                    }),
+                    V = p({}, W, {
+                        "#": 1,
+                        "?": 1,
+                        "{": 1,
+                        "}": 1
+                    }),
+                    j = p({}, V, {
+                        "/": 1,
+                        ":": 1,
+                        ";": 1,
+                        "=": 1,
+                        "@": 1,
+                        "[": 1,
+                        "\\": 1,
+                        "]": 1,
+                        "^": 1,
+                        "|": 1
+                    }),
+                    G = function (e, t) {
+                        var n = f(e, 0);
+                        return n > 32 && n < 127 && !h(t, e) ? e : encodeURIComponent(e)
+                    },
+                    H = {
+                        ftp: 21,
+                        file: null,
+                        http: 80,
+                        https: 443,
+                        ws: 80,
+                        wss: 443
+                    },
+                    q = function (e) {
+                        return h(H, e.scheme)
+                    },
+                    K = function (e) {
+                        return "" != e.username || "" != e.password
+                    },
+                    Y = function (e) {
+                        return !e.host || e.cannotBeABaseURL || "file" == e.scheme
+                    },
+                    X = function (e, t) {
+                        var n;
+                        return 2 == e.length && E.test(e.charAt(0)) && (":" == (n = e.charAt(1)) || !t && "|" == n)
+                    },
+                    Q = function (e) {
+                        var t;
+                        return e.length > 1 && X(e.slice(0, 2)) && (2 == e.length || "/" === (t = e.charAt(2)) || "\\" === t || "?" === t || "#" === t)
+                    },
+                    Z = function (e) {
+                        var t = e.path,
+                            n = t.length;
+                        !n || "file" == e.scheme && 1 == n && X(t[0], !0) || t.pop()
+                    },
+                    J = function (e) {
+                        return "." === e || "%2e" === e.toLowerCase()
+                    },
+                    ee = {},
+                    te = {},
+                    ne = {},
+                    re = {},
+                    ae = {},
+                    se = {},
+                    ie = {},
+                    oe = {},
+                    ue = {},
+                    le = {},
+                    ce = {},
+                    he = {},
+                    pe = {},
+                    de = {},
+                    fe = {},
+                    me = {},
+                    ge = {},
+                    ye = {},
+                    be = {},
+                    ve = {},
+                    xe = {},
+                    we = function (e, t, n, a) {
+                        var s, i, o, u, l, c = n || ee,
+                            p = 0,
+                            f = "",
+                            m = !1,
+                            g = !1,
+                            y = !1;
+                        for (n || (e.scheme = "", e.username = "", e.password = "", e.host = null, e.port = null, e.path = [], e.query = null, e.fragment = null, e.cannotBeABaseURL = !1, t = t.replace(O, "")), t = t.replace(L, ""), s = d(t); p <= s.length;) {
+                            switch (i = s[p], c) {
+                                case ee:
+                                    if (!i || !E.test(i)) {
+                                        if (n) return "Invalid scheme";
+                                        c = ne;
+                                        continue
+                                    }
+                                    f += i.toLowerCase(), c = te;
+                                    break;
+                                case te:
+                                    if (i && (N.test(i) || "+" == i || "-" == i || "." == i)) f += i.toLowerCase();
+                                    else {
+                                        if (":" != i) {
+                                            if (n) return "Invalid scheme";
+                                            f = "", c = ne, p = 0;
+                                            continue
+                                        }
+                                        if (n && (q(e) != h(H, f) || "file" == f && (K(e) || null !== e.port) || "file" == e.scheme && !e.host)) return;
+                                        if (e.scheme = f, n) return void(q(e) && H[e.scheme] == e.port && (e.port = null));
+                                        f = "", "file" == e.scheme ? c = de : q(e) && a && a.scheme == e.scheme ? c = re : q(e) ? c = oe : "/" == s[p + 1] ? (c = ae, p++) : (e.cannotBeABaseURL = !0, e.path.push(""), c = be)
+                                    }
+                                    break;
+                                case ne:
+                                    if (!a || a.cannotBeABaseURL && "#" != i) return "Invalid scheme";
+                                    if (a.cannotBeABaseURL && "#" == i) {
+                                        e.scheme = a.scheme, e.path = a.path.slice(), e.query = a.query, e.fragment = "", e.cannotBeABaseURL = !0, c = xe;
+                                        break
+                                    }
+                                    c = "file" == a.scheme ? de : se;
+                                    continue;
+                                case re:
+                                    if ("/" != i || "/" != s[p + 1]) {
+                                        c = se;
+                                        continue
+                                    }
+                                    c = ue, p++;
+                                    break;
+                                case ae:
+                                    if ("/" == i) {
+                                        c = le;
+                                        break
+                                    }
+                                    c = ye;
+                                    continue;
+                                case se:
+                                    if (e.scheme = a.scheme, i == r) e.username = a.username, e.password = a.password, e.host = a.host, e.port = a.port, e.path = a.path.slice(), e.query = a.query;
+                                    else if ("/" == i || "\\" == i && q(e)) c = ie;
+                                    else if ("?" == i) e.username = a.username, e.password = a.password, e.host = a.host, e.port = a.port, e.path = a.path.slice(), e.query = "", c = ve;
+                                    else {
+                                        if ("#" != i) {
+                                            e.username = a.username, e.password = a.password, e.host = a.host, e.port = a.port, e.path = a.path.slice(), e.path.pop(), c = ye;
+                                            continue
+                                        }
+                                        e.username = a.username, e.password = a.password, e.host = a.host, e.port = a.port, e.path = a.path.slice(), e.query = a.query, e.fragment = "", c = xe
+                                    }
+                                    break;
+                                case ie:
+                                    if (!q(e) || "/" != i && "\\" != i) {
+                                        if ("/" != i) {
+                                            e.username = a.username, e.password = a.password, e.host = a.host, e.port = a.port, c = ye;
+                                            continue
+                                        }
+                                        c = le
+                                    } else c = ue;
+                                    break;
+                                case oe:
+                                    if (c = ue, "/" != i || "/" != f.charAt(p + 1)) continue;
+                                    p++;
+                                    break;
+                                case ue:
+                                    if ("/" != i && "\\" != i) {
+                                        c = le;
+                                        continue
+                                    }
+                                    break;
+                                case le:
+                                    if ("@" == i) {
+                                        m && (f = "%40" + f), m = !0, o = d(f);
+                                        for (var b = 0; b < o.length; b++) {
+                                            var v = o[b];
+                                            if (":" != v || y) {
+                                                var x = G(v, j);
+                                                y ? e.password += x : e.username += x
+                                            } else y = !0
+                                        }
+                                        f = ""
+                                    } else if (i == r || "/" == i || "?" == i || "#" == i || "\\" == i && q(e)) {
+                                        if (m && "" == f) return "Invalid authority";
+                                        p -= d(f).length + 1, f = "", c = ce
+                                    } else f += i;
+                                    break;
+                                case ce:
+                                case he:
+                                    if (n && "file" == e.scheme) {
+                                        c = me;
+                                        continue
+                                    }
+                                    if (":" != i || g) {
+                                        if (i == r || "/" == i || "?" == i || "#" == i || "\\" == i && q(e)) {
+                                            if (q(e) && "" == f) return "Invalid host";
+                                            if (n && "" == f && (K(e) || null !== e.port)) return;
+                                            if (u = P(e, f)) return u;
+                                            if (f = "", c = ge, n) return;
+                                            continue
+                                        }
+                                        "[" == i ? g = !0 : "]" == i && (g = !1), f += i
+                                    } else {
+                                        if ("" == f) return "Invalid host";
+                                        if (u = P(e, f)) return u;
+                                        if (f = "", c = pe, n == he) return
+                                    }
+                                    break;
+                                case pe:
+                                    if (!C.test(i)) {
+                                        if (i == r || "/" == i || "?" == i || "#" == i || "\\" == i && q(e) || n) {
+                                            if ("" != f) {
+                                                var w = parseInt(f, 10);
+                                                if (w > 65535) return "Invalid port";
+                                                e.port = q(e) && w === H[e.scheme] ? null : w, f = ""
+                                            }
+                                            if (n) return;
+                                            c = ge;
+                                            continue
+                                        }
+                                        return "Invalid port"
+                                    }
+                                    f += i;
+                                    break;
+                                case de:
+                                    if (e.scheme = "file", "/" == i || "\\" == i) c = fe;
+                                    else {
+                                        if (!a || "file" != a.scheme) {
+                                            c = ye;
+                                            continue
+                                        }
+                                        if (i == r) e.host = a.host, e.path = a.path.slice(), e.query = a.query;
+                                        else if ("?" == i) e.host = a.host, e.path = a.path.slice(), e.query = "", c = ve;
+                                        else {
+                                            if ("#" != i) {
+                                                Q(s.slice(p).join("")) || (e.host = a.host, e.path = a.path.slice(), Z(e)), c = ye;
+                                                continue
+                                            }
+                                            e.host = a.host, e.path = a.path.slice(), e.query = a.query, e.fragment = "", c = xe
+                                        }
+                                    }
+                                    break;
+                                case fe:
+                                    if ("/" == i || "\\" == i) {
+                                        c = me;
+                                        break
+                                    }
+                                    a && "file" == a.scheme && !Q(s.slice(p).join("")) && (X(a.path[0], !0) ? e.path.push(a.path[0]) : e.host = a.host), c = ye;
+                                    continue;
+                                case me:
+                                    if (i == r || "/" == i || "\\" == i || "?" == i || "#" == i) {
+                                        if (!n && X(f)) c = ye;
+                                        else if ("" == f) {
+                                            if (e.host = "", n) return;
+                                            c = ge
+                                        } else {
+                                            if (u = P(e, f)) return u;
+                                            if ("localhost" == e.host && (e.host = ""), n) return;
+                                            f = "", c = ge
+                                        }
+                                        continue
+                                    }
+                                    f += i;
+                                    break;
+                                case ge:
+                                    if (q(e)) {
+                                        if (c = ye, "/" != i && "\\" != i) continue
+                                    } else if (n || "?" != i)
+                                        if (n || "#" != i) {
+                                            if (i != r && (c = ye, "/" != i)) continue
+                                        } else e.fragment = "", c = xe;
+                                    else e.query = "", c = ve;
+                                    break;
+                                case ye:
+                                    if (i == r || "/" == i || "\\" == i && q(e) || !n && ("?" == i || "#" == i)) {
+                                        if (".." === (l = (l = f).toLowerCase()) || "%2e." === l || ".%2e" === l || "%2e%2e" === l ? (Z(e), "/" == i || "\\" == i && q(e) || e.path.push("")) : J(f) ? "/" == i || "\\" == i && q(e) || e.path.push("") : ("file" == e.scheme && !e.path.length && X(f) && (e.host && (e.host = ""), f = f.charAt(0) + ":"), e.path.push(f)), f = "", "file" == e.scheme && (i == r || "?" == i || "#" == i))
+                                            for (; e.path.length > 1 && "" === e.path[0];) e.path.shift();
+                                        "?" == i ? (e.query = "", c = ve) : "#" == i && (e.fragment = "", c = xe)
+                                    } else f += G(i, V);
+                                    break;
+                                case be:
+                                    "?" == i ? (e.query = "", c = ve) : "#" == i ? (e.fragment = "", c = xe) : i != r && (e.path[0] += G(i, U));
+                                    break;
+                                case ve:
+                                    n || "#" != i ? i != r && ("'" == i && q(e) ? e.query += "%27" : e.query += "#" == i ? "%23" : G(i, U)) : (e.fragment = "", c = xe);
+                                    break;
+                                case xe:
+                                    i != r && (e.fragment += G(i, W))
+                            }
+                            p++
+                        }
+                    },
+                    ke = function (e) {
+                        var t, n, r = c(this, ke, "URL"),
+                            a = arguments.length > 1 ? arguments[1] : void 0,
+                            i = String(e),
+                            o = k(r, {
+                                type: "URL"
+                            });
+                        if (void 0 !== a)
+                            if (a instanceof ke) t = S(a);
+                            else if (n = we(t = {}, String(a))) throw TypeError(n);
+                        if (n = we(o, i, null, t)) throw TypeError(n);
+                        var u = o.searchParams = new x,
+                            l = w(u);
+                        l.updateSearchParams(o.query), l.updateURL = function () {
+                            o.query = String(u) || null
+                        }, s || (r.href = Ie.call(r), r.origin = Ae.call(r), r.protocol = Ee.call(r), r.username = Ne.call(r), r.password = Ce.call(r), r.host = Te.call(r), r.hostname = Re.call(r), r.port = _e.call(r), r.pathname = Fe.call(r), r.search = Me.call(r), r.searchParams = De.call(r), r.hash = Oe.call(r))
+                    },
+                    Se = ke.prototype,
+                    Ie = function () {
+                        var e = S(this),
+                            t = e.scheme,
+                            n = e.username,
+                            r = e.password,
+                            a = e.host,
+                            s = e.port,
+                            i = e.path,
+                            o = e.query,
+                            u = e.fragment,
+                            l = t + ":";
+                        return null !== a ? (l += "//", K(e) && (l += n + (r ? ":" + r : "") + "@"), l += z(a), null !== s && (l += ":" + s)) : "file" == t && (l += "//"), l += e.cannotBeABaseURL ? i[0] : i.length ? "/" + i.join("/") : "", null !== o && (l += "?" + o), null !== u && (l += "#" + u), l
+                    },
+                    Ae = function () {
+                        var e = S(this),
+                            t = e.scheme,
+                            n = e.port;
+                        if ("blob" == t) try {
+                            return new URL(t.path[0]).origin
+                        } catch (e) {
+                            return "null"
+                        }
+                        return "file" != t && q(e) ? t + "://" + z(e.host) + (null !== n ? ":" + n : "") : "null"
+                    },
+                    Ee = function () {
+                        return S(this).scheme + ":"
+                    },
+                    Ne = function () {
+                        return S(this).username
+                    },
+                    Ce = function () {
+                        return S(this).password
+                    },
+                    Te = function () {
+                        var e = S(this),
+                            t = e.host,
+                            n = e.port;
+                        return null === t ? "" : null === n ? z(t) : z(t) + ":" + n
+                    },
+                    Re = function () {
+                        var e = S(this).host;
+                        return null === e ? "" : z(e)
+                    },
+                    _e = function () {
+                        var e = S(this).port;
+                        return null === e ? "" : String(e)
+                    },
+                    Fe = function () {
+                        var e = S(this),
+                            t = e.path;
+                        return e.cannotBeABaseURL ? t[0] : t.length ? "/" + t.join("/") : ""
+                    },
+                    Me = function () {
+                        var e = S(this).query;
+                        return e ? "?" + e : ""
+                    },
+                    De = function () {
+                        return S(this).searchParams
+                    },
+                    Oe = function () {
+                        var e = S(this).fragment;
+                        return e ? "#" + e : ""
+                    },
+                    Le = function (e, t) {
+                        return {
+                            get: e,
+                            set: t,
+                            configurable: !0,
+                            enumerable: !0
+                        }
+                    };
+                if (s && u(Se, {
+                        href: Le(Ie, (function (e) {
+                            var t = S(this),
+                                n = String(e),
+                                r = we(t, n);
+                            if (r) throw TypeError(r);
+                            w(t.searchParams).updateSearchParams(t.query)
+                        })),
+                        origin: Le(Ae),
+                        protocol: Le(Ee, (function (e) {
+                            var t = S(this);
+                            we(t, String(e) + ":", ee)
+                        })),
+                        username: Le(Ne, (function (e) {
+                            var t = S(this),
+                                n = d(String(e));
+                            if (!Y(t)) {
+                                t.username = "";
+                                for (var r = 0; r < n.length; r++) t.username += G(n[r], j)
+                            }
+                        })),
+                        password: Le(Ce, (function (e) {
+                            var t = S(this),
+                                n = d(String(e));
+                            if (!Y(t)) {
+                                t.password = "";
+                                for (var r = 0; r < n.length; r++) t.password += G(n[r], j)
+                            }
+                        })),
+                        host: Le(Te, (function (e) {
+                            var t = S(this);
+                            t.cannotBeABaseURL || we(t, String(e), ce)
+                        })),
+                        hostname: Le(Re, (function (e) {
+                            var t = S(this);
+                            t.cannotBeABaseURL || we(t, String(e), he)
+                        })),
+                        port: Le(_e, (function (e) {
+                            var t = S(this);
+                            Y(t) || ("" == (e = String(e)) ? t.port = null : we(t, e, pe))
+                        })),
+                        pathname: Le(Fe, (function (e) {
+                            var t = S(this);
+                            t.cannotBeABaseURL || (t.path = [], we(t, e + "", ge))
+                        })),
+                        search: Le(Me, (function (e) {
+                            var t = S(this);
+                            "" == (e = String(e)) ? t.query = null: ("?" == e.charAt(0) && (e = e.slice(1)), t.query = "", we(t, e, ve)), w(t.searchParams).updateSearchParams(t.query)
+                        })),
+                        searchParams: Le(De),
+                        hash: Le(Oe, (function (e) {
+                            var t = S(this);
+                            "" != (e = String(e)) ? ("#" == e.charAt(0) && (e = e.slice(1)), t.fragment = "", we(t, e, xe)) : t.fragment = null
+                        }))
+                    }), l(Se, "toJSON", (function () {
+                        return Ie.call(this)
+                    }), {
+                        enumerable: !0
+                    }), l(Se, "toString", (function () {
+                        return Ie.call(this)
+                    }), {
+                        enumerable: !0
+                    }), v) {
+                    var Pe = v.createObjectURL,
+                        $e = v.revokeObjectURL;
+                    Pe && l(ke, "createObjectURL", (function (e) {
+                        return Pe.apply(v, arguments)
+                    })), $e && l(ke, "revokeObjectURL", (function (e) {
+                        return $e.apply(v, arguments)
+                    }))
+                }
+                g(ke, "URL"), a({
+                    global: !0,
+                    forced: !i,
+                    sham: !s
+                }, {
+                    URL: ke
+                })
+            }, function (e, t, n) {
+                var r = n(6),
+                    a = n(49),
+                    s = n(29),
+                    i = a("iterator");
+                e.exports = !r((function () {
+                    var e = new URL("b?a=1&b=2&c=3", "http://a"),
+                        t = e.searchParams,
+                        n = "";
+                    return e.pathname = "c%20d", t.forEach((function (e, r) {
+                        t.delete("b"), n += r + e
+                    })), s && !e.toJSON || !t.sort || "http://a/c%20d?a=1&c=3" !== e.href || "3" !== t.get("c") || "a=1" !== String(new URLSearchParams("?a=1")) || !t[i] || "a" !== new URL("https://a@b").username || "b" !== new URLSearchParams(new URLSearchParams("a=b")).get("a") || "xn--e1aybc" !== new URL("http://тест").host || "#%D0%B1" !== new URL("http://a#б").hash || "a1c3" !== n || "x" !== new URL("http://x", void 0).host
+                }))
+            }, function (e, t, n) {
+                var r = /[^\0-\u007E]/,
+                    a = /[.\u3002\uFF0E\uFF61]/g,
+                    s = "Overflow: input needs wider integers to process",
+                    i = Math.floor,
+                    o = String.fromCharCode,
+                    u = function (e) {
+                        return e + 22 + 75 * (e < 26)
+                    },
+                    l = function (e, t, n) {
+                        var r = 0;
+                        for (e = n ? i(e / 700) : e >> 1, e += i(e / t); e > 455; r += 36) e = i(e / 35);
+                        return i(r + 36 * e / (e + 38))
+                    },
+                    c = function (e) {
+                        var t, n, r = [],
+                            a = (e = function (e) {
+                                for (var t = [], n = 0, r = e.length; n < r;) {
+                                    var a = e.charCodeAt(n++);
+                                    if (a >= 55296 && a <= 56319 && n < r) {
+                                        var s = e.charCodeAt(n++);
+                                        56320 == (64512 & s) ? t.push(((1023 & a) << 10) + (1023 & s) + 65536) : (t.push(a), n--)
+                                    } else t.push(a)
+                                }
+                                return t
+                            }(e)).length,
+                            c = 128,
+                            h = 0,
+                            p = 72;
+                        for (t = 0; t < e.length; t++)(n = e[t]) < 128 && r.push(o(n));
+                        var d = r.length,
+                            f = d;
+                        for (d && r.push("-"); f < a;) {
+                            var m = 2147483647;
+                            for (t = 0; t < e.length; t++)(n = e[t]) >= c && n < m && (m = n);
+                            var g = f + 1;
+                            if (m - c > i((2147483647 - h) / g)) throw RangeError(s);
+                            for (h += (m - c) * g, c = m, t = 0; t < e.length; t++) {
+                                if ((n = e[t]) < c && ++h > 2147483647) throw RangeError(s);
+                                if (n == c) {
+                                    for (var y = h, b = 36;; b += 36) {
+                                        var v = b <= p ? 1 : b >= p + 26 ? 26 : b - p;
+                                        if (y < v) break;
+                                        var x = y - v,
+                                            w = 36 - v;
+                                        r.push(o(u(v + x % w))), y = i(x / w)
+                                    }
+                                    r.push(o(u(y))), p = l(h, g, f == d), h = 0, ++f
+                                }
+                            }++h, ++c
+                        }
+                        return r.join("")
+                    };
+                e.exports = function (e) {
+                    var t, n, s = [],
+                        i = e.toLowerCase().replace(a, ".").split(".");
+                    for (t = 0; t < i.length; t++) n = i[t], s.push(r.test(n) ? "xn--" + c(n) : n);
+                    return s.join(".")
+                }
+            }, function (e, t, n) {
+                n(89);
+                var r = n(2),
+                    a = n(34),
+                    s = n(244),
+                    i = n(21),
+                    o = n(126),
+                    u = n(95),
+                    l = n(91),
+                    c = n(25),
+                    h = n(123),
+                    p = n(15),
+                    d = n(64),
+                    f = n(84),
+                    m = n(20),
+                    g = n(14),
+                    y = n(58),
+                    b = n(8),
+                    v = n(247),
+                    x = n(83),
+                    w = n(49),
+                    k = a("fetch"),
+                    S = a("Headers"),
+                    I = w("iterator"),
+                    A = c.set,
+                    E = c.getterFor("URLSearchParams"),
+                    N = c.getterFor("URLSearchParamsIterator"),
+                    C = /\+/g,
+                    T = Array(4),
+                    R = function (e) {
+                        return T[e - 1] || (T[e - 1] = RegExp("((?:%[\\da-f]{2}){" + e + "})", "gi"))
+                    },
+                    _ = function (e) {
+                        try {
+                            return decodeURIComponent(e)
+                        } catch (t) {
+                            return e
+                        }
+                    },
+                    F = function (e) {
+                        var t = e.replace(C, " "),
+                            n = 4;
+                        try {
+                            return decodeURIComponent(t)
+                        } catch (e) {
+                            for (; n;) t = t.replace(R(n--), _);
+                            return t
+                        }
+                    },
+                    M = /[!'()~]|%20/g,
+                    D = {
+                        "!": "%21",
+                        "'": "%27",
+                        "(": "%28",
+                        ")": "%29",
+                        "~": "%7E",
+                        "%20": "+"
+                    },
+                    O = function (e) {
+                        return D[e]
+                    },
+                    L = function (e) {
+                        return encodeURIComponent(e).replace(M, O)
+                    },
+                    P = function (e, t) {
+                        if (t)
+                            for (var n, r, a = t.split("&"), s = 0; s < a.length;)(n = a[s++]).length && (r = n.split("="), e.push({
+                                key: F(r.shift()),
+                                value: F(r.join("="))
+                            }))
+                    },
+                    $ = function (e) {
+                        this.entries.length = 0, P(this.entries, e)
+                    },
+                    B = function (e, t) {
+                        if (e < t) throw TypeError("Not enough arguments")
+                    },
+                    z = l((function (e, t) {
+                        A(this, {
+                            type: "URLSearchParamsIterator",
+                            iterator: v(E(e).entries),
+                            kind: t
+                        })
+                    }), "Iterator", (function () {
+                        var e = N(this),
+                            t = e.kind,
+                            n = e.iterator.next(),
+                            r = n.value;
+                        return n.done || (n.value = "keys" === t ? r.key : "values" === t ? r.value : [r.key, r.value]), n
+                    })),
+                    U = function () {
+                        h(this, U, "URLSearchParams");
+                        var e, t, n, r, a, s, i, o, u, l = arguments.length > 0 ? arguments[0] : void 0,
+                            c = this,
+                            d = [];
+                        if (A(c, {
+                                type: "URLSearchParams",
+                                entries: d,
+                                updateURL: function () {},
+                                updateSearchParams: $
+                            }), void 0 !== l)
+                            if (g(l))
+                                if ("function" == typeof (e = x(l)))
+                                    for (n = (t = e.call(l)).next; !(r = n.call(t)).done;) {
+                                        if ((i = (s = (a = v(m(r.value))).next).call(a)).done || (o = s.call(a)).done || !s.call(a).done) throw TypeError("Expected sequence with length 2");
+                                        d.push({
+                                            key: i.value + "",
+                                            value: o.value + ""
+                                        })
+                                    } else
+                                        for (u in l) p(l, u) && d.push({
+                                            key: u,
+                                            value: l[u] + ""
+                                        });
+                                else P(d, "string" == typeof l ? "?" === l.charAt(0) ? l.slice(1) : l : l + "")
+                    },
+                    W = U.prototype;
+                o(W, {
+                    append: function (e, t) {
+                        B(arguments.length, 2);
+                        var n = E(this);
+                        n.entries.push({
+                            key: e + "",
+                            value: t + ""
+                        }), n.updateURL()
+                    },
+                    delete: function (e) {
+                        B(arguments.length, 1);
+                        for (var t = E(this), n = t.entries, r = e + "", a = 0; a < n.length;) n[a].key === r ? n.splice(a, 1) : a++;
+                        t.updateURL()
+                    },
+                    get: function (e) {
+                        B(arguments.length, 1);
+                        for (var t = E(this).entries, n = e + "", r = 0; r < t.length; r++)
+                            if (t[r].key === n) return t[r].value;
+                        return null
+                    },
+                    getAll: function (e) {
+                        B(arguments.length, 1);
+                        for (var t = E(this).entries, n = e + "", r = [], a = 0; a < t.length; a++) t[a].key === n && r.push(t[a].value);
+                        return r
+                    },
+                    has: function (e) {
+                        B(arguments.length, 1);
+                        for (var t = E(this).entries, n = e + "", r = 0; r < t.length;)
+                            if (t[r++].key === n) return !0;
+                        return !1
+                    },
+                    set: function (e, t) {
+                        B(arguments.length, 1);
+                        for (var n, r = E(this), a = r.entries, s = !1, i = e + "", o = t + "", u = 0; u < a.length; u++)(n = a[u]).key === i && (s ? a.splice(u--, 1) : (s = !0, n.value = o));
+                        s || a.push({
+                            key: i,
+                            value: o
+                        }), r.updateURL()
+                    },
+                    sort: function () {
+                        var e, t, n, r = E(this),
+                            a = r.entries,
+                            s = a.slice();
+                        for (a.length = 0, n = 0; n < s.length; n++) {
+                            for (e = s[n], t = 0; t < n; t++)
+                                if (a[t].key > e.key) {
+                                    a.splice(t, 0, e);
+                                    break
+                                } t === n && a.push(e)
+                        }
+                        r.updateURL()
+                    },
+                    forEach: function (e) {
+                        for (var t, n = E(this).entries, r = d(e, arguments.length > 1 ? arguments[1] : void 0, 3), a = 0; a < n.length;) r((t = n[a++]).value, t.key, this)
+                    },
+                    keys: function () {
+                        return new z(this, "keys")
+                    },
+                    values: function () {
+                        return new z(this, "values")
+                    },
+                    entries: function () {
+                        return new z(this, "entries")
+                    }
+                }, {
+                    enumerable: !0
+                }), i(W, I, W.entries), i(W, "toString", (function () {
+                    for (var e, t = E(this).entries, n = [], r = 0; r < t.length;) e = t[r++], n.push(L(e.key) + "=" + L(e.value));
+                    return n.join("&")
+                }), {
+                    enumerable: !0
+                }), u(U, "URLSearchParams"), r({
+                    global: !0,
+                    forced: !s
+                }, {
+                    URLSearchParams: U
+                }), s || "function" != typeof k || "function" != typeof S || r({
+                    global: !0,
+                    enumerable: !0,
+                    forced: !0
+                }, {
+                    fetch: function (e) {
+                        var t, n, r, a = [e];
+                        return arguments.length > 1 && (g(t = arguments[1]) && (n = t.body, "URLSearchParams" === f(n) && ((r = t.headers ? new S(t.headers) : new S).has("content-type") || r.set("content-type", "application/x-www-form-urlencoded;charset=UTF-8"), t = y(t, {
+                            body: b(0, String(n)),
+                            headers: b(0, r)
+                        }))), a.push(t)), k.apply(this, a)
+                    }
+                }), e.exports = {
+                    URLSearchParams: U,
+                    getState: E
+                }
+            }, function (e, t, n) {
+                var r = n(20),
+                    a = n(83);
+                e.exports = function (e) {
+                    var t = a(e);
+                    if ("function" != typeof t) throw TypeError(String(e) + " is not iterable");
+                    return r(t.call(e))
+                }
+            }, function (e, t, n) {
+                n(2)({
+                    target: "URL",
+                    proto: !0,
+                    enumerable: !0
+                }, {
+                    toJSON: function () {
+                        return URL.prototype.toString.call(this)
+                    }
+                })
+            }])
+        }(),
+        function (e) {
+            "use strict";
+            var t = "URLSearchParams" in self,
+                n = "Symbol" in self && "iterator" in Symbol,
+                r = "FileReader" in self && "Blob" in self && function () {
+                    try {
+                        return new Blob, !0
+                    } catch (e) {
+                        return !1
+                    }
+                }(),
+                a = "FormData" in self,
+                s = "ArrayBuffer" in self;
+            if (s) var i = ["[object Int8Array]", "[object Uint8Array]", "[object Uint8ClampedArray]", "[object Int16Array]", "[object Uint16Array]", "[object Int32Array]", "[object Uint32Array]", "[object Float32Array]", "[object Float64Array]"],
+                o = ArrayBuffer.isView || function (e) {
+                    return e && i.indexOf(Object.prototype.toString.call(e)) > -1
+                };
+
+            function u(e) {
+                if ("string" != typeof e && (e = String(e)), /[^a-z0-9\-#$%&'*+.^_`|~]/i.test(e)) throw new TypeError("Invalid character in header field name");
+                return e.toLowerCase()
+            }
+
+            function l(e) {
+                return "string" != typeof e && (e = String(e)), e
+            }
+
+            function c(e) {
+                var t = {
+                    next: function () {
+                        var t = e.shift();
+                        return {
+                            done: void 0 === t,
+                            value: t
+                        }
+                    }
+                };
+                return n && (t[Symbol.iterator] = function () {
+                    return t
+                }), t
+            }
+
+            function h(e) {
+                this.map = {}, e instanceof h ? e.forEach((function (e, t) {
+                    this.append(t, e)
+                }), this) : Array.isArray(e) ? e.forEach((function (e) {
+                    this.append(e[0], e[1])
+                }), this) : e && Object.getOwnPropertyNames(e).forEach((function (t) {
+                    this.append(t, e[t])
+                }), this)
+            }
+
+            function p(e) {
+                if (e.bodyUsed) return Promise.reject(new TypeError("Already read"));
+                e.bodyUsed = !0
+            }
+
+            function d(e) {
+                return new Promise((function (t, n) {
+                    e.onload = function () {
+                        t(e.result)
+                    }, e.onerror = function () {
+                        n(e.error)
+                    }
+                }))
+            }
+
+            function f(e) {
+                var t = new FileReader,
+                    n = d(t);
+                return t.readAsArrayBuffer(e), n
+            }
+
+            function m(e) {
+                if (e.slice) return e.slice(0);
+                var t = new Uint8Array(e.byteLength);
+                return t.set(new Uint8Array(e)), t.buffer
+            }
+
+            function g() {
+                return this.bodyUsed = !1, this._initBody = function (e) {
+                    var n;
+                    this._bodyInit = e, e ? "string" == typeof e ? this._bodyText = e : r && Blob.prototype.isPrototypeOf(e) ? this._bodyBlob = e : a && FormData.prototype.isPrototypeOf(e) ? this._bodyFormData = e : t && URLSearchParams.prototype.isPrototypeOf(e) ? this._bodyText = e.toString() : s && r && (n = e) && DataView.prototype.isPrototypeOf(n) ? (this._bodyArrayBuffer = m(e.buffer), this._bodyInit = new Blob([this._bodyArrayBuffer])) : s && (ArrayBuffer.prototype.isPrototypeOf(e) || o(e)) ? this._bodyArrayBuffer = m(e) : this._bodyText = e = Object.prototype.toString.call(e) : this._bodyText = "", this.headers.get("content-type") || ("string" == typeof e ? this.headers.set("content-type", "text/plain;charset=UTF-8") : this._bodyBlob && this._bodyBlob.type ? this.headers.set("content-type", this._bodyBlob.type) : t && URLSearchParams.prototype.isPrototypeOf(e) && this.headers.set("content-type", "application/x-www-form-urlencoded;charset=UTF-8"))
+                }, r && (this.blob = function () {
+                    var e = p(this);
+                    if (e) return e;
+                    if (this._bodyBlob) return Promise.resolve(this._bodyBlob);
+                    if (this._bodyArrayBuffer) return Promise.resolve(new Blob([this._bodyArrayBuffer]));
+                    if (this._bodyFormData) throw new Error("could not read FormData body as blob");
+                    return Promise.resolve(new Blob([this._bodyText]))
+                }, this.arrayBuffer = function () {
+                    return this._bodyArrayBuffer ? p(this) || Promise.resolve(this._bodyArrayBuffer) : this.blob().then(f)
+                }), this.text = function () {
+                    var e, t, n, r = p(this);
+                    if (r) return r;
+                    if (this._bodyBlob) return e = this._bodyBlob, n = d(t = new FileReader), t.readAsText(e), n;
+                    if (this._bodyArrayBuffer) return Promise.resolve(function (e) {
+                        for (var t = new Uint8Array(e), n = new Array(t.length), r = 0; r < t.length; r++) n[r] = String.fromCharCode(t[r]);
+                        return n.join("")
+                    }(this._bodyArrayBuffer));
+                    if (this._bodyFormData) throw new Error("could not read FormData body as text");
+                    return Promise.resolve(this._bodyText)
+                }, a && (this.formData = function () {
+                    return this.text().then(v)
+                }), this.json = function () {
+                    return this.text().then(JSON.parse)
+                }, this
+            }
+            h.prototype.append = function (e, t) {
+                e = u(e), t = l(t);
+                var n = this.map[e];
+                this.map[e] = n ? n + ", " + t : t
+            }, h.prototype.delete = function (e) {
+                delete this.map[u(e)]
+            }, h.prototype.get = function (e) {
+                return e = u(e), this.has(e) ? this.map[e] : null
+            }, h.prototype.has = function (e) {
+                return this.map.hasOwnProperty(u(e))
+            }, h.prototype.set = function (e, t) {
+                this.map[u(e)] = l(t)
+            }, h.prototype.forEach = function (e, t) {
+                for (var n in this.map) this.map.hasOwnProperty(n) && e.call(t, this.map[n], n, this)
+            }, h.prototype.keys = function () {
+                var e = [];
+                return this.forEach((function (t, n) {
+                    e.push(n)
+                })), c(e)
+            }, h.prototype.values = function () {
+                var e = [];
+                return this.forEach((function (t) {
+                    e.push(t)
+                })), c(e)
+            }, h.prototype.entries = function () {
+                var e = [];
+                return this.forEach((function (t, n) {
+                    e.push([n, t])
+                })), c(e)
+            }, n && (h.prototype[Symbol.iterator] = h.prototype.entries);
+            var y = ["DELETE", "GET", "HEAD", "OPTIONS", "POST", "PUT"];
+
+            function b(e, t) {
+                var n, r, a = (t = t || {}).body;
+                if (e instanceof b) {
+                    if (e.bodyUsed) throw new TypeError("Already read");
+                    this.url = e.url, this.credentials = e.credentials, t.headers || (this.headers = new h(e.headers)), this.method = e.method, this.mode = e.mode, this.signal = e.signal, a || null == e._bodyInit || (a = e._bodyInit, e.bodyUsed = !0)
+                } else this.url = String(e);
+                if (this.credentials = t.credentials || this.credentials || "same-origin", !t.headers && this.headers || (this.headers = new h(t.headers)), this.method = (r = (n = t.method || this.method || "GET").toUpperCase(), y.indexOf(r) > -1 ? r : n), this.mode = t.mode || this.mode || null, this.signal = t.signal || this.signal, this.referrer = null, ("GET" === this.method || "HEAD" === this.method) && a) throw new TypeError("Body not allowed for GET or HEAD requests");
+                this._initBody(a)
+            }
+
+            function v(e) {
+                var t = new FormData;
+                return e.trim().split("&").forEach((function (e) {
+                    if (e) {
+                        var n = e.split("="),
+                            r = n.shift().replace(/\+/g, " "),
+                            a = n.join("=").replace(/\+/g, " ");
+                        t.append(decodeURIComponent(r), decodeURIComponent(a))
+                    }
+                })), t
+            }
+
+            function x(e, t) {
+                t || (t = {}), this.type = "default", this.status = void 0 === t.status ? 200 : t.status, this.ok = this.status >= 200 && this.status < 300, this.statusText = "statusText" in t ? t.statusText : "OK", this.headers = new h(t.headers), this.url = t.url || "", this._initBody(e)
+            }
+            b.prototype.clone = function () {
+                return new b(this, {
+                    body: this._bodyInit
+                })
+            }, g.call(b.prototype), g.call(x.prototype), x.prototype.clone = function () {
+                return new x(this._bodyInit, {
+                    status: this.status,
+                    statusText: this.statusText,
+                    headers: new h(this.headers),
+                    url: this.url
+                })
+            }, x.error = function () {
+                var e = new x(null, {
+                    status: 0,
+                    statusText: ""
+                });
+                return e.type = "error", e
+            };
+            var w = [301, 302, 303, 307, 308];
+            x.redirect = function (e, t) {
+                if (-1 === w.indexOf(t)) throw new RangeError("Invalid status code");
+                return new x(null, {
+                    status: t,
+                    headers: {
+                        location: e
+                    }
+                })
+            }, e.DOMException = self.DOMException;
+            try {
+                new e.DOMException
+            } catch (t) {
+                e.DOMException = function (e, t) {
+                    this.message = e, this.name = t;
+                    var n = Error(e);
+                    this.stack = n.stack
+                }, e.DOMException.prototype = Object.create(Error.prototype), e.DOMException.prototype.constructor = e.DOMException
+            }
+
+            function k(t, n) {
+                return new Promise((function (a, s) {
+                    var i = new b(t, n);
+                    if (i.signal && i.signal.aborted) return s(new e.DOMException("Aborted", "AbortError"));
+                    var o = new XMLHttpRequest;
+
+                    function u() {
+                        o.abort()
+                    }
+                    o.onload = function () {
+                        var e, t, n = {
+                            status: o.status,
+                            statusText: o.statusText,
+                            headers: (e = o.getAllResponseHeaders() || "", t = new h, e.replace(/\r?\n[\t ]+/g, " ").split(/\r?\n/).forEach((function (e) {
+                                var n = e.split(":"),
+                                    r = n.shift().trim();
+                                if (r) {
+                                    var a = n.join(":").trim();
+                                    t.append(r, a)
+                                }
+                            })), t)
+                        };
+                        n.url = "responseURL" in o ? o.responseURL : n.headers.get("X-Request-URL");
+                        var r = "response" in o ? o.response : o.responseText;
+                        a(new x(r, n))
+                    }, o.onerror = function () {
+                        s(new TypeError("Network request failed"))
+                    }, o.ontimeout = function () {
+                        s(new TypeError("Network request failed"))
+                    }, o.onabort = function () {
+                        s(new e.DOMException("Aborted", "AbortError"))
+                    }, o.open(i.method, i.url, !0), "include" === i.credentials ? o.withCredentials = !0 : "omit" === i.credentials && (o.withCredentials = !1), "responseType" in o && r && (o.responseType = "blob"), i.headers.forEach((function (e, t) {
+                        o.setRequestHeader(t, e)
+                    })), i.signal && (i.signal.addEventListener("abort", u), o.onreadystatechange = function () {
+                        4 === o.readyState && i.signal.removeEventListener("abort", u)
+                    }), o.send(void 0 === i._bodyInit ? null : i._bodyInit)
+                }))
+            }
+            k.polyfill = !0, self.fetch || (self.fetch = k, self.Headers = h, self.Request = b, self.Response = x), e.Headers = h, e.Request = b, e.Response = x, e.fetch = k
+        }({})
+    }).call(this, n(36))
+}]);
\ No newline at end of file
diff --git a/public/acuant/11.9.2/face_landmark_68_tiny_model-weights_manifest.json b/public/acuant/11.9.3.508.1/face_landmark_68_tiny_model-weights_manifest.json
similarity index 100%
rename from public/acuant/11.9.2/face_landmark_68_tiny_model-weights_manifest.json
rename to public/acuant/11.9.3.508.1/face_landmark_68_tiny_model-weights_manifest.json
diff --git a/public/acuant/11.9.2/face_landmark_68_tiny_model.bin b/public/acuant/11.9.3.508.1/face_landmark_68_tiny_model.bin
similarity index 100%
rename from public/acuant/11.9.2/face_landmark_68_tiny_model.bin
rename to public/acuant/11.9.3.508.1/face_landmark_68_tiny_model.bin
diff --git a/public/acuant/11.9.2/html5-qrcode.min.js b/public/acuant/11.9.3.508.1/html5-qrcode.min.js
similarity index 100%
rename from public/acuant/11.9.2/html5-qrcode.min.js
rename to public/acuant/11.9.3.508.1/html5-qrcode.min.js
diff --git a/public/acuant/11.9.2/opencv.min.js b/public/acuant/11.9.3.508.1/opencv.min.js
similarity index 100%
rename from public/acuant/11.9.2/opencv.min.js
rename to public/acuant/11.9.3.508.1/opencv.min.js
diff --git a/public/acuant/11.9.2/tiny_face_detector_model-shard1 b/public/acuant/11.9.3.508.1/tiny_face_detector_model-shard1
similarity index 100%
rename from public/acuant/11.9.2/tiny_face_detector_model-shard1
rename to public/acuant/11.9.3.508.1/tiny_face_detector_model-shard1
diff --git a/public/acuant/11.9.2/tiny_face_detector_model-weights_manifest.json b/public/acuant/11.9.3.508.1/tiny_face_detector_model-weights_manifest.json
similarity index 100%
rename from public/acuant/11.9.2/tiny_face_detector_model-weights_manifest.json
rename to public/acuant/11.9.3.508.1/tiny_face_detector_model-weights_manifest.json