Add C# API for SenseVoice models (#1151)

k2-fsa · Jul 20, 2024 · e472180 · e472180
1 parent 25f0a10
commit e472180
Show file tree

Hide file tree

Showing 8 changed files with 79 additions and 28 deletions.
diff --git a/.github/scripts/test-dot-net.sh b/.github/scripts/test-dot-net.sh
@@ -2,16 +2,8 @@
 
 cd dotnet-examples/
 
-cd ./keyword-spotting-from-files
-./run.sh
-
-cd ../online-decode-files
-./run-transducer-itn.sh
-./run-zipformer2-ctc.sh
-./run-transducer.sh
-./run-paraformer.sh
-
-cd ../offline-decode-files
+cd ./offline-decode-files
+./run-sense-voice-ctc.sh
 ./run-paraformer-itn.sh
 ./run-telespeech-ctc.sh
 ./run-nemo-ctc.sh
@@ -21,6 +13,15 @@ cd ../offline-decode-files
 ./run-whisper.sh
 ./run-tdnn-yesno.sh
 
+cd ../keyword-spotting-from-files
+./run.sh
+
+cd ../online-decode-files
+./run-transducer-itn.sh
+./run-zipformer2-ctc.sh
+./run-transducer.sh
+./run-paraformer.sh
+
 cd ../vad-non-streaming-asr-paraformer
 ./run.sh
 

diff --git a/dotnet-examples/offline-decode-files/Program.cs b/dotnet-examples/offline-decode-files/Program.cs
@@ -61,6 +61,12 @@ class Options
     [Option("telespeech-ctc", Required = false, HelpText = "Path to model.onnx. Used only for TeleSpeech CTC models")]
     public string TeleSpeechCtc { get; set; } = "";
 
+    [Option("sense-voice-model", Required = false, HelpText = "Path to model.onnx. Used only for SenseVoice CTC models")]
+    public string SenseVoiceModel { get; set; } = "";
+
+    [Option("sense-voice-use-itn", Required = false, HelpText = "1 to use inverse text normalization for sense voice.")]
+    public int SenseVoiceUseItn { get; set; } = 1;
+
     [Option("num-threads", Required = false, Default = 1, HelpText = "Number of threads for computation")]
     public int NumThreads { get; set; } = 1;
 
@@ -225,6 +231,11 @@ private static void Run(Options options)
     {
       config.ModelConfig.Tdnn.Model = options.TdnnModel;
     }
+    else if (!String.IsNullOrEmpty(options.SenseVoiceModel))
+    {
+      config.ModelConfig.SenseVoice.Model = options.SenseVoiceModel;
+      config.ModelConfig.SenseVoice.UseInverseTextNormalization = options.SenseVoiceUseItn;
+    }
     else
     {
       Console.WriteLine("Please provide a model");

diff --git a/dotnet-examples/offline-decode-files/run-sense-voice-ctc.sh b/dotnet-examples/offline-decode-files/run-sense-voice-ctc.sh
@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -d ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17 ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
+  tar xvf sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
+  rm sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
+fi
+
+dotnet run \
+  --sense-voice-model=./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/model.int8.onnx \
+  --tokens=./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt \
+  --files ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/test_wavs/zh.wav
diff --git a/scripts/dart/release.sh b/scripts/dart/release.sh
@@ -36,7 +36,7 @@ windows_x64_wheel=$src_dir/$windows_x64_wheel_filename
 function process_linux() {
   mkdir -p t
   cd t
-  curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$linux_wheel_filename
+  curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$linux_wheel_filename
   unzip $linux_wheel_filename
   cp -v sherpa_onnx/lib/*.so* ../linux
   cd ..
@@ -50,7 +50,7 @@ function process_linux() {
 function process_windows_x64() {
   mkdir -p t
   cd t
-  curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$windows_x64_wheel_filename
+  curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$windows_x64_wheel_filename
   unzip $windows_x64_wheel_filename
   cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.dll ../windows
   cd ..
@@ -60,7 +60,7 @@ function process_windows_x64() {
 function process_macos() {
   mkdir -p t
   cd t
-  curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$macos_wheel_filename
+  curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$macos_wheel_filename
   unzip $macos_wheel_filename
   cp -v sherpa_onnx/lib/*.dylib ../macos
   cd ..

diff --git a/scripts/dotnet/OfflineModelConfig.cs b/scripts/dotnet/OfflineModelConfig.cs
@@ -23,6 +23,7 @@ public OfflineModelConfig()
             ModelingUnit = "cjkchar";
             BpeVocab = "";
             TeleSpeechCtc = "";
+            SenseVoice = new OfflineSenseVoiceModelConfig();
         }
         public OfflineTransducerModelConfig Transducer;
         public OfflineParaformerModelConfig Paraformer;
@@ -51,7 +52,7 @@ public OfflineModelConfig()
 
         [MarshalAs(UnmanagedType.LPStr)]
         public string TeleSpeechCtc;
-    }
-
 
-}
+        public OfflineSenseVoiceModelConfig SenseVoice;
+    }
+}
diff --git a/scripts/dotnet/OfflineSenseVoiceModelConfig.cs b/scripts/dotnet/OfflineSenseVoiceModelConfig.cs
@@ -0,0 +1,24 @@
+/// Copyright (c)  2024  Xiaomi Corporation (authors: Fangjun Kuang)
+
+using System.Runtime.InteropServices;
+
+namespace SherpaOnnx
+{
+    [StructLayout(LayoutKind.Sequential)]
+    public struct OfflineSenseVoiceModelConfig
+    {
+        public OfflineSenseVoiceModelConfig()
+        {
+            Model = "";
+            Language = "";
+            UseInverseTextNormalization = 0;
+        }
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string Model;
+
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string Language;
+
+        public int UseInverseTextNormalization;
+    }
+}
diff --git a/scripts/dotnet/run.sh b/scripts/dotnet/run.sh
@@ -55,7 +55,7 @@ if [ ! -f $src_dir/linux-x64/libsherpa-onnx-c-api.so ]; then
   if [ -f $linux_x64_wheel ]; then
     cp -v $linux_x64_wheel .
   else
-    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$linux_x64_wheel_filename
+    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$linux_x64_wheel_filename
   fi
   unzip $linux_x64_wheel_filename
   cp -v sherpa_onnx/lib/*.so* ../
@@ -73,7 +73,7 @@ if [ ! -f $src_dir/linux-arm64/libsherpa-onnx-c-api.so ]; then
   if [ -f $linux_arm64_wheel ]; then
     cp -v $linux_arm64_wheel .
   else
-    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$linux_arm64_wheel_filename
+    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$linux_arm64_wheel_filename
   fi
   unzip $linux_arm64_wheel_filename
   cp -v sherpa_onnx/lib/*.so* ../
@@ -91,7 +91,7 @@ if [ ! -f $src_dir/macos-x64/libsherpa-onnx-c-api.dylib ]; then
   if [ -f $macos_x64_wheel  ]; then
     cp -v $macos_x64_wheel .
   else
-    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$macos_x64_wheel_filename
+    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$macos_x64_wheel_filename
   fi
   unzip $macos_x64_wheel_filename
   cp -v sherpa_onnx/lib/*.dylib ../
@@ -111,7 +111,7 @@ if [ ! -f $src_dir/macos-arm64/libsherpa-onnx-c-api.dylib ]; then
   if [ -f $macos_arm64_wheel  ]; then
     cp -v $macos_arm64_wheel .
   else
-    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$macos_arm64_wheel_filename
+    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$macos_arm64_wheel_filename
   fi
   unzip $macos_arm64_wheel_filename
   cp -v sherpa_onnx/lib/*.dylib ../
@@ -131,7 +131,7 @@ if [ ! -f $src_dir/windows-x64/sherpa-onnx-c-api.dll ]; then
   if [ -f $windows_x64_wheel ]; then
     cp -v $windows_x64_wheel .
   else
-    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/$windows_x64_wheel_filename
+    curl -OL https://$HF_MIRROR/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/$windows_x64_wheel_filename
   fi
   unzip $windows_x64_wheel_filename
   cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.dll ../

diff --git a/scripts/go/release.sh b/scripts/go/release.sh
@@ -27,7 +27,7 @@ function linux() {
   dst=$(realpath sherpa-onnx-go-linux/lib/x86_64-unknown-linux-gnu)
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
 
   cp -v sherpa_onnx/lib/*.so* $dst
@@ -39,7 +39,7 @@ function linux() {
   dst=$(realpath sherpa-onnx-go-linux/lib/aarch64-unknown-linux-gnu)
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
 
   cp -v sherpa_onnx/lib/*.so* $dst
@@ -51,7 +51,7 @@ function linux() {
   dst=$(realpath sherpa-onnx-go-linux/lib/arm-unknown-linux-gnueabihf)
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-linux_armv7l.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-linux_armv7l.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-linux_armv7l.whl
 
   cp -v sherpa_onnx/lib/*.so* $dst
@@ -84,7 +84,7 @@ function osx() {
 
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp39-cp39-macosx_11_0_x86_64.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp39-cp39-macosx_11_0_x86_64.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp39-cp39-macosx_11_0_x86_64.whl
 
   cp -v sherpa_onnx/lib/*.dylib $dst/
@@ -102,7 +102,7 @@ function osx() {
 
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp39-cp39-macosx_11_0_arm64.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp39-cp39-macosx_11_0_arm64.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp39-cp39-macosx_11_0_arm64.whl
 
   cp -v sherpa_onnx/lib/*.dylib $dst/
@@ -137,7 +137,7 @@ function windows() {
   dst=$(realpath sherpa-onnx-go-windows/lib/x86_64-pc-windows-gnu)
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl
 
   cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.dll $dst
@@ -149,7 +149,7 @@ function windows() {
   dst=$(realpath sherpa-onnx-go-windows/lib/i686-pc-windows-gnu)
   mkdir t
   cd t
-  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win32.whl
+  wget -q https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/$SHERPA_ONNX_VERSION/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win32.whl
   unzip ./sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win32.whl
 
   cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.dll $dst