Merge remote-tracking branch 'upstream/develop' into fix-cumsum

Author: cangtianhuang

.github/workflows/Slice-baseline.yml

@@ -2,6 +2,13 @@ name: Slice-baseline-paddle
 
 on:
   workflow_dispatch:
+    inputs:
+      PR_ID:
+        required: false
+        type: string
+      COMMIT_ID:
+        required: false
+        type: string
   schedule:
     - cron: '0 20 * * 0'
 
@@ -43,3 +50,5 @@ jobs:
       docker_build_image: ${{ needs.build-docker.outputs.docker_build_image }}
       slice-check: 'true'
       SLICE_TEST_MODE: insert_baseline
+      MANUALLY_PR_ID: ${{ inputs.PR_ID }}
+      MANUALLY_COMMIT_ID: ${{ inputs.COMMIT_ID }}

.github/workflows/_Linux-XPU.yml

@@ -206,7 +206,7 @@ jobs:
           CCACHE_DIR: /root/.ccache
           CCACHE_MAXSIZE: 150G
           CCACHE_LIMIT_MULTIPLE: 0.8
-          IF_KUNLUN3: "OFF"
+          IF_KUNLUN3: "ON"
           GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           home_dir: ${{ github.workspace }}/../../../..
           FLAGS_use_stride_kernel: "0"

.github/workflows/_Slice.yml

@@ -20,6 +20,12 @@ on:
         type: string
         required: false
         default: 'paddle'
+      MANUALLY_PR_ID:
+        type: string
+        required: false
+      MANUALLY_COMMIT_ID:
+        type: string
+        required: false
 
 env:
   PR_ID: ${{ github.event.pull_request.number || '0' }}
@@ -47,6 +53,7 @@ jobs:
 
   slice:
     name: Slice test
+    needs: check-bypass
     if: ${{ inputs.can-skip != 'true' && needs.check-bypass.outputs.can-skip != 'true' }}
     runs-on:
       group: slice
@@ -105,7 +112,11 @@ jobs:
           if [[ "${{ inputs.SLICE_BENCHMARK_FRAMEWORKS }}" == "torch" ]];then
             python3.10 -m pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu118
           else
-            python3.10 -m pip install $wheel_link
+            if [[ "${{ inputs.MANUALLY_PR_ID }}" == "" ]]; then
+              python3.10 -m pip install $wheel_link
+            else
+              python3.10 -m pip install https://paddle-github-action.bj.bcebos.com/PR/build/${{ inputs.MANUALLY_PR_ID }}/${{ inputs.MANUALLY_COMMIT_ID }}/paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl
+            fi
           fi
           python3.10 -m pip install -r PaddleTest/framework/e2e/api_benchmark/requirement.txt
           cd PaddleTest/framework/slice_benchmark

ci/check_approval.sh

@@ -309,6 +309,12 @@ if [ "${ALL_PADDLE_ENFORCE}" != "" ] && [ "${PR_ID}" != "" ]; then
     check_approval 1 luotao1 zhangbo9674 phlrain
 fi
 
+CHINESE_CHECK=$(git diff -U0 upstream/$BRANCH |grep "^+" |grep -P '[\p{Han}]')
+if [ "${CHINESE_CHECK}" != "" ] && [ "${PR_ID}" != "" ]; then
+	echo_line="Not recommended to use Chinese. You must have one RD (tianshuo78520a or swgu98 or zhangbo9674 or risemeup1) approval."
+    check_approval 1 tianshuo78520a swgu98 zhangbo9674 risemeup1
+fi
+
 ALL_ADDED_LINES=$(git diff -U0 upstream/$BRANCH |grep "^+" || true)
 ALL_PADDLE_CHECK=$(echo $ALL_ADDED_LINES |grep -zoE "(PADDLE_ENFORCE[A-Z_]{0,9}|PADDLE_THROW)\(.[^,\);]*.[^;]*\);\s" || true)
 VALID_PADDLE_CHECK=$(echo "$ALL_PADDLE_CHECK" | grep -zoE '(PADDLE_ENFORCE[A-Z_]{0,9}|PADDLE_THROW)\(([^,;]+,)*[^";]*errors::.[^"]*".[^";]{20,}.[^;]*\);\s' || true)

ci/kunlun_test.sh

@@ -159,9 +159,10 @@ set +x
             git clone --depth 1000 https://gitee.com/paddlepaddle/PaddleX.git
             cd PaddleX
             pip install -e .
+            pip install numpy==1.24.4 pypdfium2
 
             #install paddle x dependency
-            paddlex --install PaddleClas
+            paddlex --install PaddleClas -y
 
             #download paddle dataset
             wget -q https://paddle-model-ecology.bj.bcebos.com/paddlex/data/cls_flowers_examples.tar -P ./dataset

cmake/cinn.cmake

@@ -26,12 +26,21 @@ else()
 endif()
 
 if(NOT DEFINED ENV{runtime_include_dir})
-  message(
-    STATUS
-      "set runtime_include_dir: ${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/cuda")
-  set(ENV{runtime_include_dir} "${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/cuda")
-  add_definitions(
-    -DRUNTIME_INCLUDE_DIR="${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/cuda")
+  if(WITH_GPU)
+    message(
+      STATUS
+        "set runtime_include_dir: ${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/cuda")
+    set(ENV{runtime_include_dir} "${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/cuda")
+    add_definitions(
+      -DRUNTIME_INCLUDE_DIR="${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/cuda")
+  elseif(WITH_ROCM)
+    message(
+      STATUS
+        "set runtime_include_dir: ${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/hip")
+    set(ENV{runtime_include_dir} "${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/hip")
+    add_definitions(
+      -DRUNTIME_INCLUDE_DIR="${CMAKE_SOURCE_DIR}/paddle/cinn/runtime/hip")
+  endif()
 endif()
 
 if(WITH_TESTING)
@@ -118,6 +127,10 @@ if(WITH_ROCM)
     add_definitions(-DCINN_WITH_HIP)
   endif()
   link_libraries(${ROCM_HIPRTC_LIB})
+
+  message(
+    STATUS "copy paddle/cinn/common/float16.h to $ENV{runtime_include_dir}")
+  file(COPY paddle/cinn/common/float16.h DESTINATION $ENV{runtime_include_dir})
 endif()
 
 set(cinnapi_src CACHE INTERNAL "" FORCE)

doc/README_cn.md

@@ -0,0 +1,6 @@
+# 致读者和开发者
+感谢您阅读 PaddlePaddle 文档。
+
+自 **2018年9月17日** 起，**0.15.0 及 develop** 分支的文档源码已迁移至 [FluidDoc Repo](https://github.com/PaddlePaddle/FluidDoc) 仓库 ，并将在该仓库中持续更新。
+
+请前往 FluidDoc 仓库获取最新文档。

paddle/cinn/backends/codegen_device_util.cc

@@ -44,13 +44,12 @@ ir::Module CreateSwitchWithBroadcastConditionModule(
       ir::Argument(kernel_args, ir::Argument::IO::kOutput),
       ir::Argument(kernel_args_num, ir::Argument::IO::kInput),
       ir::Argument(tensor_shape_args, ir::Argument::IO::kOutput)};
-
   const auto &symbolic_arg_define = [&]() -> std::vector<ir::Expr> {
     std::vector<ir::Expr> arg_defs;
     for (const auto &item : symbolic_shape_var_index) {
       ir::Expr call_get_value_in_kernel_args =
           ir::Call::Make(Int(64),
-                         runtime::intrinsic::get_value_in_cuda_kernel_args,
+                         runtime::intrinsic::get_value_in_kernel_args,
                          {kernel_args, ir::Expr(item.first)},
                          {},
                          ir::CallType::Extern,
@@ -384,7 +383,7 @@ void detail::CollectBucketStrategyHostFunctionVisitor::ProcessArgs(
     if (args[i].is_var()) {
       ir::Expr call_get_value_in_kernel_args =
           ir::Call::Make(Int(64),
-                         runtime::intrinsic::get_value_in_cuda_kernel_args,
+                         runtime::intrinsic::get_value_in_kernel_args,
                          {kernel_args_, ir::Expr(i)},
                          {},
                          ir::CallType::Extern,

paddle/cinn/backends/codegen_invoke_module.h

@@ -68,7 +68,7 @@ class CodeGenSwitchHost : public CodeGenInvokeModule {
       : CodeGenInvokeModule(m, b, vars) {}
   // only support call of args get function and inner case host function call
   llvm::Value *Visit(const ir::Call *op) override {
-    if (op->name == runtime::intrinsic::get_value_in_cuda_kernel_args) {
+    if (op->name == runtime::intrinsic::get_value_in_kernel_args) {
       return CodeGenLLVM::Visit(op);
     } else {
       return LowerInnerCaseCall(op);

paddle/cinn/backends/hip/codegen_hip_dev.cc

@@ -19,7 +19,10 @@ namespace backends {
 namespace hip {
 
 const std::string CodeGenHipDevice::source_header_ =  // NOLINT
-    R"(#include "cinn_hip_runtime_source.h"
+    R"(#define CINN_WITH_HIP
+     #include "float16.h"
+     using cinn::common::float16;
+     #include "cinn_hip_runtime_source.h"
 )";
 
 const std::string &CodeGenHipDevice::GetSourceHeader() {