Support windows (#17)

* add onnxruntime for windows

-name: test-linux-macos
+name: test-linux-macos-windows
 
 on:
   push:
     branches:
       - master
     paths:
-      - '.github/workflows/test-linux-macos.yaml'
+      - '.github/workflows/test-linux-macos-windows.yaml'
       - 'CMakeLists.txt'
       - 'cmake/**'
       - 'sherpa-onnx/csrc/*'
@@ -13,42 +13,54 @@ on:
     branches:
       - master
     paths:
-      - '.github/workflows/test-linux-macos.yaml'
+      - '.github/workflows/test-linux-macos-windows.yaml'
       - 'CMakeLists.txt'
       - 'cmake/**'
       - 'sherpa-onnx/csrc/*'
 
 concurrency:
-  group: test-linux-macos-${{ github.ref }}
+  group: test-linux-macos-windows-${{ github.ref }}
   cancel-in-progress: true
 
 permissions:
   contents: read
 
 jobs:
-  test-linux:
+  test-linux-macos-windows:
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest]
+        os: [ubuntu-latest, macos-latest, windows-latest]
 
     steps:
       - uses: actions/checkout@v2
         with:
           fetch-depth: 0
 
+      # see https://github.com/microsoft/setup-msbuild
+      - name: Add msbuild to PATH
+        if: startsWith(matrix.os, 'windows')
+        uses: microsoft/setup-msbuild@v1.0.2
+
       - name: Download pretrained model and test-data (English)
         shell: bash
         run: |
           git lfs install
-          git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
+          GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
+          cd icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
+          ls -lh exp/onnx/*.onnx
+          git lfs pull --include "exp/onnx/*.onnx"
+          ls -lh exp/onnx/*.onnx
 
       - name: Download pretrained model and test-data (Chinese)
         shell: bash
         run: |
-          git lfs install
-          git clone https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2
+          GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2
+          cd icefall_asr_wenetspeech_pruned_transducer_stateless2
+          ls -lh exp/*.onnx
+          git lfs pull --include "exp/*.onnx"
+          ls -lh exp/*.onnx
 
       - name: Configure CMake
         shell: bash
@@ -58,11 +70,21 @@ jobs:
           cmake -D CMAKE_BUILD_TYPE=Release ..
 
       - name: Build sherpa-onnx for ubuntu/macos
+        if: startsWith(matrix.os, 'ubuntu') || startsWith(matrix.os, 'macos')
+        shell: bash
         run: |
           cd build
           make VERBOSE=1 -j3
 
+      - name: Build sherpa-onnx for Windows
+        if: startsWith(matrix.os, 'windows')
+        shell: bash
+        run: |
+          cmake --build ./build --config Release
+
       - name: Run tests for ubuntu/macos (English)
+        if: startsWith(matrix.os, 'ubuntu') || startsWith(matrix.os, 'macos')
+        shell: bash
         run: |
           time ./build/bin/sherpa-onnx \
             ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500/tokens.txt \
@@ -91,7 +113,40 @@ jobs:
             ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_decoder_proj.onnx \
             ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0002.wav
 
+      - name: Run tests for Windows (English)
+        if: startsWith(matrix.os, 'windows')
+        shell: bash
+        run: |
+          ./build/bin/Release/sherpa-onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500/tokens.txt \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/encoder.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/decoder.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_encoder_proj.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_decoder_proj.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1089-134686-0001.wav
+
+          ./build/bin/Release/sherpa-onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500/tokens.txt \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/encoder.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/decoder.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_encoder_proj.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_decoder_proj.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0001.wav
+
+          ./build/bin/Release/sherpa-onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500/tokens.txt \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/encoder.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/decoder.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_encoder_proj.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/onnx/joiner_decoder_proj.onnx \
+            ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0002.wav
+
       - name: Run tests for ubuntu/macos (Chinese)
+        if: startsWith(matrix.os, 'ubuntu') || startsWith(matrix.os, 'macos')
+        shell: bash
         run: |
           time ./build/bin/sherpa-onnx \
             ./icefall_asr_wenetspeech_pruned_transducer_stateless2/data/lang_char/tokens.txt \
@@ -119,3 +174,34 @@ jobs:
             ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_encoder_proj-epoch-10-avg-2.onnx \
             ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_decoder_proj-epoch-10-avg-2.onnx \
             ./icefall_asr_wenetspeech_pruned_transducer_stateless2/test_wavs/DEV_T0000000002.wav
+
+      - name: Run tests for windows (Chinese)
+        if: startsWith(matrix.os, 'windows')
+        shell: bash
+        run: |
+          ./build/bin/Release/sherpa-onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/data/lang_char/tokens.txt \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/encoder-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/decoder-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_encoder_proj-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_decoder_proj-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/test_wavs/DEV_T0000000000.wav
+
+          ./build/bin/Release/sherpa-onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/data/lang_char/tokens.txt \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/encoder-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/decoder-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_encoder_proj-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_decoder_proj-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/test_wavs/DEV_T0000000001.wav
+
+          ./build/bin/Release/sherpa-onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/data/lang_char/tokens.txt \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/encoder-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/decoder-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_encoder_proj-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/joiner_decoder_proj-epoch-10-avg-2.onnx \
+            ./icefall_asr_wenetspeech_pruned_transducer_stateless2/test_wavs/DEV_T0000000002.wav

@@ -21,10 +21,6 @@ set(CMAKE_INSTALL_RPATH ${SHERPA_ONNX_RPATH_ORIGIN})
 set(CMAKE_BUILD_RPATH ${SHERPA_ONNX_RPATH_ORIGIN})
 
 set(BUILD_SHARED_LIBS ON)
-if(WIN32)
-  message(STATUS "Set BUILD_SHARED_LIBS to OFF for Windows")
-  set(BUILD_SHARED_LIBS OFF CACHE BOOL "" FORCE)
-endif()
 
 if(NOT CMAKE_BUILD_TYPE)
   message(STATUS "No CMAKE_BUILD_TYPE given, default to Release")

@@ -20,25 +20,37 @@ the following links:
 **HINT**: The script for exporting the Chinese model can be found at
 <https://github.com/k2-fsa/icefall/blob/master/egs/wenetspeech/ASR/pruned_transducer_stateless2/export.py>
 
-# Usage
+## Build for Linux/macOS
 
 ```bash
 git clone https://github.com/k2-fsa/sherpa-onnx
 cd sherpa-onnx
 mkdir build
 cd build
-cmake ..
+cmake -DCMAKE_BUILD_TYPE=Release ..
 make -j6
 cd ..
 ```
 
-## Download the pretrained model (English)
+## Build for Windows
 
-**Caution**: You have to run `git lfs install`. Otherwise, you will be **SAD** later.
+```bash
+git clone https://github.com/k2-fsa/sherpa-onnx
+cd sherpa-onnx
+mkdir build
+cd build
+cmake -DCMAKE_BUILD_TYPE=Release ..
+cmake --build . --config Release
+cd ..
+```
+
+## Download the pretrained model (English)
 
 ```bash
-git lfs install
-git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
+GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
+cd icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
+git lfs pull --include "exp/onnx/*.onnx"
+cd ..
 
 ./build/bin/sherpa-onnx --help
 
@@ -54,11 +66,11 @@ git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-trans
 
 ## Download the pretrained model (Chinese)
 
-**Caution**: You have to run `git lfs install`. Otherwise, you will be **SAD** later.
-
 ```bash
-git lfs install
-git clone https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2
+GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2
+cd icefall_asr_wenetspeech_pruned_transducer_stateless2
+git lfs pull --include "exp/*.onnx"
+cd ..
 
 ./build/bin/sherpa-onnx --help
 

@@ -9,8 +9,8 @@ function(download_kaldi_native_fbank)
 
   include(FetchContent)
 
-  set(kaldi_native_fbank_URL  "https://github.com/csukuangfj/kaldi-native-fbank/archive/refs/tags/v1.4.tar.gz")
-  set(kaldi_native_fbank_HASH "SHA256=771e08cb7edf512c828f4577d0d071a7993991d7e5415b11a843975dcf3e4d2d")
+  set(kaldi_native_fbank_URL  "https://github.com/csukuangfj/kaldi-native-fbank/archive/refs/tags/v1.5.tar.gz")
+  set(kaldi_native_fbank_HASH "SHA256=632c68adf8f6de831198a2a0e4c1920b31d5a1de263dcac5105be9da99f40bd5")
 
   set(KALDI_NATIVE_FBANK_BUILD_TESTS OFF CACHE BOOL "" FORCE)
   set(KALDI_NATIVE_FBANK_BUILD_PYTHON OFF CACHE BOOL "" FORCE)
@@ -36,4 +36,4 @@ function(download_kaldi_native_fbank)
   )
 endfunction()
 
-download_kaldi_native_fbank()
\ No newline at end of file
+download_kaldi_native_fbank()

@@ -31,6 +31,15 @@ function(download_onnxruntime)
     #
     # ./include
     #    It contains all the needed header files
+  elseif(WIN32)
+    set(onnxruntime_URL  "https://github.com/microsoft/onnxruntime/releases/download/v1.12.1/onnxruntime-win-x64-1.12.1.zip")
+    set(onnxruntime_HASH "SHA256=c69650ba14aeae5903b05256a82e77164fff2de992072bc695a3838c1830b85a")
+    # After downloading, it contains:
+    #  ./lib/onnxruntime.{dll,lib,pdb}
+    #  ./lib/onnxruntime_providers_shared.{dll,lib,pdb}
+    #
+    # ./include
+    #    It contains all the needed header files
   else()
     message(FATAL_ERROR "Only support Linux and macOS at present. Will support other OSes later")
   endif()
@@ -59,6 +68,17 @@ function(download_onnxruntime)
     IMPORTED_LOCATION ${location_onnxruntime}
     INTERFACE_INCLUDE_DIRECTORIES "${onnxruntime_SOURCE_DIR}/include"
   )
+  if(WIN32)
+    set_property(TARGET onnxruntime
+      PROPERTY
+        IMPORTED_IMPLIB "${onnxruntime_SOURCE_DIR}/lib/onnxruntime.lib"
+    )
+
+    file(COPY ${onnxruntime_SOURCE_DIR}/lib/onnxruntime.dll
+      DESTINATION
+        ${CMAKE_BINARY_DIR}/bin/${CMAKE_BUILD_TYPE}
+    )
+  endif()
 endfunction()
 
 download_onnxruntime()

@@ -21,8 +21,26 @@
 #include <utility>
 #include <vector>
 
+#ifdef _MSC_VER
+// For ToWide() below
+#include <codecvt>
+#include <locale>
+#endif
+
 namespace sherpa_onnx {
 
+#ifdef _MSC_VER
+// See
+// https://stackoverflow.com/questions/2573834/c-convert-string-or-char-to-wstring-or-wchar-t
+static std::wstring ToWide(const std::string &s) {
+  std::wstring_convert<std::codecvt_utf8_utf16<wchar_t>> converter;
+  return converter.from_bytes(s);
+}
+#define SHERPA_MAYBE_WIDE(s) ToWide(s)
+#else
+#define SHERPA_MAYBE_WIDE(s) s
+#endif
+
 /**
  * Get the input names of a model.
  *
@@ -85,8 +103,8 @@ RnntModel::RnntModel(const std::string &encoder_filename,
 }
 
 void RnntModel::InitEncoder(const std::string &filename) {
-  encoder_sess_ =
-      std::make_unique<Ort::Session>(env_, filename.c_str(), sess_opts_);
+  encoder_sess_ = std::make_unique<Ort::Session>(
+      env_, SHERPA_MAYBE_WIDE(filename).c_str(), sess_opts_);
   GetInputNames(encoder_sess_.get(), &encoder_input_names_,
                 &encoder_input_names_ptr_);
 
@@ -95,8 +113,8 @@ void RnntModel::InitEncoder(const std::string &filename) {
 }
 
 void RnntModel::InitDecoder(const std::string &filename) {
-  decoder_sess_ =
-      std::make_unique<Ort::Session>(env_, filename.c_str(), sess_opts_);
+  decoder_sess_ = std::make_unique<Ort::Session>(
+      env_, SHERPA_MAYBE_WIDE(filename).c_str(), sess_opts_);
 
   GetInputNames(decoder_sess_.get(), &decoder_input_names_,
                 &decoder_input_names_ptr_);
@@ -106,8 +124,8 @@ void RnntModel::InitDecoder(const std::string &filename) {
 }
 
 void RnntModel::InitJoiner(const std::string &filename) {
-  joiner_sess_ =
-      std::make_unique<Ort::Session>(env_, filename.c_str(), sess_opts_);
+  joiner_sess_ = std::make_unique<Ort::Session>(
+      env_, SHERPA_MAYBE_WIDE(filename).c_str(), sess_opts_);
 
   GetInputNames(joiner_sess_.get(), &joiner_input_names_,
                 &joiner_input_names_ptr_);
@@ -117,8 +135,8 @@ void RnntModel::InitJoiner(const std::string &filename) {
 }
 
 void RnntModel::InitJoinerEncoderProj(const std::string &filename) {
-  joiner_encoder_proj_sess_ =
-      std::make_unique<Ort::Session>(env_, filename.c_str(), sess_opts_);
+  joiner_encoder_proj_sess_ = std::make_unique<Ort::Session>(
+      env_, SHERPA_MAYBE_WIDE(filename).c_str(), sess_opts_);
 
   GetInputNames(joiner_encoder_proj_sess_.get(),
                 &joiner_encoder_proj_input_names_,
@@ -130,8 +148,8 @@ void RnntModel::InitJoinerEncoderProj(const std::string &filename) {
 }
 
 void RnntModel::InitJoinerDecoderProj(const std::string &filename) {
-  joiner_decoder_proj_sess_ =
-      std::make_unique<Ort::Session>(env_, filename.c_str(), sess_opts_);
+  joiner_decoder_proj_sess_ = std::make_unique<Ort::Session>(
+      env_, SHERPA_MAYBE_WIDE(filename).c_str(), sess_opts_);
 
   GetInputNames(joiner_decoder_proj_sess_.get(),
                 &joiner_decoder_proj_input_names_,