ggml : try to improve threading

.github/workflows/bindings-go.yml

@@ -1,22 +0,0 @@
-name: Bindings Tests (Go)
-on:
-  push:
-    paths:
-      - bindings/go/**
-      - whisper.h
-  pull_request:
-    paths:
-      - bindings/go/**
-      - whisper.h
-
-jobs:
-  ubuntu-latest:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/setup-go@v3
-        with:
-          go-version: '^1.19'
-      - uses: actions/checkout@v1
-      - run: |
-          cd bindings/go
-          make test

.github/workflows/bindings-ruby.yml

@@ -1,22 +0,0 @@
-name: Bindings Tests (Ruby)
-on:
-  push:
-    paths:
-      - bindings/ruby/**
-      - whisper.h
-  pull_request:
-    paths:
-      - bindings/ruby/**
-      - whisper.h
-
-jobs:
-  ubuntu-latest:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: ruby/setup-ruby@v1
-        with:
-          ruby-version: '3.0'
-      - uses: actions/checkout@v1
-      - run: |
-          cd bindings/ruby/ext
-          ruby extconf.rb && make

.github/workflows/build.yml

@@ -1,41 +1,31 @@
 name: CI
-on: [push, pull_request]
-
-env:
-  ubuntu_image: "ubuntu:22.04"
+on: [push]
 
 jobs:
     ubuntu-latest:
         runs-on: ubuntu-latest
 
-    strategy:
-      fail-fast: false
-      matrix:
-        arch: [linux/amd64, linux/arm64, linux/arm/v7, linux/ppc64le]
-
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-
-      - name: Build ${{ matrix.arch }}
+            - name: Dependencies
+              run: |
+                  sudo apt-get update
+                  sudo apt-get install build-essential
+                  sudo apt-get install libsdl2-dev
+
+            - name: Build
               run: |
-          docker run --platform ${{ matrix.arch }} --rm \
-            -v ${{ github.workspace }}:/workspace \
-            -w /workspace ${{ env.ubuntu_image }} /bin/sh -c '
-            apt update
-            apt install -y build-essential libsdl2-dev
                 make
-            make stream'
+                make stream
 
     macOS-latest:
         runs-on: macOS-latest
 
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
             - name: Dependencies
               run: |
@@ -47,104 +37,82 @@ jobs:
                 make
                 make stream
 
-  freeBSD-latest:
-    runs-on: macos-12
-
-    steps:
-      - name: Clone
-        uses: actions/checkout@v3
-
-      - name: Build
-        uses: cross-platform-actions/action@v0.15.0
-        with:
-          operating_system: freebsd
-          version: '13.2'
-          run: |
-            sudo pkg update
-            sudo pkg install -y gmake sdl2
-            gmake
-            gmake stream
-
     ubuntu-latest-gcc:
         runs-on: ubuntu-latest
 
         strategy:
-      fail-fast: false
             matrix:
                 build: [Debug, Release]
-        arch: [linux/amd64, linux/arm64, linux/arm/v7, linux/ppc64le]
 
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-
-      - name: Build ${{ matrix.arch }}
+            - name: Dependencies
+              run: |
+                  sudo apt-get update
+                  sudo apt-get install build-essential
+                  sudo apt-get install cmake
+                  sudo apt-get install libsdl2-dev
+
+            - name: Configure
+              run: cmake . -DWHISPER_SUPPORT_SDL2=ON -DCMAKE_BUILD_TYPE=${{ matrix.build }}
+
+            - name: Build
               run: |
-          docker run --platform ${{ matrix.arch }} --rm \
-            -v ${{ github.workspace }}:/workspace \
-            -w /workspace ${{ env.ubuntu_image }} /bin/sh -c '
-            apt update
-            apt install -y build-essential cmake libsdl2-dev
-            cmake . -DWHISPER_SUPPORT_SDL2=ON -DCMAKE_BUILD_TYPE=${{ matrix.build }}
                 make
-            ctest -L gh --output-on-failure'
+                ctest -L gh --output-on-failure
 
     ubuntu-latest-clang:
         runs-on: ubuntu-latest
 
         strategy:
-      fail-fast: false
             matrix:
                 build: [Debug, Release]
-        arch: [linux/amd64, linux/arm64, linux/arm/v7, linux/ppc64le]
 
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-
-      - name: Build ${{ matrix.arch }}
+            - name: Dependencies
+              run: |
+                  sudo apt-get update
+                  sudo apt-get install build-essential
+                  sudo apt-get install cmake
+                  sudo apt-get install libsdl2-dev
+
+            - name: Configure
+              run: cmake . -DWHISPER_SUPPORT_SDL2=ON -DCMAKE_BUILD_TYPE=${{ matrix.build }} -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_COMPILER=clang
+
+            - name: Build
               run: |
-          docker run --platform ${{ matrix.arch }} --rm \
-            -v ${{ github.workspace }}:/workspace \
-            -w /workspace ${{ env.ubuntu_image }} /bin/sh -c '
-            apt update
-            apt install -y build-essential cmake libsdl2-dev
-            cmake . -DWHISPER_SUPPORT_SDL2=ON -DCMAKE_BUILD_TYPE=${{ matrix.build }} -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_COMPILER=clang
                 make
-            ctest -L gh --output-on-failure'
+                ctest -L gh --output-on-failure
 
     ubuntu-latest-gcc-sanitized:
         runs-on: ubuntu-latest
 
         strategy:
-      fail-fast: false
             matrix:
                 sanitizer: [ADDRESS, THREAD, UNDEFINED]
-        arch: [linux/amd64]
 
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-
-      - name: Build ${{ matrix.arch }}
+            - name: Dependencies
+              run: |
+                  sudo apt-get update
+                  sudo apt-get install build-essential
+                  sudo apt-get install cmake
+
+            - name: Configure
+              run: cmake . -DCMAKE_BUILD_TYPE=Debug -DWHISPER_SANITIZE_${{ matrix.sanitizer }}=ON
+
+            - name: Build
               run: |
-          docker run --platform ${{ matrix.arch }} --rm \
-            -v ${{ github.workspace }}:/workspace \
-            -w /workspace ${{ env.ubuntu_image }} /bin/sh -c '
-            apt update
-            apt install -y build-essential cmake
-            cmake . -DCMAKE_BUILD_TYPE=Debug -DWHISPER_SANITIZE_${{ matrix.sanitizer }}=ON
                 make
-            ctest -L gh --output-on-failure'
+                ctest -L gh --output-on-failure
 
     windows:
         runs-on: windows-latest
@@ -157,16 +125,14 @@ jobs:
                 include:
                   - arch: Win32
                     s2arc: x86
-            jnaPath: win32-x86
                   - arch: x64
                     s2arc: x64
-            jnaPath: win32-x86-64
                   - sdl2: ON
                     s2ver: 2.26.0
 
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
             - name: Add msbuild to PATH
               uses: microsoft/setup-msbuild@v1
@@ -193,12 +159,6 @@ jobs:
               if: matrix.sdl2 == 'ON'
               run: copy "$env:SDL2_DIR/../lib/${{ matrix.s2arc }}/SDL2.dll" build/bin/${{ matrix.build }}
 
-      - name: Upload dll
-        uses: actions/upload-artifact@v3
-        with:
-          name: ${{ matrix.jnaPath }}_whisper.dll
-          path: build/bin/${{ matrix.build }}/whisper.dll
-
             - name: Upload binaries
               if: matrix.sdl2 == 'ON'
               uses: actions/upload-artifact@v1
@@ -227,7 +187,7 @@ jobs:
 
         steps:
             - name: Clone
-        uses: actions/checkout@v3
+              uses: actions/checkout@v1
 
             - name: Add msbuild to PATH
               uses: microsoft/setup-msbuild@v1
@@ -275,178 +235,3 @@ jobs:
               with:
                 name: whisper-blas-bin-${{ matrix.arch }}
                 path: build/bin/${{ matrix.build }}
-
-  windows-cublas:
-    runs-on: windows-latest
-
-    strategy:
-      matrix:
-        build: [Release]
-        arch: [x64]
-        cublas: [ON]
-        sdl2: [ON]
-        include:
-          - arch: x64
-            s2arc: x64
-          - sdl2: ON
-            s2ver: 2.26.0
-
-    steps:
-      - name: Clone
-        uses: actions/checkout@v3
-
-      - name: Add msbuild to PATH
-        uses: microsoft/setup-msbuild@v1
-
-      - name: Install CUDA Toolkit
-        id: cuda-toolkit
-        uses: Jimver/cuda-toolkit@v0.2.10
-
-      - name: Fetch SDL2 and set SDL2_DIR
-        if: matrix.sdl2 == 'ON'
-        run: |
-          C:/msys64/usr/bin/wget.exe -qO sdl2.zip https://github.com/libsdl-org/SDL/releases/download/release-${{ matrix.s2ver }}/SDL2-devel-${{ matrix.s2ver }}-VC.zip
-          7z x sdl2.zip
-          echo "SDL2_DIR=$env:GITHUB_WORKSPACE/SDL2-${{ matrix.s2ver }}/cmake" >> $env:GITHUB_ENV
-
-      - name: Configure
-        run: >
-          cmake -S . -B ./build -A ${{ matrix.arch }}
-          -DCMAKE_BUILD_TYPE=${{ matrix.build }}
-          -DWHISPER_CUBLAS=1
-
-      - name: Build
-        run: |
-          cd ./build
-          msbuild ALL_BUILD.vcxproj -t:build -p:configuration=${{ matrix.build }} -p:platform=${{ matrix.arch }}
-
-      - name: Copy SDL2.dll
-        if: matrix.sdl2 == 'ON'
-        run: copy "$env:SDL2_DIR/../lib/${{ matrix.s2arc }}/SDL2.dll" build/bin/${{ matrix.build }}
-
-      - name: Upload binaries
-        if: matrix.sdl2 == 'ON'
-        uses: actions/upload-artifact@v1
-        with:
-          name: whisper-cublas-bin-${{ matrix.arch }}
-          path: build/bin/${{ matrix.build }}
-
-  emscripten:
-    runs-on: ubuntu-latest
-
-    strategy:
-      matrix:
-        build: [Release]
-
-    steps:
-      - name: Clone
-        uses: actions/checkout@v3
-
-      - name: Setup emsdk
-        uses: mymindstorm/setup-emsdk@v12
-
-      - name: Verify
-        run: emcc -v
-
-      - name: Build
-        run: |
-          emcmake cmake . -DCMAKE_BUILD_TYPE=${{ matrix.build }}
-          make
-
-  ios:
-    runs-on: macos-latest
-
-    strategy:
-      matrix:
-        build: [Release]
-
-    steps:
-      - name: Clone
-        uses: actions/checkout@v3
-
-      - name: Configure
-        run: |
-          cp models/for-tests-ggml-base.en.bin models/ggml-base.en.bin
-          mkdir models/ggml-base.en-encoder.mlmodelc
-
-      - name: Build objc example
-        run: xcodebuild -project examples/whisper.objc/whisper.objc.xcodeproj -scheme whisper.objc -configuration ${{ matrix.build }} -sdk iphonesimulator build
-
-      - name: Build swiftui example
-        run: xcodebuild -project examples/whisper.swiftui/whisper.swiftui.xcodeproj -scheme WhisperCppDemo -configuration ${{ matrix.build }} -sdk iphonesimulator build
-
-  android:
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Clone
-        uses: actions/checkout@v3
-
-      - name: Install Java
-        uses: actions/setup-java@v3
-        with:
-          distribution: zulu
-          java-version: 17
-
-      - name: Setup Android SDK
-        uses: android-actions/setup-android@v2
-
-      - name: Build
-        run: |
-          cd examples/whisper.android
-          ./gradlew assembleRelease --no-daemon
-
-  java:
-    needs: [ 'windows' ]
-    runs-on: windows-latest
-    steps:
-      - uses: actions/checkout@v3
-
-      - name: Install Java
-        uses: actions/setup-java@v1
-        with:
-          java-version: 17
-
-      - name: Download Windows lib
-        uses: actions/download-artifact@v3
-        with:
-          name: win32-x86-64_whisper.dll
-          path: bindings/java/build/generated/resources/main/win32-x86-64
-
-      - name: Build
-        run: |
-          models\download-ggml-model.cmd tiny.en
-          cd bindings/java
-          chmod +x ./gradlew
-          ./gradlew build
-
-      - name: Upload jar
-        uses: actions/upload-artifact@v3
-        with:
-          name: whispercpp.jar
-          path: bindings/java/build/libs/whispercpp-*.jar
-
-      - name: Publish package
-        if: ${{ github.ref == 'refs/heads/master' }}
-        uses: gradle/gradle-build-action@v2
-        with:
-          arguments: publish
-          build-root-directory: bindings/java
-        env:
-          MAVEN_USERNAME: ${{ secrets.JIRA_USER }}
-          MAVEN_PASSWORD: ${{ secrets.JIRA_PASS }}
-          # MAVEN_USERNAME: ${{ secrets.OSSRH_USERNAME }}
-          # MAVEN_PASSWORD: ${{ secrets.OSSRH_TOKEN }}
-
-  quantize:
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Clone
-        uses: actions/checkout@v3
-
-      - name: Test quantize
-        run: |
-          ./models/download-ggml-model.sh tiny.en
-          make quantize
-          ./quantize models/ggml-tiny.en.bin models/ggml-tiny.en-q4_0.bin q4_0

.github/workflows/examples.yml

@@ -1,48 +0,0 @@
-name: Examples Tests
-on:
-  push:
-    paths:
-      - examples/addon.node/**
-      - whisper.h
-  pull_request:
-    paths:
-      - examples/addon.node/**
-      - whisper.h
-
-jobs:
-  addon_node-ubuntu-latest:
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        node-version: [ 16.x, 18.x ]
-    steps:
-      - name: Clone
-        uses: actions/checkout@v1
-
-      - name: Dependencies
-        run: |
-          sudo apt-get update
-          sudo apt-get install build-essential
-          sudo apt-get install cmake
-          sudo apt-get install libsdl2-dev
-
-      - name: Use Node.js ${{ matrix.node-version }}
-        uses: actions/setup-node@v1
-        with:
-          node-version: ${{ matrix.node-version }}
-          cache: 'npm'
-
-      - name: Install package.json dependencies
-        working-directory: ./examples/addon.node
-        run: npm install
-
-      - name: Compile addon.node
-        run: npx cmake-js compile -T whisper-addon -B Release
-
-      - name: Download test model
-        run: |
-          bash ./models/download-ggml-model.sh base.en
-      - name: Test
-        run: |
-          cd examples/addon.node
-          npm run test

.gitignore

@@ -1,8 +1,5 @@
 *.o
-*.a
 .cache/
-.coreml/
-.test/
 .vs/
 .vscode/
 .DS_Store
@@ -11,10 +8,6 @@ build/
 build-em/
 build-debug/
 build-release/
-build-rwdi/
-build-static/
-build-cublas/
-build-no-accel/
 build-sanitize-addr/
 build-sanitize-thread/
 
@@ -22,14 +15,9 @@ build-sanitize-thread/
 /stream
 /command
 /talk
-/talk-llama
 /bench
-/quantize
-/lsp
 
-arm_neon.h
 sync.sh
-libwhisper.a
 libwhisper.so
 compile_commands.json
 
@@ -39,10 +27,3 @@ examples/whisper.objc/whisper.objc.xcodeproj/xcuserdata/
 examples/whisper.objc/whisper.objc.xcodeproj/project.xcworkspace/xcuserdata
 
 extra/bench-gg.txt
-
-models/*.mlmodel
-models/*.mlmodelc
-models/*.mlpackage
-bindings/java/.gradle/
-bindings/java/.idea/
-.idea/

CMakeLists.txt

@@ -1,16 +1,15 @@
 cmake_minimum_required (VERSION 3.0)
 
-project(whisper.cpp VERSION 1.4.2)
-
-# Add path to modules
-list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")
+project(whisper.cpp VERSION 1.0.4)
 
+set(CMAKE_EXPORT_COMPILE_COMMANDS "on")
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib")
 
 if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
     set(WHISPER_STANDALONE ON)
-    include(GitVars)
-    include(BuildTypes)
+    include(cmake/GitVars.cmake)
+    include(cmake/BuildTypes.cmake)
 
     # configure project version
     if (EXISTS "${CMAKE_SOURCE_DIR}/bindings/ios/Makefile-tmpl")
@@ -47,26 +46,14 @@ option(WHISPER_SANITIZE_UNDEFINED     "whisper: enable undefined sanitizer" OFF)
 option(WHISPER_BUILD_TESTS             "whisper: build tests"    ${WHISPER_STANDALONE})
 option(WHISPER_BUILD_EXAMPLES          "whisper: build examples" ${WHISPER_STANDALONE})
 
-option(WHISPER_SDL2                   "whisper: support for libSDL2" OFF)
-
-option(WHISPER_NO_AVX                 "whisper: disable AVX"  OFF)
-option(WHISPER_NO_AVX2                "whisper: disable AVX2" OFF)
-option(WHISPER_NO_FMA                 "whisper: disable FMA"  OFF)
-option(WHISPER_NO_F16C                "whisper: disable F16c" OFF)
-
-option(WHISPER_OPENVINO               "whisper: support for OpenVINO" OFF)
+option(WHISPER_SUPPORT_SDL2            "whisper: support for libSDL2" OFF)
 
 if (APPLE)
     option(WHISPER_NO_ACCELERATE       "whisper: disable Accelerate framework" OFF)
-    option(WHISPER_COREML                "whisper: enable Core ML framework"     OFF)
-    option(WHISPER_COREML_ALLOW_FALLBACK "whisper: allow non-CoreML fallback"    OFF)
+    option(WHISPER_NO_AVX              "whisper: disable AVX" OFF)
+    option(WHISPER_NO_AVX2             "whisper: disable AVX2" OFF)
 else()
-    option(WHISPER_BLAS                  "whisper: use BLAS libraries"  OFF)
-    option(WHISPER_BLAS_VENDOR           "whisper: BLAS library vendor" Generic)
-    option(WHISPER_OPENBLAS              "whisper: prefer OpenBLAS"     OFF)
-    option(WHISPER_CUBLAS                "whisper: support for cuBLAS"  OFF)
-    option(WHISPER_HIPBLAS               "whisper: support for hipBLAS" OFF)
-    option(WHISPER_CLBLAST               "whisper: use CLBlast"         OFF)
+    option(WHISPER_SUPPORT_OPENBLAS    "whisper: support for OpenBLAS" OFF)
 endif()
 
 option(WHISPER_PERF                    "whisper: enable perf timings" OFF)
@@ -95,14 +82,14 @@ endif()
 
 # dependencies
 
+set(CMAKE_C_STANDARD   11)
+set(CMAKE_CXX_STANDARD 11)
+
 find_package(Threads REQUIRED)
 
-# on APPLE
-if (APPLE)
-    # include Accelerate framework
-    if (NOT WHISPER_NO_ACCELERATE)
+# on APPLE - include Accelerate framework
+if (APPLE AND NOT WHISPER_NO_ACCELERATE)
     find_library(ACCELERATE_FRAMEWORK Accelerate)
-
     if (ACCELERATE_FRAMEWORK)
         message(STATUS "Accelerate framework found")
 
@@ -111,135 +98,20 @@ if (APPLE)
     else()
         message(WARNING "Accelerate framework not found")
     endif()
-    endif()
-
-    if (WHISPER_COREML)
-        find_library(FOUNDATION_FRAMEWORK Foundation)
-        find_library(COREML_FRAMEWORK CoreML)
-
-        if (COREML_FRAMEWORK)
-            message(STATUS "CoreML framework found")
-
-            set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DWHISPER_USE_COREML)
-        else()
-            message(WARNING "CoreML framework not found")
-        endif()
-
-        if (WHISPER_COREML_ALLOW_FALLBACK)
-            set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DWHISPER_COREML_ALLOW_FALLBACK)
-        endif()
-    endif()
 endif()
 
-if (WHISPER_OPENBLAS)
-    set(WHISPER_BLAS_VENDOR "OpenBLAS")
-    set(WHISPER_BLAS ON)
-endif()
+if (WHISPER_SUPPORT_OPENBLAS)
+    find_library(OPENBLAS_LIB
+        NAMES openblas libopenblas
+        )
+    if (OPENBLAS_LIB)
+        message(STATUS "OpenBLAS found")
 
-if (WHISPER_BLAS)
-    if (WIN32)
-        if(DEFINED ENV{OPENBLAS_PATH})
-            set(BLAS_LIBRARIES $ENV{OPENBLAS_PATH}/lib/libopenblas.dll.a)
-            message(STATUS "Libraries ${BLAS_LIBRARIES}")
+        set(WHISPER_EXTRA_LIBS  ${WHISPER_EXTRA_LIBS}  ${OPENBLAS_LIB})
         set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_USE_OPENBLAS)
-            include_directories($ENV{OPENBLAS_PATH}/include)
-            set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} ${BLAS_LIBRARIES})
-        else ()
-            message(WARNING "BLAS library was not found. Environment variable OPENBLAS_PATH not defined.")
-        endif ()
-    else ()
-        set(BLA_STATIC 1)
-        set(BLA_VENDOR ${WHISPER_BLAS_VENDOR})
-        #    set(BLA_PREFER_PKGCONFIG 1)
-        set(BLA_SIZEOF_INTEGER 8)
-        find_package(BLAS)
-
-        if(BLAS_FOUND)
-            message(STATUS "BLAS compatible library found")
-            message(STATUS "Libraries ${BLAS_LIBRARIES}")
-            find_path(BLAS_INCLUDE_DIRS cblas.h /usr/include/openblas /usr/local/include/openblas $ENV{BLAS_HOME}/include)
-            set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_USE_OPENBLAS)
-            include_directories(${BLAS_INCLUDE_DIRS})
-            set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} ${BLAS_LIBRARIES})
     else()
-            message(WARNING "BLAS library was not found")
+        message(WARNING "OpenBLAS not found")
     endif()
-    endif ()
-endif ()
-
-if (WHISPER_CUBLAS)
-    cmake_minimum_required(VERSION 3.17)
-
-    find_package(CUDAToolkit)
-
-    if (CUDAToolkit_FOUND)
-        message(STATUS "cuBLAS found")
-
-        enable_language(CUDA)
-
-        set(GGML_CUDA_SOURCES ggml-cuda.cu ggml-cuda.h)
-
-        add_compile_definitions(GGML_USE_CUBLAS)
-
-        if (WHISPER_STATIC)
-            set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} CUDA::cudart_static CUDA::cublas_static CUDA::cublasLt_static)
-        else()
-            set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} CUDA::cudart CUDA::cublas CUDA::cublasLt)
-        endif()
-
-    else()
-        message(WARNING "cuBLAS not found")
-    endif()
-endif()
-
-
-if (WHISPER_HIPBLAS)
-    list(APPEND CMAKE_PREFIX_PATH /opt/rocm)
-    if (NOT ${CMAKE_C_COMPILER_ID} MATCHES "Clang")
-        message(WARNING "Only LLVM is supported for HIP, hint: CC=/opt/rocm/llvm/bin/clang")
-    endif()
-    if (NOT ${CMAKE_CXX_COMPILER_ID} MATCHES "Clang")
-        message(WARNING "Only LLVM is supported for HIP, hint: CXX=/opt/rocm/llvm/bin/clang++")
-    endif()
-
-    find_package(hip)
-    find_package(hipblas)
-    find_package(rocblas)
-
-    if (${hipblas_FOUND} AND ${hip_FOUND})
-        message(STATUS "HIP and hipBLAS found")
-        add_compile_definitions(GGML_USE_HIPBLAS GGML_USE_CUBLAS)
-        add_library(ggml-rocm OBJECT ggml-cuda.cu ggml-cuda.h)
-        set_property(TARGET ggml-rocm PROPERTY POSITION_INDEPENDENT_CODE ON)
-        set_source_files_properties(ggml-cuda.cu PROPERTIES LANGUAGE CXX)
-        target_link_libraries(ggml-rocm PRIVATE hip::device PUBLIC hip::host roc::rocblas roc::hipblas)
-
-        if (WHISPER_STATIC)
-            message(FATAL_ERROR "Static linking not supported for HIP/ROCm")
-        endif()
-        set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} ggml-rocm)
-    else()
-        message(WARNING "hipBLAS or HIP not found. Try setting CMAKE_PREFIX_PATH=/opt/rocm")
-    endif()
-endif()
-
-if (WHISPER_CLBLAST)
-    find_package(CLBlast)
-    if (CLBlast_FOUND)
-        message(STATUS "CLBlast found")
-
-        set(GGML_OPENCL_SOURCES ggml-opencl.cpp ggml-opencl.h)
-
-        add_compile_definitions(GGML_USE_CLBLAST)
-
-        set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} clblast)
-    else()
-        message(WARNING "CLBlast not found")
-    endif()
-endif()
-
-if( WHISPER_OPENVINO )
-    find_package(OpenVINO REQUIRED COMPONENTS Runtime)
 endif()
 
 # compiler flags
@@ -259,7 +131,6 @@ if (WHISPER_ALL_WARNINGS)
             -Wcast-qual                     \
             -Wstrict-prototypes             \
             -Wpointer-arith                 \
-            -Wno-unused-function            \
         ")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} \
             -Wall                           \
@@ -281,25 +152,11 @@ message(STATUS "CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
 
 if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
     message(STATUS "ARM detected")
-elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64le")
-    message(STATUS "PowerPC detected")
 else()
     message(STATUS "x86 detected")
     if (MSVC)
-        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /utf-8")
-        set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /utf-8")
-        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} /utf-8")
-        if(NOT WHISPER_NO_AVX2)
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX2")
         set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /arch:AVX2")
-            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} /arch:AVX2")
-        else()
-            if(NOT WHISPER_NO_AVX)
-                set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX")
-                set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /arch:AVX")
-                set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} /arch:AVX")
-            endif()
-        endif()
     else()
         if (EMSCRIPTEN)
             set(CMAKE_C_FLAGS   "${CMAKE_C_FLAGS}   -pthread")
@@ -311,12 +168,7 @@ else()
             if(NOT WHISPER_NO_AVX2)
                 set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx2")
             endif()
-            if(NOT WHISPER_NO_FMA)
-                set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfma")
-            endif()
-            if(NOT WHISPER_NO_F16C)
-                set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mf16c")
-            endif()
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfma -mf16c")
         endif()
     endif()
 endif()
@@ -325,51 +177,6 @@ if (WHISPER_PERF)
     set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_PERF)
 endif()
 
-#
-# whisper.coreml - Core ML support
-#
-
-if (WHISPER_COREML)
-    set(TARGET whisper.coreml)
-
-    add_library(${TARGET}
-        coreml/whisper-encoder.h
-        coreml/whisper-encoder.mm
-        coreml/whisper-encoder-impl.h
-        coreml/whisper-encoder-impl.m
-        )
-
-    include(DefaultTargetOptions)
-
-    target_include_directories(${TARGET} PUBLIC
-        .
-        )
-
-    target_link_libraries(${TARGET} PRIVATE ${FOUNDATION_FRAMEWORK} ${COREML_FRAMEWORK})
-
-    set_target_properties(${TARGET} PROPERTIES
-        COMPILE_FLAGS "-fobjc-arc"
-        )
-endif()
-
-if (WHISPER_OPENVINO)
-    set(TARGET whisper.openvino)
-
-    add_library(${TARGET} OBJECT
-        openvino/whisper-openvino-encoder.h
-        openvino/whisper-openvino-encoder.cpp
-        )
-
-    target_include_directories(${TARGET} PUBLIC
-        .
-        )
-
-    set_property(TARGET ${TARGET} PROPERTY POSITION_INDEPENDENT_CODE ON)
-    set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DWHISPER_USE_OPENVINO)
-
-    target_link_libraries(${TARGET} PRIVATE openvino::runtime)
-endif()
-
 #
 # whisper - this is the main library of the project
 #
@@ -379,26 +186,14 @@ set(TARGET whisper)
 add_library(${TARGET}
     ggml.h
     ggml.c
-    ${GGML_CUDA_SOURCES}
-    ${GGML_OPENCL_SOURCES}
     whisper.h
     whisper.cpp
     )
 
-include(DefaultTargetOptions)
-
 target_include_directories(${TARGET} PUBLIC
     .
     )
 
-if (WHISPER_COREML)
-    target_link_libraries(${TARGET} PRIVATE whisper.coreml)
-endif()
-
-if (WHISPER_OPENVINO)
-    target_link_libraries(${TARGET} PRIVATE whisper.openvino)
-endif()
-
 if (MSVC)
     target_link_libraries(${TARGET} PRIVATE ${WHISPER_EXTRA_LIBS} ${CMAKE_THREAD_LIBS_INIT})
 
@@ -414,19 +209,7 @@ if (BUILD_SHARED_LIBS)
 
     target_compile_definitions(${TARGET} PUBLIC
         WHISPER_SHARED
-        GGML_SHARED
         )
-
-    target_compile_definitions(${TARGET} PRIVATE
-        WHISPER_BUILD
-        GGML_BUILD
-        )
-endif()
-
-if (GGML_CUDA_SOURCES)
-    message(STATUS "GGML CUDA sources found, configuring CUDA architecture")
-    set_property(TARGET whisper PROPERTY CUDA_ARCHITECTURES OFF)
-    set_property(TARGET whisper PROPERTY CUDA_SELECT_NVCC_ARCH_FLAGS "Auto")
 endif()
 
 if (EMSCRIPTEN)
@@ -437,13 +220,9 @@ target_compile_definitions(${TARGET} PUBLIC
     ${WHISPER_EXTRA_FLAGS}
     )
 
-set_target_properties(${TARGET} PROPERTIES PUBLIC_HEADER "whisper.h")
-
 install(TARGETS ${TARGET}
     LIBRARY DESTINATION lib
     ARCHIVE DESTINATION lib/static
-    RUNTIME DESTINATION bin
-    PUBLIC_HEADER DESTINATION include
     )
 
 #
@@ -456,7 +235,7 @@ add_subdirectory(bindings)
 # programs, examples and tests
 #
 
-if (WHISPER_BUILD_TESTS AND NOT CMAKE_JS_VERSION)
+if (WHISPER_BUILD_TESTS)
     enable_testing()
     add_subdirectory(tests)
 endif ()

LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2023 Georgi Gerganov
+Copyright (c) 2022 Georgi Gerganov
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

Makefile

@@ -1,5 +1,3 @@
-default: main bench quantize
-
 ifndef UNAME_S
 UNAME_S := $(shell uname -s)
 endif
@@ -12,15 +10,6 @@ ifndef UNAME_M
 UNAME_M := $(shell uname -m)
 endif
 
-ifndef NVCC_VERSION
-	ifeq ($(call,$(shell which nvcc))$(.SHELLSTATUS),0)
-		NVCC_VERSION := $(shell nvcc --version | egrep -o "V[0-9]+.[0-9]+.[0-9]+" | cut -c2-)
-	endif
-endif
-
-CCV := $(shell $(CC) --version | head -n 1)
-CXXV := $(shell $(CXX) --version | head -n 1)
-
 # Mac OS + Arm can report x86_64
 # ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789
 ifeq ($(UNAME_S),Darwin)
@@ -38,26 +27,25 @@ endif
 # Compile flags
 #
 
-CFLAGS   = -I.              -O3 -DNDEBUG -std=c11   -fPIC
-CXXFLAGS = -I. -I./examples -O3 -DNDEBUG -std=c++11 -fPIC
+CFLAGS   = -I.              -O3 -std=c11   -fPIC
+CXXFLAGS = -I. -I./examples -O3 -std=c++11 -fPIC
 LDFLAGS  =
 
-# ref: https://github.com/ggerganov/whisper.cpp/issues/37
-ifneq ($(wildcard /usr/include/musl/*),)
-	CFLAGS += -D_POSIX_SOURCE -D_GNU_SOURCE
-	CXXFLAGS += -D_POSIX_SOURCE -D_GNU_SOURCE
-endif
-
-# RLIMIT_MEMLOCK came in BSD, is not specified in POSIX.1,
-# and on macOS its availability depends on enabling Darwin extensions
-ifeq ($(UNAME_S),Darwin)
-	CFLAGS   += -D_DARWIN_C_SOURCE
-	CXXFLAGS += -D_DARWIN_C_SOURCE
-endif
-
 # OS specific
 # TODO: support Windows
-ifeq ($(filter $(UNAME_S),Linux Darwin DragonFly FreeBSD NetBSD OpenBSD Haiku),$(UNAME_S))
+ifeq ($(UNAME_S),Linux)
+	CFLAGS   += -pthread
+	CXXFLAGS += -pthread
+endif
+ifeq ($(UNAME_S),Darwin)
+	CFLAGS   += -pthread
+	CXXFLAGS += -pthread
+endif
+ifeq ($(UNAME_S),FreeBSD)
+	CFLAGS   += -pthread
+	CXXFLAGS += -pthread
+endif
+ifeq ($(UNAME_S),Haiku)
 	CFLAGS   += -pthread
 	CXXFLAGS += -pthread
 endif
@@ -65,69 +53,64 @@ endif
 # Architecture specific
 # TODO: probably these flags need to be tweaked on some architectures
 #       feel free to update the Makefile for your architecture and send a pull request or issue
-ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686 amd64))
+ifeq ($(UNAME_M),x86_64)
 	ifeq ($(UNAME_S),Darwin)
-		CPUINFO_CMD := sysctl machdep.cpu.features
-	else ifeq ($(UNAME_S),Linux)
-		CPUINFO_CMD := cat /proc/cpuinfo
-	else ifneq (,$(filter MINGW32_NT% MINGW64_NT%,$(UNAME_S)))
-		CPUINFO_CMD := cat /proc/cpuinfo
-	else ifneq (,$(filter DragonFly FreeBSD,$(UNAME_S)))
-		CPUINFO_CMD := grep Features /var/run/dmesg.boot
-	else ifeq ($(UNAME_S),Haiku)
-		CPUINFO_CMD := sysinfo -cpu
-	endif
-
-	ifdef CPUINFO_CMD
-		AVX_M := $(shell $(CPUINFO_CMD) | grep -iwE 'AVX|AVX1.0')
-		ifneq (,$(AVX_M))
+		CFLAGS += -mfma -mf16c
+		AVX1_M := $(shell sysctl machdep.cpu.features)
+		ifneq (,$(findstring AVX1.0,$(AVX1_M)))
 			CFLAGS += -mavx
-			CXXFLAGS += -mavx
 		endif
-
-		AVX2_M := $(shell $(CPUINFO_CMD) | grep -iw 'AVX2')
-		ifneq (,$(AVX2_M))
+		AVX2_M := $(shell sysctl machdep.cpu.leaf7_features)
+		ifneq (,$(findstring AVX2,$(AVX2_M)))
 			CFLAGS += -mavx2
-			CXXFLAGS += -mavx2
 		endif
-
-		FMA_M := $(shell $(CPUINFO_CMD) | grep -iw 'FMA')
-		ifneq (,$(FMA_M))
+	else ifeq ($(UNAME_S),Linux)
+		AVX1_M := $(shell grep "avx " /proc/cpuinfo)
+		ifneq (,$(findstring avx,$(AVX1_M)))
+			CFLAGS += -mavx
+		endif
+		AVX2_M := $(shell grep "avx2 " /proc/cpuinfo)
+		ifneq (,$(findstring avx2,$(AVX2_M)))
+			CFLAGS += -mavx2
+		endif
+		FMA_M := $(shell grep "fma " /proc/cpuinfo)
+		ifneq (,$(findstring fma,$(FMA_M)))
 			CFLAGS += -mfma
-			CXXFLAGS += -mfma
 		endif
-
-		F16C_M := $(shell $(CPUINFO_CMD) | grep -iw 'F16C')
-		ifneq (,$(F16C_M))
+		F16C_M := $(shell grep "f16c " /proc/cpuinfo)
+		ifneq (,$(findstring f16c,$(F16C_M)))
 			CFLAGS += -mf16c
-			CXXFLAGS += -mf16c
 		endif
-
-		SSE3_M := $(shell $(CPUINFO_CMD) | grep -iwE 'PNI|SSE3')
-		ifneq (,$(SSE3_M))
-			CFLAGS   += -msse3
-			CXXFLAGS += -msse3
+	else ifeq ($(UNAME_S),Haiku)
+		AVX1_M := $(shell sysinfo -cpu | grep "AVX ")
+		ifneq (,$(findstring avx,$(AVX1_M)))
+			CFLAGS += -mavx
 		endif
-
-		SSSE3_M := $(shell $(CPUINFO_CMD) | grep -iw 'SSSE3')
-		ifneq (,$(SSSE3_M))
-			CFLAGS   += -mssse3
-			CXXFLAGS += -mssse3
+		AVX2_M := $(shell sysinfo -cpu | grep "AVX2 ")
+		ifneq (,$(findstring avx2,$(AVX2_M)))
+			CFLAGS += -mavx2
 		endif
+		FMA_M := $(shell sysinfo -cpu | grep "FMA ")
+		ifneq (,$(findstring fma,$(FMA_M)))
+			CFLAGS += -mfma
+		endif
+		F16C_M := $(shell sysinfo -cpu | grep "F16C ")
+		ifneq (,$(findstring f16c,$(F16C_M)))
+			CFLAGS += -mf16c
+		endif
+	else
+		CFLAGS += -mfma -mf16c -mavx -mavx2
 	endif
 endif
-
-ifneq ($(filter ppc64%,$(UNAME_M)),)
+ifeq ($(UNAME_M),amd64)
+	CFLAGS += -mavx -mavx2 -mfma -mf16c
+endif
+ifeq ($(UNAME_M),ppc64le)
 	POWER9_M := $(shell grep "POWER9" /proc/cpuinfo)
 	ifneq (,$(findstring POWER9,$(POWER9_M)))
 		CFLAGS += -mpower9-vector
 	endif
-	# Require c++23's std::byteswap for big-endian support.
-	ifeq ($(UNAME_M),ppc64)
-		CXXFLAGS += -std=c++23 -DGGML_BIG_ENDIAN
-	endif
 endif
-
 ifndef WHISPER_NO_ACCELERATE
 	# Mac M1 - include Accelerate framework
 	ifeq ($(UNAME_S),Darwin)
@@ -135,145 +118,49 @@ ifndef WHISPER_NO_ACCELERATE
 		LDFLAGS += -framework Accelerate
 	endif
 endif
-
-ifdef WHISPER_COREML
-	CXXFLAGS += -DWHISPER_USE_COREML
-	LDFLAGS  += -framework Foundation -framework CoreML
-
-ifdef WHISPER_COREML_ALLOW_FALLBACK
-	CXXFLAGS += -DWHISPER_COREML_ALLOW_FALLBACK
-endif
-endif
-
 ifdef WHISPER_OPENBLAS
-	CFLAGS  += -DGGML_USE_OPENBLAS -I/usr/local/include/openblas -I/usr/include/openblas
+	CFLAGS  += -DGGML_USE_OPENBLAS -I/usr/local/include/openblas
 	LDFLAGS += -lopenblas
 endif
-
-ifdef WHISPER_CUBLAS
-	ifeq ($(shell expr $(NVCC_VERSION) \>= 11.6), 1)
-		CUDA_ARCH_FLAG=native
-	else
-		CUDA_ARCH_FLAG=all
-	endif
-
-	CFLAGS      += -DGGML_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include
-	CXXFLAGS    += -DGGML_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include
-	LDFLAGS     += -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L/usr/local/cuda/lib64 -L/opt/cuda/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib
-	WHISPER_OBJ += ggml-cuda.o
-	NVCC        = nvcc
-	NVCCFLAGS   = --forward-unknown-to-host-compiler -arch=$(CUDA_ARCH_FLAG)
-
-ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
-	$(NVCC) $(NVCCFLAGS) $(CXXFLAGS) -Wno-pedantic -c $< -o $@
-endif
-
-ifdef WHISPER_HIPBLAS
-	ROCM_PATH   ?= /opt/rocm
-	HIPCC       ?= $(ROCM_PATH)/bin/hipcc
-	GPU_TARGETS ?= $(shell $(ROCM_PATH)/llvm/bin/amdgpu-arch)
-	CFLAGS      += -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS
-	CXXFLAGS    += -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS
-	LDFLAGS     += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib
-	LDFLAGS     += -lhipblas -lamdhip64 -lrocblas
-	HIPFLAGS    += $(addprefix --offload-arch=,$(GPU_TARGETS))
-	WHISPER_OBJ += ggml-cuda.o
-
-ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
-	$(HIPCC) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
-endif
-
-ifdef WHISPER_CLBLAST
-	CFLAGS 		+= -DGGML_USE_CLBLAST
-	CXXFLAGS 	+= -DGGML_USE_CLBLAST
-	LDFLAGS	 	+= -lclblast
-	ifeq ($(UNAME_S),Darwin)
-		LDFLAGS	 	+= -framework OpenCL
-	else
-		LDFLAGS	    += -lOpenCL
-	endif
-	WHISPER_OBJ	+= ggml-opencl.o
-
-ggml-opencl.o: ggml-opencl.cpp ggml-opencl.h
-	$(CXX) $(CXXFLAGS) -c $< -o $@
-endif
-
 ifdef WHISPER_GPROF
 	CFLAGS  += -pg
 	CXXFLAGS  += -pg
 endif
-
 ifneq ($(filter aarch64%,$(UNAME_M)),)
-	CFLAGS   += -mcpu=native
-	CXXFLAGS += -mcpu=native
 endif
-
 ifneq ($(filter armv6%,$(UNAME_M)),)
-	# 32-bit Raspberry Pi 1, 2, 3
-	CFLAGS += -mfpu=neon -mfp16-format=ieee -mno-unaligned-access
+	# Raspberry Pi 1, 2, 3
+	CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
 endif
-
 ifneq ($(filter armv7%,$(UNAME_M)),)
-	# 32-bit ARM, for example on Armbian or possibly raspbian
-	#CFLAGS   += -mfpu=neon -mfp16-format=ieee -funsafe-math-optimizations -mno-unaligned-access
-	#CXXFLAGS += -mfpu=neon -mfp16-format=ieee -funsafe-math-optimizations -mno-unaligned-access
-
-	# 64-bit ARM on 32-bit OS, use these (TODO: auto-detect 64-bit)
-	CFLAGS   += -mfpu=neon-fp-armv8 -mfp16-format=ieee -funsafe-math-optimizations -mno-unaligned-access
-	CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -funsafe-math-optimizations -mno-unaligned-access
+	# Raspberry Pi 4
+	CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
 endif
-
 ifneq ($(filter armv8%,$(UNAME_M)),)
 	# Raspberry Pi 4
-	CFLAGS   += -mfpu=neon-fp-armv8 -mfp16-format=ieee -funsafe-math-optimizations -mno-unaligned-access
-	CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -funsafe-math-optimizations -mno-unaligned-access
+	CFLAGS += -mfp16-format=ieee -mno-unaligned-access
 endif
 
-#
-# Print build information
-#
-
-$(info I whisper.cpp build info: )
-$(info I UNAME_S:  $(UNAME_S))
-$(info I UNAME_P:  $(UNAME_P))
-$(info I UNAME_M:  $(UNAME_M))
-$(info I CFLAGS:   $(CFLAGS))
-$(info I CXXFLAGS: $(CXXFLAGS))
-$(info I LDFLAGS:  $(LDFLAGS))
-$(info I CC:       $(CCV))
-$(info I CXX:      $(CXXV))
-$(info )
+default: main
 
 #
 # Build library
 #
 
-ggml.o: ggml.c ggml.h ggml-cuda.h
-	$(CC)  $(CFLAGS)   -c $< -o $@
+ggml.o: ggml.c ggml.h
+	$(CC)  $(CFLAGS)   -c ggml.c -o ggml.o
 
-whisper.o: whisper.cpp whisper.h ggml.h ggml-cuda.h
-	$(CXX) $(CXXFLAGS) -c $< -o $@
+whisper.o: whisper.cpp whisper.h
+	$(CXX) $(CXXFLAGS) -c whisper.cpp -o whisper.o
 
-ifndef WHISPER_COREML
-WHISPER_OBJ += whisper.o
-else
-whisper-encoder.o: coreml/whisper-encoder.mm coreml/whisper-encoder.h
-	$(CXX) -O3 -I . -fobjc-arc -c coreml/whisper-encoder.mm -o whisper-encoder.o
+libwhisper.a: ggml.o whisper.o
+	$(AR) rcs libwhisper.a ggml.o whisper.o
 
-whisper-encoder-impl.o: coreml/whisper-encoder-impl.m coreml/whisper-encoder-impl.h
-	$(CXX) -O3 -I . -fobjc-arc -c coreml/whisper-encoder-impl.m -o whisper-encoder-impl.o
-
-WHISPER_OBJ += whisper.o whisper-encoder.o whisper-encoder-impl.o
-endif
-
-libwhisper.a: ggml.o $(WHISPER_OBJ)
-	$(AR) rcs libwhisper.a ggml.o $(WHISPER_OBJ)
-
-libwhisper.so: ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) -shared -o libwhisper.so ggml.o $(WHISPER_OBJ) $(LDFLAGS)
+libwhisper.so: ggml.o whisper.o
+	$(CXX) $(CXXFLAGS) -shared -o libwhisper.so ggml.o whisper.o $(LDFLAGS)
 
 clean:
-	rm -f *.o main stream command talk talk-llama bench quantize lsp libwhisper.a libwhisper.so
+	rm -f *.o main stream command talk bench libwhisper.a libwhisper.so
 
 #
 # Examples
@@ -281,33 +168,21 @@ clean:
 
 CC_SDL=`sdl2-config --cflags --libs`
 
-SRC_COMMON     = examples/common.cpp examples/common-ggml.cpp
-SRC_COMMON_SDL = examples/common-sdl.cpp
-
-main: examples/main/main.cpp $(SRC_COMMON) ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/main/main.cpp $(SRC_COMMON) ggml.o $(WHISPER_OBJ) -o main $(LDFLAGS)
+main: examples/main/main.cpp ggml.o whisper.o
+	$(CXX) $(CXXFLAGS) examples/main/main.cpp ggml.o whisper.o -o main $(LDFLAGS)
 	./main -h
 
-bench: examples/bench/bench.cpp ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/bench/bench.cpp ggml.o $(WHISPER_OBJ) -o bench $(LDFLAGS)
+stream: examples/stream/stream.cpp ggml.o whisper.o
+	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp ggml.o whisper.o -o stream $(CC_SDL) $(LDFLAGS)
 
-quantize: examples/quantize/quantize.cpp ggml.o $(WHISPER_OBJ) $(SRC_COMMON)
-	$(CXX) $(CXXFLAGS) examples/quantize/quantize.cpp $(SRC_COMMON) ggml.o $(WHISPER_OBJ) -o quantize $(LDFLAGS)
+command: examples/command/command.cpp ggml.o whisper.o
+	$(CXX) $(CXXFLAGS) examples/command/command.cpp ggml.o whisper.o -o command $(CC_SDL) $(LDFLAGS)
 
-stream: examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ) -o stream $(CC_SDL) $(LDFLAGS)
+talk: examples/talk/talk.cpp  examples/talk/gpt-2.cpp ggml.o whisper.o
+	$(CXX) $(CXXFLAGS) examples/talk/talk.cpp examples/talk/gpt-2.cpp ggml.o whisper.o -o talk $(CC_SDL) $(LDFLAGS)
 
-command: examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ) -o command $(CC_SDL) $(LDFLAGS)
-
-lsp: examples/lsp/lsp.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/lsp/lsp.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ) -o lsp $(CC_SDL) $(LDFLAGS)
-
-talk: examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ) -o talk $(CC_SDL) $(LDFLAGS)
-
-talk-llama: examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) ggml.o $(WHISPER_OBJ) -o talk-llama $(CC_SDL) $(LDFLAGS)
+bench: examples/bench/bench.cpp ggml.o whisper.o
+	$(CXX) $(CXXFLAGS) examples/bench/bench.cpp ggml.o whisper.o -o bench $(LDFLAGS)
 
 #
 # Audio samples
@@ -322,19 +197,12 @@ samples:
 	@wget --quiet --show-progress -O samples/gb1.ogg https://upload.wikimedia.org/wikipedia/commons/1/1f/George_W_Bush_Columbia_FINAL.ogg
 	@wget --quiet --show-progress -O samples/hp0.ogg https://upload.wikimedia.org/wikipedia/en/d/d4/En.henryfphillips.ogg
 	@wget --quiet --show-progress -O samples/mm1.wav https://cdn.openai.com/whisper/draft-20220913a/micro-machines.wav
-	@wget --quiet --show-progress -O samples/a13.mp3 https://upload.wikimedia.org/wikipedia/commons/transcoded/6/6f/Apollo13-wehaveaproblem.ogg/Apollo13-wehaveaproblem.ogg.mp3
-	@wget --quiet --show-progress -O samples/diffusion2023-07-03.flac https://archive.org/download/diffusion2023-07-03/diffusion2023-07-03.flac
 	@echo "Converting to 16-bit WAV ..."
 	@ffmpeg -loglevel -0 -y -i samples/gb0.ogg -ar 16000 -ac 1 -c:a pcm_s16le samples/gb0.wav
 	@ffmpeg -loglevel -0 -y -i samples/gb1.ogg -ar 16000 -ac 1 -c:a pcm_s16le samples/gb1.wav
 	@ffmpeg -loglevel -0 -y -i samples/hp0.ogg -ar 16000 -ac 1 -c:a pcm_s16le samples/hp0.wav
-	@rm samples/*.ogg
 	@ffmpeg -loglevel -0 -y -i samples/mm1.wav -ar 16000 -ac 1 -c:a pcm_s16le samples/mm0.wav
 	@rm samples/mm1.wav
-	@ffmpeg -loglevel -0 -y -i samples/a13.mp3 -ar 16000 -ac 1 -c:a pcm_s16le -ss 00:00:00 -to 00:00:30 samples/a13.wav
-	@rm samples/a13.mp3
-	@ffmpeg -loglevel -0 -y -i samples/diffusion2023-07-03.flac -ar 16000 -ac 1 -c:a pcm_s16le samples/diffusion2023-07-03.wav
-	@rm samples/diffusion2023-07-03.flac
 
 #
 # Models
@@ -376,4 +244,4 @@ tiny.en tiny base.en base small.en small medium.en medium large-v1 large: main
 
 .PHONY: tests
 tests:
-	bash ./tests/run-tests.sh $(word 2, $(MAKECMDGOALS))
+	bash ./tests/run-tests.sh

README.md

@@ -1,28 +1,20 @@
 # whisper.cpp
 
-![whisper.cpp](https://user-images.githubusercontent.com/1991296/235238348-05d0f6a4-da44-4900-a1de-d0707e75b763.jpeg)
-
 [![Actions Status](https://github.com/ggerganov/whisper.cpp/workflows/CI/badge.svg)](https://github.com/ggerganov/whisper.cpp/actions)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 [![npm](https://img.shields.io/npm/v/whisper.cpp.svg)](https://www.npmjs.com/package/whisper.cpp/)
 
-Beta: [v1.4.2](https://github.com/ggerganov/whisper.cpp/releases/tag/v1.4.2) / Stable: [v1.2.1](https://github.com/ggerganov/whisper.cpp/releases/tag/v1.2.1) / [Roadmap | F.A.Q.](https://github.com/ggerganov/whisper.cpp/discussions/126)
+[Roadmap | F.A.Q.](https://github.com/ggerganov/whisper.cpp/discussions/126)
 
 High-performance inference of [OpenAI's Whisper](https://github.com/openai/whisper) automatic speech recognition (ASR) model:
 
 - Plain C/C++ implementation without dependencies
-- Apple silicon first-class citizen - optimized via ARM NEON, Accelerate framework and [Core ML](https://github.com/ggerganov/whisper.cpp#core-ml-support)
+- Apple silicon first-class citizen - optimized via Arm Neon and Accelerate framework
 - AVX intrinsics support for x86 architectures
-- VSX intrinsics support for POWER architectures
 - Mixed F16 / F32 precision
-- [4-bit and 5-bit integer quantization support](https://github.com/ggerganov/whisper.cpp#quantization)
-- Low memory usage (Flash Attention)
+- Low memory usage (Flash Attention + Flash Forward)
 - Zero memory allocations at runtime
 - Runs on the CPU
-- [Partial GPU support for NVIDIA via cuBLAS](https://github.com/ggerganov/whisper.cpp#nvidia-gpu-support-via-cublas)
-- [Partial OpenCL GPU support via CLBlast](https://github.com/ggerganov/whisper.cpp#opencl-gpu-support-via-clblast)
-- [BLAS CPU support via OpenBLAS](https://github.com/ggerganov/whisper.cpp#blas-cpu-support-via-openblas)
-- [OpenVINO Support](https://github.com/ggerganov/whisper.cpp#openvino-support)
 - [C-style API](https://github.com/ggerganov/whisper.cpp/blob/master/whisper.h)
 
 Supported platforms:
@@ -30,7 +22,6 @@ Supported platforms:
 - [x] Mac OS (Intel and Arm)
 - [x] [iOS](examples/whisper.objc)
 - [x] [Android](examples/whisper.android)
-- [x] [Java](bindings/java/README.md)
 - [x] Linux / [FreeBSD](https://github.com/ggerganov/whisper.cpp/issues/56#issuecomment-1350920264)
 - [x] [WebAssembly](examples/whisper.wasm)
 - [x] Windows ([MSVC](https://github.com/ggerganov/whisper.cpp/blob/master/.github/workflows/build.yml#L117-L144) and [MinGW](https://github.com/ggerganov/whisper.cpp/issues/168)]
@@ -61,21 +52,17 @@ Or you can even run it straight in the browser: [talk.wasm](examples/talk.wasm)
 - Various other examples are available in the [examples](examples) folder
 
 The tensor operators are optimized heavily for Apple silicon CPUs. Depending on the computation size, Arm Neon SIMD
-intrinsics or CBLAS Accelerate framework routines are used. The latter are especially effective for bigger sizes since
+instrisics or CBLAS Accelerate framework routines are used. The latter are especially effective for bigger sizes since
 the Accelerate framework utilizes the special-purpose AMX coprocessor available in modern Apple products.
 
 ## Quick start
 
-First clone the repository.
-
-Then, download one of the Whisper models converted in [ggml format](models). For example:
+First, download one of the Whisper models converted in [ggml format](models). For example:
 
 ```bash
 bash ./models/download-ggml-model.sh base.en
 ```
 
-If you wish to convert the Whisper models to ggml format yourself, instructions are in [models/README.md](models/README.md).
-
 Now build the [main](examples/main) example and transcribe an audio file like this:
 
 ```bash
@@ -83,7 +70,7 @@ Now build the [main](examples/main) example and transcribe an audio file like th
 make
 
 # transcribe an audio file
-./main -f samples/jfk.wav
+./main -f input.wav
 ```
 
 ---
@@ -109,32 +96,20 @@ options:
   -d  N,    --duration N    [0      ] duration of audio to process in milliseconds
   -mc N,    --max-context N [-1     ] maximum number of text context tokens to store
   -ml N,    --max-len N     [0      ] maximum segment length in characters
-  -bo N,     --best-of N         [5      ] number of best candidates to keep
-  -bs N,     --beam-size N       [-1     ] beam size for beam search
   -wt N,    --word-thold N  [0.01   ] word timestamp probability threshold
-  -et N,     --entropy-thold N   [2.40   ] entropy threshold for decoder fail
-  -lpt N,    --logprob-thold N   [-1.00  ] log probability threshold for decoder fail
   -su,      --speed-up      [false  ] speed up audio by x2 (reduced accuracy)
   -tr,      --translate     [false  ] translate from source language to english
-  -tdrz,     --tinydiarize       [false  ] enable tinydiarize (requires a tdrz model)
-  -di,       --diarize           [false  ] stereo audio diarization
-  -nf,       --no-fallback       [false  ] do not use temperature fallback while decoding
   -otxt,    --output-txt    [false  ] output result in a text file
   -ovtt,    --output-vtt    [false  ] output result in a vtt file
   -osrt,    --output-srt    [false  ] output result in a srt file
   -owts,    --output-words  [false  ] output script for generating karaoke video
-  -ocsv,     --output-csv        [false  ] output result in a CSV file
-  -of FNAME, --output-file FNAME [       ] output file path (without file extension)
   -ps,      --print-special [false  ] print special tokens
   -pc,      --print-colors  [false  ] print colors
-  -pp,       --print-progress    [false  ] print progress
   -nt,      --no-timestamps [true   ] do not print timestamps
-  -l LANG,   --language LANG     [en     ] spoken language ('auto' for auto-detect)
-             --prompt PROMPT     [       ] initial prompt
+  -l LANG,  --language LANG [en     ] spoken language
   -m FNAME, --model FNAME   [models/ggml-base.en.bin] model path
   -f FNAME, --file FNAME    [       ] input WAV file path
 
-
 bash ./models/download-ggml-model.sh base.en
 Downloading ggml model base.en ...
 ggml-base.en.bin               100%[========================>] 141.11M  6.34MB/s    in 24s
@@ -152,8 +127,7 @@ Running base.en on all samples in ./samples ...
 [+] Running base.en on samples/jfk.wav ... (run 'ffplay samples/jfk.wav' to listen)
 ----------------------------------------------
 
-whisper_init_from_file: loading model from 'models/ggml-base.en.bin'
-whisper_model_load: loading model
+whisper_model_load: loading model from 'models/ggml-base.en.bin'
 whisper_model_load: n_vocab       = 51864
 whisper_model_load: n_audio_ctx   = 1500
 whisper_model_load: n_audio_state = 512
@@ -166,14 +140,13 @@ whisper_model_load: n_text_layer  = 6
 whisper_model_load: n_mels        = 80
 whisper_model_load: f16           = 1
 whisper_model_load: type          = 2
-whisper_model_load: mem required  =  215.00 MB (+    6.00 MB per decoder)
-whisper_model_load: kv self size  =    5.25 MB
-whisper_model_load: kv cross size =   17.58 MB
 whisper_model_load: adding 1607 extra tokens
-whisper_model_load: model ctx     =  140.60 MB
+whisper_model_load: mem_required  =  506.00 MB
+whisper_model_load: ggml ctx size =  140.60 MB
+whisper_model_load: memory size   =   22.83 MB
 whisper_model_load: model size    =  140.54 MB
 
-system_info: n_threads = 4 / 10 | AVX = 0 | AVX2 = 0 | AVX512 = 0 | FMA = 0 | NEON = 1 | ARM_FMA = 1 | F16C = 0 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | VSX = 0 |
+system_info: n_threads = 4 / 10 | AVX = 0 | AVX2 = 0 | AVX512 = 0 | NEON = 1 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 |
 
 main: processing 'samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...
 
@@ -181,13 +154,12 @@ main: processing 'samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 proc
 [00:00:00.000 --> 00:00:11.000]   And so my fellow Americans, ask not what your country can do for you, ask what you can do for your country.
 
 
-whisper_print_timings:     fallbacks =   0 p /   0 h
-whisper_print_timings:     load time =   113.81 ms
-whisper_print_timings:      mel time =    15.40 ms
-whisper_print_timings:   sample time =    11.58 ms /    27 runs (    0.43 ms per run)
-whisper_print_timings:   encode time =   266.60 ms /     1 runs (  266.60 ms per run)
-whisper_print_timings:   decode time =    66.11 ms /    27 runs (    2.45 ms per run)
-whisper_print_timings:    total time =   476.31 ms
+whisper_print_timings:     load time =   105.91 ms
+whisper_print_timings:      mel time =    24.62 ms
+whisper_print_timings:   sample time =     3.63 ms
+whisper_print_timings:   encode time =   324.71 ms / 54.12 ms per layer
+whisper_print_timings:   decode time =    83.58 ms / 13.93 ms per layer
+whisper_print_timings:    total time =   542.81 ms
 ```
 
 The command downloads the `base.en` model converted to custom `ggml` format and runs the inference on all `.wav` samples in the folder `samples`.
@@ -230,217 +202,26 @@ make large
 
 | Model  | Disk   | Mem     | SHA                                        |
 | ---    | ---    | ---     | ---                                        |
-| tiny   |  75 MB | ~125 MB | `bd577a113a864445d4c299885e0cb97d4ba92b5f` |
-| base   | 142 MB | ~210 MB | `465707469ff3a37a2b9b8d8f89f2f99de7299dac` |
-| small  | 466 MB | ~600 MB | `55356645c2b361a969dfd0ef2c5a50d530afd8d5` |
-| medium | 1.5 GB | ~1.7 GB | `fd9727b6e1217c2f614f9b698455c4ffd82463b4` |
-| large  | 2.9 GB | ~3.3 GB | `0f4c8e34f21cf1a914c59d8b3ce882345ad349d6` |
-
-## Quantization
-
-`whisper.cpp` supports integer quantization of the Whisper `ggml` models.
-Quantized models require less memory and disk space and depending on the hardware can be processed more efficiently.
-
-Here are the steps for creating and using a quantized model:
-
-```bash
-# quantize a model with Q5_0 method
-make quantize
-./quantize models/ggml-base.en.bin models/ggml-base.en-q5_0.bin q5_0
-
-# run the examples as usual, specifying the quantized model file
-./main -m models/ggml-base.en-q5_0.bin ./samples/gb0.wav
-```
-
-## Core ML support
-
-On Apple Silicon devices, the Encoder inference can be executed on the Apple Neural Engine (ANE) via Core ML. This can result in significant
-speed-up - more than x3 faster compared with CPU-only execution. Here are the instructions for generating a Core ML model and using it with `whisper.cpp`:
-
-- Install Python dependencies needed for the creation of the Core ML model:
-
-  ```bash
-  pip install ane_transformers
-  pip install openai-whisper
-  pip install coremltools
-  ```
-
-  - To ensure `coremltools` operates correctly, please confirm that [Xcode](https://developer.apple.com/xcode/) is installed and execute `xcode-select --install` to install the command-line tools.
-  - Python 3.10 is recommended.
-  - [OPTIONAL] It is recommended to utilize a Python version management system, such as [Miniconda](https://docs.conda.io/en/latest/miniconda.html)  for this step:
-    - To create an environment, use: `conda create -n py310-whisper python=3.10 -y`
-    - To activate the environment, use: `conda activate py310-whisper`
-
-- Generate a Core ML model. For example, to generate a `base.en` model, use:
-
-  ```bash
-  ./models/generate-coreml-model.sh base.en
-  ```
-
-  This will generate the folder `models/ggml-base.en-encoder.mlmodelc`
-
-- Build `whisper.cpp` with Core ML support:
-
-  ```bash
-  # using Makefile
-  make clean
-  WHISPER_COREML=1 make -j
-
-  # using CMake
-  cmake -B build -DWHISPER_COREML=1
-  cmake --build build -j --config Release
-  ```
-
-- Run the examples as usual. For example:
-
-  ```bash
-  ./main -m models/ggml-base.en.bin -f samples/jfk.wav
-
-  ...
-
-  whisper_init_state: loading Core ML model from 'models/ggml-base.en-encoder.mlmodelc'
-  whisper_init_state: first run on a device may take a while ...
-  whisper_init_state: Core ML model loaded
-
-  system_info: n_threads = 4 / 10 | AVX = 0 | AVX2 = 0 | AVX512 = 0 | FMA = 0 | NEON = 1 | ARM_FMA = 1 | F16C = 0 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | VSX = 0 | COREML = 1 |
-
-  ...
-  ```
-
-  The first run on a device is slow, since the ANE service compiles the Core ML model to some device-specific format.
-  Next runs are faster.
-
-For more information about the Core ML implementation please refer to PR [#566](https://github.com/ggerganov/whisper.cpp/pull/566).
-
-## OpenVINO support
-
-On platforms that support [OpenVINO](https://github.com/openvinotoolkit/openvino), the Encoder inference can be executed
-on OpenVINO-supported devices including x86 CPUs and Intel GPUs (integrated & discrete).
-
-This can result in significant speedup in encoder performance. Here are the instructions for generating the OpenVINO model and using it with `whisper.cpp`:
-
-- First, setup python virtual env. and install python dependencies. Python 3.10 is recommended.
-
-  Windows:
-  ```
-  cd models
-  python -m venv openvino_conv_env
-  openvino_conv_env\Scripts\activate
-  python -m pip install --upgrade pip
-  pip install -r openvino-conversion-requirements.txt
-  ```
-
-  Linux and macOS:
-  ```
-  cd models
-  python3 -m venv openvino_conv_env
-  source openvino_conv_env/bin/activate
-  python -m pip install --upgrade pip
-  pip install -r openvino-conversion-requirements.txt
-  ```
-
-- Generate an OpenVINO encoder model. For example, to generate a `base.en` model, use:
-
-  ```
-  python convert-whisper-to-openvino.py --model base.en
-  ```
-
-  This will produce ggml-base.en-encoder-openvino.xml/.bin IR model files. It's recommended to relocate these to the same folder as ggml models, as that
-  is the default location that the OpenVINO extension will search at runtime.
-
-- Build `whisper.cpp` with OpenVINO support:
-
-  Download OpenVINO package from [release page](https://github.com/openvinotoolkit/openvino/releases). The recommended version to use is [2023.0.0](https://github.com/openvinotoolkit/openvino/releases/tag/2023.0.0).
-
-  After downloading & extracting package onto your development system, set up required environment by sourcing setupvars script. For example:
-
-  Linux:
-  ```bash
-  source /path/to/l_openvino_toolkit_ubuntu22_2023.0.0.10926.b4452d56304_x86_64/setupvars.sh
-  ```
-
-  Windows (cmd):
-  ```
-  C:\Path\To\w_openvino_toolkit_windows_2023.0.0.10926.b4452d56304_x86_64\setupvars.bat
-  ```
-
-  And then build the project using cmake:
-  ```bash
-  cmake -B build -DWHISPER_OPENVINO=1
-  cmake --build build -j --config Release
-  ```
-
-- Run the examples as usual. For example:
-  ```bash
-  ./main -m models/ggml-base.en.bin -f samples/jfk.wav
-
-  ...
-
-  whisper_ctx_init_openvino_encoder: loading OpenVINO model from 'models/ggml-base.en-encoder-openvino.xml'
-  whisper_ctx_init_openvino_encoder: first run on a device may take a while ...
-  whisper_openvino_init: path_model = models/ggml-base.en-encoder-openvino.xml, device = GPU, cache_dir = models/ggml-base.en-encoder-openvino-cache
-  whisper_ctx_init_openvino_encoder: OpenVINO model loaded
-
-  system_info: n_threads = 4 / 8 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | COREML = 0 | OPENVINO = 1 |
-
-  ...
-  ```
-
-  The first time run on an OpenVINO device is slow, since the OpenVINO framework will compile the IR (Intermediate Representation) model to a device-specific 'blob'. This device-specific blob will get
-  cached for the next run.
-  
-For more information about the Core ML implementation please refer to PR [#1037](https://github.com/ggerganov/whisper.cpp/pull/1037).
-
-## NVIDIA GPU support via cuBLAS
-
-With NVIDIA cards the Encoder processing can to a large extent be offloaded to the GPU through cuBLAS.
-First, make sure you have installed `cuda`: https://developer.nvidia.com/cuda-downloads
-
-Now build `whisper.cpp` with cuBLAS support:
-
-```
-make clean
-WHISPER_CUBLAS=1 make -j
-```
-
-## OpenCL GPU support via CLBlast
-
-For cards and integrated GPUs that support OpenCL, the Encoder processing can be largely offloaded to the GPU through CLBlast. This is especially useful for users with AMD APUs or low end devices for up to ~2x speedup.
-
-First, make sure you have installed `CLBlast` for your OS or Distribution: https://github.com/CNugteren/CLBlast
-
-Now build `whisper.cpp` with CLBlast support:
-
-```
-Makefile:
-cd whisper.cpp
-make clean
-WHISPER_CLBLAST=1 make -j
-
-CMake:
-cd whisper.cpp
-cmake -B build -DWHISPER_CLBLAST=ON
-cmake --build build -j --config Release
-```
-
-
-Run all the examples as usual.
-
-## BLAS CPU support via OpenBLAS
-
-Encoder processing can be accelerated on the CPU via OpenBLAS.
-First, make sure you have installed `openblas`: https://www.openblas.net/
-
-Now build `whisper.cpp` with OpenBLAS support:
-
-```
-make clean
-WHISPER_OPENBLAS=1 make -j
-```
+| tiny   |  75 MB | ~390 MB | `bd577a113a864445d4c299885e0cb97d4ba92b5f` |
+| base   | 142 MB | ~500 MB | `465707469ff3a37a2b9b8d8f89f2f99de7299dac` |
+| small  | 466 MB | ~1.0 GB | `55356645c2b361a969dfd0ef2c5a50d530afd8d5` |
+| medium | 1.5 GB | ~2.6 GB | `fd9727b6e1217c2f614f9b698455c4ffd82463b4` |
+| large  | 2.9 GB | ~4.7 GB | `0f4c8e34f21cf1a914c59d8b3ce882345ad349d6` |
 
 ## Limitations
 
 - Inference only
+- No GPU support
+- Very basic greedy sampling scheme - always pick up the token with highest probability.
+  This should be similar to the [GreedyDecoder](https://github.com/openai/whisper/blob/main/whisper/decoding.py#L249-L274)
+  from the original python implementation, so in order to make a fair comparison between the 2 implementations, make sure
+  to run the python code with the following parameters:
+
+  ```
+  whisper --best_of None --beam_size None ...
+  ```
+
+  In the future, `whisper.cpp` will support more sampling strategies.
 
 ## Another example
 
@@ -453,8 +234,7 @@ in about half a minute on a MacBook M1 Pro, using `medium.en` model:
 ```java
 $ ./main -m models/ggml-medium.en.bin -f samples/gb1.wav -t 8
 
-whisper_init_from_file: loading model from 'models/ggml-medium.en.bin'
-whisper_model_load: loading model
+whisper_model_load: loading model from 'models/ggml-medium.en.bin'
 whisper_model_load: n_vocab       = 51864
 whisper_model_load: n_audio_ctx   = 1500
 whisper_model_load: n_audio_state = 1024
@@ -467,71 +247,65 @@ whisper_model_load: n_text_layer  = 24
 whisper_model_load: n_mels        = 80
 whisper_model_load: f16           = 1
 whisper_model_load: type          = 4
-whisper_model_load: mem required  = 1720.00 MB (+   43.00 MB per decoder)
-whisper_model_load: kv self size  =   42.00 MB
-whisper_model_load: kv cross size =  140.62 MB
+whisper_model_load: mem_required  = 2610.00 MB
 whisper_model_load: adding 1607 extra tokens
-whisper_model_load: model ctx     = 1462.35 MB
+whisper_model_load: ggml ctx size = 1644.97 MB
+whisper_model_load: memory size =   182.62 MB
 whisper_model_load: model size  =  1462.12 MB
 
-system_info: n_threads = 8 / 10 | AVX = 0 | AVX2 = 0 | AVX512 = 0 | FMA = 0 | NEON = 1 | ARM_FMA = 1 | F16C = 0 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | VSX = 0 |
+main: processing 'samples/gb1.wav' (3179750 samples, 198.7 sec), 8 threads, lang = en, task = transcribe, timestamps = 1 ...
 
-main: processing 'samples/gb1.wav' (3179750 samples, 198.7 sec), 8 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...
+[00:00.000 --> 00:08.000]   My fellow Americans, this day has brought terrible news and great sadness to our country.
+[00:08.000 --> 00:17.000]   At nine o'clock this morning, Mission Control in Houston lost contact with our Space Shuttle Columbia.
+[00:17.000 --> 00:23.000]   A short time later, debris was seen falling from the skies above Texas.
+[00:23.000 --> 00:29.000]   The Columbia's lost. There are no survivors.
+[00:29.000 --> 00:32.000]   On board was a crew of seven.
+[00:32.000 --> 00:39.000]   Colonel Rick Husband, Lieutenant Colonel Michael Anderson, Commander Laurel Clark,
+[00:39.000 --> 00:48.000]   Captain David Brown, Commander William McCool, Dr. Kultna Shavla, and Ilan Ramon,
+[00:48.000 --> 00:52.000]   a colonel in the Israeli Air Force.
+[00:52.000 --> 00:58.000]   These men and women assumed great risk in the service to all humanity.
+[00:58.000 --> 01:03.000]   In an age when space flight has come to seem almost routine,
+[01:03.000 --> 01:07.000]   it is easy to overlook the dangers of travel by rocket
+[01:07.000 --> 01:12.000]   and the difficulties of navigating the fierce outer atmosphere of the Earth.
+[01:12.000 --> 01:18.000]   These astronauts knew the dangers, and they faced them willingly,
+[01:18.000 --> 01:23.000]   knowing they had a high and noble purpose in life.
+[01:23.000 --> 01:31.000]   Because of their courage and daring and idealism, we will miss them all the more.
+[01:31.000 --> 01:36.000]   All Americans today are thinking as well of the families of these men and women
+[01:36.000 --> 01:40.000]   who have been given this sudden shock and grief.
+[01:40.000 --> 01:45.000]   You're not alone. Our entire nation grieves with you,
+[01:45.000 --> 01:52.000]   and those you love will always have the respect and gratitude of this country.
+[01:52.000 --> 01:56.000]   The cause in which they died will continue.
+[01:56.000 --> 02:04.000]   Mankind is led into the darkness beyond our world by the inspiration of discovery
+[02:04.000 --> 02:11.000]   and the longing to understand. Our journey into space will go on.
+[02:11.000 --> 02:16.000]   In the skies today, we saw destruction and tragedy.
+[02:16.000 --> 02:22.000]   Yet farther than we can see, there is comfort and hope.
+[02:22.000 --> 02:29.000]   In the words of the prophet Isaiah, "Lift your eyes and look to the heavens
+[02:29.000 --> 02:35.000]   who created all these. He who brings out the starry hosts one by one
+[02:35.000 --> 02:39.000]   and calls them each by name."
+[02:39.000 --> 02:46.000]   Because of His great power and mighty strength, not one of them is missing.
+[02:46.000 --> 02:55.000]   The same Creator who names the stars also knows the names of the seven souls we mourn today.
+[02:55.000 --> 03:01.000]   The crew of the shuttle Columbia did not return safely to earth,
+[03:01.000 --> 03:05.000]   yet we can pray that all are safely home.
+[03:05.000 --> 03:13.000]   May God bless the grieving families, and may God continue to bless America.
+[03:13.000 --> 03:41.000]   Audio
 
 
-[00:00:00.000 --> 00:00:08.000]   My fellow Americans, this day has brought terrible news and great sadness to our country.
-[00:00:08.000 --> 00:00:17.000]   At nine o'clock this morning, Mission Control in Houston lost contact with our Space Shuttle Columbia.
-[00:00:17.000 --> 00:00:23.000]   A short time later, debris was seen falling from the skies above Texas.
-[00:00:23.000 --> 00:00:29.000]   The Columbia's lost. There are no survivors.
-[00:00:29.000 --> 00:00:32.000]   On board was a crew of seven.
-[00:00:32.000 --> 00:00:39.000]   Colonel Rick Husband, Lieutenant Colonel Michael Anderson, Commander Laurel Clark,
-[00:00:39.000 --> 00:00:48.000]   Captain David Brown, Commander William McCool, Dr. Kultna Shavla, and Ilan Ramon,
-[00:00:48.000 --> 00:00:52.000]   a colonel in the Israeli Air Force.
-[00:00:52.000 --> 00:00:58.000]   These men and women assumed great risk in the service to all humanity.
-[00:00:58.000 --> 00:01:03.000]   In an age when space flight has come to seem almost routine,
-[00:01:03.000 --> 00:01:07.000]   it is easy to overlook the dangers of travel by rocket
-[00:01:07.000 --> 00:01:12.000]   and the difficulties of navigating the fierce outer atmosphere of the Earth.
-[00:01:12.000 --> 00:01:18.000]   These astronauts knew the dangers, and they faced them willingly,
-[00:01:18.000 --> 00:01:23.000]   knowing they had a high and noble purpose in life.
-[00:01:23.000 --> 00:01:31.000]   Because of their courage and daring and idealism, we will miss them all the more.
-[00:01:31.000 --> 00:01:36.000]   All Americans today are thinking as well of the families of these men and women
-[00:01:36.000 --> 00:01:40.000]   who have been given this sudden shock and grief.
-[00:01:40.000 --> 00:01:45.000]   You're not alone. Our entire nation grieves with you,
-[00:01:45.000 --> 00:01:52.000]   and those you love will always have the respect and gratitude of this country.
-[00:01:52.000 --> 00:01:56.000]   The cause in which they died will continue.
-[00:01:56.000 --> 00:02:04.000]   Mankind is led into the darkness beyond our world by the inspiration of discovery
-[00:02:04.000 --> 00:02:11.000]   and the longing to understand. Our journey into space will go on.
-[00:02:11.000 --> 00:02:16.000]   In the skies today, we saw destruction and tragedy.
-[00:02:16.000 --> 00:02:22.000]   Yet farther than we can see, there is comfort and hope.
-[00:02:22.000 --> 00:02:29.000]   In the words of the prophet Isaiah, "Lift your eyes and look to the heavens
-[00:02:29.000 --> 00:02:35.000]   who created all these. He who brings out the starry hosts one by one
-[00:02:35.000 --> 00:02:39.000]   and calls them each by name."
-[00:02:39.000 --> 00:02:46.000]   Because of His great power and mighty strength, not one of them is missing.
-[00:02:46.000 --> 00:02:55.000]   The same Creator who names the stars also knows the names of the seven souls we mourn today.
-[00:02:55.000 --> 00:03:01.000]   The crew of the shuttle Columbia did not return safely to earth,
-[00:03:01.000 --> 00:03:05.000]   yet we can pray that all are safely home.
-[00:03:05.000 --> 00:03:13.000]   May God bless the grieving families, and may God continue to bless America.
-[00:03:13.000 --> 00:03:19.000]   [Silence]
-
-
-whisper_print_timings:     fallbacks =   1 p /   0 h
-whisper_print_timings:     load time =   569.03 ms
-whisper_print_timings:      mel time =   146.85 ms
-whisper_print_timings:   sample time =   238.66 ms /   553 runs (    0.43 ms per run)
-whisper_print_timings:   encode time = 18665.10 ms /     9 runs ( 2073.90 ms per run)
-whisper_print_timings:   decode time = 13090.93 ms /   549 runs (   23.85 ms per run)
-whisper_print_timings:    total time = 32733.52 ms
+whisper_print_timings:     load time =   575.92 ms
+whisper_print_timings:      mel time =   230.60 ms
+whisper_print_timings:   sample time =    73.19 ms
+whisper_print_timings:   encode time = 19552.61 ms / 814.69 ms per layer
+whisper_print_timings:   decode time = 13249.96 ms / 552.08 ms per layer
+whisper_print_timings:    total time = 33686.27 ms
 ```
 </details>
 
 ## Real-time audio input example
 
 This is a naive example of performing real-time inference on audio from your microphone.
-The [stream](examples/stream) tool samples the audio every half a second and runs the transcription continuously.
+The [stream](examples/stream) tool samples the audio every half a second and runs the transcription continously.
 More info is available in [issue #10](https://github.com/ggerganov/whisper.cpp/issues/10).
 
 ```java
-make stream
 ./stream -m ./models/ggml-base.en.bin -t 8 --step 500 --length 5000
 ```
 
@@ -542,10 +316,6 @@ https://user-images.githubusercontent.com/1991296/194935793-76afede7-cfa8-48d8-a
 Adding the `--print-colors` argument will print the transcribed text using an experimental color coding strategy
 to highlight words with high or low confidence:
 
-```java
-./main -m models/ggml-base.en.bin -f samples/gb0.wav --print-colors
-```
-
 <img width="965" alt="image" src="https://user-images.githubusercontent.com/1991296/197356445-311c8643-9397-4e5e-b46e-0b4b4daa2530.png">
 
 ## Controlling the length of the generated text segments (experimental)
@@ -572,7 +342,7 @@ main: processing './samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 pr
 [00:00:10.020 --> 00:00:11.000]   country.
 ```
 
-## Word-level timestamp (experimental)
+## Word-level timestamp
 
 The `--max-len` argument can be used to obtain word-level timestamps. Simply use `-ml 1`:
 
@@ -613,32 +383,6 @@ main: processing './samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 pr
 [00:00:10.510 --> 00:00:11.000]  .
 ```
 
-## Speaker segmentation via tinydiarize (experimental)
-
-More information about this approach is available here: https://github.com/ggerganov/whisper.cpp/pull/1058
-
-Sample usage:
-
-```py
-# download a tinydiarize compatible model
-./models/download-ggml-model.sh small.en-tdrz
-
-# run as usual, adding the "-tdrz" command-line argument
-./main -f ./samples/a13.wav -m ./models/ggml-small.en-tdrz.bin -tdrz
-...
-main: processing './samples/a13.wav' (480000 samples, 30.0 sec), 4 threads, 1 processors, lang = en, task = transcribe, tdrz = 1, timestamps = 1 ...
-...
-[00:00:00.000 --> 00:00:03.800]   Okay Houston, we've had a problem here. [SPEAKER_TURN]
-[00:00:03.800 --> 00:00:06.200]   This is Houston. Say again please. [SPEAKER_TURN]
-[00:00:06.200 --> 00:00:08.260]   Uh Houston we've had a problem.
-[00:00:08.260 --> 00:00:11.320]   We've had a main beam up on a volt. [SPEAKER_TURN]
-[00:00:11.320 --> 00:00:13.820]   Roger main beam interval. [SPEAKER_TURN]
-[00:00:13.820 --> 00:00:15.100]   Uh uh [SPEAKER_TURN]
-[00:00:15.100 --> 00:00:18.020]   So okay stand, by thirteen we're looking at it. [SPEAKER_TURN]
-[00:00:18.020 --> 00:00:25.740]   Okay uh right now uh Houston the uh voltage is uh is looking good um.
-[00:00:27.620 --> 00:00:29.940]   And we had a a pretty large bank or so.
-```
-
 ## Karaoke-style movie generation (experimental)
 
 The [main](examples/main) example provides support for output of karaoke-style movies, where the
@@ -677,19 +421,6 @@ https://user-images.githubusercontent.com/1991296/199337538-b7b0c7a3-2753-4a88-a
 
 ---
 
-## Video comparison of different models
-
-Use the [extra/bench-wts.sh](https://github.com/ggerganov/whisper.cpp/blob/master/extra/bench-wts.sh) script to generate a video in the following format:
-
-```java
-./extra/bench-wts.sh samples/jfk.wav
-ffplay ./samples/jfk.wav.all.mp4
-```
-
-https://user-images.githubusercontent.com/1991296/223206245-2d36d903-cf8e-4f09-8c3b-eb9f9c39d6fc.mp4
-
----
-
 ## Benchmarks
 
 In order to have an objective comparison of the performance of the inference across different system configurations,
@@ -710,7 +441,7 @@ The original models are converted to a custom binary format. This allows to pack
 You can download the converted models using the [models/download-ggml-model.sh](models/download-ggml-model.sh) script
 or manually from here:
 
-- https://huggingface.co/ggerganov/whisper.cpp
+- https://huggingface.co/datasets/ggerganov/whisper.cpp
 - https://ggml.ggerganov.com
 
 For more details, see the conversion script [models/convert-pt-to-ggml.py](models/convert-pt-to-ggml.py) or the README
@@ -720,21 +451,9 @@ in [models](models).
 
 - [X] Rust: [tazz4843/whisper-rs](https://github.com/tazz4843/whisper-rs) | [#310](https://github.com/ggerganov/whisper.cpp/discussions/310)
 - [X] Javascript: [bindings/javascript](bindings/javascript) | [#309](https://github.com/ggerganov/whisper.cpp/discussions/309)
-  - React Native (iOS / Android): [whisper.rn](https://github.com/mybigday/whisper.rn)
 - [X] Go: [bindings/go](bindings/go) | [#312](https://github.com/ggerganov/whisper.cpp/discussions/312)
-- [X] Java:
-  - [GiviMAD/whisper-jni](https://github.com/GiviMAD/whisper-jni)
-- [X] Ruby: [bindings/ruby](bindings/ruby) | [#507](https://github.com/ggerganov/whisper.cpp/discussions/507)
 - [X] Objective-C / Swift: [ggerganov/whisper.spm](https://github.com/ggerganov/whisper.spm) | [#313](https://github.com/ggerganov/whisper.cpp/discussions/313)
-  - [exPHAT/SwiftWhisper](https://github.com/exPHAT/SwiftWhisper)
-- [X] .NET: | [#422](https://github.com/ggerganov/whisper.cpp/discussions/422)
-  - [sandrohanea/whisper.net](https://github.com/sandrohanea/whisper.net)
-  - [NickDarvey/whisper](https://github.com/NickDarvey/whisper)
-- [X] Python: | [#9](https://github.com/ggerganov/whisper.cpp/issues/9)
-  - [stlukey/whispercpp.py](https://github.com/stlukey/whispercpp.py) (Cython)
-  - [aarnphm/whispercpp](https://github.com/aarnphm/whispercpp) (Pybind11)
-- [X] R: [bnosac/audio.whisper](https://github.com/bnosac/audio.whisper)
-- [X] Unity: [macoron/whisper.unity](https://github.com/Macoron/whisper.unity)
+- [ ] Python: soon | [WIP](https://github.com/ggerganov/whisper.cpp/issues/9)
 
 ## Examples
 
@@ -748,7 +467,6 @@ Some of the examples are even ported to run in the browser using WebAssembly. Ch
 | [stream](examples/stream) | [stream.wasm](examples/stream.wasm) | Real-time transcription of raw microphone capture |
 | [command](examples/command) | [command.wasm](examples/command.wasm) | Basic voice assistant example for receiving voice commands from the mic |
 | [talk](examples/talk) | [talk.wasm](examples/talk.wasm) | Talk with a GPT-2 bot |
-| [talk-llama](examples/talk-llama) | | Talk with a LLaMA bot |
 | [whisper.objc](examples/whisper.objc) | | iOS mobile application using whisper.cpp |
 | [whisper.swiftui](examples/whisper.swiftui) | | SwiftUI iOS / macOS application using whisper.cpp |
 | [whisper.android](examples/whisper.android) | | Android mobile application using whisper.cpp |

bindings/go/.gitignore

@@ -1,2 +1,3 @@
 build
 models
+go.sum

bindings/go/Makefile

@@ -1,27 +1,28 @@
-BUILD_DIR := build
-MODELS_DIR := models
+CMAKE := $(shell which cmake)
+BUILD_DIR := "build"
+MODELS_DIR := "models"
 EXAMPLES_DIR := $(wildcard examples/*)
-INCLUDE_PATH := $(abspath ../..)
-LIBRARY_PATH := $(abspath ../..)
+C_INCLUDE_PATH := "../.."
 
 all: clean whisper examples
 
 whisper: mkdir
 	@echo Build whisper
-	@${MAKE} -C ../.. libwhisper.a
+	@${CMAKE} -S ../.. -B ${BUILD_DIR} -D BUILD_SHARED_LIBS=off -D WHISPER_NO_AVX2=on
+	@${CMAKE} --build ${BUILD_DIR} --target whisper
 
 test: model-small whisper modtidy
-	@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} go test -v .
-	@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} go test -v ./pkg/whisper/...
+	@go test -v .
+	@go test -v ./pkg/whisper/...
 
 examples: $(EXAMPLES_DIR)
 
 model-small: mkdir examples/go-model-download
-	@${BUILD_DIR}/go-model-download -out models ggml-small.en.bin
+	@${BUILD_DIR}/go-model-download -out models small.en
 
 $(EXAMPLES_DIR): mkdir whisper modtidy
 	@echo Build example $(notdir $@)
-	@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} go build ${BUILD_FLAGS} -o ${BUILD_DIR}/$(notdir $@) ./$@
+	@go build ${BUILD_FLAGS} -o ${BUILD_DIR}/$(notdir $@) ./$@
 
 mkdir:
 	@echo Mkdir ${BUILD_DIR}

bindings/go/README.md

@@ -31,7 +31,7 @@ func main() {
 	if err != nil {
 		panic(err)
 	}
-	if err := context.Process(samples, nil, nil); err != nil {
+	if err := context.Process(samples, nil); err != nil {
 		return err
 	}
 
@@ -74,27 +74,4 @@ And you can then test a model against samples with the following command:
 ./build/go-whisper -model models/ggml-tiny.en.bin samples/jfk.wav 
 ```
 
-## Using the bindings
-
-To use the bindings in your own software,
-
-  1. Import `github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper` (or `github.com/ggerganov/whisper.cpp/bindings/go` into your package;
-  2. Compile `libwhisper.a` (you can use `make whisper` in the `bindings/go` directory);
-  3. Link your go binary against whisper by setting the environment variables `C_INCLUDE_PATH` and `LIBRARY_PATH`
-     to point to the `whisper.h` file directory and `libwhisper.a` file directory respectively.
-
-Look at the `Makefile` in the `bindings/go` directory for an example.
-
-The API Documentation:
-
-  * https://pkg.go.dev/github.com/ggerganov/whisper.cpp/bindings/go
-  * https://pkg.go.dev/github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper
-
-Getting help:
-
-  * Follow the discussion for the go bindings [here](https://github.com/ggerganov/whisper.cpp/discussions/312)
-
-## License
-
-The license for the Go bindings is the same as the license for the rest of the whisper.cpp project, which is the MIT License. See the `LICENSE` file for more details.
 

bindings/go/examples/go-model-download/main.go

@@ -17,14 +17,15 @@ import (
 // CONSTANTS
 
 const (
-	srcUrl  = "https://huggingface.co/ggerganov/whisper.cpp/resolve/main" // The location of the models
+	srcUrl        = "https://huggingface.co/"                           // The location of the models
+	srcPathPrefix = "/datasets/ggerganov/whisper.cpp/resolve/main/ggml" // Filename prefix
 	srcExt        = ".bin"                                              // Filename extension
 	bufSize       = 1024 * 64                                           // Size of the buffer used for downloading the model
 )
 
 var (
 	// The models which will be downloaded, if no model is specified as an argument
-	modelNames = []string{"ggml-tiny.en", "ggml-tiny", "ggml-base.en", "ggml-base", "ggml-small.en", "ggml-small", "ggml-medium.en", "ggml-medium", "ggml-large-v1", "ggml-large"}
+	modelNames = []string{"tiny.en", "tiny", "base.en", "base", "small.en", "small", "medium.en", "medium", "large-v1", "large"}
 )
 
 var (
@@ -122,14 +123,11 @@ func GetModels() []string {
 
 // URLForModel returns the URL for the given model on huggingface.co
 func URLForModel(model string) (string, error) {
-	if filepath.Ext(model) != srcExt {
-		model += srcExt
-	}
 	url, err := url.Parse(srcUrl)
 	if err != nil {
 		return "", err
 	} else {
-		url.Path = filepath.Join(url.Path, model)
+		url.Path = srcPathPrefix + "-" + model + srcExt
 	}
 	return url.String(), nil
 }

bindings/go/examples/go-whisper/color.go

@@ -1,22 +0,0 @@
-package main
-
-import "fmt"
-
-///////////////////////////////////////////////////////////////////////////////
-// CONSTANTS
-
-const (
-	Reset     = "\033[0m"
-	RGBPrefix = "\033[38;5;" // followed by RGB values in decimal format separated by colons
-	RGBSuffix = "m"
-)
-
-///////////////////////////////////////////////////////////////////////////////
-// PUBLIC METHODS
-
-// Colorize text with RGB values, from 0 to 23
-func Colorize(text string, v int) string {
-	// https://en.wikipedia.org/wiki/ANSI_escape_code#8-bit
-	// Grayscale colors are in the range 232-255
-	return RGBPrefix + fmt.Sprint(v%24+232) + RGBSuffix + text + Reset
-}

bindings/go/examples/go-whisper/flags.go

@@ -2,12 +2,6 @@ package main
 
 import (
 	"flag"
-	"fmt"
-	"strings"
-	"time"
-
-	// Packages
-	whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
 )
 
 ///////////////////////////////////////////////////////////////////////////////
@@ -48,26 +42,6 @@ func (flags *Flags) GetLanguage() string {
 	return flags.Lookup("language").Value.String()
 }
 
-func (flags *Flags) IsTranslate() bool {
-	return flags.Lookup("translate").Value.(flag.Getter).Get().(bool)
-}
-
-func (flags *Flags) GetOffset() time.Duration {
-	return flags.Lookup("offset").Value.(flag.Getter).Get().(time.Duration)
-}
-
-func (flags *Flags) GetDuration() time.Duration {
-	return flags.Lookup("duration").Value.(flag.Getter).Get().(time.Duration)
-}
-
-func (flags *Flags) GetThreads() uint {
-	return flags.Lookup("threads").Value.(flag.Getter).Get().(uint)
-}
-
-func (flags *Flags) GetOut() string {
-	return strings.ToLower(flags.Lookup("out").Value.String())
-}
-
 func (flags *Flags) IsSpeedup() bool {
 	return flags.Lookup("speedup").Value.String() == "true"
 }
@@ -76,81 +50,12 @@ func (flags *Flags) IsTokens() bool {
 	return flags.Lookup("tokens").Value.String() == "true"
 }
 
-func (flags *Flags) IsColorize() bool {
-	return flags.Lookup("colorize").Value.String() == "true"
-}
-
-func (flags *Flags) GetMaxLen() uint {
-	return flags.Lookup("max-len").Value.(flag.Getter).Get().(uint)
-}
-
-func (flags *Flags) GetMaxTokens() uint {
-	return flags.Lookup("max-tokens").Value.(flag.Getter).Get().(uint)
-}
-
-func (flags *Flags) GetWordThreshold() float32 {
-	return float32(flags.Lookup("word-thold").Value.(flag.Getter).Get().(float64))
-}
-
-func (flags *Flags) SetParams(context whisper.Context) error {
-	if lang := flags.GetLanguage(); lang != "" && lang != "auto" {
-		fmt.Fprintf(flags.Output(), "Setting language to %q\n", lang)
-		if err := context.SetLanguage(lang); err != nil {
-			return err
-		}
-	}
-	if flags.IsTranslate() && context.IsMultilingual() {
-		fmt.Fprintf(flags.Output(), "Setting translate to true\n")
-		context.SetTranslate(true)
-	}
-	if offset := flags.GetOffset(); offset != 0 {
-		fmt.Fprintf(flags.Output(), "Setting offset to %v\n", offset)
-		context.SetOffset(offset)
-	}
-	if duration := flags.GetDuration(); duration != 0 {
-		fmt.Fprintf(flags.Output(), "Setting duration to %v\n", duration)
-		context.SetDuration(duration)
-	}
-	if flags.IsSpeedup() {
-		fmt.Fprintf(flags.Output(), "Setting speedup to true\n")
-		context.SetSpeedup(true)
-	}
-	if threads := flags.GetThreads(); threads != 0 {
-		fmt.Fprintf(flags.Output(), "Setting threads to %d\n", threads)
-		context.SetThreads(threads)
-	}
-	if max_len := flags.GetMaxLen(); max_len != 0 {
-		fmt.Fprintf(flags.Output(), "Setting max_segment_length to %d\n", max_len)
-		context.SetMaxSegmentLength(max_len)
-	}
-	if max_tokens := flags.GetMaxTokens(); max_tokens != 0 {
-		fmt.Fprintf(flags.Output(), "Setting max_tokens to %d\n", max_tokens)
-		context.SetMaxTokensPerSegment(max_tokens)
-	}
-	if word_threshold := flags.GetWordThreshold(); word_threshold != 0 {
-		fmt.Fprintf(flags.Output(), "Setting word_threshold to %f\n", word_threshold)
-		context.SetTokenThreshold(word_threshold)
-	}
-
-	// Return success
-	return nil
-}
-
 ///////////////////////////////////////////////////////////////////////////////
 // PRIVATE METHODS
 
 func registerFlags(flag *Flags) {
 	flag.String("model", "", "Path to the model file")
-	flag.String("language", "", "Spoken language")
-	flag.Bool("translate", false, "Translate from source language to english")
-	flag.Duration("offset", 0, "Time offset")
-	flag.Duration("duration", 0, "Duration of audio to process")
-	flag.Uint("threads", 0, "Number of threads to use")
+	flag.String("language", "", "Language")
 	flag.Bool("speedup", false, "Enable speedup")
-	flag.Uint("max-len", 0, "Maximum segment length in characters")
-	flag.Uint("max-tokens", 0, "Maximum tokens per segment")
-	flag.Float64("word-thold", 0, "Maximum segment score")
 	flag.Bool("tokens", false, "Display tokens")
-	flag.Bool("colorize", false, "Colorize tokens")
-	flag.String("out", "", "Output format (srt, none or leave as empty string)")
 }

bindings/go/examples/go-whisper/main.go

@@ -35,7 +35,8 @@ func main() {
 
 	// Process files
 	for _, filename := range flags.Args() {
-		if err := Process(model, filename, flags); err != nil {
+		fmt.Println("Processing", filename)
+		if err := Process(model, filename, flags.GetLanguage(), flags.IsSpeedup(), flags.IsTokens()); err != nil {
 			fmt.Fprintln(os.Stderr, err)
 			continue
 		}

bindings/go/examples/go-whisper/process.go

@@ -11,7 +11,7 @@ import (
 	wav "github.com/go-audio/wav"
 )
 
-func Process(model whisper.Model, path string, flags *Flags) error {
+func Process(model whisper.Model, path string, lang string, speedup, tokens bool) error {
 	var data []float32
 
 	// Create processing context
@@ -20,22 +20,14 @@ func Process(model whisper.Model, path string, flags *Flags) error {
 		return err
 	}
 
-	// Set the parameters
-	if err := flags.SetParams(context); err != nil {
-		return err
-	}
-
-	fmt.Printf("\n%s\n", context.SystemInfo())
-
 	// Open the file
-	fmt.Fprintf(flags.Output(), "Loading %q\n", path)
 	fh, err := os.Open(path)
 	if err != nil {
 		return err
 	}
 	defer fh.Close()
 
-	// Decode the WAV file - load the full buffer
+	// Decode the WAV file
 	dec := wav.NewDecoder(fh)
 	if buf, err := dec.FullPCMBuffer(); err != nil {
 		return err
@@ -47,86 +39,42 @@ func Process(model whisper.Model, path string, flags *Flags) error {
 		data = buf.AsFloat32Buffer().Data
 	}
 
-	// Segment callback when -tokens is specified
+	// Set the parameters
 	var cb whisper.SegmentCallback
-	if flags.IsTokens() {
+	if lang != "" {
+		if err := context.SetLanguage(lang); err != nil {
+			return err
+		}
+	}
+	if speedup {
+		context.SetSpeedup(true)
+	}
+	if tokens {
 		cb = func(segment whisper.Segment) {
-			fmt.Fprintf(flags.Output(), "%02d [%6s->%6s] ", segment.Num, segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
+			fmt.Printf("%02d [%6s->%6s] ", segment.Num, segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
 			for _, token := range segment.Tokens {
-				if flags.IsColorize() && context.IsText(token) {
-					fmt.Fprint(flags.Output(), Colorize(token.Text, int(token.P*24.0)), " ")
-				} else {
-					fmt.Fprint(flags.Output(), token.Text, " ")
+				fmt.Printf("%q ", token.Text)
 			}
-			}
-			fmt.Fprintln(flags.Output(), "")
-			fmt.Fprintln(flags.Output(), "")
+			fmt.Println("")
 		}
 	}
 
 	// Process the data
-	fmt.Fprintf(flags.Output(), "  ...processing %q\n", path)
-	context.ResetTimings()
-	if err := context.Process(data, cb, nil); err != nil {
+	if err := context.Process(data, cb); err != nil {
 		return err
 	}
 
-	context.PrintTimings()
-
 	// Print out the results
-	switch {
-	case flags.GetOut() == "srt":
-		return OutputSRT(os.Stdout, context)
-	case flags.GetOut() == "none":
-		return nil
-	default:
-		return Output(os.Stdout, context, flags.IsColorize())
-	}
-}
-
-// Output text as SRT file
-func OutputSRT(w io.Writer, context whisper.Context) error {
-	n := 1
 	for {
 		segment, err := context.NextSegment()
 		if err == io.EOF {
-			return nil
+			break
 		} else if err != nil {
 			return err
 		}
-		fmt.Fprintln(w, n)
-		fmt.Fprintln(w, srtTimestamp(segment.Start), " --> ", srtTimestamp(segment.End))
-		fmt.Fprintln(w, segment.Text)
-		fmt.Fprintln(w, "")
-		n++
+		fmt.Printf("[%6s->%6s] %s\n", segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond), segment.Text)
 	}
-}
 
-// Output text to terminal
-func Output(w io.Writer, context whisper.Context, colorize bool) error {
-	for {
-		segment, err := context.NextSegment()
-		if err == io.EOF {
+	// Return success
 	return nil
-		} else if err != nil {
-			return err
-		}
-		fmt.Fprintf(w, "[%6s->%6s]", segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
-		if colorize {
-			for _, token := range segment.Tokens {
-				if !context.IsText(token) {
-					continue
-				}
-				fmt.Fprint(w, " ", Colorize(token.Text, int(token.P*24.0)))
-			}
-			fmt.Fprint(w, "\n")
-		} else {
-			fmt.Fprintln(w, " ", segment.Text)
-		}
-	}
-}
-
-// Return srtTimestamp
-func srtTimestamp(t time.Duration) string {
-	return fmt.Sprintf("%02d:%02d:%02d,%03d", t/time.Hour, (t%time.Hour)/time.Minute, (t%time.Minute)/time.Second, (t%time.Second)/time.Millisecond)
 }

bindings/go/go.sum

@@ -1,23 +0,0 @@
-github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
-github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/go-audio/audio v1.0.0 h1:zS9vebldgbQqktK4H0lUqWrG8P0NxCJVqcj7ZpNnwd4=
-github.com/go-audio/audio v1.0.0/go.mod h1:6uAu0+H2lHkwdGsAY+j2wHPNPpPoeg5AaEFh9FlA+Zs=
-github.com/go-audio/riff v1.0.0 h1:d8iCGbDvox9BfLagY94fBynxSPHO80LmZCaOsmKxokA=
-github.com/go-audio/riff v1.0.0/go.mod h1:l3cQwc85y79NQFCRB7TiPoNiaijp6q8Z0Uv38rVG498=
-github.com/go-audio/wav v1.1.0 h1:jQgLtbqBzY7G+BM8fXF7AHUk1uHUviWS4X39d5rsL2g=
-github.com/go-audio/wav v1.1.0/go.mod h1:mpe9qfwbScEbkd8uybLuIpTgHyrISw/OTuvjUW2iGtE=
-github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
-github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
-github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSSt89Yw=
-github.com/stretchr/objx v0.5.0/go.mod h1:Yh+to48EsGEfYuaHDzXPcE3xhTkx73EhmCGUpEOglKo=
-github.com/stretchr/testify v1.7.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-github.com/stretchr/testify v1.8.0/go.mod h1:yNjHg4UonilssWZ8iaSj1OCr/vHnekPRkoO+kdMU+MU=
-github.com/stretchr/testify v1.8.1 h1:w7B6lhMri9wdJUVmEZPGGhZzrYTPvgJArz7wNPgYKsk=
-github.com/stretchr/testify v1.8.1/go.mod h1:w2LPCIKwWwSfY2zedu0+kehJoqGctiVI29o6fzry7u4=
-gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
-gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
-gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
-gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
-gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

bindings/go/params.go

@@ -1,5 +1,8 @@
 package whisper
 
+// This file defines the whisper_token, whisper_token_data and whisper_full_params
+// structures, which are used by the whisper_full() function.
+
 import (
 	"fmt"
 )
@@ -19,10 +22,6 @@ func (p *Params) SetTranslate(v bool) {
 	p.translate = toBool(v)
 }
 
-func (p *Params) SetSplitOnWord(v bool) {
-	p.split_on_word = toBool(v)
-}
-
 func (p *Params) SetNoContext(v bool) {
 	p.no_context = toBool(v)
 }
@@ -51,12 +50,7 @@ func (p *Params) SetSpeedup(v bool) {
 	p.speed_up = toBool(v)
 }
 
-// Set language id
 func (p *Params) SetLanguage(lang int) error {
-	if lang == -1 {
-		p.language = nil
-		return nil
-	}
 	str := C.whisper_lang_str(C.int(lang))
 	if str == nil {
 		return ErrInvalidLanguage
@@ -66,7 +60,6 @@ func (p *Params) SetLanguage(lang int) error {
 	return nil
 }
 
-// Get language id
 func (p *Params) Language() int {
 	if p.language == nil {
 		return -1
@@ -74,50 +67,18 @@ func (p *Params) Language() int {
 	return int(C.whisper_lang_id(p.language))
 }
 
-// Threads available
-func (p *Params) Threads() int {
-	return int(p.n_threads)
-}
-
-// Set number of threads to use
 func (p *Params) SetThreads(threads int) {
 	p.n_threads = C.int(threads)
 }
 
-// Set start offset in ms
 func (p *Params) SetOffset(offset_ms int) {
 	p.offset_ms = C.int(offset_ms)
 }
 
-// Set audio duration to process in ms
 func (p *Params) SetDuration(duration_ms int) {
 	p.duration_ms = C.int(duration_ms)
 }
 
-// Set timestamp token probability threshold (~0.01)
-func (p *Params) SetTokenThreshold(t float32) {
-	p.thold_pt = C.float(t)
-}
-
-// Set timestamp token sum probability threshold (~0.01)
-func (p *Params) SetTokenSumThreshold(t float32) {
-	p.thold_ptsum = C.float(t)
-}
-
-// Set max segment length in characters
-func (p *Params) SetMaxSegmentLength(n int) {
-	p.max_len = C.int(n)
-}
-
-func (p *Params) SetTokenTimestamps(b bool) {
-	p.token_timestamps = toBool(b)
-}
-
-// Set max tokens per segment (0 = no limit)
-func (p *Params) SetMaxTokensPerSegment(n int) {
-	p.max_tokens = C.int(n)
-}
-
 ///////////////////////////////////////////////////////////////////////////////
 // PRIVATE METHODS
 

bindings/go/pkg/whisper/consts.go

@@ -15,7 +15,6 @@ var (
 	ErrInternalAppError    = errors.New("internal application error")
 	ErrProcessingFailed    = errors.New("processing failed")
 	ErrUnsupportedLanguage = errors.New("unsupported language")
-	ErrModelNotMultilingual = errors.New("model is not multilingual")
 )
 
 ///////////////////////////////////////////////////////////////////////////////

bindings/go/pkg/whisper/context.go

@@ -1,9 +1,7 @@
 package whisper
 
 import (
-	"fmt"
 	"io"
-	"runtime"
 	"strings"
 	"time"
 
@@ -26,7 +24,7 @@ var _ Context = (*context)(nil)
 ///////////////////////////////////////////////////////////////////////////////
 // LIFECYCLE
 
-func newContext(model *model, params whisper.Params) (Context, error) {
+func NewContext(model *model, params whisper.Params) (Context, error) {
 	context := new(context)
 	context.model = model
 	context.params = params
@@ -43,13 +41,7 @@ func (context *context) SetLanguage(lang string) error {
 	if context.model.ctx == nil {
 		return ErrInternalAppError
 	}
-	if !context.model.IsMultilingual() {
-		return ErrModelNotMultilingual
-	}
-
-	if lang == "auto" {
-		context.params.SetLanguage(-1)
-	} else if id := context.model.ctx.Whisper_lang_id(lang); id < 0 {
+	if id := context.model.ctx.Whisper_lang_id(lang); id < 0 {
 		return ErrUnsupportedLanguage
 	} else if err := context.params.SetLanguage(id); err != nil {
 		return err
@@ -58,114 +50,23 @@ func (context *context) SetLanguage(lang string) error {
 	return nil
 }
 
-func (context *context) IsMultilingual() bool {
-	return context.model.IsMultilingual()
-}
-
 // Get language
 func (context *context) Language() string {
-	id := context.params.Language()
-	if id == -1 {
-		return "auto"
-	}
 	return whisper.Whisper_lang_str(context.params.Language())
 }
 
-// Set translate flag
-func (context *context) SetTranslate(v bool) {
-	context.params.SetTranslate(v)
-}
-
 // Set speedup flag
 func (context *context) SetSpeedup(v bool) {
 	context.params.SetSpeedup(v)
 }
 
-func (context *context) SetSplitOnWord(v bool) {
-        context.params.SetSplitOnWord(v)
-}
-
-// Set number of threads to use
-func (context *context) SetThreads(v uint) {
-	context.params.SetThreads(int(v))
-}
-
-// Set time offset
-func (context *context) SetOffset(v time.Duration) {
-	context.params.SetOffset(int(v.Milliseconds()))
-}
-
-// Set duration of audio to process
-func (context *context) SetDuration(v time.Duration) {
-	context.params.SetDuration(int(v.Milliseconds()))
-}
-
-// Set timestamp token probability threshold (~0.01)
-func (context *context) SetTokenThreshold(t float32) {
-	context.params.SetTokenThreshold(t)
-}
-
-// Set timestamp token sum probability threshold (~0.01)
-func (context *context) SetTokenSumThreshold(t float32) {
-	context.params.SetTokenSumThreshold(t)
-}
-
-// Set max segment length in characters
-func (context *context) SetMaxSegmentLength(n uint) {
-	context.params.SetMaxSegmentLength(int(n))
-}
-
-// Set token timestamps flag
-func (context *context) SetTokenTimestamps(b bool) {
-	context.params.SetTokenTimestamps(b)
-}
-
-// Set max tokens per segment (0 = no limit)
-func (context *context) SetMaxTokensPerSegment(n uint) {
-	context.params.SetMaxTokensPerSegment(int(n))
-}
-
-// ResetTimings resets the mode timings. Should be called before processing
-func (context *context) ResetTimings() {
-	context.model.ctx.Whisper_reset_timings()
-}
-
-// PrintTimings prints the model timings to stdout.
-func (context *context) PrintTimings() {
-	context.model.ctx.Whisper_print_timings()
-}
-
-// SystemInfo returns the system information
-func (context *context) SystemInfo() string {
-	return fmt.Sprintf("system_info: n_threads = %d / %d | %s\n",
-		context.params.Threads(),
-		runtime.NumCPU(),
-		whisper.Whisper_print_system_info(),
-	)
-}
-
-// Use mel data at offset_ms to try and auto-detect the spoken language
-// Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
-// Returns the probabilities of all languages.
-func (context *context) WhisperLangAutoDetect(offset_ms int, n_threads int) ([]float32, error) {
-	langProbs, err := context.model.ctx.Whisper_lang_auto_detect(offset_ms, n_threads)
-	if err != nil {
-		return nil, err
-	}
-	return langProbs, nil
-}
-
 // Process new sample data and return any errors
-func (context *context) Process(
-	data []float32,
-	callNewSegment SegmentCallback,
-	callProgress ProgressCallback,
-) error {
+func (context *context) Process(data []float32, cb SegmentCallback) error {
 	if context.model.ctx == nil {
 		return ErrInternalAppError
 	}
 	// If the callback is defined then we force on single_segment mode
-	if callNewSegment != nil {
+	if cb != nil {
 		context.params.SetSingleSegment(true)
 	}
 
@@ -173,28 +74,24 @@ func (context *context) Process(
 	processors := 0
 	if processors > 1 {
 		if err := context.model.ctx.Whisper_full_parallel(context.params, data, processors, nil, func(new int) {
-			if callNewSegment != nil {
+			if cb != nil {
 				num_segments := context.model.ctx.Whisper_full_n_segments()
 				s0 := num_segments - new
 				for i := s0; i < num_segments; i++ {
-					callNewSegment(toSegment(context.model.ctx, i))
+					cb(toSegment(context.model.ctx, i))
 				}
 			}
 		}); err != nil {
 			return err
 		}
 	} else if err := context.model.ctx.Whisper_full(context.params, data, nil, func(new int) {
-		if callNewSegment != nil {
+		if cb != nil {
 			num_segments := context.model.ctx.Whisper_full_n_segments()
 			s0 := num_segments - new
 			for i := s0; i < num_segments; i++ {
-				callNewSegment(toSegment(context.model.ctx, i))
+				cb(toSegment(context.model.ctx, i))
 			}
 		}
-	}, func(progress int) {
-		if callProgress != nil {
-			callProgress(progress)
-		}
 	}); err != nil {
 		return err
 	}
@@ -222,65 +119,6 @@ func (context *context) NextSegment() (Segment, error) {
 	return result, nil
 }
 
-// Test for text tokens
-func (context *context) IsText(t Token) bool {
-	switch {
-	case context.IsBEG(t):
-		return false
-	case context.IsSOT(t):
-		return false
-	case whisper.Token(t.Id) >= context.model.ctx.Whisper_token_eot():
-		return false
-	case context.IsPREV(t):
-		return false
-	case context.IsSOLM(t):
-		return false
-	case context.IsNOT(t):
-		return false
-	default:
-		return true
-	}
-}
-
-// Test for "begin" token
-func (context *context) IsBEG(t Token) bool {
-	return whisper.Token(t.Id) == context.model.ctx.Whisper_token_beg()
-}
-
-// Test for "start of transcription" token
-func (context *context) IsSOT(t Token) bool {
-	return whisper.Token(t.Id) == context.model.ctx.Whisper_token_sot()
-}
-
-// Test for "end of transcription" token
-func (context *context) IsEOT(t Token) bool {
-	return whisper.Token(t.Id) == context.model.ctx.Whisper_token_eot()
-}
-
-// Test for "start of prev" token
-func (context *context) IsPREV(t Token) bool {
-	return whisper.Token(t.Id) == context.model.ctx.Whisper_token_prev()
-}
-
-// Test for "start of lm" token
-func (context *context) IsSOLM(t Token) bool {
-	return whisper.Token(t.Id) == context.model.ctx.Whisper_token_solm()
-}
-
-// Test for "No timestamps" token
-func (context *context) IsNOT(t Token) bool {
-	return whisper.Token(t.Id) == context.model.ctx.Whisper_token_not()
-}
-
-// Test for token associated with a specific language
-func (context *context) IsLANG(t Token, lang string) bool {
-	if id := context.model.ctx.Whisper_lang_id(lang); id >= 0 {
-		return whisper.Token(t.Id) == context.model.ctx.Whisper_token_lang(id)
-	} else {
-		return false
-	}
-}
-
 ///////////////////////////////////////////////////////////////////////////////
 // PRIVATE METHODS
 
@@ -297,14 +135,10 @@ func toSegment(ctx *whisper.Context, n int) Segment {
 func toTokens(ctx *whisper.Context, n int) []Token {
 	result := make([]Token, ctx.Whisper_full_n_tokens(n))
 	for i := 0; i < len(result); i++ {
-		data := ctx.Whisper_full_get_token_data(n, i)
-
 		result[i] = Token{
 			Id:   int(ctx.Whisper_full_get_token_id(n, i)),
-			Text:  ctx.Whisper_full_get_token_text(n, i),
+			Text: strings.TrimSpace(ctx.Whisper_full_get_token_text(n, i)),
 			P:    ctx.Whisper_full_get_token_p(n, i),
-			Start: time.Duration(data.T0()) * time.Millisecond * 10,
-			End:   time.Duration(data.T1()) * time.Millisecond * 10,
 		}
 	}
 	return result

bindings/go/pkg/whisper/interface.go

@@ -12,10 +12,6 @@ import (
 // time. It is called during the Process function
 type SegmentCallback func(Segment)
 
-// ProgressCallback is the callback function for reporting progress during
-// processing. It is called during the Process function
-type ProgressCallback func(int)
-
 // Model is the interface to a whisper model. Create a new model with the
 // function whisper.New(string)
 type Model interface {
@@ -24,54 +20,24 @@ type Model interface {
 	// Return a new speech-to-text context.
 	NewContext() (Context, error)
 
-	// Return true if the model is multilingual.
-	IsMultilingual() bool
-
 	// Return all languages supported.
 	Languages() []string
 }
 
 // Context is the speach recognition context.
 type Context interface {
-	SetLanguage(string) error // Set the language to use for speech recognition, use "auto" for auto detect language.
-	SetTranslate(bool)        // Set translate flag
-	IsMultilingual() bool     // Return true if the model is multilingual.
+	SetLanguage(string) error // Set the language to use for speech recognition.
 	Language() string         // Get language
-
-	SetOffset(time.Duration)      // Set offset
-	SetDuration(time.Duration)    // Set duration
-	SetThreads(uint)              // Set number of threads to use
 	SetSpeedup(bool)          // Set speedup flag
-	SetSplitOnWord(bool)          // Set split on word flag
-	SetTokenThreshold(float32)    // Set timestamp token probability threshold
-	SetTokenSumThreshold(float32) // Set timestamp token sum probability threshold
-	SetMaxSegmentLength(uint)     // Set max segment length in characters
-	SetTokenTimestamps(bool)      // Set token timestamps flag
-	SetMaxTokensPerSegment(uint)  // Set max tokens per segment (0 = no limit)
 
 	// Process mono audio data and return any errors.
 	// If defined, newly generated segments are passed to the
 	// callback function during processing.
-	Process([]float32, SegmentCallback, ProgressCallback) error
+	Process([]float32, SegmentCallback) error
 
 	// After process is called, return segments until the end of the stream
 	// is reached, when io.EOF is returned.
 	NextSegment() (Segment, error)
-
-	IsBEG(Token) bool          // Test for "begin" token
-	IsSOT(Token) bool          // Test for "start of transcription" token
-	IsEOT(Token) bool          // Test for "end of transcription" token
-	IsPREV(Token) bool         // Test for "start of prev" token
-	IsSOLM(Token) bool         // Test for "start of lm" token
-	IsNOT(Token) bool          // Test for "No timestamps" token
-	IsLANG(Token, string) bool // Test for token associated with a specific language
-	IsText(Token) bool         // Test for text token
-
-	// Timings
-	PrintTimings()
-	ResetTimings()
-
-	SystemInfo() string
 }
 
 // Segment is the text result of a speech recognition.
@@ -94,5 +60,4 @@ type Token struct {
 	Id   int
 	Text string
 	P    float32
-	Start, End time.Duration
 }

bindings/go/pkg/whisper/model.go

@@ -23,7 +23,7 @@ var _ Model = (*model)(nil)
 ///////////////////////////////////////////////////////////////////////////////
 // LIFECYCLE
 
-func New(path string) (Model, error) {
+func New(path string) (*model, error) {
 	model := new(model)
 	if _, err := os.Stat(path); err != nil {
 		return nil, err
@@ -64,11 +64,6 @@ func (model *model) String() string {
 ///////////////////////////////////////////////////////////////////////////////
 // PUBLIC METHODS
 
-// Return true if model is multilingual (language and translation options are supported)
-func (model *model) IsMultilingual() bool {
-	return model.ctx.Whisper_is_multilingual() != 0
-}
-
 // Return all recognized languages. Initially it is set to auto-detect
 func (model *model) Languages() []string {
 	result := make([]string, 0, whisper.Whisper_lang_max_id())
@@ -94,8 +89,7 @@ func (model *model) NewContext() (Context, error) {
 	params.SetPrintRealtime(false)
 	params.SetPrintTimestamps(false)
 	params.SetThreads(runtime.NumCPU())
-	params.SetNoContext(true)
 
 	// Return new context
-	return newContext(model, params)
+	return NewContext(model, params)
 }

bindings/go/whisper.go

@@ -9,37 +9,28 @@ import (
 // CGO
 
 /*
-#cgo LDFLAGS: -lwhisper -lm -lstdc++
+#cgo CFLAGS: -I${SRCDIR}/../..
+#cgo LDFLAGS: -L${SRCDIR}/build -lwhisper -lm -lstdc++
 #cgo darwin LDFLAGS: -framework Accelerate
 #include <whisper.h>
 #include <stdlib.h>
 
 extern void callNewSegment(void* user_data, int new);
-extern void callProgress(void* user_data, int progress);
 extern bool callEncoderBegin(void* user_data);
 
 // Text segment callback
 // Called on every newly generated text segment
 // Use the whisper_full_...() functions to obtain the text segments
-static void whisper_new_segment_cb(struct whisper_context* ctx, struct whisper_state* state, int n_new, void* user_data) {
+static void whisper_new_segment_cb(struct whisper_context* ctx, int n_new, void* user_data) {
     if(user_data != NULL && ctx != NULL) {
         callNewSegment(user_data, n_new);
     }
 }
 
-// Progress callback
-// Called on every newly generated text segment
-// Use the whisper_full_...() functions to obtain the text segments
-static void whisper_progress_cb(struct whisper_context* ctx, struct whisper_state* state, int progress, void* user_data) {
-    if(user_data != NULL && ctx != NULL) {
-        callProgress(user_data, progress);
-    }
-}
-
 // Encoder begin callback
 // If not NULL, called before the encoder starts
 // If it returns false, the computation is aborted
-static bool whisper_encoder_begin_cb(struct whisper_context* ctx, struct whisper_state* state, void* user_data) {
+static bool whisper_encoder_begin_cb(struct whisper_context* ctx, void* user_data) {
     if(user_data != NULL && ctx != NULL) {
         return callEncoderBegin(user_data);
     }
@@ -53,8 +44,6 @@ static struct whisper_full_params whisper_full_default_params_cb(struct whisper_
 	params.new_segment_callback_user_data = (void*)(ctx);
 	params.encoder_begin_callback = whisper_encoder_begin_cb;
 	params.encoder_begin_callback_user_data = (void*)(ctx);
-	params.progress_callback = whisper_progress_cb;
-	params.progress_callback_user_data = (void*)(ctx);
 	return params;
 }
 */
@@ -103,7 +92,7 @@ var (
 func Whisper_init(path string) *Context {
 	cPath := C.CString(path)
 	defer C.free(unsafe.Pointer(cPath))
-	if ctx := C.whisper_init_from_file(cPath); ctx != nil {
+	if ctx := C.whisper_init(cPath); ctx != nil {
 		return (*Context)(ctx)
 	} else {
 		return nil
@@ -159,6 +148,16 @@ func (ctx *Context) Whisper_decode(tokens []Token, past, threads int) error {
 	}
 }
 
+// whisper_sample_best() returns the token with the highest probability
+func (ctx *Context) Whisper_sample_best() TokenData {
+	return TokenData(C.whisper_sample_best((*C.struct_whisper_context)(ctx)))
+}
+
+// whisper_sample_timestamp() returns the most probable timestamp token
+func (ctx *Context) Whisper_sample_timestamp(is_initial bool) TokenData {
+	return TokenData(C.whisper_sample_timestamp((*C.struct_whisper_context)(ctx), C.bool(is_initial)))
+}
+
 // Convert the provided text into tokens. The tokens pointer must be large enough to hold the resulting tokens.
 // Returns the number of tokens on success
 func (ctx *Context) Whisper_tokenize(text string, tokens []Token) (int, error) {
@@ -172,10 +171,6 @@ func (ctx *Context) Whisper_tokenize(text string, tokens []Token) (int, error) {
 }
 
 // Return the id of the specified language, returns -1 if not found
-// Examples:
-//
-//	"de" -> 2
-//	"german" -> 2
 func (ctx *Context) Whisper_lang_id(lang string) int {
 	return int(C.whisper_lang_id(C.CString(lang)))
 }
@@ -216,10 +211,6 @@ func (ctx *Context) Whisper_n_text_ctx() int {
 	return int(C.whisper_n_text_ctx((*C.struct_whisper_context)(ctx)))
 }
 
-func (ctx *Context) Whisper_n_audio_ctx() int {
-	return int(C.whisper_n_audio_ctx((*C.struct_whisper_context)(ctx)))
-}
-
 func (ctx *Context) Whisper_is_multilingual() int {
 	return int(C.whisper_is_multilingual((*C.struct_whisper_context)(ctx)))
 }
@@ -270,13 +261,13 @@ func (ctx *Context) Whisper_token_lang(lang_id int) Token {
 }
 
 // Task tokens
-func (ctx *Context) Whisper_token_translate() Token {
-	return Token(C.whisper_token_translate((*C.struct_whisper_context)(ctx)))
+func Whisper_token_translate() Token {
+	return Token(C.whisper_token_translate())
 }
 
 // Task tokens
-func (ctx *Context) Whisper_token_transcribe() Token {
-	return Token(C.whisper_token_transcribe((*C.struct_whisper_context)(ctx)))
+func Whisper_token_transcribe() Token {
+	return Token(C.whisper_token_transcribe())
 }
 
 // Performance information
@@ -302,19 +293,11 @@ func (ctx *Context) Whisper_full_default_params(strategy SamplingStrategy) Param
 
 // Run the entire model: PCM -> log mel spectrogram -> encoder -> decoder -> text
 // Uses the specified decoding strategy to obtain the text.
-func (ctx *Context) Whisper_full(
-	params Params,
-	samples []float32,
-	encoderBeginCallback func() bool,
-	newSegmentCallback func(int),
-	progressCallback func(int),
-) error {
+func (ctx *Context) Whisper_full(params Params, samples []float32, encoderBeginCallback func() bool, newSegmentCallback func(int)) error {
 	registerEncoderBeginCallback(ctx, encoderBeginCallback)
 	registerNewSegmentCallback(ctx, newSegmentCallback)
-	registerProgressCallback(ctx, progressCallback)
 	defer registerEncoderBeginCallback(ctx, nil)
 	defer registerNewSegmentCallback(ctx, nil)
-	defer registerProgressCallback(ctx, nil)
 	if C.whisper_full((*C.struct_whisper_context)(ctx), (C.struct_whisper_full_params)(params), (*C.float)(&samples[0]), C.int(len(samples))) == 0 {
 		return nil
 	} else {
@@ -338,18 +321,6 @@ func (ctx *Context) Whisper_full_parallel(params Params, samples []float32, proc
 	}
 }
 
-// Return the id of the autodetected language, returns -1 if not found
-// Added to whisper.cpp in
-// https://github.com/ggerganov/whisper.cpp/commit/a1c1583cc7cd8b75222857afc936f0638c5683d6
-//
-// Examples:
-//
-//	"de" -> 2
-//	"german" -> 2
-func (ctx *Context) Whisper_full_lang_id() int {
-	return int(C.whisper_full_lang_id((*C.struct_whisper_context)(ctx)))
-}
-
 // Number of generated text segments.
 // A segment can be a few words, a sentence, or even a paragraph.
 func (ctx *Context) Whisper_full_n_segments() int {
@@ -388,7 +359,7 @@ func (ctx *Context) Whisper_full_get_token_id(segment int, token int) Token {
 
 // Get token data for the specified token in the specified segment.
 // This contains probabilities, timestamps, etc.
-func (ctx *Context) Whisper_full_get_token_data(segment int, token int) TokenData {
+func (ctx *Context) whisper_full_get_token_data(segment int, token int) TokenData {
 	return TokenData(C.whisper_full_get_token_data((*C.struct_whisper_context)(ctx), C.int(segment), C.int(token)))
 }
 
@@ -402,7 +373,6 @@ func (ctx *Context) Whisper_full_get_token_p(segment int, token int) float32 {
 
 var (
 	cbNewSegment   = make(map[unsafe.Pointer]func(int))
-	cbProgress     = make(map[unsafe.Pointer]func(int))
 	cbEncoderBegin = make(map[unsafe.Pointer]func() bool)
 )
 
@@ -414,14 +384,6 @@ func registerNewSegmentCallback(ctx *Context, fn func(int)) {
 	}
 }
 
-func registerProgressCallback(ctx *Context, fn func(int)) {
-	if fn == nil {
-		delete(cbProgress, unsafe.Pointer(ctx))
-	} else {
-		cbProgress[unsafe.Pointer(ctx)] = fn
-	}
-}
-
 func registerEncoderBeginCallback(ctx *Context, fn func() bool) {
 	if fn == nil {
 		delete(cbEncoderBegin, unsafe.Pointer(ctx))
@@ -437,13 +399,6 @@ func callNewSegment(user_data unsafe.Pointer, new C.int) {
 	}
 }
 
-//export callProgress
-func callProgress(user_data unsafe.Pointer, progress C.int) {
-	if fn, ok := cbProgress[user_data]; ok {
-		fn(int(progress))
-	}
-}
-
 //export callEncoderBegin
 func callEncoderBegin(user_data unsafe.Pointer) C.bool {
 	if fn, ok := cbEncoderBegin[user_data]; ok {
@@ -455,15 +410,3 @@ func callEncoderBegin(user_data unsafe.Pointer) C.bool {
 	}
 	return true
 }
-
-func (t TokenData) T0() int64 {
-	return int64(t.t0)
-}
-
-func (t TokenData) T1() int64 {
-	return int64(t.t1)
-}
-
-func (t TokenData) Id() Token {
-	return Token(t.id)
-}

bindings/go/whisper_test.go

@@ -50,10 +50,7 @@ func Test_Whisper_001(t *testing.T) {
 	ctx := whisper.Whisper_init(ModelPath)
 	assert.NotNil(ctx)
 	defer ctx.Whisper_free()
-	params := ctx.Whisper_full_default_params(whisper.SAMPLING_GREEDY)
-	data := buf.AsFloat32Buffer().Data
-	err = ctx.Whisper_full(params, data, nil, nil, nil)
-	assert.NoError(err)
+	assert.NoError(ctx.Whisper_full(ctx.Whisper_full_default_params(whisper.SAMPLING_GREEDY), buf.AsFloat32Buffer().Data, nil, nil))
 
 	// Print out tokens
 	num_segments := ctx.Whisper_full_n_segments()

bindings/java/.idea/uiDesigner.xml

@@ -1,124 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<project version="4">
-  <component name="Palette2">
-    <group name="Swing">
-      <item class="com.intellij.uiDesigner.HSpacer" tooltip-text="Horizontal Spacer" icon="/com/intellij/uiDesigner/icons/hspacer.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="1" hsize-policy="6" anchor="0" fill="1" />
-      </item>
-      <item class="com.intellij.uiDesigner.VSpacer" tooltip-text="Vertical Spacer" icon="/com/intellij/uiDesigner/icons/vspacer.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="6" hsize-policy="1" anchor="0" fill="2" />
-      </item>
-      <item class="javax.swing.JPanel" icon="/com/intellij/uiDesigner/icons/panel.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="3" hsize-policy="3" anchor="0" fill="3" />
-      </item>
-      <item class="javax.swing.JScrollPane" icon="/com/intellij/uiDesigner/icons/scrollPane.svg" removable="false" auto-create-binding="false" can-attach-label="true">
-        <default-constraints vsize-policy="7" hsize-policy="7" anchor="0" fill="3" />
-      </item>
-      <item class="javax.swing.JButton" icon="/com/intellij/uiDesigner/icons/button.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="3" anchor="0" fill="1" />
-        <initial-values>
-          <property name="text" value="Button" />
-        </initial-values>
-      </item>
-      <item class="javax.swing.JRadioButton" icon="/com/intellij/uiDesigner/icons/radioButton.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="3" anchor="8" fill="0" />
-        <initial-values>
-          <property name="text" value="RadioButton" />
-        </initial-values>
-      </item>
-      <item class="javax.swing.JCheckBox" icon="/com/intellij/uiDesigner/icons/checkBox.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="3" anchor="8" fill="0" />
-        <initial-values>
-          <property name="text" value="CheckBox" />
-        </initial-values>
-      </item>
-      <item class="javax.swing.JLabel" icon="/com/intellij/uiDesigner/icons/label.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="0" anchor="8" fill="0" />
-        <initial-values>
-          <property name="text" value="Label" />
-        </initial-values>
-      </item>
-      <item class="javax.swing.JTextField" icon="/com/intellij/uiDesigner/icons/textField.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1">
-          <preferred-size width="150" height="-1" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JPasswordField" icon="/com/intellij/uiDesigner/icons/passwordField.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1">
-          <preferred-size width="150" height="-1" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JFormattedTextField" icon="/com/intellij/uiDesigner/icons/formattedTextField.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1">
-          <preferred-size width="150" height="-1" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JTextArea" icon="/com/intellij/uiDesigner/icons/textArea.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
-          <preferred-size width="150" height="50" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JTextPane" icon="/com/intellij/uiDesigner/icons/textPane.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
-          <preferred-size width="150" height="50" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JEditorPane" icon="/com/intellij/uiDesigner/icons/editorPane.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
-          <preferred-size width="150" height="50" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JComboBox" icon="/com/intellij/uiDesigner/icons/comboBox.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="0" hsize-policy="2" anchor="8" fill="1" />
-      </item>
-      <item class="javax.swing.JTable" icon="/com/intellij/uiDesigner/icons/table.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
-          <preferred-size width="150" height="50" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JList" icon="/com/intellij/uiDesigner/icons/list.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="6" hsize-policy="2" anchor="0" fill="3">
-          <preferred-size width="150" height="50" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JTree" icon="/com/intellij/uiDesigner/icons/tree.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
-          <preferred-size width="150" height="50" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JTabbedPane" icon="/com/intellij/uiDesigner/icons/tabbedPane.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="3" hsize-policy="3" anchor="0" fill="3">
-          <preferred-size width="200" height="200" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JSplitPane" icon="/com/intellij/uiDesigner/icons/splitPane.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="3" hsize-policy="3" anchor="0" fill="3">
-          <preferred-size width="200" height="200" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JSpinner" icon="/com/intellij/uiDesigner/icons/spinner.svg" removable="false" auto-create-binding="true" can-attach-label="true">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1" />
-      </item>
-      <item class="javax.swing.JSlider" icon="/com/intellij/uiDesigner/icons/slider.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1" />
-      </item>
-      <item class="javax.swing.JSeparator" icon="/com/intellij/uiDesigner/icons/separator.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3" />
-      </item>
-      <item class="javax.swing.JProgressBar" icon="/com/intellij/uiDesigner/icons/progressbar.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="0" fill="1" />
-      </item>
-      <item class="javax.swing.JToolBar" icon="/com/intellij/uiDesigner/icons/toolbar.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="6" anchor="0" fill="1">
-          <preferred-size width="-1" height="20" />
-        </default-constraints>
-      </item>
-      <item class="javax.swing.JToolBar$Separator" icon="/com/intellij/uiDesigner/icons/toolbarSeparator.svg" removable="false" auto-create-binding="false" can-attach-label="false">
-        <default-constraints vsize-policy="0" hsize-policy="0" anchor="0" fill="1" />
-      </item>
-      <item class="javax.swing.JScrollBar" icon="/com/intellij/uiDesigner/icons/scrollbar.svg" removable="false" auto-create-binding="true" can-attach-label="false">
-        <default-constraints vsize-policy="6" hsize-policy="0" anchor="0" fill="2" />
-      </item>
-    </group>
-  </component>
-</project>

bindings/java/README.md

@@ -1,71 +0,0 @@
-# Java JNI bindings for Whisper
-
-This package provides Java JNI bindings for whisper.cpp. They have been tested on:
-
-  * <strike>Darwin (OS X) 12.6 on x64_64</strike>
-  * Ubuntu on x86_64
-  * Windows on x86_64
-
-The "low level" bindings are in `WhisperCppJnaLibrary`. The most simple usage is as follows:
-
-JNA will attempt to load the `whispercpp` shared library from:
-
-- jna.library.path
-- jna.platform.library
-- ~/Library/Frameworks
-- /Library/Frameworks
-- /System/Library/Frameworks
-- classpath
-
-```java
-import io.github.ggerganov.whispercpp.WhisperCpp;
-
-public class Example {
-
-    public static void main(String[] args) {
-        WhisperCpp whisper = new WhisperCpp();
-        // By default, models are loaded from ~/.cache/whisper/ and are usually named "ggml-${name}.bin"
-        // or you can provide the absolute path to the model file.
-        long context = whisper.initContext("base.en");
-        try {
-            var whisperParams = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
-            // custom configuration if required
-            whisperParams.temperature_inc = 0f;
-            
-            var samples = readAudio(); // divide each value by 32767.0f
-            whisper.fullTranscribe(whisperParams, samples);
-            
-            int segmentCount = whisper.getTextSegmentCount(context);
-            for (int i = 0; i < segmentCount; i++) {
-                String text = whisper.getTextSegment(context, i);
-                System.out.println(segment.getText());
-            }
-        } finally {
-             whisper.freeContext(context);
-        }
-     }
-}
-```
-
-## Building & Testing
-
-In order to build, you need to have the JDK 8 or higher installed. Run the tests with:
-
-```bash
-git clone https://github.com/ggerganov/whisper.cpp.git
-cd whisper.cpp/bindings/java
-
-./gradlew build
-```
-
-You need to have the `whisper` library in your [JNA library path](https://java-native-access.github.io/jna/4.2.1/com/sun/jna/NativeLibrary.html). On Windows the dll is included in the jar and you can update it:
-
-```bash
-copy /y ..\..\build\bin\Release\whisper.dll build\generated\resources\main\win32-x86-64\whisper.dll
-```
-
-
-## License
-
-The license for the Go bindings is the same as the license for the rest of the whisper.cpp project, which is the MIT License. See the `LICENSE` file for more details.
-

bindings/java/build.gradle

@@ -1,112 +0,0 @@
-plugins {
-    id 'java'
-    id 'java-library'
-    id 'maven-publish'
-}
-
-archivesBaseName = 'whispercpp'
-group = 'io.github.ggerganov'
-version = '1.4.0'
-
-sourceCompatibility = 1.8
-targetCompatibility = 1.8
-
-sourceSets {
-    main {
-        resources {
-            srcDirs = ['src/main/resources', 'build/generated/resources/main']
-        }
-    }
-    test {
-        runtimeClasspath += files('build/generated/resources/main')
-    }
-}
-
-tasks.register('copyLibwhisperDynlib', Copy) {
-    from '../../build'
-    include 'libwhisper.dynlib'
-    into 'build/generated/resources/main/darwin'
-}
-
-tasks.register('copyLibwhisperSo', Copy) {
-    from '../../build'
-    include 'libwhisper.so'
-    into 'build/generated/resources/main/linux-x86-64'
-}
-
-tasks.register('copyWhisperDll', Copy) {
-    from '../../build/Release'
-    include 'whisper.dll'
-    into 'build/generated/resources/main/windows-x86-64'
-}
-
-tasks.register('copyLibs') {
-    dependsOn copyLibwhisperDynlib, copyLibwhisperSo, copyWhisperDll
-}
-
-test {
-    systemProperty 'jna.library.path', project.file('build/generated/resources/main').absolutePath
-}
-
-java {
-    withSourcesJar()
-    withJavadocJar()
-}
-
-jar {
-    exclude '**/whisper_java.exp', '**/whisper_java.lib'
-}
-
-javadoc {
-    options.addStringOption('Xdoclint:none', '-quiet')
-}
-
-tasks.withType(Test) {
-    useJUnitPlatform()
-}
-
-dependencies {
-    implementation "net.java.dev.jna:jna:5.13.0"
-    testImplementation "org.junit.jupiter:junit-jupiter:5.9.2"
-    testImplementation "org.assertj:assertj-core:3.24.2"
-}
-
-repositories {
-    mavenCentral()
-}
-
-publishing {
-    publications {
-        mavenJava(MavenPublication) {
-            artifactId = 'whispercpp'
-            from components.java
-            pom {
-                name = 'whispercpp'
-                description = "Java JNA bindings for OpenAI's Whisper model, implemented in C/C++"
-                url = 'https://github.com/ggerganov/whisper.cpp'
-                licenses {
-                    license {
-                        name = 'MIT licence'
-                        url = 'https://raw.githubusercontent.com/ggerganov/whisper.cpp/master/LICENSE'
-                    }
-                }
-                developers {
-                    developer {
-                        id = 'ggerganov'
-                        name = 'Georgi Gerganov'
-                        email = 'ggerganov@gmail.com'
-                    }
-                    developer {
-                        id = 'nalbion'
-                        name = 'Nicholas Albion'
-                        email = 'nalbion@yahoo.com'
-                    }
-                }
-                scm {
-                    connection = 'scm:git:git://github.com/ggerganov/whisper.cpp.git'
-                    url = 'https://github.com/ggerganov/whisper.cpp'
-                }
-            }
-        }
-    }
-}

bindings/java/gradle.properties

@@ -1,6 +0,0 @@
-org.gradle.jvmargs=-Xms256m -Xmx1024m
-system.include.dir=/usr/include
-#system.local.include.dir=../../include
-system.local.include.dir=./build/generated/sources/headers/java/main
-jni.include.dir=/usr/lib/jvm/java-8-openjdk-amd64/include/
-jni.lib.dir=/usr/lib/jvm/java-8-openjdk-amd64/lib/

bindings/java/gradle/wrapper/gradle-wrapper.properties

@@ -1,6 +0,0 @@
-distributionBase=GRADLE_USER_HOME
-distributionPath=wrapper/dists
-distributionUrl=https\://services.gradle.org/distributions/gradle-8.1-bin.zip
-networkTimeout=10000
-zipStoreBase=GRADLE_USER_HOME
-zipStorePath=wrapper/dists

bindings/java/gradlew

@@ -1,244 +0,0 @@
-#!/bin/sh
-
-#
-# Copyright © 2015-2021 the original authors.
-#
-# 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.
-#
-
-##############################################################################
-#
-#   Gradle start up script for POSIX generated by Gradle.
-#
-#   Important for running:
-#
-#   (1) You need a POSIX-compliant shell to run this script. If your /bin/sh is
-#       noncompliant, but you have some other compliant shell such as ksh or
-#       bash, then to run this script, type that shell name before the whole
-#       command line, like:
-#
-#           ksh Gradle
-#
-#       Busybox and similar reduced shells will NOT work, because this script
-#       requires all of these POSIX shell features:
-#         * functions;
-#         * expansions «$var», «${var}», «${var:-default}», «${var+SET}»,
-#           «${var#prefix}», «${var%suffix}», and «$( cmd )»;
-#         * compound commands having a testable exit status, especially «case»;
-#         * various built-in commands including «command», «set», and «ulimit».
-#
-#   Important for patching:
-#
-#   (2) This script targets any POSIX shell, so it avoids extensions provided
-#       by Bash, Ksh, etc; in particular arrays are avoided.
-#
-#       The "traditional" practice of packing multiple parameters into a
-#       space-separated string is a well documented source of bugs and security
-#       problems, so this is (mostly) avoided, by progressively accumulating
-#       options in "$@", and eventually passing that to Java.
-#
-#       Where the inherited environment variables (DEFAULT_JVM_OPTS, JAVA_OPTS,
-#       and GRADLE_OPTS) rely on word-splitting, this is performed explicitly;
-#       see the in-line comments for details.
-#
-#       There are tweaks for specific operating systems such as AIX, CygWin,
-#       Darwin, MinGW, and NonStop.
-#
-#   (3) This script is generated from the Groovy template
-#       https://github.com/gradle/gradle/blob/HEAD/subprojects/plugins/src/main/resources/org/gradle/api/internal/plugins/unixStartScript.txt
-#       within the Gradle project.
-#
-#       You can find Gradle at https://github.com/gradle/gradle/.
-#
-##############################################################################
-
-# Attempt to set APP_HOME
-
-# Resolve links: $0 may be a link
-app_path=$0
-
-# Need this for daisy-chained symlinks.
-while
-    APP_HOME=${app_path%"${app_path##*/}"}  # leaves a trailing /; empty if no leading path
-    [ -h "$app_path" ]
-do
-    ls=$( ls -ld "$app_path" )
-    link=${ls#*' -> '}
-    case $link in             #(
-      /*)   app_path=$link ;; #(
-      *)    app_path=$APP_HOME$link ;;
-    esac
-done
-
-# This is normally unused
-# shellcheck disable=SC2034
-APP_BASE_NAME=${0##*/}
-APP_HOME=$( cd "${APP_HOME:-./}" && pwd -P ) || exit
-
-# Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
-DEFAULT_JVM_OPTS='"-Xmx64m" "-Xms64m"'
-
-# Use the maximum available, or set MAX_FD != -1 to use that value.
-MAX_FD=maximum
-
-warn () {
-    echo "$*"
-} >&2
-
-die () {
-    echo
-    echo "$*"
-    echo
-    exit 1
-} >&2
-
-# OS specific support (must be 'true' or 'false').
-cygwin=false
-msys=false
-darwin=false
-nonstop=false
-case "$( uname )" in                #(
-  CYGWIN* )         cygwin=true  ;; #(
-  Darwin* )         darwin=true  ;; #(
-  MSYS* | MINGW* )  msys=true    ;; #(
-  NONSTOP* )        nonstop=true ;;
-esac
-
-CLASSPATH=$APP_HOME/gradle/wrapper/gradle-wrapper.jar
-
-
-# Determine the Java command to use to start the JVM.
-if [ -n "$JAVA_HOME" ] ; then
-    if [ -x "$JAVA_HOME/jre/sh/java" ] ; then
-        # IBM's JDK on AIX uses strange locations for the executables
-        JAVACMD=$JAVA_HOME/jre/sh/java
-    else
-        JAVACMD=$JAVA_HOME/bin/java
-    fi
-    if [ ! -x "$JAVACMD" ] ; then
-        die "ERROR: JAVA_HOME is set to an invalid directory: $JAVA_HOME
-
-Please set the JAVA_HOME variable in your environment to match the
-location of your Java installation."
-    fi
-else
-    JAVACMD=java
-    which java >/dev/null 2>&1 || die "ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
-
-Please set the JAVA_HOME variable in your environment to match the
-location of your Java installation."
-fi
-
-# Increase the maximum file descriptors if we can.
-if ! "$cygwin" && ! "$darwin" && ! "$nonstop" ; then
-    case $MAX_FD in #(
-      max*)
-        # In POSIX sh, ulimit -H is undefined. That's why the result is checked to see if it worked.
-        # shellcheck disable=SC3045
-        MAX_FD=$( ulimit -H -n ) ||
-            warn "Could not query maximum file descriptor limit"
-    esac
-    case $MAX_FD in  #(
-      '' | soft) :;; #(
-      *)
-        # In POSIX sh, ulimit -n is undefined. That's why the result is checked to see if it worked.
-        # shellcheck disable=SC3045
-        ulimit -n "$MAX_FD" ||
-            warn "Could not set maximum file descriptor limit to $MAX_FD"
-    esac
-fi
-
-# Collect all arguments for the java command, stacking in reverse order:
-#   * args from the command line
-#   * the main class name
-#   * -classpath
-#   * -D...appname settings
-#   * --module-path (only if needed)
-#   * DEFAULT_JVM_OPTS, JAVA_OPTS, and GRADLE_OPTS environment variables.
-
-# For Cygwin or MSYS, switch paths to Windows format before running java
-if "$cygwin" || "$msys" ; then
-    APP_HOME=$( cygpath --path --mixed "$APP_HOME" )
-    CLASSPATH=$( cygpath --path --mixed "$CLASSPATH" )
-
-    JAVACMD=$( cygpath --unix "$JAVACMD" )
-
-    # Now convert the arguments - kludge to limit ourselves to /bin/sh
-    for arg do
-        if
-            case $arg in                                #(
-              -*)   false ;;                            # don't mess with options #(
-              /?*)  t=${arg#/} t=/${t%%/*}              # looks like a POSIX filepath
-                    [ -e "$t" ] ;;                      #(
-              *)    false ;;
-            esac
-        then
-            arg=$( cygpath --path --ignore --mixed "$arg" )
-        fi
-        # Roll the args list around exactly as many times as the number of
-        # args, so each arg winds up back in the position where it started, but
-        # possibly modified.
-        #
-        # NB: a `for` loop captures its iteration list before it begins, so
-        # changing the positional parameters here affects neither the number of
-        # iterations, nor the values presented in `arg`.
-        shift                   # remove old arg
-        set -- "$@" "$arg"      # push replacement arg
-    done
-fi
-
-# Collect all arguments for the java command;
-#   * $DEFAULT_JVM_OPTS, $JAVA_OPTS, and $GRADLE_OPTS can contain fragments of
-#     shell script including quotes and variable substitutions, so put them in
-#     double quotes to make sure that they get re-expanded; and
-#   * put everything else in single quotes, so that it's not re-expanded.
-
-set -- \
-        "-Dorg.gradle.appname=$APP_BASE_NAME" \
-        -classpath "$CLASSPATH" \
-        org.gradle.wrapper.GradleWrapperMain \
-        "$@"
-
-# Stop when "xargs" is not available.
-if ! command -v xargs >/dev/null 2>&1
-then
-    die "xargs is not available"
-fi
-
-# Use "xargs" to parse quoted args.
-#
-# With -n1 it outputs one arg per line, with the quotes and backslashes removed.
-#
-# In Bash we could simply go:
-#
-#   readarray ARGS < <( xargs -n1 <<<"$var" ) &&
-#   set -- "${ARGS[@]}" "$@"
-#
-# but POSIX shell has neither arrays nor command substitution, so instead we
-# post-process each arg (as a line of input to sed) to backslash-escape any
-# character that might be a shell metacharacter, then use eval to reverse
-# that process (while maintaining the separation between arguments), and wrap
-# the whole thing up as a single "set" statement.
-#
-# This will of course break if any of these variables contains a newline or
-# an unmatched quote.
-#
-
-eval "set -- $(
-        printf '%s\n' "$DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS" |
-        xargs -n1 |
-        sed ' s~[^-[:alnum:]+,./:=@_]~\\&~g; ' |
-        tr '\n' ' '
-    )" '"$@"'
-
-exec "$JAVACMD" "$@"

bindings/java/gradlew.bat

@@ -1,92 +0,0 @@
-@rem
-@rem Copyright 2015 the original author or authors.
-@rem
-@rem Licensed under the Apache License, Version 2.0 (the "License");
-@rem you may not use this file except in compliance with the License.
-@rem You may obtain a copy of the License at
-@rem
-@rem      https://www.apache.org/licenses/LICENSE-2.0
-@rem
-@rem Unless required by applicable law or agreed to in writing, software
-@rem distributed under the License is distributed on an "AS IS" BASIS,
-@rem WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-@rem See the License for the specific language governing permissions and
-@rem limitations under the License.
-@rem
-
-@if "%DEBUG%"=="" @echo off
-@rem ##########################################################################
-@rem
-@rem  Gradle startup script for Windows
-@rem
-@rem ##########################################################################
-
-@rem Set local scope for the variables with windows NT shell
-if "%OS%"=="Windows_NT" setlocal
-
-set DIRNAME=%~dp0
-if "%DIRNAME%"=="" set DIRNAME=.
-@rem This is normally unused
-set APP_BASE_NAME=%~n0
-set APP_HOME=%DIRNAME%
-
-@rem Resolve any "." and ".." in APP_HOME to make it shorter.
-for %%i in ("%APP_HOME%") do set APP_HOME=%%~fi
-
-@rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
-set DEFAULT_JVM_OPTS="-Xmx64m" "-Xms64m"
-
-@rem Find java.exe
-if defined JAVA_HOME goto findJavaFromJavaHome
-
-set JAVA_EXE=java.exe
-%JAVA_EXE% -version >NUL 2>&1
-if %ERRORLEVEL% equ 0 goto execute
-
-echo.
-echo ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
-echo.
-echo Please set the JAVA_HOME variable in your environment to match the
-echo location of your Java installation.
-
-goto fail
-
-:findJavaFromJavaHome
-set JAVA_HOME=%JAVA_HOME:"=%
-set JAVA_EXE=%JAVA_HOME%/bin/java.exe
-
-if exist "%JAVA_EXE%" goto execute
-
-echo.
-echo ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME%
-echo.
-echo Please set the JAVA_HOME variable in your environment to match the
-echo location of your Java installation.
-
-goto fail
-
-:execute
-@rem Setup the command line
-
-set CLASSPATH=%APP_HOME%\gradle\wrapper\gradle-wrapper.jar
-
-
-@rem Execute Gradle
-"%JAVA_EXE%" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% "-Dorg.gradle.appname=%APP_BASE_NAME%" -classpath "%CLASSPATH%" org.gradle.wrapper.GradleWrapperMain %*
-
-:end
-@rem End local scope for the variables with windows NT shell
-if %ERRORLEVEL% equ 0 goto mainEnd
-
-:fail
-rem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of
-rem the _cmd.exe /c_ return code!
-set EXIT_CODE=%ERRORLEVEL%
-if %EXIT_CODE% equ 0 set EXIT_CODE=1
-if not ""=="%GRADLE_EXIT_CONSOLE%" exit %EXIT_CODE%
-exit /b %EXIT_CODE%
-
-:mainEnd
-if "%OS%"=="Windows_NT" endlocal
-
-:omega

bindings/java/settings.gradle

@@ -1 +0,0 @@
-rootProject.name = "whispercpp"

bindings/java/src/main/java/io/github/ggerganov/whispercpp/WhisperContext.java

@@ -1,39 +0,0 @@
-package io.github.ggerganov.whispercpp;
-
-import com.sun.jna.Structure;
-import com.sun.jna.ptr.PointerByReference;
-import io.github.ggerganov.whispercpp.ggml.GgmlType;
-import io.github.ggerganov.whispercpp.WhisperModel;
-
-import java.util.List;
-
-public class WhisperContext extends Structure {
-    int t_load_us = 0;
-    int t_start_us = 0;
-
-    /** weight type (FP32 / FP16 / QX) */
-    GgmlType wtype = GgmlType.GGML_TYPE_F16;
-    /** intermediate type (FP32 or FP16) */
-    GgmlType itype = GgmlType.GGML_TYPE_F16;
-
-//    WhisperModel model;
-    public PointerByReference model;
-//    whisper_vocab vocab;
-//    whisper_state * state = nullptr;
-    public PointerByReference vocab;
-    public PointerByReference state;
-
-    /** populated by whisper_init_from_file() */
-    String path_model;
-
-//    public static class ByReference extends WhisperContext implements Structure.ByReference {
-//    }
-//
-//    public static class ByValue extends WhisperContext implements Structure.ByValue {
-//    }
-//
-//    @Override
-//    protected List<String> getFieldOrder() {
-//        return List.of("t_load_us", "t_start_us", "wtype", "itype", "model", "vocab", "state", "path_model");
-//    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/WhisperCpp.java

@@ -1,151 +0,0 @@
-package io.github.ggerganov.whispercpp;
-
-import com.sun.jna.Native;
-import com.sun.jna.Pointer;
-import io.github.ggerganov.whispercpp.params.WhisperFullParams;
-import io.github.ggerganov.whispercpp.params.WhisperSamplingStrategy;
-
-import java.io.File;
-import java.io.FileNotFoundException;
-import java.io.IOException;
-
-/**
- * Before calling most methods, you must call `initContext(modelPath)` to initialise the `ctx` Pointer.
- */
-public class WhisperCpp implements AutoCloseable {
-    private WhisperCppJnaLibrary lib = WhisperCppJnaLibrary.instance;
-    private Pointer ctx = null;
-    private Pointer greedyPointer = null;
-    private Pointer beamPointer = null;
-
-    public File modelDir() {
-        String modelDirPath = System.getenv("XDG_CACHE_HOME");
-        if (modelDirPath == null) {
-            modelDirPath = System.getProperty("user.home") + "/.cache";
-        }
-
-        return new File(modelDirPath, "whisper");
-    }
-
-    /**
-     * @param modelPath - absolute path, or just the name (eg: "base", "base-en" or "base.en")
-     */
-    public void initContext(String modelPath) throws FileNotFoundException {
-        if (ctx != null) {
-            lib.whisper_free(ctx);
-        }
-
-        if (!modelPath.contains("/") && !modelPath.contains("\\")) {
-            if (!modelPath.endsWith(".bin")) {
-                modelPath = "ggml-" + modelPath.replace("-", ".") + ".bin";
-            }
-
-            modelPath = new File(modelDir(), modelPath).getAbsolutePath();
-        }
-
-        ctx = lib.whisper_init_from_file(modelPath);
-
-        if (ctx == null) {
-            throw new FileNotFoundException(modelPath);
-        }
-    }
-
-    /**
-     * Provides default params which can be used with `whisper_full()` etc.
-     * Because this function allocates memory for the params, the caller must call either:
-     * - call `whisper_free_params()`
-     * - `Native.free(Pointer.nativeValue(pointer));`
-     *
-     * @param strategy - GREEDY
-     */
-    public WhisperFullParams getFullDefaultParams(WhisperSamplingStrategy strategy) {
-        Pointer pointer;
-
-        // whisper_full_default_params_by_ref allocates memory which we need to delete, so only create max 1 pointer for each strategy.
-        if (strategy == WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY) {
-            if (greedyPointer == null) {
-                greedyPointer = lib.whisper_full_default_params_by_ref(strategy.ordinal());
-            }
-            pointer = greedyPointer;
-        } else {
-            if (beamPointer == null) {
-                beamPointer = lib.whisper_full_default_params_by_ref(strategy.ordinal());
-            }
-            pointer = beamPointer;
-        }
-
-        WhisperFullParams params = new WhisperFullParams(pointer);
-        params.read();
-        return params;
-    }
-
-    @Override
-    public void close() {
-        freeContext();
-        freeParams();
-        System.out.println("Whisper closed");
-    }
-
-    private void freeContext() {
-        if (ctx != null) {
-            lib.whisper_free(ctx);
-        }
-    }
-
-    private void freeParams() {
-        if (greedyPointer != null) {
-            Native.free(Pointer.nativeValue(greedyPointer));
-            greedyPointer = null;
-        }
-        if (beamPointer != null) {
-            Native.free(Pointer.nativeValue(beamPointer));
-            beamPointer = null;
-        }
-    }
-
-    /**
-     * Run the entire model: PCM -> log mel spectrogram -> encoder -> decoder -> text.
-     * Not thread safe for same context
-     * Uses the specified decoding strategy to obtain the text.
-     */
-    public String fullTranscribe(WhisperFullParams whisperParams, float[] audioData) throws IOException {
-        if (ctx == null) {
-            throw new IllegalStateException("Model not initialised");
-        }
-
-        if (lib.whisper_full(ctx, whisperParams, audioData, audioData.length) != 0) {
-            throw new IOException("Failed to process audio");
-        }
-
-        int nSegments = lib.whisper_full_n_segments(ctx);
-
-        StringBuilder str = new StringBuilder();
-
-        for (int i = 0; i < nSegments; i++) {
-            String text = lib.whisper_full_get_segment_text(ctx, i);
-            System.out.println("Segment:" + text);
-            str.append(text);
-        }
-
-        return str.toString().trim();
-    }
-
-//    public int getTextSegmentCount(Pointer ctx) {
-//        return lib.whisper_full_n_segments(ctx);
-//    }
-//    public String getTextSegment(Pointer ctx, int index) {
-//        return lib.whisper_full_get_segment_text(ctx, index);
-//    }
-
-    public String getSystemInfo() {
-        return lib.whisper_print_system_info();
-    }
-
-    public int benchMemcpy(int nthread) {
-        return lib.whisper_bench_memcpy(nthread);
-    }
-
-    public int benchGgmlMulMat(int nthread) {
-        return lib.whisper_bench_ggml_mul_mat(nthread);
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/WhisperCppJnaLibrary.java

@@ -1,376 +0,0 @@
-package io.github.ggerganov.whispercpp;
-
-import com.sun.jna.Library;
-import com.sun.jna.Native;
-import com.sun.jna.Pointer;
-import io.github.ggerganov.whispercpp.model.WhisperModelLoader;
-import io.github.ggerganov.whispercpp.model.WhisperTokenData;
-import io.github.ggerganov.whispercpp.params.WhisperFullParams;
-
-public interface WhisperCppJnaLibrary extends Library {
-    WhisperCppJnaLibrary instance = Native.load("whisper", WhisperCppJnaLibrary.class);
-
-    String whisper_print_system_info();
-
-    /**
-     * Allocate (almost) all memory needed for the model by loading from a file.
-     *
-     * @param path_model Path to the model file
-     * @return Whisper context on success, null on failure
-     */
-    Pointer whisper_init_from_file(String path_model);
-
-    /**
-     * Allocate (almost) all memory needed for the model by loading from a buffer.
-     *
-     * @param buffer       Model buffer
-     * @param buffer_size  Size of the model buffer
-     * @return Whisper context on success, null on failure
-     */
-    Pointer whisper_init_from_buffer(Pointer buffer, int buffer_size);
-
-    /**
-     * Allocate (almost) all memory needed for the model using a model loader.
-     *
-     * @param loader Model loader
-     * @return Whisper context on success, null on failure
-     */
-    Pointer whisper_init(WhisperModelLoader loader);
-
-    /**
-     * Allocate (almost) all memory needed for the model by loading from a file without allocating the state.
-     *
-     * @param path_model Path to the model file
-     * @return Whisper context on success, null on failure
-     */
-    Pointer whisper_init_from_file_no_state(String path_model);
-
-    /**
-     * Allocate (almost) all memory needed for the model by loading from a buffer without allocating the state.
-     *
-     * @param buffer       Model buffer
-     * @param buffer_size  Size of the model buffer
-     * @return Whisper context on success, null on failure
-     */
-    Pointer whisper_init_from_buffer_no_state(Pointer buffer, int buffer_size);
-
-//    Pointer whisper_init_from_buffer_no_state(Pointer buffer, long buffer_size);
-
-    /**
-     * Allocate (almost) all memory needed for the model using a model loader without allocating the state.
-     *
-     * @param loader Model loader
-     * @return Whisper context on success, null on failure
-     */
-    Pointer whisper_init_no_state(WhisperModelLoader loader);
-
-    /**
-     * Allocate memory for the Whisper state.
-     *
-     * @param ctx Whisper context
-     * @return Whisper state on success, null on failure
-     */
-    Pointer whisper_init_state(Pointer ctx);
-
-    /**
-     * Free all allocated memory associated with the Whisper context.
-     *
-     * @param ctx Whisper context
-     */
-    void whisper_free(Pointer ctx);
-
-    /**
-     * Free all allocated memory associated with the Whisper state.
-     *
-     * @param state Whisper state
-     */
-    void whisper_free_state(Pointer state);
-
-
-    /**
-     * Convert RAW PCM audio to log mel spectrogram.
-     * The resulting spectrogram is stored inside the default state of the provided whisper context.
-     *
-     * @param ctx - Pointer to a WhisperContext
-     * @return 0 on success
-     */
-    int whisper_pcm_to_mel(Pointer ctx, final float[] samples, int n_samples, int n_threads);
-
-    /**
-     * @param ctx Pointer to a WhisperContext
-     * @param state Pointer to WhisperState
-     * @param n_samples
-     * @param n_threads
-     * @return 0 on success
-     */
-    int whisper_pcm_to_mel_with_state(Pointer ctx, Pointer state, final float[] samples, int n_samples, int n_threads);
-
-    /**
-     * This can be used to set a custom log mel spectrogram inside the default state of the provided whisper context.
-     * Use this instead of whisper_pcm_to_mel() if you want to provide your own log mel spectrogram.
-     * n_mel must be 80
-     * @return 0 on success
-     */
-    int whisper_set_mel(Pointer ctx, final float[] data, int n_len, int n_mel);
-    int whisper_set_mel_with_state(Pointer ctx, Pointer state, final float[] data, int n_len, int n_mel);
-
-    /**
-     * Run the Whisper encoder on the log mel spectrogram stored inside the default state in the provided whisper context.
-     * Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
-     * Offset can be used to specify the offset of the first frame in the spectrogram.
-     * @return 0 on success
-     */
-    int whisper_encode(Pointer ctx, int offset, int n_threads);
-
-    int whisper_encode_with_state(Pointer ctx, Pointer state, int offset, int n_threads);
-
-    /**
-     * Run the Whisper decoder to obtain the logits and probabilities for the next token.
-     * Make sure to call whisper_encode() first.
-     * tokens + n_tokens is the provided context for the decoder.
-     * n_past is the number of tokens to use from previous decoder calls.
-     * Returns 0 on success
-     * TODO: add support for multiple decoders
-     */
-    int whisper_decode(Pointer ctx, Pointer tokens, int n_tokens, int n_past, int n_threads);
-
-    /**
-     * @param ctx
-     * @param state
-     * @param tokens Pointer to int tokens
-     * @param n_tokens
-     * @param n_past
-     * @param n_threads
-     * @return
-     */
-    int whisper_decode_with_state(Pointer ctx, Pointer state, Pointer tokens, int n_tokens, int n_past, int n_threads);
-
-    /**
-     * Convert the provided text into tokens.
-     * The tokens pointer must be large enough to hold the resulting tokens.
-     * Returns the number of tokens on success, no more than n_max_tokens
-     * Returns -1 on failure
-     * TODO: not sure if correct
-     */
-    int whisper_tokenize(Pointer ctx, String text, Pointer tokens, int n_max_tokens);
-
-    /** Largest language id (i.e. number of available languages - 1) */
-    int whisper_lang_max_id();
-
-    /**
-     * @return the id of the specified language, returns -1 if not found.
-     * Examples:
-     *   "de" -> 2
-     *   "german" -> 2
-     */
-    int whisper_lang_id(String lang);
-
-    /** @return the short string of the specified language id (e.g. 2 -> "de"), returns nullptr if not found */
-    String whisper_lang_str(int id);
-
-    /**
-     * Use mel data at offset_ms to try and auto-detect the spoken language.
-     * Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first
-     * Returns the top language id or negative on failure
-     * If not null, fills the lang_probs array with the probabilities of all languages
-     * The array must be whisper_lang_max_id() + 1 in size
-     *
-     * ref: https://github.com/openai/whisper/blob/main/whisper/decoding.py#L18-L69
-     */
-    int whisper_lang_auto_detect(Pointer ctx, int offset_ms, int n_threads, float[] lang_probs);
-
-    int whisper_lang_auto_detect_with_state(Pointer ctx, Pointer state, int offset_ms, int n_threads, float[] lang_probs);
-
-    int whisper_n_len           (Pointer ctx); // mel length
-    int whisper_n_len_from_state(Pointer state); // mel length
-    int whisper_n_vocab         (Pointer ctx);
-    int whisper_n_text_ctx      (Pointer ctx);
-    int whisper_n_audio_ctx     (Pointer ctx);
-    int whisper_is_multilingual (Pointer ctx);
-
-    int whisper_model_n_vocab      (Pointer ctx);
-    int whisper_model_n_audio_ctx  (Pointer ctx);
-    int whisper_model_n_audio_state(Pointer ctx);
-    int whisper_model_n_audio_head (Pointer ctx);
-    int whisper_model_n_audio_layer(Pointer ctx);
-    int whisper_model_n_text_ctx   (Pointer ctx);
-    int whisper_model_n_text_state (Pointer ctx);
-    int whisper_model_n_text_head  (Pointer ctx);
-    int whisper_model_n_text_layer (Pointer ctx);
-    int whisper_model_n_mels       (Pointer ctx);
-    int whisper_model_ftype        (Pointer ctx);
-    int whisper_model_type         (Pointer ctx);
-
-    /**
-     * Token logits obtained from the last call to whisper_decode().
-     * The logits for the last token are stored in the last row
-     * Rows: n_tokens
-     * Cols: n_vocab
-     */
-    float[] whisper_get_logits           (Pointer ctx);
-    float[] whisper_get_logits_from_state(Pointer state);
-
-    // Token Id -> String. Uses the vocabulary in the provided context
-    String whisper_token_to_str(Pointer ctx, int token);
-    String whisper_model_type_readable(Pointer ctx);
-
-    // Special tokens
-    int whisper_token_eot (Pointer ctx);
-    int whisper_token_sot (Pointer ctx);
-    int whisper_token_prev(Pointer ctx);
-    int whisper_token_solm(Pointer ctx);
-    int whisper_token_not (Pointer ctx);
-    int whisper_token_beg (Pointer ctx);
-    int whisper_token_lang(Pointer ctx, int lang_id);
-
-    // Task tokens
-    int whisper_token_translate (Pointer ctx);
-    int whisper_token_transcribe(Pointer ctx);
-
-    // Performance information from the default state.
-    void whisper_print_timings(Pointer ctx);
-    void whisper_reset_timings(Pointer ctx);
-
-    // Note: Even if `whisper_full_params is stripped back to just 4 ints, JNA throws "Invalid memory access"
-    //       when `whisper_full_default_params()` tries to return a struct.
-    // WhisperFullParams whisper_full_default_params(int strategy);
-
-    /**
-     * Provides default params which can be used with `whisper_full()` etc.
-     * Because this function allocates memory for the params, the caller must call either:
-     * - call `whisper_free_params()`
-     * - `Native.free(Pointer.nativeValue(pointer));`
-     *
-     * @param strategy - WhisperSamplingStrategy.value
-     */
-    Pointer whisper_full_default_params_by_ref(int strategy);
-
-    void whisper_free_params(Pointer params);
-
-    /**
-     * Run the entire model: PCM -> log mel spectrogram -> encoder -> decoder -> text
-     * Not thread safe for same context
-     * Uses the specified decoding strategy to obtain the text.
-     */
-    int whisper_full(Pointer ctx, WhisperFullParams params, final float[] samples, int n_samples);
-
-    int whisper_full_with_state(Pointer ctx, Pointer state, WhisperFullParams params, final float[] samples, int n_samples);
-
-    // Split the input audio in chunks and process each chunk separately using whisper_full_with_state()
-    // Result is stored in the default state of the context
-    // Not thread safe if executed in parallel on the same context.
-    // It seems this approach can offer some speedup in some cases.
-    // However, the transcription accuracy can be worse at the beginning and end of each chunk.
-    int whisper_full_parallel(Pointer ctx, WhisperFullParams params, final float[] samples, int n_samples, int n_processors);
-
-    /**
-     * Number of generated text segments.
-     * A segment can be a few words, a sentence, or even a paragraph.
-     * @param ctx Pointer to WhisperContext
-     */
-    int whisper_full_n_segments (Pointer ctx);
-
-    /**
-     * @param state Pointer to WhisperState
-     */
-    int whisper_full_n_segments_from_state(Pointer state);
-
-    /**
-     * Language id associated with the context's default state.
-     * @param ctx Pointer to WhisperContext
-     */
-    int whisper_full_lang_id(Pointer ctx);
-
-    /** Language id associated with the provided state */
-    int whisper_full_lang_id_from_state(Pointer state);
-
-    /**
-     * Convert RAW PCM audio to log mel spectrogram but applies a Phase Vocoder to speed up the audio x2.
-     * The resulting spectrogram is stored inside the default state of the provided whisper context.
-     * @return 0 on success
-     */
-    int whisper_pcm_to_mel_phase_vocoder(Pointer ctx, final float[] samples, int n_samples, int n_threads);
-
-    int whisper_pcm_to_mel_phase_vocoder_with_state(Pointer ctx, Pointer state, final float[] samples, int n_samples, int n_threads);
-
-    /** Get the start time of the specified segment. */
-    long whisper_full_get_segment_t0(Pointer ctx, int i_segment);
-
-    /** Get the start time of the specified segment from the state. */
-    long whisper_full_get_segment_t0_from_state(Pointer state, int i_segment);
-
-    /** Get the end time of the specified segment. */
-    long whisper_full_get_segment_t1(Pointer ctx, int i_segment);
-
-    /** Get the end time of the specified segment from the state. */
-    long whisper_full_get_segment_t1_from_state(Pointer state, int i_segment);
-
-    /** Get the text of the specified segment. */
-    String whisper_full_get_segment_text(Pointer ctx, int i_segment);
-
-    /** Get the text of the specified segment from the state. */
-    String whisper_full_get_segment_text_from_state(Pointer state, int i_segment);
-
-    /** Get the number of tokens in the specified segment. */
-    int whisper_full_n_tokens(Pointer ctx, int i_segment);
-
-    /** Get the number of tokens in the specified segment from the state. */
-    int whisper_full_n_tokens_from_state(Pointer state, int i_segment);
-
-    /** Get the token text of the specified token in the specified segment. */
-    String whisper_full_get_token_text(Pointer ctx, int i_segment, int i_token);
-
-
-    /** Get the token text of the specified token in the specified segment from the state. */
-    String whisper_full_get_token_text_from_state(Pointer ctx, Pointer state, int i_segment, int i_token);
-
-    /** Get the token ID of the specified token in the specified segment. */
-    int whisper_full_get_token_id(Pointer ctx, int i_segment, int i_token);
-
-    /** Get the token ID of the specified token in the specified segment from the state. */
-    int whisper_full_get_token_id_from_state(Pointer state, int i_segment, int i_token);
-
-    /** Get token data for the specified token in the specified segment. */
-    WhisperTokenData whisper_full_get_token_data(Pointer ctx, int i_segment, int i_token);
-
-    /** Get token data for the specified token in the specified segment from the state. */
-    WhisperTokenData whisper_full_get_token_data_from_state(Pointer state, int i_segment, int i_token);
-
-    /** Get the probability of the specified token in the specified segment. */
-    float whisper_full_get_token_p(Pointer ctx, int i_segment, int i_token);
-
-    /** Get the probability of the specified token in the specified segment from the state. */
-    float whisper_full_get_token_p_from_state(Pointer state, int i_segment, int i_token);
-
-    /**
-     * Benchmark function for memcpy.
-     *
-     * @param nThreads Number of threads to use for the benchmark.
-     * @return The result of the benchmark.
-     */
-    int whisper_bench_memcpy(int nThreads);
-
-    /**
-     * Benchmark function for memcpy as a string.
-     *
-     * @param nThreads Number of threads to use for the benchmark.
-     * @return The result of the benchmark as a string.
-     */
-    String whisper_bench_memcpy_str(int nThreads);
-
-    /**
-     * Benchmark function for ggml_mul_mat.
-     *
-     * @param nThreads Number of threads to use for the benchmark.
-     * @return The result of the benchmark.
-     */
-    int whisper_bench_ggml_mul_mat(int nThreads);
-
-    /**
-     * Benchmark function for ggml_mul_mat as a string.
-     *
-     * @param nThreads Number of threads to use for the benchmark.
-     * @return The result of the benchmark as a string.
-     */
-    String whisper_bench_ggml_mul_mat_str(int nThreads);
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/callbacks/WhisperEncoderBeginCallback.java

@@ -1,24 +0,0 @@
-package io.github.ggerganov.whispercpp.callbacks;
-
-import com.sun.jna.Callback;
-import com.sun.jna.Pointer;
-import io.github.ggerganov.whispercpp.WhisperContext;
-import io.github.ggerganov.whispercpp.model.WhisperState;
-
-/**
- * Callback before the encoder starts.
- * If not null, called before the encoder starts.
- * If it returns false, the computation is aborted.
- */
-public interface WhisperEncoderBeginCallback extends Callback {
-
-    /**
-     * Callback method before the encoder starts.
-     *
-     * @param ctx        The whisper context.
-     * @param state      The whisper state.
-     * @param user_data  User data.
-     * @return True if the computation should proceed, false otherwise.
-     */
-    boolean callback(Pointer ctx, Pointer state, Pointer user_data);
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/callbacks/WhisperLogitsFilterCallback.java

@@ -1,25 +0,0 @@
-package io.github.ggerganov.whispercpp.callbacks;
-
-import com.sun.jna.Callback;
-import com.sun.jna.Pointer;
-import io.github.ggerganov.whispercpp.model.WhisperTokenData;
-
-/**
- * Callback to filter logits.
- * Can be used to modify the logits before sampling.
- * If not null, called after applying temperature to logits.
- */
-public interface WhisperLogitsFilterCallback extends Callback {
-
-    /**
-     * Callback method to filter logits.
-     *
-     * @param ctx        The whisper context.
-     * @param state      The whisper state.
-     * @param tokens     The array of whisper_token_data.
-     * @param n_tokens   The number of tokens.
-     * @param logits     The array of logits.
-     * @param user_data  User data.
-     */
-    void callback(Pointer ctx, Pointer state, WhisperTokenData[] tokens, int n_tokens, float[] logits, Pointer user_data);
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/callbacks/WhisperNewSegmentCallback.java

@@ -1,24 +0,0 @@
-package io.github.ggerganov.whispercpp.callbacks;
-
-import com.sun.jna.Callback;
-import com.sun.jna.Pointer;
-import io.github.ggerganov.whispercpp.WhisperContext;
-import io.github.ggerganov.whispercpp.model.WhisperState;
-
-/**
- * Callback for the text segment.
- * Called on every newly generated text segment.
- * Use the whisper_full_...() functions to obtain the text segments.
- */
-public interface WhisperNewSegmentCallback extends Callback {
-
-    /**
-     * Callback method for the text segment.
-     *
-     * @param ctx        The whisper context.
-     * @param state      The whisper state.
-     * @param n_new      The number of newly generated text segments.
-     * @param user_data  User data.
-     */
-    void callback(Pointer ctx, Pointer state, int n_new, Pointer user_data);
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/callbacks/WhisperProgressCallback.java

@@ -1,22 +0,0 @@
-package io.github.ggerganov.whispercpp.callbacks;
-
-import com.sun.jna.Callback;
-import com.sun.jna.Pointer;
-import io.github.ggerganov.whispercpp.WhisperContext;
-import io.github.ggerganov.whispercpp.model.WhisperState;
-
-/**
- * Callback for progress updates.
- */
-public interface WhisperProgressCallback extends Callback {
-
-    /**
-     * Callback method for progress updates.
-     *
-     * @param ctx        The whisper context.
-     * @param state      The whisper state.
-     * @param progress   The progress value.
-     * @param user_data  User data.
-     */
-    void callback(Pointer ctx, Pointer state, int progress, Pointer user_data);
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/ggml/GgmlTensor.java

@@ -1,4 +0,0 @@
-package io.github.ggerganov.whispercpp.ggml;
-
-public class GgmlTensor {
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/ggml/GgmlType.java

@@ -1,18 +0,0 @@
-package io.github.ggerganov.whispercpp.ggml;
-
-public enum GgmlType {
-    GGML_TYPE_F32,
-    GGML_TYPE_F16,
-    GGML_TYPE_Q4_0,
-    GGML_TYPE_Q4_1,
-    REMOVED_GGML_TYPE_Q4_2,  // support has been removed
-    REMOVED_GGML_TYPE_Q4_3, // support has been removed
-    GGML_TYPE_Q5_0,
-    GGML_TYPE_Q5_1,
-    GGML_TYPE_Q8_0,
-    GGML_TYPE_Q8_1,
-    GGML_TYPE_I8,
-    GGML_TYPE_I16,
-    GGML_TYPE_I32,
-    GGML_TYPE_COUNT,
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/model/EModel.java

@@ -1,10 +0,0 @@
-package io.github.ggerganov.whispercpp.model;
-
-public enum EModel {
-    MODEL_UNKNOWN,
-    MODEL_TINY,
-    MODEL_BASE,
-    MODEL_SMALL,
-    MODEL_MEDIUM,
-    MODEL_LARGE,
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/model/WhisperModel.java

@@ -1,49 +0,0 @@
-package io.github.ggerganov.whispercpp;
-
-import io.github.ggerganov.whispercpp.ggml.GgmlTensor;
-import io.github.ggerganov.whispercpp.model.EModel;
-
-public class WhisperModel {
-//    EModel type = EModel.MODEL_UNKNOWN;
-//
-//    WhisperHParams hparams;
-//    WhisperFilters filters;
-//
-//    // encoder.positional_embedding
-//    GgmlTensor e_pe;
-//
-//    // encoder.conv1
-//    GgmlTensor e_conv_1_w;
-//    GgmlTensor e_conv_1_b;
-//
-//    // encoder.conv2
-//    GgmlTensor e_conv_2_w;
-//    GgmlTensor e_conv_2_b;
-//
-//    // encoder.ln_post
-//    GgmlTensor e_ln_w;
-//    GgmlTensor e_ln_b;
-//
-//    // decoder.positional_embedding
-//    GgmlTensor d_pe;
-//
-//    // decoder.token_embedding
-//    GgmlTensor d_te;
-//
-//    // decoder.ln
-//    GgmlTensor d_ln_w;
-//    GgmlTensor d_ln_b;
-//
-//    std::vector<whisper_layer_encoder> layers_encoder;
-//    std::vector<whisper_layer_decoder> layers_decoder;
-//
-//    // context
-//    struct ggml_context * ctx;
-//
-//    // the model memory buffer is read-only and can be shared between processors
-//    std::vector<uint8_t> * buf;
-//
-//    // tensors
-//    int n_loaded;
-//    Map<String, GgmlTensor> tensors;
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/model/WhisperModelLoader.java

@@ -1,62 +0,0 @@
-package io.github.ggerganov.whispercpp.model;
-
-import com.sun.jna.Callback;
-import com.sun.jna.Pointer;
-import com.sun.jna.Structure;
-
-
-public class WhisperModelLoader extends Structure {
-    public Pointer context;
-    public ReadFunction read;
-    public EOFFunction eof;
-    public CloseFunction close;
-
-    public static class ReadFunction implements Callback {
-        public Pointer invoke(Pointer ctx, Pointer output, int readSize) {
-            // TODO
-            return ctx;
-        }
-    }
-
-    public static class EOFFunction implements Callback {
-        public boolean invoke(Pointer ctx) {
-            // TODO
-            return false;
-        }
-    }
-
-    public static class CloseFunction implements Callback {
-        public void invoke(Pointer ctx) {
-            // TODO
-        }
-    }
-
-//    public WhisperModelLoader(Pointer p) {
-//        super(p);
-//        read = new ReadFunction();
-//        eof = new EOFFunction();
-//        close = new CloseFunction();
-//        read.setCallback(this);
-//        eof.setCallback(this);
-//        close.setCallback(this);
-//        read.write();
-//        eof.write();
-//        close.write();
-//    }
-
-    public WhisperModelLoader() {
-        super();
-    }
-
-    public interface ReadCallback extends Callback {
-        Pointer invoke(Pointer ctx, Pointer output, int readSize);
-    }
-
-    public interface EOFCallback extends Callback {
-        boolean invoke(Pointer ctx);
-    }
-
-    public interface CloseCallback extends Callback {
-        void invoke(Pointer ctx);
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/model/WhisperState.java

@@ -1,4 +0,0 @@
-package io.github.ggerganov.whispercpp.model;
-
-public class WhisperState {
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/model/WhisperTokenData.java

@@ -1,50 +0,0 @@
-package io.github.ggerganov.whispercpp.model;
-
-import com.sun.jna.Structure;
-
-import java.util.Arrays;
-import java.util.List;
-
-/**
- * Structure representing token data.
- */
-public class WhisperTokenData extends Structure {
-
-    /** Token ID. */
-    public int id;
-
-    /** Forced timestamp token ID. */
-    public int tid;
-
-    /** Probability of the token. */
-    public float p;
-
-    /** Log probability of the token. */
-    public float plog;
-
-    /** Probability of the timestamp token. */
-    public float pt;
-
-    /** Sum of probabilities of all timestamp tokens. */
-    public float ptsum;
-
-    /**
-     * Start time of the token (token-level timestamp data).
-     * Do not use if you haven't computed token-level timestamps.
-     */
-    public long t0;
-
-    /**
-     * End time of the token (token-level timestamp data).
-     * Do not use if you haven't computed token-level timestamps.
-     */
-    public long t1;
-
-    /** Voice length of the token. */
-    public float vlen;
-
-    @Override
-    protected List<String> getFieldOrder() {
-        return Arrays.asList("id", "tid", "p", "plog", "pt", "ptsum", "t0", "t1", "vlen");
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/BeamSearchParams.java

@@ -1,19 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-import com.sun.jna.Structure;
-
-import java.util.Arrays;
-import java.util.List;
-
-public class BeamSearchParams extends Structure {
-    /** ref: <a href="https://github.com/openai/whisper/blob/f82bc59f5ea234d4b97fb2860842ed38519f7e65/whisper/transcribe.py#L265">...</a> */
-    public int beam_size;
-
-    /** ref: <a href="https://arxiv.org/pdf/2204.05424.pdf">...</a> */
-    public float patience;
-
-    @Override
-    protected List<String> getFieldOrder() {
-        return Arrays.asList("beam_size", "patience");
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/CBool.java

@@ -1,30 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-import com.sun.jna.IntegerType;
-
-import java.util.function.BooleanSupplier;
-
-public class CBool extends IntegerType implements BooleanSupplier {
-    public static final int SIZE = 1;
-    public static final CBool FALSE = new CBool(0);
-    public static final CBool TRUE = new CBool(1);
-
-
-    public CBool() {
-        this(0);
-    }
-
-    public CBool(long value) {
-        super(SIZE, value, true);
-    }
-
-    @Override
-    public boolean getAsBoolean() {
-        return intValue() == 1;
-    }
-
-    @Override
-    public String toString() {
-        return intValue() == 1 ? "true" : "false";
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/GreedyParams.java

@@ -1,16 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-import com.sun.jna.Structure;
-
-import java.util.Collections;
-import java.util.List;
-
-public class GreedyParams extends Structure {
-    /** <a href="https://github.com/openai/whisper/blob/f82bc59f5ea234d4b97fb2860842ed38519f7e65/whisper/transcribe.py#L264">...</a> */
-    public int best_of;
-
-    @Override
-    protected List<String> getFieldOrder() {
-        return Collections.singletonList("best_of");
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/WhisperFilters.java

@@ -1,10 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-import java.util.List;
-
-public class WhisperFilters {
-    int n_mel;
-    int n_fft;
-
-    List<Float> data;
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/WhisperFullParams.java

@@ -1,321 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-import com.sun.jna.*;
-import io.github.ggerganov.whispercpp.callbacks.WhisperEncoderBeginCallback;
-import io.github.ggerganov.whispercpp.callbacks.WhisperLogitsFilterCallback;
-import io.github.ggerganov.whispercpp.callbacks.WhisperNewSegmentCallback;
-import io.github.ggerganov.whispercpp.callbacks.WhisperProgressCallback;
-
-import java.util.Arrays;
-import java.util.List;
-
-/**
- * Parameters for the whisper_full() function.
- * If you change the order or add new parameters, make sure to update the default values in whisper.cpp:
- * whisper_full_default_params()
- */
-public class WhisperFullParams extends Structure {
-
-    public WhisperFullParams(Pointer p) {
-        super(p);
-//        super(p, ALIGN_MSVC);
-//        super(p, ALIGN_GNUC);
-    }
-
-    /** Sampling strategy for whisper_full() function. */
-    public int strategy;
-
-    /** Number of threads. (default = 4) */
-    public int n_threads;
-
-    /** Maximum tokens to use from past text as a prompt for the decoder. (default = 16384) */
-    public int n_max_text_ctx;
-
-    /** Start offset in milliseconds. (default = 0) */
-    public int offset_ms;
-
-    /** Audio duration to process in milliseconds. (default = 0) */
-    public int duration_ms;
-
-    /** Translate flag. (default = false) */
-    public CBool translate;
-
-    /** The compliment of translateMode() */
-    public void transcribeMode() {
-        translate = CBool.FALSE;
-    }
-
-    /** The compliment of transcribeMode() */
-    public void translateMode() {
-        translate = CBool.TRUE;
-    }
-
-    /** Flag to indicate whether to use past transcription (if any) as an initial prompt for the decoder. (default = true) */
-    public CBool no_context;
-
-    /** Flag to indicate whether to use past transcription (if any) as an initial prompt for the decoder. (default = true) */
-    public void enableContext(boolean enable) {
-        no_context = enable ? CBool.FALSE : CBool.TRUE;
-    }
-
-    /** Flag to force single segment output (useful for streaming). (default = false) */
-    public CBool single_segment;
-
-    /** Flag to force single segment output (useful for streaming). (default = false) */
-    public void singleSegment(boolean single) {
-        single_segment = single ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Flag to print special tokens (e.g., <SOT>, <EOT>, <BEG>, etc.). (default = false) */
-    public CBool print_special;
-
-    /** Flag to print special tokens (e.g., <SOT>, <EOT>, <BEG>, etc.). (default = false) */
-    public void printSpecial(boolean enable) {
-        print_special = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Flag to print progress information. (default = true) */
-    public CBool print_progress;
-
-    /** Flag to print progress information. (default = true) */
-    public void printProgress(boolean enable) {
-        print_progress = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Flag to print results from within whisper.cpp (avoid it, use callback instead). (default = true) */
-    public CBool print_realtime;
-
-    /** Flag to print results from within whisper.cpp (avoid it, use callback instead). (default = true) */
-    public void printRealtime(boolean enable) {
-        print_realtime = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Flag to print timestamps for each text segment when printing realtime. (default = true) */
-    public CBool print_timestamps;
-
-    /** Flag to print timestamps for each text segment when printing realtime. (default = true) */
-    public void printTimestamps(boolean enable) {
-        print_timestamps = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** [EXPERIMENTAL] Flag to enable token-level timestamps. (default = false) */
-    public CBool token_timestamps;
-
-    /** [EXPERIMENTAL] Flag to enable token-level timestamps. (default = false) */
-    public void tokenTimestamps(boolean enable) {
-        token_timestamps = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** [EXPERIMENTAL] Timestamp token probability threshold (~0.01). (default = 0.01) */
-    public float thold_pt;
-
-    /** [EXPERIMENTAL] Timestamp token sum probability threshold (~0.01). */
-    public float thold_ptsum;
-
-    /** Maximum segment length in characters. (default = 0) */
-    public int max_len;
-
-    /** Flag to split on word rather than on token (when used with max_len). (default = false) */
-    public CBool split_on_word;
-
-    /** Flag to split on word rather than on token (when used with max_len). (default = false) */
-    public void splitOnWord(boolean enable) {
-        split_on_word = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Maximum tokens per segment (0, default = no limit) */
-    public int max_tokens;
-
-    /** Flag to speed up the audio by 2x using Phase Vocoder. (default = false) */
-    public CBool speed_up;
-
-    /** Flag to speed up the audio by 2x using Phase Vocoder. (default = false) */
-    public void speedUp(boolean enable) {
-        speed_up = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Overwrite the audio context size (0 = use default). */
-    public int audio_ctx;
-
-    /** Enable tinydiarize (default = false) */
-    public CBool tdrz_enable;
-
-    /** Enable tinydiarize (default = false) */
-    public void tdrzEnable(boolean enable) {
-        tdrz_enable = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Tokens to provide to the whisper decoder as an initial prompt.
-     * These are prepended to any existing text context from a previous call. */
-    public String initial_prompt;
-
-    /** Prompt tokens. (int*) */
-    public Pointer prompt_tokens;
-
-    public void setPromptTokens(int[] tokens) {
-        Memory mem = new Memory(tokens.length * 4L);
-        mem.write(0, tokens, 0, tokens.length);
-        prompt_tokens = mem;
-    }
-
-    /** Number of prompt tokens. */
-    public int prompt_n_tokens;
-
-    /** Language for auto-detection.
-     * For auto-detection, set to `null`, `""`, or "auto". */
-    public String language;
-
-    /** Flag to indicate whether to detect language automatically. */
-    public CBool detect_language;
-
-    /** Flag to indicate whether to detect language automatically. */
-    public void detectLanguage(boolean enable) {
-        detect_language = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    // Common decoding parameters.
-
-    /** Flag to suppress blank tokens. */
-    public CBool suppress_blank;
-
-    public void suppressBlanks(boolean enable) {
-        suppress_blank = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Flag to suppress non-speech tokens. */
-    public CBool suppress_non_speech_tokens;
-
-    /** Flag to suppress non-speech tokens. */
-    public void suppressNonSpeechTokens(boolean enable) {
-        suppress_non_speech_tokens = enable ? CBool.TRUE : CBool.FALSE;
-    }
-
-    /** Initial decoding temperature. */
-    public float temperature;
-
-    /** Maximum initial timestamp. */
-    public float max_initial_ts;
-
-    /** Length penalty. */
-    public float length_penalty;
-
-    // Fallback parameters.
-
-    /** Temperature increment. */
-    public float temperature_inc;
-
-    /** Entropy threshold (similar to OpenAI's "compression_ratio_threshold"). */
-    public float entropy_thold;
-
-    /** Log probability threshold. */
-    public float logprob_thold;
-
-    /** No speech threshold. */
-    public float no_speech_thold;
-
-    /** Greedy decoding parameters. */
-    public GreedyParams greedy;
-
-    /**
-     * Beam search decoding parameters.
-     */
-    public BeamSearchParams beam_search;
-
-    public void setBestOf(int bestOf) {
-        if (greedy == null) {
-            greedy = new GreedyParams();
-        }
-        greedy.best_of = bestOf;
-    }
-
-    public void setBeamSize(int beamSize) {
-        if (beam_search == null) {
-            beam_search = new BeamSearchParams();
-        }
-        beam_search.beam_size = beamSize;
-    }
-
-    public void setBeamSizeAndPatience(int beamSize, float patience) {
-        if (beam_search == null) {
-            beam_search = new BeamSearchParams();
-        }
-        beam_search.beam_size = beamSize;
-        beam_search.patience = patience;
-    }
-
-    /**
-     * Callback for every newly generated text segment.
-     * WhisperNewSegmentCallback
-     */
-    public Pointer new_segment_callback;
-
-    /**
-     * User data for the new_segment_callback.
-     */
-    public Pointer new_segment_callback_user_data;
-
-    /**
-     * Callback on each progress update.
-     * WhisperProgressCallback
-     */
-    public Pointer progress_callback;
-
-    /**
-     * User data for the progress_callback.
-     */
-    public Pointer progress_callback_user_data;
-
-    /**
-     * Callback each time before the encoder starts.
-     * WhisperEncoderBeginCallback
-     */
-    public Pointer encoder_begin_callback;
-
-    /**
-     * User data for the encoder_begin_callback.
-     */
-    public Pointer encoder_begin_callback_user_data;
-
-    /**
-     * Callback by each decoder to filter obtained logits.
-     * WhisperLogitsFilterCallback
-     */
-    public Pointer logits_filter_callback;
-
-    /**
-     * User data for the logits_filter_callback.
-     */
-    public Pointer logits_filter_callback_user_data;
-
-
-    public void setNewSegmentCallback(WhisperNewSegmentCallback callback) {
-        new_segment_callback = CallbackReference.getFunctionPointer(callback);
-    }
-
-    public void setProgressCallback(WhisperProgressCallback callback) {
-        progress_callback = CallbackReference.getFunctionPointer(callback);
-    }
-
-    public void setEncoderBeginCallbackeginCallbackCallback(WhisperEncoderBeginCallback callback) {
-        encoder_begin_callback = CallbackReference.getFunctionPointer(callback);
-    }
-
-    public void setLogitsFilterCallback(WhisperLogitsFilterCallback callback) {
-        logits_filter_callback = CallbackReference.getFunctionPointer(callback);
-    }
-
-    @Override
-    protected List<String> getFieldOrder() {
-        return Arrays.asList("strategy", "n_threads", "n_max_text_ctx", "offset_ms", "duration_ms", "translate",
-                "no_context", "single_segment",
-                "print_special", "print_progress", "print_realtime", "print_timestamps",  "token_timestamps",
-                "thold_pt", "thold_ptsum", "max_len", "split_on_word", "max_tokens", "speed_up", "audio_ctx",
-                "tdrz_enable", "initial_prompt", "prompt_tokens", "prompt_n_tokens", "language", "detect_language",
-                "suppress_blank", "suppress_non_speech_tokens", "temperature", "max_initial_ts", "length_penalty",
-                "temperature_inc", "entropy_thold", "logprob_thold", "no_speech_thold", "greedy", "beam_search",
-                "new_segment_callback", "new_segment_callback_user_data",
-                "progress_callback", "progress_callback_user_data",
-                "encoder_begin_callback", "encoder_begin_callback_user_data",
-                "logits_filter_callback", "logits_filter_callback_user_data");
-    }
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/WhisperHParams.java

@@ -1,15 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-public class WhisperHParams {
-    int n_vocab       = 51864;
-    int n_audio_ctx   = 1500;
-    int n_audio_state = 384;
-    int n_audio_head  = 6;
-    int n_audio_layer = 4;
-    int n_text_ctx    = 448;
-    int n_text_state  = 384;
-    int n_text_head   = 6;
-    int n_text_layer  = 4;
-    int n_mels        = 80;
-    int ftype         = 1;
-}

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/WhisperSamplingStrategy.java

@@ -1,10 +0,0 @@
-package io.github.ggerganov.whispercpp.params;
-
-/** Available sampling strategies */
-public enum WhisperSamplingStrategy {
-    /** similar to OpenAI's GreedyDecoder */
-    WHISPER_SAMPLING_GREEDY,
-
-    /** similar to OpenAI's BeamSearchDecoder */
-    WHISPER_SAMPLING_BEAM_SEARCH
-}

bindings/java/src/test/java/io/github/ggerganov/whispercpp/WhisperCppTest.java

@@ -1,102 +0,0 @@
-package io.github.ggerganov.whispercpp;
-
-import static org.junit.jupiter.api.Assertions.*;
-
-import io.github.ggerganov.whispercpp.params.CBool;
-import io.github.ggerganov.whispercpp.params.WhisperFullParams;
-import io.github.ggerganov.whispercpp.params.WhisperSamplingStrategy;
-import org.junit.jupiter.api.BeforeAll;
-import org.junit.jupiter.api.Test;
-import javax.sound.sampled.AudioInputStream;
-import javax.sound.sampled.AudioSystem;
-import java.io.File;
-import java.io.FileNotFoundException;
-
-class WhisperCppTest {
-    private static WhisperCpp whisper = new WhisperCpp();
-    private static boolean modelInitialised = false;
-
-    @BeforeAll
-    static void init() throws FileNotFoundException {
-        // By default, models are loaded from ~/.cache/whisper/ and are usually named "ggml-${name}.bin"
-        // or you can provide the absolute path to the model file.
-        String modelName = "../../models/ggml-tiny.en.bin";
-        try {
-            whisper.initContext(modelName);
-//            whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
-//            whisper.getJavaDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
-            modelInitialised = true;
-        } catch (FileNotFoundException ex) {
-            System.out.println("Model " + modelName + " not found");
-        }
-    }
-
-    @Test
-    void testGetDefaultFullParams_BeamSearch() {
-        // When
-        WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
-
-        // Then
-        assertEquals(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH.ordinal(), params.strategy);
-        assertNotEquals(0, params.n_threads);
-        assertEquals(16384, params.n_max_text_ctx);
-        assertFalse(params.translate);
-        assertEquals(0.01f, params.thold_pt);
-        assertEquals(2, params.beam_search.beam_size);
-        assertEquals(-1.0f, params.beam_search.patience);
-    }
-
-    @Test
-    void testGetDefaultFullParams_Greedy() {
-        // When
-        WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
-
-        // Then
-        assertEquals(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY.ordinal(), params.strategy);
-        assertNotEquals(0, params.n_threads);
-        assertEquals(16384, params.n_max_text_ctx);
-        assertEquals(2, params.greedy.best_of);
-    }
-
-    @Test
-    void testFullTranscribe() throws Exception {
-        if (!modelInitialised) {
-            System.out.println("Model not initialised, skipping test");
-            return;
-        }
-
-        // Given
-        File file = new File(System.getProperty("user.dir"), "../../samples/jfk.wav");
-        AudioInputStream audioInputStream = AudioSystem.getAudioInputStream(file);
-
-        byte[] b = new byte[audioInputStream.available()];
-        float[] floats = new float[b.length / 2];
-
-//        WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
-        WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
-        params.setProgressCallback((ctx, state, progress, user_data) -> System.out.println("progress: " + progress));
-        params.print_progress = CBool.FALSE;
-//        params.initial_prompt = "and so my fellow Americans um, like";
-
-
-        try {
-            audioInputStream.read(b);
-
-            for (int i = 0, j = 0; i < b.length; i += 2, j++) {
-                int intSample = (int) (b[i + 1]) << 8 | (int) (b[i]) & 0xFF;
-                floats[j] = intSample / 32767.0f;
-            }
-
-            // When
-            String result = whisper.fullTranscribe(params, floats);
-
-            // Then
-            System.err.println(result);
-            assertEquals("And so my fellow Americans ask not what your country can do for you " +
-                    "ask what you can do for your country.",
-                    result.replace(",", ""));
-        } finally {
-            audioInputStream.close();
-        }
-    }
-}

bindings/java/src/test/java/io/github/ggerganov/whispercpp/WhisperJnaLibraryTest.java

@@ -1,17 +0,0 @@
-package io.github.ggerganov.whispercpp;
-
-import static org.junit.jupiter.api.Assertions.*;
-
-import org.junit.jupiter.api.Test;
-
-class WhisperJnaLibraryTest {
-
-    @Test
-    void testWhisperPrint_system_info() {
-        String systemInfo = WhisperCppJnaLibrary.instance.whisper_print_system_info();
-        // eg: "AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0
-        //    | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | COREML = 0 | "
-        System.out.println("System info: " + systemInfo);
-        assertTrue(systemInfo.length() > 10);
-    }
-}

bindings/javascript/emscripten.cpp

@@ -20,7 +20,7 @@ struct whisper_context * g_context;
 EMSCRIPTEN_BINDINGS(whisper) {
     emscripten::function("init", emscripten::optional_override([](const std::string & path_model) {
         if (g_context == nullptr) {
-            g_context = whisper_init_from_file(path_model.c_str());
+            g_context = whisper_init(path_model.c_str());
             if (g_context != nullptr) {
                 return true;
             } else {

bindings/javascript/libwhisper.worker.js

@@ -1 +1 @@
-"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:f=>(0,eval)(fs.readFileSync(f,"utf8")+"//# sourceURL="+f),postMessage:msg=>parentPort.postMessage(msg),performance:global.performance||{now:Date.now}})}var initializedJS=false;function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"\n");return}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;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var module=Module["wasmModule"];Module["wasmModule"]=null;var instance=new WebAssembly.Instance(module,info);return receiveInstance(instance)};self.onunhandledrejection=e=>{throw e.reason||e};function handleMessage(e){try{if(e.data.cmd==="load"){let messageQueue=[];self.onmessage=e=>messageQueue.push(e);self.startWorker=instance=>{Module=instance;postMessage({"cmd":"loaded"});for(let msg of messageQueue){handleMessage(msg)}self.onmessage=handleMessage};Module["wasmModule"]=e.data.wasmModule;for(const handler of e.data.handlers){Module[handler]=(...args)=>{postMessage({cmd:"callHandler",handler:handler,args:args})}}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)}whisper_factory(Module)}else if(e.data.cmd==="run"){Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["__emscripten_thread_mailbox_await"](e.data.pthread_ptr);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){Module["__embind_initialize_bindings"]();initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){throw ex}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="checkMailbox"){if(initializedJS){Module["checkMailbox"]()}}else if(e.data.cmd){err(`worker.js received unknown command ${e.data.cmd}`);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}}self.onmessage=handleMessage;
+"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8")+"//# sourceURL="+f)},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"\n");return}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;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;for(const handler of e.data.handlers){Module[handler]=function(){postMessage({cmd:"callHandler",handler:handler,args:[...arguments]})}}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)}whisper_factory(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){Module["__embind_initialize_bindings"]();pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};

bindings/javascript/package.json

@@ -1,6 +1,6 @@
 {
   "name": "whisper.cpp",
-  "version": "1.4.2",
+  "version": "1.0.4",
   "description": "Whisper speech recognition",
   "main": "whisper.js",
   "scripts": {

bindings/ruby/ext/.gitignore

@@ -1,7 +0,0 @@
-Makefile
-ggml.c
-ggml.h
-whisper.bundle
-whisper.cpp
-whisper.h
-dr_wav.h

bindings/ruby/ext/extconf.rb

@@ -1,21 +0,0 @@
-require 'mkmf'
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','whisper.cpp')} .")
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','whisper.h')} .")
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml.h')} .")
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml.c')} .")
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','examples','dr_wav.h')} .")
-
-
-# need to use c++ compiler flags
-$CXXFLAGS << ' -std=c++11'
-# Set to true when building binary gems
-if enable_config('static-stdlib', false)
-  $LDFLAGS << ' -static-libgcc -static-libstdc++'
-end
-
-if enable_config('march-tune-native', false)
-  $CFLAGS << ' -march=native -mtune=native'
-  $CXXFLAGS << ' -march=native -mtune=native'
-end
-
-create_makefile('whisper')

bindings/ruby/ext/ruby_whisper.cpp

@@ -1,426 +0,0 @@
-#include <ruby.h>
-#include "ruby_whisper.h"
-#define DR_WAV_IMPLEMENTATION
-#include "dr_wav.h"
-#include <cmath>
-#include <fstream>
-#include <cstdio>
-#include <string>
-#include <thread>
-#include <vector>
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define BOOL_PARAMS_SETTER(self, prop, value) \
-  ruby_whisper_params *rwp; \
-  Data_Get_Struct(self, ruby_whisper_params, rwp); \
-  if (value == Qfalse || value == Qnil) { \
-    rwp->params.prop = false; \
-  } else { \
-    rwp->params.prop = true; \
-  } \
-  return value; \
-
-#define BOOL_PARAMS_GETTER(self,  prop) \
-  ruby_whisper_params *rwp; \
-  Data_Get_Struct(self, ruby_whisper_params, rwp); \
-  if (rwp->params.prop) { \
-    return Qtrue; \
-  } else { \
-    return Qfalse; \
-  }
-
-VALUE mWhisper;
-VALUE cContext;
-VALUE cParams;
-
-static void ruby_whisper_free(ruby_whisper *rw) {
-  if (rw->context) {
-    whisper_free(rw->context);
-    rw->context = NULL;
-  }
-}
-static void ruby_whisper_params_free(ruby_whisper_params *rwp) {
-}
-
-void rb_whisper_mark(ruby_whisper *rw) {
-  // call rb_gc_mark on any ruby references in rw
-}
-
-void rb_whisper_free(ruby_whisper *rw) {
-  ruby_whisper_free(rw);
-  free(rw);
-}
-
-void rb_whisper_params_mark(ruby_whisper_params *rwp) {
-}
-
-void rb_whisper_params_free(ruby_whisper_params *rwp) {
-  ruby_whisper_params_free(rwp);
-  free(rwp);
-}
-
-static VALUE ruby_whisper_allocate(VALUE klass) {
-  ruby_whisper *rw;
-  rw = ALLOC(ruby_whisper);
-  rw->context = NULL;
-  return Data_Wrap_Struct(klass, rb_whisper_mark, rb_whisper_free, rw);
-}
-
-static VALUE ruby_whisper_params_allocate(VALUE klass) {
-  ruby_whisper_params *rwp;
-  rwp = ALLOC(ruby_whisper_params);
-  rwp->params = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
-  return Data_Wrap_Struct(klass, rb_whisper_params_mark, rb_whisper_params_free, rwp);
-}
-
-static VALUE ruby_whisper_initialize(int argc, VALUE *argv, VALUE self) {
-  ruby_whisper *rw;
-  VALUE whisper_model_file_path;
-
-  // TODO: we can support init from buffer here too maybe another ruby object to expose
-  rb_scan_args(argc, argv, "01", &whisper_model_file_path);
-  Data_Get_Struct(self, ruby_whisper, rw);
-
-  if (!rb_respond_to(whisper_model_file_path, rb_intern("to_s"))) {
-    rb_raise(rb_eRuntimeError, "Expected file path to model to initialize Whisper::Context");
-  }
-  rw->context = whisper_init_from_file(StringValueCStr(whisper_model_file_path));
-  if (rw->context == nullptr) {
-    rb_raise(rb_eRuntimeError, "error: failed to initialize whisper context");
-  }
-  return self;
-}
-
-/*
- * transcribe a single file
- * can emit to a block results
- *
- **/
-static VALUE ruby_whisper_transcribe(int argc, VALUE *argv, VALUE self) {
-  ruby_whisper *rw;
-  ruby_whisper_params *rwp;
-  VALUE wave_file_path, blk, params;
-
-  rb_scan_args(argc, argv, "02&", &wave_file_path, &params, &blk);
-  Data_Get_Struct(self, ruby_whisper, rw);
-  Data_Get_Struct(params, ruby_whisper_params, rwp);
-
-  if (!rb_respond_to(wave_file_path, rb_intern("to_s"))) {
-    rb_raise(rb_eRuntimeError, "Expected file path to wave file");
-  }
-
-  std::string fname_inp = StringValueCStr(wave_file_path);
-
-  std::vector<float> pcmf32; // mono-channel F32 PCM
-  std::vector<std::vector<float>> pcmf32s; // stereo-channel F32 PCM
-
-  // WAV input - this is directly from main.cpp example
-  {
-    drwav wav;
-    std::vector<uint8_t> wav_data; // used for pipe input from stdin
-
-    if (fname_inp == "-") {
-      {
-        uint8_t buf[1024];
-        while (true) {
-          const size_t n = fread(buf, 1, sizeof(buf), stdin);
-          if (n == 0) {
-            break;
-          }
-          wav_data.insert(wav_data.end(), buf, buf + n);
-        }
-      }
-
-      if (drwav_init_memory(&wav, wav_data.data(), wav_data.size(), nullptr) == false) {
-        fprintf(stderr, "error: failed to open WAV file from stdin\n");
-        return self;
-      }
-
-      fprintf(stderr, "%s: read %zu bytes from stdin\n", __func__, wav_data.size());
-    } else if (drwav_init_file(&wav, fname_inp.c_str(), nullptr) == false) {
-      fprintf(stderr, "error: failed to open '%s' as WAV file\n", fname_inp.c_str());
-      return self;
-    }
-
-    if (wav.channels != 1 && wav.channels != 2) {
-      fprintf(stderr, "WAV file '%s' must be mono or stereo\n", fname_inp.c_str());
-      return self;
-    }
-
-    if (rwp->diarize && wav.channels != 2 && rwp->params.print_timestamps == false) {
-      fprintf(stderr, "WAV file '%s' must be stereo for diarization and timestamps have to be enabled\n", fname_inp.c_str());
-      return self;
-    }
-
-    if (wav.sampleRate != WHISPER_SAMPLE_RATE) {
-      fprintf(stderr, "WAV file '%s' must be %i kHz\n", fname_inp.c_str(), WHISPER_SAMPLE_RATE/1000);
-      return self;
-    }
-
-    if (wav.bitsPerSample != 16) {
-      fprintf(stderr, "WAV file '%s' must be 16-bit\n", fname_inp.c_str());
-      return self;
-    }
-
-    const uint64_t n = wav_data.empty() ? wav.totalPCMFrameCount : wav_data.size()/(wav.channels*wav.bitsPerSample/8);
-
-    std::vector<int16_t> pcm16;
-    pcm16.resize(n*wav.channels);
-    drwav_read_pcm_frames_s16(&wav, n, pcm16.data());
-    drwav_uninit(&wav);
-
-    // convert to mono, float
-    pcmf32.resize(n);
-    if (wav.channels == 1) {
-      for (uint64_t i = 0; i < n; i++) {
-        pcmf32[i] = float(pcm16[i])/32768.0f;
-      }
-    } else {
-      for (uint64_t i = 0; i < n; i++) {
-        pcmf32[i] = float(pcm16[2*i] + pcm16[2*i + 1])/65536.0f;
-      }
-    }
-
-    if (rwp->diarize) {
-      // convert to stereo, float
-      pcmf32s.resize(2);
-
-      pcmf32s[0].resize(n);
-      pcmf32s[1].resize(n);
-      for (uint64_t i = 0; i < n; i++) {
-        pcmf32s[0][i] = float(pcm16[2*i])/32768.0f;
-        pcmf32s[1][i] = float(pcm16[2*i + 1])/32768.0f;
-      }
-    }
-  }
-  {
-    static bool is_aborted = false; // NOTE: this should be atomic to avoid data race
-
-    rwp->params.encoder_begin_callback = [](struct whisper_context * /*ctx*/, struct whisper_state * /*state*/, void * user_data) {
-      bool is_aborted = *(bool*)user_data;
-      return !is_aborted;
-    };
-    rwp->params.encoder_begin_callback_user_data = &is_aborted;
-  }
-
-  if (whisper_full_parallel(rw->context, rwp->params, pcmf32.data(), pcmf32.size(), 1) != 0) {
-    fprintf(stderr, "failed to process audio\n");
-    return self;
-  }
-  const int n_segments = whisper_full_n_segments(rw->context);
-  VALUE output = rb_str_new2("");
-  for (int i = 0; i < n_segments; ++i) {
-    const char * text = whisper_full_get_segment_text(rw->context, i);
-    output = rb_str_concat(output, rb_str_new2(text));
-  }
-  VALUE idCall = rb_intern("call");
-  if (blk != Qnil) {
-    rb_funcall(blk, idCall, 1, output);
-  }
-  return self;
-}
-
-/*
- * params.language = "auto" | "en", etc...
- */
-static VALUE ruby_whisper_params_set_language(VALUE self, VALUE value) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  if (value == Qfalse || value == Qnil) {
-    rwp->params.language = "auto";
-  } else {
-    rwp->params.language = StringValueCStr(value);
-  }
-  return value;
-}
-static VALUE ruby_whisper_params_get_language(VALUE self) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  if (rwp->params.language) {
-    return rb_str_new2(rwp->params.language);
-  } else {
-    return rb_str_new2("auto");
-  }
-}
-static VALUE ruby_whisper_params_set_translate(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, translate, value)
-}
-static VALUE ruby_whisper_params_get_translate(VALUE self) {
-  BOOL_PARAMS_GETTER(self, translate)
-}
-static VALUE ruby_whisper_params_set_no_context(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, no_context, value)
-}
-static VALUE ruby_whisper_params_get_no_context(VALUE self) {
-  BOOL_PARAMS_GETTER(self, no_context)
-}
-static VALUE ruby_whisper_params_set_single_segment(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, single_segment, value)
-}
-static VALUE ruby_whisper_params_get_single_segment(VALUE self) {
-  BOOL_PARAMS_GETTER(self, single_segment)
-}
-static VALUE ruby_whisper_params_set_print_special(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, print_special, value)
-}
-static VALUE ruby_whisper_params_get_print_special(VALUE self) {
-  BOOL_PARAMS_GETTER(self, print_special)
-}
-static VALUE ruby_whisper_params_set_print_progress(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, print_progress, value)
-}
-static VALUE ruby_whisper_params_get_print_progress(VALUE self) {
-  BOOL_PARAMS_GETTER(self, print_progress)
-}
-static VALUE ruby_whisper_params_set_print_realtime(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, print_realtime, value)
-}
-static VALUE ruby_whisper_params_get_print_realtime(VALUE self) {
-  BOOL_PARAMS_GETTER(self, print_realtime)
-}
-static VALUE ruby_whisper_params_set_print_timestamps(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, print_timestamps, value)
-}
-static VALUE ruby_whisper_params_get_print_timestamps(VALUE self) {
-  BOOL_PARAMS_GETTER(self, print_timestamps)
-}
-static VALUE ruby_whisper_params_set_suppress_blank(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, suppress_blank, value)
-}
-static VALUE ruby_whisper_params_get_suppress_blank(VALUE self) {
-  BOOL_PARAMS_GETTER(self, suppress_blank)
-}
-static VALUE ruby_whisper_params_set_suppress_non_speech_tokens(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, suppress_non_speech_tokens, value)
-}
-static VALUE ruby_whisper_params_get_suppress_non_speech_tokens(VALUE self) {
-  BOOL_PARAMS_GETTER(self, suppress_non_speech_tokens)
-}
-static VALUE ruby_whisper_params_get_token_timestamps(VALUE self) {
-  BOOL_PARAMS_GETTER(self, token_timestamps)
-}
-static VALUE ruby_whisper_params_set_token_timestamps(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, token_timestamps, value)
-}
-static VALUE ruby_whisper_params_get_split_on_word(VALUE self) {
-  BOOL_PARAMS_GETTER(self, split_on_word)
-}
-static VALUE ruby_whisper_params_set_split_on_word(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, split_on_word, value)
-}
-static VALUE ruby_whisper_params_get_speed_up(VALUE self) {
-  BOOL_PARAMS_GETTER(self, speed_up)
-}
-static VALUE ruby_whisper_params_set_speed_up(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, speed_up, value)
-}
-static VALUE ruby_whisper_params_get_diarize(VALUE self) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  if (rwp->diarize) {
-    return Qtrue;
-  } else {
-    return Qfalse;
-  }
-}
-static VALUE ruby_whisper_params_set_diarize(VALUE self, VALUE value) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  if (value == Qfalse || value == Qnil) {
-    rwp->diarize = false;
-  } else {
-    rwp->diarize = true;
-  } \
-  return value;
-}
-
-static VALUE ruby_whisper_params_get_offset(VALUE self) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  return INT2NUM(rwp->params.offset_ms);
-}
-static VALUE ruby_whisper_params_set_offset(VALUE self, VALUE value) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  rwp->params.offset_ms = NUM2INT(value);
-  return value;
-}
-static VALUE ruby_whisper_params_get_duration(VALUE self) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  return INT2NUM(rwp->params.duration_ms);
-}
-static VALUE ruby_whisper_params_set_duration(VALUE self, VALUE value) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  rwp->params.duration_ms = NUM2INT(value);
-  return value;
-}
-
-static VALUE ruby_whisper_params_get_max_text_tokens(VALUE self) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  return INT2NUM(rwp->params.n_max_text_ctx);
-}
-static VALUE ruby_whisper_params_set_max_text_tokens(VALUE self, VALUE value) {
-  ruby_whisper_params *rwp;
-  Data_Get_Struct(self, ruby_whisper_params, rwp);
-  rwp->params.n_max_text_ctx = NUM2INT(value);
-  return value;
-}
-
-void Init_whisper() {
-  mWhisper = rb_define_module("Whisper");
-  cContext = rb_define_class_under(mWhisper, "Context", rb_cObject);
-  cParams  = rb_define_class_under(mWhisper, "Params", rb_cObject);
-
-  rb_define_alloc_func(cContext, ruby_whisper_allocate);
-  rb_define_method(cContext, "initialize", ruby_whisper_initialize, -1);
-
-  rb_define_method(cContext, "transcribe", ruby_whisper_transcribe, -1);
-
-  rb_define_alloc_func(cParams, ruby_whisper_params_allocate);
-
-  rb_define_method(cParams, "language=", ruby_whisper_params_set_language, 1);
-  rb_define_method(cParams, "language", ruby_whisper_params_get_language, 0);
-  rb_define_method(cParams, "translate=", ruby_whisper_params_set_translate, 1);
-  rb_define_method(cParams, "translate", ruby_whisper_params_get_translate, 0);
-  rb_define_method(cParams, "no_context=", ruby_whisper_params_set_no_context, 1);
-  rb_define_method(cParams, "no_context", ruby_whisper_params_get_no_context, 0);
-  rb_define_method(cParams, "single_segment=", ruby_whisper_params_set_single_segment, 1);
-  rb_define_method(cParams, "single_segment", ruby_whisper_params_get_single_segment, 0);
-  rb_define_method(cParams, "print_special", ruby_whisper_params_get_print_special, 0);
-  rb_define_method(cParams, "print_special=", ruby_whisper_params_set_print_special, 1);
-  rb_define_method(cParams, "print_progress", ruby_whisper_params_get_print_progress, 0);
-  rb_define_method(cParams, "print_progress=", ruby_whisper_params_set_print_progress, 1);
-  rb_define_method(cParams, "print_realtime", ruby_whisper_params_get_print_realtime, 0);
-  rb_define_method(cParams, "print_realtime=", ruby_whisper_params_set_print_realtime, 1);
-  rb_define_method(cParams, "print_timestamps", ruby_whisper_params_get_print_timestamps, 0);
-  rb_define_method(cParams, "print_timestamps=", ruby_whisper_params_set_print_timestamps, 1);
-  rb_define_method(cParams, "suppress_blank", ruby_whisper_params_get_suppress_blank, 0);
-  rb_define_method(cParams, "suppress_blank=", ruby_whisper_params_set_suppress_blank, 1);
-  rb_define_method(cParams, "suppress_non_speech_tokens", ruby_whisper_params_get_suppress_non_speech_tokens, 0);
-  rb_define_method(cParams, "suppress_non_speech_tokens=", ruby_whisper_params_set_suppress_non_speech_tokens, 1);
-  rb_define_method(cParams, "token_timestamps", ruby_whisper_params_get_token_timestamps, 0);
-  rb_define_method(cParams, "token_timestamps=", ruby_whisper_params_set_token_timestamps, 1);
-  rb_define_method(cParams, "split_on_word", ruby_whisper_params_get_split_on_word, 0);
-  rb_define_method(cParams, "split_on_word=", ruby_whisper_params_set_split_on_word, 1);
-  rb_define_method(cParams, "speed_up", ruby_whisper_params_get_speed_up, 0);
-  rb_define_method(cParams, "speed_up=", ruby_whisper_params_set_speed_up, 1);
-  rb_define_method(cParams, "diarize", ruby_whisper_params_get_diarize, 0);
-  rb_define_method(cParams, "diarize=", ruby_whisper_params_set_diarize, 1);
-
-  rb_define_method(cParams, "offset", ruby_whisper_params_get_offset, 0);
-  rb_define_method(cParams, "offset=", ruby_whisper_params_set_offset, 1);
-  rb_define_method(cParams, "duration", ruby_whisper_params_get_duration, 0);
-  rb_define_method(cParams, "duration=", ruby_whisper_params_set_duration, 1);
-
-  rb_define_method(cParams, "max_text_tokens", ruby_whisper_params_get_max_text_tokens, 0);
-  rb_define_method(cParams, "max_text_tokens=", ruby_whisper_params_set_max_text_tokens, 1);
-}
-#ifdef __cplusplus
-}
-#endif

bindings/ruby/ext/ruby_whisper.h

@@ -1,15 +0,0 @@
-#ifndef __RUBY_WHISPER_H
-#define __RUBY_WHISPER_H
-
-#include "whisper.h"
-
-typedef struct {
-  struct whisper_context *context;
-} ruby_whisper;
-
-typedef struct {
-  struct whisper_full_params params;
-  bool diarize;
-} ruby_whisper_params;
-
-#endif

bindings/ruby/tests/test_whisper.rb

@@ -1,138 +0,0 @@
-TOPDIR = File.expand_path(File.join(File.dirname(__FILE__), '..'))
-EXTDIR = File.join(TOPDIR, 'ext')
-#$LIBDIR = File.join(TOPDIR, 'lib')
-#$:.unshift(LIBDIR)
-$:.unshift(EXTDIR)
-
-require 'whisper'
-require 'test/unit'
-
-class TestWhisper < Test::Unit::TestCase
-  def setup
-    @params  = Whisper::Params.new
-  end
-
-  def test_language
-    @params.language = "en"
-    assert_equal @params.language, "en"
-    @params.language = "auto"
-    assert_equal @params.language, "auto"
-  end
-
-  def test_offset
-    @params.offset = 10_000
-    assert_equal @params.offset, 10_000
-    @params.offset = 0
-    assert_equal @params.offset, 0
-  end
-
-  def test_duration
-    @params.duration = 60_000
-    assert_equal @params.duration, 60_000
-    @params.duration = 0
-    assert_equal @params.duration, 0
-  end
-
-  def test_max_text_tokens
-    @params.max_text_tokens = 300
-    assert_equal @params.max_text_tokens, 300
-    @params.max_text_tokens = 0
-    assert_equal @params.max_text_tokens, 0
-  end
-
-  def test_translate
-    @params.translate = true
-    assert @params.translate
-    @params.translate = false
-    assert !@params.translate
-  end
-
-  def test_no_context
-    @params.no_context = true
-    assert @params.no_context
-    @params.no_context = false
-    assert !@params.no_context
-  end
-
-  def test_single_segment
-    @params.single_segment = true
-    assert @params.single_segment
-    @params.single_segment = false
-    assert !@params.single_segment
-  end
-
-  def test_print_special
-    @params.print_special = true
-    assert @params.print_special
-    @params.print_special = false
-    assert !@params.print_special
-  end
-
-  def test_print_progress
-    @params.print_progress = true
-    assert @params.print_progress
-    @params.print_progress = false
-    assert !@params.print_progress
-  end
-
-  def test_print_realtime
-    @params.print_realtime = true
-    assert @params.print_realtime
-    @params.print_realtime = false
-    assert !@params.print_realtime
-  end
-
-  def test_print_timestamps
-    @params.print_timestamps = true
-    assert @params.print_timestamps
-    @params.print_timestamps = false
-    assert !@params.print_timestamps
-  end
-
-  def test_suppress_blank
-    @params.suppress_blank = true
-    assert @params.suppress_blank
-    @params.suppress_blank = false
-    assert !@params.suppress_blank
-  end
-
-  def test_suppress_non_speech_tokens
-    @params.suppress_non_speech_tokens = true
-    assert @params.suppress_non_speech_tokens
-    @params.suppress_non_speech_tokens = false
-    assert !@params.suppress_non_speech_tokens
-  end
-
-  def test_token_timestamps
-    @params.token_timestamps = true
-    assert @params.token_timestamps
-    @params.token_timestamps = false
-    assert !@params.token_timestamps
-  end
-
-  def test_split_on_word
-    @params.split_on_word = true
-    assert @params.split_on_word
-    @params.split_on_word = false
-    assert !@params.split_on_word
-  end
-
-  def test_speed_up
-    @params.speed_up = true
-    assert @params.speed_up
-    @params.speed_up = false
-    assert !@params.speed_up
-  end
-
-  def test_whisper
-    @whisper = Whisper::Context.new(File.join(TOPDIR, '..', '..', 'models', 'ggml-base.en.bin'))
-    params  = Whisper::Params.new
-    params.print_timestamps = false
-
-    jfk = File.join(TOPDIR, '..', '..', 'samples', 'jfk.wav')
-    @whisper.transcribe(jfk, params) {|text|
-      assert_match /ask not what your country can do for you, ask what you can do for your country/, text
-    }
-  end
-
-end

cmake/DefaultTargetOptions.cmake

@@ -1,17 +0,0 @@
-# Set the default compile features and properties for a target.
-
-if (NOT TARGET)
-    message(FATAL_ERROR "TARGET not set before including DefaultTargetOptions")
-endif()
-
-target_compile_features(${TARGET}
-    PRIVATE
-        cxx_std_11
-    )
-
-set_target_properties(${TARGET}
-    PROPERTIES
-        EXPORT_COMPILE_COMMANDS ON
-        RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/bin"
-        INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib"
-)

coreml/whisper-decoder-impl.h

@@ -1,146 +0,0 @@
-//
-// whisper-decoder-impl.h
-//
-// This file was automatically generated and should not be edited.
-//
-
-#import <Foundation/Foundation.h>
-#import <CoreML/CoreML.h>
-#include <stdint.h>
-#include <os/log.h>
-
-NS_ASSUME_NONNULL_BEGIN
-
-
-/// Model Prediction Input Type
-API_AVAILABLE(macos(12.0), ios(15.0), watchos(8.0), tvos(15.0)) __attribute__((visibility("hidden")))
-@interface whisper_decoder_implInput : NSObject<MLFeatureProvider>
-
-/// token_data as 1 by 1 matrix of 32-bit integers
-@property (readwrite, nonatomic, strong) MLMultiArray * token_data;
-
-/// audio_data as 1 × 384 × 1 × 1500 4-dimensional array of floats
-@property (readwrite, nonatomic, strong) MLMultiArray * audio_data;
-- (instancetype)init NS_UNAVAILABLE;
-- (instancetype)initWithToken_data:(MLMultiArray *)token_data audio_data:(MLMultiArray *)audio_data NS_DESIGNATED_INITIALIZER;
-
-@end
-
-
-/// Model Prediction Output Type
-API_AVAILABLE(macos(12.0), ios(15.0), watchos(8.0), tvos(15.0)) __attribute__((visibility("hidden")))
-@interface whisper_decoder_implOutput : NSObject<MLFeatureProvider>
-
-/// var_1346 as multidimensional array of floats
-@property (readwrite, nonatomic, strong) MLMultiArray * var_1346;
-- (instancetype)init NS_UNAVAILABLE;
-- (instancetype)initWithVar_1346:(MLMultiArray *)var_1346 NS_DESIGNATED_INITIALIZER;
-
-@end
-
-
-/// Class for model loading and prediction
-API_AVAILABLE(macos(12.0), ios(15.0), watchos(8.0), tvos(15.0)) __attribute__((visibility("hidden")))
-@interface whisper_decoder_impl : NSObject
-@property (readonly, nonatomic, nullable) MLModel * model;
-
-/**
-    URL of the underlying .mlmodelc directory.
-*/
-+ (nullable NSURL *)URLOfModelInThisBundle;
-
-/**
-    Initialize whisper_decoder_impl instance from an existing MLModel object.
-
-    Usually the application does not use this initializer unless it makes a subclass of whisper_decoder_impl.
-    Such application may want to use `-[MLModel initWithContentsOfURL:configuration:error:]` and `+URLOfModelInThisBundle` to create a MLModel object to pass-in.
-*/
-- (instancetype)initWithMLModel:(MLModel *)model NS_DESIGNATED_INITIALIZER;
-
-/**
-    Initialize whisper_decoder_impl instance with the model in this bundle.
-*/
-- (nullable instancetype)init;
-
-/**
-    Initialize whisper_decoder_impl instance with the model in this bundle.
-
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithConfiguration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Initialize whisper_decoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_decoder_impl.
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Initialize whisper_decoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_decoder_impl.
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Construct whisper_decoder_impl instance asynchronously with configuration.
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_decoder_impl instance or NSError object.
-*/
-+ (void)loadWithConfiguration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_decoder_impl * _Nullable model, NSError * _Nullable error))handler;
-
-/**
-    Construct whisper_decoder_impl instance asynchronously with URL of .mlmodelc directory and optional configuration.
-
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param modelURL The model URL.
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_decoder_impl instance or NSError object.
-*/
-+ (void)loadContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_decoder_impl * _Nullable model, NSError * _Nullable error))handler;
-
-/**
-    Make a prediction using the standard interface
-    @param input an instance of whisper_decoder_implInput to predict from
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the prediction as whisper_decoder_implOutput
-*/
-- (nullable whisper_decoder_implOutput *)predictionFromFeatures:(whisper_decoder_implInput *)input error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Make a prediction using the standard interface
-    @param input an instance of whisper_decoder_implInput to predict from
-    @param options prediction options
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the prediction as whisper_decoder_implOutput
-*/
-- (nullable whisper_decoder_implOutput *)predictionFromFeatures:(whisper_decoder_implInput *)input options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Make a prediction using the convenience interface
-    @param token_data as 1 by 1 matrix of 32-bit integers:
-    @param audio_data as 1 × 384 × 1 × 1500 4-dimensional array of floats:
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the prediction as whisper_decoder_implOutput
-*/
-- (nullable whisper_decoder_implOutput *)predictionFromToken_data:(MLMultiArray *)token_data audio_data:(MLMultiArray *)audio_data error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Batch prediction
-    @param inputArray array of whisper_decoder_implInput instances to obtain predictions from
-    @param options prediction options
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the predictions as NSArray<whisper_decoder_implOutput *>
-*/
-- (nullable NSArray<whisper_decoder_implOutput *> *)predictionsFromInputs:(NSArray<whisper_decoder_implInput*> *)inputArray options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-@end
-
-NS_ASSUME_NONNULL_END

coreml/whisper-decoder-impl.m

@@ -1,201 +0,0 @@
-//
-// whisper-decoder-impl.m
-//
-// This file was automatically generated and should not be edited.
-//
-
-#if !__has_feature(objc_arc)
-#error This file must be compiled with automatic reference counting enabled (-fobjc-arc)
-#endif
-
-#import "whisper-decoder-impl.h"
-
-@implementation whisper_decoder_implInput
-
-- (instancetype)initWithToken_data:(MLMultiArray *)token_data audio_data:(MLMultiArray *)audio_data {
-    self = [super init];
-    if (self) {
-        _token_data = token_data;
-        _audio_data = audio_data;
-    }
-    return self;
-}
-
-- (NSSet<NSString *> *)featureNames {
-    return [NSSet setWithArray:@[@"token_data", @"audio_data"]];
-}
-
-- (nullable MLFeatureValue *)featureValueForName:(NSString *)featureName {
-    if ([featureName isEqualToString:@"token_data"]) {
-        return [MLFeatureValue featureValueWithMultiArray:self.token_data];
-    }
-    if ([featureName isEqualToString:@"audio_data"]) {
-        return [MLFeatureValue featureValueWithMultiArray:self.audio_data];
-    }
-    return nil;
-}
-
-@end
-
-@implementation whisper_decoder_implOutput
-
-- (instancetype)initWithVar_1346:(MLMultiArray *)var_1346 {
-    self = [super init];
-    if (self) {
-        _var_1346 = var_1346;
-    }
-    return self;
-}
-
-- (NSSet<NSString *> *)featureNames {
-    return [NSSet setWithArray:@[@"var_1346"]];
-}
-
-- (nullable MLFeatureValue *)featureValueForName:(NSString *)featureName {
-    if ([featureName isEqualToString:@"var_1346"]) {
-        return [MLFeatureValue featureValueWithMultiArray:self.var_1346];
-    }
-    return nil;
-}
-
-@end
-
-@implementation whisper_decoder_impl
-
-
-/**
-    URL of the underlying .mlmodelc directory.
-*/
-+ (nullable NSURL *)URLOfModelInThisBundle {
-    NSString *assetPath = [[NSBundle bundleForClass:[self class]] pathForResource:@"whisper_decoder_impl" ofType:@"mlmodelc"];
-    if (nil == assetPath) { os_log_error(OS_LOG_DEFAULT, "Could not load whisper-decoder-impl.mlmodelc in the bundle resource"); return nil; }
-    return [NSURL fileURLWithPath:assetPath];
-}
-
-
-/**
-    Initialize whisper_decoder_impl instance from an existing MLModel object.
-
-    Usually the application does not use this initializer unless it makes a subclass of whisper_decoder_impl.
-    Such application may want to use `-[MLModel initWithContentsOfURL:configuration:error:]` and `+URLOfModelInThisBundle` to create a MLModel object to pass-in.
-*/
-- (instancetype)initWithMLModel:(MLModel *)model {
-    self = [super init];
-    if (!self) { return nil; }
-    _model = model;
-    if (_model == nil) { return nil; }
-    return self;
-}
-
-
-/**
-    Initialize whisper_decoder_impl instance with the model in this bundle.
-*/
-- (nullable instancetype)init {
-    return [self initWithContentsOfURL:(NSURL * _Nonnull)self.class.URLOfModelInThisBundle error:nil];
-}
-
-
-/**
-    Initialize whisper_decoder_impl instance with the model in this bundle.
-
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithConfiguration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    return [self initWithContentsOfURL:(NSURL * _Nonnull)self.class.URLOfModelInThisBundle configuration:configuration error:error];
-}
-
-
-/**
-    Initialize whisper_decoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_decoder_impl.
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    MLModel *model = [MLModel modelWithContentsOfURL:modelURL error:error];
-    if (model == nil) { return nil; }
-    return [self initWithMLModel:model];
-}
-
-
-/**
-    Initialize whisper_decoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_decoder_impl.
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    MLModel *model = [MLModel modelWithContentsOfURL:modelURL configuration:configuration error:error];
-    if (model == nil) { return nil; }
-    return [self initWithMLModel:model];
-}
-
-
-/**
-    Construct whisper_decoder_impl instance asynchronously with configuration.
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_decoder_impl instance or NSError object.
-*/
-+ (void)loadWithConfiguration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_decoder_impl * _Nullable model, NSError * _Nullable error))handler {
-    [self loadContentsOfURL:(NSURL * _Nonnull)[self URLOfModelInThisBundle]
-              configuration:configuration
-          completionHandler:handler];
-}
-
-
-/**
-    Construct whisper_decoder_impl instance asynchronously with URL of .mlmodelc directory and optional configuration.
-
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param modelURL The model URL.
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_decoder_impl instance or NSError object.
-*/
-+ (void)loadContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_decoder_impl * _Nullable model, NSError * _Nullable error))handler {
-    [MLModel loadContentsOfURL:modelURL
-                 configuration:configuration
-             completionHandler:^(MLModel *model, NSError *error) {
-        if (model != nil) {
-            whisper_decoder_impl *typedModel = [[whisper_decoder_impl alloc] initWithMLModel:model];
-            handler(typedModel, nil);
-        } else {
-            handler(nil, error);
-        }
-    }];
-}
-
-- (nullable whisper_decoder_implOutput *)predictionFromFeatures:(whisper_decoder_implInput *)input error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    return [self predictionFromFeatures:input options:[[MLPredictionOptions alloc] init] error:error];
-}
-
-- (nullable whisper_decoder_implOutput *)predictionFromFeatures:(whisper_decoder_implInput *)input options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    id<MLFeatureProvider> outFeatures = [self.model predictionFromFeatures:input options:options error:error];
-    if (!outFeatures) { return nil; }
-    return [[whisper_decoder_implOutput alloc] initWithVar_1346:(MLMultiArray *)[outFeatures featureValueForName:@"var_1346"].multiArrayValue];
-}
-
-- (nullable whisper_decoder_implOutput *)predictionFromToken_data:(MLMultiArray *)token_data audio_data:(MLMultiArray *)audio_data error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    whisper_decoder_implInput *input_ = [[whisper_decoder_implInput alloc] initWithToken_data:token_data audio_data:audio_data];
-    return [self predictionFromFeatures:input_ error:error];
-}
-
-- (nullable NSArray<whisper_decoder_implOutput *> *)predictionsFromInputs:(NSArray<whisper_decoder_implInput*> *)inputArray options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    id<MLBatchProvider> inBatch = [[MLArrayBatchProvider alloc] initWithFeatureProviderArray:inputArray];
-    id<MLBatchProvider> outBatch = [self.model predictionsFromBatch:inBatch options:options error:error];
-    if (!outBatch) { return nil; }
-    NSMutableArray<whisper_decoder_implOutput*> *results = [NSMutableArray arrayWithCapacity:(NSUInteger)outBatch.count];
-    for (NSInteger i = 0; i < outBatch.count; i++) {
-        id<MLFeatureProvider> resultProvider = [outBatch featuresAtIndex:i];
-        whisper_decoder_implOutput * result = [[whisper_decoder_implOutput alloc] initWithVar_1346:(MLMultiArray *)[resultProvider featureValueForName:@"var_1346"].multiArrayValue];
-        [results addObject:result];
-    }
-    return results;
-}
-
-@end

coreml/whisper-encoder-impl.h

@@ -1,142 +0,0 @@
-//
-// whisper-encoder-impl.h
-//
-// This file was automatically generated and should not be edited.
-//
-
-#import <Foundation/Foundation.h>
-#import <CoreML/CoreML.h>
-#include <stdint.h>
-#include <os/log.h>
-
-NS_ASSUME_NONNULL_BEGIN
-
-
-/// Model Prediction Input Type
-API_AVAILABLE(macos(12.0), ios(15.0), watchos(8.0), tvos(15.0)) __attribute__((visibility("hidden")))
-@interface whisper_encoder_implInput : NSObject<MLFeatureProvider>
-
-/// logmel_data as 1 × 80 × 3000 3-dimensional array of floats
-@property (readwrite, nonatomic, strong) MLMultiArray * logmel_data;
-- (instancetype)init NS_UNAVAILABLE;
-- (instancetype)initWithLogmel_data:(MLMultiArray *)logmel_data NS_DESIGNATED_INITIALIZER;
-
-@end
-
-
-/// Model Prediction Output Type
-API_AVAILABLE(macos(12.0), ios(15.0), watchos(8.0), tvos(15.0)) __attribute__((visibility("hidden")))
-@interface whisper_encoder_implOutput : NSObject<MLFeatureProvider>
-
-/// output as multidimensional array of floats
-@property (readwrite, nonatomic, strong) MLMultiArray * output;
-- (instancetype)init NS_UNAVAILABLE;
-- (instancetype)initWithOutput:(MLMultiArray *)output NS_DESIGNATED_INITIALIZER;
-
-@end
-
-
-/// Class for model loading and prediction
-API_AVAILABLE(macos(12.0), ios(15.0), watchos(8.0), tvos(15.0)) __attribute__((visibility("hidden")))
-@interface whisper_encoder_impl : NSObject
-@property (readonly, nonatomic, nullable) MLModel * model;
-
-/**
-    URL of the underlying .mlmodelc directory.
-*/
-+ (nullable NSURL *)URLOfModelInThisBundle;
-
-/**
-    Initialize whisper_encoder_impl instance from an existing MLModel object.
-
-    Usually the application does not use this initializer unless it makes a subclass of whisper_encoder_impl.
-    Such application may want to use `-[MLModel initWithContentsOfURL:configuration:error:]` and `+URLOfModelInThisBundle` to create a MLModel object to pass-in.
-*/
-- (instancetype)initWithMLModel:(MLModel *)model NS_DESIGNATED_INITIALIZER;
-
-/**
-    Initialize whisper_encoder_impl instance with the model in this bundle.
-*/
-- (nullable instancetype)init;
-
-/**
-    Initialize whisper_encoder_impl instance with the model in this bundle.
-
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithConfiguration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Initialize whisper_encoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_encoder_impl.
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Initialize whisper_encoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_encoder_impl.
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Construct whisper_encoder_impl instance asynchronously with configuration.
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_encoder_impl instance or NSError object.
-*/
-+ (void)loadWithConfiguration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_encoder_impl * _Nullable model, NSError * _Nullable error))handler;
-
-/**
-    Construct whisper_encoder_impl instance asynchronously with URL of .mlmodelc directory and optional configuration.
-
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param modelURL The model URL.
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_encoder_impl instance or NSError object.
-*/
-+ (void)loadContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_encoder_impl * _Nullable model, NSError * _Nullable error))handler;
-
-/**
-    Make a prediction using the standard interface
-    @param input an instance of whisper_encoder_implInput to predict from
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the prediction as whisper_encoder_implOutput
-*/
-- (nullable whisper_encoder_implOutput *)predictionFromFeatures:(whisper_encoder_implInput *)input error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Make a prediction using the standard interface
-    @param input an instance of whisper_encoder_implInput to predict from
-    @param options prediction options
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the prediction as whisper_encoder_implOutput
-*/
-- (nullable whisper_encoder_implOutput *)predictionFromFeatures:(whisper_encoder_implInput *)input options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Make a prediction using the convenience interface
-    @param logmel_data as 1 × 80 × 3000 3-dimensional array of floats:
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the prediction as whisper_encoder_implOutput
-*/
-- (nullable whisper_encoder_implOutput *)predictionFromLogmel_data:(MLMultiArray *)logmel_data error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-
-/**
-    Batch prediction
-    @param inputArray array of whisper_encoder_implInput instances to obtain predictions from
-    @param options prediction options
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-    @return the predictions as NSArray<whisper_encoder_implOutput *>
-*/
-- (nullable NSArray<whisper_encoder_implOutput *> *)predictionsFromInputs:(NSArray<whisper_encoder_implInput*> *)inputArray options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error;
-@end
-
-NS_ASSUME_NONNULL_END

coreml/whisper-encoder-impl.m

@@ -1,197 +0,0 @@
-//
-// whisper-encoder-impl.m
-//
-// This file was automatically generated and should not be edited.
-//
-
-#if !__has_feature(objc_arc)
-#error This file must be compiled with automatic reference counting enabled (-fobjc-arc)
-#endif
-
-#import "whisper-encoder-impl.h"
-
-@implementation whisper_encoder_implInput
-
-- (instancetype)initWithLogmel_data:(MLMultiArray *)logmel_data {
-    self = [super init];
-    if (self) {
-        _logmel_data = logmel_data;
-    }
-    return self;
-}
-
-- (NSSet<NSString *> *)featureNames {
-    return [NSSet setWithArray:@[@"logmel_data"]];
-}
-
-- (nullable MLFeatureValue *)featureValueForName:(NSString *)featureName {
-    if ([featureName isEqualToString:@"logmel_data"]) {
-        return [MLFeatureValue featureValueWithMultiArray:self.logmel_data];
-    }
-    return nil;
-}
-
-@end
-
-@implementation whisper_encoder_implOutput
-
-- (instancetype)initWithOutput:(MLMultiArray *)output {
-    self = [super init];
-    if (self) {
-        _output = output;
-    }
-    return self;
-}
-
-- (NSSet<NSString *> *)featureNames {
-    return [NSSet setWithArray:@[@"output"]];
-}
-
-- (nullable MLFeatureValue *)featureValueForName:(NSString *)featureName {
-    if ([featureName isEqualToString:@"output"]) {
-        return [MLFeatureValue featureValueWithMultiArray:self.output];
-    }
-    return nil;
-}
-
-@end
-
-@implementation whisper_encoder_impl
-
-
-/**
-    URL of the underlying .mlmodelc directory.
-*/
-+ (nullable NSURL *)URLOfModelInThisBundle {
-    NSString *assetPath = [[NSBundle bundleForClass:[self class]] pathForResource:@"whisper_encoder_impl" ofType:@"mlmodelc"];
-    if (nil == assetPath) { os_log_error(OS_LOG_DEFAULT, "Could not load whisper-encoder-impl.mlmodelc in the bundle resource"); return nil; }
-    return [NSURL fileURLWithPath:assetPath];
-}
-
-
-/**
-    Initialize whisper_encoder_impl instance from an existing MLModel object.
-
-    Usually the application does not use this initializer unless it makes a subclass of whisper_encoder_impl.
-    Such application may want to use `-[MLModel initWithContentsOfURL:configuration:error:]` and `+URLOfModelInThisBundle` to create a MLModel object to pass-in.
-*/
-- (instancetype)initWithMLModel:(MLModel *)model {
-    self = [super init];
-    if (!self) { return nil; }
-    _model = model;
-    if (_model == nil) { return nil; }
-    return self;
-}
-
-
-/**
-    Initialize whisper_encoder_impl instance with the model in this bundle.
-*/
-- (nullable instancetype)init {
-    return [self initWithContentsOfURL:(NSURL * _Nonnull)self.class.URLOfModelInThisBundle error:nil];
-}
-
-
-/**
-    Initialize whisper_encoder_impl instance with the model in this bundle.
-
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithConfiguration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    return [self initWithContentsOfURL:(NSURL * _Nonnull)self.class.URLOfModelInThisBundle configuration:configuration error:error];
-}
-
-
-/**
-    Initialize whisper_encoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_encoder_impl.
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    MLModel *model = [MLModel modelWithContentsOfURL:modelURL error:error];
-    if (model == nil) { return nil; }
-    return [self initWithMLModel:model];
-}
-
-
-/**
-    Initialize whisper_encoder_impl instance from the model URL.
-
-    @param modelURL URL to the .mlmodelc directory for whisper_encoder_impl.
-    @param configuration The model configuration object
-    @param error If an error occurs, upon return contains an NSError object that describes the problem. If you are not interested in possible errors, pass in NULL.
-*/
-- (nullable instancetype)initWithContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    MLModel *model = [MLModel modelWithContentsOfURL:modelURL configuration:configuration error:error];
-    if (model == nil) { return nil; }
-    return [self initWithMLModel:model];
-}
-
-
-/**
-    Construct whisper_encoder_impl instance asynchronously with configuration.
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_encoder_impl instance or NSError object.
-*/
-+ (void)loadWithConfiguration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_encoder_impl * _Nullable model, NSError * _Nullable error))handler {
-    [self loadContentsOfURL:(NSURL * _Nonnull)[self URLOfModelInThisBundle]
-              configuration:configuration
-          completionHandler:handler];
-}
-
-
-/**
-    Construct whisper_encoder_impl instance asynchronously with URL of .mlmodelc directory and optional configuration.
-
-    Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.
-
-    @param modelURL The model URL.
-    @param configuration The model configuration
-    @param handler When the model load completes successfully or unsuccessfully, the completion handler is invoked with a valid whisper_encoder_impl instance or NSError object.
-*/
-+ (void)loadContentsOfURL:(NSURL *)modelURL configuration:(MLModelConfiguration *)configuration completionHandler:(void (^)(whisper_encoder_impl * _Nullable model, NSError * _Nullable error))handler {
-    [MLModel loadContentsOfURL:modelURL
-                 configuration:configuration
-             completionHandler:^(MLModel *model, NSError *error) {
-        if (model != nil) {
-            whisper_encoder_impl *typedModel = [[whisper_encoder_impl alloc] initWithMLModel:model];
-            handler(typedModel, nil);
-        } else {
-            handler(nil, error);
-        }
-    }];
-}
-
-- (nullable whisper_encoder_implOutput *)predictionFromFeatures:(whisper_encoder_implInput *)input error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    return [self predictionFromFeatures:input options:[[MLPredictionOptions alloc] init] error:error];
-}
-
-- (nullable whisper_encoder_implOutput *)predictionFromFeatures:(whisper_encoder_implInput *)input options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    id<MLFeatureProvider> outFeatures = [self.model predictionFromFeatures:input options:options error:error];
-    if (!outFeatures) { return nil; }
-    return [[whisper_encoder_implOutput alloc] initWithOutput:(MLMultiArray *)[outFeatures featureValueForName:@"output"].multiArrayValue];
-}
-
-- (nullable whisper_encoder_implOutput *)predictionFromLogmel_data:(MLMultiArray *)logmel_data error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    whisper_encoder_implInput *input_ = [[whisper_encoder_implInput alloc] initWithLogmel_data:logmel_data];
-    return [self predictionFromFeatures:input_ error:error];
-}
-
-- (nullable NSArray<whisper_encoder_implOutput *> *)predictionsFromInputs:(NSArray<whisper_encoder_implInput*> *)inputArray options:(MLPredictionOptions *)options error:(NSError * _Nullable __autoreleasing * _Nullable)error {
-    id<MLBatchProvider> inBatch = [[MLArrayBatchProvider alloc] initWithFeatureProviderArray:inputArray];
-    id<MLBatchProvider> outBatch = [self.model predictionsFromBatch:inBatch options:options error:error];
-    if (!outBatch) { return nil; }
-    NSMutableArray<whisper_encoder_implOutput*> *results = [NSMutableArray arrayWithCapacity:(NSUInteger)outBatch.count];
-    for (NSInteger i = 0; i < outBatch.count; i++) {
-        id<MLFeatureProvider> resultProvider = [outBatch featuresAtIndex:i];
-        whisper_encoder_implOutput * result = [[whisper_encoder_implOutput alloc] initWithOutput:(MLMultiArray *)[resultProvider featureValueForName:@"output"].multiArrayValue];
-        [results addObject:result];
-    }
-    return results;
-}
-
-@end

coreml/whisper-encoder.h

@@ -1,22 +0,0 @@
-// Wrapper of the Core ML Whisper Encoder model
-//
-// Code is derived from the work of Github user @wangchou
-// ref: https://github.com/wangchou/callCoreMLFromCpp
-
-#if __cplusplus
-extern "C" {
-#endif
-
-struct whisper_coreml_context;
-
-struct whisper_coreml_context * whisper_coreml_init(const char * path_model);
-void whisper_coreml_free(struct whisper_coreml_context * ctx);
-
-void whisper_coreml_encode(
-        const whisper_coreml_context * ctx,
-                               float * mel,
-                               float * out);
-
-#if __cplusplus
-}
-#endif

coreml/whisper-encoder.mm

@@ -1,65 +0,0 @@
-#if !__has_feature(objc_arc)
-#error This file must be compiled with automatic reference counting enabled (-fobjc-arc)
-#endif
-
-#import "whisper-encoder.h"
-#import "whisper-encoder-impl.h"
-
-#import <CoreML/CoreML.h>
-
-#include <stdlib.h>
-
-#if __cplusplus
-extern "C" {
-#endif
-
-struct whisper_coreml_context {
-    const void * data;
-};
-
-struct whisper_coreml_context * whisper_coreml_init(const char * path_model) {
-    NSString * path_model_str = [[NSString alloc] initWithUTF8String:path_model];
-
-    NSURL * url_model = [NSURL fileURLWithPath: path_model_str];
-
-    const void * data = CFBridgingRetain([[whisper_encoder_impl alloc] initWithContentsOfURL:url_model error:nil]);
-
-    if (data == NULL) {
-        return NULL;
-    }
-
-    whisper_coreml_context * ctx = new whisper_coreml_context;
-
-    ctx->data = data;
-
-    return ctx;
-}
-
-void whisper_coreml_free(struct whisper_coreml_context * ctx) {
-    CFRelease(ctx->data);
-    delete ctx;
-}
-
-void whisper_coreml_encode(
-        const whisper_coreml_context * ctx,
-                               float * mel,
-                               float * out) {
-    MLMultiArray * inMultiArray = [
-        [MLMultiArray alloc] initWithDataPointer: mel
-                                           shape: @[@1, @80, @3000]
-                                        dataType: MLMultiArrayDataTypeFloat32
-                                         strides: @[@(240000), @(3000), @1]
-                                     deallocator: nil
-                                           error: nil
-    ];
-
-    @autoreleasepool {
-        whisper_encoder_implOutput * outCoreML = [(__bridge id) ctx->data predictionFromLogmel_data:inMultiArray error:nil];
-
-        memcpy(out, outCoreML.output.dataPointer, outCoreML.output.count * sizeof(float));
-    }
-}
-
-#if __cplusplus
-}
-#endif

examples/CMakeLists.txt

@@ -4,7 +4,7 @@ find_package(Threads REQUIRED)
 
 # third-party
 
-if (WHISPER_SDL2)
+if (WHISPER_SUPPORT_SDL2)
     # SDL2
     find_package(SDL2 REQUIRED)
 
@@ -14,42 +14,6 @@ if (WHISPER_SDL2)
     message(STATUS "SDL2_LIBRARIES = ${SDL2_LIBRARIES}")
 endif()
 
-# common
-
-set(TARGET common)
-
-add_library(${TARGET} STATIC
-    common.h
-    common.cpp
-    common-ggml.h
-    common-ggml.cpp
-    grammar-parser.cpp
-    )
-
-include(DefaultTargetOptions)
-
-target_link_libraries(${TARGET} PRIVATE whisper)
-
-set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
-
-if (WHISPER_SDL2)
-    # common-sdl
-
-    set(TARGET common-sdl)
-
-    add_library(${TARGET} STATIC
-        common-sdl.h
-        common-sdl.cpp
-        )
-
-    include(DefaultTargetOptions)
-
-    target_include_directories(${TARGET} PUBLIC ${SDL2_INCLUDE_DIRS})
-    target_link_libraries(${TARGET} PRIVATE ${SDL2_LIBRARIES})
-
-    set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
-endif()
-
 # examples
 
 include_directories(${CMAKE_CURRENT_SOURCE_DIR})
@@ -60,15 +24,10 @@ if (EMSCRIPTEN)
     add_subdirectory(command.wasm)
     add_subdirectory(talk.wasm)
     add_subdirectory(bench.wasm)
-elseif(CMAKE_JS_VERSION)
-    add_subdirectory(addon.node)
 else()
     add_subdirectory(main)
     add_subdirectory(stream)
     add_subdirectory(command)
     add_subdirectory(bench)
-    add_subdirectory(quantize)
     add_subdirectory(talk)
-    add_subdirectory(talk-llama)
-    add_subdirectory(lsp)
 endif()

examples/addon.node/.gitignore

@@ -1,3 +0,0 @@
-.idea
-node_modules
-build

examples/addon.node/CMakeLists.txt

@@ -1,31 +0,0 @@
-set(TARGET whisper-addon)
-
-# Base settings
-#==================================================================
-# env var supported by cmake-js
-add_definitions(-DNAPI_VERSION=4)
-include_directories(${CMAKE_JS_INC})
-#==================================================================
-
-add_library(${TARGET} SHARED ${CMAKE_JS_SRC} addon.cpp)
-set_target_properties(${TARGET} PROPERTIES PREFIX "" SUFFIX ".node")
-
-include(DefaultTargetOptions)
-
-# Include N-API wrappers
-#==================================================================
-execute_process(COMMAND node -p "require('node-addon-api').include"
-        WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
-        OUTPUT_VARIABLE NODE_ADDON_API_DIR
-        )
-string(REPLACE "\n" "" NODE_ADDON_API_DIR ${NODE_ADDON_API_DIR})
-string(REPLACE "\"" "" NODE_ADDON_API_DIR ${NODE_ADDON_API_DIR})
-target_include_directories(${TARGET} PRIVATE ${NODE_ADDON_API_DIR})
-#==================================================================
-
-target_link_libraries(${TARGET} ${CMAKE_JS_LIB} common whisper ${CMAKE_THREAD_LIBS_INIT})
-
-if(MSVC AND CMAKE_JS_NODELIB_DEF AND CMAKE_JS_NODELIB_TARGET)
-    # Generate node.lib
-    execute_process(COMMAND ${CMAKE_AR} /def:${CMAKE_JS_NODELIB_DEF} /out:${CMAKE_JS_NODELIB_TARGET} ${CMAKE_STATIC_LINKER_FLAGS})
-endif()

examples/addon.node/README.md

@@ -1,37 +0,0 @@
-# addon
-
-This is an addon demo that can **perform whisper model reasoning in `node` and `electron` environments**, based on [cmake-js](https://github.com/cmake-js/cmake-js).
-It can be used as a reference for using the whisper.cpp project in other node projects.
-
-## Install
-
-```shell
-npm install
-```
-
-## Compile
-
-Make sure it is in the project root directory and compiled with make-js.
-
-```shell
-npx cmake-js compile -T whisper-addon -B Release
-```
-
-For Electron addon and cmake-js options, you can see [cmake-js](https://github.com/cmake-js/cmake-js) and make very few configuration changes.
-
-> Such as appointing special cmake path:
-> ```shell
-> npx cmake-js compile -c 'xxx/cmake' -T whisper-addon -B Release
-> ```
-
-## Run
-
-```shell
-cd examples/addon.node
-
-node index.js --language='language' --model='model-path' --fname_inp='file-path'
-```
-
-Because this is a simple Demo, only the above parameters are set in the node environment.
-
-Other parameters can also be specified in the node environment.

examples/addon.node/__test__/whisper.spec.js

@@ -1,23 +0,0 @@
-const path = require("path");
-const { whisper } = require(path.join(
-  __dirname,
-  "../../../build/Release/whisper-addon"
-));
-const { promisify } = require("util");
-
-const whisperAsync = promisify(whisper);
-
-const whisperParamsMock = {
-  language: "en",
-  model: path.join(__dirname, "../../../models/ggml-base.en.bin"),
-  fname_inp: path.join(__dirname, "../../../samples/jfk.wav"),
-};
-
-describe("Run whisper.node", () => {
-    test("it should receive a non-empty value", async () => {
-        let result = await whisperAsync(whisperParamsMock);
-
-        expect(result.length).toBeGreaterThan(0);
-    }, 10000);
-});
-

examples/addon.node/addon.cpp

@@ -1,338 +0,0 @@
-#include "napi.h"
-#include "common.h"
-
-#include "whisper.h"
-
-#include <string>
-#include <thread>
-#include <vector>
-#include <cmath>
-#include <cstdint>
-
-struct whisper_params {
-    int32_t n_threads    = std::min(4, (int32_t) std::thread::hardware_concurrency());
-    int32_t n_processors = 1;
-    int32_t offset_t_ms  = 0;
-    int32_t offset_n     = 0;
-    int32_t duration_ms  = 0;
-    int32_t max_context  = -1;
-    int32_t max_len      = 0;
-    int32_t best_of      = 5;
-    int32_t beam_size    = -1;
-
-    float word_thold    = 0.01f;
-    float entropy_thold = 2.4f;
-    float logprob_thold = -1.0f;
-
-    bool speed_up       = false;
-    bool translate      = false;
-    bool diarize        = false;
-    bool output_txt     = false;
-    bool output_vtt     = false;
-    bool output_srt     = false;
-    bool output_wts     = false;
-    bool output_csv     = false;
-    bool print_special  = false;
-    bool print_colors   = false;
-    bool print_progress = false;
-    bool no_timestamps  = false;
-
-    std::string language = "en";
-    std::string prompt;
-    std::string model    = "../../ggml-large.bin";
-
-    std::vector<std::string> fname_inp = {};
-    std::vector<std::string> fname_out = {};
-};
-
-struct whisper_print_user_data {
-    const whisper_params * params;
-
-    const std::vector<std::vector<float>> * pcmf32s;
-};
-
-//  500 -> 00:05.000
-// 6000 -> 01:00.000
-std::string to_timestamp(int64_t t, bool comma = false) {
-    int64_t msec = t * 10;
-    int64_t hr = msec / (1000 * 60 * 60);
-    msec = msec - hr * (1000 * 60 * 60);
-    int64_t min = msec / (1000 * 60);
-    msec = msec - min * (1000 * 60);
-    int64_t sec = msec / 1000;
-    msec = msec - sec * 1000;
-
-    char buf[32];
-    snprintf(buf, sizeof(buf), "%02d:%02d:%02d%s%03d", (int) hr, (int) min, (int) sec, comma ? "," : ".", (int) msec);
-
-    return std::string(buf);
-}
-
-int timestamp_to_sample(int64_t t, int n_samples) {
-    return std::max(0, std::min((int) n_samples - 1, (int) ((t*WHISPER_SAMPLE_RATE)/100)));
-}
-
-void whisper_print_segment_callback(struct whisper_context * ctx, struct whisper_state * state, int n_new, void * user_data) {
-    const auto & params  = *((whisper_print_user_data *) user_data)->params;
-    const auto & pcmf32s = *((whisper_print_user_data *) user_data)->pcmf32s;
-
-    const int n_segments = whisper_full_n_segments(ctx);
-
-    std::string speaker = "";
-
-    int64_t t0;
-    int64_t t1;
-
-    // print the last n_new segments
-    const int s0 = n_segments - n_new;
-
-    if (s0 == 0) {
-        printf("\n");
-    }
-
-    for (int i = s0; i < n_segments; i++) {
-        if (!params.no_timestamps || params.diarize) {
-            t0 = whisper_full_get_segment_t0(ctx, i);
-            t1 = whisper_full_get_segment_t1(ctx, i);
-        }
-
-        if (!params.no_timestamps) {
-            printf("[%s --> %s]  ", to_timestamp(t0).c_str(), to_timestamp(t1).c_str());
-        }
-
-        if (params.diarize && pcmf32s.size() == 2) {
-            const int64_t n_samples = pcmf32s[0].size();
-
-            const int64_t is0 = timestamp_to_sample(t0, n_samples);
-            const int64_t is1 = timestamp_to_sample(t1, n_samples);
-
-            double energy0 = 0.0f;
-            double energy1 = 0.0f;
-
-            for (int64_t j = is0; j < is1; j++) {
-                energy0 += fabs(pcmf32s[0][j]);
-                energy1 += fabs(pcmf32s[1][j]);
-            }
-
-            if (energy0 > 1.1*energy1) {
-                speaker = "(speaker 0)";
-            } else if (energy1 > 1.1*energy0) {
-                speaker = "(speaker 1)";
-            } else {
-                speaker = "(speaker ?)";
-            }
-
-            //printf("is0 = %lld, is1 = %lld, energy0 = %f, energy1 = %f, %s\n", is0, is1, energy0, energy1, speaker.c_str());
-        }
-
-        // colorful print bug
-        //
-        const char * text = whisper_full_get_segment_text(ctx, i);
-        printf("%s%s", speaker.c_str(), text);
-
-
-        // with timestamps or speakers: each segment on new line
-        if (!params.no_timestamps || params.diarize) {
-            printf("\n");
-        }
-
-        fflush(stdout);
-    }
-}
-
-int run(whisper_params &params, std::vector<std::vector<std::string>> &result) {
-    if (params.fname_inp.empty()) {
-        fprintf(stderr, "error: no input files specified\n");
-        return 2;
-    }
-
-    if (params.language != "auto" && whisper_lang_id(params.language.c_str()) == -1) {
-        fprintf(stderr, "error: unknown language '%s'\n", params.language.c_str());
-        exit(0);
-    }
-
-    // whisper init
-
-    struct whisper_context * ctx = whisper_init_from_file(params.model.c_str());
-
-    if (ctx == nullptr) {
-        fprintf(stderr, "error: failed to initialize whisper context\n");
-        return 3;
-    }
-
-    for (int f = 0; f < (int) params.fname_inp.size(); ++f) {
-        const auto fname_inp = params.fname_inp[f];
-        const auto fname_out = f < (int)params.fname_out.size() && !params.fname_out[f].empty() ? params.fname_out[f] : params.fname_inp[f];
-
-        std::vector<float> pcmf32; // mono-channel F32 PCM
-        std::vector<std::vector<float>> pcmf32s; // stereo-channel F32 PCM
-
-        if (!::read_wav(fname_inp, pcmf32, pcmf32s, params.diarize)) {
-            fprintf(stderr, "error: failed to read WAV file '%s'\n", fname_inp.c_str());
-            continue;
-        }
-
-        // print system information
-        {
-            fprintf(stderr, "\n");
-            fprintf(stderr, "system_info: n_threads = %d / %d | %s\n",
-                    params.n_threads*params.n_processors, std::thread::hardware_concurrency(), whisper_print_system_info());
-        }
-
-        // print some info about the processing
-        {
-            fprintf(stderr, "\n");
-            if (!whisper_is_multilingual(ctx)) {
-                if (params.language != "en" || params.translate) {
-                    params.language = "en";
-                    params.translate = false;
-                    fprintf(stderr, "%s: WARNING: model is not multilingual, ignoring language and translation options\n", __func__);
-                }
-            }
-            fprintf(stderr, "%s: processing '%s' (%d samples, %.1f sec), %d threads, %d processors, lang = %s, task = %s, timestamps = %d ...\n",
-                    __func__, fname_inp.c_str(), int(pcmf32.size()), float(pcmf32.size())/WHISPER_SAMPLE_RATE,
-                    params.n_threads, params.n_processors,
-                    params.language.c_str(),
-                    params.translate ? "translate" : "transcribe",
-                    params.no_timestamps ? 0 : 1);
-
-            fprintf(stderr, "\n");
-        }
-
-        // run the inference
-        {
-            whisper_full_params wparams = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
-
-            wparams.strategy = params.beam_size > 1 ? WHISPER_SAMPLING_BEAM_SEARCH : WHISPER_SAMPLING_GREEDY;
-
-            wparams.print_realtime   = false;
-            wparams.print_progress   = params.print_progress;
-            wparams.print_timestamps = !params.no_timestamps;
-            wparams.print_special    = params.print_special;
-            wparams.translate        = params.translate;
-            wparams.language         = params.language.c_str();
-            wparams.n_threads        = params.n_threads;
-            wparams.n_max_text_ctx   = params.max_context >= 0 ? params.max_context : wparams.n_max_text_ctx;
-            wparams.offset_ms        = params.offset_t_ms;
-            wparams.duration_ms      = params.duration_ms;
-
-            wparams.token_timestamps = params.output_wts || params.max_len > 0;
-            wparams.thold_pt         = params.word_thold;
-            wparams.entropy_thold    = params.entropy_thold;
-            wparams.logprob_thold    = params.logprob_thold;
-            wparams.max_len          = params.output_wts && params.max_len == 0 ? 60 : params.max_len;
-
-            wparams.speed_up         = params.speed_up;
-
-            wparams.greedy.best_of        = params.best_of;
-            wparams.beam_search.beam_size = params.beam_size;
-
-            wparams.initial_prompt   = params.prompt.c_str();
-
-            whisper_print_user_data user_data = { &params, &pcmf32s };
-
-            // this callback is called on each new segment
-            if (!wparams.print_realtime) {
-                wparams.new_segment_callback           = whisper_print_segment_callback;
-                wparams.new_segment_callback_user_data = &user_data;
-            }
-
-            // example for abort mechanism
-            // in this example, we do not abort the processing, but we could if the flag is set to true
-            // the callback is called before every encoder run - if it returns false, the processing is aborted
-            {
-                static bool is_aborted = false; // NOTE: this should be atomic to avoid data race
-
-                wparams.encoder_begin_callback = [](struct whisper_context * /*ctx*/, struct whisper_state * /*state*/, void * user_data) {
-                    bool is_aborted = *(bool*)user_data;
-                    return !is_aborted;
-                };
-                wparams.encoder_begin_callback_user_data = &is_aborted;
-            }
-
-            if (whisper_full_parallel(ctx, wparams, pcmf32.data(), pcmf32.size(), params.n_processors) != 0) {
-                fprintf(stderr, "failed to process audio\n");
-                return 10;
-            }
-        }
-    }
-
-    const int n_segments = whisper_full_n_segments(ctx);
-    result.resize(n_segments);
-    for (int i = 0; i < n_segments; ++i) {
-        const char * text = whisper_full_get_segment_text(ctx, i);
-        const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
-        const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
-
-        result[i].emplace_back(to_timestamp(t0, true));
-        result[i].emplace_back(to_timestamp(t1, true));
-        result[i].emplace_back(text);
-    }
-
-    whisper_print_timings(ctx);
-    whisper_free(ctx);
-
-    return 0;
-}
-
-class Worker : public Napi::AsyncWorker {
- public:
-  Worker(Napi::Function& callback, whisper_params params)
-      : Napi::AsyncWorker(callback), params(params) {}
-
-  void Execute() override {
-    run(params, result);
-  }
-
-  void OnOK() override {
-    Napi::HandleScope scope(Env());
-    Napi::Object res = Napi::Array::New(Env(), result.size());
-    for (uint64_t i = 0; i < result.size(); ++i) {
-      Napi::Object tmp = Napi::Array::New(Env(), 3);
-      for (uint64_t j = 0; j < 3; ++j) {
-        tmp[j] = Napi::String::New(Env(), result[i][j]);
-      }
-      res[i] = tmp;
-    }
-    Callback().Call({Env().Null(), res});
-  }
-
- private:
-  whisper_params params;
-  std::vector<std::vector<std::string>> result;
-};
-
-
-
-Napi::Value whisper(const Napi::CallbackInfo& info) {
-  Napi::Env env = info.Env();
-  if (info.Length() <= 0 || !info[0].IsObject()) {
-    Napi::TypeError::New(env, "object expected").ThrowAsJavaScriptException();
-  }
-  whisper_params params;
-
-  Napi::Object whisper_params = info[0].As<Napi::Object>();
-  std::string language = whisper_params.Get("language").As<Napi::String>();
-  std::string model = whisper_params.Get("model").As<Napi::String>();
-  std::string input = whisper_params.Get("fname_inp").As<Napi::String>();
-
-  params.language = language;
-  params.model = model;
-  params.fname_inp.emplace_back(input);
-
-  Napi::Function callback = info[1].As<Napi::Function>();
-  Worker* worker = new Worker(callback, params);
-  worker->Queue();
-  return env.Undefined();
-}
-
-
-Napi::Object Init(Napi::Env env, Napi::Object exports) {
-  exports.Set(
-      Napi::String::New(env, "whisper"),
-      Napi::Function::New(env, whisper)
-  );
-  return exports;
-}
-
-NODE_API_MODULE(whisper, Init);

examples/addon.node/index.js

@@ -1,36 +0,0 @@
-const path = require("path");
-const { whisper } = require(path.join(
-  __dirname,
-  "../../build/Release/whisper-addon"
-));
-const { promisify } = require("util");
-
-const whisperAsync = promisify(whisper);
-
-const whisperParams = {
-  language: "en",
-  model: path.join(__dirname, "../../models/ggml-base.en.bin"),
-  fname_inp: "../../samples/jfk.wav",
-};
-
-const arguments = process.argv.slice(2);
-const params = Object.fromEntries(
-  arguments.reduce((pre, item) => {
-    if (item.startsWith("--")) {
-      return [...pre, item.slice(2).split("=")];
-    }
-    return pre;
-  }, [])
-);
-
-for (const key in params) {
-  if (whisperParams.hasOwnProperty(key)) {
-    whisperParams[key] = params[key];
-  }
-}
-
-console.log("whisperParams =", whisperParams);
-
-whisperAsync(whisperParams).then((result) => {
-  console.log(`Result from whisper: ${result}`);
-});

examples/addon.node/package.json

@@ -1,16 +0,0 @@
-{
-  "name": "whisper-addon",
-  "version": "0.0.0",
-  "description": "",
-  "main": "index.js",
-  "author": "Qanhe Chen",
-  "license": "MIT",
-  "scripts": {
-    "test": "jest"
-  },
-  "devDependencies": {
-    "cmake-js": "^7.1.1",
-    "jest": "^29.4.0",
-    "node-addon-api": "^5.0.0"
-  }
-}

examples/bench.wasm/CMakeLists.txt

@@ -8,8 +8,6 @@ add_executable(${TARGET}
     emscripten.cpp
     )
 
-include(DefaultTargetOptions)
-
 target_link_libraries(${TARGET} PRIVATE
     whisper
     )
@@ -31,9 +29,9 @@ endif()
 set_target_properties(${TARGET} PROPERTIES LINK_FLAGS " \
     --bind \
     -s USE_PTHREADS=1 \
-    -s PTHREAD_POOL_SIZE_STRICT=0 \
-    -s INITIAL_MEMORY=2000MB \
-    -s TOTAL_MEMORY=2000MB \
+    -s PTHREAD_POOL_SIZE=8 \
+    -s INITIAL_MEMORY=1024MB \
+    -s TOTAL_MEMORY=1024MB \
     -s FORCE_FILESYSTEM=1 \
     -s EXPORTED_RUNTIME_METHODS=\"['print', 'printErr', 'ccall', 'cwrap']\" \
     ${EXTRA_FLAGS} \

examples/bench.wasm/emscripten.cpp

@@ -28,11 +28,6 @@ void bench_main(size_t index) {
         return;
     }
 
-    {
-        fprintf(stderr, "\n");
-        fprintf(stderr, "system_info: n_threads = %d / %d | %s\n", n_threads, std::thread::hardware_concurrency(), whisper_print_system_info());
-    }
-
     if (int ret = whisper_encode(ctx, 0, n_threads) != 0) {
         fprintf(stderr, "error: failed to encode model: %d\n", ret);
         return;
@@ -57,7 +52,7 @@ EMSCRIPTEN_BINDINGS(bench) {
     emscripten::function("init", emscripten::optional_override([](const std::string & path_model) {
         for (size_t i = 0; i < g_contexts.size(); ++i) {
             if (g_contexts[i] == nullptr) {
-                g_contexts[i] = whisper_init_from_file(path_model.c_str());
+                g_contexts[i] = whisper_init(path_model.c_str());
                 if (g_contexts[i] != nullptr) {
                     if (g_worker.joinable()) {
                         g_worker.join();

examples/bench.wasm/index-tmpl.html

@@ -35,15 +35,6 @@
 
             <br><br>
 
-            <b>More examples:</b>
-                <a href="https://whisper.ggerganov.com/">main</a> |
-                <a href="https://whisper.ggerganov.com/bench">bench</a> |
-                <a href="https://whisper.ggerganov.com/stream">stream</a> |
-                <a href="https://whisper.ggerganov.com/command">command</a> |
-                <a href="https://whisper.ggerganov.com/talk">talk</a> |
-
-            <br><br>
-
             <hr>
 
             Select the model you would like to use and click the "Bench" button.<br>
@@ -55,16 +46,9 @@
                 Whisper model: <span id="model-whisper-status"></span>
                 <button id="fetch-whisper-tiny-en" onclick="loadWhisper('tiny.en')">tiny.en (75 MB)</button>
                 <button id="fetch-whisper-base-en" onclick="loadWhisper('base.en')">base.en (142 MB)</button>
-                <button id="fetch-whisper-small-en" onclick="loadWhisper('small.en')">small.en (466 MB)</button>
-                <input type="file" id="whisper-file" name="file" onchange="loadFile(event, 'whisper.bin')" />
-                <br><br>
-                Quantized models:<br><br>
-                <button id="fetch-whisper-tiny-en-q5_1"   onclick="loadWhisper('tiny-en-q5_1')">tiny.en (Q5_1, 31 MB)</button>
-                <button id="fetch-whisper-base-en-q5_1"   onclick="loadWhisper('base-en-q5_1')">base.en (Q5_1, 57 MB)</button>
-                <button id="fetch-whisper-small-en-q5_1"  onclick="loadWhisper('small-en-q5_1')">small.en (Q5_1, 182 MB)</button>
-                <button id="fetch-whisper-medium-en-q5_0" onclick="loadWhisper('medium-en-q5_0')">medium.en (Q5_0, 515 MB)</button>
-                <button id="fetch-whisper-large-q5_0"     onclick="loadWhisper('large-q5_0')">large (Q5_0, 1030 MB)</button>
                 <span id="fetch-whisper-progress"></span>
+
+                <input type="file" id="whisper-file" name="file" onchange="loadFile(event, 'whisper.bin')" />
             </div>
 
             <br>
@@ -176,14 +160,6 @@
 
                 document.getElementById('fetch-whisper-tiny-en').style.display = 'none';
                 document.getElementById('fetch-whisper-base-en').style.display = 'none';
-                document.getElementById('fetch-whisper-small-en').style.display = 'none';
-
-                document.getElementById('fetch-whisper-tiny-en-q5_1'  ).style.display = 'none';
-                document.getElementById('fetch-whisper-base-en-q5_1'  ).style.display = 'none';
-                document.getElementById('fetch-whisper-small-en-q5_1' ).style.display = 'none';
-                document.getElementById('fetch-whisper-medium-en-q5_0').style.display = 'none';
-                document.getElementById('fetch-whisper-large-q5_0'    ).style.display = 'none';
-
                 document.getElementById('whisper-file'         ).style.display = 'none';
                 document.getElementById('model-whisper-status' ).innerHTML = 'loaded model: ' + file.name;
             }
@@ -192,25 +168,11 @@
                 let urls = {
                     'tiny.en': 'https://whisper.ggerganov.com/ggml-model-whisper-tiny.en.bin',
                     'base.en': 'https://whisper.ggerganov.com/ggml-model-whisper-base.en.bin',
-                    'small.en': 'https://whisper.ggerganov.com/ggml-model-whisper-small.en.bin',
-
-                    'tiny-en-q5_1':  'https://whisper.ggerganov.com/ggml-model-whisper-tiny.en-q5_1.bin',
-                    'base-en-q5_1':  'https://whisper.ggerganov.com/ggml-model-whisper-base.en-q5_1.bin',
-                    'small-en-q5_1': 'https://whisper.ggerganov.com/ggml-model-whisper-small.en-q5_1.bin',
-                    'medium-en-q5_0':'https://whisper.ggerganov.com/ggml-model-whisper-medium.en-q5_0.bin',
-                    'large-q5_0':    'https://whisper.ggerganov.com/ggml-model-whisper-large-q5_0.bin',
                 };
 
                 let sizes = {
                     'tiny.en': 75,
                     'base.en': 142,
-                    'small.en': 466,
-
-                    'tiny-en-q5_1':   31,
-                    'base-en-q5_1':   57,
-                    'small-en-q5_1':  182,
-                    'medium-en-q5_0': 515,
-                    'large-q5_0':     1030,
                 };
 
                 let url     = urls[model];
@@ -219,15 +181,6 @@
 
                 document.getElementById('fetch-whisper-tiny-en').style.display = 'none';
                 document.getElementById('fetch-whisper-base-en').style.display = 'none';
-                document.getElementById('fetch-whisper-small-en').style.display = 'none';
-
-                document.getElementById('fetch-whisper-tiny-en-q5_1'  ).style.display = 'none';
-                document.getElementById('fetch-whisper-base-en-q5_1'  ).style.display = 'none';
-                document.getElementById('fetch-whisper-small-en-q5_1' ).style.display = 'none';
-                document.getElementById('fetch-whisper-medium-en-q5_0').style.display = 'none';
-                document.getElementById('fetch-whisper-large-q5_0'    ).style.display = 'none';
-
-                document.getElementById('whisper-file'        ).style.display = 'none';
                 document.getElementById('model-whisper-status').innerHTML = 'loading "' + model + '" ... ';
 
                 cbProgress = function(p) {
@@ -239,15 +192,6 @@
                     var el;
                     el = document.getElementById('fetch-whisper-tiny-en'); if (el) el.style.display = 'inline-block';
                     el = document.getElementById('fetch-whisper-base-en'); if (el) el.style.display = 'inline-block';
-                    el = document.getElementById('fetch-whisper-small-en'); if (el) el.style.display = 'inline-block';
-
-                    el = document.getElementById('fetch-whisper-tiny-en-q5_1'  ); if (el) el.style.display = 'inline-block';
-                    el = document.getElementById('fetch-whisper-base-en-q5_1'  ); if (el) el.style.display = 'inline-block';
-                    el = document.getElementById('fetch-whisper-small-en-q5_1' ); if (el) el.style.display = 'inline-block';
-                    el = document.getElementById('fetch-whisper-medium-en-q5_0'); if (el) el.style.display = 'inline-block';
-                    el = document.getElementById('fetch-whisper-large-q5_0'    ); if (el) el.style.display = 'inline-block';
-
-                    el = document.getElementById('whisper-file'        ); if (el) el.style.display = 'inline-block';
                     el = document.getElementById('model-whisper-status');  if (el) el.innerHTML = '';
                 };
 

examples/bench/CMakeLists.txt

@@ -1,6 +1,3 @@
 set(TARGET bench)
 add_executable(${TARGET} bench.cpp)
-
-include(DefaultTargetOptions)
-
 target_link_libraries(${TARGET} PRIVATE whisper ${CMAKE_THREAD_LIBS_INIT})

examples/bench/bench.cpp

@@ -7,7 +7,6 @@
 // command-line parameters
 struct whisper_params {
     int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
-    int32_t what = 0; // what to benchmark: 0 - whisper ecoder, 1 - memcpy, 2 - ggml_mul_mat
 
     std::string model = "models/ggml-base.en.bin";
 };
@@ -24,7 +23,6 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
         }
         else if (arg == "-t" || arg == "--threads") { params.n_threads = std::stoi(argv[++i]); }
         else if (arg == "-m" || arg == "--model")   { params.model     = argv[++i]; }
-        else if (arg == "-w" || arg == "--what")    { params.what     = atoi(argv[++i]); }
         else {
             fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
             whisper_print_usage(argc, argv, params);
@@ -43,17 +41,19 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
     fprintf(stderr, "  -h,       --help        [default] show this help message and exit\n");
     fprintf(stderr, "  -t N,     --threads N   [%-7d] number of threads to use during computation\n", params.n_threads);
     fprintf(stderr, "  -m FNAME, --model FNAME [%-7s] model path\n",                                  params.model.c_str());
-    fprintf(stderr, "  -w N,     --what N      [%-7d] what to benchmark:\n",                          params.what);
-    fprintf(stderr, "                           %-7s  0 - whisper encoder\n",                         "");
-    fprintf(stderr, "                           %-7s  1 - memcpy\n",                                  "");
-    fprintf(stderr, "                           %-7s  2 - ggml_mul_mat\n",                            "");
     fprintf(stderr, "\n");
 }
 
-int whisper_bench_encoder(const whisper_params & params) {
+int main(int argc, char ** argv) {
+    whisper_params params;
+
+    if (whisper_params_parse(argc, argv, params) == false) {
+        return 1;
+    }
+
     // whisper init
 
-    struct whisper_context * ctx = whisper_init_from_file(params.model.c_str());
+    struct whisper_context * ctx = whisper_init(params.model.c_str());
 
     {
         fprintf(stderr, "\n");
@@ -92,22 +92,3 @@ int whisper_bench_encoder(const whisper_params & params) {
 
     return 0;
 }
-
-int main(int argc, char ** argv) {
-    whisper_params params;
-
-    if (whisper_params_parse(argc, argv, params) == false) {
-        return 1;
-    }
-
-    int ret = -1;
-
-    switch (params.what) {
-        case 0: ret = whisper_bench_encoder(params);                break;
-        case 1: ret = whisper_bench_memcpy(params.n_threads);       break;
-        case 2: ret = whisper_bench_ggml_mul_mat(params.n_threads); break;
-        default: fprintf(stderr, "error: unknown benchmark: %d\n", params.what); break;
-    }
-
-    return ret;
-}

examples/command.wasm/CMakeLists.txt

@@ -8,10 +8,7 @@ add_executable(${TARGET}
     emscripten.cpp
     )
 
-include(DefaultTargetOptions)
-
 target_link_libraries(${TARGET} PRIVATE
-    common
     whisper
     )
 

examples/command.wasm/emscripten.cpp

@@ -1,5 +1,4 @@
 #include "ggml.h"
-#include "common.h"
 #include "whisper.h"
 
 #include <emscripten.h>
@@ -28,11 +27,92 @@ std::string g_transcribed   = "";
 
 std::vector<float> g_pcmf32;
 
+static std::string trim(const std::string & s) {
+    std::regex e("^\\s+|\\s+$");
+    return std::regex_replace(s, e, "");
+}
+
+static void high_pass_filter(std::vector<float> & data, float cutoff, float sample_rate) {
+    const float rc = 1.0f / (2.0f * M_PI * cutoff);
+    const float dt = 1.0f / sample_rate;
+    const float alpha = dt / (rc + dt);
+
+    float y = data[0];
+
+    for (size_t i = 1; i < data.size(); i++) {
+        y = alpha * (y + data[i] - data[i - 1]);
+        data[i] = y;
+    }
+}
+
+// compute similarity between two strings using Levenshtein distance
+static float similarity(const std::string & s0, const std::string & s1) {
+    const size_t len0 = s0.size() + 1;
+    const size_t len1 = s1.size() + 1;
+
+    std::vector<int> col(len1, 0);
+    std::vector<int> prevCol(len1, 0);
+
+    for (size_t i = 0; i < len1; i++) {
+        prevCol[i] = i;
+    }
+
+    for (size_t i = 0; i < len0; i++) {
+        col[0] = i;
+        for (size_t j = 1; j < len1; j++) {
+            col[j] = std::min(std::min(1 + col[j - 1], 1 + prevCol[j]), prevCol[j - 1] + (s0[i - 1] == s1[j - 1] ? 0 : 1));
+        }
+        col.swap(prevCol);
+    }
+
+    const float dist = prevCol[len1 - 1];
+
+    return 1.0f - (dist / std::max(s0.size(), s1.size()));
+}
+
 void command_set_status(const std::string & status) {
     std::lock_guard<std::mutex> lock(g_mutex);
     g_status = status;
 }
 
+bool command_vad_simple(std::vector<float> & pcmf32, int sample_rate, int last_ms, float vad_thold, float freq_thold, bool verbose) {
+    const int n_samples      = pcmf32.size();
+    const int n_samples_last = (sample_rate * last_ms) / 1000;
+
+    if (n_samples_last >= n_samples) {
+        // not enough samples - assume no speech
+        return false;
+    }
+
+    if (freq_thold > 0.0f) {
+        high_pass_filter(pcmf32, freq_thold, sample_rate);
+    }
+
+    float energy_all  = 0.0f;
+    float energy_last = 0.0f;
+
+    for (size_t i = 0; i < n_samples; i++) {
+        energy_all += fabsf(pcmf32[i]);
+
+        if (i >= n_samples - n_samples_last) {
+            energy_last += fabsf(pcmf32[i]);
+        }
+    }
+
+    energy_all  /= n_samples;
+    energy_last /= n_samples_last;
+
+    if (verbose) {
+        fprintf(stderr, "%s: energy_all: %f, energy_last: %f, vad_thold: %f, freq_thold: %f\n", __func__, energy_all, energy_last, vad_thold, freq_thold);
+    }
+
+    if (energy_last > vad_thold*energy_all) {
+        return false;
+    }
+
+    return true;
+}
+
 std::string command_transcribe(whisper_context * ctx, const whisper_full_params & wparams, const std::vector<float> & pcmf32, float & prob, int64_t & t_ms) {
     const auto t_start = std::chrono::high_resolution_clock::now();
 
@@ -75,7 +155,7 @@ void command_get_audio(int ms, int sample_rate, std::vector<float> & audio) {
     const int64_t n_samples = (ms * sample_rate) / 1000;
 
     int64_t n_take = 0;
-    if (n_samples > (int) g_pcmf32.size()) {
+    if (g_pcmf32.size() < n_samples) {
         n_take = g_pcmf32.size();
     } else {
         n_take = n_samples;
@@ -107,6 +187,7 @@ void command_main(size_t index) {
 
     printf("command: using %d threads\n", wparams.n_threads);
 
+    bool is_running   = true;
     bool have_prompt  = false;
     bool ask_prompt   = true;
     bool print_energy = false;
@@ -152,7 +233,7 @@ void command_main(size_t index) {
         {
             command_get_audio(vad_ms, WHISPER_SAMPLE_RATE, pcmf32_cur);
 
-            if (::vad_simple(pcmf32_cur, WHISPER_SAMPLE_RATE, 1000, vad_thold, freq_thold, print_energy)) {
+            if (command_vad_simple(pcmf32_cur, WHISPER_SAMPLE_RATE, 1000, vad_thold, freq_thold, print_energy)) {
                 fprintf(stdout, "%s: Speech detected! Processing ...\n", __func__);
                 command_set_status("Speech detected! Processing ...");
 
@@ -243,7 +324,7 @@ EMSCRIPTEN_BINDINGS(command) {
     emscripten::function("init", emscripten::optional_override([](const std::string & path_model) {
         for (size_t i = 0; i < g_contexts.size(); ++i) {
             if (g_contexts[i] == nullptr) {
-                g_contexts[i] = whisper_init_from_file(path_model.c_str());
+                g_contexts[i] = whisper_init(path_model.c_str());
                 if (g_contexts[i] != nullptr) {
                     g_running = true;
                     if (g_worker.joinable()) {

examples/command.wasm/index-tmpl.html

@@ -35,15 +35,6 @@
 
             <br><br>
 
-            <b>More examples:</b>
-                <a href="https://whisper.ggerganov.com/">main</a> |
-                <a href="https://whisper.ggerganov.com/bench">bench</a> |
-                <a href="https://whisper.ggerganov.com/stream">stream</a> |
-                <a href="https://whisper.ggerganov.com/command">command</a> |
-                <a href="https://whisper.ggerganov.com/talk">talk</a> |
-
-            <br><br>
-
             <hr>
 
             Select the model you would like to use, click the "Start" button and follow the instructions.
@@ -54,10 +45,6 @@
                 Whisper model: <span id="model-whisper-status"></span>
                 <button id="fetch-whisper-tiny-en" onclick="loadWhisper('tiny.en')">tiny.en (75 MB)</button>
                 <button id="fetch-whisper-base-en" onclick="loadWhisper('base.en')">base.en (142 MB)</button>
-                <br><br>
-                Quantized models:<br><br>
-                <button id="fetch-whisper-tiny-en-q5_1"   onclick="loadWhisper('tiny-en-q5_1')">tiny.en (Q5_1, 31 MB)</button>
-                <button id="fetch-whisper-base-en-q5_1"   onclick="loadWhisper('base-en-q5_1')">base.en (Q5_1, 57 MB)</button>
                 <span id="fetch-whisper-progress"></span>
 
                 <!--
@@ -175,17 +162,11 @@
                 let urls = {
                     'tiny.en': 'https://whisper.ggerganov.com/ggml-model-whisper-tiny.en.bin',
                     'base.en': 'https://whisper.ggerganov.com/ggml-model-whisper-base.en.bin',
-
-                    'tiny-en-q5_1':  'https://whisper.ggerganov.com/ggml-model-whisper-tiny.en-q5_1.bin',
-                    'base-en-q5_1':  'https://whisper.ggerganov.com/ggml-model-whisper-base.en-q5_1.bin',
                 };
 
                 let sizes = {
                     'tiny.en': 75,
                     'base.en': 142,
-
-                    'tiny-en-q5_1':   31,
-                    'base-en-q5_1':   57,
                 };
 
                 let url     = urls[model];
@@ -196,10 +177,6 @@
 
                 document.getElementById('fetch-whisper-tiny-en').style.display = 'none';
                 document.getElementById('fetch-whisper-base-en').style.display = 'none';
-
-                document.getElementById('fetch-whisper-tiny-en-q5_1').style.display = 'none';
-                document.getElementById('fetch-whisper-base-en-q5_1').style.display = 'none';
-
                 document.getElementById('model-whisper-status').innerHTML = 'loading "' + model + '" ... ';
 
                 cbProgress = function(p) {
@@ -211,10 +188,6 @@
                     var el;
                     el = document.getElementById('fetch-whisper-tiny-en'); if (el) el.style.display = 'inline-block';
                     el = document.getElementById('fetch-whisper-base-en'); if (el) el.style.display = 'inline-block';
-
-                    el = document.getElementById('fetch-whisper-tiny-en-q5_1'); if (el) el.style.display = 'inline-block';
-                    el = document.getElementById('fetch-whisper-base-en-q5_1'); if (el) el.style.display = 'inline-block';
-
                     el = document.getElementById('model-whisper-status');  if (el) el.innerHTML = '';
                 };
 

examples/command/CMakeLists.txt

@@ -1,9 +1,7 @@
-if (WHISPER_SDL2)
+if (WHISPER_SUPPORT_SDL2)
     # command
     set(TARGET command)
     add_executable(${TARGET} command.cpp)
-
-    include(DefaultTargetOptions)
-
-    target_link_libraries(${TARGET} PRIVATE common common-sdl whisper ${CMAKE_THREAD_LIBS_INIT})
+    target_include_directories(${TARGET} PRIVATE ${SDL2_INCLUDE_DIRS})
+    target_link_libraries(${TARGET} PRIVATE whisper ${SDL2_LIBRARIES} ${CMAKE_THREAD_LIBS_INIT})
 endif ()

examples/command/README.md

@@ -9,19 +9,7 @@ More info is available in [issue #171](https://github.com/ggerganov/whisper.cpp/
 
 # On Raspberry Pi, use tiny or base models + "-ac 768" for better performance
 ./command -m ./models/ggml-tiny.en.bin -ac 768 -t 3 -c 0
-```
 
-https://user-images.githubusercontent.com/1991296/204038393-2f846eae-c255-4099-a76d-5735c25c49da.mp4
-
-Web version: [examples/command.wasm](/examples/command.wasm)
-
-## Guided mode
-
-"Guided mode" allows you to specify a list of commands (i.e. strings) and the transcription will be guided to classify your command into one from the list. This can be useful in situations where a device is listening only for a small subset of commands.
-
-Initial tests show that this approach might be extremely efficient in terms of performance, since it integrates very well with the "partial Encoder" idea from #137.
-
-```bash
 # Run in guided mode, the list of allowed commands is in commands.txt
 ./command -m ./models/ggml-base.en.bin -cmd ./examples/command/commands.txt
 
@@ -29,8 +17,9 @@ Initial tests show that this approach might be extremely efficient in terms of p
 ./command -m ./models/ggml-tiny.en.bin -cmd ./examples/command/commands.txt -ac 128 -t 3 -c 0
 ```
 
-https://user-images.githubusercontent.com/1991296/207435352-8fc4ed3f-bde5-4555-9b8b-aeeb76bee969.mp4
+https://user-images.githubusercontent.com/1991296/204038393-2f846eae-c255-4099-a76d-5735c25c49da.mp4
 
+Web version: [examples/command.wasm](/examples/command.wasm)
 
 ## Building
 

examples/common-ggml.cpp

@@ -1,246 +0,0 @@
-#include "common-ggml.h"
-
-#include <regex>
-#include <map>
-
-static const std::map<std::string, enum ggml_ftype> GGML_FTYPE_MAP = {
-    {"q4_0", GGML_FTYPE_MOSTLY_Q4_0},
-    {"q4_1", GGML_FTYPE_MOSTLY_Q4_1},
-    {"q5_0", GGML_FTYPE_MOSTLY_Q5_0},
-    {"q5_1", GGML_FTYPE_MOSTLY_Q5_1},
-    {"q8_0", GGML_FTYPE_MOSTLY_Q8_0},
-};
-
-void ggml_print_ftypes(FILE * fp) {
-    for (auto it = GGML_FTYPE_MAP.begin(); it != GGML_FTYPE_MAP.end(); it++) {
-        fprintf(fp, "  type = \"%s\" or %d\n", it->first.c_str(), it->second);
-    }
-}
-
-enum ggml_ftype ggml_parse_ftype(const char * str) {
-    enum ggml_ftype ftype;
-    if (str[0] == 'q') {
-        const auto it = GGML_FTYPE_MAP.find(str);
-        if (it == GGML_FTYPE_MAP.end()) {
-            fprintf(stderr, "%s: unknown ftype '%s'\n", __func__, str);
-            return GGML_FTYPE_UNKNOWN;
-        }
-        ftype = it->second;
-    } else {
-        ftype = (enum ggml_ftype) atoi(str);
-    }
-
-    return ftype;
-}
-
-bool ggml_common_quantize_0(
-        std::ifstream & finp,
-        std::ofstream & fout,
-        const ggml_ftype ftype,
-        const std::vector<std::string> & to_quant,
-        const std::vector<std::string> & to_skip) {
-
-    ggml_type qtype = GGML_TYPE_F32;
-
-    switch (ftype) {
-        case GGML_FTYPE_MOSTLY_Q4_0: qtype = GGML_TYPE_Q4_0; break;
-        case GGML_FTYPE_MOSTLY_Q4_1: qtype = GGML_TYPE_Q4_1; break;
-        case GGML_FTYPE_MOSTLY_Q5_0: qtype = GGML_TYPE_Q5_0; break;
-        case GGML_FTYPE_MOSTLY_Q5_1: qtype = GGML_TYPE_Q5_1; break;
-        case GGML_FTYPE_MOSTLY_Q8_0: qtype = GGML_TYPE_Q8_0; break;
-        case GGML_FTYPE_UNKNOWN:
-        case GGML_FTYPE_ALL_F32:
-        case GGML_FTYPE_MOSTLY_F16:
-        case GGML_FTYPE_MOSTLY_Q4_1_SOME_F16:
-        case GGML_FTYPE_MOSTLY_Q2_K:
-        case GGML_FTYPE_MOSTLY_Q3_K:
-        case GGML_FTYPE_MOSTLY_Q4_K:
-        case GGML_FTYPE_MOSTLY_Q5_K:
-        case GGML_FTYPE_MOSTLY_Q6_K:
-                {
-                    fprintf(stderr, "%s: invalid model type %d\n", __func__, ftype);
-                    return false;
-                }
-    };
-
-    if (!ggml_is_quantized(qtype)) {
-        fprintf(stderr, "%s: invalid quantization type %d (%s)\n", __func__, qtype, ggml_type_name(qtype));
-        return false;
-    }
-
-    size_t total_size_org = 0;
-    size_t total_size_new = 0;
-
-    std::vector<float> work;
-
-    std::vector<uint8_t>     data_u8;
-    std::vector<ggml_fp16_t> data_f16;
-    std::vector<float>       data_f32;
-
-    std::vector<int64_t> hist_all(1 << 4, 0);
-
-    while (true) {
-        int32_t n_dims;
-        int32_t length;
-        int32_t ttype;
-
-        finp.read(reinterpret_cast<char *>(&n_dims), sizeof(n_dims));
-        finp.read(reinterpret_cast<char *>(&length), sizeof(length));
-        finp.read(reinterpret_cast<char *>(&ttype),  sizeof(ttype));
-
-        if (finp.eof()) {
-            break;
-        }
-
-        int32_t nelements = 1;
-        int32_t ne[4] = { 1, 1, 1, 1 };
-        for (int i = 0; i < n_dims; ++i) {
-            finp.read (reinterpret_cast<char *>(&ne[i]), sizeof(ne[i]));
-            nelements *= ne[i];
-        }
-
-        std::string name(length, 0);
-        finp.read (&name[0], length);
-
-        printf("%64s - [%5d, %5d, %5d], type = %6s ", name.data(), ne[0], ne[1], ne[2], ggml_type_name((ggml_type) ttype));
-
-        bool quantize = false;
-
-        // check if we should quantize this tensor
-        for (const auto & s : to_quant) {
-            if (std::regex_match(name, std::regex(s))) {
-                quantize = true;
-                break;
-            }
-        }
-
-        // check if we should skip this tensor
-        for (const auto & s : to_skip) {
-            if (std::regex_match(name, std::regex(s))) {
-                quantize = false;
-                break;
-            }
-        }
-
-        // quantize only 2D tensors
-        quantize &= (n_dims == 2);
-
-        if (quantize) {
-            if (ttype != GGML_TYPE_F32 && ttype != GGML_TYPE_F16) {
-                fprintf(stderr, "%s: unsupported ttype %d (%s) for integer quantization\n", __func__, ttype, ggml_type_name((ggml_type) ttype));
-                return false;
-            }
-
-            if (ttype == GGML_TYPE_F16) {
-                data_f16.resize(nelements);
-                finp.read(reinterpret_cast<char *>(data_f16.data()), nelements * sizeof(ggml_fp16_t));
-                data_f32.resize(nelements);
-                for (int i = 0; i < nelements; ++i) {
-                    data_f32[i] = ggml_fp16_to_fp32(data_f16[i]);
-                }
-            } else {
-                data_f32.resize(nelements);
-                finp.read(reinterpret_cast<char *>(data_f32.data()), nelements * sizeof(float));
-            }
-
-            ttype = qtype;
-        } else {
-            const int bpe = (ttype == 0) ? sizeof(float) : sizeof(uint16_t);
-
-            data_u8.resize(nelements*bpe);
-            finp.read(reinterpret_cast<char *>(data_u8.data()), nelements * bpe);
-        }
-
-        fout.write(reinterpret_cast<char *>(&n_dims), sizeof(n_dims));
-        fout.write(reinterpret_cast<char *>(&length), sizeof(length));
-        fout.write(reinterpret_cast<char *>(&ttype),  sizeof(ttype));
-        for (int i = 0; i < n_dims; ++i) {
-            fout.write(reinterpret_cast<char *>(&ne[i]), sizeof(ne[i]));
-        }
-        fout.write(&name[0], length);
-
-        if (quantize) {
-            work.resize(nelements); // for quantization
-
-            size_t cur_size = 0;
-            std::vector<int64_t> hist_cur(1 << 4, 0);
-
-            switch ((ggml_type) ttype) {
-                case GGML_TYPE_Q4_0:
-                    {
-                        cur_size = ggml_quantize_q4_0(data_f32.data(), work.data(), nelements, ne[0], hist_cur.data());
-                    } break;
-                case GGML_TYPE_Q4_1:
-                    {
-                        cur_size = ggml_quantize_q4_1(data_f32.data(), work.data(), nelements, ne[0], hist_cur.data());
-                    } break;
-                case GGML_TYPE_Q5_0:
-                    {
-                        cur_size = ggml_quantize_q5_0(data_f32.data(), work.data(), nelements, ne[0], hist_cur.data());
-                    } break;
-                case GGML_TYPE_Q5_1:
-                    {
-                        cur_size = ggml_quantize_q5_1(data_f32.data(), work.data(), nelements, ne[0], hist_cur.data());
-                    } break;
-                case GGML_TYPE_Q8_0:
-                    {
-                        cur_size = ggml_quantize_q8_0(data_f32.data(), work.data(), nelements, ne[0], hist_cur.data());
-                    } break;
-                case GGML_TYPE_F32:
-                case GGML_TYPE_F16:
-                case GGML_TYPE_I8:
-                case GGML_TYPE_I16:
-                case GGML_TYPE_I32:
-                case GGML_TYPE_Q8_1:
-                case GGML_TYPE_Q2_K:
-                case GGML_TYPE_Q3_K:
-                case GGML_TYPE_Q4_K:
-                case GGML_TYPE_Q5_K:
-                case GGML_TYPE_Q6_K:
-                case GGML_TYPE_Q8_K:
-                case GGML_TYPE_COUNT:
-                    {
-                        fprintf(stderr, "%s: unsupported quantization type %d (%s)\n", __func__, ttype, ggml_type_name((ggml_type) ttype));
-                        return false;
-                    }
-            }
-
-            fout.write(reinterpret_cast<char *>(work.data()), cur_size);
-            total_size_new += cur_size;
-
-            printf("size = %8.2f MB -> %8.2f MB | hist: ", nelements * sizeof(float)/1024.0/1024.0, cur_size/1024.0/1024.0);
-            for (int i = 0; i < (int) hist_cur.size(); ++i) {
-                hist_all[i] += hist_cur[i];
-            }
-
-            for (int i = 0; i < (int) hist_cur.size(); ++i) {
-                printf("%5.3f ", hist_cur[i] / (float)nelements);
-            }
-            printf("\n");
-        } else {
-            printf("size = %8.3f MB\n", data_u8.size()/1024.0/1024.0);
-            fout.write(reinterpret_cast<char *>(data_u8.data()), data_u8.size());
-            total_size_new += data_u8.size();
-        }
-
-        total_size_org += nelements * sizeof(float);
-    }
-
-    printf("%s: model size  = %8.2f MB\n", __func__, total_size_org/1024.0/1024.0);
-    printf("%s: quant size  = %8.2f MB | ftype = %d (%s)\n", __func__, total_size_new/1024.0/1024.0, ftype, ggml_type_name(qtype));
-
-    {
-        int64_t sum_all = 0;
-        for (int i = 0; i < (int) hist_all.size(); ++i) {
-            sum_all += hist_all[i];
-        }
-
-        printf("%s: hist: ", __func__);
-        for (int i = 0; i < (int) hist_all.size(); ++i) {
-            printf("%5.3f ", hist_all[i] / (float)sum_all);
-        }
-        printf("\n");
-    }
-
-    return true;
-}

examples/common-ggml.h

@@ -1,18 +0,0 @@
-#pragma once
-
-#include "ggml.h"
-
-#include <fstream>
-#include <vector>
-#include <string>
-
-enum ggml_ftype ggml_parse_ftype(const char * str);
-
-void ggml_print_ftypes(FILE * fp = stderr);
-
-bool ggml_common_quantize_0(
-        std::ifstream & finp,
-        std::ofstream & fout,
-        const ggml_ftype ftype,
-        const std::vector<std::string> & to_quant,
-        const std::vector<std::string> & to_skip);

examples/common-sdl.cpp

@@ -1,226 +0,0 @@
-#include "common-sdl.h"
-
-audio_async::audio_async(int len_ms) {
-    m_len_ms = len_ms;
-
-    m_running = false;
-}
-
-audio_async::~audio_async() {
-    if (m_dev_id_in) {
-        SDL_CloseAudioDevice(m_dev_id_in);
-    }
-}
-
-bool audio_async::init(int capture_id, int sample_rate) {
-    SDL_LogSetPriority(SDL_LOG_CATEGORY_APPLICATION, SDL_LOG_PRIORITY_INFO);
-
-    if (SDL_Init(SDL_INIT_AUDIO) < 0) {
-        SDL_LogError(SDL_LOG_CATEGORY_APPLICATION, "Couldn't initialize SDL: %s\n", SDL_GetError());
-        return false;
-    }
-
-    SDL_SetHintWithPriority(SDL_HINT_AUDIO_RESAMPLING_MODE, "medium", SDL_HINT_OVERRIDE);
-
-    {
-        int nDevices = SDL_GetNumAudioDevices(SDL_TRUE);
-        fprintf(stderr, "%s: found %d capture devices:\n", __func__, nDevices);
-        for (int i = 0; i < nDevices; i++) {
-            fprintf(stderr, "%s:    - Capture device #%d: '%s'\n", __func__, i, SDL_GetAudioDeviceName(i, SDL_TRUE));
-        }
-    }
-
-    SDL_AudioSpec capture_spec_requested;
-    SDL_AudioSpec capture_spec_obtained;
-
-    SDL_zero(capture_spec_requested);
-    SDL_zero(capture_spec_obtained);
-
-    capture_spec_requested.freq     = sample_rate;
-    capture_spec_requested.format   = AUDIO_F32;
-    capture_spec_requested.channels = 1;
-    capture_spec_requested.samples  = 1024;
-    capture_spec_requested.callback = [](void * userdata, uint8_t * stream, int len) {
-        audio_async * audio = (audio_async *) userdata;
-        audio->callback(stream, len);
-    };
-    capture_spec_requested.userdata = this;
-
-    if (capture_id >= 0) {
-        fprintf(stderr, "%s: attempt to open capture device %d : '%s' ...\n", __func__, capture_id, SDL_GetAudioDeviceName(capture_id, SDL_TRUE));
-        m_dev_id_in = SDL_OpenAudioDevice(SDL_GetAudioDeviceName(capture_id, SDL_TRUE), SDL_TRUE, &capture_spec_requested, &capture_spec_obtained, 0);
-    } else {
-        fprintf(stderr, "%s: attempt to open default capture device ...\n", __func__);
-        m_dev_id_in = SDL_OpenAudioDevice(nullptr, SDL_TRUE, &capture_spec_requested, &capture_spec_obtained, 0);
-    }
-
-    if (!m_dev_id_in) {
-        fprintf(stderr, "%s: couldn't open an audio device for capture: %s!\n", __func__, SDL_GetError());
-        m_dev_id_in = 0;
-
-        return false;
-    } else {
-        fprintf(stderr, "%s: obtained spec for input device (SDL Id = %d):\n", __func__, m_dev_id_in);
-        fprintf(stderr, "%s:     - sample rate:       %d\n",                   __func__, capture_spec_obtained.freq);
-        fprintf(stderr, "%s:     - format:            %d (required: %d)\n",    __func__, capture_spec_obtained.format,
-                capture_spec_requested.format);
-        fprintf(stderr, "%s:     - channels:          %d (required: %d)\n",    __func__, capture_spec_obtained.channels,
-                capture_spec_requested.channels);
-        fprintf(stderr, "%s:     - samples per frame: %d\n",                   __func__, capture_spec_obtained.samples);
-    }
-
-    m_sample_rate = capture_spec_obtained.freq;
-
-    m_audio.resize((m_sample_rate*m_len_ms)/1000);
-
-    return true;
-}
-
-bool audio_async::resume() {
-    if (!m_dev_id_in) {
-        fprintf(stderr, "%s: no audio device to resume!\n", __func__);
-        return false;
-    }
-
-    if (m_running) {
-        fprintf(stderr, "%s: already running!\n", __func__);
-        return false;
-    }
-
-    SDL_PauseAudioDevice(m_dev_id_in, 0);
-
-    m_running = true;
-
-    return true;
-}
-
-bool audio_async::pause() {
-    if (!m_dev_id_in) {
-        fprintf(stderr, "%s: no audio device to pause!\n", __func__);
-        return false;
-    }
-
-    if (!m_running) {
-        fprintf(stderr, "%s: already paused!\n", __func__);
-        return false;
-    }
-
-    SDL_PauseAudioDevice(m_dev_id_in, 1);
-
-    m_running = false;
-
-    return true;
-}
-
-bool audio_async::clear() {
-    if (!m_dev_id_in) {
-        fprintf(stderr, "%s: no audio device to clear!\n", __func__);
-        return false;
-    }
-
-    if (!m_running) {
-        fprintf(stderr, "%s: not running!\n", __func__);
-        return false;
-    }
-
-    {
-        std::lock_guard<std::mutex> lock(m_mutex);
-
-        m_audio_pos = 0;
-        m_audio_len = 0;
-    }
-
-    return true;
-}
-
-// callback to be called by SDL
-void audio_async::callback(uint8_t * stream, int len) {
-    if (!m_running) {
-        return;
-    }
-
-    const size_t n_samples = len / sizeof(float);
-
-    m_audio_new.resize(n_samples);
-    memcpy(m_audio_new.data(), stream, n_samples * sizeof(float));
-
-    //fprintf(stderr, "%s: %zu samples, pos %zu, len %zu\n", __func__, n_samples, m_audio_pos, m_audio_len);
-
-    {
-        std::lock_guard<std::mutex> lock(m_mutex);
-
-        if (m_audio_pos + n_samples > m_audio.size()) {
-            const size_t n0 = m_audio.size() - m_audio_pos;
-
-            memcpy(&m_audio[m_audio_pos], stream, n0 * sizeof(float));
-            memcpy(&m_audio[0], &stream[n0], (n_samples - n0) * sizeof(float));
-
-            m_audio_pos = (m_audio_pos + n_samples) % m_audio.size();
-            m_audio_len = m_audio.size();
-        } else {
-            memcpy(&m_audio[m_audio_pos], stream, n_samples * sizeof(float));
-
-            m_audio_pos = (m_audio_pos + n_samples) % m_audio.size();
-            m_audio_len = std::min(m_audio_len + n_samples, m_audio.size());
-        }
-    }
-}
-
-void audio_async::get(int ms, std::vector<float> & result) {
-    if (!m_dev_id_in) {
-        fprintf(stderr, "%s: no audio device to get audio from!\n", __func__);
-        return;
-    }
-
-    if (!m_running) {
-        fprintf(stderr, "%s: not running!\n", __func__);
-        return;
-    }
-
-    result.clear();
-
-    {
-        std::lock_guard<std::mutex> lock(m_mutex);
-
-        if (ms <= 0) {
-            ms = m_len_ms;
-        }
-
-        size_t n_samples = (m_sample_rate * ms) / 1000;
-        if (n_samples > m_audio_len) {
-            n_samples = m_audio_len;
-        }
-
-        result.resize(n_samples);
-
-        int s0 = m_audio_pos - n_samples;
-        if (s0 < 0) {
-            s0 += m_audio.size();
-        }
-
-        if (s0 + n_samples > m_audio.size()) {
-            const size_t n0 = m_audio.size() - s0;
-
-            memcpy(result.data(), &m_audio[s0], n0 * sizeof(float));
-            memcpy(&result[n0], &m_audio[0], (n_samples - n0) * sizeof(float));
-        } else {
-            memcpy(result.data(), &m_audio[s0], n_samples * sizeof(float));
-        }
-    }
-}
-
-bool sdl_poll_events() {
-    SDL_Event event;
-    while (SDL_PollEvent(&event)) {
-        switch (event.type) {
-            case SDL_QUIT:
-                {
-                    return false;
-                } break;
-            default:
-                break;
-        }
-    }
-
-    return true;
-}

examples/common-sdl.h

@@ -1,50 +0,0 @@
-#pragma once
-
-#include <SDL.h>
-#include <SDL_audio.h>
-
-#include <atomic>
-#include <cstdint>
-#include <vector>
-#include <mutex>
-
-//
-// SDL Audio capture
-//
-
-class audio_async {
-public:
-    audio_async(int len_ms);
-    ~audio_async();
-
-    bool init(int capture_id, int sample_rate);
-
-    // start capturing audio via the provided SDL callback
-    // keep last len_ms seconds of audio in a circular buffer
-    bool resume();
-    bool pause();
-    bool clear();
-
-    // callback to be called by SDL
-    void callback(uint8_t * stream, int len);
-
-    // get audio data from the circular buffer
-    void get(int ms, std::vector<float> & audio);
-
-private:
-    SDL_AudioDeviceID m_dev_id_in = 0;
-
-    int m_len_ms = 0;
-    int m_sample_rate = 0;
-
-    std::atomic_bool m_running;
-    std::mutex       m_mutex;
-
-    std::vector<float> m_audio;
-    std::vector<float> m_audio_new;
-    size_t             m_audio_pos = 0;
-    size_t             m_audio_len = 0;
-};
-
-// Return false if need to quit
-bool sdl_poll_events();

examples/common.cpp

@@ -1,809 +0,0 @@
-#define _USE_MATH_DEFINES // for M_PI
-
-#include "common.h"
-
-// third-party utilities
-// use your favorite implementations
-#define DR_WAV_IMPLEMENTATION
-#include "dr_wav.h"
-
-#include <cmath>
-#include <cstring>
-#include <fstream>
-#include <regex>
-#include <locale>
-#include <codecvt>
-#include <sstream>
-
-#if defined(_MSC_VER)
-#pragma warning(disable: 4244 4267) // possible loss of data
-#endif
-
-// Function to check if the next argument exists
-std::string get_next_arg(int& i, int argc, char** argv, const std::string& flag, gpt_params& params) {
-    if (i + 1 < argc && argv[i + 1][0] != '-') {
-        return argv[++i];
-    } else {
-        fprintf(stderr, "error: %s requires one argument.\n", flag.c_str());
-        gpt_print_usage(argc, argv, params);
-        exit(0);
-    }
-}
-
-bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
-    for (int i = 1; i < argc; i++) {
-        std::string arg = argv[i];
-
-        if (arg == "-s" || arg == "--seed") {
-            params.seed = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "-t" || arg == "--threads") {
-            params.n_threads = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "-ngl" || arg == "--gpu-layers" || arg == "--n-gpu-layers") {
-            params.n_gpu_layers = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "-p" || arg == "--prompt") {
-            params.prompt = get_next_arg(i, argc, argv, arg, params);
-        } else if (arg == "-n" || arg == "--n_predict") {
-            params.n_predict = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "--top_k") {
-            params.top_k = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "--top_p") {
-            params.top_p = std::stof(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "--temp") {
-            params.temp = std::stof(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "--repeat-last-n") {
-            params.repeat_last_n = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "--repeat-penalty") {
-            params.repeat_penalty = std::stof(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "-b" || arg == "--batch_size") {
-            params.n_batch= std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "-m" || arg == "--model") {
-            params.model = get_next_arg(i, argc, argv, arg, params);
-        } else if (arg == "-i" || arg == "--interactive") {
-            params.interactive = true;
-        } else if (arg == "-ip" || arg == "--interactive-port") {
-            params.interactive = true;
-            params.interactive_port = std::stoi(get_next_arg(i, argc, argv, arg, params));
-        } else if (arg == "-h" || arg == "--help") {
-            gpt_print_usage(argc, argv, params);
-            exit(0);
-        } else if (arg == "-f" || arg == "--file") {
-            get_next_arg(i, argc, argv, arg, params);
-            std::ifstream file(argv[i]);
-            if (!file) {
-                fprintf(stderr, "error: failed to open file '%s'\n", argv[i]);
-                break;
-            }
-            std::copy(std::istreambuf_iterator<char>(file), std::istreambuf_iterator<char>(), back_inserter(params.prompt));
-            if (params.prompt.back() == '\n') {
-                params.prompt.pop_back();
-            }
-        } else if (arg == "-tt" || arg == "--token_test") {
-            params.token_test = get_next_arg(i, argc, argv, arg, params);
-        }
-        else {
-            fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
-            gpt_print_usage(argc, argv, params);
-            exit(0);
-        }
-    }
-
-    return true;
-}
-
-void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
-    fprintf(stderr, "usage: %s [options]\n", argv[0]);
-    fprintf(stderr, "\n");
-    fprintf(stderr, "options:\n");
-    fprintf(stderr, "  -h, --help            show this help message and exit\n");
-    fprintf(stderr, "  -s SEED, --seed SEED  RNG seed (default: -1)\n");
-    fprintf(stderr, "  -t N, --threads N     number of threads to use during computation (default: %d)\n", params.n_threads);
-    fprintf(stderr, "  -ngl N, --gpu-layers N  number of layers to offload to GPU on supported models (default: %d)\n", params.n_gpu_layers);
-    fprintf(stderr, "  -p PROMPT, --prompt PROMPT\n");
-    fprintf(stderr, "                        prompt to start generation with (default: random)\n");
-    fprintf(stderr, "  -f FNAME, --file FNAME\n");
-    fprintf(stderr, "                        load prompt from a file\n");
-    fprintf(stderr, "  -tt TOKEN_TEST, --token_test TOKEN_TEST\n");
-    fprintf(stderr, "                        test tokenization\n");
-    fprintf(stderr, "  -n N, --n_predict N   number of tokens to predict (default: %d)\n", params.n_predict);
-    fprintf(stderr, "  --top_k N             top-k sampling (default: %d)\n", params.top_k);
-    fprintf(stderr, "  --top_p N             top-p sampling (default: %.1f)\n", params.top_p);
-    fprintf(stderr, "  --temp N              temperature (default: %.1f)\n", params.temp);
-    fprintf(stderr, "  --repeat-last-n N     last n tokens to consider for penalize (default: %d, 0 = disabled)\n", params.repeat_last_n);
-    fprintf(stderr, "  --repeat-penalty N    penalize repeat sequence of tokens (default: %.2f, 1.0 = disabled)\n", (double)params.repeat_penalty);
-    fprintf(stderr, "  -b N, --batch_size N  batch size for prompt processing (default: %d)\n", params.n_batch);
-    fprintf(stderr, "  -m FNAME, --model FNAME\n");
-    fprintf(stderr, "                        model path (default: %s)\n", params.model.c_str());
-    fprintf(stderr, "\n");
-}
-
-std::string gpt_random_prompt(std::mt19937 & rng) {
-    const int r = rng() % 10;
-    switch (r) {
-        case 0: return "So";
-        case 1: return "Once upon a time";
-        case 2: return "When";
-        case 3: return "The";
-        case 4: return "After";
-        case 5: return "If";
-        case 6: return "import";
-        case 7: return "He";
-        case 8: return "She";
-        case 9: return "They";
-        default: return "To";
-    }
-
-    return "The";
-}
-
-std::string trim(const std::string & s) {
-    std::regex e("^\\s+|\\s+$");
-    return std::regex_replace(s, e, "");
-}
-
-std::string replace(const std::string & s, const std::string & from, const std::string & to) {
-    std::string result = s;
-    size_t pos = 0;
-    while ((pos = result.find(from, pos)) != std::string::npos) {
-        result.replace(pos, from.length(), to);
-        pos += to.length();
-    }
-    return result;
-}
-
-void gpt_vocab::add_special_token(const std::string & token) {
-    special_tokens.push_back(token);
-}
-
-std::map<std::string, int32_t> json_parse(const std::string & fname) {
-    std::map<std::string, int32_t> result;
-
-    // read file into string
-    std::string json;
-    {
-        std::ifstream ifs(fname);
-        if (!ifs) {
-            fprintf(stderr, "Failed to open %s\n", fname.c_str());
-            exit(1);
-        }
-
-        json = std::string((std::istreambuf_iterator<char>(ifs)),
-                (std::istreambuf_iterator<char>()));
-    }
-
-    if (json[0] != '{') {
-        return result;
-    }
-
-    // parse json
-    {
-        bool has_key  = false;
-        bool in_token = false;
-
-        std::string str_key = "";
-        std::string str_val = "";
-
-        int n = json.size();
-        for (int i = 1; i < n; ++i) {
-            if (!in_token) {
-                if (json[i] == ' ') continue;
-                if (json[i] == '"') {
-                    in_token = true;
-                    continue;
-                }
-            } else {
-                if (json[i] == '\\' && i+1 < n) {
-                    if (has_key == false) {
-                        str_key += json[i];
-                    } else {
-                        str_val += json[i];
-                    }
-                    ++i;
-                } else if (json[i] == '"') {
-                    if (has_key == false) {
-                        has_key = true;
-                        ++i;
-                        while (json[i] == ' ') ++i;
-                        ++i; // :
-                        while (json[i] == ' ') ++i;
-                        if (json[i] != '\"') {
-                            while (json[i] != ',' && json[i] != '}') {
-                                str_val += json[i++];
-                            }
-                            has_key = false;
-                        } else {
-                            in_token = true;
-                            continue;
-                        }
-                    } else {
-                        has_key = false;
-                    }
-
-                    str_key = ::replace(str_key, "\\u0120", " " ); // \u0120 -> space
-                    str_key = ::replace(str_key, "\\u010a", "\n"); // \u010a -> new line
-                    str_key = ::replace(str_key, "\\\"",    "\""); // \\\"   -> "
-
-                    try {
-                        result[str_key] = std::stoi(str_val);
-                    } catch (...) {
-                        //fprintf(stderr, "%s: ignoring key '%s' with value '%s'\n", fname.c_str(), str_key.c_str(), str_val.c_str());
-
-                    }
-                    str_key = "";
-                    str_val = "";
-                    in_token = false;
-                    continue;
-                }
-                if (has_key == false) {
-                    str_key += json[i];
-                } else {
-                    str_val += json[i];
-                }
-            }
-        }
-    }
-
-    return result;
-}
-
-std::string convert_to_utf8(const std::wstring & input) {
-    std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
-    return converter.to_bytes(input);
-}
-
-
-std::wstring convert_to_wstring(const std::string & input) {
-    std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
-    return converter.from_bytes(input);
-}
-
-void gpt_split_words(std::string str, std::vector<std::string>& words) {
-    const std::string pattern = R"('s|'t|'re|'ve|'m|'ll|'d| ?[[:alpha:]]+| ?[[:digit:]]+| ?[^\s[:alpha:][:digit:]]+|\s+(?!\S)|\s+)";
-    const std::regex re(pattern);
-    std::smatch m;
-
-    while (std::regex_search(str, m, re)) {
-        for (auto x : m) {
-            words.push_back(x);
-        }
-        str = m.suffix();
-    }
-}
-
-std::vector<gpt_vocab::id> gpt_tokenize(const gpt_vocab & vocab, const std::string & text) {
-    std::vector<std::string> words;
-
-    // first split the text into words
-    {
-        std::string str = text;
-
-        // Generate the subpattern from the special_tokens vector if it's not empty
-        if (!vocab.special_tokens.empty()) {
-            const std::regex escape(R"([\[\\\^\$\.\|\?\*\+\(\)\{\}])");
-            std::string special_tokens_subpattern;
-            for (const auto & token : vocab.special_tokens) {
-                if (!special_tokens_subpattern.empty()) {
-                    special_tokens_subpattern += "|";
-                }
-                special_tokens_subpattern += std::regex_replace(token, escape, R"(\$&)");
-            }
-
-            std::regex re(special_tokens_subpattern);
-            std::smatch m;
-            // Split the text by special tokens.
-            while (std::regex_search(str, m, re)) {
-                // Split the substrings in-between special tokens into words.
-                gpt_split_words(m.prefix(), words);
-                // Add matched special tokens as words.
-                for (auto x : m) {
-                    words.push_back(x);
-                }
-                str = m.suffix();
-            }
-            // Remaining text without special tokens will be handled below.
-        }
-
-        gpt_split_words(str, words);
-    }
-
-    // find the longest token that forms each word in words:
-    std::vector<gpt_vocab::id> tokens;
-    for (const auto & word : words) {
-        for (int i = 0; i < (int) word.size(); ){
-            for (int j = word.size() - 1; j >= i; j--){
-                auto cand = word.substr(i, j-i+1);
-                auto it = vocab.token_to_id.find(cand);
-                if (it != vocab.token_to_id.end()){ // word.substr(i, j-i+1) in vocab
-                    tokens.push_back(it->second);
-                    i = j + 1;
-                    break;
-                }
-                else if (j == i){ // word.substr(i, 1) has no matching
-                    fprintf(stderr, "%s: unknown token '%s'\n", __func__, word.substr(i, 1).data());
-                    i++;
-                }
-            }
-        }
-    }
-
-    return tokens;
-}
-
-std::vector<gpt_vocab::id> parse_tokens_from_string(const std::string& input, char delimiter) {
-    std::vector<gpt_vocab::id> output;
-    std::stringstream ss(input);
-    std::string token;
-
-    while (std::getline(ss, token, delimiter)) {
-        output.push_back(std::stoi(token));
-    }
-
-    return output;
-}
-
-std::map<std::string, std::vector<gpt_vocab::id>> extract_tests_from_file(const std::string & fpath_test){
-    if (fpath_test.empty()){
-        fprintf(stderr, "%s : No test file found.\n", __func__);
-        return std::map<std::string, std::vector<gpt_vocab::id>>();
-    }
-
-    std::map<std::string, std::vector<gpt_vocab::id>> tests;
-
-    auto fin = std::ifstream(fpath_test, std::ios_base::in);
-    const char * delimeter = " => ";
-    const char del_tok = ',';
-    std::string line;
-    while (std::getline(fin, line)) {
-        size_t delimiterPos = line.find(delimeter);
-        if (delimiterPos != std::string::npos) {
-            std::string text = line.substr(0, delimiterPos);
-            std::string s_tokens = line.substr(delimiterPos + std::strlen(delimeter));
-            tests[text] = parse_tokens_from_string(s_tokens, del_tok);
-        }
-    }
-    return tests;
-}
-
-void test_gpt_tokenizer(gpt_vocab & vocab, const std::string & fpath_test){
-    std::map<std::string, std::vector<gpt_vocab::id>> tests = extract_tests_from_file(fpath_test);
-
-    size_t n_fails = 0;
-
-    for (const auto & test : tests) {
-        std::vector<gpt_vocab::id> tokens = gpt_tokenize(vocab, test.first);
-
-        if (tokens != test.second){
-            n_fails++;
-
-            // print out failure cases
-            fprintf(stderr, "%s : failed test: '%s'\n", __func__, test.first.c_str());
-            fprintf(stderr, "%s : tokens in hf:   ", __func__);
-            for (const auto & t : test.second) {
-                fprintf(stderr, "%s(%d), ", vocab.id_to_token[t].c_str(), t);
-            }
-            fprintf(stderr, "\n");
-            fprintf(stderr, "%s : tokens in ggml: ", __func__);
-            for (const auto & t : tokens) {
-                fprintf(stderr, "%s(%d), ", vocab.id_to_token[t].c_str(), t);
-            }
-            fprintf(stderr, "\n");
-        }
-    }
-
-    fprintf(stderr, "%s : %zu tests failed out of %zu tests.\n", __func__, n_fails, tests.size());
-}
-
-bool gpt_vocab_init(const std::string & fname, gpt_vocab & vocab) {
-    printf("%s: loading vocab from '%s'\n", __func__, fname.c_str());
-
-    vocab.token_to_id = ::json_parse(fname);
-
-    for (const auto & kv : vocab.token_to_id) {
-        vocab.id_to_token[kv.second] = kv.first;
-    }
-
-    printf("%s: vocab size = %d\n", __func__, (int) vocab.token_to_id.size());
-
-    // print the vocabulary
-    //for (auto kv : vocab.token_to_id) {
-    //    printf("'%s' -> %d\n", kv.first.data(), kv.second);
-    //}
-
-    return true;
-}
-
-gpt_vocab::id gpt_sample_top_k_top_p(
-        const gpt_vocab & vocab,
-        const float * logits,
-        int    top_k,
-        double top_p,
-        double temp,
-        std::mt19937 & rng) {
-    int n_logits = vocab.id_to_token.size();
-
-    std::vector<std::pair<double, gpt_vocab::id>> logits_id;
-    logits_id.reserve(n_logits);
-
-    {
-        const double scale = 1.0/temp;
-        for (int i = 0; i < n_logits; ++i) {
-            logits_id.push_back(std::make_pair(logits[i]*scale, i));
-        }
-    }
-
-    // find the top K tokens
-    std::partial_sort(
-            logits_id.begin(),
-            logits_id.begin() + top_k, logits_id.end(),
-            [](const std::pair<double, gpt_vocab::id> & a, const std::pair<double, gpt_vocab::id> & b) {
-        return a.first > b.first;
-    });
-
-    logits_id.resize(top_k);
-
-    double maxl = -INFINITY;
-    for (const auto & kv : logits_id) {
-        maxl = std::max(maxl, kv.first);
-    }
-
-    // compute probs for the top K tokens
-    std::vector<double> probs;
-    probs.reserve(logits_id.size());
-
-    double sum = 0.0;
-    for (const auto & kv : logits_id) {
-        double p = exp(kv.first - maxl);
-        probs.push_back(p);
-        sum += p;
-    }
-
-    // normalize the probs
-    for (auto & p : probs) {
-        p /= sum;
-    }
-
-    if (top_p < 1.0f) {
-        double cumsum = 0.0f;
-        for (int i = 0; i < top_k; i++) {
-            cumsum += probs[i];
-            if (cumsum >= top_p) {
-                top_k = i + 1;
-                probs.resize(top_k);
-                logits_id.resize(top_k);
-                break;
-            }
-        }
-
-        cumsum = 1.0/cumsum;
-        for (int i = 0; i < (int) probs.size(); i++) {
-            probs[i] *= cumsum;
-        }
-    }
-
-    //printf("\n");
-    //for (int i = 0; i < (int) probs.size(); i++) {
-    //    printf("%d: '%s' %f\n", i, vocab.id_to_token.at(logits_id[i].second).c_str(), probs[i]);
-    //}
-    //exit(0);
-
-    std::discrete_distribution<> dist(probs.begin(), probs.end());
-    int idx = dist(rng);
-
-    return logits_id[idx].second;
-}
-
-gpt_vocab::id gpt_sample_top_k_top_p_repeat(
-        const gpt_vocab & vocab,
-        const float * logits,
-        const int32_t * last_n_tokens_data,
-        size_t last_n_tokens_data_size,
-        int    top_k,
-        double top_p,
-        double temp,
-        int repeat_last_n,
-        float repeat_penalty,
-        std::mt19937 & rng) {
-
-    int n_logits = vocab.id_to_token.size();
-
-    const auto * plogits = logits;
-
-    const auto last_n_tokens = std::vector<int32_t>(last_n_tokens_data, last_n_tokens_data + last_n_tokens_data_size);
-
-    if (temp <= 0) {
-        // select the token with the highest logit directly
-        float max_logit = plogits[0];
-        gpt_vocab::id max_id = 0;
-
-        for (int i = 1; i < n_logits; ++i) {
-            if (plogits[i] > max_logit) {
-                max_logit = plogits[i];
-                max_id = i;
-            }
-        }
-        return max_id;
-    }
-
-
-    std::vector<std::pair<double, gpt_vocab::id>> logits_id;
-    logits_id.reserve(n_logits);
-
-    {
-        const float scale = 1.0f/temp;
-        for (int i = 0; i < n_logits; ++i) {
-            // repetition penalty from ctrl paper (https://arxiv.org/abs/1909.05858)
-            // credit https://github.com/facebookresearch/llama/compare/main...shawwn:llama:main
-            if (repeat_last_n > 0 && std::find(last_n_tokens.end()-repeat_last_n, last_n_tokens.end(), i) != last_n_tokens.end()) {
-                // if score < 0 then repetition penalty has to multiplied to reduce the previous token probability
-                if (plogits[i] < 0.0f) {
-                    logits_id.push_back(std::make_pair(plogits[i]*scale*repeat_penalty, i));
-                } else {
-                    logits_id.push_back(std::make_pair(plogits[i]*scale/repeat_penalty, i));
-                }
-            } else {
-                logits_id.push_back(std::make_pair(plogits[i]*scale, i));
-            }
-        }
-    }
-
-    // find the top K tokens
-    std::partial_sort(
-            logits_id.begin(),
-            logits_id.begin() + top_k, logits_id.end(),
-            [](const std::pair<double, gpt_vocab::id> & a, const std::pair<double, gpt_vocab::id> & b) {
-        return a.first > b.first;
-    });
-
-    logits_id.resize(top_k);
-
-    double maxl = -INFINITY;
-    for (const auto & kv : logits_id) {
-        maxl = std::max(maxl, kv.first);
-    }
-
-    // compute probs for the top K tokens
-    std::vector<double> probs;
-    probs.reserve(logits_id.size());
-
-    double sum = 0.0;
-    for (const auto & kv : logits_id) {
-        double p = exp(kv.first - maxl);
-        probs.push_back(p);
-        sum += p;
-    }
-
-    // normalize the probs
-    for (auto & p : probs) {
-        p /= sum;
-    }
-
-    if (top_p < 1.0f) {
-        double cumsum = 0.0f;
-        for (int i = 0; i < top_k; i++) {
-            cumsum += probs[i];
-            if (cumsum >= top_p) {
-                top_k = i + 1;
-                probs.resize(top_k);
-                logits_id.resize(top_k);
-                break;
-            }
-        }
-
-        cumsum = 1.0/cumsum;
-        for (int i = 0; i < (int) probs.size(); i++) {
-            probs[i] *= cumsum;
-        }
-    }
-
-//    printf("\n");
-//    for (int i = 0; i < (int) probs.size(); i++) {
-//    for (int i = 0; i < 10; i++) {
-//        printf("%d: '%s' %f\n", i, vocab.id_to_token.at(logits_id[i].second).c_str(), probs[i]);
-//    }
-
-    std::discrete_distribution<> dist(probs.begin(), probs.end());
-    int idx = dist(rng);
-
-    return logits_id[idx].second;
-
-}
-
-bool read_wav(const std::string & fname, std::vector<float>& pcmf32, std::vector<std::vector<float>>& pcmf32s, bool stereo) {
-    drwav wav;
-    std::vector<uint8_t> wav_data; // used for pipe input from stdin
-
-    if (fname == "-") {
-        {
-            uint8_t buf[1024];
-            while (true)
-            {
-                const size_t n = fread(buf, 1, sizeof(buf), stdin);
-                if (n == 0) {
-                    break;
-                }
-                wav_data.insert(wav_data.end(), buf, buf + n);
-            }
-        }
-
-        if (drwav_init_memory(&wav, wav_data.data(), wav_data.size(), nullptr) == false) {
-            fprintf(stderr, "error: failed to open WAV file from stdin\n");
-            return false;
-        }
-
-        fprintf(stderr, "%s: read %zu bytes from stdin\n", __func__, wav_data.size());
-    }
-    else if (drwav_init_file(&wav, fname.c_str(), nullptr) == false) {
-        fprintf(stderr, "error: failed to open '%s' as WAV file\n", fname.c_str());
-        return false;
-    }
-
-    if (wav.channels != 1 && wav.channels != 2) {
-        fprintf(stderr, "%s: WAV file '%s' must be mono or stereo\n", __func__, fname.c_str());
-        return false;
-    }
-
-    if (stereo && wav.channels != 2) {
-        fprintf(stderr, "%s: WAV file '%s' must be stereo for diarization\n", __func__, fname.c_str());
-        return false;
-    }
-
-    if (wav.sampleRate != COMMON_SAMPLE_RATE) {
-        fprintf(stderr, "%s: WAV file '%s' must be %i kHz\n", __func__, fname.c_str(), COMMON_SAMPLE_RATE/1000);
-        return false;
-    }
-
-    if (wav.bitsPerSample != 16) {
-        fprintf(stderr, "%s: WAV file '%s' must be 16-bit\n", __func__, fname.c_str());
-        return false;
-    }
-
-    const uint64_t n = wav_data.empty() ? wav.totalPCMFrameCount : wav_data.size()/(wav.channels*wav.bitsPerSample/8);
-
-    std::vector<int16_t> pcm16;
-    pcm16.resize(n*wav.channels);
-    drwav_read_pcm_frames_s16(&wav, n, pcm16.data());
-    drwav_uninit(&wav);
-
-    // convert to mono, float
-    pcmf32.resize(n);
-    if (wav.channels == 1) {
-        for (uint64_t i = 0; i < n; i++) {
-            pcmf32[i] = float(pcm16[i])/32768.0f;
-        }
-    } else {
-        for (uint64_t i = 0; i < n; i++) {
-            pcmf32[i] = float(pcm16[2*i] + pcm16[2*i + 1])/65536.0f;
-        }
-    }
-
-    if (stereo) {
-        // convert to stereo, float
-        pcmf32s.resize(2);
-
-        pcmf32s[0].resize(n);
-        pcmf32s[1].resize(n);
-        for (uint64_t i = 0; i < n; i++) {
-            pcmf32s[0][i] = float(pcm16[2*i])/32768.0f;
-            pcmf32s[1][i] = float(pcm16[2*i + 1])/32768.0f;
-        }
-    }
-
-    return true;
-}
-
-void high_pass_filter(std::vector<float> & data, float cutoff, float sample_rate) {
-    const float rc = 1.0f / (2.0f * M_PI * cutoff);
-    const float dt = 1.0f / sample_rate;
-    const float alpha = dt / (rc + dt);
-
-    float y = data[0];
-
-    for (size_t i = 1; i < data.size(); i++) {
-        y = alpha * (y + data[i] - data[i - 1]);
-        data[i] = y;
-    }
-}
-
-bool vad_simple(std::vector<float> & pcmf32, int sample_rate, int last_ms, float vad_thold, float freq_thold, bool verbose) {
-    const int n_samples      = pcmf32.size();
-    const int n_samples_last = (sample_rate * last_ms) / 1000;
-
-    if (n_samples_last >= n_samples) {
-        // not enough samples - assume no speech
-        return false;
-    }
-
-    if (freq_thold > 0.0f) {
-        high_pass_filter(pcmf32, freq_thold, sample_rate);
-    }
-
-    float energy_all  = 0.0f;
-    float energy_last = 0.0f;
-
-    for (int i = 0; i < n_samples; i++) {
-        energy_all += fabsf(pcmf32[i]);
-
-        if (i >= n_samples - n_samples_last) {
-            energy_last += fabsf(pcmf32[i]);
-        }
-    }
-
-    energy_all  /= n_samples;
-    energy_last /= n_samples_last;
-
-    if (verbose) {
-        fprintf(stderr, "%s: energy_all: %f, energy_last: %f, vad_thold: %f, freq_thold: %f\n", __func__, energy_all, energy_last, vad_thold, freq_thold);
-    }
-
-    if (energy_last > vad_thold*energy_all) {
-        return false;
-    }
-
-    return true;
-}
-
-float similarity(const std::string & s0, const std::string & s1) {
-    const size_t len0 = s0.size() + 1;
-    const size_t len1 = s1.size() + 1;
-
-    std::vector<int> col(len1, 0);
-    std::vector<int> prevCol(len1, 0);
-
-    for (size_t i = 0; i < len1; i++) {
-        prevCol[i] = i;
-    }
-
-    for (size_t i = 0; i < len0; i++) {
-        col[0] = i;
-        for (size_t j = 1; j < len1; j++) {
-            col[j] = std::min(std::min(1 + col[j - 1], 1 + prevCol[j]), prevCol[j - 1] + (i > 0 && s0[i - 1] == s1[j - 1] ? 0 : 1));
-        }
-        col.swap(prevCol);
-    }
-
-    const float dist = prevCol[len1 - 1];
-
-    return 1.0f - (dist / std::max(s0.size(), s1.size()));
-}
-
-bool sam_params_parse(int argc, char ** argv, sam_params & params) {
-    for (int i = 1; i < argc; i++) {
-        std::string arg = argv[i];
-
-        if (arg == "-s" || arg == "--seed") {
-            params.seed = std::stoi(argv[++i]);
-        } else if (arg == "-t" || arg == "--threads") {
-            params.n_threads = std::stoi(argv[++i]);
-        } else if (arg == "-m" || arg == "--model") {
-            params.model = argv[++i];
-        } else if (arg == "-i" || arg == "--inp") {
-            params.fname_inp = argv[++i];
-        } else if (arg == "-o" || arg == "--out") {
-            params.fname_out = argv[++i];
-        } else if (arg == "-h" || arg == "--help") {
-            sam_print_usage(argc, argv, params);
-            exit(0);
-        } else {
-            fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
-            sam_print_usage(argc, argv, params);
-            exit(0);
-        }
-    }
-
-    return true;
-}
-
-void sam_print_usage(int argc, char ** argv, const sam_params & params) {
-    fprintf(stderr, "usage: %s [options]\n", argv[0]);
-    fprintf(stderr, "\n");
-    fprintf(stderr, "options:\n");
-    fprintf(stderr, "  -h, --help            show this help message and exit\n");
-    fprintf(stderr, "  -s SEED, --seed SEED  RNG seed (default: -1)\n");
-    fprintf(stderr, "  -t N, --threads N     number of threads to use during computation (default: %d)\n", params.n_threads);
-    fprintf(stderr, "  -m FNAME, --model FNAME\n");
-    fprintf(stderr, "                        model path (default: %s)\n", params.model.c_str());
-    fprintf(stderr, "  -i FNAME, --inp FNAME\n");
-    fprintf(stderr, "                        input file (default: %s)\n", params.fname_inp.c_str());
-    fprintf(stderr, "  -o FNAME, --out FNAME\n");
-    fprintf(stderr, "                        output file (default: %s)\n", params.fname_out.c_str());
-    fprintf(stderr, "\n");
-}

examples/common.h

@@ -1,176 +0,0 @@
-// Various helper functions and utilities
-
-#pragma once
-
-#include <string>
-#include <map>
-#include <vector>
-#include <random>
-#include <thread>
-
-#define COMMON_SAMPLE_RATE 16000
-
-//
-// GPT CLI argument parsing
-//
-
-struct gpt_params {
-    int32_t seed      = -1;  // RNG seed
-    int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
-    int32_t n_predict = 200; // new tokens to predict
-    int32_t n_batch   = 8;   // batch size for prompt processing
-
-    // sampling parameters
-    int32_t top_k          = 40;
-    float   top_p          = 0.9f;
-    float   temp           = 0.9f;
-    int32_t repeat_last_n  = 64;
-    float   repeat_penalty = 1.00f;
-
-    std::string model      = "models/gpt-2-117M/ggml-model.bin"; // model path
-    std::string prompt     = "";
-    std::string token_test = "";
-
-    bool    interactive      = false;
-    int32_t interactive_port = -1;
-
-    int32_t n_gpu_layers     = 0;
-};
-
-bool gpt_params_parse(int argc, char ** argv, gpt_params & params);
-
-void gpt_print_usage(int argc, char ** argv, const gpt_params & params);
-
-std::string gpt_random_prompt(std::mt19937 & rng);
-
-//
-// Vocab utils
-//
-
-std::string trim(const std::string & s);
-
-std::string replace(
-        const std::string & s,
-        const std::string & from,
-        const std::string & to);
-
-struct gpt_vocab {
-    using id    = int32_t;
-    using token = std::string;
-
-    std::map<token, id> token_to_id;
-    std::map<id, token> id_to_token;
-    std::vector<std::string> special_tokens;
-
-    void add_special_token(const std::string & token);
-};
-
-// poor-man's JSON parsing
-std::map<std::string, int32_t> json_parse(const std::string & fname);
-
-std::string convert_to_utf8(const std::wstring & input);
-
-std::wstring convert_to_wstring(const std::string & input);
-
-void gpt_split_words(std::string str, std::vector<std::string>& words);
-
-// split text into tokens
-//
-// ref: https://github.com/openai/gpt-2/blob/a74da5d99abaaba920de8131d64da2862a8f213b/src/encoder.py#L53
-//
-// Regex (Python):
-// r"""'s|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+"""
-//
-// Regex (C++):
-// R"('s|'t|'re|'ve|'m|'ll|'d| ?[[:alpha:]]+| ?[[:digit:]]+| ?[^\s[:alpha:][:digit:]]+|\s+(?!\S)|\s+)"
-//
-std::vector<gpt_vocab::id> gpt_tokenize(const gpt_vocab & vocab, const std::string & text);
-
-// test outputs of gpt_tokenize
-//
-//   - compare with tokens generated by the huggingface tokenizer
-//   - test cases are chosen based on the model's main language (under 'prompt' directory)
-//   - if all sentences are tokenized identically, print 'All tests passed.'
-//   - otherwise, print sentence, huggingface tokens, ggml tokens
-//
-void test_gpt_tokenizer(gpt_vocab & vocab, const std::string & fpath_test);
-
-// load the tokens from encoder.json
-bool gpt_vocab_init(const std::string & fname, gpt_vocab & vocab);
-
-// sample next token given probabilities for each embedding
-//
-//   - consider only the top K tokens
-//   - from them, consider only the top tokens with cumulative probability > P
-//
-// TODO: not sure if this implementation is correct
-// TODO: temperature is not implemented
-//
-gpt_vocab::id gpt_sample_top_k_top_p(
-        const gpt_vocab & vocab,
-        const float * logits,
-        int    top_k,
-        double top_p,
-        double temp,
-        std::mt19937 & rng);
-
-gpt_vocab::id gpt_sample_top_k_top_p_repeat(
-        const gpt_vocab & vocab,
-        const float * logits,
-        const int32_t * last_n_tokens_data,
-        size_t last_n_tokens_data_size,
-        int    top_k,
-        double top_p,
-        double temp,
-        int repeat_last_n,
-        float repeat_penalty,
-        std::mt19937 & rng);
-
-//
-// Audio utils
-//
-
-// Read WAV audio file and store the PCM data into pcmf32
-// The sample rate of the audio must be equal to COMMON_SAMPLE_RATE
-// If stereo flag is set and the audio has 2 channels, the pcmf32s will contain 2 channel PCM
-bool read_wav(
-        const std::string & fname,
-        std::vector<float> & pcmf32,
-        std::vector<std::vector<float>> & pcmf32s,
-        bool stereo);
-
-// Apply a high-pass frequency filter to PCM audio
-// Suppresses frequencies below cutoff Hz
-void high_pass_filter(
-        std::vector<float> & data,
-        float cutoff,
-        float sample_rate);
-
-// Basic voice activity detection (VAD) using audio energy adaptive threshold
-bool vad_simple(
-        std::vector<float> & pcmf32,
-        int   sample_rate,
-        int   last_ms,
-        float vad_thold,
-        float freq_thold,
-        bool  verbose);
-
-// compute similarity between two strings using Levenshtein distance
-float similarity(const std::string & s0, const std::string & s1);
-
-//
-// SAM argument parsing
-//
-
-struct sam_params {
-    int32_t seed      = -1; // RNG seed
-    int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
-
-    std::string model     = "models/sam-vit-b/ggml-model-f16.bin"; // model path
-    std::string fname_inp = "img.jpg";
-    std::string fname_out = "img.out";
-};
-
-bool sam_params_parse(int argc, char ** argv, sam_params & params);
-
-void sam_print_usage(int argc, char ** argv, const sam_params & params);

examples/grammar-parser.cpp

@@ -1,423 +0,0 @@
-#include "grammar-parser.h"
-#include <cstdint>
-#include <cwchar>
-#include <string>
-#include <utility>
-#include <stdexcept>
-#include <exception>
-
-namespace grammar_parser {
-    // NOTE: assumes valid utf8 (but checks for overrun)
-    // copied from whisper.cpp
-    std::pair<uint32_t, const char *> decode_utf8(const char * src) {
-        static const int lookup[] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 3, 4 };
-        uint8_t  first_byte = static_cast<uint8_t>(*src);
-        uint8_t  highbits   = first_byte >> 4;
-        int      len        = lookup[highbits];
-        uint8_t  mask       = (1 << (8 - len)) - 1;
-        uint32_t value      = first_byte & mask;
-        const char * end    = src + len; // may overrun!
-        const char * pos    = src + 1;
-        for ( ; pos < end && *pos; pos++) {
-            value = (value << 6) + (static_cast<uint8_t>(*pos) & 0x3F);
-        }
-        return std::make_pair(value, pos);
-    }
-
-    uint32_t get_symbol_id(parse_state & state, const char * src, size_t len) {
-        uint32_t next_id = static_cast<uint32_t>(state.symbol_ids.size());
-        auto result = state.symbol_ids.insert(std::make_pair(std::string(src, len), next_id));
-        return result.first->second;
-    }
-
-    uint32_t generate_symbol_id(parse_state & state, const std::string & base_name) {
-        uint32_t next_id = static_cast<uint32_t>(state.symbol_ids.size());
-        state.symbol_ids[base_name + '_' + std::to_string(next_id)] = next_id;
-        return next_id;
-    }
-
-    void add_rule(
-            parse_state & state,
-            uint32_t      rule_id,
-            const std::vector<whisper_grammar_element> & rule) {
-        if (state.rules.size() <= rule_id) {
-            state.rules.resize(rule_id + 1);
-        }
-        state.rules[rule_id] = rule;
-    }
-
-    bool is_word_char(char c) {
-        return ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || c == '-' || ('0' <= c && c <= '9');
-    }
-
-    std::pair<uint32_t, const char *> parse_hex(const char * src, int size) {
-        const char * pos   = src;
-        const char * end   = src + size;
-        uint32_t     value = 0;
-        for ( ; pos < end && *pos; pos++) {
-            value <<= 4;
-            char c = *pos;
-            if ('a' <= c && c <= 'f') {
-                value += c - 'a' + 10;
-            } else if ('A' <= c && c <= 'F') {
-                value += c - 'A' + 10;
-            } else if ('0' <= c && c <= '9') {
-                value += c - '0';
-            } else {
-                break;
-            }
-        }
-        if (pos != end) {
-            throw std::runtime_error("expecting " + std::to_string(size) + " hex chars at " + src);
-        }
-        return std::make_pair(value, pos);
-    }
-
-    const char * parse_space(const char * src, bool newline_ok) {
-        const char * pos = src;
-        while (*pos == ' ' || *pos == '\t' || *pos == '#' ||
-                (newline_ok && (*pos == '\r' || *pos == '\n'))) {
-            if (*pos == '#') {
-                while (*pos && *pos != '\r' && *pos != '\n') {
-                    pos++;
-                }
-            } else {
-                pos++;
-            }
-        }
-        return pos;
-    }
-
-    const char * parse_name(const char * src) {
-        const char * pos = src;
-        while (is_word_char(*pos)) {
-            pos++;
-        }
-        if (pos == src) {
-            throw std::runtime_error(std::string("expecting name at ") + src);
-        }
-        return pos;
-    }
-
-    std::pair<uint32_t, const char *> parse_char(const char * src) {
-        if (*src == '\\') {
-            switch (src[1]) {
-                case 'x': return parse_hex(src + 2, 2);
-                case 'u': return parse_hex(src + 2, 4);
-                case 'U': return parse_hex(src + 2, 8);
-                case 't': return std::make_pair('\t', src + 2);
-                case 'r': return std::make_pair('\r', src + 2);
-                case 'n': return std::make_pair('\n', src + 2);
-                case '\\':
-                case '"':
-                case '[':
-                case ']':
-                    return std::make_pair(src[1], src + 2);
-                default:
-                    throw std::runtime_error(std::string("unknown escape at ") + src);
-            }
-        } else if (*src) {
-            return decode_utf8(src);
-        }
-        throw std::runtime_error("unexpected end of input");
-    }
-
-    const char * parse_alternates(
-            parse_state       & state,
-            const char        * src,
-            const std::string & rule_name,
-            uint32_t            rule_id,
-            bool                is_nested);
-
-    const char * parse_sequence(
-            parse_state                        & state,
-            const char                         * src,
-            const std::string                  & rule_name,
-            std::vector<whisper_grammar_element> & out_elements,
-            bool                                 is_nested) {
-        size_t last_sym_start = out_elements.size();
-        const char * pos = src;
-        while (*pos) {
-            if (*pos == '"') { // literal string
-                pos++;
-                last_sym_start = out_elements.size();
-                while (*pos != '"') {
-                    auto char_pair = parse_char(pos);
-                         pos       = char_pair.second;
-                    out_elements.push_back({WHISPER_GRETYPE_CHAR, char_pair.first});
-                }
-                pos = parse_space(pos + 1, is_nested);
-            } else if (*pos == '[') { // char range(s)
-                pos++;
-                enum whisper_gretype start_type = WHISPER_GRETYPE_CHAR;
-                if (*pos == '^') {
-                    pos++;
-                    start_type = WHISPER_GRETYPE_CHAR_NOT;
-                }
-                last_sym_start = out_elements.size();
-                while (*pos != ']') {
-                    auto char_pair = parse_char(pos);
-                         pos       = char_pair.second;
-                    enum whisper_gretype type = last_sym_start < out_elements.size()
-                        ? WHISPER_GRETYPE_CHAR_ALT
-                        : start_type;
-
-                    out_elements.push_back({type, char_pair.first});
-                    if (pos[0] == '-' && pos[1] != ']') {
-                        auto endchar_pair = parse_char(pos + 1);
-                             pos          = endchar_pair.second;
-                        out_elements.push_back({WHISPER_GRETYPE_CHAR_RNG_UPPER, endchar_pair.first});
-                    }
-                }
-                pos = parse_space(pos + 1, is_nested);
-            } else if (is_word_char(*pos)) { // rule reference
-                const char * name_end    = parse_name(pos);
-                uint32_t     ref_rule_id = get_symbol_id(state, pos, name_end - pos);
-                pos = parse_space(name_end, is_nested);
-                last_sym_start = out_elements.size();
-                out_elements.push_back({WHISPER_GRETYPE_RULE_REF, ref_rule_id});
-            } else if (*pos == '(') { // grouping
-                // parse nested alternates into synthesized rule
-                pos = parse_space(pos + 1, true);
-                uint32_t sub_rule_id = generate_symbol_id(state, rule_name);
-                pos = parse_alternates(state, pos, rule_name, sub_rule_id, true);
-                last_sym_start = out_elements.size();
-                // output reference to synthesized rule
-                out_elements.push_back({WHISPER_GRETYPE_RULE_REF, sub_rule_id});
-                if (*pos != ')') {
-                    throw std::runtime_error(std::string("expecting ')' at ") + pos);
-                }
-                pos = parse_space(pos + 1, is_nested);
-            } else if (*pos == '*' || *pos == '+' || *pos == '?') { // repetition operator
-                if (last_sym_start == out_elements.size()) {
-                    throw std::runtime_error(std::string("expecting preceeding item to */+/? at ") + pos);
-                }
-
-                // apply transformation to previous symbol (last_sym_start to end) according to
-                // rewrite rules:
-                // S* --> S' ::= S S' |
-                // S+ --> S' ::= S S' | S
-                // S? --> S' ::= S |
-                uint32_t sub_rule_id = generate_symbol_id(state, rule_name);
-                std::vector<whisper_grammar_element> sub_rule;
-                // add preceding symbol to generated rule
-                sub_rule.insert(
-                    sub_rule.end(), out_elements.begin() + last_sym_start, out_elements.end());
-                if (*pos == '*' || *pos == '+') {
-                    // cause generated rule to recurse
-                    sub_rule.push_back({WHISPER_GRETYPE_RULE_REF, sub_rule_id});
-                }
-                // mark start of alternate def
-                sub_rule.push_back({WHISPER_GRETYPE_ALT, 0});
-                if (*pos == '+') {
-                    // add preceding symbol as alternate only for '+' (otherwise empty)
-                    sub_rule.insert(
-                        sub_rule.end(), out_elements.begin() + last_sym_start, out_elements.end());
-                }
-                sub_rule.push_back({WHISPER_GRETYPE_END, 0});
-                add_rule(state, sub_rule_id, sub_rule);
-
-                // in original rule, replace previous symbol with reference to generated rule
-                out_elements.resize(last_sym_start);
-                out_elements.push_back({WHISPER_GRETYPE_RULE_REF, sub_rule_id});
-
-                pos = parse_space(pos + 1, is_nested);
-            } else {
-                break;
-            }
-        }
-        return pos;
-    }
-
-    const char * parse_alternates(
-            parse_state       & state,
-            const char        * src,
-            const std::string & rule_name,
-            uint32_t            rule_id,
-            bool                is_nested) {
-        std::vector<whisper_grammar_element> rule;
-        const char * pos = parse_sequence(state, src, rule_name, rule, is_nested);
-        while (*pos == '|') {
-            rule.push_back({WHISPER_GRETYPE_ALT, 0});
-            pos = parse_space(pos + 1, true);
-            pos = parse_sequence(state, pos, rule_name, rule, is_nested);
-        }
-        rule.push_back({WHISPER_GRETYPE_END, 0});
-        add_rule(state, rule_id, rule);
-        return pos;
-    }
-
-    const char * parse_rule(parse_state & state, const char * src) {
-        const char * name_end = parse_name(src);
-        const char * pos      = parse_space(name_end, false);
-        size_t       name_len = name_end - src;
-        uint32_t     rule_id  = get_symbol_id(state, src, name_len);
-        const std::string name(src, name_len);
-
-        if (!(pos[0] == ':' && pos[1] == ':' && pos[2] == '=')) {
-            throw std::runtime_error(std::string("expecting ::= at ") + pos);
-        }
-        pos = parse_space(pos + 3, true);
-
-        pos = parse_alternates(state, pos, name, rule_id, false);
-
-        if (*pos == '\r') {
-            pos += pos[1] == '\n' ? 2 : 1;
-        } else if (*pos == '\n') {
-            pos++;
-        } else if (*pos) {
-            throw std::runtime_error(std::string("expecting newline or end at ") + pos);
-        }
-        return parse_space(pos, true);
-    }
-
-    parse_state parse(const char * src) {
-        try {
-            parse_state state;
-            const char * pos = parse_space(src, true);
-            while (*pos) {
-                pos = parse_rule(state, pos);
-            }
-            return state;
-        } catch (const std::exception & err) {
-            fprintf(stderr, "%s: error parsing grammar: %s\n", __func__, err.what());
-            return parse_state();
-        }
-    }
-
-    void print_grammar_char(FILE * file, uint32_t c) {
-        if (0x20 <= c && c <= 0x7f) {
-            fprintf(file, "%c", static_cast<char>(c));
-        } else {
-            // cop out of encoding UTF-8
-            fprintf(file, "<U+%04X>", c);
-        }
-    }
-
-    bool is_char_element(whisper_grammar_element elem) {
-        switch (elem.type) {
-            case WHISPER_GRETYPE_CHAR:           return true;
-            case WHISPER_GRETYPE_CHAR_NOT:       return true;
-            case WHISPER_GRETYPE_CHAR_ALT:       return true;
-            case WHISPER_GRETYPE_CHAR_RNG_UPPER: return true;
-            default:                           return false;
-        }
-    }
-
-    void print_rule_binary(FILE * file, const std::vector<whisper_grammar_element> & rule) {
-        for (auto elem : rule) {
-            switch (elem.type) {
-                case WHISPER_GRETYPE_END:            fprintf(file, "END");            break;
-                case WHISPER_GRETYPE_ALT:            fprintf(file, "ALT");            break;
-                case WHISPER_GRETYPE_RULE_REF:       fprintf(file, "RULE_REF");       break;
-                case WHISPER_GRETYPE_CHAR:           fprintf(file, "CHAR");           break;
-                case WHISPER_GRETYPE_CHAR_NOT:       fprintf(file, "CHAR_NOT");       break;
-                case WHISPER_GRETYPE_CHAR_RNG_UPPER: fprintf(file, "CHAR_RNG_UPPER"); break;
-                case WHISPER_GRETYPE_CHAR_ALT:       fprintf(file, "CHAR_ALT");       break;
-            }
-            switch (elem.type) {
-                case WHISPER_GRETYPE_END:
-                case WHISPER_GRETYPE_ALT:
-                case WHISPER_GRETYPE_RULE_REF:
-                    fprintf(file, "(%u) ", elem.value);
-                    break;
-                case WHISPER_GRETYPE_CHAR:
-                case WHISPER_GRETYPE_CHAR_NOT:
-                case WHISPER_GRETYPE_CHAR_RNG_UPPER:
-                case WHISPER_GRETYPE_CHAR_ALT:
-                    fprintf(file, "(\"");
-                    print_grammar_char(file, elem.value);
-                    fprintf(file, "\") ");
-                    break;
-            }
-        }
-        fprintf(file, "\n");
-    }
-
-    void print_rule(
-            FILE     * file,
-            uint32_t   rule_id,
-            const std::vector<whisper_grammar_element> & rule,
-            const std::map<uint32_t, std::string>    & symbol_id_names) {
-        if (rule.empty() || rule.back().type != WHISPER_GRETYPE_END) {
-            throw std::runtime_error(
-                "malformed rule, does not end with WHISPER_GRETYPE_END: " + std::to_string(rule_id));
-        }
-        fprintf(file, "%s ::= ", symbol_id_names.at(rule_id).c_str());
-        for (size_t i = 0, end = rule.size() - 1; i < end; i++) {
-            whisper_grammar_element elem = rule[i];
-            switch (elem.type) {
-                case WHISPER_GRETYPE_END:
-                    throw std::runtime_error(
-                        "unexpected end of rule: " + std::to_string(rule_id) + "," +
-                        std::to_string(i));
-                case WHISPER_GRETYPE_ALT:
-                    fprintf(file, "| ");
-                    break;
-                case WHISPER_GRETYPE_RULE_REF:
-                    fprintf(file, "%s ", symbol_id_names.at(elem.value).c_str());
-                    break;
-                case WHISPER_GRETYPE_CHAR:
-                    fprintf(file, "[");
-                    print_grammar_char(file, elem.value);
-                    break;
-                case WHISPER_GRETYPE_CHAR_NOT:
-                    fprintf(file, "[^");
-                    print_grammar_char(file, elem.value);
-                    break;
-                case WHISPER_GRETYPE_CHAR_RNG_UPPER:
-                    if (i == 0 || !is_char_element(rule[i - 1])) {
-                        throw std::runtime_error(
-                            "WHISPER_GRETYPE_CHAR_RNG_UPPER without preceding char: " +
-                            std::to_string(rule_id) + "," + std::to_string(i));
-                    }
-                    fprintf(file, "-");
-                    print_grammar_char(file, elem.value);
-                    break;
-                case WHISPER_GRETYPE_CHAR_ALT:
-                    if (i == 0 || !is_char_element(rule[i - 1])) {
-                        throw std::runtime_error(
-                            "WHISPER_GRETYPE_CHAR_ALT without preceding char: " +
-                            std::to_string(rule_id) + "," + std::to_string(i));
-                    }
-                    print_grammar_char(file, elem.value);
-                    break;
-            }
-            if (is_char_element(elem)) {
-                switch (rule[i + 1].type) {
-                    case WHISPER_GRETYPE_CHAR_ALT:
-                    case WHISPER_GRETYPE_CHAR_RNG_UPPER:
-                        break;
-                    default:
-                        fprintf(file, "] ");
-                }
-            }
-        }
-        fprintf(file, "\n");
-    }
-
-    void print_grammar(FILE * file, const parse_state & state) {
-        try {
-            std::map<uint32_t, std::string> symbol_id_names;
-            for (auto kv : state.symbol_ids) {
-                symbol_id_names[kv.second] = kv.first;
-            }
-            for (size_t i = 0, end = state.rules.size(); i < end; i++) {
-                // fprintf(file, "%zu: ", i);
-                // print_rule_binary(file, state.rules[i]);
-                print_rule(file, uint32_t(i), state.rules[i], symbol_id_names);
-                // fprintf(file, "\n");
-            }
-        } catch (const std::exception & err) {
-            fprintf(stderr, "\n%s: error printing grammar: %s\n", __func__, err.what());
-        }
-    }
-
-    std::vector<const whisper_grammar_element *> parse_state::c_rules() const{
-        std::vector<const whisper_grammar_element *> ret;
-        for (const auto & rule : rules) {
-            ret.push_back(rule.data());
-        }
-        return ret;
-    }
-}