compile.yml

name: Update Binaries

on:
  workflow_dispatch:
    inputs:
      llama_cpp_commit:
        description: 'Branch, tag, or commit to use for llama.cpp'
        required: true
        default: 'master'
  push:
    branches: [cron_job]
  #schedule:
  #  - cron: "22 22 * * 2"
concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.event.inputs.llama_cpp_commit }}
  cancel-in-progress: true

env:
  # Compiler defines common to all platforms
  COMMON_DEFINE: -DGGML_NATIVE=OFF -DLLAMA_BUILD_TESTS=OFF -DLLAMA_BUILD_EXAMPLES=ON -DLLAMA_BUILD_SERVER=OFF -DBUILD_SHARED_LIBS=ON -DGGML_RPC=ON

jobs:
  compile-linux:
    name: Compile (Linux)
    strategy:
      fail-fast: true
      matrix:
        include:
          - build: 'noavx'
            defines: '-DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF'
          - build: 'avx2'
            defines: ''
          - build: 'avx'
            defines: '-DGGML_AVX2=OFF'
          - build: 'avx512'
            defines: '-DGGML_AVX512=ON'
    runs-on: ubuntu-20.04
    steps:
      - uses: actions/checkout@v4
        with:
          repository: ggerganov/llama.cpp
          fetch-depth: 0
          ref: '${{ github.event.inputs.llama_cpp_commit }}'
      - name: Build
        id: cmake_build
        run: |
          mkdir build
          cd build
          cmake .. ${{ env.COMMON_DEFINE }} ${{ matrix.defines }}
          cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
          ls -R
      - uses: actions/upload-artifact@v4
        with:
          path: ./build/src/libllama.so
          name: llama-bin-linux-${{ matrix.build }}-x64.so
          if-no-files-found: error
      - uses: actions/upload-artifact@v4
        with:
          path: ./build/ggml/src/libggml.so
          name: ggml-bin-linux-${{ matrix.build }}-x64.so
          if-no-files-found: error
      - name: Upload Llava
        uses: actions/upload-artifact@v4
        with:
          path: ./build/examples/llava/libllava_shared.so
          name: llava-bin-linux-${{ matrix.build }}-x64.so
          if-no-files-found: error

  compile-windows:
    name: Compile (Windows)
    strategy:
      fail-fast: true
      matrix:
        include:
          - build: 'noavx'
            defines: '-DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF'
          - build: 'avx2'
            defines: ''
          - build: 'avx'
            defines: '-DGGML_AVX2=OFF'
          - build: 'avx512'
            defines: '-DGGML_AVX512=ON -DGGML_AVX512_VBMI=ON -DGGML_AVX512_VNNI=ON'
    runs-on: windows-latest
    steps:
      - uses: actions/checkout@v4
        with:
          repository: ggerganov/llama.cpp
          fetch-depth: 0
          ref: '${{ github.event.inputs.llama_cpp_commit }}'

      - name: Build
        id: cmake_build
        run: |
          mkdir build
          cd build
          cmake .. ${{ env.COMMON_DEFINE }} ${{ matrix.defines }}
          cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
          tree /f

      - name: Upload artifacts (llama)
        uses: actions/upload-artifact@v4
        with:
          path: .\build\bin\Release\llama.dll
          name: llama-bin-win-${{ matrix.build }}-x64.dll
          if-no-files-found: error
      - name: Upload artifacts (ggml)
        uses: actions/upload-artifact@v4
        with:
          path: .\build\bin\Release\ggml.dll
          name: ggml-bin-win-${{ matrix.build }}-x64.dll
          if-no-files-found: error

      - name: Upload artifacts (llava)
        uses: actions/upload-artifact@v4
        with:
          path: .\build\bin\Release\llava_shared.dll
          name: llava-bin-win-${{ matrix.build }}-x64.dll
          if-no-files-found: error
          
  compile-vulkan:
      name: Compile (vulkan) - ${{ matrix.os }}
      strategy:
          fail-fast: false
          matrix:
              os: [
                  ubuntu-22.04,
                  windows-latest
              ]
      env:
          VULKAN_VERSION: 1.3.261.1
      runs-on: ${{ matrix.os }}
      steps:
          - name: Clone
            id: checkout
            uses: actions/checkout@v4
            with:
                repository: ggerganov/llama.cpp
                fetch-depth: 0
                ref: '${{ github.event.inputs.llama_cpp_commit }}'

          - name: Download dependencies - Linux
            if: ${{ matrix.os == 'ubuntu-22.04' }}
            run: |
                wget -qO- https://packages.lunarg.com/lunarg-signing-key-pub.asc | sudo tee /etc/apt/trusted.gpg.d/lunarg.asc
                sudo wget -qO /etc/apt/sources.list.d/lunarg-vulkan-jammy.list http://packages.lunarg.com/vulkan/lunarg-vulkan-jammy.list
                sudo apt update
                sudo apt install vulkan-sdk
          - name: Download dependencies - Windows
            id: get_vulkan
            if: ${{ matrix.os == 'windows-latest' }}
            run: |
                curl.exe -o $env:RUNNER_TEMP/VulkanSDK-Installer.exe -L "https://sdk.lunarg.com/sdk/download/${env:VULKAN_VERSION}/windows/VulkanSDK-${env:VULKAN_VERSION}-Installer.exe"
                & "$env:RUNNER_TEMP\VulkanSDK-Installer.exe" --accept-licenses --default-answer --confirm-command install
                Add-Content $env:GITHUB_ENV "VULKAN_SDK=C:\VulkanSDK\${env:VULKAN_VERSION}"
                Add-Content $env:GITHUB_PATH "C:\VulkanSDK\${env:VULKAN_VERSION}\bin"
          - name: Build
            id: cmake_build
            if: ${{ matrix.os == 'windows-latest' }}
            run: |
                mkdir build
                cd build
                cmake .. ${{ env.COMMON_DEFINE }} -DGGML_VULKAN=ON
                cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
                ls -R
          - name: Build
            if: ${{ matrix.os == 'ubuntu-22.04' }}
            run: |
                mkdir build
                cd build
                cmake .. ${{ env.COMMON_DEFINE }} -DGGML_VULKAN=ON
                cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
                ls -R
          - name: Upload llama artifacts (Windows)
            if: ${{ matrix.os == 'windows-latest' }}
            uses: actions/upload-artifact@v4
            with:
                path: .\build\bin\Release\llama.dll
                name: llama-bin-win-vulkan-x64.dll
                if-no-files-found: error
          - name: Upload llama artifacts (Windows)
            if: ${{ matrix.os == 'windows-latest' }}
            uses: actions/upload-artifact@v4
            with:
                path: .\build\bin\Release\ggml.dll
                name: ggml-bin-win-vulkan-x64.dll
                if-no-files-found: error
          - name: Upload llava artifacts (Windows)
            if: ${{ matrix.os == 'windows-latest' }}
            uses: actions/upload-artifact@v4
            with:
                path: .\build\bin\Release\llava_shared.dll
                name: llava-bin-win-vulkan-x64.dll
                if-no-files-found: error
          - name: Upload llama artifacts (Linux)
            if: ${{ matrix.os == 'ubuntu-22.04' }}
            uses: actions/upload-artifact@v4
            with:
                path: ./build/src/libllama.so
                name: llama-bin-linux-vulkan-x64.so
                if-no-files-found: error
          - name: Upload ggml artifacts (Linux)
            if: ${{ matrix.os == 'ubuntu-22.04' }}
            uses: actions/upload-artifact@v4
            with:
              path: ./build/ggml/src/libggml.so
              name: ggml-bin-linux-vulkan-x64.so
              if-no-files-found: error
          - name: Upload llava artifacts (Linux)
            if: ${{ matrix.os == 'ubuntu-22.04' }}
            uses: actions/upload-artifact@v4
            with:
                path: ./build/examples/llava/libllava_shared.so
                name: llava-bin-linux-vulkan-x64.so
                if-no-files-found: error
                
  compile-cublas:
    name: Compile (cublas)
    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-20.04, windows-2019]
        cuda: ['12.2.0', '11.7.1']
    runs-on: ${{ matrix.os }}
    steps:
      - name: Clone
        id: checkout
        uses: actions/checkout@v4
        with:
          repository: ggerganov/llama.cpp
          fetch-depth: 0
          ref: '${{ github.event.inputs.llama_cpp_commit }}'

      - uses: Jimver/cuda-toolkit@v0.2.15
        if: runner.os == 'Windows'
        id: cuda-toolkit-windows
        with:
          cuda: ${{ matrix.cuda }}
          method: 'network'
          sub-packages: '["nvcc", "cudart", "cublas", "cublas_dev", "thrust", "visual_studio_integration"]'

      - uses: Jimver/cuda-toolkit@v0.2.15
        if: runner.os == 'Linux'
        id: cuda-toolkit-linux
        with:
          cuda: ${{ matrix.cuda }}
          method: 'network'
          linux-local-args: '["--toolkit"]'

      - name: Build
        id: cmake_build
        run: |
          mkdir build
          cd build
          cmake .. ${{ env.COMMON_DEFINE }} -DGGML_CUDA=ON
          cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
          ls -R

      - name: Upload artifacts (Windows)
        if: ${{ matrix.os == 'windows-2019' }}
        uses: actions/upload-artifact@v4
        with:
          path: .\build\bin\Release\llama.dll
          name: llama-bin-win-cublas-cu${{ matrix.cuda }}-x64.dll
          if-no-files-found: error
      - name: Upload artifacts (ggml)
        if: ${{ matrix.os == 'windows-2019' }}
        uses: actions/upload-artifact@v4
        with:
          path: .\build\bin\Release\ggml.dll
          name: ggml-bin-win-cublas-cu${{ matrix.cuda }}-x64.dll
          if-no-files-found: error
      - name: Upload llava artifacts (Windows)
        if: ${{ matrix.os == 'windows-2019' }}
        uses: actions/upload-artifact@v4
        with:
          path: .\build\bin\Release\llava_shared.dll
          name: llava-bin-win-cublas-cu${{ matrix.cuda }}-x64.dll
          if-no-files-found: error
      - name: Upload artifacts (Linux)
        if: ${{ matrix.os == 'ubuntu-20.04' }}
        uses: actions/upload-artifact@v4
        with:
          path: ./build/src/libllama.so
          name: llama-bin-linux-cublas-cu${{ matrix.cuda }}-x64.so
          if-no-files-found: error
      - name: Upload artifacts ggml (Linux)
        if: ${{ matrix.os == 'ubuntu-20.04' }}
        uses: actions/upload-artifact@v4
        with:
          path: ./build/ggml/src/libggml.so
          name: ggml-bin-linux-cublas-cu${{ matrix.cuda }}-x64.so
          if-no-files-found: error
      - name: Upload llava artifacts (Linux)
        if: ${{ matrix.os == 'ubuntu-20.04' }}
        uses: actions/upload-artifact@v4
        with:
          path: ./build/examples/llava/libllava_shared.so
          name: llava-bin-linux-cublas-cu${{ matrix.cuda }}-x64.so
          if-no-files-found: error
 
  compile-macos:
    name: Compile (MacOS)
    strategy:
      fail-fast: true
      matrix:
        include:
          - build: 'arm64'
            defines: '-DCMAKE_OSX_ARCHITECTURES=arm64 -DGGML_METAL_EMBED_LIBRARY=ON'
          - build: 'x64'
            defines: '-DCMAKE_OSX_ARCHITECTURES=x86_64 -DGGML_METAL=OFF -DGGML_AVX=ON -DGGML_AVX2=ON'
          - build: 'x64-rosetta2'
            defines: '-DCMAKE_OSX_ARCHITECTURES=x86_64 -DGGML_METAL=OFF -DGGML_AVX=OFF -DGGML_AVX2=OFF'
    runs-on: macos-latest   
    steps:
      - uses: actions/checkout@v4
        with:
          repository: ggerganov/llama.cpp
          fetch-depth: 0
          ref: '${{ github.event.inputs.llama_cpp_commit }}'
      - name: Dependencies
        continue-on-error: true
        run: |
          brew update
      - name: Build
        id: cmake_build
        run: |
          mkdir build
          cd build
          cmake .. ${{ env.COMMON_DEFINE }} ${{ matrix.defines }}
          cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
      - name: Upload ggml
        uses: actions/upload-artifact@v4
        with:
            path: ./build/ggml/src/libggml.dylib
            name: ggml-bin-osx-${{ matrix.build }}.dylib
            if-no-files-found: error
      - name: Upload llama
        uses: actions/upload-artifact@v4
        with:
          path: ./build/src/libllama.dylib
          name: llama-bin-osx-${{ matrix.build }}.dylib
          if-no-files-found: error
      - name: Upload Llava
        uses: actions/upload-artifact@v4
        with:
          path: ./build/examples/llava/libllava_shared.dylib
          name: llava-bin-osx-${{ matrix.build }}.dylib
          if-no-files-found: error
      - name: Upload Metal
        if: ${{ matrix.build == 'arm64' }}
        uses: actions/upload-artifact@v4
        with:
          path: ./build/bin/ggml-metal.metal
          name: ggml-metal.metal
          if-no-files-found: error
  
  compile-android:
    strategy:
      fail-fast: true
      matrix:
        include:
          - build: 'x86'
            defines: '-DANDROID_ABI=x86'
          - build: 'x86_64'
            defines: '-DANDROID_ABI=x86_64'
          - build: 'arm64-v8a'
            defines: '-DANDROID_ABI=arm64-v8a'
    runs-on: ubuntu-20.04
    steps:
      - uses: actions/checkout@v4
        with:
          repository: ggerganov/llama.cpp
          fetch-depth: 0
          ref: '${{ github.event.inputs.llama_cpp_commit }}'
      - uses: nttld/setup-ndk@v1
        id: setup-ndk
        with:
          ndk-version: r26d
          add-to-path: false
      - name: Build
        id: cmake_build
        env:
          CMAKE_FLAGS: '-DCMAKE_TOOLCHAIN_FILE=${{ steps.setup-ndk.outputs.ndk-path }}/build/cmake/android.toolchain.cmake -DANDROID_PLATFORM=android-23'
        run: |
          mkdir build
          cd build
          cmake .. ${{ env.COMMON_DEFINE }} ${{ env.CMAKE_FLAGS }} ${{ matrix.defines }}
          cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
          cd ..
          ls -R
      - name: Upload Llama
        uses: actions/upload-artifact@v4
        with:
          path: ./build/src/libllama.so
          name: llama-bin-android-${{ matrix.build }}.so
      - uses: actions/upload-artifact@v4
        with:
          path: ./build/ggml/src/libggml.so
          name: ggml-bin-android-${{ matrix.build }}.so
          if-no-files-found: error
      - name: Upload Llava
        uses: actions/upload-artifact@v4
        with:
          path: ./build/examples/llava/libllava_shared.so
          name: llava-bin-android-${{ matrix.build }}.so

  build-deps:
    runs-on: ubuntu-latest
    name: "Gather Binaries"
    if: ${{ always() }}
    needs: [
      "compile-linux",
      "compile-windows",
      "compile-vulkan",
      "compile-cublas",
      "compile-macos",
      "compile-android"
    ]
    steps:      
      - uses: actions/download-artifact@v4
        with:
          path: artifacts
      - name: List Files
        run: ls -R
      - name: Rearrange Files
        run: |
          # Make all directories at once
          mkdir --parents deps/{avx,avx2,avx512,osx-arm64,osx-x64,osx-x64-rosetta2,cu11.7.1,cu12.2.0,vulkan,android-arm64-v8a,android-x86,android-x86_64}

          # Linux
          cp artifacts/ggml-bin-linux-noavx-x64.so/libggml.so           deps/libggml.so
          cp artifacts/llama-bin-linux-noavx-x64.so/libllama.so         deps/libllama.so
          cp artifacts/llava-bin-linux-noavx-x64.so/libllava_shared.so  deps/libllava_shared.so

          cp artifacts/ggml-bin-linux-avx-x64.so/libggml.so             deps/avx/libggml.so
          cp artifacts/llama-bin-linux-avx-x64.so/libllama.so           deps/avx/libllama.so
          cp artifacts/llava-bin-linux-avx-x64.so/libllava_shared.so    deps/avx/libllava_shared.so

          cp artifacts/ggml-bin-linux-avx2-x64.so/libggml.so            deps/avx2/libggml.so
          cp artifacts/llama-bin-linux-avx2-x64.so/libllama.so          deps/avx2/libllama.so
          cp artifacts/llava-bin-linux-avx2-x64.so/libllava_shared.so   deps/avx2/libllava_shared.so

          cp artifacts/ggml-bin-linux-avx512-x64.so/libggml.so          deps/avx512/libggml.so
          cp artifacts/llama-bin-linux-avx512-x64.so/libllama.so        deps/avx512/libllama.so
          cp artifacts/llava-bin-linux-avx512-x64.so/libllava_shared.so deps/avx512/libllava_shared.so

          # Windows
          cp artifacts/ggml-bin-win-noavx-x64.dll/ggml.dll            deps/ggml.dll
          cp artifacts/llama-bin-win-noavx-x64.dll/llama.dll          deps/llama.dll
          cp artifacts/llava-bin-win-noavx-x64.dll/llava_shared.dll   deps/llava_shared.dll

          cp artifacts/ggml-bin-win-avx-x64.dll/ggml.dll              deps/avx/ggml.dll
          cp artifacts/llama-bin-win-avx-x64.dll/llama.dll            deps/avx/llama.dll
          cp artifacts/llava-bin-win-avx-x64.dll/llava_shared.dll     deps/avx/llava_shared.dll

          cp artifacts/ggml-bin-win-avx2-x64.dll/ggml.dll             deps/avx2/ggml.dll
          cp artifacts/llama-bin-win-avx2-x64.dll/llama.dll           deps/avx2/llama.dll
          cp artifacts/llava-bin-win-avx2-x64.dll/llava_shared.dll    deps/avx2/llava_shared.dll

          cp artifacts/ggml-bin-win-avx512-x64.dll/ggml.dll           deps/avx512/ggml.dll
          cp artifacts/llama-bin-win-avx512-x64.dll/llama.dll         deps/avx512/llama.dll
          cp artifacts/llava-bin-win-avx512-x64.dll/llava_shared.dll  deps/avx512/llava_shared.dll

          # MacOS
          cp artifacts/ggml-bin-osx-arm64.dylib/libggml.dylib           deps/osx-arm64/libggml.dylib
          cp artifacts/llama-bin-osx-arm64.dylib/libllama.dylib         deps/osx-arm64/libllama.dylib
          cp artifacts/llava-bin-osx-arm64.dylib/libllava_shared.dylib  deps/osx-arm64/libllava_shared.dylib
          cp artifacts/ggml-metal.metal/ggml-metal.metal                deps/osx-arm64/ggml-metal.metal
            
          cp artifacts/ggml-bin-osx-x64.dylib/libggml.dylib             deps/osx-x64/libggml.dylib
          cp artifacts/llama-bin-osx-x64.dylib/libllama.dylib           deps/osx-x64/libllama.dylib
          cp artifacts/llava-bin-osx-x64.dylib/libllava_shared.dylib    deps/osx-x64/libllava_shared.dylib
            
          cp artifacts/ggml-bin-osx-x64-rosetta2.dylib/libggml.dylib          deps/osx-x64-rosetta2/libggml.dylib
          cp artifacts/llama-bin-osx-x64-rosetta2.dylib/libllama.dylib        deps/osx-x64-rosetta2/libllama.dylib
          cp artifacts/llava-bin-osx-x64-rosetta2.dylib/libllava_shared.dylib deps/osx-x64-rosetta2/libllava_shared.dylib

          # Android
          cp artifacts/ggml-bin-android-arm64-v8a.so/libggml.so           deps/android-arm64-v8a/libggml.so
          cp artifacts/llama-bin-android-arm64-v8a.so/libllama.so         deps/android-arm64-v8a/libllama.so
          cp artifacts/llava-bin-android-arm64-v8a.so/libllava_shared.so  deps/android-arm64-v8a/libllava_shared.so

          cp artifacts/ggml-bin-android-x86.so/libggml.so                 deps/android-x86/libggml.so
          cp artifacts/llama-bin-android-x86.so/libllama.so               deps/android-x86/libllama.so
          cp artifacts/llava-bin-android-x86.so/libllava_shared.so        deps/android-x86/libllava_shared.so

          cp artifacts/ggml-bin-android-x86_64.so/libggml.so              deps/android-x86_64/libggml.so
          cp artifacts/llama-bin-android-x86_64.so/libllama.so            deps/android-x86_64/libllama.so
          cp artifacts/llava-bin-android-x86_64.so/libllava_shared.so     deps/android-x86_64/libllava_shared.so

          # Windows CUDA
          cp artifacts/ggml-bin-win-cublas-cu11.7.1-x64.dll/ggml.dll            deps/cu11.7.1/ggml.dll
          cp artifacts/llama-bin-win-cublas-cu11.7.1-x64.dll/llama.dll          deps/cu11.7.1/llama.dll
          cp artifacts/llava-bin-win-cublas-cu11.7.1-x64.dll/llava_shared.dll   deps/cu11.7.1/llava_shared.dll
          
          cp artifacts/ggml-bin-win-cublas-cu12.2.0-x64.dll/ggml.dll            deps/cu12.2.0/ggml.dll
          cp artifacts/llama-bin-win-cublas-cu12.2.0-x64.dll/llama.dll          deps/cu12.2.0/llama.dll
          cp artifacts/llava-bin-win-cublas-cu12.2.0-x64.dll/llava_shared.dll   deps/cu12.2.0/llava_shared.dll

          # Linux CUDA
          cp artifacts/ggml-bin-linux-cublas-cu11.7.1-x64.so/libggml.so           deps/cu11.7.1/libggml.so
          cp artifacts/llama-bin-linux-cublas-cu11.7.1-x64.so/libllama.so         deps/cu11.7.1/libllama.so
          cp artifacts/llava-bin-linux-cublas-cu11.7.1-x64.so/libllava_shared.so  deps/cu11.7.1/libllava_shared.so

          cp artifacts/ggml-bin-linux-cublas-cu12.2.0-x64.so/libggml.so           deps/cu12.2.0/libggml.so
          cp artifacts/llama-bin-linux-cublas-cu12.2.0-x64.so/libllama.so         deps/cu12.2.0/libllama.so
          cp artifacts/llava-bin-linux-cublas-cu12.2.0-x64.so/libllava_shared.so  deps/cu12.2.0/libllava_shared.so

          # Windows Vulkan
          cp artifacts/ggml-bin-win-vulkan-x64.dll/ggml.dll           deps/vulkan/ggml.dll
          cp artifacts/llama-bin-win-vulkan-x64.dll/llama.dll         deps/vulkan/llama.dll
          cp artifacts/llava-bin-win-vulkan-x64.dll/llava_shared.dll  deps/vulkan/llava_shared.dll
            
          # Linux Vulkan
          cp artifacts/ggml-bin-linux-vulkan-x64.so/libggml.so           deps/vulkan/libggml.so
          cp artifacts/llama-bin-linux-vulkan-x64.so/libllama.so         deps/vulkan/libllama.so
          cp artifacts/llava-bin-linux-vulkan-x64.so/libllava_shared.so  deps/vulkan/libllava_shared.so

      - name: Upload artifacts
        uses: actions/upload-artifact@v4
        with:
          path: deps/
          name: deps


      - name: Remove Artifacts
        uses: geekyeggo/delete-artifact@v5
        with:
          name: |
            llama-*
            llava-*
            *.metal
            ggml-*