feat: first commit #27
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: ci | |
| permissions: | |
| contents: read | |
| pull-requests: read | |
| actions: read | |
| env: | |
| VERSION: "${{ github.ref_name }}" | |
| on: | |
| workflow_dispatch: { } | |
| push: | |
| tags: | |
| - "v*.*.*" | |
| branches: | |
| - main | |
| paths-ignore: | |
| - "docs/**" | |
| - "**.md" | |
| - "**.mdx" | |
| - "**.png" | |
| - "**.jpg" | |
| pull_request: | |
| branches: | |
| - main | |
| paths-ignore: | |
| - "docs/**" | |
| - "**.md" | |
| - "**.mdx" | |
| - "**.png" | |
| - "**.jpg" | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }} | |
| cancel-in-progress: true | |
| jobs: | |
| darwin-metal: | |
| if: ${{ false }} | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| arch: [ amd64, arm64 ] | |
| # see https://github.com/actions/runner-images?tab=readme-ov-file#available-images, | |
| # https://support.apple.com/en-us/102894. | |
| runs-on: ${{ matrix.arch == 'amd64' && 'macos-13' || 'macos-14' }} | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v4 | |
| with: | |
| fetch-depth: 0 | |
| submodules: 'recursive' | |
| - name: Setup XCode | |
| uses: maxim-lobanov/setup-xcode@v1 | |
| with: | |
| xcode-version: '15.2' | |
| - name: Setup Cache | |
| timeout-minutes: 5 | |
| uses: actions/cache@v3 | |
| with: | |
| key: cache-darwin-metal-${{ matrix.arch }}-${{ hashFiles('**/.gitmodules') }} | |
| path: | | |
| ${{ github.workspace }}/.cache | |
| - name: Deps | |
| run: | | |
| brew update && brew install ccache | |
| - name: Build | |
| run: | | |
| echo "===== BUILD =====" | |
| mkdir -p ${{ github.workspace }}/.cache | |
| CCACHE_DIR=${{ github.workspace }}/.cache/ccache make -j LLAMA_METAL=1 LLAMA_NO_OPENMP=1 ${{ matrix.arch == 'amd64' && 'CFLAGS="-mfma -mf16c -mavx -mavx2" CXXFLAGS="-mfma -mf16c -mavx -mavx2"' || '' }} | |
| echo "===== RESULT =====" | |
| if [ -f ${{ github.workspace }}/bin/llama-box ]; then | |
| ldd ${{ github.workspace }}/bin/llama-box | |
| else | |
| exit 1 | |
| fi | |
| echo "===== PACKAGE =====" | |
| mkdir -p ${{ github.workspace }}/out | |
| zip -j ${{ github.workspace }}/out/llama-box-darwin-${{ matrix.arch }}-metal.zip ${{ github.workspace }}/bin/* | |
| - name: Upload Artifact | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| path: ${{ github.workspace }}/out/*.zip | |
| name: llama-box-darwin-${{ matrix.arch }}-metal | |
| linux-hip: | |
| if: ${{ false }} | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| arch: [ amd64 ] | |
| version: [ '6.0.2' ] | |
| runs-on: ubuntu-22.04 | |
| steps: | |
| - name: Maximize Space | |
| # see https://github.com/easimon/maximize-build-space/blob/master/action.yml. | |
| run: | | |
| sudo rm -rf /usr/share/dotnet | |
| sudo rm -rf /usr/local/lib/android | |
| sudo rm -rf /opt/ghc | |
| sudo rm -rf /opt/hostedtoolcache/CodeQL | |
| sudo docker image prune --all --force | |
| - name: Clone | |
| uses: actions/checkout@v4 | |
| with: | |
| fetch-depth: 0 | |
| submodules: 'recursive' | |
| - name: Setup Cache | |
| timeout-minutes: 5 | |
| uses: actions/cache@v3 | |
| with: | |
| key: cache-linux-hip-${{ matrix.arch }}-${{ matrix.version }}-${{ hashFiles('**/.gitmodules') }} | |
| path: | | |
| ${{ github.workspace }}/.cache | |
| - name: Setup QEMU | |
| if: ${{ matrix.arch == 'arm64' }} | |
| uses: docker/setup-qemu-action@v3 | |
| with: | |
| image: tonistiigi/binfmt:qemu-v7.0.0 | |
| platforms: "arm64" | |
| - name: Build | |
| # disable OpenMP to support static linking, | |
| # see https://github.com/ggerganov/llama.cpp/issues/7743#issuecomment-2148342691, | |
| # https://github.com/ggerganov/llama.cpp/issues/7719#issuecomment-2147631216. | |
| # build fat binary, | |
| # see https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878, | |
| # https://llvm.org/docs/AMDGPUUsage.html. | |
| env: | |
| GPU_TARGETS: "gfx803 gfx900 gfx906 gfx908 gfx90a gfx940 gfx941 gfx942 gfx1010 gfx1030 gfx1100 gfx1101 gfx1102" | |
| run: | | |
| echo "===== SCRIPT =====" | |
| cat <<EOF > /tmp/entrypoint.sh | |
| #!/bin/bash | |
| apt-get update && apt-get install -y build-essential git libgomp1 ccache | |
| git config --global --add safe.directory /workspace/llama.cpp | |
| make -j LLAMA_HIPBLAS=1 LLAMA_NO_OPENMP=1 ${{ matrix.arch == 'amd64' && 'CFLAGS="-mfma -mf16c -mavx -mavx2" CXXFLAGS="-mfma -mf16c -mavx -mavx2"' || '' }} | |
| EOF | |
| chmod +x /tmp/entrypoint.sh | |
| cat /tmp/entrypoint.sh | |
| echo "===== BUILD =====" | |
| mkdir -p ${{ github.workspace }}/.cache | |
| docker run \ | |
| --rm \ | |
| --privileged \ | |
| --platform linux/${{ matrix.arch }} \ | |
| --volume ${{ github.workspace }}:/workspace \ | |
| --volume /tmp/entrypoint.sh:/entrypoint.sh \ | |
| --entrypoint /entrypoint.sh \ | |
| --workdir /workspace \ | |
| --env CCACHE_DIR=/workspace/.cache/ccache \ | |
| --env CC=/opt/rocm/llvm/bin/clang \ | |
| --env CXX=/opt/rocm/llvm/bin/clang++ \ | |
| --env GPU_TARGETS="${{ env.GPU_TARGETS }}" \ | |
| --env AMDGPU_TARGETS="${{ env.GPU_TARGETS }}" \ | |
| rocm/dev-ubuntu-22.04:${{ matrix.version }}-complete | |
| echo "===== RESULT =====" | |
| if [ -f ${{ github.workspace }}/bin/llama-box ]; then | |
| ldd ${{ github.workspace }}/bin/llama-box | |
| else | |
| exit 1 | |
| fi | |
| echo "===== PACKAGE =====" | |
| mkdir -p ${{ github.workspace }}/out | |
| zip -j ${{ github.workspace }}/out/llama-box-linux-${{ matrix.arch }}-hip-${{ matrix.version }}.zip ${{ github.workspace }}/bin/* | |
| - name: Upload Artifact | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| path: ${{ github.workspace }}/out/*.zip | |
| name: llama-box-linux-${{ matrix.arch }}-hip-${{ matrix.version }} | |
| linux-cuda: | |
| if: ${{ false }} | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| arch: [ amd64 ] | |
| version: [ '12.2.0', '11.7.1' ] | |
| runs-on: ubuntu-22.04 | |
| steps: | |
| - name: Maximize Space | |
| # see https://github.com/easimon/maximize-build-space/blob/master/action.yml. | |
| run: | | |
| sudo rm -rf /usr/share/dotnet | |
| sudo rm -rf /usr/local/lib/android | |
| sudo rm -rf /opt/ghc | |
| sudo rm -rf /opt/hostedtoolcache/CodeQL | |
| sudo docker image prune --all --force | |
| - name: Clone | |
| uses: actions/checkout@v4 | |
| with: | |
| fetch-depth: 0 | |
| submodules: 'recursive' | |
| - name: Setup Cache | |
| timeout-minutes: 5 | |
| uses: actions/cache@v3 | |
| with: | |
| key: cache-linux-cuda-${{ matrix.arch }}-${{ matrix.version }}-${{ hashFiles('**/.gitmodules') }} | |
| path: | | |
| ${{ github.workspace }}/.cache | |
| - name: Setup QEMU | |
| if: ${{ matrix.arch == 'arm64' }} | |
| uses: docker/setup-qemu-action@v3 | |
| with: | |
| image: tonistiigi/binfmt:qemu-v7.0.0 | |
| platforms: "arm64" | |
| - name: Build | |
| # disable OpenMP to support static linking, | |
| # see https://github.com/ggerganov/llama.cpp/issues/7743#issuecomment-2148342691, | |
| # https://github.com/ggerganov/llama.cpp/issues/7719#issuecomment-2147631216. | |
| run: | | |
| echo "===== SCRIPT =====" | |
| cat <<EOF > /tmp/entrypoint.sh | |
| #!/bin/bash | |
| apt-get update && apt-get install -y build-essential git libgomp1 ccache | |
| git config --global --add safe.directory /workspace/llama.cpp | |
| make -j LLAMA_CUDA=1 LLAMA_NO_OPENMP=1 LLAMA_CUDA_FORCE_MMQ=1 ${{ matrix.arch == 'amd64' && 'CFLAGS="-mfma -mf16c -mavx -mavx2" CXXFLAGS="-mfma -mf16c -mavx -mavx2"' || '' }} | |
| EOF | |
| chmod +x /tmp/entrypoint.sh | |
| cat /tmp/entrypoint.sh | |
| echo "===== BUILD =====" | |
| mkdir -p ./.cache | |
| docker run \ | |
| --rm \ | |
| --privileged \ | |
| --platform linux/${{ matrix.arch }} \ | |
| --volume ${{ github.workspace }}:/workspace \ | |
| --workdir /workspace \ | |
| --volume /tmp/entrypoint.sh:/entrypoint.sh \ | |
| --entrypoint /entrypoint.sh \ | |
| --env CCACHE_DIR=/workspace/.cache/ccache \ | |
| --env CUDA_DOCKER_ARCH=all \ | |
| nvidia/cuda:${{ matrix.version }}-devel-ubuntu22.04 | |
| echo "===== RESULT =====" | |
| if [ -f ${{ github.workspace }}/bin/llama-box ]; then | |
| ldd ${{ github.workspace }}/bin/llama-box | |
| else | |
| exit 1 | |
| fi | |
| echo "===== PACKAGE =====" | |
| mkdir -p ${{ github.workspace }}/out | |
| zip -j ${{ github.workspace }}/out/llama-box-linux-${{ matrix.arch }}-cuda-${{ matrix.version }}.zip ${{ github.workspace }}/bin/* | |
| - name: Upload Artifact | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| path: ${{ github.workspace }}/out/*.zip | |
| name: llama-box-linux-${{ matrix.arch }}-cuda-${{ matrix.version }} | |
| windows-hip: | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| arch: [ amd64 ] | |
| # see https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html. | |
| version: [ '6.0.2' ] | |
| runs-on: windows-2022 | |
| steps: | |
| - name: Clone | |
| uses: actions/checkout@v4 | |
| with: | |
| fetch-depth: 0 | |
| submodules: 'recursive' | |
| - name: Setup Cache | |
| timeout-minutes: 5 | |
| uses: actions/cache@v3 | |
| with: | |
| key: cache-windows-hip-${{ matrix.arch }}-${{ matrix.version }}-${{ hashFiles('**/.gitmodules') }} | |
| path: | | |
| ${{ github.workspace }}/.cache | |
| - name: Deps | |
| run: | | |
| $ErrorActionPreference = "Stop" | |
| $WarningPreference = 'SilentlyContinue' | |
| $VerbosePreference = 'SilentlyContinue' | |
| $DebugPreference = 'SilentlyContinue' | |
| $ProgressPreference = 'SilentlyContinue' | |
| Write-Host "I install make" | |
| choco install make | |
| Write-Host "I install ccache" | |
| choco install make ccache | |
| Write-Host "I install AMD HIP SDK" | |
| Invoke-WebRequest -Uri "https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-${{ startsWith(matrix.version, '6') && '23.Q4' || '23.Q3' }}-WinSvr2022-For-HIP.exe" -OutFile "${env:RUNNER_TEMP}\rocm-install.exe" | |
| Start-Process "${env:RUNNER_TEMP}\rocm-install.exe" -ArgumentList '-install' -NoNewWindow -Wait | |
| Write-Host "I configure the PATH environment variable" | |
| $path = $(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | Split-Path | Split-Path) | |
| $newMachinePath = $path + ";" + [System.Environment]::GetEnvironmentVariable("Path","Machine") | |
| [Environment]::SetEnvironmentVariable("Path", $newMachinePath, [System.EnvironmentVariableTarget]::Machine) | |
| $env:Path = $path + ";" + $env:Path | |
| Write-Host "I verify Clang" | |
| clang.exe --version | |
| - name: Build | |
| env: | |
| GPU_TARGETS: "gfx803 gfx900 gfx906 gfx908 gfx90a gfx940 gfx941 gfx942 gfx1010 gfx1030 gfx1100 gfx1101 gfx1102" | |
| run: | | |
| Write-Host "===== BUILD =====" | |
| New-Item -Force -ItemType Directory -Path "${{ github.workspace }}/.cache" -ErrorAction Ignore | Out-Null | |
| $env:CCACHE_DIR = "${{ github.workspace }}/.cache/ccache" | |
| make -j LLAMA_HIPBLAS=1 LLAMA_NO_OPENMP=1 ${{ matrix.arch == 'amd64' && 'CFLAGS="-mfma -mf16c -mavx -mavx2"' || '' }} | |
| Write-Host "===== RESULT =====" | |
| if (Test-Path -Path "${{ github.workspace }}/bin/llama-box.exe") { | |
| llvm-objdump.exe -p "${{ github.workspace }}/bin/llama-box.exe" | |
| } else { | |
| exit 1 | |
| } | |
| release: | |
| if: ${{ startsWith(github.ref, 'refs/tags/') }} | |
| permissions: | |
| contents: write | |
| actions: read | |
| id-token: write | |
| runs-on: ubuntu-22.04 | |
| needs: | |
| - darwin-metal | |
| - linux-hip | |
| - linux-cuda | |
| steps: | |
| - name: Download Artifact | |
| uses: actions/download-artifact@v4 | |
| with: | |
| path: ${{ github.workspace }}/out | |
| merge-multiple: true | |
| - name: Release | |
| uses: softprops/action-gh-release@v1 | |
| with: | |
| fail_on_unmatched_files: true | |
| tag_name: "${{ env.VERSION }}" | |
| prerelease: ${{ contains(github.ref, 'rc') }} | |
| files: ${{ github.workspace }}/out/* |