Upload AVX512 workflow

.github/workflows/build-wheels-0.1.66-62-AVX512.yml

@@ -0,0 +1,78 @@
+name: Build Wheels 0.1.66/62 AVX512
+
+on: workflow_dispatch
+
+permissions:
+ contents: write
+
+jobs:
+ build_wheels:
+ name: Build wheels for ${{ matrix.os }}
+ runs-on: ${{ matrix.os }}
+ strategy:
+ matrix:
+ os: [ubuntu-20.04, windows-latest]
+ pyver: ["3.7", "3.8", "3.9", "3.10", "3.11"]
+ cuda: ["11.6.2", "11.7.1", "11.8.0", "12.0.1", "12.1.0"]
+ llamacpp: ['v0.1.66','v0.1.62']
+ defaults:
+ run:
+ shell: pwsh
+ env:
+ CUDAVER: ${{ matrix.cuda }}
+ LLAMAVER: ${{ matrix.llamacpp }}
+
+ steps:
+ - uses: actions/checkout@v3
+ - uses: actions/checkout@v3
+ with:
+ repository: 'abetlen/llama-cpp-python'
+ ref: ${{ matrix.llamacpp }}
+ path: 'llama-cpp-python'
+ submodules: 'recursive'
+
+ - uses: actions/setup-python@v3
+ with:
+ python-version: ${{ matrix.pyver }}
+
+ - name: Setup Mamba
+ uses: conda-incubator/[email protected]
+ with:
+ activate-environment: "build"
+ python-version: ${{ matrix.pyver }}
+ miniforge-variant: Mambaforge
+ miniforge-version: latest
+ use-mamba: true
+ add-pip-as-python-dependency: true
+ auto-activate-base: false
+
+ - name: Install Dependencies
+ run: |
+ $cudaVersion = $env:CUDAVER
+ $cudaChannels = ''
+ $cudaNum = [int]$cudaVersion.substring($cudaVersion.LastIndexOf('.')+1)
+ while ($cudaNum -ge 0) { $cudaChannels += '-c nvidia/label/cuda-' + $cudaVersion.Remove($cudaVersion.LastIndexOf('.')+1) + $cudaNum + ' '; $cudaNum-- }
+ mamba install -y 'cuda' $cudaChannels.TrimEnd().Split()
+ python -m pip install build wheel
+ 
+ - name: Build Wheel
+ run: |
+ Set-Location 'llama-cpp-python'
+ $cudaVersion = $env:CUDAVER.Remove($env:CUDAVER.LastIndexOf('.')).Replace('.','')
+ $env:CUDA_PATH = $env:CONDA_PREFIX
+ $env:CUDA_HOME = $env:CONDA_PREFIX
+ if ($IsLinux) {$env:LD_LIBRARY_PATH = $env:CONDA_PREFIX + '/lib:' + $env:LD_LIBRARY_PATH}
+ $env:VERBOSE = '1'
+ $env:FORCE_CMAKE = '1'
+ $env:CMAKE_ARGS = '-DLLAMA_CUBLAS=on -DLLAMA_AVX512=on'
+ if ($env:LLAMAVER -eq 'v0.1.66') {$env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DCMAKE_CUDA_ARCHITECTURES=all'}
+ if ($env:LLAMAVER -eq 'v0.1.62') {$env:CUDAFLAGS = '-arch=all'}
+ python -m build --wheel -C--build-option=egg_info "-C--build-option=--tag-build=+cu$cudaVersion"
+ 
+ - name: Upload files to a GitHub release
+ uses: svenstaro/[email protected]
+ with:
+ file: ./llama-cpp-python/dist/*.whl
+ tag: AVX512
+ file_glob: true
+ overwrite: true