Upload 0.1.72 workflow

Author: jllllll

.github/workflows/build-all-wheels.yml

@@ -34,3 +34,8 @@ jobs:
     name: Build Wheels 0.1.71
     needs: run_workflow_0-1-70
     uses: ./.github/workflows/build-wheels-0.1.71.yml
+    
+  run_workflow_0-1-72:
+    name: Build Wheels 0.1.72
+    needs: run_workflow_0-1-71
+    uses: ./.github/workflows/build-wheels-0.1.72.yml

.github/workflows/build-wheels-0.1.72.yml

@@ -0,0 +1,88 @@
+name: Build Wheels 0.1.72
+
+on:
+  workflow_dispatch:
+  workflow_call:
+
+permissions:
+  contents: write
+
+jobs:
+  build_wheels:
+    name: ${{ matrix.os }} ${{ matrix.pyver }} ${{ matrix.cuda }} ${{ matrix.releasetag == 'wheels' && 'AVX2' || matrix.releasetag }}
+    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", "12.1.1", "12.2.0"]
+        releasetag: ["AVX","wheels","AVX512"]
+    defaults:
+      run:
+        shell: pwsh
+    env:
+      CUDAVER: ${{ matrix.cuda }}
+      AVXVER: ${{ matrix.releasetag }}
+
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          repository: 'abetlen/llama-cpp-python'
+          ref: 'v0.1.72'
+          submodules: 'recursive'
+          
+      - uses: actions/setup-python@v3
+        with:
+          python-version: ${{ matrix.pyver }}
+        
+      - name: Setup Mamba
+        uses: conda-incubator/setup-miniconda@v2.2.0
+        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: |
+          $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'
+          $env:CMAKE_ARGS = '-DLLAMA_CUBLAS=on -DCMAKE_CUDA_ARCHITECTURES=35-real;37-real;52;61-real;70-real;72-real;75-real;80-real;86-real;89-real;90'
+          if ([version]$env:CUDAVER -ge [version]'12.0') {$env:CMAKE_ARGS = '-DLLAMA_CUBLAS=on -DCMAKE_CUDA_ARCHITECTURES=52;61-real;70-real;72-real;75-real;80-real;86-real;89-real;90'}
+          if ([version]$env:CUDAVER -lt [version]'11.8') {$env:CMAKE_ARGS = '-DLLAMA_CUBLAS=on -DCMAKE_CUDA_ARCHITECTURES=35-real;37-real;52;61-real;70-real;72-real;75-real;80-real;86'}
+          if ($env:AVXVER -eq 'AVX') {$env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DLLAMA_AVX2=off'}
+          if ($env:AVXVER -eq 'AVX512') {$env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DLLAMA_AVX512=on'}
+          python -m build --wheel -C--build-option=egg_info "-C--build-option=--tag-build=+cu$cudaVersion"
+          
+      - name: Upload files to a GitHub release
+        id: upload-release
+        uses: svenstaro/upload-release-action@2.6.1
+        continue-on-error: true
+        with:
+          file: ./dist/*.whl
+          tag: ${{ matrix.releasetag }}
+          file_glob: true
+          make_latest: false
+          overwrite: true
+        
+      - uses: actions/upload-artifact@v3
+        if: steps.upload-release.outcome == 'failure'
+        with:
+          name: ${{ matrix.releasetag == 'wheels' && 'AVX2' || matrix.releasetag }}
+          path: ./dist/*.whl