wheels.yml

name: qwentts-cpp-python wheels

"on":
  workflow_dispatch:
    inputs:
      qwentts_ref:
        description: qwentts.cpp commit, tag, or branch to build
        required: true
        default: 6fae92914045cd83364d2845ceaa0f7969727319

jobs:
  macos-metal:
    uses: ./.github/workflows/metal-wheel.yml
    with:
      qwentts_ref: ${{ inputs.qwentts_ref || '6fae92914045cd83364d2845ceaa0f7969727319' }}

  linux-cuda:
    name: Linux ${{ matrix.arch }} CUDA ${{ matrix.cuda_version }} wheel
    runs-on: ${{ matrix.runner }}
    strategy:
      fail-fast: false
      matrix:
        include:
          - arch: x86_64
            runner: ubuntu-22.04
            cuda_repo_distro: ubuntu2204
            cuda_repo_arch: x86_64
            cuda_version: "12.4"
            cuda_package_suffix: "12-4"
            cuda_home: /usr/local/cuda-12.4
            wheel_build_tag: 1cu124
            cuda_architectures: 75-virtual;80-real;86-real;90-real
            manylinux_platform: manylinux_2_35_x86_64
            build_jobs: "2"
          - arch: aarch64
            runner: ubuntu-22.04-arm
            cuda_repo_distro: ubuntu2204
            cuda_repo_arch: sbsa
            cuda_version: "12.4"
            cuda_package_suffix: "12-4"
            cuda_home: /usr/local/cuda-12.4
            wheel_build_tag: 1cu124
            cuda_architectures: 75-virtual;80-real;86-real;90-real
            manylinux_platform: manylinux_2_35_aarch64
            build_jobs: "2"
          - arch: x86_64
            runner: ubuntu-24.04
            cuda_repo_distro: ubuntu2404
            cuda_repo_arch: x86_64
            cuda_version: "12.8"
            cuda_package_suffix: "12-8"
            cuda_home: /usr/local/cuda-12.8
            wheel_build_tag: 1cu128
            cuda_architectures: 75-virtual;86-real;90-real;120-real;120-virtual
            manylinux_platform: manylinux_2_39_x86_64
            build_jobs: "2"
          - arch: aarch64
            runner: ubuntu-24.04-arm
            cuda_repo_distro: ubuntu2404
            cuda_repo_arch: sbsa
            cuda_version: "12.8"
            cuda_package_suffix: "12-8"
            cuda_home: /usr/local/cuda-12.8
            wheel_build_tag: 1cu128
            cuda_architectures: 75-virtual;86-real;90-real;120-real;120-virtual
            manylinux_platform: manylinux_2_39_aarch64
            build_jobs: "2"
          - arch: x86_64
            runner: ubuntu-24.04
            cuda_repo_distro: ubuntu2404
            cuda_repo_arch: x86_64
            cuda_version: "13.0"
            cuda_package_suffix: "13-0"
            cuda_home: /usr/local/cuda-13.0
            wheel_build_tag: 1cu130
            cuda_architectures: 75-virtual;80-real;86-real;90-real;121-real
            manylinux_platform: manylinux_2_39_x86_64
            build_jobs: "2"
          - arch: aarch64
            runner: ubuntu-24.04-arm
            cuda_repo_distro: ubuntu2404
            cuda_repo_arch: sbsa
            cuda_version: "13.0"
            cuda_package_suffix: "13-0"
            cuda_home: /usr/local/cuda-13.0
            wheel_build_tag: 1cu130
            cuda_architectures: 75-virtual;80-real;86-real;90-real;121-real
            manylinux_platform: manylinux_2_39_aarch64
            build_jobs: "2"
    env:
      QWENTTS_CPP_BACKEND: cuda
      QWENTTS_CPP_BUILD_JOBS: ${{ matrix.build_jobs }}
      QWENTTS_CPP_WHEEL_BUILD_TAG: ${{ matrix.wheel_build_tag }}
      QWENTTS_REF: ${{ inputs.qwentts_ref || '6fae92914045cd83364d2845ceaa0f7969727319' }}
      CUDA_ARCHITECTURES: ${{ matrix.cuda_architectures }}
    steps:
      - uses: actions/checkout@v4

      - uses: actions/setup-python@v5
        with:
          python-version: "3.12"

      - name: Free disk space
        run: |
          sudo rm -rf /usr/local/lib/android /opt/ghc /usr/share/dotnet
          sudo apt-get clean

      - name: Install build tools
        run: |
          python -m pip install --upgrade pip auditwheel build wheel
          sudo apt-get update
          sudo apt-get install -y --no-install-recommends cmake ninja-build patchelf wget ca-certificates

      - name: Install CUDA toolkit
        run: |
          wget -q https://developer.download.nvidia.com/compute/cuda/repos/${{ matrix.cuda_repo_distro }}/${{ matrix.cuda_repo_arch }}/cuda-keyring_1.1-1_all.deb
          sudo dpkg -i cuda-keyring_1.1-1_all.deb
          sudo apt-get update
          sudo apt-get install -y --no-install-recommends \
            cuda-nvcc-${{ matrix.cuda_package_suffix }} \
            cuda-cudart-dev-${{ matrix.cuda_package_suffix }} \
            libcublas-dev-${{ matrix.cuda_package_suffix }}
          echo "${{ matrix.cuda_home }}/bin" >> "$GITHUB_PATH"
          echo "CUDA_HOME=${{ matrix.cuda_home }}" >> "$GITHUB_ENV"
          echo "CMAKE_CUDA_COMPILER=${{ matrix.cuda_home }}/bin/nvcc" >> "$GITHUB_ENV"

      - name: Fetch qwentts.cpp
        run: |
          git init third_party/qwentts.cpp
          git -C third_party/qwentts.cpp remote add origin https://github.com/ServeurpersoCom/qwentts.cpp
          git -C third_party/qwentts.cpp fetch --depth 1 origin "$QWENTTS_REF"
          git -C third_party/qwentts.cpp checkout --detach FETCH_HEAD
          git -C third_party/qwentts.cpp submodule update --init --recursive

      - name: Build wheel
        run: |
          python scripts/build_native.py \
            --source third_party/qwentts.cpp \
            --backend cuda \
            --clean \
            --cmake-arg=-G \
            --cmake-arg=Ninja \
            --cmake-arg="-DCMAKE_CUDA_ARCHITECTURES=$CUDA_ARCHITECTURES"
          python -m build --wheel
          mkdir -p wheelhouse
          python -m auditwheel repair \
            --plat ${{ matrix.manylinux_platform }} \
            --exclude libcudart.so.12 \
            --exclude libcudart.so.13 \
            --exclude libcublas.so.12 \
            --exclude libcublas.so.13 \
            --exclude libcublasLt.so.12 \
            --exclude libcublasLt.so.13 \
            --exclude libcuda.so.1 \
            -w wheelhouse \
            dist/*.whl

      - uses: actions/upload-artifact@v4
        with:
          name: qwentts-cpp-python-linux-${{ matrix.arch }}-cuda-${{ matrix.wheel_build_tag }}
          path: wheelhouse/*.whl

  linux-cpu:
    name: Linux ${{ matrix.arch }} CPU wheel
    runs-on: ${{ matrix.runner }}
    strategy:
      fail-fast: false
      matrix:
        include:
          - arch: x86_64
            runner: ubuntu-24.04
          - arch: aarch64
            runner: ubuntu-24.04-arm
    env:
      QWENTTS_CPP_BACKEND: cpu
      QWENTTS_CPP_BUILD_JOBS: "2"
      QWENTTS_CPP_WHEEL_BUILD_TAG: 1cpu
      QWENTTS_REF: ${{ inputs.qwentts_ref || '6fae92914045cd83364d2845ceaa0f7969727319' }}
    steps:
      - uses: actions/checkout@v4

      - uses: actions/setup-python@v5
        with:
          python-version: "3.12"

      - name: Install build tools
        run: |
          python -m pip install --upgrade pip auditwheel build wheel
          sudo apt-get update
          sudo apt-get install -y --no-install-recommends cmake ninja-build patchelf ca-certificates

      - name: Fetch qwentts.cpp
        run: |
          git init third_party/qwentts.cpp
          git -C third_party/qwentts.cpp remote add origin https://github.com/ServeurpersoCom/qwentts.cpp
          git -C third_party/qwentts.cpp fetch --depth 1 origin "$QWENTTS_REF"
          git -C third_party/qwentts.cpp checkout --detach FETCH_HEAD
          git -C third_party/qwentts.cpp submodule update --init --recursive

      - name: Build wheel
        run: |
          python scripts/build_native.py \
            --source third_party/qwentts.cpp \
            --backend cpu \
            --clean \
            --cmake-arg=-G \
            --cmake-arg=Ninja
          python -m build --wheel
          mkdir -p wheelhouse
          python -m auditwheel repair \
            --plat manylinux_2_39_${{ matrix.arch }} \
            -w wheelhouse \
            dist/*.whl

      - uses: actions/upload-artifact@v4
        with:
          name: qwentts-cpp-python-linux-${{ matrix.arch }}-cpu
          path: wheelhouse/*.whl