nightly-release.yml

name: Nightly-Release

on:
  workflow_dispatch:
    inputs:
      repo:
        description: "repo"
        default: "flashinfer-ai/flashinfer"
      branch:
        description: "branch"
        default: "main"
  schedule:
    - cron: '0 2 * * *'

env:
  TORCH_CUDA_ARCH_LIST: "7.5 8.0 8.9 9.0+PTX"

jobs:
  build:
    outputs:
      version: ${{ steps.get_version.outputs.version }}
    strategy:
      fail-fast: false
      matrix:
        cuda: ["12.1", "12.4"]
        torch: ["2.4"]
    runs-on: [self-hosted]
    steps:
      - name: Set default values
        id: set_defaults
        run: |
          echo "[INFO] Event name is '${{ github.event_name }}'"
          if [[ "${{ github.event_name }}" == "workflow_dispatch" ]]; then
            echo "REPO=${{ inputs.repo }}" >> $GITHUB_ENV
            echo "BRANCH=${{ inputs.branch }}" >> $GITHUB_ENV
          elif [[ "${{ github.event_name }}" == "schedule" ]]; then
            echo "REPO=flashinfer-ai/flashinfer" >> $GITHUB_ENV
            echo "BRANCH=main" >> $GITHUB_ENV
          else
            echo "[ERROR] Unsupported event '${{ github.event_name }}'." >&2
            exit 1
          fi

      - run: |
          rm -rf flashinfer
          git clone -b ${{ env.BRANCH }} --recurse-submodules --depth=1 https://github.com/${{ env.REPO }}
          pushd flashinfer
          sed -i 's|/ci-cache|/opt/dlami/nvme/flashinfer/github|g' scripts/run-ci-build-wheel.sh
          sed -i '/mkdir -p "\$CONDA_pkgs_dirs" "\$XDG_CACHE_HOME"/d' scripts/run-ci-build-wheel.sh
          version="$(cat version.txt)"+"$(git rev-parse HEAD | cut -c1-7)".cu${{ matrix.cuda }}torch${{ matrix.torch }}
          echo "VERSION=$version" >> $GITHUB_ENV
          echo "version=$version" >> $GITHUB_OUTPUT
        id: get_version
      - name: Build wheel
        run: |
          pushd flashinfer
          docker run --rm -t \
              -v "$(pwd)":/app \
              -e FLASHINFER_CI_CUDA_VERSION=${{ matrix.cuda }} \
              -e FLASHINFER_CI_TORCH_VERSION=${{ matrix.torch }} \
              -e FLASHINFER_BUILD_VERSION=${{ env.VERSION }} \
              -e TORCH_CUDA_ARCH_LIST="$TORCH_CUDA_ARCH_LIST" \
              -e MAX_JOBS=192 \
              --user $(id -u):$(id -g) \
              pytorch/manylinux-builder:cuda${{ matrix.cuda }} \
              bash /app/scripts/run-ci-build-wheel.sh
          popd
        timeout-minutes: 30
      - run: du -h flashinfer/dist/*
      - uses: actions/upload-artifact@v4
        with:
          name: wheel-cuda${{ matrix.cuda }}-torch${{ matrix.torch }}
          path: flashinfer/dist/*

  release:
    needs: build
    runs-on: [self-hosted]
    steps:
      - uses: actions/download-artifact@v4
        with:
          path: dist/
          merge-multiple: true
          pattern: wheel-*
      - run: ls -lah dist/
      - run: |
          version="${{ needs.build.outputs.version }}"
          tag_version=$(echo $version | sed 's/\.cu[0-9.]*torch[0-9.]*$//')
          echo "TAG_VERSION=$tag_version" >> $GITHUB_ENV
          find dist -type f ! -name "*${tag_version}*" -delete
      - uses: softprops/action-gh-release@v1
        with:
          tag_name: ${{ env.TAG_VERSION }}
          files: dist/flashinfer*.whl