build-ngc-xformers.yaml

name: xformers for NGC
run-name: xformers via ${{ github.event_name }}

on:
  push:
    branches:
      - "main"
      - "release"
    paths:
      - ".github/workflows/build-ngc-xformers.yaml"
      - "docker/xformers/*"
      - "docker-bake.xformers.hcl"
      - "docker-bake.ngc.hcl"
      - "!**.md"

  workflow_dispatch:
    inputs:
      force-push:
        description: "push to GHCR"
        type: boolean
        required: true
        default: false

  workflow_call:
    inputs:
      force-push:
        description: "push to GHCR"
        type: boolean
        required: true
        default: false

  pull_request:
    paths:
      - ".github/workflows/build-ngc-xformers.yaml"
      - "docker/xformers/*"
      - "docker-bake.xformers.hcl"
      - "docker-bake.ngc.hcl"
      - "!**.md"

defaults:
  run:
    shell: bash

concurrency:
  group: ${{ github.workflow }}-${{ github.event_name }}-${{ github.ref }}-xformers
  cancel-in-progress: true

env:
  IMAGE_REGISTRY: ghcr.io
  IMAGE_NAME: xformers
  IMAGE_NAMESPACE: neggles/tensorpods
  IMAGE_PLATFORMS: "linux/amd64"

  # sorry pascal users but your cards are no good here
  TORCH_CUDA_ARCH_LIST: "7.0;7.5;8.0;8.6;8.9;9.0"
  # for ninja so the runner doesn't explode
  MAX_JOBS: 1
  # NVCC my behated
  NVCC_THREADS: 1

jobs:
  build:
    name: NGC
    runs-on: ubuntu-latest
    permissions:
      packages: write
      contents: read

    strategy:
      fail-fast: false
      max-parallel: 1
      matrix:
        include:
          - target: "xformers"
            ngc-ver: "23.08"
            xformers-ver: "tensorpods-v0.0.21"

          - target: "xformers"
            ngc-ver: "23.08"
            xformers-ver: "tensorpods"

          - target: "xformers"
            ngc-ver: "23.07"
            xformers-ver: "tensorpods"

    steps:
      - name: Checkout
        id: checkout
        uses: actions/checkout@v3
        with:
          submodules: recursive

      - name: Set up Docker Buildx
        id: buildx
        uses: docker/setup-buildx-action@v2
        with:
          platforms: ${{ env.IMAGE_PLATFORMS }}

      - name: Login to GHCR
        uses: docker/login-action@v2
        with:
          registry: ${{ env.IMAGE_REGISTRY }}
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - name: Generate docker tags
        id: meta
        uses: docker/metadata-action@v4
        with:
          flavor: |
            suffix=-ngc${{ matrix.ngc-ver }}
          images: |
            ${{ env.IMAGE_REGISTRY }}/${{ github.repository }}/${{ matrix.target }}
          tags: |
            type=raw,value=latest,enable={{is_default_branch}}
            type=raw,value=${{ matrix.xformers-ver }},enable={{is_default_branch}}
            type=schedule,pattern={{date 'YYYYMMDD-hhmm' tz='UTC'}},enable={{is_default_branch}}
            type=sha,format=short
            type=ref,event=tag
            type=ref,event=pr
          labels: |
            org.opencontainers.image.title=xformers-ngc
            org.opencontainers.image.description=xformers wheel for NGC pytorch containers
            org.opencontainers.image.vendor=neggles.dev

      - name: Free disk space
        id: free-disk-space
        run: |
          df -h .
          sudo find "$AGENT_TOOLSDIRECTORY" -delete
          sudo find /usr/share/dotnet -delete
          sudo find /usr/local/lib/android -delete
          df -h .

      - name: Build & Push Image
        id: build-push
        uses: docker/bake-action@v3
        env:
          NGC_VERSION: ${{ matrix.ngc-ver }}
          XFORMERS_REF: ${{ matrix.xformers-ver }}
        with:
          targets: ${{ matrix.target }}
          files: |
            ./docker-bake.xformers.hcl
            ${{ steps.meta.outputs.bake-file }}
          push: ${{ contains(fromJSON('["push", "schedule"]'), github.event_name) || inputs.force-push }}
          set: |
            *.cache-from=type=gha
            *.cache-to=type=gha,mode=min