E2E on AWS CUDA - [DO NOT MERGE] exploration #5061
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: E2E on AWS CUDA | |
| run-name: E2E on AWS CUDA - ${{ github.event.workflow_run.display_title }} | |
| # We have to keep pre-commit AWS CUDA testing in a separate workflow because we | |
| # need access to AWS secret and that isn't available on pull_request jobs for | |
| # PRs from forks. And github's "require approval for all outside collaborators" | |
| # is bypassed on pull_request_target. | |
| # | |
| # Also, we use commit status and not check suite/run (which, in theory, is more | |
| # powerful) due to https://github.com/orgs/community/discussions/24616. | |
| on: | |
| workflow_run: | |
| workflows: [SYCL Pre Commit on Linux] | |
| types: | |
| - completed | |
| permissions: | |
| contents: read | |
| jobs: | |
| create-check: | |
| runs-on: [Linux, build] | |
| permissions: | |
| checks: write | |
| statuses: write | |
| steps: | |
| - uses: actions/github-script@v7 | |
| with: | |
| script: | | |
| const sha = context.payload.workflow_run.head_sha | |
| const run_id = '${{ github.run_id }}' | |
| const this_run_url = 'https://github.com/' + context.repo.owner + '/' + context.repo.repo + '/actions/runs/' + run_id | |
| // Create commit status. | |
| await github.request('POST /repos/{owner}/{repo}/statuses/{sha}', { | |
| owner: context.repo.owner, | |
| repo: context.repo.repo, | |
| sha: sha, | |
| state: 'pending', | |
| target_url: this_run_url, | |
| description: 'SYCL E2E on AWS CUDA', | |
| context: 'SYCL E2E on AWS CUDA', | |
| }) | |
| aws-start: | |
| runs-on: ubuntu-20.04 | |
| environment: aws | |
| steps: | |
| - uses: actions/checkout@v4 | |
| with: | |
| sparse-checkout: devops/actions/aws-ec2 | |
| - run: npm install ./devops/actions/aws-ec2 | |
| - uses: ./devops/actions/aws-ec2 | |
| with: | |
| mode: start | |
| runs-on-list: '[{"runs-on":"aws_cuda-${{ github.event.workflow_run.id }}-${{ github.event.workflow_run.run_attempt }}","aws-ami":"ami-01cb0573cb039ab24","aws-type":["g5.2xlarge","g5.4xlarge"],"aws-disk":"/dev/sda1:64","aws-spot":"false"}]' | |
| GH_PERSONAL_ACCESS_TOKEN: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }} | |
| AWS_ACCESS_KEY: ${{ secrets.AWS_ACCESS_KEY }} | |
| AWS_SECRET_KEY: ${{ secrets.AWS_SECRET_KEY }} | |
| e2e-cuda: | |
| needs: [aws-start] | |
| uses: ./.github/workflows/sycl-linux-run-tests.yml | |
| with: | |
| name: CUDA E2E | |
| runner: '["aws_cuda-${{ github.event.workflow_run.id }}-${{ github.event.workflow_run.run_attempt }}"]' | |
| image: ghcr.io/intel/llvm/ubuntu2204_build:latest | |
| image_options: -u 1001 --gpus all --cap-add SYS_ADMIN --env NVIDIA_DISABLE_REQUIRE=1 | |
| target_devices: ext_oneapi_cuda:gpu | |
| # No idea why but that seems to work and be in sync with the main | |
| # pre-commit workflow. | |
| ref: ${{ github.event.workflow_run.referenced_workflows[0].sha }} | |
| merge_ref: '' | |
| sycl_toolchain_artifact: sycl_linux_default | |
| sycl_toolchain_archive: llvm_sycl.tar.zst | |
| sycl_toolchain_decompress_command: zstd | |
| update-check: | |
| needs: [create-check, e2e-cuda] | |
| if: always() | |
| runs-on: [Linux, build] | |
| permissions: | |
| checks: write | |
| statuses: write | |
| steps: | |
| - uses: actions/github-script@v7 | |
| with: | |
| script: | | |
| const sha = context.payload.workflow_run.head_sha | |
| const run_id = '${{ github.run_id }}' | |
| const this_run_url = 'https://github.com/' + context.repo.owner + '/' + context.repo.repo + '/actions/runs/' + run_id | |
| // Update commit status. | |
| await github.request('POST /repos/{owner}/{repo}/statuses/{sha}', { | |
| owner: context.repo.owner, | |
| repo: context.repo.repo, | |
| sha: sha, | |
| state: '${{ needs.e2e-cuda.result }}', | |
| target_url: this_run_url, | |
| description: 'SYCL E2E on AWS CUDA', | |
| context: 'SYCL E2E on AWS CUDA', | |
| }) | |
| aws-stop: | |
| needs: [aws-start, e2e-cuda] | |
| if: always() | |
| runs-on: ubuntu-20.04 | |
| environment: aws | |
| steps: | |
| - uses: actions/checkout@v4 | |
| with: | |
| sparse-checkout: devops/actions/aws-ec2 | |
| - run: npm install ./devops/actions/aws-ec2 | |
| - uses: ./devops/actions/aws-ec2 | |
| with: | |
| mode: stop | |
| runs-on-list: '[{"runs-on":"aws_cuda-${{ github.event.workflow_run.id }}-${{ github.event.workflow_run.run_attempt }}","aws-ami":"ami-01cb0573cb039ab24","aws-type":["g5.2xlarge","g5.4xlarge"],"aws-disk":"/dev/sda1:64","aws-spot":"false"}]' | |
| GH_PERSONAL_ACCESS_TOKEN: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }} | |
| AWS_ACCESS_KEY: ${{ secrets.AWS_ACCESS_KEY }} | |
| AWS_SECRET_KEY: ${{ secrets.AWS_SECRET_KEY }} |