Nightly Rosetta T5x build and test #141
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: Nightly Rosetta T5x build and test | |
| on: | |
| workflow_run: | |
| workflows: [Nightly T5X build] | |
| types: [completed] | |
| branches: [main] | |
| workflow_dispatch: | |
| inputs: | |
| BASE_IMAGE: | |
| type: string | |
| description: 'T5x image built by NVIDIA/JAX-Toolbox' | |
| default: 'ghcr.io/nvidia/upstream-t5x:latest' | |
| required: true | |
| PUBLISH: | |
| type: boolean | |
| description: Publish dated images and update the 'latest' tag? | |
| default: false | |
| required: false | |
| env: | |
| BASE_LIBRARY: t5x | |
| DOCKER_REGISTRY: ghcr.io/nvidia | |
| permissions: | |
| contents: read # to fetch code | |
| actions: write # to cancel previous workflows | |
| packages: write # to upload container | |
| jobs: | |
| metadata: | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch' | |
| runs-on: ubuntu-22.04 | |
| outputs: | |
| BUILD_DATE: ${{ steps.meta-vars.outputs.BUILD_DATE }} | |
| BASE_LIBRARY: ${{ steps.meta-vars.outputs.BASE_LIBRARY }} | |
| BASE_IMAGE: ${{ steps.meta-vars.outputs.BASE_IMAGE }} | |
| PUBLISH: ${{ steps.meta-vars.outputs.PUBLISH }} | |
| steps: | |
| - name: Set build metadata | |
| id: meta-vars | |
| shell: bash -x -e {0} | |
| run: | | |
| BUILD_DATE=$(TZ='US/Los_Angeles' date '+%Y-%m-%d') | |
| if [[ -z "${{ inputs.BASE_IMAGE }}" ]]; then | |
| BASE_IMAGE=${{ env.DOCKER_REGISTRY }}/upstream-${{ env.BASE_LIBRARY }}:latest | |
| else | |
| BASE_IMAGE=${{ inputs.BASE_IMAGE }} | |
| fi | |
| echo "BUILD_DATE=${BUILD_DATE}" >> $GITHUB_OUTPUT | |
| echo "BASE_LIBRARY=${{ env.BASE_LIBRARY }}" >> $GITHUB_OUTPUT | |
| echo "BASE_IMAGE=${BASE_IMAGE}" >> $GITHUB_OUTPUT | |
| echo "PUBLISH=${{ inputs.PUBLISH }}" >> $GITHUB_OUTPUT | |
| build: | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch' | |
| needs: metadata | |
| uses: ./.github/workflows/_build_rosetta.yaml | |
| with: | |
| BUILD_DATE: ${{ needs.metadata.outputs.BUILD_DATE }} | |
| BASE_LIBRARY: ${{ needs.metadata.outputs.BASE_LIBRARY }} | |
| BASE_IMAGE: ${{ needs.metadata.outputs.BASE_IMAGE }} | |
| # TODO: Can't build ARM until https://github.com/NVIDIA/JAX-Toolbox/pull/252 is available | |
| PLATFORMS: '["amd64"]' | |
| secrets: inherit | |
| publish-build: | |
| needs: [metadata, build] | |
| uses: ./.github/workflows/_publish_badge.yaml | |
| if: ( success() || failure() ) && (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch' | |
| secrets: inherit | |
| with: | |
| ENDPOINT_FILENAME: 'rosetta-t5x-build-status.json' | |
| PUBLISH: ${{ github.event_name == 'workflow_run' || needs.metadata.outputs.PUBLISH == 'true' }} | |
| SCRIPT: | | |
| if [[ ${{ needs.build.result }} == "success" ]]; then | |
| BADGE_COLOR=brightgreen | |
| MSG=passing | |
| else | |
| BADGE_COLOR=red | |
| MSG=failing | |
| fi | |
| echo "LABEL='nightly'" >> $GITHUB_OUTPUT | |
| echo "MESSAGE='${MSG}'" >> $GITHUB_OUTPUT | |
| echo "COLOR='${BADGE_COLOR}'" >> $GITHUB_OUTPUT | |
| test-unit: | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch' | |
| needs: build | |
| uses: ./.github/workflows/_test_rosetta.yaml | |
| with: | |
| ROSETTA_IMAGE: ${{ needs.build.outputs.DOCKER_TAGS }} | |
| secrets: inherit | |
| test-t5x: | |
| needs: build | |
| uses: ./.github/workflows/_test_t5x.yaml | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch' | |
| with: | |
| T5X_IMAGE: ${{ needs.build.outputs.DOCKER_TAGS }} | |
| # Disable packing b/c rosetta-t5x images run with TE by default, and TE does not currently support packing | |
| EXTRA_GIN_ARGS: "--gin.train/utils.DatasetConfig.pack=False --gin.train_eval/utils.DatasetConfig.pack=False" | |
| secrets: inherit | |
| publish-t5x: | |
| needs: [metadata, test-t5x] | |
| uses: ./.github/workflows/_publish_t5x_results.yaml | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch' | |
| with: | |
| BUILD_DATE: ${{ needs.metadata.outputs.BUILD_DATE }} | |
| EXPERIMENT_SUBDIR: ROSETTA_T5X | |
| secrets: inherit | |
| publish-test: | |
| needs: [metadata, build, test-unit, test-t5x] | |
| uses: ./.github/workflows/_publish_badge.yaml | |
| if: ( always() ) | |
| secrets: inherit | |
| with: | |
| ENDPOINT_FILENAME: 'rosetta-t5x-overall-test-status.json' | |
| PUBLISH: ${{ github.event_name == 'workflow_run' || needs.metadata.outputs.PUBLISH == 'true' }} | |
| SCRIPT: | | |
| UNIT_STATUS=${{ needs.test-unit.outputs.TEST_STATUS }} | |
| T5X_STATUS=${{ needs.test-t5x.outputs.TEST_STATUS }} | |
| echo "LABEL='Tests'" >> $GITHUB_OUTPUT | |
| if [[ ${{ needs.build.result }} == "success" ]]; then | |
| if [[ $UNIT_STATUS == "success" ]] && [[ $T5X_STATUS == "success" ]]; then | |
| COLOR=brightgreen | |
| MESSAGE="Unit passed / MGMN passed" | |
| elif [[ $UNIT_STATUS == "success" ]]; then | |
| COLOR=yellow | |
| MESSAGE="Unit passed / MGMN failed" | |
| elif [[ $T5X_STATUS == "success" ]]; then | |
| COLOR=yellow | |
| MESSAGE="Unit failed / MGMN passed" | |
| else | |
| COLOR=red | |
| MESSAGE="Unit failed / MGMN failed" | |
| fi | |
| else | |
| MESSAGE="n/a" | |
| COLOR="red" | |
| fi | |
| echo "MESSAGE='${MESSAGE}'" >> $GITHUB_OUTPUT | |
| echo "COLOR='${COLOR}'" >> $GITHUB_OUTPUT | |
| publish-latest-container: | |
| needs: [metadata, build, test-t5x, test-unit] | |
| if: ( needs.test-unit.outputs.TEST_STATUS == 'success' && needs.test-t5x.outputs.TEST_STATUS == 'success' ) && ((github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || (github.event_name == 'workflow_dispatch' && inputs.PUBLISH)) | |
| uses: ./.github/workflows/_publish_container.yaml | |
| secrets: inherit | |
| with: | |
| SOURCE_IMAGE: ${{ needs.build.outputs.DOCKER_TAGS }} | |
| TARGET_IMAGE: t5x | |
| TARGET_TAGS: | | |
| type=raw,value=latest,priority=1000 | |
| publish-container: | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || (github.event_name == 'workflow_dispatch' && inputs.PUBLISH) | |
| needs: [metadata, build] | |
| uses: ./.github/workflows/_publish_container.yaml | |
| secrets: inherit | |
| with: | |
| SOURCE_IMAGE: ${{ needs.build.outputs.DOCKER_TAGS }} | |
| TARGET_IMAGE: t5x | |
| TARGET_TAGS: | | |
| type=raw,value=nightly-${{ needs.metadata.outputs.BUILD_DATE }},priority=900 | |
| if-upstream-failed: | |
| runs-on: ubuntu-latest | |
| if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'failure') && github.event_name != 'workflow_dispatch' | |
| steps: | |
| - run: echo 'Upstream workflow failed, aborting run' && exit 1 |