nightly-te-test.yaml

name: Nightly Transformer Engine test
run-name: Nightly Transformer Engine test (${{ github.event_name == 'workflow_run' && format('nightly {0}', github.event.workflow_run.created_at) || github.event_name }})

on:
  workflow_run:
    workflows: [Nightly Transformer Engine build]
    types: [completed]
    branches: [main]
  workflow_dispatch:
    inputs:
      JAX_TE_IMAGE:
        type: string
        description: 'JAX-TE image build by NVIDIA/JAX-Toolbox'
        required: true
        default: 'ghcr.io/nvidia/jax-te:latest'
      PUBLISH:
        type: boolean
        description: Update status badge?
        default: false
        required: true

permissions:
  contents: read  # to fetch code
  actions:  write # to cancel previous workflows
  packages: write # to upload container

env:
  DEFAULT_JAX_TE_IMAGE: 'ghcr.io/nvidia/jax-te:latest'

jobs:

  metadata:
    runs-on: ubuntu-22.04
    outputs:
      JAX_TE_IMAGE: ${{ steps.meta.outputs.JAX_TE_IMAGE }}
      PUBLISH: ${{ steps.meta.outputs.PUBLISH }}
    steps:
      - name: Set metadata
        id: meta
        shell: bash -x -e {0}
        run: |
          if [[ -z "${{ inputs.JAX_TE_IMAGE }}" ]]; then
            JAX_TE_IMAGE=${{ env.DEFAULT_JAX_TE_IMAGE }}
          else
            JAX_TE_IMAGE=${{ inputs.JAX_TE_IMAGE }}
          fi
          echo "JAX_TE_IMAGE=${JAX_TE_IMAGE}" >> $GITHUB_OUTPUT
          echo "PUBLISH=${{ inputs.PUBLISH }}" >> $GITHUB_OUTPUT

  run-jobs:
    needs: metadata
    uses: ./.github/workflows/_test_te.yaml
    if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch'
    with:
      JAX_TE_IMAGE: ${{ needs.metadata.outputs.JAX_TE_IMAGE }}
    secrets: inherit

  publish-unit:
    needs: [metadata, run-jobs]
    uses: ./.github/workflows/_publish_badge.yaml
    if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch'
    secrets: inherit
    with:
      ENDPOINT_FILENAME: 'te-unit-test-status.json'
      PUBLISH: ${{ github.event_name == 'workflow_run' || needs.metadata.outputs.PUBLISH == 'true' }}
      SCRIPT: |
        ARTIFACTS="${{ needs.run-jobs.outputs.UNIT_TEST_ARTIFACT_NAME }}/*.jsonl"
        all_outcomes() {
          cat $ARTIFACTS | jq -r '. | select((.["$report_type"] == "TestReport") and (.when == "call")) | .outcome'
        }
        cnt_type() {
          cat $ARTIFACTS | jq '. | select((.["$report_type"] == "TestReport") and (.when == "call") and (.outcome | contains("'${1}'"))) | .outcome' | wc -l
        }
        SKIPPED_TESTS=$(cnt_type skipped)
        FAILED_TESTS=$(cnt_type failed)
        PASSED_TESTS=$(cnt_type passed)
        TOTAL_TESTS=$(all_outcomes | wc -l)
        echo "Unit test breakdown:"
        all_outcomes | sort | uniq -c
        if [[ $FAILED_TESTS -eq 0 ]] && [[ $TOTAL_TESTS -gt 0 ]]; then
          BADGE_COLOR=brightgreen
        else
          if [[ $PASSED_TESTS -eq 0 ]]; then
            BADGE_COLOR=red
          else
            BADGE_COLOR=yellow
          fi
        fi
        echo "LABEL='Unit'" >> $GITHUB_OUTPUT
        echo "MESSAGE='${PASSED_TESTS}/${SKIPPED_TESTS}/${FAILED_TESTS} pass/skip/fail'" >> $GITHUB_OUTPUT
        echo "COLOR='${BADGE_COLOR}'" >> $GITHUB_OUTPUT

  publish-integration:
    needs: [metadata, run-jobs]
    uses: ./.github/workflows/_publish_badge.yaml
    if: (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success') || github.event_name == 'workflow_dispatch'
    secrets: inherit
    with:
      ENDPOINT_FILENAME: 'te-integration-test-status.json'
      PUBLISH: ${{ github.event_name == 'workflow_run' || needs.metadata.outputs.PUBLISH == 'true' }}
      SCRIPT: |
        ARTIFACTS="${{ needs.run-jobs.outputs.INTEGRATION_TEST_ARTIFACT_NAME }}/*.jsonl"
        all_outcomes() {
          cat $ARTIFACTS | jq -r '. | select((.["$report_type"] == "TestReport") and (.when == "call")) | .outcome'
        }
        cnt_type() {
          cat $ARTIFACTS | jq '. | select((.["$report_type"] == "TestReport") and (.when == "call") and (.outcome | contains("'${1}'"))) | .outcome' | wc -l
        }
        SKIPPED_TESTS=$(cnt_type skipped)
        FAILED_TESTS=$(cnt_type failed)
        PASSED_TESTS=$(cnt_type passed)
        TOTAL_TESTS=$(all_outcomes | wc -l)
        echo "Unit test breakdown:"
        all_outcomes | sort | uniq -c
        if [[ $FAILED_TESTS -eq 0 ]] && [[ $TOTAL_TESTS -gt 0 ]]; then
          BADGE_COLOR=brightgreen
        else
          if [[ $PASSED_TESTS -eq 0 ]]; then
            BADGE_COLOR=red
          else
            BADGE_COLOR=yellow
          fi
        fi
        echo "LABEL='Integration'" >> $GITHUB_OUTPUT
        echo "MESSAGE='${PASSED_TESTS}/${SKIPPED_TESTS}/${FAILED_TESTS} pass/skip/fail'" >> $GITHUB_OUTPUT
        echo "COLOR='${BADGE_COLOR}'" >> $GITHUB_OUTPUT

  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