|
| 1 | +name: PR comment GitHub CI |
| 2 | + |
| 3 | +on: |
| 4 | + issue_comment: |
| 5 | + types: |
| 6 | + - created |
| 7 | + branches-ignore: |
| 8 | + - main |
| 9 | +concurrency: |
| 10 | + group: ${{ github.workflow }}-${{ github.event.issue.number }}-${{ startsWith(github.event.comment.body, 'run-slow') || startsWith(github.event.comment.body, 'run slow') || startsWith(github.event.comment.body, 'run_slow') }} |
| 11 | + cancel-in-progress: true |
| 12 | +permissions: read-all |
| 13 | + |
| 14 | +env: |
| 15 | + HF_HOME: /mnt/cache |
| 16 | + TRANSFORMERS_IS_CI: yes |
| 17 | + OMP_NUM_THREADS: 8 |
| 18 | + MKL_NUM_THREADS: 8 |
| 19 | + RUN_SLOW: yes |
| 20 | + # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access. |
| 21 | + # This token is created under the bot `hf-transformers-bot`. |
| 22 | + HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }} |
| 23 | + SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }} |
| 24 | + TF_FORCE_GPU_ALLOW_GROWTH: true |
| 25 | + RUN_PT_TF_CROSS_TESTS: 1 |
| 26 | + CUDA_VISIBLE_DEVICES: 0,1 |
| 27 | + |
| 28 | +jobs: |
| 29 | + get-pr-number: |
| 30 | + runs-on: ubuntu-22.04 |
| 31 | + name: Get PR number |
| 32 | + # For security: only allow team members to run |
| 33 | + if: ${{ github.event.issue.state == 'open' && contains(fromJSON('["ydshieh", "ArthurZucker", "zucchini-nlp", "qubvel", "molbap", "gante", "LysandreJik", "Cyrilvallez"]'), github.actor) && (startsWith(github.event.comment.body, 'run-slow') || startsWith(github.event.comment.body, 'run slow') || startsWith(github.event.comment.body, 'run_slow')) }} |
| 34 | + outputs: |
| 35 | + PR_NUMBER: ${{ steps.set_pr_number.outputs.PR_NUMBER }} |
| 36 | + steps: |
| 37 | + - name: Get PR number |
| 38 | + shell: bash |
| 39 | + run: | |
| 40 | + if [[ "${{ github.event.issue.number }}" != "" && "${{ github.event.issue.pull_request }}" != "" ]]; then |
| 41 | + echo "PR_NUMBER=${{ github.event.issue.number }}" >> $GITHUB_ENV |
| 42 | + else |
| 43 | + echo "PR_NUMBER=" >> $GITHUB_ENV |
| 44 | + fi |
| 45 | +
|
| 46 | + - name: Check PR number |
| 47 | + shell: bash |
| 48 | + run: | |
| 49 | + echo "${{ env.PR_NUMBER }}" |
| 50 | +
|
| 51 | + - name: Set PR number |
| 52 | + id: set_pr_number |
| 53 | + run: echo "PR_NUMBER=${{ env.PR_NUMBER }}" >> "$GITHUB_OUTPUT" |
| 54 | + |
| 55 | + get-sha: |
| 56 | + runs-on: ubuntu-22.04 |
| 57 | + needs: get-pr-number |
| 58 | + if: ${{ needs.get-pr-number.outputs.PR_NUMBER != ''}} |
| 59 | + outputs: |
| 60 | + PR_HEAD_SHA: ${{ steps.get_sha.outputs.PR_HEAD_SHA }} |
| 61 | + steps: |
| 62 | + - uses: actions/checkout@v4 |
| 63 | + with: |
| 64 | + fetch-depth: "0" |
| 65 | + ref: "refs/pull/${{needs.get-pr-number.outputs.PR_NUMBER}}/merge" |
| 66 | + |
| 67 | + - name: Get SHA (and verify timestamps against the issue comment date) |
| 68 | + id: get_sha |
| 69 | + env: |
| 70 | + PR_NUMBER: ${{ needs.get-pr-number.outputs.PR_NUMBER }} |
| 71 | + COMMENT_DATE: ${{ github.event.comment.created_at }} |
| 72 | + run: | |
| 73 | + git fetch origin refs/pull/$PR_NUMBER/head:refs/remotes/pull/$PR_NUMBER/head |
| 74 | + git checkout refs/remotes/pull/$PR_NUMBER/head |
| 75 | + echo "PR_HEAD_SHA: $(git log -1 --format=%H)" |
| 76 | + echo "PR_HEAD_SHA=$(git log -1 --format=%H)" >> "$GITHUB_OUTPUT" |
| 77 | + git fetch origin refs/pull/$PR_NUMBER/merge:refs/remotes/pull/$PR_NUMBER/merge |
| 78 | + git checkout refs/remotes/pull/$PR_NUMBER/merge |
| 79 | + PR_MERGE_COMMIT_TIMESTAMP=$(git log -1 --date=unix --format=%cd) |
| 80 | + echo "PR_MERGE_COMMIT_TIMESTAMP: $PR_MERGE_COMMIT_TIMESTAMP" |
| 81 | + COMMENT_TIMESTAMP=$(date -d "${COMMENT_DATE}" +"%s") |
| 82 | + echo "PR_HEAD_SHA: $COMMENT_DATE" |
| 83 | + echo "COMMENT_TIMESTAMP: $COMMENT_TIMESTAMP" |
| 84 | + if [ $COMMENT_TIMESTAMP -le $PR_MERGE_COMMIT_TIMESTAMP ]; then |
| 85 | + echo "Last commit on the pull request is newer than the issue comment triggering this run! Abort!"; |
| 86 | + exit -1; |
| 87 | + fi |
| 88 | +
|
| 89 | + # use a python script to handle this complex logic |
| 90 | + # case 1: `run-slow` (auto. infer with limited number of models, but in particular, new model) |
| 91 | + # case 2: `run-slow model_1, model_2` |
| 92 | + get-tests: |
| 93 | + runs-on: ubuntu-22.04 |
| 94 | + needs: get-pr-number |
| 95 | + if: ${{ needs.get-pr-number.outputs.PR_NUMBER != ''}} |
| 96 | + outputs: |
| 97 | + models: ${{ steps.models_to_run.outputs.models }} |
| 98 | + steps: |
| 99 | + - uses: actions/checkout@v4 |
| 100 | + with: |
| 101 | + fetch-depth: "0" |
| 102 | + ref: "refs/pull/${{needs.get-pr-number.outputs.PR_NUMBER}}/merge" |
| 103 | + |
| 104 | + - name: Get models to test |
| 105 | + env: |
| 106 | + PR_COMMENT: ${{ github.event.comment.body }} |
| 107 | + run: | |
| 108 | + python -m pip install GitPython |
| 109 | + python utils/pr_slow_ci_models.py --message "$PR_COMMENT" | tee output.txt |
| 110 | + echo "models=$(tail -n 1 output.txt)" >> $GITHUB_ENV |
| 111 | +
|
| 112 | + - name: Show models to test |
| 113 | + id: models_to_run |
| 114 | + run: | |
| 115 | + echo "${{ env.models }}" |
| 116 | + echo "models=${{ env.models }}" >> $GITHUB_ENV |
| 117 | + echo "models=${{ env.models }}" >> $GITHUB_OUTPUT |
| 118 | +
|
| 119 | + reply_to_comment: |
| 120 | + name: Reply to the comment |
| 121 | + if: ${{ needs.get-tests.outputs.models != '[]' }} |
| 122 | + needs: [get-pr-number, get-tests] |
| 123 | + permissions: |
| 124 | + pull-requests: write |
| 125 | + runs-on: ubuntu-22.04 |
| 126 | + steps: |
| 127 | + - name: Reply to the comment |
| 128 | + env: |
| 129 | + GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 130 | + MODELS: ${{ needs.get-tests.outputs.models }} |
| 131 | + run: | |
| 132 | + gh api \ |
| 133 | + --method POST \ |
| 134 | + -H "Accept: application/vnd.github+json" \ |
| 135 | + -H "X-GitHub-Api-Version: 2022-11-28" \ |
| 136 | + repos/${{ github.repository }}/issues/${{ needs.get-pr-number.outputs.PR_NUMBER }}/comments \ |
| 137 | + -f "body=This comment contains run-slow, running the specified jobs: ${{ env.MODELS }} ..." |
| 138 | +
|
| 139 | + create_run: |
| 140 | + name: Create run |
| 141 | + if: ${{ needs.get-tests.outputs.models != '[]' }} |
| 142 | + needs: [get-sha, get-tests, reply_to_comment] |
| 143 | + permissions: |
| 144 | + statuses: write |
| 145 | + runs-on: ubuntu-22.04 |
| 146 | + steps: |
| 147 | + - name: Create Run |
| 148 | + id: create_run |
| 149 | + env: |
| 150 | + GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 151 | + # Create a commit status (pending) for a run of this workflow. The status has to be updated later in `update_run_status`. |
| 152 | + # See https://docs.github.com/en/rest/commits/statuses?apiVersion=2022-11-28#create-a-commit-status |
| 153 | + GITHUB_RUN_URL: https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }} |
| 154 | + run: | |
| 155 | + gh api \ |
| 156 | + --method POST \ |
| 157 | + -H "Accept: application/vnd.github+json" \ |
| 158 | + -H "X-GitHub-Api-Version: 2022-11-28" \ |
| 159 | + repos/${{ github.repository }}/statuses/${{ needs.get-sha.outputs.PR_HEAD_SHA }} \ |
| 160 | + -f "target_url=$GITHUB_RUN_URL" -f "state=pending" -f "description=Slow CI job" -f "context=pytest/custom-tests" |
| 161 | +
|
| 162 | + run_models_gpu: |
| 163 | + name: Run all tests for the model |
| 164 | + if: ${{ needs.get-tests.outputs.models != '[]' }} |
| 165 | + needs: [get-pr-number, get-tests, create_run] |
| 166 | + strategy: |
| 167 | + fail-fast: false |
| 168 | + matrix: |
| 169 | + folders: ${{ fromJson(needs.get-tests.outputs.models) }} |
| 170 | + machine_type: [aws-g4dn-2xlarge-cache, aws-g4dn-12xlarge-cache] |
| 171 | + runs-on: |
| 172 | + group: '${{ matrix.machine_type }}' |
| 173 | + container: |
| 174 | + image: huggingface/transformers-all-latest-gpu |
| 175 | + options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ |
| 176 | + steps: |
| 177 | + - name: Echo input and matrix info |
| 178 | + shell: bash |
| 179 | + run: | |
| 180 | + echo "${{ matrix.folders }}" |
| 181 | +
|
| 182 | + - name: Echo folder ${{ matrix.folders }} |
| 183 | + shell: bash |
| 184 | + # For folders like `models/bert`, set an env. var. (`matrix_folders`) to `models_bert`, which will be used to |
| 185 | + # set the artifact folder names (because the character `/` is not allowed). |
| 186 | + run: | |
| 187 | + echo "${{ matrix.folders }}" |
| 188 | + matrix_folders=${{ matrix.folders }} |
| 189 | + matrix_folders=${matrix_folders/'models/'/'models_'} |
| 190 | + echo "$matrix_folders" |
| 191 | + echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV |
| 192 | +
|
| 193 | + - name: Checkout to PR merge commit |
| 194 | + working-directory: /transformers |
| 195 | + run: | |
| 196 | + git fetch origin refs/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge:refs/remotes/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge |
| 197 | + git checkout refs/remotes/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge |
| 198 | + git log -1 --format=%H |
| 199 | +
|
| 200 | + - name: Reinstall transformers in edit mode (remove the one installed during docker image build) |
| 201 | + working-directory: /transformers |
| 202 | + run: python3 -m pip uninstall -y transformers && python3 -m pip install -e . |
| 203 | + |
| 204 | + - name: NVIDIA-SMI |
| 205 | + run: | |
| 206 | + nvidia-smi |
| 207 | +
|
| 208 | + - name: Set `machine_type` for report and artifact names |
| 209 | + working-directory: /transformers |
| 210 | + shell: bash |
| 211 | + run: | |
| 212 | + echo "${{ matrix.machine_type }}" |
| 213 | + if [ "${{ matrix.machine_type }}" = "aws-g4dn-2xlarge-cache" ]; then |
| 214 | + machine_type=single-gpu |
| 215 | + elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then |
| 216 | + machine_type=multi-gpu |
| 217 | + else |
| 218 | + machine_type=${{ matrix.machine_type }} |
| 219 | + fi |
| 220 | + echo "$machine_type" |
| 221 | + echo "machine_type=$machine_type" >> $GITHUB_ENV |
| 222 | +
|
| 223 | + - name: Environment |
| 224 | + working-directory: /transformers |
| 225 | + run: | |
| 226 | + python3 utils/print_env.py |
| 227 | +
|
| 228 | + - name: Show installed libraries and their versions |
| 229 | + working-directory: /transformers |
| 230 | + run: pip freeze |
| 231 | + |
| 232 | + - name: Run all tests on GPU |
| 233 | + working-directory: /transformers |
| 234 | + run: | |
| 235 | + export CUDA_VISIBLE_DEVICES="$(python3 utils/set_cuda_devices_for_ci.py --test_folder ${{ matrix.folders }})" |
| 236 | + echo $CUDA_VISIBLE_DEVICES |
| 237 | + python3 -m pytest -v -rsfE --make-reports=${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }} |
| 238 | +
|
| 239 | + - name: Failure short reports |
| 240 | + if: ${{ failure() }} |
| 241 | + continue-on-error: true |
| 242 | + run: cat /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt |
| 243 | + |
| 244 | + - name: Make sure report directory exists |
| 245 | + shell: bash |
| 246 | + run: | |
| 247 | + mkdir -p /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports |
| 248 | + echo "hello" > /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/hello.txt |
| 249 | + echo "${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports" |
| 250 | +
|
| 251 | + - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports" |
| 252 | + if: ${{ always() }} |
| 253 | + uses: actions/upload-artifact@v4 |
| 254 | + with: |
| 255 | + name: ${{ env.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports |
| 256 | + path: /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports |
| 257 | + |
| 258 | + update_run_status: |
| 259 | + name: Update Check Run Status |
| 260 | + needs: [get-sha, create_run, run_models_gpu] |
| 261 | + permissions: |
| 262 | + statuses: write |
| 263 | + if: ${{ always() && needs.create_run.result == 'success' }} |
| 264 | + runs-on: ubuntu-22.04 |
| 265 | + env: |
| 266 | + GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 267 | + GITHUB_RUN_URL: https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }} |
| 268 | + steps: |
| 269 | + - name: Get `run_models_gpu` job status |
| 270 | + run: | |
| 271 | + echo "${{ needs.run_models_gpu.result }}" |
| 272 | + if [ "${{ needs.run_models_gpu.result }}" = "cancelled" ]; then |
| 273 | + echo "STATUS=failure" >> $GITHUB_ENV |
| 274 | + elif [ "${{ needs.run_models_gpu.result }}" = "skipped" ]; then |
| 275 | + echo "STATUS=success" >> $GITHUB_ENV |
| 276 | + else |
| 277 | + echo "STATUS=${{ needs.run_models_gpu.result }}" >> $GITHUB_ENV |
| 278 | + fi |
| 279 | +
|
| 280 | + - name: Update PR commit statuses |
| 281 | + run: | |
| 282 | + echo "${{ needs.run_models_gpu.result }}" |
| 283 | + echo "${{ env.STATUS }}" |
| 284 | + gh api \ |
| 285 | + --method POST \ |
| 286 | + -H "Accept: application/vnd.github+json" \ |
| 287 | + -H "X-GitHub-Api-Version: 2022-11-28" \ |
| 288 | + repos/${{ github.repository }}/statuses/${{ needs.get-sha.outputs.PR_HEAD_SHA }} \ |
| 289 | + -f "target_url=$GITHUB_RUN_URL" -f "state=${{ env.STATUS }}" -f "description=Slow CI job" -f "context=pytest/custom-tests" |
0 commit comments