diff --git a/.github/workflows/self-push-amd-mi210-caller.yml b/.github/workflows/self-push-amd-mi210-caller.yml index 561230438..2fb075eb2 100644 --- a/.github/workflows/self-push-amd-mi210-caller.yml +++ b/.github/workflows/self-push-amd-mi210-caller.yml @@ -19,7 +19,7 @@ jobs: run_amd_ci: name: AMD mi210 if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller'))) - uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_push.yaml@main + uses: ./.github/workflows/self-push-amd.yml with: gpu_flavor: mi210 secrets: inherit diff --git a/.github/workflows/self-push-amd-mi250-caller.yml b/.github/workflows/self-push-amd-mi250-caller.yml index ea2801f2c..2485a78c3 100644 --- a/.github/workflows/self-push-amd-mi250-caller.yml +++ b/.github/workflows/self-push-amd-mi250-caller.yml @@ -19,7 +19,7 @@ jobs: run_amd_ci: name: AMD mi250 if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller'))) - uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_push.yaml@main + uses: ./.github/workflows/self-push-amd.yml with: gpu_flavor: mi250 secrets: inherit diff --git a/.github/workflows/self-push-amd-mi300-caller.yml b/.github/workflows/self-push-amd-mi300-caller.yml index 08d915541..a8ee4e540 100644 --- a/.github/workflows/self-push-amd-mi300-caller.yml +++ b/.github/workflows/self-push-amd-mi300-caller.yml @@ -19,7 +19,7 @@ jobs: run_amd_ci: name: AMD mi300 if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && (startsWith(github.ref_name, 'run_amd_push_ci_caller') || startsWith(github.ref_name, 'mi300-ci')))) - uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_push.yaml@main + uses: ./.github/workflows/self-push-amd.yml with: gpu_flavor: mi300 secrets: inherit diff --git a/.github/workflows/self-push-amd.yml b/.github/workflows/self-push-amd.yml new file mode 100644 index 000000000..6931c2f3e --- /dev/null +++ b/.github/workflows/self-push-amd.yml @@ -0,0 +1,335 @@ +name: Self-hosted runner AMD GPU (push) + +on: + workflow_call: + inputs: + gpu_flavor: + required: true + type: string + +env: + HF_HOME: /mnt/cache + TRANSFORMERS_IS_CI: yes + OMP_NUM_THREADS: 8 + MKL_NUM_THREADS: 8 + PYTEST_TIMEOUT: 60 + TF_FORCE_GPU_ALLOW_GROWTH: true + RUN_PT_TF_CROSS_TESTS: 1 + HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }} + +jobs: + check_runner_status: + name: Check Runner Status + runs-on: ubuntu-22.04 + steps: + - name: Checkout transformers + uses: actions/checkout@v4 + with: + fetch-depth: 2 + + - name: Check Runner Status + run: python utils/check_self_hosted_runner.py --target_runners amd-mi210-single-gpu-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }} + + check_runners: + name: Check Runners + needs: check_runner_status + strategy: + matrix: + machine_type: [single-gpu, multi-gpu] + runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}'] + container: + image: huggingface/transformers-pytorch-amd-gpu-push-ci # <--- We test only for PyTorch for now + options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ + steps: + - name: ROCM-SMI + run: | + rocm-smi + - name: ROCM-INFO + run: | + rocminfo | grep "Agent" -A 14 + - name: Show ROCR environment + run: | + echo "ROCR: $ROCR_VISIBLE_DEVICES" + + setup_gpu: + name: Setup + needs: check_runners + strategy: + matrix: + machine_type: [single-gpu, multi-gpu] + runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}'] + container: + image: huggingface/transformers-pytorch-amd-gpu-push-ci # <--- We test only for PyTorch for now + options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ + outputs: + matrix: ${{ steps.set-matrix.outputs.matrix }} + test_map: ${{ steps.set-matrix.outputs.test_map }} + env: + # `CI_BRANCH_PUSH`: The branch name from the push event + # `CI_BRANCH_WORKFLOW_RUN`: The name of the branch on which this workflow is triggered by `workflow_run` event + # `CI_SHA_PUSH`: The commit SHA from the push event + # `CI_SHA_WORKFLOW_RUN`: The commit SHA that triggers this workflow by `workflow_run` event + CI_BRANCH_PUSH: ${{ github.event.ref }} + CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }} + CI_SHA_PUSH: ${{ github.event.head_commit.id }} + CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }} + steps: + # Necessary to get the correct branch name and commit SHA for `workflow_run` event + # We also take into account the `push` event (we might want to test some changes in a branch) + - name: Prepare custom environment variables + shell: bash + # `CI_BRANCH`: The non-empty branch name from the above two (one and only one of them is empty) + # `CI_SHA`: The non-empty commit SHA from the above two (one and only one of them is empty) + run: | + CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''} + echo $CI_BRANCH_PUSH + echo $CI_BRANCH_WORKFLOW_RUN + echo $CI_SHA_PUSH + echo $CI_SHA_WORKFLOW_RUN + [[ ! -z "$CI_BRANCH_PUSH" ]] && echo "CI_BRANCH=$CI_BRANCH_PUSH" >> $GITHUB_ENV || echo "CI_BRANCH=$CI_BRANCH_WORKFLOW_RUN" >> $GITHUB_ENV + [[ ! -z "$CI_SHA_PUSH" ]] && echo "CI_SHA=$CI_SHA_PUSH" >> $GITHUB_ENV || echo "CI_SHA=$CI_SHA_WORKFLOW_RUN" >> $GITHUB_ENV + + - name: print environment variables + run: | + echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}" + echo "env.CI_SHA = ${{ env.CI_SHA }}" + + - name: Update clone using environment variables + working-directory: /transformers + run: | + echo "original branch = $(git branch --show-current)" + git fetch && git checkout ${{ env.CI_BRANCH }} + echo "updated branch = $(git branch --show-current)" + git checkout ${{ env.CI_SHA }} + echo "log = $(git log -n 1)" + + - name: Cleanup + working-directory: /transformers + run: | + rm -rf tests/__pycache__ + rm -rf tests/models/__pycache__ + rm -rf reports + + - name: Show installed libraries and their versions + working-directory: /transformers + run: pip freeze + + - name: Fetch the tests to run + working-directory: /transformers + # TODO: add `git-python` in the docker images + run: | + pip install --upgrade git-python + python3 utils/tests_fetcher.py --diff_with_last_commit | tee test_preparation.txt + + - name: Report fetched tests + uses: actions/upload-artifact@v4 + with: + name: test_fetched + path: /transformers/test_preparation.txt + + - id: set-matrix + name: Organize tests into models + working-directory: /transformers + # The `keys` is used as GitHub actions matrix for jobs, i.e. `models/bert`, `tokenization`, `pipeline`, etc. + # The `test_map` is used to get the actual identified test files under each key. + # If no test to run (so no `test_map.json` file), create a dummy map (empty matrix will fail) + run: | + if [ -f test_map.json ]; then + keys=$(python3 -c 'import json; fp = open("test_map.json"); test_map = json.load(fp); fp.close(); d = list(test_map.keys()); print(d)') + test_map=$(python3 -c 'import json; fp = open("test_map.json"); test_map = json.load(fp); fp.close(); print(test_map)') + else + keys=$(python3 -c 'keys = ["dummy"]; print(keys)') + test_map=$(python3 -c 'test_map = {"dummy": []}; print(test_map)') + fi + echo $keys + echo $test_map + echo "matrix=$keys" >> $GITHUB_OUTPUT + echo "test_map=$test_map" >> $GITHUB_OUTPUT + + run_models_gpu: + name: Model tests + needs: setup_gpu + # `dummy` means there is no test to run + if: contains(fromJson(needs.setup_gpu.outputs.matrix), 'dummy') != true + strategy: + fail-fast: false + matrix: + folders: ${{ fromJson(needs.setup_gpu.outputs.matrix) }} + machine_type: [single-gpu, multi-gpu] + runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}'] + container: + image: huggingface/transformers-pytorch-amd-gpu-push-ci # <--- We test only for PyTorch for now + options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ + env: + # For the meaning of these environment variables, see the job `Setup` + CI_BRANCH_PUSH: ${{ github.event.ref }} + CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }} + CI_SHA_PUSH: ${{ github.event.head_commit.id }} + CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }} + steps: + # Necessary to get the correct branch name and commit SHA for `workflow_run` event + # We also take into account the `push` event (we might want to test some changes in a branch) + - name: Prepare custom environment variables + shell: bash + # For the meaning of these environment variables, see the job `Setup` + run: | + CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''} + echo $CI_BRANCH_PUSH + echo $CI_BRANCH_WORKFLOW_RUN + echo $CI_SHA_PUSH + echo $CI_SHA_WORKFLOW_RUN + [[ ! -z "$CI_BRANCH_PUSH" ]] && echo "CI_BRANCH=$CI_BRANCH_PUSH" >> $GITHUB_ENV || echo "CI_BRANCH=$CI_BRANCH_WORKFLOW_RUN" >> $GITHUB_ENV + [[ ! -z "$CI_SHA_PUSH" ]] && echo "CI_SHA=$CI_SHA_PUSH" >> $GITHUB_ENV || echo "CI_SHA=$CI_SHA_WORKFLOW_RUN" >> $GITHUB_ENV + + - name: print environment variables + run: | + echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}" + echo "env.CI_SHA = ${{ env.CI_SHA }}" + + - name: Update clone using environment variables + working-directory: /transformers + run: | + echo "original branch = $(git branch --show-current)" + git fetch && git checkout ${{ env.CI_BRANCH }} + echo "updated branch = $(git branch --show-current)" + git checkout ${{ env.CI_SHA }} + echo "log = $(git log -n 1)" + + - name: Reinstall transformers in edit mode (remove the one installed during docker image build) + working-directory: /transformers + run: python3 -m pip uninstall -y transformers && python3 -m pip install -e . + + - name: Echo folder ${{ matrix.folders }} + shell: bash + # For folders like `models/bert`, set an env. var. (`matrix_folders`) to `models_bert`, which will be used to + # set the artifact folder names (because the character `/` is not allowed). + run: | + echo "${{ matrix.folders }}" + echo "${{ fromJson(needs.setup_gpu.outputs.test_map)[matrix.folders] }}" + matrix_folders=${{ matrix.folders }} + matrix_folders=${matrix_folders/'models/'/'models_'} + echo "$matrix_folders" + echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV + + - name: ROCM-SMI + run: | + rocm-smi + - name: ROCM-INFO + run: | + rocminfo | grep "Agent" -A 14 + - name: Show ROCR environment + run: | + echo "ROCR: $ROCR_VISIBLE_DEVICES" + + - name: Environment + working-directory: /transformers + run: | + python3 utils/print_env.py + + - name: Show installed libraries and their versions + working-directory: /transformers + run: pip freeze + + - name: Run all non-slow selected tests on GPU + working-directory: /transformers + run: | + python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports ${{ fromJson(needs.setup_gpu.outputs.test_map)[matrix.folders] }} -m "not not_device_test" + + - name: Failure short reports + if: ${{ failure() }} + continue-on-error: true + run: cat /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt + + - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports" + if: ${{ always() }} + uses: actions/upload-artifact@v4 + with: + name: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports + path: /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports + + send_results: + name: Send results to webhook + runs-on: ubuntu-22.04 + if: always() + needs: [ + check_runner_status, + check_runners, + setup_gpu, + run_models_gpu, +# run_tests_torch_cuda_extensions_single_gpu, +# run_tests_torch_cuda_extensions_multi_gpu + ] + env: + # For the meaning of these environment variables, see the job `Setup` + CI_BRANCH_PUSH: ${{ github.event.ref }} + CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }} + CI_SHA_PUSH: ${{ github.event.head_commit.id }} + CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }} + steps: + - name: Preliminary job status + shell: bash + # For the meaning of these environment variables, see the job `Setup` + run: | + echo "Runner availability: ${{ needs.check_runner_status.result }}" + echo "Setup status: ${{ needs.setup_gpu.result }}" + echo "Runner status: ${{ needs.check_runners.result }}" + + # Necessary to get the correct branch name and commit SHA for `workflow_run` event + # We also take into account the `push` event (we might want to test some changes in a branch) + - name: Prepare custom environment variables + shell: bash + # For the meaning of these environment variables, see the job `Setup` + run: | + CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''} + echo $CI_BRANCH_PUSH + echo $CI_BRANCH_WORKFLOW_RUN + echo $CI_SHA_PUSH + echo $CI_SHA_WORKFLOW_RUN + [[ ! -z "$CI_BRANCH_PUSH" ]] && echo "CI_BRANCH=$CI_BRANCH_PUSH" >> $GITHUB_ENV || echo "CI_BRANCH=$CI_BRANCH_WORKFLOW_RUN" >> $GITHUB_ENV + [[ ! -z "$CI_SHA_PUSH" ]] && echo "CI_SHA=$CI_SHA_PUSH" >> $GITHUB_ENV || echo "CI_SHA=$CI_SHA_WORKFLOW_RUN" >> $GITHUB_ENV + + - name: print environment variables + run: | + echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}" + echo "env.CI_SHA = ${{ env.CI_SHA }}" + + - uses: actions/checkout@v4 + # To avoid failure when multiple commits are merged into `main` in a short period of time. + # Checking out to an old commit beyond the fetch depth will get an error `fatal: reference is not a tree: ... + # (Only required for `workflow_run` event, where we get the latest HEAD on `main` instead of the event commit) + with: + fetch-depth: 20 + + - name: Update clone using environment variables + run: | + echo "original branch = $(git branch --show-current)" + git fetch && git checkout ${{ env.CI_BRANCH }} + echo "updated branch = $(git branch --show-current)" + git checkout ${{ env.CI_SHA }} + echo "log = $(git log -n 1)" + + - uses: actions/download-artifact@v4 + - name: Send message to Slack + env: + CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }} + CI_SLACK_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID }} + CI_SLACK_CHANNEL_ID_DAILY: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }} + CI_SLACK_CHANNEL_ID_AMD: ${{ secrets.CI_SLACK_CHANNEL_ID_AMD }} + CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }} + CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_AMD }} + ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }} + CI_EVENT: Push CI (AMD) - ${{ inputs.gpu_flavor }} + CI_TITLE_PUSH: ${{ github.event.head_commit.message }} + CI_TITLE_WORKFLOW_RUN: ${{ github.event.workflow_run.head_commit.message }} + CI_SHA: ${{ env.CI_SHA }} + RUNNER_STATUS: ${{ needs.check_runner_status.result }} + RUNNER_ENV_STATUS: ${{ needs.check_runners.result }} + SETUP_STATUS: ${{ needs.setup_gpu.result }} + + # We pass `needs.setup_gpu.outputs.matrix` as the argument. A processing in `notification_service.py` to change + # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`. + run: | + pip install huggingface_hub + pip install slack_sdk + pip show slack_sdk + python utils/notification_service.py "${{ needs.setup_gpu.outputs.matrix }}"