fix

.github/workflows/build-docker-images.yml

@@ -3,7 +3,7 @@ name: Build docker images (scheduled)
 on:
   push:
     branches:
-      - build_ci_docker_image*
+      - build_docker_on_kube_2
   repository_dispatch:
   workflow_call:
     inputs:
@@ -18,199 +18,9 @@ concurrency:
   cancel-in-progress: false
 
 jobs:
-  latest-docker:
-    name: "Latest PyTorch + TensorFlow [dev]"
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Cleanup disk
-        run: |
-          sudo ls -l /usr/local/lib/
-          sudo ls -l /usr/share/
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-          sudo rm -rf /usr/local/lib/android
-          sudo rm -rf /usr/share/dotnet
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      -
-        name: Check out code
-        uses: actions/checkout@v3
-      -
-        name: Login to DockerHub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      -
-        name: Build and push
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-all-latest-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-all-latest-gpu${{ inputs.image_postfix }}
-      # Push CI images still need to be re-built daily
-      -
-        name: Build and push (for Push CI) in a daily basis
-        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-        if: inputs.image_postfix != '-push-ci'
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-all-latest-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-all-latest-gpu-push-ci
-
-  latest-torch-deepspeed-docker:
-    name: "Latest PyTorch + DeepSpeed"
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Cleanup disk
-        run: |
-          sudo ls -l /usr/local/lib/
-          sudo ls -l /usr/share/
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-          sudo rm -rf /usr/local/lib/android
-          sudo rm -rf /usr/share/dotnet
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      -
-        name: Check out code
-        uses: actions/checkout@v3
-      -
-        name: Login to DockerHub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      -
-        name: Build and push
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-pytorch-deepspeed-latest-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-pytorch-deepspeed-latest-gpu${{ inputs.image_postfix }}
-
-  # Can't build 2 images in a single job `latest-torch-deepspeed-docker` (for `nvcr.io/nvidia`)
-  latest-torch-deepspeed-docker-for-push-ci-daily-build:
-    name: "Latest PyTorch + DeepSpeed (Push CI - Daily Build)"
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Cleanup disk
-        run: |
-          sudo ls -l /usr/local/lib/
-          sudo ls -l /usr/share/
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-          sudo rm -rf /usr/local/lib/android
-          sudo rm -rf /usr/share/dotnet
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      -
-        name: Check out code
-        uses: actions/checkout@v3
-      -
-        name: Login to DockerHub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      # Push CI images still need to be re-built daily
-      -
-        name: Build and push (for Push CI) in a daily basis
-        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-        if: inputs.image_postfix != '-push-ci'
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-pytorch-deepspeed-latest-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
-
-  doc-builder:
-    name: "Doc builder"
-    # Push CI doesn't need this image
-    if: inputs.image_postfix != '-push-ci'
-    runs-on: ubuntu-22.04
-    steps:
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      -
-        name: Check out code
-        uses: actions/checkout@v3
-      -
-        name: Login to DockerHub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      -
-        name: Build and push
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-doc-builder
-          push: true
-          tags: huggingface/transformers-doc-builder
-
-  latest-pytorch:
-    name: "Latest PyTorch [dev]"
-    # Push CI doesn't need this image
-    if: inputs.image_postfix != '-push-ci'
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Cleanup disk
-        run: |
-          sudo ls -l /usr/local/lib/
-          sudo ls -l /usr/share/
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-          sudo rm -rf /usr/local/lib/android
-          sudo rm -rf /usr/share/dotnet
-          sudo du -sh /usr/local/lib/
-          sudo du -sh /usr/share/
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      -
-        name: Check out code
-        uses: actions/checkout@v3
-      -
-        name: Login to DockerHub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      -
-        name: Build and push
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-pytorch-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-pytorch-gpu
-
   latest-pytorch-amd:
     name: "Latest PyTorch (AMD) [dev]"
-    runs-on: [self-hosted, docker-gpu, amd-gpu, single-gpu, mi210]
+    runs-on: [self-hosted, yih-dar-shieh-debug-doctest]
     steps:
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
@@ -228,45 +38,4 @@ jobs:
           build-args: |
             REF=main
           push: true
-          tags: huggingface/transformers-pytorch-amd-gpu${{ inputs.image_postfix }}
-      # Push CI images still need to be re-built daily
-      -
-        name: Build and push (for Push CI) in a daily basis
-        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-        if: inputs.image_postfix != '-push-ci'
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-pytorch-amd-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-pytorch-amd-gpu-push-ci
-
-  latest-tensorflow:
-    name: "Latest TensorFlow [dev]"
-    # Push CI doesn't need this image
-    if: inputs.image_postfix != '-push-ci'
-    runs-on: ubuntu-22.04
-    steps:
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      -
-        name: Check out code
-        uses: actions/checkout@v3
-      -
-        name: Login to DockerHub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      -
-        name: Build and push
-        uses: docker/build-push-action@v5
-        with:
-          context: ./docker/transformers-tensorflow-gpu
-          build-args: |
-            REF=main
-          push: true
-          tags: huggingface/transformers-tensorflow-gpu
+          tags: huggingface/transformers-pytorch-amd-gpu${{ inputs.image_postfix }}

.github/workflows/self-push-amd-mi210-caller.yml

@@ -18,7 +18,7 @@ on:
 jobs:
   run_amd_ci:
     name: AMD mi210
-    if: (cancelled() != true) && ((github.event_name != 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
+    if: (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
     uses: ./.github/workflows/self-push-amd.yml
     with:
       gpu_flavor: mi210

.github/workflows/self-push-amd-mi250-caller.yml

@@ -18,7 +18,7 @@ on:
 jobs:
   run_amd_ci:
     name: AMD mi250
-    if: (cancelled() != true) && ((github.event_name != 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
+    if: (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
     uses: ./.github/workflows/self-push-amd.yml
     with:
       gpu_flavor: mi250

.github/workflows/self-scheduled-amd-caller.yml

@@ -0,0 +1,25 @@
+name: Self-hosted runner (AMD scheduled CI caller)
+
+on:
+  schedule:
+    - cron: "17 2 * * *"
+  push:
+    branches:
+      - run_amd_scheduled_ci_caller*
+
+jobs:
+  run_amd_ci_mi210:
+    name: AMD mi210
+    if: (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_scheduled_ci_caller')))
+    uses: ./.github/workflows/self-scheduled-amd.yml
+    with:
+      gpu_flavor: mi210
+    secrets: inherit
+
+  run_amd_ci_mi250:
+    name: AMD mi250
+    if: (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_scheduled_ci_caller')))
+    uses: ./.github/workflows/self-scheduled-amd.yml
+    with:
+      gpu_flavor: mi250
+    secrets: inherit

.github/workflows/self-scheduled-amd.yml

@@ -0,0 +1,443 @@
+name: Self-hosted runner (scheduled-amd)
+
+# Note: For the AMD CI, we rely on a caller workflow and on the workflow_call event to trigger the
+# CI in order to run it on both MI210 and MI250, without having to use matrix here which pushes
+# us towards the limit of allowed jobs on GitHub Actions.
+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
+  RUN_SLOW: yes
+  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
+
+
+# Important note: each job (run_tests_single_gpu, run_tests_multi_gpu, run_examples_gpu, run_pipelines_torch_gpu) requires all the previous jobs before running.
+# This is done so that we avoid parallelizing the scheduled tests, to leave available
+# runners for the push CI that is running on the same machine.
+jobs:
+  check_runner_status:
+    name: Check Runner Status
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 2
+
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners hf-amd-mi210-ci-1gpu-1,hf-amd-mi250-ci-1gpu-1 --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+  check_runners:
+    name: Check Runners
+    needs: check_runner_status
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: Show HIP environment
+        run: |
+          echo "HIP: $HIP_VISIBLE_DEVICES"
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+  setup:
+    name: Setup
+    needs: check_runners
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    outputs:
+      matrix: ${{ steps.set-matrix.outputs.matrix }}
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: |
+          git fetch && git checkout ${{ github.sha }}
+
+      - name: Cleanup
+        working-directory: /transformers
+        run: |
+          rm -rf tests/__pycache__
+          rm -rf tests/models/__pycache__
+          rm -rf reports
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - id: set-matrix
+        name: Identify models to test
+        working-directory: /transformers/tests
+        run: |
+          echo "matrix=$(python3 -c 'import os; tests = os.getcwd(); model_tests = os.listdir(os.path.join(tests, "models")); d1 = sorted(list(filter(os.path.isdir, os.listdir(tests)))); d2 = sorted(list(filter(os.path.isdir, [f"models/{x}" for x in model_tests]))); d1.remove("models"); d = d2 + d1; print(d)')" >> $GITHUB_OUTPUT
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: Show HIP environment
+        run: |
+          echo "HIP: $HIP_VISIBLE_DEVICES"
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+  run_tests_single_gpu:
+    name: Single GPU tests
+    strategy:
+      max-parallel: 1  # For now, not to parallelize. Can change later if it works well.
+      fail-fast: false
+      matrix:
+        folders: ${{ fromJson(needs.setup.outputs.matrix) }}
+        machine_type: [single-gpu]
+    runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - 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 }}"
+          matrix_folders=${{ matrix.folders }}
+          matrix_folders=${matrix_folders/'models/'/'models_'}
+          echo "$matrix_folders"
+          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
+
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - 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: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: Show HIP environment
+        run: |
+          echo "HIP: $HIP_VISIBLE_DEVICES"
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all tests on GPU
+        working-directory: /transformers
+        run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} tests/${{ matrix.folders }}
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
+
+      - name: Test suite reports artifacts
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v3
+        with:
+          name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
+
+  run_tests_multi_gpu:
+    name: Multi GPU tests
+    strategy:
+      max-parallel: 1
+      fail-fast: false
+      matrix:
+        folders: ${{ fromJson(needs.setup.outputs.matrix) }}
+        machine_type: [multi-gpu]
+    runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - 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 }}"
+          matrix_folders=${{ matrix.folders }}
+          matrix_folders=${matrix_folders/'models/'/'models_'}
+          echo "$matrix_folders"
+          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
+
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - 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: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: Show HIP environment
+        run: |
+          echo "HIP: $HIP_VISIBLE_DEVICES"
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all tests on GPU
+        working-directory: /transformers
+        run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} tests/${{ matrix.folders }}
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
+
+      - name: Test suite reports artifacts
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v3
+        with:
+          name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
+
+  run_examples_gpu:
+    name: Examples tests
+    strategy:
+      fail-fast: false
+      matrix:
+        machine_type: [single-gpu]
+    runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - 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: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: Show HIP environment
+        run: |
+          echo "HIP: $HIP_VISIBLE_DEVICES"
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run examples tests on GPU
+        working-directory: /transformers
+        run: |
+          pip install -r examples/pytorch/_tests_requirements.txt
+          python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_examples_gpu examples/pytorch
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_examples_gpu/failures_short.txt
+
+      - name: Test suite reports artifacts
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v3
+        with:
+          name: ${{ matrix.machine_type }}_run_examples_gpu
+          path: /transformers/reports/${{ matrix.machine_type }}_examples_gpu
+
+  run_pipelines_torch_gpu:
+    name: PyTorch pipelines tests
+    strategy:
+      fail-fast: false
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - 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: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: Show HIP environment
+        run: |
+          echo "HIP: $HIP_VISIBLE_DEVICES"
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all pipeline tests on GPU
+        working-directory: /transformers
+        run: |
+          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ matrix.machine_type }}_tests_torch_pipeline_gpu tests/pipelines
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_tests_torch_pipeline_gpu/failures_short.txt
+
+      - name: Test suite reports artifacts
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v3
+        with:
+          name: ${{ matrix.machine_type }}_run_tests_torch_pipeline_gpu
+          path: /transformers/reports/${{ matrix.machine_type }}_tests_torch_pipeline_gpu
+
+  run_extract_warnings:
+    name: Extract warnings in CI artifacts
+    runs-on: ubuntu-22.04
+    if: always()
+    needs: [
+      check_runner_status,
+      check_runners,
+      setup,
+      run_tests_single_gpu,
+      run_tests_multi_gpu,
+      run_examples_gpu,
+      run_pipelines_torch_gpu,
+      # run_all_tests_torch_cuda_extensions_gpu
+    ]
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 2
+
+      - name: Install transformers
+        run: pip install transformers
+
+      - name: Show installed libraries and their versions
+        run: pip freeze
+
+      - name: Create output directory
+        run: mkdir warnings_in_ci
+
+      - uses: actions/download-artifact@v3
+        with:
+          path: warnings_in_ci
+
+      - name: Show artifacts
+        run: echo "$(python3 -c 'import os; d = os.listdir(); print(d)')"
+        working-directory: warnings_in_ci
+
+      - name: Extract warnings in CI artifacts
+        run: |
+          python3 utils/extract_warnings.py --workflow_run_id ${{ github.run_id }} --output_dir warnings_in_ci --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }} --from_gh
+          echo "$(python3 -c 'import os; import json; fp = open("warnings_in_ci/selected_warnings.json"); d = json.load(fp); d = "\n".join(d) ;print(d)')"
+
+      - name: Upload artifact
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v3
+        with:
+          name: warnings_in_ci
+          path: warnings_in_ci/selected_warnings.json
+
+  send_results:
+    name: Send results to webhook
+    runs-on: ubuntu-22.04
+    if: always()
+    needs: [
+      check_runner_status,
+      check_runners,
+      setup,
+      run_tests_single_gpu,
+      run_tests_multi_gpu,
+      run_examples_gpu,
+      run_pipelines_torch_gpu,
+      # run_all_tests_torch_cuda_extensions_gpu,
+      run_extract_warnings
+    ]
+    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 "Runner status: ${{ needs.check_runners.result }}"
+          echo "Setup status: ${{ needs.setup.result }}"
+
+      - uses: actions/checkout@v3
+      - uses: actions/download-artifact@v3
+      - name: Send message to Slack
+        env:
+          CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
+          CI_SLACK_CHANNEL_ID_DAILY_AMD: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY_AMD }}
+          CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
+          CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY_AMD }}
+          ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+          CI_EVENT: Scheduled CI (AMD)
+          CI_SHA: ${{ github.sha }}
+          CI_WORKFLOW_REF: ${{ github.workflow_ref }}
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup.result }}
+        # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
+        # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
+        run: |
+          sudo apt-get install -y curl
+          pip install slack_sdk
+          pip show slack_sdk
+          python utils/notification_service.py "${{ needs.setup.outputs.matrix }}"
+
+      # Upload complete failure tables, as they might be big and only truncated versions could be sent to Slack.
+      - name: Failure table artifacts
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v3
+        with:
+          name: test_failure_tables
+          path: test_failure_tables

docker/transformers-pytorch-amd-gpu/Dockerfile

@@ -1,24 +1,23 @@
-FROM rocm/pytorch:rocm5.6_ubuntu20.04_py3.8_pytorch_2.0.1
+FROM rocm/dev-ubuntu-20.04:5.6
+# rocm/pytorch has no version with 2.1.0
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
+ARG PYTORCH='2.1.0'
+ARG TORCH_VISION='0.16.0'
+ARG TORCH_AUDIO='2.1.0'
+ARG ROCM='5.6'
+
 RUN apt update && \
-    apt install -y --no-install-recommends git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-pip ffmpeg && \
+    apt install -y --no-install-recommends git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-dev python3-pip ffmpeg && \
     apt clean && \
     rm -rf /var/lib/apt/lists/*
 
+RUN python3 -m pip install torch==$PYTORCH torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm$ROCM
+
 RUN python3 -m pip install --no-cache-dir --upgrade pip setuptools ninja git+https://github.com/facebookresearch/detectron2.git pytesseract "itsdangerous<2.1.0"
 
-# If set to nothing, will install the latest version
-ARG PYTORCH='2.0.1'
-ARG TORCH_VISION='0.15.2'
-ARG TORCH_AUDIO='2.0.2'
-ARG ROCM='5.6'
-
-RUN git clone --depth 1 --branch v$TORCH_AUDIO https://github.com/pytorch/audio.git
-RUN cd audio && USE_ROCM=1 USE_CUDA=0 python setup.py install
-
 ARG REF=main
 WORKDIR /
 RUN git clone https://github.com/huggingface/transformers && cd transformers && git checkout $REF

utils/notification_service.py

@@ -968,10 +968,15 @@ if __name__ == "__main__":
         "Torch CUDA extension tests": "run_tests_torch_cuda_extensions_gpu_test_reports",
     }
 
-    if ci_event in ["push", "Nightly CI"] or ci_event.startswith("Past CI") or ci_event.startswith("Push CI (AMD)"):
+    if ci_event in ["push", "Nightly CI"] or ci_event.startswith("Past CI"):
         del additional_files["Examples directory"]
         del additional_files["PyTorch pipelines"]
         del additional_files["TensorFlow pipelines"]
+    elif ci_event.startswith("Scheduled CI (AMD)"):
+        del additional_files["TensorFlow pipelines"]
+        del additional_files["Torch CUDA extension tests"]
+    elif ci_event.startswith("Push CI (AMD)"):
+        additional_files = {}
 
     additional_results = {
         key: {