SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix missing job action button in CI report (#17270)

Browse Source 
* use matrix.machine_type

* fix job names used in job_link

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
This commit is contained in:
Yih-Dar
2022-05-17 08:31:06 +02:00
committed by GitHub
parent 5a9957358c
commit 1ac2b8fa7f
3 changed files with 62 additions and 106 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
.github/workflows/self-push.yml vendored
View File

@@ -73,7 +73,7 @@ jobs:
echo "::set-output name=test_map::$test_map" echo "::set-output name=test_map::$test_map"
run_tests_single_gpu: run_tests_single_gpu:
name: Model Tests on single GPU name: Model tests
needs: setup needs: setup
# `dummy` means there is no test to run # `dummy` means there is no test to run
if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true
@@ -81,8 +81,8 @@ jobs:
fail-fast: false fail-fast: false
matrix: matrix:
folders: ${{ fromJson(needs.setup.outputs.matrix) }} folders: ${{ fromJson(needs.setup.outputs.matrix) }}
machines: [single-gpu] machine_type: [single-gpu]
runs-on: [self-hosted, docker-gpu, '${{ matrix.machines }}'] runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
container: container:
image: huggingface/transformers-all-latest-gpu image: huggingface/transformers-all-latest-gpu
options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -117,22 +117,22 @@ jobs:
- name: Run all non-slow selected tests on GPU - name: Run all non-slow selected tests on GPU
working-directory: /transformers working-directory: /transformers
run: | run: |
python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machines }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }} python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat /transformers/reports/${{ matrix.machines }}_tests_gpu_${{ matrix.folders }}/failures_short.txt run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ matrix.machines }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
path: /transformers/reports/${{ matrix.machines }}_tests_gpu_${{ matrix.folders }} path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
run_tests_multi_gpu: run_tests_multi_gpu:
name: Model Tests on multi GPUs name: Model tests
needs: setup needs: setup
# `dummy` means there is no test to run # `dummy` means there is no test to run
if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true
@@ -140,8 +140,8 @@ jobs:
fail-fast: false fail-fast: false
matrix: matrix:
folders: ${{ fromJson(needs.setup.outputs.matrix) }} folders: ${{ fromJson(needs.setup.outputs.matrix) }}
machines: [multi-gpu] machine_type: [multi-gpu]
runs-on: [self-hosted, docker-gpu, '${{ matrix.machines }}'] runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
container: container:
image: huggingface/transformers-all-latest-gpu image: huggingface/transformers-all-latest-gpu
options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -178,29 +178,29 @@ jobs:
MKL_SERVICE_FORCE_INTEL: 1 MKL_SERVICE_FORCE_INTEL: 1
working-directory: /transformers working-directory: /transformers
run: | run: |
python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machines }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }} python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat /transformers/reports/${{ matrix.machines }}_tests_gpu_${{ matrix.folders }}/failures_short.txt run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ matrix.machines }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
path: /transformers/reports/${{ matrix.machines }}_tests_gpu_${{ matrix.folders }} path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
run_tests_torch_cuda_extensions_single_gpu: run_tests_torch_cuda_extensions_single_gpu:
name: Torch CUDA extension tests on single GPU name: Torch CUDA extension tests
needs: setup needs: setup
if: contains(fromJson(needs.setup.outputs.matrix), 'deepspeed') || contains(fromJson(needs.setup.outputs.matrix), 'extended') if: contains(fromJson(needs.setup.outputs.matrix), 'deepspeed') || contains(fromJson(needs.setup.outputs.matrix), 'extended')
strategy: strategy:
fail-fast: false fail-fast: false
matrix: matrix:
machines: [single-gpu] machine_type: [single-gpu]
runs-on: [self-hosted, docker-gpu, '${{ matrix.machines }}'] runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
container: container:
image: nvcr.io/nvidia/pytorch:21.03-py3 image: nvcr.io/nvidia/pytorch:21.03-py3
options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -227,29 +227,29 @@ jobs:
- name: Run all non-slow selected tests on GPU - name: Run all non-slow selected tests on GPU
# TODO: Here we pass all tests in the 2 folders for simplicity. It's better to pass only the identified tests. # TODO: Here we pass all tests in the 2 folders for simplicity. It's better to pass only the identified tests.
run: | run: |
python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ matrix.machines }}_tests_torch_cuda_extensions_gpu tests/deepspeed tests/extended python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu tests/deepspeed tests/extended
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat reports/${{ matrix.machines }}_tests_torch_cuda_extensions_gpu/failures_short.txt run: cat reports/${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ matrix.machines }}_run_tests_torch_cuda_extensions_gpu_test_reports name: ${{ matrix.machine_type }}_run_tests_torch_cuda_extensions_gpu_test_reports
path: reports/${{ matrix.machines }}_tests_torch_cuda_extensions_gpu path: reports/${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu
run_tests_torch_cuda_extensions_multi_gpu: run_tests_torch_cuda_extensions_multi_gpu:
name: Torch CUDA extension tests on multi GPUs name: Torch CUDA extension tests
needs: setup needs: setup
if: contains(fromJson(needs.setup.outputs.matrix), 'deepspeed') || contains(fromJson(needs.setup.outputs.matrix), 'extended') if: contains(fromJson(needs.setup.outputs.matrix), 'deepspeed') || contains(fromJson(needs.setup.outputs.matrix), 'extended')
strategy: strategy:
fail-fast: false fail-fast: false
matrix: matrix:
machines: [multi-gpu] machine_type: [multi-gpu]
runs-on: [self-hosted, docker-gpu, '${{ matrix.machines }}'] runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
container: container:
image: nvcr.io/nvidia/pytorch:21.03-py3 image: nvcr.io/nvidia/pytorch:21.03-py3
options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -277,19 +277,19 @@ jobs:
- name: Run all non-slow selected tests on GPU - name: Run all non-slow selected tests on GPU
# TODO: Here we pass all tests in the 2 folders for simplicity. It's better to pass only the identified tests. # TODO: Here we pass all tests in the 2 folders for simplicity. It's better to pass only the identified tests.
run: | run: |
python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ matrix.machines }}_tests_torch_cuda_extensions_gpu tests/deepspeed tests/extended python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu tests/deepspeed tests/extended
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat reports/${{ matrix.machines }}_tests_torch_cuda_extensions_gpu/failures_short.txt run: cat reports/${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ matrix.machines }}_run_tests_torch_cuda_extensions_gpu_test_reports name: ${{ matrix.machine_type }}_run_tests_torch_cuda_extensions_gpu_test_reports
path: reports/${{ matrix.machines }}_tests_torch_cuda_extensions_gpu path: reports/${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu
send_results: send_results:
name: Send results to webhook name: Send results to webhook

109
.github/workflows/self-scheduled.yml vendored
View File

@@ -26,8 +26,8 @@ jobs:
name: Setup name: Setup
strategy: strategy:
matrix: matrix:
machines: [multi-gpu-docker, single-gpu-docker] machine_type: [single-gpu, multi-gpu]
runs-on: ${{ matrix.machines }} runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
container: container:
image: huggingface/transformers-all-latest-gpu image: huggingface/transformers-all-latest-gpu
options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -69,8 +69,8 @@ jobs:
fail-fast: false fail-fast: false
matrix: matrix:
folders: ${{ fromJson(needs.setup.outputs.matrix) }} folders: ${{ fromJson(needs.setup.outputs.matrix) }}
machines: [single-gpu-docker] machine_type: [single-gpu]
runs-on: ${{ matrix.machines }} runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
container: container:
image: huggingface/transformers-all-latest-gpu image: huggingface/transformers-all-latest-gpu
options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -87,34 +87,25 @@ jobs:
echo "$matrix_folders" echo "$matrix_folders"
echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
# Set machine type, i.e. `single-gpu` or `multi-gpu`. Here we just remove `-docker`.
- name: Set machine type from ${{ matrix.machines }}
shell: bash
run: |
machine_type=${{ matrix.machines }}
machine_type=${machine_type/'-docker'/''}
echo "machine_type=$machine_type"
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Update clone - name: Update clone
working-directory: /transformers working-directory: /transformers
run: git fetch && git checkout ${{ github.sha }} run: git fetch && git checkout ${{ github.sha }}
- name: Run all tests on GPU - name: Run all tests on GPU
working-directory: /transformers working-directory: /transformers
run: python3 -m pytest -v --make-reports=${{ env.machine_type }}_tests_gpu_${{ matrix.folders }} tests/${{ matrix.folders }} run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} tests/${{ matrix.folders }}
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ env.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
path: /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }} path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
run_tests_multi_gpu: run_tests_multi_gpu:
name: Model tests name: Model tests
@@ -122,8 +113,8 @@ jobs:
fail-fast: false fail-fast: false
matrix: matrix:
folders: ${{ fromJson(needs.setup.outputs.matrix) }} folders: ${{ fromJson(needs.setup.outputs.matrix) }}
machines: [multi-gpu-docker] machine_type: [multi-gpu]
runs-on: ${{ matrix.machines }} runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
container: container:
image: huggingface/transformers-all-latest-gpu image: huggingface/transformers-all-latest-gpu
options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -140,34 +131,25 @@ jobs:
echo "$matrix_folders" echo "$matrix_folders"
echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
# Set machine type, i.e. `single-gpu` or `multi-gpu`. Here we just remove `-docker`.
- name: Set machine type from ${{ matrix.machines }}
shell: bash
run: |
machine_type=${{ matrix.machines }}
machine_type=${machine_type/'-docker'/''}
echo "machine_type=$machine_type"
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Update clone - name: Update clone
working-directory: /transformers working-directory: /transformers
run: git fetch && git checkout ${{ github.sha }} run: git fetch && git checkout ${{ github.sha }}
- name: Run all tests on GPU - name: Run all tests on GPU
working-directory: /transformers working-directory: /transformers
run: python3 -m pytest -v --make-reports=${{ env.machine_type }}_tests_gpu_${{ matrix.folders }} tests/${{ matrix.folders }} run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} tests/${{ matrix.folders }}
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ env.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
path: /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }} path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
run_examples_gpu: run_examples_gpu:
name: Examples directory name: Examples directory
@@ -204,22 +186,13 @@ jobs:
strategy: strategy:
fail-fast: false fail-fast: false
matrix: matrix:
machines: [multi-gpu-docker, single-gpu-docker] machine_type: [single-gpu, multi-gpu]
runs-on: ${{ matrix.machines }} runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
container: container:
image: huggingface/transformers-pytorch-gpu image: huggingface/transformers-pytorch-gpu
options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
needs: setup needs: setup
steps: steps:
# Set machine type, i.e. `single-gpu` or `multi-gpu`. Here we just remove `-docker`.
- name: Set machine type from ${{ matrix.machines }}
shell: bash
run: |
machine_type=${{ matrix.machines }}
machine_type=${machine_type/'-docker'/''}
echo "machine_type=$machine_type"
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Update clone - name: Update clone
working-directory: /transformers working-directory: /transformers
run: git fetch && git checkout ${{ github.sha }} run: git fetch && git checkout ${{ github.sha }}
@@ -229,41 +202,32 @@ jobs:
env: env:
RUN_PIPELINE_TESTS: yes RUN_PIPELINE_TESTS: yes
run: | run: |
python3 -m pytest -n 1 -v --dist=loadfile -m is_pipeline_test --make-reports=${{ env.machine_type }}_tests_torch_pipeline_gpu tests python3 -m pytest -n 1 -v --dist=loadfile -m is_pipeline_test --make-reports=${{ matrix.machine_type }}_tests_torch_pipeline_gpu tests
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat /transformers/reports/${{ env.machine_type }}_tests_torch_pipeline_gpu/failures_short.txt run: cat /transformers/reports/${{ matrix.machine_type }}_tests_torch_pipeline_gpu/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ env.machine_type }}_run_tests_torch_pipeline_gpu name: ${{ matrix.machine_type }}_run_tests_torch_pipeline_gpu
path: /transformers/reports/${{ env.machine_type }}_tests_torch_pipeline_gpu path: /transformers/reports/${{ matrix.machine_type }}_tests_torch_pipeline_gpu
run_pipelines_tf_gpu: run_pipelines_tf_gpu:
name: TensorFlow pipelines name: TensorFlow pipelines
strategy: strategy:
fail-fast: false fail-fast: false
matrix: matrix:
machines: [multi-gpu-docker, single-gpu-docker] machine_type: [single-gpu, multi-gpu]
runs-on: ${{ matrix.machines }} runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
container: container:
image: huggingface/transformers-tensorflow-gpu image: huggingface/transformers-tensorflow-gpu
options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
needs: setup needs: setup
steps: steps:
# Set machine type, i.e. `single-gpu` or `multi-gpu`. Here we just remove `-docker`.
- name: Set machine type from ${{ matrix.machines }}
shell: bash
run: |
machine_type=${{ matrix.machines }}
machine_type=${machine_type/'-docker'/''}
echo "machine_type=$machine_type"
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Update clone - name: Update clone
working-directory: /transformers working-directory: /transformers
run: | run: |
@@ -274,41 +238,32 @@ jobs:
env: env:
RUN_PIPELINE_TESTS: yes RUN_PIPELINE_TESTS: yes
run: | run: |
python3 -m pytest -n 1 -v --dist=loadfile -m is_pipeline_test --make-reports=${{ env.machine_type }}_tests_tf_pipeline_gpu tests python3 -m pytest -n 1 -v --dist=loadfile -m is_pipeline_test --make-reports=${{ matrix.machine_type }}_tests_tf_pipeline_gpu tests
- name: Failure short reports - name: Failure short reports
if: ${{ always() }} if: ${{ always() }}
run: | run: |
cat /transformers/reports/${{ env.machine_type }}_tests_tf_pipeline_gpu/failures_short.txt cat /transformers/reports/${{ matrix.machine_type }}_tests_tf_pipeline_gpu/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ env.machine_type }}_run_tests_tf_pipeline_gpu name: ${{ matrix.machine_type }}_run_tests_tf_pipeline_gpu
path: /transformers/reports/${{ env.machine_type }}_tests_tf_pipeline_gpu path: /transformers/reports/${{ matrix.machine_type }}_tests_tf_pipeline_gpu
run_all_tests_torch_cuda_extensions_gpu: run_all_tests_torch_cuda_extensions_gpu:
name: Torch CUDA extension tests name: Torch CUDA extension tests
strategy: strategy:
fail-fast: false fail-fast: false
matrix: matrix:
machines: [multi-gpu-docker, single-gpu-docker] machine_type: [single-gpu, multi-gpu]
runs-on: ${{ matrix.machines }} runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
needs: setup needs: setup
container: container:
image: huggingface/transformers-pytorch-deepspeed-latest-gpu image: huggingface/transformers-pytorch-deepspeed-latest-gpu
options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/ options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
steps: steps:
# Set machine type, i.e. `single-gpu` or `multi-gpu`. Here we just remove `-docker`.
- name: Set machine type from ${{ matrix.machines }}
shell: bash
run: |
machine_type=${{ matrix.machines }}
machine_type=${machine_type/'-docker'/''}
echo "machine_type=$machine_type"
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Update clone - name: Update clone
working-directory: /workspace/transformers working-directory: /workspace/transformers
run: git fetch && git checkout ${{ github.sha }} run: git fetch && git checkout ${{ github.sha }}
@@ -324,19 +279,19 @@ jobs:
- name: Run all tests on GPU - name: Run all tests on GPU
working-directory: /workspace/transformers working-directory: /workspace/transformers
run: | run: |
python -m pytest -v --make-reports=${{ env.machine_type }}_tests_torch_cuda_extensions_gpu tests/deepspeed tests/extended python -m pytest -v --make-reports=${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu tests/deepspeed tests/extended
- name: Failure short reports - name: Failure short reports
if: ${{ failure() }} if: ${{ failure() }}
continue-on-error: true continue-on-error: true
run: cat /workspace/transformers/reports/${{ env.machine_type }}_tests_torch_cuda_extensions_gpu/failures_short.txt run: cat /workspace/transformers/reports/${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu/failures_short.txt
- name: Test suite reports artifacts - name: Test suite reports artifacts
if: ${{ always() }} if: ${{ always() }}
uses: actions/upload-artifact@v2 uses: actions/upload-artifact@v2
with: with:
name: ${{ env.machine_type }}_run_tests_torch_cuda_extensions_gpu_test_reports name: ${{ matrix.machine_type }}_run_tests_torch_cuda_extensions_gpu_test_reports
path: /workspace/transformers/reports/${{ env.machine_type }}_tests_torch_cuda_extensions_gpu path: /workspace/transformers/reports/${{ matrix.machine_type }}_tests_torch_cuda_extensions_gpu
send_results: send_results:

3
utils/notification_service.py
View File

@@ -621,7 +621,8 @@ if __name__ == "__main__":
if "stats" in artifact: if "stats" in artifact:
# Link to the GitHub Action job # Link to the GitHub Action job
model_results[model]["job_link"] = github_actions_job_links.get( model_results[model]["job_link"] = github_actions_job_links.get(
f"Model tests ({model}, {artifact_path['gpu']}-gpu)" # The job names use `matrix.folder` which contain things like `models/bert` instead of `models_bert`
f"Model tests ({model.replace('models_', 'models/')}, {artifact_path['gpu']}-gpu)"
) )
failed, success, time_spent = handle_test_results(artifact["stats"]) failed, success, time_spent = handle_test_results(artifact["stats"])