Compare commits

...

1 Commits

Author SHA1 Message Date
c5b6f0b8fb Add initial version for run_tests_multi_gpu 2023-09-22 16:58:13 +02:00

View File

@ -244,6 +244,101 @@ jobs:
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: Model tests
needs: setup
# `dummy` means there is no test to run
if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true
strategy:
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 }}' ]
container:
image: transformers-pytorch-amd-gpu-push-ci
options: --device /dev/kfd "$(./get_gpu.sh)" --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
steps:
# Necessary to get the correct branch name and commit SHA for `workflow_run` event
# We also take into account the `push` event (we might want to test some changes in a branch)
- name: Prepare custom environment variables
shell: bash
# For the meaning of these environment variables, see the job `Setup`
run: |
CI_BRANCH_PUSH=${{ github.event.ref }}
CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
CI_SHA_PUSH=${{ github.event.head_commit.id }}
CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
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 "${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}"
matrix_folders=${{ matrix.folders }}
matrix_folders=${matrix_folders/'models/'/'models_'}
echo "$matrix_folders"
echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
- name: ROCM-SMI
run: |
rocm-smi
- name: Environment
working-directory: /transformers
run: |
python3 utils/print_env.py
- name: Show installed libraries and their versions
working-directory: /transformers
run: pip freeze
- name: Run all non-slow selected tests on GPU
env:
MKL_SERVICE_FORCE_INTEL: 1
working-directory: /transformers
run: |
python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
- name: Failure short reports
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 }}
send_results:
name: Send results to webhook
runs-on: ubuntu-latest