transformers/.github/workflows/self-scheduled-intel-gaudi.yml
2025-06-23 10:56:51 +02:00

346 lines
12 KiB
YAML

name: Self-hosted runner (scheduled-intel-gaudi)
on:
workflow_call:
inputs:
job:
required: true
type: string
slack_report_channel:
required: true
type: string
runner_scale_set:
required: true
type: string
ci_event:
required: true
type: string
report_repo_id:
required: true
type: string
env:
NUM_SLICES: 2
RUN_SLOW: yes
PT_HPU_LAZY_MODE: 0
TRANSFORMERS_IS_CI: yes
PT_ENABLE_INT64_SUPPORT: 1
HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
HF_HOME: /mnt/cache/.cache/huggingface
jobs:
setup:
if: contains(fromJSON('["run_models_gpu", "run_trainer_and_fsdp_gpu"]'), inputs.job)
name: Setup
runs-on: ubuntu-latest
outputs:
slice_ids: ${{ steps.set-matrix.outputs.slice_ids }}
folder_slices: ${{ steps.set-matrix.outputs.folder_slices }}
quantization_matrix: ${{ steps.set-matrix.outputs.quantization_matrix }}
steps:
- name: Checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: "3.10"
- id: set-matrix
if: contains(fromJSON('["run_models_gpu", "run_trainer_and_fsdp_gpu"]'), inputs.job)
name: Identify models to test
working-directory: tests
run: |
if [ "${{ inputs.job }}" = "run_models_gpu" ]; then
echo "folder_slices=$(python3 ../utils/split_model_tests.py --num_splits ${{ env.NUM_SLICES }})" >> $GITHUB_OUTPUT
echo "slice_ids=$(python3 -c 'd = list(range(${{ env.NUM_SLICES }})); print(d)')" >> $GITHUB_OUTPUT
elif [ "${{ inputs.job }}" = "run_trainer_and_fsdp_gpu" ]; then
echo "folder_slices=[['trainer'], ['fsdp']]" >> $GITHUB_OUTPUT
echo "slice_ids=[0, 1]" >> $GITHUB_OUTPUT
fi
- id: set-matrix-quantization
if: ${{ inputs.job == 'run_quantization_torch_gpu' }}
name: Identify quantization method to test
working-directory: tests
run: |
echo "quantization_matrix=$(python3 -c 'import os; tests = os.getcwd(); quantization_tests = os.listdir(os.path.join(tests, "quantization")); d = sorted(list(filter(os.path.isdir, [f"quantization/{x}" for x in quantization_tests]))) ; print(d)')" >> $GITHUB_OUTPUT
run_models_gpu:
if: ${{ inputs.job == 'run_models_gpu' }}
name: " "
needs: setup
strategy:
fail-fast: false
matrix:
machine_type: [1gaudi, 2gaudi]
slice_id: ${{ fromJSON(needs.setup.outputs.slice_ids) }}
uses: ./.github/workflows/model_jobs_intel_gaudi.yml
with:
slice_id: ${{ matrix.slice_id }}
machine_type: ${{ matrix.machine_type }}
folder_slices: ${{ needs.setup.outputs.folder_slices }}
runner: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
report_name_prefix: run_models_gpu
secrets: inherit
run_trainer_and_fsdp_gpu:
if: ${{ inputs.job == 'run_trainer_and_fsdp_gpu' }}
name: " "
needs: setup
strategy:
fail-fast: false
matrix:
machine_type: [1gaudi, 2gaudi]
slice_id: ${{ fromJSON(needs.setup.outputs.slice_ids) }}
uses: ./.github/workflows/model_jobs_intel_gaudi.yml
with:
slice_id: ${{ matrix.slice_id }}
machine_type: ${{ matrix.machine_type }}
folder_slices: ${{ needs.setup.outputs.folder_slices }}
runner: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
report_name_prefix: run_trainer_and_fsdp_gpu
secrets: inherit
run_pipelines_gpu:
if: ${{ inputs.job == 'run_pipelines_gpu' }}
name: Pipelines
strategy:
fail-fast: false
matrix:
machine_type: [1gaudi, 2gaudi]
runs-on:
group: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
container:
image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
options: --runtime=habana
-v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
--env OMPI_MCA_btl_vader_single_copy_mechanism=none
--env HABANA_VISIBLE_DEVICES
--env HABANA_VISIBLE_MODULES
--cap-add=sys_nice
--shm-size=64G
steps:
- name: Checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install dependencies
run: |
pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn librosa soundfile
- name: HL-SMI
run: |
hl-smi
echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
- name: Environment
run: python3 utils/print_env.py
- name: Show installed libraries and their versions
run: pip freeze
- name: Set `machine_type` for report and artifact names
shell: bash
run: |
if [ "${{ matrix.machine_type }}" = "1gaudi" ]; then
machine_type=single-gpu
elif [ "${{ matrix.machine_type }}" = "2gaudi" ]; then
machine_type=multi-gpu
else
machine_type=${{ matrix.machine_type }}
fi
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Run all pipeline tests on Intel Gaudi
run: |
python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_pipelines_gpu_test_reports tests/pipelines -m "not not_device_test"
- name: Failure short reports
if: ${{ failure() }}
continue-on-error: true
run: |
cat reports/${{ env.machine_type }}_run_pipelines_gpu_test_reports/failures_short.txt
- name: "Test suite reports artifacts: ${{ env.machine_type }}_run_pipelines_gpu_test_reports"
if: ${{ always() }}
uses: actions/upload-artifact@v4
with:
name: ${{ env.machine_type }}_run_pipelines_gpu_test_reports
path: reports/${{ env.machine_type }}_run_pipelines_gpu_test_reports
run_examples_gpu:
if: ${{ inputs.job == 'run_examples_gpu' }}
name: Examples directory
strategy:
fail-fast: false
matrix:
machine_type: [1gaudi]
runs-on:
group: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
container:
image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
options: --runtime=habana
-v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
--env OMPI_MCA_btl_vader_single_copy_mechanism=none
--env HABANA_VISIBLE_DEVICES
--env HABANA_VISIBLE_MODULES
--cap-add=sys_nice
--shm-size=64G
steps:
- name: Checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install dependencies
run: |
pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn librosa soundfile
- name: HL-SMI
run: |
hl-smi
echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
- name: Environment
run: |
python3 utils/print_env.py
- name: Show installed libraries and their versions
run: |
pip freeze
- name: Set `machine_type` for report and artifact names
shell: bash
run: |
if [ "${{ matrix.machine_type }}" = "1gaudi" ]; then
machine_type=single-gpu
elif [ "${{ matrix.machine_type }}" = "2gaudi" ]; then
machine_type=multi-gpu
else
machine_type=${{ matrix.machine_type }}
fi
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Run examples tests on Intel Gaudi
run: |
pip install -r examples/pytorch/_tests_requirements.txt
python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_examples_gpu_test_reports examples/pytorch -m "not not_device_test"
- name: Failure short reports
if: ${{ failure() }}
continue-on-error: true
run: |
cat reports/${{ env.machine_type }}_run_examples_gpu_test_reports/failures_short.txt
- name: "Test suite reports artifacts: ${{ env.machine_type }}_run_examples_gpu_test_reports"
if: ${{ always() }}
uses: actions/upload-artifact@v4
with:
name: ${{ env.machine_type }}_run_examples_gpu_test_reports
path: reports/${{ env.machine_type }}_run_examples_gpu_test_reports
run_deepspeed_gpu:
if: ${{ inputs.job == 'run_deepspeed_gpu' }}
name: Intel Gaudi deepspeed tests
strategy:
fail-fast: false
matrix:
machine_type: [1gaudi, 2gaudi]
runs-on:
group: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
container:
image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
options: --runtime=habana
-v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
--env OMPI_MCA_btl_vader_single_copy_mechanism=none
--env HABANA_VISIBLE_DEVICES
--env HABANA_VISIBLE_MODULES
--cap-add=sys_nice
--shm-size=64G
steps:
- name: Checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install dependencies
run: |
pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn librosa soundfile
pip install git+https://github.com/HabanaAI/DeepSpeed.git@1.20.0
- name: HL-SMI
run: |
hl-smi
echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
- name: Environment
run: |
python3 utils/print_env.py
- name: Show installed libraries and their versions
run: |
pip freeze
- name: Set `machine_type` for report and artifact names
shell: bash
run: |
if [ "${{ matrix.machine_type }}" = "1gaudi" ]; then
machine_type=single-gpu
elif [ "${{ matrix.machine_type }}" = "2gaudi" ]; then
machine_type=multi-gpu
else
machine_type=${{ matrix.machine_type }}
fi
echo "machine_type=$machine_type" >> $GITHUB_ENV
- name: Run all deepspeed tests on intel Gaudi
run: |
python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_deepspeed_gpu_test_reports tests/deepspeed -m "not not_device_test"
- name: Failure short reports
if: ${{ failure() }}
continue-on-error: true
run: |
cat reports/${{ env.machine_type }}_run_deepspeed_gpu_test_reports/failures_short.txt
- name: "Test suite reports artifacts: ${{ env.machine_type }}_run_deepspeed_gpu_test_reports"
if: ${{ always() }}
uses: actions/upload-artifact@v4
with:
name: ${{ env.machine_type }}_run_deepspeed_gpu_test_reports
path: reports/${{ env.machine_type }}_run_deepspeed_gpu_test_reports
send_results:
name: Slack Report
needs:
[
setup,
run_models_gpu,
run_examples_gpu,
run_pipelines_gpu,
run_deepspeed_gpu,
run_trainer_and_fsdp_gpu,
]
if: ${{ always() }}
uses: ./.github/workflows/slack-report.yml
with:
job: ${{ inputs.job }}
setup_status: ${{ needs.setup.result }}
slack_report_channel: ${{ inputs.slack_report_channel }}
quantization_matrix: ${{ needs.setup.outputs.quantization_matrix }}
folder_slices: ${{ needs.setup.outputs.folder_slices }}
report_repo_id: ${{ inputs.report_repo_id }}
ci_event: ${{ inputs.ci_event }}
secrets: inherit