[CI] Add profile test #2
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Linux Profile Test | |
| on: | |
| schedule: | |
| # GMT+8 0:00 Sunday | |
| - cron: '0 16 * * 6' | |
| pull_request: | |
| branches: | |
| - main | |
| workflow_dispatch: | |
| inputs: | |
| pytorch: | |
| required: false | |
| type: string | |
| default: 'nightly' | |
| description: Pytorch branch/commit | |
| python: | |
| required: false | |
| type: string | |
| default: '3.10' | |
| description: Python version | |
| runner: | |
| required: true | |
| type: string | |
| default: 'linux.idc.xpu' | |
| description: Runner label | |
| driver: | |
| required: false | |
| type: string | |
| default: 'lts' | |
| description: Driver lts/rolling | |
| nightly_whl: | |
| required: false | |
| type: string | |
| default: '' | |
| description: Pytorch nightly wheel version | |
| permissions: read-all | |
| jobs: | |
| Torch-XPU-Profile-Tests: | |
| runs-on: ${{ inputs.runner != '' && inputs.runner || 'linux.idc.xpu' }} | |
| # Don't run on forked repos | |
| if: ${{ github.repository_owner == 'intel' }} | |
| env: | |
| python: ${{ inputs.python != '' && inputs.python || '3.10' }} | |
| pytorch: ${{ inputs.pytorch != '' && inputs.pytorch || 'nightly' }} | |
| steps: | |
| - name: Checkout torch-xpu-ops | |
| uses: actions/checkout@v4 | |
| - name: Create unique Conda ENV name | |
| run: | | |
| echo "CONDA_ENV_NAME=profile_test_${ZE_AFFINITY_MASK}" >> $GITHUB_ENV | |
| - name: Prepare Conda ENV | |
| run: | | |
| which conda && conda clean -ay | |
| conda remove --all -y -n $CONDA_ENV_NAME || rm -rf $(dirname ${CONDA_EXE})/../envs/$CONDA_ENV_NAME | |
| conda create -n $CONDA_ENV_NAME python=${{ env.python }} cmake ninja -y | |
| source activate $CONDA_ENV_NAME | |
| pip install mkl-static==2025.0.1 mkl-include==2025.0.1 | |
| pip install pandas scipy tqdm | |
| - name: Prepare Stock Pytorch | |
| id: installed | |
| run: | | |
| pwd | |
| source activate $CONDA_ENV_NAME | |
| if [ -z "${{ inputs.nightly_whl }}" ]; then | |
| pip install torch torchvision torchaudio --pre --index-url https://download.pytorch.org/whl/nightly/xpu | |
| else | |
| pip install torch==$(echo ${{ inputs.nightly_whl }}) torchvision torchaudio --pre --index-url https://download.pytorch.org/whl/nightly/xpu | |
| fi | |
| echo "TORCH_BRANCH_ID=$(python -c 'import torch; print(torch.__version__)')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}" | |
| TORCH_COMMIT_ID=$(python -c 'import torch; print(torch.version.git_version)') | |
| echo "TORCH_COMMIT_ID=${TORCH_COMMIT_ID}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}" | |
| cd ../ && rm -rf pytorch | |
| git clone https://github.com/pytorch/pytorch pytorch | |
| cd pytorch && git checkout ${TORCH_COMMIT_ID} | |
| # apply PRs for stock pytorch | |
| pip install requests | |
| # python ../torch-xpu-ops/.github/scripts/apply_torch_pr.py | |
| git status && git show -s | |
| pip install -r requirements.txt | |
| TORCH_XPU_OPS_COMMIT=$(<third_party/xpu.txt) | |
| echo "TORCH_XPU_OPS_COMMIT=${TORCH_XPU_OPS_COMMIT}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}" | |
| rm -rf third_party/torch-xpu-ops | |
| git clone https://github.com/intel/torch-xpu-ops.git third_party/torch-xpu-ops | |
| cd third_party/torch-xpu-ops | |
| git checkout ${TORCH_XPU_OPS_COMMIT} | |
| cd ../../ | |
| python third_party/torch-xpu-ops/.github/scripts/apply_torch_pr.py | |
| - name: Show GITHUB_ENV | |
| run: | | |
| echo "$GITHUB_ENV" | |
| rm -rf ../pytorch/inductor_log | |
| rm -rf /tmp/torchinductor_* | |
| - name: Model Test | |
| if: github.event_name == 'schedule' && github.event.schedule == '0 17 * * 5' | |
| run: | | |
| source activate $CONDA_ENV_NAME | |
| PROFILE=1 python -u .github/scripts/profile/rn50.py -a resnet50 --dummy ./ --num-iterations 20 --xpu 0 | |
| - name: Official UT Test | |
| if: github.event_name == 'schedule' && github.event.schedule == '0 17 * * 5' | |
| run: | | |
| cd ${{ github.workspace }} | |
| rm -rf profile_test && mkdir -p profile_test | |
| source activate $CONDA_ENV_NAME | |
| cd ../pytorch/test/profiler | |
| python -m pytest -vs test_cpp_thread.py | tee ${{ github.workspace }}/profile_test/test_cpp_thread.log | |
| python -m pytest -vs test_execution_trace.py | tee ${{ github.workspace }}/profile_test/test_execution_trace.log | |
| python -m pytest -vs test_memory_profiler.py | tee ${{ github.workspace }}/profile_test/test_memory_profiler.log | |
| python -m pytest -vs test_profiler_tree.py | tee ${{ github.workspace }}/profile_test/test_profiler_tree.log | |
| - name: Issue Reproduce UT Test | |
| if: github.event_name == 'schedule' && github.event.schedule == '0 17 * * 5' | |
| run: | | |
| source activate $CONDA_ENV_NAME | |
| mkdir -p profile_test/issue_reproduce | |
| python -u torch-xpu-ops/.github/scripts/profile/correlation_id_mixed.py | tee ${{ github.workspace }}/profile_test/issue_reproduce/correlation_id_mixed.log | |
| python -u torch-xpu-ops/.github/scripts/profile/reproducer.missing.gpu.kernel.time.py | tee ${{ github.workspace }}/profile_test/issue_reproduce/reproducer.missing.gpu.kernel.time.log | |
| python -u torch-xpu-ops/.github/scripts/profile/time_precision_in_profile.py | tee ${{ github.workspace }}/profile_test/issue_reproduce/time_precision_in_profile.log | |
| python -u torch-xpu-ops/.github/scripts/profile/profile_partial_runtime_ops.py | tee ${{ github.workspace }}/profile_test/issue_reproduce/profile_partial_runtime_ops.log | |
| python -u torch-xpu-ops/.github/scripts/profile/triton_xpu_ops_time.py | tee ${{ github.workspace }}/profile_test/issue_reproduce/triton_xpu_ops_time.log | |
| - name: Print model results | |
| if: ${{ ! cancelled() }} | |
| run: | | |
| { | |
| sed -i '1d' .github/scripts/profile/profiling.fp32.train.pt | |
| head -n -2 .github/scripts/profile/profiling.fp32.train.pt > ${{ github.workspace }}/profile_test/profiling.rn50.fp32.pt | |
| echo "### Model Test" | |
| echo "| Name | Self CPU % | Self CPU | CPU total % | CPU total | CPU time avg | Self XPU | Self XPU % | XPU total | XPU time avg | of Calls |" | |
| echo "| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |" | |
| while read line; do | |
| if [[ "$(echo $line | cut -f2 -d" ")" =~ [0-9] ]]; then | |
| Name=$(echo $line | cut -f1 -d" ") | |
| else | |
| Name=$(echo $line | cut -f 1,2 -d" ") | |
| fi | |
| if [[ "$Name" == "$(echo $line | cut -f 1,2 -d" ")" ]]; then | |
| Self_CPU_percent=$(echo $line | cut -f3 -d" ") | |
| Self_CPU=$(echo $line | cut -f4 -d" ") | |
| CPU_total_percent=$(echo $line | cut -f5 -d" ") | |
| CPU_total=$(echo $line | cut -f6 -d" ") | |
| CPU_time_avg=$(echo $line | cut -f7 -d" ") | |
| Self_XPU=$(echo $line | cut -f8 -d" ") | |
| Self_XPU_percent=$(echo $line | cut -f9 -d" ") | |
| XPU_total=$(echo $line | cut -f10 -d" ") | |
| XPU_time_avg=$(echo $line | cut -f11 -d" ") | |
| of_Calls=$(echo $line | cut -f12 -d" ") | |
| else | |
| Self_CPU_percent=$(echo $line | cut -f2 -d" ") | |
| Self_CPU=$(echo $line | cut -f3 -d" ") | |
| CPU_total_percent=$(echo $line | cut -f4 -d" ") | |
| CPU_total=$(echo $line | cut -f5 -d" ") | |
| CPU_time_avg=$(echo $line | cut -f6 -d" ") | |
| Self_XPU=$(echo $line | cut -f7 -d" ") | |
| Self_XPU_percent=$(echo $line | cut -f8 -d" ") | |
| XPU_total=$(echo $line | cut -f9 -d" ") | |
| XPU_time_avg=$(echo $line | cut -f10 -d" ") | |
| of_Calls=$(echo $line | cut -f11 -d" ") | |
| fi | |
| echo "| $Name | $Self_CPU_percent | $Self_CPU | $CPU_total_percent | $CPU_total | $CPU_time_avg | $Self_XPU | $Self_XPU_percent | $XPU_total | $XPU_time_avg | $of_Calls |" | |
| done <${{ github.workspace }}/profile_test/profiling.rn50.fp32.pt | |
| } >> $GITHUB_STEP_SUMMARY | |
| - name: Print Issue Reproduce UT results | |
| if: ${{ ! cancelled() }} | |
| run: | | |
| cd ${{ github.workspace }}/profile_test/issue_reproduce | |
| { | |
| echo "### Issue reproduce Test" | |
| filelist=`ls *.log` | |
| for log in $filelist; do | |
| echo "$log" | |
| echo "| Name | Self CPU % | Self CPU | CPU total % | CPU total | CPU time avg | Self XPU | Self XPU % | XPU total | XPU time avg | of Calls |" | |
| echo "| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |" | |
| while read line; do | |
| if [[ "$(echo $line | cut -f2 -d" ")" =~ [0-9] ]]; then | |
| Name=$(echo $line | cut -f1 -d" ") | |
| else | |
| Name=$(echo $line | cut -f 1,2 -d" ") | |
| fi | |
| if [[ "$Name" == "$(echo $line | cut -f 1,2 -d" ")" ]]; then | |
| Self_CPU_percent=$(echo $line | cut -f3 -d" ") | |
| Self_CPU=$(echo $line | cut -f4 -d" ") | |
| CPU_total_percent=$(echo $line | cut -f5 -d" ") | |
| CPU_total=$(echo $line | cut -f6 -d" ") | |
| CPU_time_avg=$(echo $line | cut -f7 -d" ") | |
| Self_XPU=$(echo $line | cut -f8 -d" ") | |
| Self_XPU_percent=$(echo $line | cut -f9 -d" ") | |
| XPU_total=$(echo $line | cut -f10 -d" ") | |
| XPU_time_avg=$(echo $line | cut -f11 -d" ") | |
| of_Calls=$(echo $line | cut -f12 -d" ") | |
| else | |
| Self_CPU_percent=$(echo $line | cut -f2 -d" ") | |
| Self_CPU=$(echo $line | cut -f3 -d" ") | |
| CPU_total_percent=$(echo $line | cut -f4 -d" ") | |
| CPU_total=$(echo $line | cut -f5 -d" ") | |
| CPU_time_avg=$(echo $line | cut -f6 -d" ") | |
| Self_XPU=$(echo $line | cut -f7 -d" ") | |
| Self_XPU_percent=$(echo $line | cut -f8 -d" ") | |
| XPU_total=$(echo $line | cut -f9 -d" ") | |
| XPU_time_avg=$(echo $line | cut -f10 -d" ") | |
| of_Calls=$(echo $line | cut -f11 -d" ") | |
| fi | |
| echo "| $Name | $Self_CPU_percent | $Self_CPU | $CPU_total_percent | $CPU_total | $CPU_time_avg | $Self_XPU | $Self_XPU_percent | $XPU_total | $XPU_time_avg | $of_Calls |" | |
| done <$log | |
| done | |
| } >> $GITHUB_STEP_SUMMARY | |
| - name: Upload Profile Test log | |
| if: ${{ ! cancelled() }} | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| name: Torch-XPU-Profile-Log-${{ github.event.pull_request.number || github.sha }} | |
| path: | | |
| ${{ github.workspace }}/profile_test | |