AMD PyTorch Job #36
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: AMD PyTorch Job | |
| on: | |
| workflow_dispatch: | |
| inputs: | |
| script_content: | |
| description: 'Content of Python script' | |
| required: true | |
| type: string | |
| filename: | |
| description: 'Name of Python script' | |
| required: true | |
| type: string | |
| reference_content: | |
| description: 'Content of the reference code script (optional)' | |
| required: false | |
| type: string | |
| reference_filename: | |
| description: 'Name of reference script (supports .py or .cu)' | |
| required: false | |
| type: string | |
| eval_content: | |
| description: 'Content of the outer eval code script (optional)' | |
| required: false | |
| type: string | |
| eval_filename: | |
| description: 'Name of outer eval script (supports .py or .cu)' | |
| required: false | |
| type: string | |
| jobs: | |
| train: | |
| runs-on: [amdgpu-mi250-x86-64] | |
| timeout-minutes: 10 | |
| env: | |
| VENV_DIR: /groups/aig_sharks/pytorch_venv | |
| steps: | |
| - name: Setup Python | |
| uses: actions/setup-python@v5 | |
| with: | |
| python-version: '3.10' | |
| - name: Create script | |
| shell: python | |
| run: | | |
| with open('${{ github.event.inputs.filename }}', 'w') as f: | |
| f.write('''${{ github.event.inputs.script_content }}''') | |
| - name: Create reference scripts if provided | |
| shell: bash | |
| run: | | |
| if [[ -n "${{ github.event.inputs.reference_filename }}" ]]; then | |
| echo "Creating reference script..." | |
| cat > "${{ github.event.inputs.reference_filename }}" <<EOL | |
| ${{ github.event.inputs.reference_content }} | |
| EOL | |
| cat "${{ github.event.inputs.reference_filename }}" # Debug: Show file contents | |
| else | |
| echo "No reference content provided." | |
| fi | |
| - name: Create eval scripts if provided | |
| shell: bash | |
| run: | | |
| if [[ -n "${{ github.event.inputs.eval_filename }}" ]]; then | |
| echo "Creating reference script..." | |
| cat > "${{ github.event.inputs.eval_filename }}" <<EOL | |
| ${{ github.event.inputs.eval_content }} | |
| EOL | |
| cat "${{ github.event.inputs.eval_filename }}" # Debug: Show file contents | |
| else | |
| echo "No eval content provided." | |
| fi | |
| - name: Setup Virtual Environment and Install Dependencies | |
| run: | | |
| python -m venv ${VENV_DIR} | |
| source ${VENV_DIR}/bin/activate | |
| pip install --upgrade pip | |
| pip install --pre pytorch-triton-rocm==3.1.0+cf34004b8a torch==2.6.0.dev20241023+rocm6.2 --index-url https://download.pytorch.org/whl/nightly/rocm6.2 | |
| - name: Run script | |
| shell: bash | |
| run: | | |
| if [[ -n "${{ github.event.inputs.eval_content }}" ]]; then | |
| echo "Running Python file..." | |
| python3 "${{ github.event.inputs.eval_filename }}" > training.log 2>&1 | |
| cat training.log # Debug: show output | |
| else | |
| echo "Running Python file..." | |
| python3 "${{ github.event.inputs.filename }}" > training.log 2>&1 | |
| cat training.log # Debug: show output | |
| fi | |
| - name: Extract score | |
| shell: bash | |
| run: | | |
| # Extract only the last occurrence of "score:" | |
| score=$(grep -oP 'score:\s*\d+(\.\d+)?' training.log | tail -n 1 | awk '{print $2}') | |
| echo "Score extracted: $score" | |
| echo "::set-output name=score::$score" | |
| - name: Upload training artifacts | |
| uses: actions/upload-artifact@v4 | |
| if: always() | |
| with: | |
| name: training-artifacts | |
| path: | | |
| training.log | |
| ${{ github.event.inputs.filename }} |