BioNeMo Model Convergence Tests #238
This file contains hidden or 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: "BioNeMo Model Convergence Tests" | |
| on: | |
| workflow_dispatch: | |
| inputs: | |
| gpu_type: | |
| description: "GPU type to use" | |
| required: true | |
| default: "h100-sxm" | |
| type: choice | |
| options: | |
| - h200 | |
| - h100-sxm | |
| - a100-80gb | |
| model_config: | |
| description: "Model configuration to use" | |
| required: true | |
| default: "esm2_native_te_650m" | |
| type: choice | |
| options: | |
| - esm2_native_te_650m | |
| - esm2_native_te_3b | |
| - esm2_native_te_15b | |
| - codonfm_ptl_te | |
| branch: | |
| description: "Branch to use (ignored if commit SHA is provided)" | |
| required: true | |
| default: "main" | |
| type: string | |
| commit_sha: | |
| description: "Commit SHA (optional - overrides branch if provided)" | |
| required: false | |
| type: string | |
| schedule: | |
| - cron: "0 8 * * *" # everyday at 1am PST | |
| jobs: | |
| submit-lepton-jobs: | |
| runs-on: ubuntu-latest | |
| strategy: | |
| matrix: | |
| model_config: ${{ github.event_name == 'schedule' && fromJSON('["esm2_native_te_650m", "esm2_native_te_3b", "esm2_native_te_15b", "codonfm_ptl_te"]') || fromJSON(format('["{0}"]', github.event.inputs.model_config)) }} | |
| fail-fast: false | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v4 | |
| with: | |
| ref: ${{ github.event.inputs.commit_sha || github.event.inputs.branch || 'main' }} | |
| - name: Set up Python | |
| uses: actions/setup-python@v5 | |
| with: | |
| python-version: "3.11" | |
| cache: "pip" | |
| cache-dependency-path: | | |
| ci/lepton/requirements.txt | |
| - name: Install dependencies | |
| run: | | |
| python -m pip install --upgrade pip | |
| pip install -r ci/lepton/requirements.txt | |
| - name: Submit Lepton Jobs | |
| env: | |
| LEP_LOGIN_CREDENTIALS: ${{ secrets.LEP_LOGIN_CREDENTIALS }} | |
| GPU_TYPE: ${{ github.event.inputs.gpu_type || 'h100-sxm' }} | |
| BRANCH: ${{ github.event.inputs.branch || 'main' }} | |
| COMMIT_SHA: ${{ github.event.inputs.commit_sha || '' }} | |
| MODEL_CONFIG: ${{ matrix.model_config }} | |
| run: | | |
| set -euo pipefail | |
| lep login -c "$LEP_LOGIN_CREDENTIALS" || true | |
| # Map GPU type to node group | |
| if [ "$GPU_TYPE" = "h200" ]; then | |
| NODE_GROUP="nv-int-multiteam-nebius-h200-01" | |
| elif [ "$GPU_TYPE" = "h100-sxm" ]; then | |
| NODE_GROUP="yo-bom-lepton-001" | |
| elif [ "$GPU_TYPE" = "a100-80gb" ]; then | |
| NODE_GROUP="az-sat-lepton-001" | |
| else | |
| echo "Error: Unknown GPU type: $GPU_TYPE" | |
| exit 1 | |
| fi | |
| python ci/lepton/core/launch_job.py \ | |
| --config-path ../model_convergence/configs \ | |
| --config-name "recipes/$MODEL_CONFIG" \ | |
| branch="$BRANCH" \ | |
| commit_sha=$COMMIT_SHA \ | |
| node_group=$NODE_GROUP \ | |
| gpu_type="$GPU_TYPE" |