Upstream/main nxp/eiex 490 upstream contiguous memory format support for aten.clone #705
Workflow file for this run
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
| # Test ExecuTorch CUDA Build Compatibility | |
| # This workflow tests whether ExecuTorch can be successfully built with CUDA support | |
| # across different CUDA versions (12.6, 12.8, 12.9) using the command: | |
| # CMAKE_ARGS="-DEXECUTORCH_BUILD_CUDA=ON" ./install_executorch.sh | |
| # | |
| # Note: ExecuTorch automatically detects the system CUDA version using nvcc and | |
| # installs the appropriate PyTorch wheel. No manual CUDA/PyTorch installation needed. | |
| name: Test CUDA Builds | |
| on: | |
| pull_request: | |
| push: | |
| branches: | |
| - main | |
| - release/* | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}-${{ github.event_name == 'schedule' }} | |
| cancel-in-progress: false | |
| jobs: | |
| test-cuda-builds: | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| cuda-version: ["12.6", "12.8", "13.0"] | |
| name: test-executorch-cuda-build-${{ matrix.cuda-version }} | |
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main | |
| permissions: | |
| id-token: write | |
| contents: read | |
| with: | |
| timeout: 90 | |
| runner: linux.g5.4xlarge.nvidia.gpu | |
| gpu-arch-type: cuda | |
| gpu-arch-version: ${{ matrix.cuda-version }} | |
| use-custom-docker-registry: false | |
| submodules: recursive | |
| ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }} | |
| script: | | |
| set -eux | |
| # Test ExecuTorch CUDA build - ExecuTorch will automatically detect CUDA version | |
| # and install the appropriate PyTorch wheel when CMAKE_ARGS="-DEXECUTORCH_BUILD_CUDA=ON" | |
| source .ci/scripts/test-cuda-build.sh "${{ matrix.cuda-version }}" | |
| # This job will fail if any of the CUDA versions fail | |
| check-all-cuda-builds: | |
| needs: test-cuda-builds | |
| runs-on: ubuntu-latest | |
| if: always() | |
| steps: | |
| - name: Check if all CUDA builds succeeded | |
| run: | | |
| if [[ "${{ needs.test-cuda-builds.result }}" != "success" ]]; then | |
| echo "ERROR: One or more ExecuTorch CUDA builds failed!" | |
| echo "CUDA build results: ${{ needs.test-cuda-builds.result }}" | |
| exit 1 | |
| else | |
| echo "SUCCESS: All ExecuTorch CUDA builds (12.6, 12.8, 12.9) completed successfully!" | |
| fi | |
| test-models-cuda: | |
| name: test-models-cuda | |
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main | |
| permissions: | |
| id-token: write | |
| contents: read | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| model: [linear, add, add_mul, resnet18, conv1d] | |
| with: | |
| timeout: 90 | |
| runner: linux.g5.4xlarge.nvidia.gpu | |
| gpu-arch-type: cuda | |
| gpu-arch-version: 12.6 | |
| use-custom-docker-registry: false | |
| submodules: recursive | |
| ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }} | |
| script: | | |
| set -eux | |
| PYTHON_EXECUTABLE=python CMAKE_ARGS="-DEXECUTORCH_BUILD_CUDA=ON" ./install_executorch.sh | |
| export LD_LIBRARY_PATH=/opt/conda/lib:$LD_LIBRARY_PATH | |
| PYTHON_EXECUTABLE=python source .ci/scripts/test_model.sh "${{ matrix.model }}" cmake cuda | |
| export-voxtral-cuda-artifact: | |
| name: export-voxtral-cuda-artifact | |
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main | |
| permissions: | |
| id-token: write | |
| contents: read | |
| secrets: inherit | |
| strategy: | |
| fail-fast: false | |
| with: | |
| timeout: 90 | |
| secrets-env: EXECUTORCH_HF_TOKEN | |
| runner: linux.g5.4xlarge.nvidia.gpu | |
| gpu-arch-type: cuda | |
| gpu-arch-version: 12.6 | |
| use-custom-docker-registry: false | |
| submodules: recursive | |
| upload-artifact: voxtral-cuda-export | |
| ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }} | |
| script: | | |
| set -eux | |
| echo "::group::Setup ExecuTorch" | |
| CMAKE_ARGS="-DEXECUTORCH_BUILD_CUDA=ON" ./install_executorch.sh | |
| echo "::endgroup::" | |
| echo "::group::Setup Huggingface" | |
| pip install -U "huggingface_hub[cli]" accelerate | |
| huggingface-cli login --token $SECRET_EXECUTORCH_HF_TOKEN | |
| OPTIMUM_ET_VERSION=$(cat .ci/docker/ci_commit_pins/optimum-executorch.txt) | |
| pip install git+https://github.com/huggingface/optimum-executorch.git@${OPTIMUM_ET_VERSION} | |
| pip install mistral-common librosa | |
| pip list | |
| echo "::endgroup::" | |
| echo "::group::Export Voxtral" | |
| optimum-cli export executorch \ | |
| --model "mistralai/Voxtral-Mini-3B-2507" \ | |
| --task "multimodal-text-to-text" \ | |
| --recipe "cuda" \ | |
| --dtype bfloat16 \ | |
| --device cuda \ | |
| --max_seq_len 1024 \ | |
| --output_dir ./ | |
| python -m executorch.extension.audio.mel_spectrogram \ | |
| --feature_size 128 \ | |
| --stack_output \ | |
| --max_audio_len 300 \ | |
| --output_file voxtral_preprocessor.pte | |
| test -f model.pte | |
| test -f aoti_cuda_blob.ptd | |
| test -f voxtral_preprocessor.pte | |
| echo "::endgroup::" | |
| echo "::group::Store Voxtral Artifacts" | |
| mkdir -p "${RUNNER_ARTIFACT_DIR}" | |
| cp model.pte "${RUNNER_ARTIFACT_DIR}/" | |
| cp aoti_cuda_blob.ptd "${RUNNER_ARTIFACT_DIR}/" | |
| cp voxtral_preprocessor.pte "${RUNNER_ARTIFACT_DIR}/" | |
| ls -al "${RUNNER_ARTIFACT_DIR}" | |
| echo "::endgroup::" | |
| benchmark-voxtral-cuda: | |
| name: benchmark-voxtral-cuda | |
| needs: export-voxtral-cuda-artifact | |
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main | |
| permissions: | |
| id-token: write | |
| contents: read | |
| strategy: | |
| fail-fast: false | |
| with: | |
| timeout: 90 | |
| runner: linux.g5.4xlarge.nvidia.gpu | |
| gpu-arch-type: cuda | |
| gpu-arch-version: 12.6 | |
| use-custom-docker-registry: false | |
| submodules: recursive | |
| download-artifact: voxtral-cuda-export | |
| ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }} | |
| script: | | |
| set -eux | |
| echo "::group::Setup ExecuTorch Requirements" | |
| CMAKE_ARGS="-DEXECUTORCH_BUILD_CUDA=ON" ./install_requirements.sh | |
| pip list | |
| echo "::endgroup::" | |
| echo "::group::Prepare Voxtral Artifacts" | |
| cp "${RUNNER_ARTIFACT_DIR}/model.pte" . | |
| cp "${RUNNER_ARTIFACT_DIR}/aoti_cuda_blob.ptd" . | |
| ls -al model.pte aoti_cuda_blob.ptd | |
| echo "::endgroup::" | |
| echo "::group::Build Voxtral Benchmark" | |
| cmake -DCMAKE_BUILD_TYPE=Release \ | |
| -DEXECUTORCH_BUILD_CUDA=ON \ | |
| -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \ | |
| -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \ | |
| -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \ | |
| -DEXECUTORCH_BUILD_TESTS=ON \ | |
| -Bcmake-out . | |
| cmake --build cmake-out -j$(( $(nproc) - 1 )) --target voxtral_runner | |
| echo "::endgroup::" | |
| echo "::group::Run Voxtral Benchmark" | |
| export LD_LIBRARY_PATH=/opt/conda/lib:$LD_LIBRARY_PATH | |
| cmake-out/backends/cuda/voxtral_runner model.pte aoti_cuda_blob.ptd | |
| echo "::endgroup::" | |
| test-voxtral-cuda-e2e: | |
| name: test-voxtral-cuda-e2e | |
| needs: export-voxtral-cuda-artifact | |
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main | |
| permissions: | |
| id-token: write | |
| contents: read | |
| strategy: | |
| fail-fast: false | |
| with: | |
| timeout: 90 | |
| runner: linux.g5.4xlarge.nvidia.gpu | |
| gpu-arch-type: cuda | |
| gpu-arch-version: 12.6 | |
| use-custom-docker-registry: false | |
| submodules: recursive | |
| download-artifact: voxtral-cuda-export | |
| ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }} | |
| script: | | |
| set -eux | |
| echo "::group::Setup ExecuTorch Requirements" | |
| CMAKE_ARGS="-DEXECUTORCH_BUILD_CUDA=ON" ./install_requirements.sh | |
| pip list | |
| echo "::endgroup::" | |
| echo "::group::Prepare Voxtral Artifacts" | |
| cp "${RUNNER_ARTIFACT_DIR}/model.pte" . | |
| cp "${RUNNER_ARTIFACT_DIR}/aoti_cuda_blob.ptd" . | |
| cp "${RUNNER_ARTIFACT_DIR}/voxtral_preprocessor.pte" . | |
| TOKENIZER_URL="https://huggingface.co/mistralai/Voxtral-Mini-3B-2507/resolve/main/tekken.json" | |
| curl -L $TOKENIZER_URL -o tekken.json | |
| ls -al model.pte aoti_cuda_blob.ptd voxtral_preprocessor.pte tekken.json | |
| echo "::endgroup::" | |
| echo "::group::Download Test Audio File" | |
| AUDIO_URL="https://github.com/voxserv/audio_quality_testing_samples/raw/refs/heads/master/testaudio/16000/test01_20s.wav" | |
| curl -L $AUDIO_URL -o poem.wav | |
| echo "::endgroup::" | |
| echo "::group::Build Voxtral Runner" | |
| cmake --preset llm \ | |
| -DEXECUTORCH_BUILD_CUDA=ON \ | |
| -DCMAKE_INSTALL_PREFIX=cmake-out \ | |
| -DCMAKE_BUILD_TYPE=Release \ | |
| -Bcmake-out -S. | |
| cmake --build cmake-out -j$(( $(nproc) - 1 )) --target install --config Release | |
| cmake -DEXECUTORCH_BUILD_CUDA=ON \ | |
| -DCMAKE_BUILD_TYPE=Release \ | |
| -Sexamples/models/voxtral \ | |
| -Bcmake-out/examples/models/voxtral/ | |
| cmake --build cmake-out/examples/models/voxtral --target voxtral_runner --config Release | |
| echo "::endgroup::" | |
| echo "::group::Run Voxtral Runner" | |
| set +e | |
| export LD_LIBRARY_PATH=/opt/conda/lib:$LD_LIBRARY_PATH | |
| OUTPUT=$(cmake-out/examples/models/voxtral/voxtral_runner \ | |
| --model_path model.pte \ | |
| --data_path aoti_cuda_blob.ptd \ | |
| --tokenizer_path tekken.json \ | |
| --audio_path poem.wav \ | |
| --processor_path voxtral_preprocessor.pte \ | |
| --temperature 0 2>&1) | |
| EXIT_CODE=$? | |
| set -e | |
| echo "$OUTPUT" | |
| if ! echo "$OUTPUT" | grep -iq "poem"; then | |
| echo "Expected output 'poem' not found in output" | |
| exit 1 | |
| fi | |
| if [ $EXIT_CODE -ne 0 ]; then | |
| echo "Unexpected exit code: $EXIT_CODE" | |
| exit $EXIT_CODE | |
| fi | |
| echo "::endgroup::" |