CI - Bazel Optional H100 and B200 CUDA tests #666
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: CI - Bazel Optional H100 and B200 CUDA tests | |
| on: | |
| # Runs on PR if label "CI Optional GPU Presubmit" is present. | |
| workflow_dispatch: | |
| inputs: | |
| halt-for-connection: | |
| description: 'Should this workflow run wait for a remote connection?' | |
| type: choice | |
| required: true | |
| default: 'no' | |
| options: | |
| - 'yes' | |
| - 'no' | |
| pull_request: | |
| branches: | |
| - main | |
| types: [ labeled, synchronize ] | |
| schedule: | |
| - cron: "0 */2 * * *" # Run once every 2 hours | |
| permissions: | |
| contents: read | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.head_ref || github.ref }} | |
| # Don't cancel in-progress jobs for main/release branches. | |
| cancel-in-progress: ${{ !contains(github.ref, 'release/') && github.ref != 'main' }} | |
| jobs: | |
| run_tests: | |
| if: ${{ github.event.repository.fork == false && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch' || contains(github.event.pull_request.labels.*.name, 'CI Optional GPU Presubmit')) }} | |
| runs-on: linux-x86-a4-224-b200-1gpu | |
| container: 'us-docker.pkg.dev/ml-oss-artifacts-published/ml-public-container/ml-build-cuda12.8-cudnn9.8:latest' | |
| name: "Bazel single B200 CUDA tests" | |
| # End Presubmit Naming Check github-cuda-presubmits | |
| steps: | |
| - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2 | |
| with: | |
| persist-credentials: false | |
| - name: Wait For Connection | |
| uses: google-ml-infra/actions/ci_connection@7f5ca0c263a81ed09ea276524c1b9192f1304e3c | |
| with: | |
| halt-dispatch-input: ${{ inputs.halt-for-connection }} | |
| - name: Run Bazel single B200 CUDA Tests | |
| run: | | |
| nvidia-smi | |
| bazel test --config=rbe_linux_x86_64_cuda \ | |
| --config=resultstore \ | |
| --config=rbe_cache \ | |
| --repo_env=HERMETIC_CUDA_VERSION="12.8.0" \ | |
| --repo_env=HERMETIC_CUDNN_VERSION="9.8.0" \ | |
| --repo_env=HERMETIC_PYTHON_VERSION="3.13" \ | |
| --test_env=XLA_PYTHON_CLIENT_ALLOCATOR=platform \ | |
| --run_under "$(pwd)/build/parallel_accelerator_execute.sh" \ | |
| --test_output=errors \ | |
| --test_tag_filters=-multiaccelerator \ | |
| --test_env=JAX_ACCELERATOR_COUNT=1 \ | |
| --test_env=JAX_TESTS_PER_ACCELERATOR=8 \ | |
| --strategy=TestRunner=local \ | |
| --local_test_jobs=8 \ | |
| --test_env=JAX_EXCLUDE_TEST_TARGETS='PmapTest.testSizeOverflow|.*InterpretTest.*' \ | |
| --test_env=TF_CPP_MIN_LOG_LEVEL=0 \ | |
| --test_env=JAX_SKIP_SLOW_TESTS=true \ | |
| --action_env=JAX_ENABLE_X64="1" \ | |
| --action_env=NCCL_DEBUG=WARN \ | |
| --flaky_test_attempts=1 \ | |
| --test_timeout=420 \ | |
| --color=yes \ | |
| //tests:cudnn_fusion_test_gpu \ | |
| //tests:scaled_matmul_stablehlo_test_gpu \ | |
| //tests:fused_attention_stablehlo_test_gpu \ | |
| //tests:nn_test_gpu \ | |
| //tests/pallas:gpu_tests \ | |
| //tests/mosaic:gpu_tests | |
| run_multiaccelerator_tests: | |
| if: ${{ github.event.repository.fork == false && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch' || contains(github.event.pull_request.labels.*.name, 'CI Optional GPU Presubmit')) }} | |
| runs-on: linux-x86-a3-8g-h100-8gpu | |
| container: 'us-docker.pkg.dev/ml-oss-artifacts-published/ml-public-container/ml-build-cuda12.8-cudnn9.8:latest' | |
| name: "Bazel multiple H100 CUDA tests" | |
| steps: | |
| - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2 | |
| with: | |
| persist-credentials: false | |
| - name: Wait For Connection | |
| uses: google-ml-infra/actions/ci_connection@7f5ca0c263a81ed09ea276524c1b9192f1304e3c | |
| with: | |
| halt-dispatch-input: ${{ inputs.halt-for-connection }} | |
| - name: Run Bazel multiple H100 CUDA Tests | |
| run: | | |
| nvidia-smi | |
| bazel test --config=rbe_linux_x86_64_cuda \ | |
| --config=resultstore \ | |
| --config=rbe_cache \ | |
| --repo_env=HERMETIC_CUDA_VERSION="12.8.0" \ | |
| --repo_env=HERMETIC_CUDNN_VERSION="9.8.0" \ | |
| --repo_env=HERMETIC_PYTHON_VERSION="3.13" \ | |
| --test_env=XLA_PYTHON_CLIENT_ALLOCATOR=platform \ | |
| --test_output=errors \ | |
| --strategy=TestRunner=local \ | |
| --local_test_jobs=8 \ | |
| --test_env=JAX_EXCLUDE_TEST_TARGETS='PmapTest.testSizeOverflow|.*InterpretTest.*' \ | |
| --test_tag_filters=multiaccelerator \ | |
| --test_env=TF_CPP_MIN_LOG_LEVEL=0 \ | |
| --test_env=JAX_SKIP_SLOW_TESTS=true \ | |
| --action_env=JAX_ENABLE_X64="1" \ | |
| --action_env=NCCL_DEBUG=WARN \ | |
| --flaky_test_attempts=1 \ | |
| --color=yes \ | |
| //tests/mosaic:gpu_tests \ | |
| //tests/pallas:gpu_tests \ | |
| //tests:array_interoperability_test_gpu \ | |
| //tests:cudnn_fusion_test_gpu \ | |
| //tests:fused_attention_stablehlo_test_gpu \ | |
| //tests:gpu_tests \ | |
| //tests:python_callback_test_gpu \ | |
| //tests:ragged_collective_test_gpu |