ci: Enable GB200 runners #2017
+86
−96
There are no files selected for viewing
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
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
| Original file line number | Diff line number | Diff line change | ||||
|---|---|---|---|---|---|---|
|
|
@@ -14,13 +14,10 @@ | |||||
| name: "CICD NeMo RL" | ||||||
|
|
||||||
| on: | ||||||
| push: | ||||||
| branches: | ||||||
| - main | ||||||
| - "pull-request/[0-9]+" | ||||||
|
Contributor
There was a problem hiding this comment. 🧩 Analysis chain🏁 Script executed: find . -name "cicd-main.yml" -type fRepository: NVIDIA-NeMo/RL Length of output: 91 🏁 Script executed: cat -n .github/workflows/cicd-main.yml | head -30Repository: NVIDIA-NeMo/RL Length of output: 1155 🌐 Web query:
💡 Result:
Example: on:
push:
branches:
- "releases/**"
- "!releases/**-alpha"Source: GitHub Docs — Workflow syntax for GitHub Actions (see References Fix push branch pattern: glob patterns don't interpret Line 20 uses 🔧 Suggested change- - "pull-request/[0-9]+"
+ - "pull-request/**"📝 Committable suggestion
Suggested change
🤖 Prompt for AI Agents |
||||||
| schedule: | ||||||
| - cron: "0 9 * * *" | ||||||
| workflow_dispatch: | ||||||
|
|
@@ -128,6 +125,18 @@ jobs: | |||||
| fi | ||||||
| echo "image_tag=$IMAGE_TAG" | tee -a "$GITHUB_OUTPUT" | ||||||
|
|
||||||
| org-member-pre-flight: | ||||||
| uses: NVIDIA-NeMo/FW-CI-templates/.github/workflows/_cicd_preflight.yml@fd82c6b23b5987d226f00d0719560f6e91210021 | ||||||
| with: | ||||||
| default_runner_prefix: ${{ vars.DEFAULT_RUNNER_PREFIX }} | ||||||
| non_nvidia_runner_prefix: ${{ vars.NON_NVIDIA_RUNNER_PREFIX }} | ||||||
| default_test_data_path: ${{ vars.DEFAULT_TEST_DATA_PATH }} | ||||||
| non_nvidia_test_data_path: ${{ vars.NON_NVIDIA_TEST_DATA_PATH }} | ||||||
| default_registry: ${{ vars.DEFAULT_CONTAINER_REGISTRY }} | ||||||
| non_nvidia_registry: ${{ vars.NON_NVIDIA_CONTAINER_REGISTRY }} | ||||||
| secrets: | ||||||
| NVIDIA_MANAGEMENT_ORG_PAT: ${{ secrets.NVIDIA_MANAGEMENT_ORG_PAT }} | ||||||
|
|
||||||
|
Comment on lines
+128
to
+139
Contributor
There was a problem hiding this comment. Include This new job now provides core runner/registry/data-path orchestration, but 🔧 Suggested change CI_QA_Gate:
@@
needs:
- pre-flight
+ - org-member-pre-flight
- pr-branch-up-to-date-check
- lint-check
@@
ALL_SUCCESS: >-
${{
+ needs.org-member-pre-flight.result == 'success' &&
needs.lint-check.result == 'success' &&
(needs.pr-branch-up-to-date-check.result == 'success' || needs.pr-branch-up-to-date-check.result == 'skipped') &&🤖 Prompt for AI Agents |
||||||
| pr-branch-up-to-date-check: | ||||||
| name: Check if PR branch is up to date | ||||||
| needs: [pre-flight] | ||||||
|
|
@@ -227,14 +236,16 @@ jobs: | |||||
|
|
||||||
| build-container: | ||||||
| if: ${{ needs.pre-flight.outputs.test_level != 'none' && needs.pre-flight.outputs.image_tag == '' }} | ||||||
| needs: [pre-flight, org-member-pre-flight] | ||||||
| uses: NVIDIA-NeMo/FW-CI-templates/.github/workflows/_build_container.yml@44284233576b11eb867ae55ac41fb291debc414d | ||||||
| with: | ||||||
| build-ref: ${{ github.sha }} | ||||||
| image-name: ${{ vars.CI_CONTAINER_NAME }} | ||||||
| dockerfile: docker/Dockerfile | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| image-label: ${{ vars.CI_CONTAINER_NAME }} | ||||||
| target: release | ||||||
| registry: ${{ needs.org-member-pre-flight.outputs.registry }} | ||||||
| build-contexts: | | ||||||
| nemo-rl=${{ github.run_id }}/ | ||||||
| build-args: | | ||||||
|
|
@@ -247,8 +258,8 @@ jobs: | |||||
| matrix: | ||||||
| include: | ||||||
| - script: Docs_Tests | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| needs: [pre-flight, build-container, org-member-pre-flight] | ||||||
| if: ${{ contains('docs L0 L1 L2', needs.pre-flight.outputs.test_level) }} | ||||||
| runs-on: ${{ matrix.runner }} | ||||||
| name: ${{ matrix.is_optional && 'PLEASEFIXME_' || '' }}${{ matrix.script }} | ||||||
|
|
@@ -260,6 +271,9 @@ jobs: | |||||
| uses: ./.github/actions/test-template | ||||||
| with: | ||||||
| runner: ${{ runner.name }} | ||||||
| registry: ${{ needs.org-member-pre-flight.outputs.registry }} | ||||||
| image: ${{ vars.CI_CONTAINER_NAME }} | ||||||
| test_data_path: ${{ needs.org-member-pre-flight.outputs.test_data_path }} | ||||||
| script: ${{ matrix.script }} | ||||||
| is_doc_test: "true" | ||||||
| is_fork_pr: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name != github.event.pull_request.base.repo.full_name }} | ||||||
|
|
@@ -270,12 +284,12 @@ jobs: | |||||
| matrix: | ||||||
| include: | ||||||
| - script: L0_Unit_Tests_Generation | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| - script: L0_Unit_Tests_Policy | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| - script: L0_Unit_Tests_Other | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| needs: [pre-flight, build-container, cicd-doc-tests, org-member-pre-flight] | ||||||
| if: >- | ||||||
| ${{ | ||||||
| ( | ||||||
|
|
@@ -298,6 +312,9 @@ jobs: | |||||
| with: | ||||||
| runner: ${{ runner.name }} | ||||||
| script: ${{ matrix.script }} | ||||||
| registry: ${{ needs.org-member-pre-flight.outputs.registry }} | ||||||
| test_data_path: ${{ needs.org-member-pre-flight.outputs.test_data_path }} | ||||||
| image: ${{ vars.CI_CONTAINER_NAME }} | ||||||
| image-tag: ${{ needs.pre-flight.outputs.image_tag }} | ||||||
| is_unit_test: "true" | ||||||
| cpu-only: ${{ matrix.cpu-only || false }} | ||||||
|
|
@@ -309,8 +326,8 @@ jobs: | |||||
| matrix: | ||||||
| include: | ||||||
| - script: L1_Functional_Tests_GPU | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| needs: [pre-flight, build-container, cicd-unit-tests, org-member-pre-flight] | ||||||
| runs-on: ${{ matrix.runner }} | ||||||
| if: ${{ contains('L1 L2', needs.pre-flight.outputs.test_level) }} | ||||||
| name: ${{ matrix.is_optional && 'PLEASEFIXME_' || '' }}${{ matrix.script }} | ||||||
|
|
@@ -324,6 +341,9 @@ jobs: | |||||
| HF_TOKEN: ${{ secrets.HF_TOKEN }} | ||||||
| with: | ||||||
| runner: ${{ runner.name }} | ||||||
| registry: ${{ needs.org-member-pre-flight.outputs.registry }} | ||||||
| image: ${{ vars.CI_CONTAINER_NAME }} | ||||||
| test_data_path: ${{ needs.org-member-pre-flight.outputs.test_data_path }} | ||||||
| script: ${{ matrix.script }} | ||||||
| is_fork_pr: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name != github.event.pull_request.base.repo.full_name }} | ||||||
|
|
||||||
|
|
@@ -333,8 +353,8 @@ jobs: | |||||
| matrix: | ||||||
| include: | ||||||
| - script: L1_Functional_Tests_GPU | ||||||
| runner: ${{ needs.org-member-pre-flight.outputs.runner_prefix }}-gpu-x2 | ||||||
| needs: [pre-flight, build-container, org-member-pre-flight] | ||||||
| if: ${{ needs.pre-flight.outputs.test_level == 'Lfast' }} | ||||||
| runs-on: ${{ matrix.runner }} | ||||||
| name: fast_${{ matrix.script }} | ||||||
|
|
@@ -350,6 +370,9 @@ jobs: | |||||
| runner: ${{ runner.name }} | ||||||
| script: ${{ matrix.script }} | ||||||
| image-tag: ${{ needs.pre-flight.outputs.image_tag }} | ||||||
| registry: ${{ needs.org-member-pre-flight.outputs.registry }} | ||||||
| image: ${{ vars.CI_CONTAINER_NAME }} | ||||||
| test_data_path: ${{ needs.org-member-pre-flight.outputs.test_data_path }} | ||||||
| is_fork_pr: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name != github.event.pull_request.base.repo.full_name }} | ||||||
|
|
||||||
| CI_QA_Gate: | ||||||
|
|
||||||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -52,8 +52,8 @@ run_test uv run --no-sync bash ./tests/functional/grpo_automodel_lora_async | |
| run_test uv run --no-sync bash ./tests/functional/grpo_automodel_lora_non_colocated.sh | ||
| run_test uv run --no-sync bash ./tests/functional/grpo_megatron.sh | ||
| run_test uv run --no-sync bash ./tests/functional/grpo_megatron_generation.sh | ||
| # run_test uv run --no-sync bash ./tests/functional/grpo_megatron_lora.sh | ||
| # run_test uv run --no-sync bash ./tests/functional/grpo_megatron_lora_async.sh | ||
| run_test uv run --no-sync bash ./tests/functional/grpo_multiple_dataloaders.sh | ||
| run_test uv run --no-sync bash ./tests/functional/grpo_multiturn.sh | ||
| run_test uv run --no-sync bash ./tests/functional/grpo_non_colocated.sh | ||
|
|
||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -27,4 +27,5 @@ uv run coverage run -a --data-file=$PROJECT_ROOT/tests/.coverage --source=$PROJE | |
| cat $RUN_LOG | grep "score=" | sed 's/.*score=\([^ ]*\).*/{"score": \1}/' > $JSON_METRICS | ||
|
|
||
| uv run tests/check_metrics.py $JSON_METRICS \ | ||
| 'data["score"] >= 0.1' \ | ||
| 'data["score"] < 0.14' | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -29,4 +29,5 @@ uv run coverage run -a --data-file=$PROJECT_ROOT/tests/.coverage --source=$PROJE | |
| cat $RUN_LOG | grep "score=" | sed 's/.*score=\([^ ]*\).*/{"score": \1}/' > $JSON_METRICS | ||
|
|
||
| uv run tests/check_metrics.py $JSON_METRICS \ | ||
| 'data["score"] >= 0.1' \ | ||
| 'data["score"] < 0.14' | ||
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -916,6 +916,10 @@ async def test_vllm_generation_with_hf_training_colocated( | |
| f"Skipping FP8 test. GPU compute capability {major_capability}.0 is < 9.0 (H100 required)." | ||
| ) | ||
|
|
||
| device_name = torch.cuda.get_device_name(0) | ||
| if "GB200" in device_name: | ||
| pytest.skip("Skipping FP8 test on GB200 until fixed.") | ||
|
|
||
| # Create VllmGeneration Policy | ||
| print("Creating vLLM policy...") | ||
| vllm_config = deepcopy(basic_vllm_test_config) | ||
|
|
@@ -984,6 +988,9 @@ async def test_vllm_generation_with_hf_training_non_colocated( | |
| pytest.skip( | ||
| f"Skipping FP8 test. GPU compute capability {major_capability}.0 is < 9.0 (H100 required)." | ||
| ) | ||
| device_name = torch.cuda.get_device_name(0) | ||
| if "GB200" in device_name: | ||
| pytest.skip("Skipping FP8 test on GB200 until fixed.") | ||
|
|
||
| """This test validates that DTensor policy can work together with non-colocated vLLM policy.""" | ||
| generation_cluster_separate = get_generation_cluster_separate(1) | ||
|
|
@@ -1624,6 +1631,10 @@ def test_vllm_weight_update_and_prefix_cache_reset( | |
| f"Skipping FP8 test. GPU compute capability {major_capability}.0 is < 9.0 (H100 required)." | ||
| ) | ||
|
|
||
| device_name = torch.cuda.get_device_name(0) | ||
| if "GB200" in device_name: | ||
| pytest.skip("Skipping FP8 test on GB200 until fixed.") | ||
|
|
||
| from nemo_rl.models.policy.lm_policy import Policy | ||
|
|
||
| # Create configs | ||
|
|
@@ -2038,6 +2049,10 @@ def test_vllm_generation_with_megatron_training( | |
| f"Skipping FP8 test. GPU compute capability {major_capability}.0 is < 9.0 (H100 required)." | ||
| ) | ||
|
|
||
| device_name = torch.cuda.get_device_name(0) | ||
| if "GB200" in device_name: | ||
| pytest.skip("Skipping FP8 test on GB200 until fixed.") | ||
|
|
||
| if cluster.num_gpus_per_node < tensor_parallel_size: | ||
| pytest.skip(f"Need at least {tensor_parallel_size} GPUs for this test") | ||
|
|
||
|
|
@@ -2208,6 +2223,10 @@ def test_vllm_generation_with_megatron_training_moe_model( | |
| f"Skipping FP8 test. GPU compute capability {major_capability}.0 is < 9.0 (H100 required)." | ||
| ) | ||
|
|
||
| device_name = torch.cuda.get_device_name(0) | ||
| if "GB200" in device_name: | ||
| pytest.skip("Skipping FP8 test on GB200 until fixed.") | ||
|
|
||
| model_name = "moonshotai/Moonlight-16B-A3B-Instruct" | ||
| expert_parallel_size = 8 | ||
|
|
||
|
|
||
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
registryis effectively required but declared optional.Line 97 and Line 136 always build image refs from
inputs.registry, so an empty value produces invalid image names. Make the input required (or provide a safe default).🔧 Suggested change
registry: description: "Registry to use for test" - required: false + required: trueAlso applies to: 97-98, 136-136
🤖 Prompt for AI Agents