[None][test] Run thop custom-op schema checks in the CPU pre-merge stage

[brnguyen2]

Description

tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py (custom-op schema checks, including test_register_fake) is fully GPU-free: it only imports tensorrt_llm and walks the torch schema registry. Today it runs only through the H100 test list, so a missing fake/meta implementation for a new C++ custom op is caught late and only on Hopper stages (this happened with a recently added op, where every non-H100 stage stayed green).

This PR marks the module cpu_only and adds it to l0_cpu.yml, so the check runs in the CPU-Generic pre-merge stages on every pipeline, independent of GPU pool availability. The l0_h100.yml directory entry is unchanged: GPU stages deselect the file via -m "not cpu_only", and the CPU-stage collection hook picks it up via the module marker, following the existing CPU-only test migration pattern (#17290).

Test Coverage

• unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py now collected by the CPU-Generic stages (-m cpu_only); sibling GPU-dependent files in the directory remain H100-only.

PR Checklist

• PR title and description conventions followed
• Test coverage: this PR is itself a test-coverage change

Dev Engineer Review

• Added tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py to tests/integration/test_lists/test-db/l0_cpu.yml.
• Marked the test module with pytest.mark.cpu_only.
• The test-list entry uses the expected format and limits execution to CPU stages.
• No code regressions or API changes identified.

Verdict: sufficient.

QA Engineer Review

• Modified test module: tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py.
• No test functions were added, modified, or removed.
• The existing custom-op validation tests are covered by tests/integration/test_lists/test-db/l0_cpu.yml.
• CPU-Generic x86 and ARM CI completed successfully.

Verdict: sufficient.

[brnguyen2]

/bot run --stage-list "CPU-Generic-x86-1, CPU-Generic-arm-1"

[tensorrt-cicd]

PR_Github #64128 [ run ] triggered by Bot. Commit: 620a967 Link to invocation

[tensorrt-cicd]

PR_Github #64128 [ run ] completed with state SUCCESS. Commit: 620a967
/LLM/main/L0_MergeRequest_PR pipeline #52050 (Partly Tested) completed with status: 'SUCCESS'

CI Report

Link to invocation

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 26f8b8a5-b620-4e2b-89e3-9e0474461285

📥 Commits

Reviewing files that changed from the base of the PR and between 620a967 and 0a78721.

📒 Files selected for processing (1)

• tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py

🚧 Files skipped from review as they are similar to previous changes (1)

• tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py

---

Walkthrough

The custom operations test is marked for CPU-only execution and added to the l0_cpu pre-merge test suite.

Changes

CPU test integration

Layer / File(s)	Summary
Mark and register the CPU test tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py, tests/integration/test_lists/test-db/l0_cpu.yml	The test module uses the cpu_only pytest marker. The test path is included in the l0_cpu test list.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#16498: Both changes mark CPU-only tests and add them to the l0_cpu suite.
• NVIDIA/TensorRT-LLM#17290: Both changes mark CPU-compatible tests for CPU-only execution and add them to the l0_cpu suite.

Suggested reviewers: qijune, stanleysun639

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the test change and its move to CPU pre-merge coverage.
Description check	✅ Passed	The description explains the problem, solution, test coverage, and relevant checklist items with sufficient detail.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py`:
- Around line 7-8: Add the standard NVIDIA copyright header at the top of
test_custom_ops.py, using 2026 as the latest meaningful modification year and
preserving the existing pytestmark declaration.

🪄 Autofix

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 0b03b2ff-159e-4ab1-a044-2602088d3866

📥 Commits

Reviewing files that changed from the base of the PR and between a23e8de and 620a967.

📒 Files selected for processing (2)

• tests/integration/test_lists/test-db/l0_cpu.yml
• tests/unittest/_torch/thop/parallel_hw_agnostic/test_custom_ops.py

[brnguyen2]

/bot run --stage-list "CPU-Generic-x86-1, CPU-Generic-arm-1"

[brnguyen2]

/bot run --stage-list "CPU-Generic-x86-1,CPU-Generic-arm-1"

[tensorrt-cicd]

PR_Github #64220 [ run ] triggered by Bot. Commit: 0a78721 Link to invocation

[tensorrt-cicd]

PR_Github #64220 [ run ] completed with state SUCCESS. Commit: 0a78721
/LLM/main/L0_MergeRequest_PR pipeline #52127 (Partly Tested) completed with status: 'SUCCESS'

CI Report

Link to invocation

[brnguyen2]

/bot skip --comment "Two-file test-infra change: marks the hw-agnostic custom-op schema test module cpu_only and lists it in l0_cpu.yml. The stages that newly run it (CPU-Generic-x86-1, CPU-Generic-arm-1) passed on this exact head in pipeline 52127; the only effect on GPU stages is standard -m 'not cpu_only' deselection of one module, the same mechanism used by all existing cpu_only tests (pattern of #17290). No production code changed."

[tensorrt-cicd]

PR_Github #64335 [ skip ] triggered by Bot. Commit: 0a78721 Link to invocation

[tensorrt-cicd]

PR_Github #64335 [ skip ] completed with state SUCCESS. Commit: 0a78721
Skipping testing for commit 0a78721

Link to invocation