[https://nvbugs/6619882][test] Unwaive unittest/_torch/sampler on DGX_B200

[zhaoyangwang-nvidia]

Dev Engineer Review

• Removed the _torch/sampler -k "not test_speculative_d2h_parity_real_predictor" waiver from tests/integration/test_lists/waives.txt.
• The change restores DGX_B200 sampler coverage without changing test code or public entities.
• The waiver format and scope are consistent with the stated objective.
• A local 8xB200 run passed 1,147 tests, skipped 699, and deselected 2.

QA Engineer Review

• Modified tests/integration/test_lists/waives.txt.
• Removed the _torch/sampler waiver entry.
• CBTS coverage data is unavailable.

Verdict: needs follow-up

Description

unittest/_torch/sampler -k "not test_speculative_d2h_parity_real_predictor" was waived on DGX_B200 by #17782 (a098858cef) because of an intermittent failure that shows up roughly 4 times per month.

The reported failure is not an assertion failure. In the CI log the inner pytest process dies without printing a summary or a traceback, right after the KV cache allocation of a test_beam_search_e2e[multi_process-...] case; the outer wrapper then only surfaces the generic AssertionError: failure reported in unittests.

Waiving the whole entry drops coverage for the entire sampler directory on Blackwell, which is far broader than the intermittent failure itself. This PR removes the waive so the case runs in CI again and the intermittent failure can be characterized on current main.

Test Coverage

unittest/_torch/sampler -k "not test_speculative_d2h_parity_real_predictor" (DGX_B200-PyTorch-1) is re-enabled; no other test is affected.

Local run on an 8xB200 node at bd90276f6b, bypassing the waive list:

1147 passed, 699 skipped, 2 deselected in 995.93s (0:16:35)

PR Checklist

Please review the following before submitting your PR:

• PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

• PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

• Test cases are provided for new code paths (see test instructions)

• If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• Update tava architecture diagram if there is a significant design change in PR.

• The reviewers assigned automatically/manually are appropriate for the PR.

• Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

[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: 8d13e97e-977c-4a48-8693-1244f6945408

📥 Commits

Reviewing files that changed from the base of the PR and between 18b0eaf and 33a6a44.

📒 Files selected for processing (1)

• tests/integration/test_lists/waives.txt

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/waives.txt

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

Walkthrough

The integration waiver list removes the skip entry for test_speculative_d2h_parity_real_predictor in the _torch/sampler test group.

Changes

Sampler waiver update

Layer / File(s)	Summary
Remove sampler test skip tests/integration/test_lists/waives.txt	Removes the waiver that excluded test_speculative_d2h_parity_real_predictor from the _torch/sampler test group.

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

Merge Risk: ⚪ Minimal · up to 33a6a

This change removes a test waiver and restores sampler test coverage on DGX_B200 without changing product behavior; no actionable merge-blocking risk remains beyond normal checks and review.

Possibly related PRs

• NVIDIA/TensorRT-LLM#17600: Updates the same waiver list by removing an obsolete test skip entry.
• NVIDIA/TensorRT-LLM#17943: Updates skip entries in the same waiver list for different test cases.
• NVIDIA/TensorRT-LLM#17828: Changes sampler behavior and removes related sampler waivers.

Suggested labels: ci: full pre-merge approved

Suggested reviewers: bowenfu

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the NVBugs issue, test change, and affected sampler test group.
Description check	✅ Passed	The description explains the intermittent failure, the waiver removal, the restored coverage, and the relevant test results.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.}

[zhaoyangwang-nvidia]

/bot run

[zhaoyangwang-nvidia]

/bot run --stage-list "DGX_B200-PyTorch-1"

[tensorrt-cicd]

PR_Github #67299 [ run ] triggered by Bot. Commit: 33a6a44 Link to invocation

[tensorrt-cicd]

PR_Github #67300 [ run ] triggered by Bot. Commit: 33a6a44 Link to invocation

[tensorrt-cicd]

PR_Github #67299 [ run ] completed with state ABORTED. Commit: 33a6a44

Link to invocation

[tensorrt-cicd]

PR_Github #67300 [ run ] completed with state SUCCESS. Commit: 33a6a44
/LLM/main/L0_MergeRequest_PR pipeline #54821 (Partly Tested) completed with status: 'SUCCESS'

CI Report

Link to invocation

[zhaoyangwang-nvidia]

/bot run --stage-list "DGX_B200-PyTorch-1"

[tensorrt-cicd]

PR_Github #67374 [ run ] triggered by Bot. Commit: 33a6a44 Link to invocation