[https://nvbugs/6427411][chore] Waive remaining PP sampler regressions

[chienchunhung]

Summary by CodeRabbit

• Tests
  • Updated test waiver coverage for several multi-GPU and parametrized cases.
  • Added skips for a few specific model and inference scenarios, including DeepSeek, Llama, Nemotron, TinyLlama, and LLM stats checks.
  • This helps keep the test suite stable by excluding known flaky or unsupported combinations.

Description

PR #15920 moved greedy stop-criteria handling to the host, but did not propagate use_host_stop_criteria to non-last pipeline-parallel ranks. Those ranks can attempt to read an intentionally absent finish_reasons array and fail with IndexError: list index out of range.

This PR temporarily waives the six remaining observed failures so the known regression does not block unrelated CI runs.

Scope

• Add six exact test waivers under https://nvbugs/6427411.
• Keep the broad unittest/llmapi/test_llm_multi_gpu_pytorch.py -m "gpu4" wrapper enabled so unrelated failures remain visible.
• Change CI coverage only; runtime behavior is unchanged.

Waived tests

accuracy/test_llm_api_pytorch.py::TestNemotronV3Super::test_nvfp4_parallelism[TP4_PP2]
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_fp8_block_scales_4gpus[pp4-mtp_nextn=0-fp8kv=True-attention_dp=False-cuda_graph=True-overlap_scheduler=True-torch_compile=True-sampler_async_worker=False]
accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_guided_decoding_4gpus[xgrammar]
disaggregated/test_disaggregated.py::test_disaggregated_ctxpp2_gentp2[TinyLlama-1.1B-Chat-v1.0]
disaggregated/test_disaggregated.py::test_disaggregated_ctxtp2_genpp2[TinyLlama-1.1B-Chat-v1.0]
unittest/llmapi/test_llm_multi_gpu_pytorch.py::test_llm_get_stats_pp4[False-False-True]

Follow-up

Remove these waivers after #16096 lands and the affected PP configurations pass.

Related PRs

• Regression: [None][perf] Move greedy stop checks to host #15920
• Durable fix: [None][fix] Propagate host stop criteria across PP ranks #16096
• Existing waivers for the same regression: [https://nvbugs/6427411][chore] Waive failed tests in Pre-merge #16103 and [https://nvbugs/6427411][chore] Waive a failed test in Pre-merge #16105

Test Coverage

• python3 scripts/check_test_list.py --validate --check-duplicate-waives
• Repository pre-commit and commit-message hooks

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.

[chienchunhung]

/bot skip --comment "Temporarily waiving tests to unblock CI and PR merging."

[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: 60a433c5-1355-459a-ab64-2e380d22329d

📥 Commits

Reviewing files that changed from the base of the PR and between 5042300 and f51c02a.

📒 Files selected for processing (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

This PR adds four new SKIP/waive entries to tests/integration/test_lists/waives.txt, covering DeepSeekV3Lite fp8 block scales, Llama3_1_8BInstruct guided decoding, NemotronV3Super nvfp4 parallelism, disaggregated TinyLlama context/generation configurations, and a multi-GPU stats test.

Changes

Test waive list updates

Layer / File(s)	Summary
Add new test waivers tests/integration/test_lists/waives.txt	Adds waive entries for TestDeepSeekV3Lite fp8 block scales (pp4), TestLlama3_1_8BInstruct guided decoding (xgrammar), TestNemotronV3Super nvfp4 parallelism (TP4_PP2), disaggregated TinyLlama ctxpp2/gentp2 and ctxtp2/genpp2 cases, and test_llm_get_stats_pp4.

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

Possibly related PRs

• NVIDIA/TensorRT-LLM#16093: Both PRs add new waive entries to the same tests/integration/test_lists/waives.txt file for newly failing test parametrizations.
• NVIDIA/TensorRT-LLM#16103: Both PRs modify waives.txt with entries for the same categories of disaggregated multi-GPU test cases.
• NVIDIA/TensorRT-LLM#16105: Both PRs add new waiver entries to waives.txt for specific parametrized LLM/PT tests.

Suggested reviewers: mzweilz, yuxianq

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	The PR only adds waivers and does not implement the host-stop-criteria propagation or regression coverage required by #16096.	Add the PP host-stop-criteria code fix and regression tests from #16096, or relink this PR to the waiver-only tracking issue.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: waiving the remaining PP sampler regressions.
Description check	✅ Passed	The description follows the template and includes the issue, scope, test coverage, follow-up, and checklist status.
Out of Scope Changes check	✅ Passed	The change set is limited to waiving the listed failing tests and does not introduce unrelated code or behavior changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[brb-nv]

LGTM.

[tensorrt-cicd]

PR_Github #58281 [ skip ] triggered by Bot. Commit: f51c02a Link to invocation

[tensorrt-cicd]

PR_Github #58281 [ skip ] completed with state SUCCESS. Commit: f51c02a
Skipping testing for commit f51c02a

Link to invocation