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

[chienchunhung]

Summary

Temporarily waive five exact pipeline-parallel test variants affected by NVBug 6427411:

• accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_fp8_block_scales_4gpus[tp2pp2-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::TestDeepSeekV3Lite::test_fp8_block_scales_4gpus[tp2pp2-mtp_nextn=0-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False-sampler_async_worker=False]
• accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_fp8_block_scales_4gpus[tp2pp2-mtp_nextn=0-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True-sampler_async_worker=False]
• accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_nvfp4_4gpus[moe_backend=CUTLASS-mtp_nextn=0-pp4-fp8kv=True-attention_dp=True-cuda_graph=True-overlap_scheduler=True-low_precision_combine=False-torch_compile=False]
• unittest/llmapi/test_llm_multi_gpu_pytorch.py::test_llm_get_stats_pp4[False-True-True]

Only these test IDs are waived; their broader GPU4 suites remain enabled.

Background

These cases surfaced or were identified after the earlier waivers in #16127 landed. A failing PP test can terminate its pytest/MPI process before later parameterizations run or write JUnit results, masking additional affected cases.

The observed failures share the pipeline-parallel sampler regression introduced by #15920. Confirmed cases fail in finish_if_reason with IndexError: list index out of range, and the remaining queued variant exercises the same PP sampler path. The corrective change is tracked by #16096.

This PR contains no runtime changes. These waivers can be removed after #16096 or an equivalent fix lands.

Validation

• python3 scripts/check_test_list.py --check-duplicate-waives
• python3 scripts/check_test_list.py --validate
• pre-commit run --files tests/integration/test_lists/waives.txt

Summary by CodeRabbit

• Bug Fixes
  • Updated automated test exclusions to better account for known multi-GPU coverage gaps.
  • Added additional safeguards for specific distributed inference and statistics scenarios, helping keep validation runs stable.

[chienchunhung]

/bot skip --comment "test waiving to unblock CI"

[tensorrt-cicd]

PR_Github #58304 [ skip ] triggered by Bot. Commit: a5ef292 Link to invocation

[chienchunhung]

/bot skip --comment "test waiving to unblock CI"

[tensorrt-cicd]

PR_Github #58304 [ skip ] completed with state SUCCESS. Commit: a5ef292
Skipping testing for commit a5ef292

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: 9d828680-ce47-4ebb-8cf2-09befadd73d0

📥 Commits

Reviewing files that changed from the base of the PR and between 5ce293f and b2c2831.

📒 Files selected for processing (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

This PR adds new SKIP waiver entries to the integration test waives list, covering additional TestDeepSeekV3Lite accuracy test configurations for test_fp8_block_scales_4gpus and test_nvfp4_4gpus, plus a new waiver for test_llm_get_stats_pp4 in the unittest multi-GPU pytorch tests.

Changes

Test waivers update

Layer / File(s)	Summary
Add new SKIP waiver entries tests/integration/test_lists/waives.txt	Adds three new test_fp8_block_scales_4gpus SKIP entries and one test_nvfp4_4gpus SKIP entry for TestDeepSeekV3Lite, plus a test_llm_get_stats_pp4[False-True-True] SKIP entry.

Estimated code review effort: 1 (Trivial) | ~3 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#15774: Both PRs extend SKIP waivers in waives.txt for TestDeepSeekV3Lite fp8kv=True cases including test_nvfp4.
• NVIDIA/TensorRT-LLM#16127: Both PRs add related SKIP waivers for TestDeepSeekV3Lite::test_fp8_block_scales_4gpus and test_llm_get_stats_pp4.
• NVIDIA/TensorRT-LLM#16103: Both PRs adjust SKIP waivers for the same class of LLM API multi-GPU test cases including DeepSeekV3Lite parameters.

Suggested reviewers: shuyixiong, mingyangHao, brb-nv

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is concise and accurately reflects the PR's temporary test-waiver change.
Description check	✅ Passed	The description covers summary, background, and validation, and is sufficiently complete despite missing explicit Test Coverage and PR Checklist sections.
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.}

[tensorrt-cicd]

PR_Github #58306 [ skip ] triggered by Bot. Commit: b2c2831 Link to invocation

[tensorrt-cicd]

PR_Github #58306 [ skip ] completed with state SUCCESS. Commit: b2c2831
Skipping testing for commit b2c2831

Link to invocation