[None][infra] Waive 2 failed cases for main in pre-merge 52091

[trtllm-agent]

Auto-generated Waive PR

Created by: TensorRT LLM CI Report (requested by @shuyixiong)
Target branch: main
Bug(s): 6567731

Waive entries added

accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[v1_kv_cache-True-True-cutlass-auto] SKIP (https://nvbugs/6567731)
accuracy/test_llm_api_pytorch.py::TestQwen3_8B::test_fp8_block_scales[latency] SKIP (https://nvbugs/6567731)

Already waived (skipped)

• accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[v2_kv_cache-True-True-cutlass-auto]

---

This PR was auto-generated by TensorRT LLM CI Report. Please review the waive entries before merging.

Dev Engineer Review

• Added two correctly formatted waiver entries in tests/integration/test_lists/waives.txt.
• The entries match the reported GPTOSS and Qwen3 8B test cases.
• Both entries reference NVBUG 6567731.
• The change has no API, performance, or error-handling impact.
• No duplicate or unintended scope changes were identified.

QA Engineer Review

• No test-db/ or qa/ files were modified.
• Added waivers for:
  • TestGPTOSS::test_w4_1gpu[v1_kv_cache-True-True-cutlass-auto]
  • TestQwen3_8B::test_fp8_block_scales[latency]
• CBTS coverage data is unavailable for this test-list-only change.
• Verdict: needs follow-up.

[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: 4c0dd4c4-1b53-4606-8cb5-b96c532f765e

📥 Commits

Reviewing files that changed from the base of the PR and between 77bc7f0 and fcdd282.

📒 Files selected for processing (1)

• tests/integration/test_lists/waives.txt

---

Walkthrough

The integration waiver list adds skip entries for two PyTorch accuracy tests. Both entries reference NVBUG 6567731.

Changes

PyTorch accuracy test waivers

Layer / File(s)	Summary
Add PyTorch accuracy test waivers tests/integration/test_lists/waives.txt	Adds skip records for the GPTOSS W4 single-GPU CUTLASS auto test and the Qwen3 8B FP8 block-scales latency test. Both records reference NVBUG 6567731.

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

Possibly related PRs

• NVIDIA/TensorRT-LLM#17348: Adds skip entries for other failing PyTorch integration tests.
• NVIDIA/TensorRT-LLM#17340: Adds waiver entries for failing PyTorch accuracy tests.
• NVIDIA/TensorRT-LLM#17335: Adds SKIP entries for other failing tests.

Suggested reviewers: bowenfu, brnguyen2, xinhe-nv

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the infrastructure change: waiving two failed pre-merge cases on main.
Description check	✅ Passed	The description explains the waiver purpose, identifies both test cases, links the bug, and notes the existing waiver.
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.}

[shuyixiong]

/bot skip --comment "Waive tests"

[tensorrt-cicd]

PR_Github #64265 [ skip ] triggered by Bot. Commit: fcdd282 Link to invocation

[tensorrt-cicd]

PR_Github #64265 [ skip ] completed with state SUCCESS. Commit: fcdd282
Skipping testing for commit fcdd282

Link to invocation

[brnguyen2]

Approving — the comments below are optional touch-ups, not blockers.

Entries are well-formed, correctly sorted, and not duplicates of existing waives for these tests. One scoping question inline. Also worth noting the failure signature here isn't specific to these two tests — if it recurs, chasing the underlying infra issue will be cheaper than accumulating per-test waives.

[brnguyen2]

Both entries are unqualified, so they skip on every platform, but the reported failures came from a single stage (GB300-PyTorch-1) and the failure mode is a session/worker bootstrap timeout rather than anything specific to GPT-OSS w4 or Qwen3 fp8 block scales. Consider a full:GB300/-prefixed waive (as done elsewhere in this file, e.g. lines 259, 290) so these keep running on other GPUs where they pass.