[None][test] Un-waive K2.5 Thinking FP4 disagg-NIXL e2e/gen_only tests

[chenfeiz0326]

Summary

Removes 3 SKIP entries from tests/integration/test_lists/waives.txt for the K2.5 Thinking FP4 disagg-NIXL family — the underlying issues are resolved on current ToT main.

Tests removed from waives.txt

Test	Bug	Validation
disagg_upload-e2e-gb300_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL	https://nvbugs/6280649	5/5 PASS, 119242–120884 tok/s, 1.38% spread
disagg_upload-gen_only-gb200_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL	https://nvbugs/6221022	5/5 PASS, 92186–94509 tok/s, 2.50% spread
disagg_upload-e2e-gb200_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL	https://nvbugs/6179661	Same model+config family as above; unblocked by the same ToT fixes

Both validated cases were run as 5 separate Slurm jobs (per-run output dir) on Lyris against a freshly built ToT main wheel:

• GB300 (sm_103): partition=gb300, dep4 ctx + dep16 gen
• GB200 (sm_100): partition=gb200, dep4 ctx + dep16 gen, gen_only mode

Spread is well under the 5% CV threshold used by perf-sanity.

Test plan

• CI perf-sanity job picks up the un-waived tests
• All 3 cases PASS in the next post-merge perf-sanity run

🤖 Generated with Claude Code

Summary by CodeRabbit

• Tests
  • Removed waive entries for performance sanity tests, enabling previously skipped test configurations to run.

[coderabbitai]

📝 Walkthrough

Walkthrough

Removes three SKIP waive entries from tests/integration/test_lists/waives.txt: two disagg_upload-e2e entries for gb200_kimi-k25-thinking and gb300_kimi-k25-thinking, and one disagg_upload-gen_only entry for gb200_kimi-k25-thinking, all under perf/test_perf_sanity.py::test_e2e.

Changes

kimi-k25-thinking waive removal

Layer / File(s)	Summary
Remove kimi-k25-thinking disagg_upload SKIP entries tests/integration/test_lists/waives.txt	Deletes two disagg_upload-e2e SKIP lines for gb200/gb300_kimi-k25-thinking (lines 301–304 region) and one disagg_upload-gen_only SKIP line for gb200_kimi-k25-thinking (line 325 region), leaving the qwen3-235b, deepseek-r1, and glm-5 entries untouched.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#15373: Modifies the same perf/test_perf_sanity.py::test_e2e disagg/gen-only regions in waives.txt, including gb200/gb300_kimi-k25-thinking and nearby model/mode waives.
• NVIDIA/TensorRT-LLM#15395: Changes disagg_upload-* waives for GB200/GB300 and gen-only/e2e variants in the same perf/test_perf_sanity.py section of waives.txt.
• NVIDIA/TensorRT-LLM#15269: Also removes/updates perf/test_perf_sanity.py::test_e2e disagg upload waive entries in the same disagg_upload-e2e and disagg_upload-gen_only sections.

Suggested reviewers

• mzweilz
• 2ez4bz

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically summarizes the main change: removing waive entries for K2.5 Thinking FP4 disagg-NIXL tests, with relevant context about test types (e2e/gen_only). It is concise and directly related to the changeset.
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.
Description check	✅ Passed	The PR description clearly explains the purpose, provides detailed validation results, and documents the testing methodology comprehensively.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[chenfeiz0326]

/bot run --disable-fail-fast --stage-list "GB300-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-4,GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-2,GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-4"

[tensorrt-cicd]

PR_Github #54756 [ run ] triggered by Bot. Commit: ce1e98c Link to invocation

[tensorrt-cicd]

PR_Github #54756 [ run ] completed with state FAILURE. Commit: ce1e98c
/LLM/main/L0_MergeRequest_PR pipeline #43774 (Partly Tested) completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[chenfeiz0326]

/bot run --disable-fail-fast --stage-list "GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-2"

[tensorrt-cicd]

PR_Github #54859 [ run ] triggered by Bot. Commit: 12dd871 Link to invocation

[tensorrt-cicd]

PR_Github #54859 [ run ] completed with state SUCCESS. Commit: 12dd871
/LLM/main/L0_MergeRequest_PR pipeline #43863 (Partly Tested) completed with status: 'SUCCESS'

CI Report

Link to invocation

[chenfeiz0326]

/bot skip --comment "Only unwaive perf tests, no need to run the whole CI pipeline"

[tensorrt-cicd]

PR_Github #55127 [ skip ] triggered by Bot. Commit: 12dd871 Link to invocation

[tensorrt-cicd]

PR_Github #55127 [ skip ] completed with state SUCCESS. Commit: 12dd871
Skipping testing for commit 12dd871

Link to invocation