[https://nvbugs/6305365][chore] Unwaive piecewise cudagraph related tests

[pengbowang-nv]

Summary by CodeRabbit

• Tests
  • Re-enabled four previously skipped accuracy test cases covering bfloat16 configurations for DeepSeek V3 Lite and Llama 3.1 8B Instruct.

Description

Root cause was stale torch-compile piecewise CUDA Graph state surviving ModelEngine/Python executor teardown. The previous cleanup released the regular CUDA Graph runners but did not clear PiecewiseRunner captures, cached input/output addresses, static outputs, or their private graph-pool state. A later executor could therefore replay a graph captured against the previous executor lifetime, resulting in cudaErrorLaunchFailure during attention warmup.

This was fixed by commit 211482b, see #15923. The fix invokes Backend.clear_piecewise_cuda_graphs() from ModelEngine._release_cuda_graphs(), resets all tracked PiecewiseRunner capture state, and rotates dead private graph-pool handles.

This PR removes related tests from waives.txt.

Test Coverage

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.

[pengbowang-nv]

/bot run --stage-list "DGX_H100-PyTorch-Post-Merge-1, DGX_H100-PyTorch-Post-Merge-2"

[coderabbitai]

📝 Walkthrough

Walkthrough

Removed four SKIP waivers from the integration test list, re-enabling selected DeepSeekV3Lite and Llama3.1 8B Instruct bfloat16 test configurations.

Changes

Accuracy test waiver updates

Layer / File(s)	Summary
Remove selected bfloat16 waivers tests/integration/test_lists/waives.txt	Removes two DeepSeekV3Lite and two Llama3.1 8B Instruct bfloat16 SKIP entries.

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

Possibly related PRs

• NVIDIA/TensorRT-LLM#16576: Updates related bfloat16 waivers for the same model tests.
• NVIDIA/TensorRT-LLM#16585: Adjusts waivers for related accuracy test parameterizations.
• NVIDIA/TensorRT-LLM#16587: Changes SKIP entries for related DeepSeekV3Lite and Llama3 test cases.

Suggested reviewers: bowenfu

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
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.
Title check	✅ Passed	The title is concise and accurately reflects removing piecewise cudagraph test waivers.
Description check	✅ Passed	The description covers the issue, fix, and PR checklist, though the Test Coverage section is left blank.

✨ 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 #60269 [ run ] triggered by Bot. Commit: e0f9b21 Link to invocation

[tensorrt-cicd]

PR_Github #60269 [ run ] completed with state FAILURE. Commit: e0f9b21
/LLM/main/L0_MergeRequest_PR pipeline #48628 (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

[pengbowang-nv]

/bot run --stage-list "DGX_H100-PyTorch-Post-Merge-1, DGX_H100-PyTorch-Post-Merge-2"

[tensorrt-cicd]

PR_Github #60279 [ run ] triggered by Bot. Commit: e0f9b21 Link to invocation

[tensorrt-cicd]

PR_Github #60279 [ run ] completed with state FAILURE. Commit: e0f9b21
/LLM/main/L0_MergeRequest_PR pipeline #48636 (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

[pengbowang-nv]

/bot run --stage-list "DGX_H100-PyTorch-Post-Merge-1, DGX_H100-PyTorch-Post-Merge-2"

[tensorrt-cicd]

PR_Github #60292 [ run ] triggered by Bot. Commit: e0f9b21 Link to invocation

[tensorrt-cicd]

PR_Github #60292 [ run ] completed with state SUCCESS. Commit: e0f9b21
/LLM/main/L0_MergeRequest_PR pipeline #48645 (Partly Tested) completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

CI Report

Link to invocation

[pengbowang-nv]

/bot run --disable-fail-fast

[BowenFu]

LGTM — waiver removal; the piecewise-cudagraph fix already landed (#15923), re-enabling the DSV3Lite/Llama3.1 tests. No product code.

[tensorrt-cicd]

PR_Github #60337 [ run ] triggered by Bot. Commit: e0f9b21 Link to invocation

[tensorrt-cicd]

PR_Github #60337 [ run ] completed with state SUCCESS. Commit: e0f9b21
/LLM/main/L0_MergeRequest_PR pipeline #48682 completed with status: 'SUCCESS'

CI Report

Link to invocation