[None][fix] fix oom

[xinhe-nv]

Dev Engineer Review

• Updated free_gpu_memory_fraction in two integration tests.
• Set the DeepSeek-V3-Lite FP8 block-scale multi-GPU test to 0.6.
• Set the Nemotron-Ultra V3 MARLIN test to 0.5.
• The values are valid fractional settings.
• The change is limited to test configuration.
• No public API, error handling, or unrelated configuration changes were identified.

QA Engineer Review

• Modified two test configurations:
  • DeepSeek-V3-Lite FP8 block-scale multi-GPU integration test.
  • Nemotron-Ultra V3 MARLIN integration test.
• No test-list changes were reported.
• Test-list coverage could not be verified from the available change summary.
• Verdict: needs follow-up.

Description

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.

[coderabbitai]

Walkthrough

Two PyTorch accuracy tests now use lower KV-cache memory fractions. One test changes an existing value, and the other adds an explicit value.

Changes

KV-cache test configuration

Layer / File(s)	Summary
Adjust test KV-cache fractions tests/integration/defs/accuracy/test_llm_api_pytorch.py	The DeepSeek-V3-Lite test changes free_gpu_memory_fraction from 0.75 to 0.6. The Nemotron-Ultra V3 test sets it to 0.5.

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

Possibly related PRs

• NVIDIA/TensorRT-LLM#17470: Adjusts KV-cache configuration in the same test file for GPU memory use.

Suggested reviewers: larryxfly, qijune, tburt-nv

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	⚠️ Warning	The description omits the issue, solution, and test coverage details required by the template.	Add a brief issue and solution summary, list the relevant tests, and complete the applicable checklist items.

✅ Passed checks (3 passed)

Check name	Status	Explanation
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 identifies a fix for out-of-memory errors, which matches the primary change to reduce GPU memory usage.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[xinhe-nv]

/bot skip --comment "only update free_gpu_memory_fraction value, test is pass on QA CI"

[tensorrt-cicd]

PR_Github #65239 [ skip ] triggered by Bot. Commit: 0bb8384 Link to invocation

[tensorrt-cicd]

PR_Github #65239 [ skip ] completed with state SUCCESS. Commit: 0bb8384
Skipping testing for commit 0bb8384

Link to invocation