[https://nvbugs/6478692][fix] Pass max_workers=16 to _build in EPDVariant.nano_omni_fp8 (mirroring…

[trtllm-agent]

Summary

• Root cause: nano_omni_fp8 did not override max_workers, defaulting to VideoMME.MAX_BATCH_SIZE=128, which oversubscribed the CPU-side multimodal preprocessing/hashing path on H100 and tripped a native-library SIGABRT.
• Fix: Pass max_workers=16 to _build in EPDVariant.nano_omni_fp8 (mirroring qwen3vl_2b), and remove the corresponding entry from waives.txt.
• Automated fix generated by repair-bot

Test plan

• Verify fix on the same GPU type as the original failure
• Check for regressions in related tests

Links

• Bug: https://nvbugs/6478692

Summary by CodeRabbit

• Bug Fixes
  • Improved stability for multimodal processing on H100 hardware by limiting parallel workers.
  • Re-enabled the disaggregated video multimodal accuracy test so it runs as part of validation.

[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: 043d0d47-ae7f-426b-9db5-15ebe72f8662

📥 Commits

Reviewing files that changed from the base of the PR and between d75cb89 and 537c3eb.

📒 Files selected for processing (2)

• tests/integration/defs/accuracy/test_epd_disagg_multimodal.py
• tests/integration/test_lists/waives.txt

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

The nano omni FP8 multimodal EPD test now explicitly limits its worker pool to 16, and the corresponding VideoMME test waiver is removed.

Changes

Multimodal EPD test enablement

Layer / File(s)	Summary
Worker limit and waiver removal tests/integration/defs/accuracy/test_epd_disagg_multimodal.py, tests/integration/test_lists/waives.txt	The nano_omni_fp8 variant sets max_workers=16, and its disaggregated VideoMME test is removed from the waiver list.

Estimated code review effort: 2 (Simple) | ~5 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#16638: Removes waived or unwaived test entries in the same waiver list.

Suggested reviewers: 2ez4bz

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title matches the main fix and follows the ticket/type pattern, with specific context about max_workers in nano_omni_fp8.
Description check	✅ Passed	The description covers the issue, fix, test plan, and bug link, though it doesn't use the template's exact section headings.
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.}

[xinhe-nv]

/bot run --only-qa-verify test accuracy/test_epd_disagg_multimodal.py::TestVideoMMEEPD::test_disaggregated_videomme[nemotron_nano_v3_omni_fp8]

[tensorrt-cicd]

PR_Github #60696 [ run ] triggered by Bot. Commit: 537c3eb Link to invocation

[tensorrt-cicd]

PR_Github #60696 [ run ] completed with state FAILURE. Commit: 537c3eb
LLM_FUNCTION_AUTO_V2C #299 completed with status: 'UNSTABLE'
QA verify test: accuracy/test_epd_disagg_multimodal.py::TestVideoMMEEPD::test_disaggregated_videomme[nemotron_nano_v3_omni_fp8] (NVBug 6478692, branch repair-bot-bug6478692, fork tensorrt-cicd, dry_run_close=true)

Link to invocation

[mzweilz]

LGTM if the tests pass.

[StanleySun639]

LGTM if the test can pass.

[2ez4bz]

/bot run

[tensorrt-cicd]

PR_Github #63095 [ run ] triggered by Bot. Commit: 57ade4d Link to invocation

[tensorrt-cicd]

PR_Github #63095 [ run ] completed with state SUCCESS. Commit: 57ade4d
/LLM/main/L0_MergeRequest_PR pipeline #51187 completed with status: 'SUCCESS'

CI Report

Link to invocation