[https://nvbugs/6445322][fix] Restore the pre-#16108 baseline - accuracy: 56.667 alongside the existing…

[trtllm-agent]

Summary

• Root cause: PR [https://nvbugs/6379636][fix] Fix Gemma4 MoE weight loading #16108 replaced the plain bf16 entry for google/gemma-4-26B-A4B-it in tests/integration/defs/accuracy/references/mmmu.yaml with only an NVFP4 entry, so the autodeploy bf16 test's default acc_specs matched no row and raised ValueError: Not registered specs.
• Fix: Restore the pre-[https://nvbugs/6379636][fix] Fix Gemma4 MoE weight loading #16108 baseline - accuracy: 56.667 alongside the existing NVFP4 entry (same coexistence pattern used for gemma-3-27b-it / gemma-3-12b-it in the same file).
• 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/6445322

Summary by CodeRabbit

• Tests
  • Updated MMMU integration reference data with an accuracy benchmark for the Gemma 4 26B A4B IT model.

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds a 56.667 accuracy reference for google/gemma-4-26B-A4B-it to the MMMU integration data.

Changes

MMMU reference data

Layer / File(s)	Summary
Add Gemma 4 accuracy reference tests/integration/defs/accuracy/references/mmmu.yaml	Adds an accuracy: 56.667 entry under the Gemma 4 model reference list.

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

Possibly related issues

• Issue 16275: Adds the missing Gemma 4 MMMU accuracy reference coverage.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The description covers summary, test plan, and link, but it omits the required Description, Test Coverage, and PR Checklist sections.	Add the missing template sections and fill in a short Description, Test Coverage details, and the PR Checklist confirmation.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is specific and matches the main fix to restore the missing bf16 accuracy entry.
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.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@tests/integration/defs/accuracy/references/mmmu.yaml`:
- Line 17: Add the standard SPDX copyright and license comment header to
mmmu.yaml, matching the format and values used by neighboring accuracy reference
YAML files, before the existing accuracy entry.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 83f41425-04fc-4af7-8ce5-4bbff913c7b7

📥 Commits

Reviewing files that changed from the base of the PR and between 53d2bde and 0cc697f.

📒 Files selected for processing (1)

• tests/integration/defs/accuracy/references/mmmu.yaml

[2ez4bz]

/bot run

[tensorrt-cicd]

PR_Github #58999 [ run ] triggered by Bot. Commit: e5350dc Link to invocation

[tensorrt-cicd]

PR_Github #58999 [ run ] completed with state SUCCESS. Commit: e5350dc
/LLM/main/L0_MergeRequest_PR pipeline #47528 completed with status: 'SUCCESS'

CI Report

Link to invocation