[https://nvbugs/6490033][fix] Relaxed the assertion to accept both is_eagle3() and…

[trtllm-agent]

Summary

• Root cause: Static-tree assertion in update_spec_dec_param only accepted EAGLE3 mode, but Eagle3OneModelSpecMetadata legitimately sets is_spec_dec_tree=True for EAGLE3_ONE_MODEL when eagle_choices is provided.
• Fix: Relaxed the assertion to accept both is_eagle3() and is_eagle3_one_model(); removed the two nvbugs/6490033 waivers in 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/6490033

Dev Engineer Review

• Updated update_spec_dec_param to relax the static-tree assertion so spec_metadata.spec_dec_mode may be either is_eagle3() or is_eagle3_one_model() (enables Eagle3OneModelSpecMetadata / is_spec_dec_tree=True behavior).
• Removed two corresponding waivers from tests/integration/test_lists/waives.txt tied to Eagle3 draft token tree quickstart coverage.
• The change is small and targeted; however, the assertion message still says “Tree decoding is only supported for Eagle3 now” despite now allowing Eagle3-one-model, which may be worth revising for clarity.

QA Engineer Review

• Modified: tests/integration/test_lists/waives.txt
  • Removed waivers for:
    • test_e2e.py::test_draft_token_tree_quickstart_advanced_eagle3 (Llama-3.1-8B Instruct / Eagle3)
    • test_e2e.py::test_draft_token_tree_quickstart_advanced_eagle3_depth_1_tree (Llama-3.1-8B Instruct / Eagle3)
• Test-db/qa files: none changed.
• Verdict: needs follow-up

[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: 527605c3-9ed3-436a-8472-3356a2f5f5ee

📥 Commits

Reviewing files that changed from the base of the PR and between d037559 and cb90128.

📒 Files selected for processing (1)

• tests/integration/test_lists/waives.txt

🚧 Files skipped from review as they are similar to previous changes (1)

• tests/integration/test_lists/waives.txt

---

Walkthrough

Tree-decoding validation now accepts Eagle3-one-model mode. Two corresponding Eagle3 quickstart integration tests are removed from the waiver list.

Changes

Eagle3 tree decoding

Layer / File(s)	Summary
Expand tree-mode validation and test coverage tensorrt_llm/_torch/attention_backend/trtllm.py, tests/integration/test_lists/waives.txt	The static-tree assertion accepts Eagle3-one-model, and two Eagle3 quickstart tests are removed from the integration waiver list.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#12062: Adds Eagle3 dynamic-tree handling related to Eagle3-one-model mode support.

Suggested reviewers: jieli-matrix, perkzzheng, bowenfu, brnguyen2, crazydemo

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title matches the bug fix and includes the required ticket/type format, though it is slightly truncated.
Description check	✅ Passed	The description covers the issue, fix, test plan, and link, with only minor template sections like PR checklist left implicit.
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 test_e2e.py::test_draft_token_tree_quickstart_advanced_eagle3[Llama-3.1-8b-Instruct-llama-3.1-model/Llama-3.1-8B-Instruct-EAGLE3-LLaMA3.1-Instruct-8B]

[tensorrt-cicd]

PR_Github #61276 [ run ] triggered by Bot. Commit: cb90128 Link to invocation

[tensorrt-cicd]

PR_Github #61276 [ run ] completed with state SUCCESS. Commit: cb90128
LLM_FUNCTION_AUTO_V2C #316 completed with status: 'SUCCESS'
QA verify test: test_e2e.py::test_draft_token_tree_quickstart_advanced_eagle3[Llama-3.1-8b-Instruct-llama-3.1-model/Llama-3.1-8B-Instruct-EAGLE3-LLaMA3.1-Instruct-8B] (NVBug 6490033, branch repair-bot-bug6490033, fork tensorrt-cicd, dry_run_close=true)

Link to invocation