[https://nvbugs/6336682][fix] When spec_config is not None, set piecewise_named_args_fn to a lambda…

[tensorrt-cicd]

Summary

• Root cause: piecewise_named_args_fn returned cm.named_args without cache_seq_interface, so EagleWrapper.forward fell back to the export-time prefill path which can't satisfy the post-transform target_model GraphModule's 458 cached state placeholders.
• Fix: When spec_config is not None, set piecewise_named_args_fn to a lambda returning {**cm.named_args, "cache_seq_interface": cm}; otherwise leave the original behavior. Also drop the now-redundant nvbugs/6336682 waiver.
• 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/6336682

Summary by CodeRabbit

• Refactor

  • Improved model compilation configuration handling to more effectively manage batch size and maximum draft length parameters, with enhanced support for speculative decoding and piecewise compilation workflows.

• Tests

  • Re-enabled accuracy validation test for TensorRT-LLM auto-deployment scenarios that was previously waived due to implementation issues, now indicating successful resolution and improved stability.

[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: 17002e4b-9f2a-4ea6-ba7e-4f4265e592e0

📥 Commits

Reviewing files that changed from the base of the PR and between 07270b7 and 0fb399c.

📒 Files selected for processing (2)

• tensorrt_llm/_torch/auto_deploy/transform/library/compile_model.py
• tests/integration/test_lists/waives.txt

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

In CompileModel._apply_to_full_model, a new _named_args() helper conditionally adds cache_seq_interface to named args when spec_config is present. _get_args_kwargs is updated to pass max_draft_len to set_capture_batch when spec_config exists. The piecewise compilation path is updated to use _named_args. One CI test waiver for TestNemotronSuperV3::test_mtp is removed.

Changes

Speculative decoding compile-model fix and waiver cleanup

Layer / File(s)	Summary
spec_config-aware named args and capture batch tensorrt_llm/_torch/auto_deploy/transform/library/compile_model.py, tests/integration/test_lists/waives.txt	_named_args() helper conditionally augments named args with cache_seq_interface when spec_config is present; _get_args_kwargs passes max_draft_len to set_capture_batch; piecewise path switches from a plain cm.named_args lambda to _named_args; the waiver for TestNemotronSuperV3::test_mtp[fp8_ws4_80gb-trtllm] is removed.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~5 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#15510: Modifies tests/integration/test_lists/waives.txt to adjust CI waiver entries for test_llm_api_autodeploy.py variants, the same file modified here.
• NVIDIA/TensorRT-LLM#15511: Removes SKIP waiver entries for test_llm_api_autodeploy.py in waives.txt, directly overlapping with this PR's waiver removal.

Suggested reviewers

• crazydemo
• jieli-matrix
• LarryXFly
• StanleySun639
• xinhe-nv

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

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.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is specific and clearly summarizes the main change: setting piecewise_named_args_fn to include cache_seq_interface when spec_config is present.
Description check	✅ Passed	The PR description includes a clear summary of the root cause, the fix, test coverage verification, and links to the relevant bug. It covers all essential sections required by the template.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[hnover-nv]

/bot kill

[hnover-nv]

We disabled piecewise cuda graphs for spec dec for now, so the test was already unwaived. team doesn't have bandwidth to investiagte if this is the suficient fix, so just closing.

[tensorrt-cicd]

PR_Github #56128 [ ] completed with state FAILURE. Commit: 7a298c8
Not allowed on merged PR

Link to invocation