[https://nvbugs/5938603][fix] Fix E/PD disagg chunked prefill bug

[2ez4bz]

Summary by CodeRabbit

Release Notes

• New Features

  • Added enable_chunked_prefill parameter to enable chunked prefill processing, allowing more efficient token handling during multimodal operations.
  • Added max_num_tokens parameter to specify a token budget constraint for multimodal processing workflows.

• Bug Fixes

  • Fixed multimodal embedding detection logic to correctly identify embeddings in edge cases where values may be falsy.

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)

• 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]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between bb5cf9b and 7b2f92b.

📒 Files selected for processing (3)

• tensorrt_llm/_torch/models/modeling_qwen2vl.py
• tensorrt_llm/_torch/models/modeling_qwen3vl.py
• tests/unittest/_torch/multimodal/test_mm_encoder_standalone.py

---

📝 Walkthrough

Walkthrough

These changes modify multimodal data handling in Qwen model implementations to use explicit None checks instead of truthiness checks for multimodal_embedding, and add enable_chunked_prefill and max_num_tokens parameters to multimodal encoder test cases.

Changes

Cohort / File(s)	Summary
Qwen Model Multimodal Data Checks tensorrt_llm/_torch/models/modeling_qwen2vl.py, tensorrt_llm/_torch/models/modeling_qwen3vl.py	Changed multimodal_embedding readiness conditions from truthiness checks to explicit is not None comparisons, ensuring falsy but non-None values (e.g., empty tensors) are still treated as present multimodal inputs.
Multimodal Encoder Test Parameters tests/unittest/_torch/multimodal/test_mm_encoder_standalone.py	Added enable_chunked_prefill and max_num_tokens keyword arguments to LLM constructor calls in multimodal encoder tests for both primary and optional decode paths.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 1 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The PR description is empty except for the template structure; no concrete explanation of the issue, solution, or test coverage is provided despite template sections being available.	Fill in the Description section explaining the bug and fix, and the Test Coverage section listing relevant tests that validate the changes.
Docstring Coverage	⚠️ Warning	Docstring coverage is 33.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the bug fix (E/PD disagg chunked prefill) and includes the NVBugs ticket reference, providing meaningful context about the changeset.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

Tip

Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs).
Share your feedback on Discord.

---

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

[2ez4bz]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #37157 [ run ] triggered by Bot. Commit: 7b2f92b Link to invocation

[tensorrt-cicd]

PR_Github #37157 [ run ] completed with state SUCCESS. Commit: 7b2f92b
/LLM/main/L0_MergeRequest_PR pipeline #28768 completed with status: 'FAILURE'

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[pcastonguay]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #37228 [ run ] triggered by Bot. Commit: 7b2f92b Link to invocation

[tensorrt-cicd]

PR_Github #37228 [ run ] completed with state SUCCESS. Commit: 7b2f92b
/LLM/main/L0_MergeRequest_PR pipeline #28814 completed with status: 'FAILURE'

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[2ez4bz]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #37275 [ run ] triggered by Bot. Commit: 8af575f Link to invocation

[tensorrt-cicd]

PR_Github #37275 [ run ] completed with state SUCCESS. Commit: 8af575f
/LLM/main/L0_MergeRequest_PR pipeline #28848 completed with status: 'FAILURE'

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[2ez4bz]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #37363 [ run ] triggered by Bot. Commit: 34f612c Link to invocation

[brb-nv]

LGTM.

[tensorrt-cicd]

PR_Github #37363 [ run ] completed with state SUCCESS. Commit: 34f612c
/LLM/main/L0_MergeRequest_PR pipeline #28921 completed with status: 'FAILURE'

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[2ez4bz]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #37388 [ run ] triggered by Bot. Commit: 70528c1 Link to invocation

[tensorrt-cicd]

PR_Github #37388 [ run ] completed with state SUCCESS. Commit: 70528c1
/LLM/main/L0_MergeRequest_PR pipeline #28939 completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

Link to invocation