[https://nvbugs/6428124][fix] Include use_host_stop_criteria in the PP send/recv payload in…

[trtllm-agent]

Summary

• Root cause: On non-last PP ranks, sample_state.use_host_stop_criteria stayed False while the last rank's fast host-stop path produced host.finish_reasons=None, so update_requests entered process_draft_tokens with an empty finish_reasons list and raised IndexError.
• Fix: Include use_host_stop_criteria in the PP send/recv payload in _ring_broadcast_sample_state and restore it on the receiver, using hasattr/getattr to stay compatible with non-Torch sampler flavors; remove the corresponding waives.txt entry.
• 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/6428124

Summary by CodeRabbit

• Bug Fixes

  • Improved pipeline-parallel generation to keep stop-criteria handling consistent across stages, helping requests stop at the expected time.

• Chores

  • Removed one outdated test waiver from the skip list.

[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: bcdc74a8-0578-4eaa-bd94-3eb1980eb70a

📥 Commits

Reviewing files that changed from the base of the PR and between 5ce293f and 26e810c.

📒 Files selected for processing (2)

• tensorrt_llm/_torch/pyexecutor/py_executor.py
• tests/integration/test_lists/waives.txt

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

Modified the pipeline-parallel ring broadcast of SampleState to propagate an additional use_host_stop_criteria flag alongside existing data on both receive and send paths. Also removed one test waiver entry unrelated to this change.

Changes

PP Ring Broadcast Flag Propagation

Layer / File(s)	Summary
Ring broadcast flag propagation tensorrt_llm/_torch/pyexecutor/py_executor.py	Receive path now unpacks a 3-tuple including use_host_stop_criteria and conditionally applies it to sample_state; send path appends the flag via getattr(sample_state, "use_host_stop_criteria", False) to the isend_object payload.
Test waiver removal tests/integration/test_lists/waives.txt	Removed a waiver entry for test_lora_pp2_sanity previously skipped with reference to nvbugs/6428124.

Estimated code review effort: 2 (Simple) | ~10 minutes

Sequence Diagram(s)

sequenceDiagram
  participant PrevPPRank
  participant CurrentPPRank
  participant NextPPRank
  PrevPPRank->>CurrentPPRank: recv_object(host, py_result_diffs, use_host_stop_criteria)
  CurrentPPRank->>CurrentPPRank: set sample_state.use_host_stop_criteria if attribute exists
  CurrentPPRank->>NextPPRank: isend_object(host, py_result_diffs, use_host_stop_criteria)

Loading

Suggested reviewers: yuxianq, pcastonguay, dongxuy04

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is specific and matches the bug fix, ticket, and payload change described in the PR.
Description check	✅ Passed	The PR explains the issue, fix, test plan, and bug link, but it does not follow the template headings exactly.
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.}

[chienchunhung]

A fix was being worked on: #16096

[BowenFu]

LGTM — adds use_host_stop_criteria to the PP ring send/recv payload symmetrically (send uses getattr(...,False), recv guards with hasattr); only affects multi-PP-rank sample-state broadcast, and PP=1/default bypasses the loop entirely.