Skip to content

[TRTLLM-11288][fix] Adapt LTX2 pipeline to CompilationConfig warmup interface#12232

Merged
luyiyun1021 merged 2 commits into
NVIDIA:mainfrom
luyiyun1021:fix/trtllm-11288-ltx2-warmup-config
Mar 16, 2026
Merged

[TRTLLM-11288][fix] Adapt LTX2 pipeline to CompilationConfig warmup interface#12232
luyiyun1021 merged 2 commits into
NVIDIA:mainfrom
luyiyun1021:fix/trtllm-11288-ltx2-warmup-config

Conversation

@luyiyun1021
Copy link
Copy Markdown
Collaborator

@luyiyun1021 luyiyun1021 commented Mar 16, 2026 •
edited by coderabbitai Bot
Loading

Summary by CodeRabbit

  • Refactor
    • Updated the LTX2 visual generation pipeline warmup mechanism to support explicit per-call parameter specification (height, width, frame count, and steps) instead of batch-based predefined configurations.

Description

Adapt LTX2 pipeline to the new CompilationConfig warmup interface introduced in #12107.

LTX2 was merged after #12107 and still uses the old warmup interface (common_warmup_shapes + _run_warmup(warmup_steps)). This PR migrates it to the new interface:

  • common_warmup_shapes → default_warmup_resolutions + default_warmup_num_frames
  • _run_warmup(warmup_steps) → _run_warmup(height, width, num_frames, steps)
  • warmup seed 0 → 42 (consistency with other pipelines)

Test Coverage

  • Existing test_warmup.py tests cover the base pipeline warmup logic
  • LTX2-specific tests in CI use skip_warmup=True and are not affected

PR Checklist

  • PR description clearly explains what and why.

  • PR Follows TRT-LLM CODING GUIDELINES.

  • Test cases are provided for new code paths.

  • Any new dependencies have been scanned for license and vulnerabilities.

  • CODEOWNERS updated if ownership changes.

  • Documentation updated as needed.

  • Update tava architecture diagram if significant design change.

  • 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.

@luyiyun1021
[TRTLLM-11288][fix] Adapt LTX2 pipeline to CompilationConfig warmup i…
8eb9c14 
…nterface

LTX2 was merged after NVIDIA#12107 and still uses the old warmup interface
(common_warmup_shapes + _run_warmup(warmup_steps)). Migrate to the
new interface: default_warmup_resolutions, default_warmup_num_frames,
and _run_warmup(height, width, num_frames, steps).

Signed-off-by: Yiyun Lu <[email protected]>
@luyiyun1021 luyiyun1021 requested a review from a team as a code owner March 16, 2026 03:29
@luyiyun1021
Copy link
Copy Markdown
Collaborator Author

luyiyun1021 commented Mar 16, 2026

/bot run --disable-fail-fast

@coderabbitai
Copy link
Copy Markdown
Contributor

coderabbitai Bot commented Mar 16, 2026 •
edited
Loading

📝 Walkthrough

Walkthrough

This change refactors the warmup mechanism in the LTX2Pipeline class from a batch approach iterating over predefined resolution shapes to an explicit per-call warmup pattern. The _run_warmup() method now accepts explicit height, width, num_frames, and steps parameters instead of a single warmup_steps parameter. Two new properties are introduced to provide default warmup configurations.

Changes

Cohort / File(s) Summary
Warmup Mechanism Refactor
tensorrt_llm/_torch/visual_gen/models/ltx2/pipeline_ltx2.py		 Replaced batch warmup iteration with per-call warmup. Added default_warmup_resolutions() method and default_warmup_num_frames property. Updated _run_warmup() signature to accept explicit height, width, num_frames, and steps parameters. Changed default seed from 0 to 42.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 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 (2 passed)
Check name Status Explanation
Title check ✅ Passed The title clearly and specifically describes the main change: adapting LTX2 pipeline to the new CompilationConfig warmup interface, with proper JIRA ticket reference and type label.
Description check ✅ Passed The PR description clearly explains the migration from old to new CompilationConfig warmup interface with specific API changes, rationale, and test coverage details.

✏️ 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
📝 Coding Plan
  • Generate coding plan for human review comments

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

@tensorrt-cicd
Copy link
Copy Markdown
Collaborator

tensorrt-cicd commented Mar 16, 2026

PR_Github #39020 [ run ] triggered by Bot. Commit: 8eb9c14 Link to invocation

@luyiyun1021 luyiyun1021 requested a review from zhenhuaw-me March 16, 2026 03:39
@luyiyun1021
Copy link
Copy Markdown
Collaborator Author

luyiyun1021 commented Mar 16, 2026

/bot run --disable-fail-fast

@tensorrt-cicd
Copy link
Copy Markdown
Collaborator

tensorrt-cicd commented Mar 16, 2026

PR_Github #39036 [ run ] triggered by Bot. Commit: 8dad25d Link to invocation

@tensorrt-cicd
Copy link
Copy Markdown
Collaborator

tensorrt-cicd commented Mar 16, 2026

PR_Github #39036 [ run ] completed with state SUCCESS. Commit: 8dad25d
/LLM/main/L0_MergeRequest_PR pipeline #30311 completed with status: 'SUCCESS'

CI Report

Link to invocation

@luyiyun1021 luyiyun1021 merged commit 1797e3f into NVIDIA:main Mar 16, 2026
5 checks passed
reasonsolo pushed a commit to reasonsolo/TensorRT-LLM that referenced this pull request Mar 17, 2026
@luyiyun1021 @reasonsolo
[TRTLLM-11288][fix] Adapt LTX2 pipeline to CompilationConfig warmup i…
9b42acb 
…nterface (NVIDIA#12232)

Signed-off-by: Yiyun Lu <[email protected]>
limin2021 pushed a commit to limin2021/TensorRT-LLM that referenced this pull request Mar 19, 2026
@luyiyun1021
[TRTLLM-11288][fix] Adapt LTX2 pipeline to CompilationConfig warmup i…
82f0a9a 
…nterface (NVIDIA#12232)

Signed-off-by: Yiyun Lu <[email protected]>
longcheng-nv pushed a commit to longcheng-nv/TensorRT-LLM that referenced this pull request Mar 31, 2026
@luyiyun1021 @longcheng-nv
[TRTLLM-11288][fix] Adapt LTX2 pipeline to CompilationConfig warmup i…
1208b38 
…nterface (NVIDIA#12232)

Signed-off-by: Yiyun Lu <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@zhenhuaw-me zhenhuaw-me zhenhuaw-me approved these changes

Assignees

@luyiyun1021 luyiyun1021

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@luyiyun1021 @tensorrt-cicd @zhenhuaw-me