LTXV: Accumulate VAE decode results on intermediate_device

[kijai]

Move VAE decode results to intermediate_device (generally CPU).

Since the model outputs can be huge, this should reduce the peak VRAM consumption by considerate amount.

[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: Pro

Run ID: 16be4501-2f60-4bb6-88f8-ecc3bbbe555d

📥 Commits

Reviewing files that changed from the base of the PR and between 4941cd0 and dd9f714.

📒 Files selected for processing (1)

• comfy/ldm/lightricks/vae/causal_video_autoencoder.py

---

📝 Walkthrough

Walkthrough

The pull request modifies the causal video autoencoder implementation to adjust device handling. An import for comfy.model_management has been added. In the run_up method, tensor samples are now explicitly transferred to an intermediate device via comfy.model_management.intermediate_device() before being appended to the output list. This changes where output samples reside during processing. The modification involves minimal code changes with no alterations to public method signatures.

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% 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 accurately describes the main change: moving VAE decode results to intermediate_device in the LTXV VAE component.
Description check	✅ Passed	The description clearly explains the purpose of the change: reducing peak VRAM consumption by moving VAE outputs to intermediate_device.

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

📝 Coding Plan

• Generate coding plan for human review comments

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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