Skip to content

WAN 2.2 5B it2v Support: Add support for WAN 2.2 VAE and WAN 2.2 it2v 5B model, Fix hacky code in WAN 2.1 VAE#381

Open
jzr99 wants to merge 1 commit intomainfrom
wan2p2vae
Open

WAN 2.2 5B it2v Support: Add support for WAN 2.2 VAE and WAN 2.2 it2v 5B model, Fix hacky code in WAN 2.1 VAE#381
jzr99 wants to merge 1 commit intomainfrom
wan2p2vae

Conversation

@jzr99
Copy link
Copy Markdown
Collaborator

@jzr99 jzr99 commented Apr 16, 2026

No description provided.

@jzr99 jzr99 requested a review from entrpn as a code owner April 16, 2026 22:35
@github-actions
Copy link
Copy Markdown

@github-actions
Copy link
Copy Markdown

🤖 Hi @entrpn, I've received your request, and I'm working on it now! You can track my progress in the logs for more details.

Copy link
Copy Markdown

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

## 📋 Review Summary

The pull request introduces support for WAN 2.2 VAE and WAN 2.2 it2v 5B models, while also addressing previously "hacky" frame-reordering logic in the WAN 2.1 VAE. The changes demonstrate a good understanding of the model architecture and JAX/Flax conventions, but some issues were identified in the new autoencoder_kl_wan_2p2.py file, including a critical type-hint/default-value error and leftover debug code.

🔍 General Feedback

  • Code Cleanliness: The new autoencoder_kl_wan_2p2.py file contains an excessive amount of commented-out code. Removing these will significantly improve maintainability and readability.
  • Simplification: Updates to embeddings_flax.py correctly enable multi-dimensional timestep support, which can be further leveraged to simplify the transformer logic.
  • Robustness: Fixed a frame-reordering issue in WAN 2.1 VAE which improves the overall correctness of the video generation pipeline.

Comment thread src/maxdiffusion/models/wan/transformers/transformer_wan.py
Comment thread src/maxdiffusion/models/wan/autoencoder_kl_wan_2p2.py
Comment thread src/maxdiffusion/models/wan/autoencoder_kl_wan_2p2.py Outdated
Comment thread src/maxdiffusion/models/wan/autoencoder_kl_wan_2p2.py
Comment thread src/maxdiffusion/models/wan/autoencoder_kl_wan_2p2.py
Comment thread src/maxdiffusion/models/wan/autoencoder_kl_wan_2p2.py
@jzr99 jzr99 force-pushed the wan2p2vae branch 2 times, most recently from 6efb123 to ab10595 Compare April 16, 2026 23:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants