modelscope / modelscope/DiffSynth-Studio

Flash attention question

Open
#14 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
13.1k
Forks
1.3k
Avg merge
13h 12m
Merged PRs (30d)
45

Description

Hi, Great work!!
I have one question: in the paper, you said that "we adopt flash attention [6] in all attention layers, including the text encoder, UNet, VAE, ControlNet models, and motion modules". I found the xformers_forward() function in the Attention module. However, this function is never called during the whole process of "diffutoon_toon_shading.py". It is very strange, since it still can generate high resolution videos. I am very confused how does this work?
Thanks!

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with diffutoon_toon_shading.py and trace the attention path into the Attention module. Compare that path with Attention.xformers_forward() and determine whether flash attention is invoked indirectly or bypassed. Done means documenting the actual call path and explaining how the reported high-resolution video generation is possible.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.