Split attention #1218
Draft
Split attention #1218
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
… attention update merge with attention selection, add FLASH_SPLIT
dxqb
force-pushed
the
split_attention
branch
from
4884551 to
ccc07bf
Compare
Collaborator
Author
|
this should now also work on windows, with similar speed-up. Though I have no way to test |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Split attention can significantly speed up higher-resolution training, and somewhat low resolution training:
To test:
then update.
limitations:
probably only for Linux, unless torch have recently improved SDPA in torch 2.8 or 2.9. To be testedOn linux, select
On Windows, install Flash attention and select FLASH_SPLIT
Here are pre-built wheels for Windows by @zzlol63 https://github.com/zzlol63/flash-attention-prebuild-wheels/releases
uses huggingface/diffusers#12870