Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add ZipLora #5930

Closed
wants to merge 5 commits into from
Closed

Add ZipLora #5930

wants to merge 5 commits into from

Conversation

shauray8
Copy link
Contributor

What does this PR do?

Adds ZipLoRA: Any Subject in Any Style by Effectively Merging LoRAs

Fixes #5925

Who can review?

@sanchit-gandhi @patrickvonplaten

@AmericanPresidentJimmyCarter

@patrickvonplaten
Copy link
Contributor

patrickvonplaten commented Nov 27, 2023

Should we maybe try to directly add this to PEFT cc @younesbelkada @pacman100 @sayakpaul

@sayakpaul
Copy link
Member

Adding to peft sounds good to me. But for training, I think we should encourage the community to use the scripts developed by @mkshing here: https://github.com/mkshing/ziplora-pytorch. They're quite elegant and are fully diffusers compatible.

@shauray8
Copy link
Contributor Author

The training script by @mkshing is pretty neat actually, I'm not sure if you want this to be an inference-only addition to PEFT or if it is something else.

@younesbelkada
Copy link
Contributor

Yes it would make sense to add ZipLora directly in PEFT and use it in diffusers. You can take a look at a similar PR here: huggingface/peft#1160
How does that to you @shauray8 ?

@shauray8
Copy link
Contributor Author

shauray8 commented Dec 4, 2023

Sounds good, I might need a little help moving forward @younesbelkada. Closing this and opening up a PR in PEFT.

@shauray8 shauray8 closed this Dec 4, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Add ZipLora to enhance multi-lora inference
4 participants