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.

Sign up for GitHub

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

Jump to bottom

Native Flux nf4 LoRA #3461

Merged
merged 7 commits into from
Sep 30, 2024
Merged

Native Flux nf4 LoRA #3461

merged 7 commits into from
Sep 30, 2024
+279 −23

Conversation

AI-Casanova
Copy link
Collaborator

@AI-Casanova AI-Casanova commented Sep 28, 2024 •
edited
Loading

Flux native LoRA loading

  • Works with Kohya keys and OneTrainer keys
  • Tested with FLUX1 dev nf4 checkpoint
  • Tested for backwards compatibility with SD1.5, SDXL
  • Tested ONLY on T4 GPU (cuda)
  • Option added for requantizing as nf4 or fp4, the latter matches Diffusers results better, the former looks weaker but higher quality at times. Switching requires new lora parameters to be passed (to clear cache)

@AI-Casanova AI-Casanova marked this pull request as ready for review September 29, 2024 21:06
@AI-Casanova
Copy link
Collaborator Author

Note, Dora not yet supported, as that requires more quant manipulation.

@vladmandic
Copy link
Owner

merging as-is, will test and perform edits as needed in-place.

@vladmandic vladmandic merged commit e6ff17d into vladmandic:dev Sep 30, 2024
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@vladmandic vladmandic Awaiting requested review from vladmandic

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@AI-Casanova @vladmandic