-
Notifications
You must be signed in to change notification settings - Fork 2.4k
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
dep(diffusers): upgrade diffusers to 0.20 #4311
Conversation
Removed `is_safetensors_available` as safetensors is now a required dependency of diffusers.
I did a quick test on Linux + xformers of generation stuff for SD 1.x, 2.x, and SDXL, and some inpainting, and that all seemed fine. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There is still code in backend/model_management/convert_ckpt_to_diffusers that uses the diffusers is_safetensors_available() method. I'm guessing those will currently break model manager when conversion to diffusers is needed.
…is now a required dependency.
oh, ckpts, those things. Good catch, Gregg! I've removed those references to the function and did another search to be sure. I didn't find any |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM!
[TAESD - Tiny Autoencoder for Stable Diffusion](https://github.com/madebyollin/taesd) - is a tiny VAE that provides significantly better results than my single-multiplication hack but is still very fast. The entire TAESD model weights are under 10 MB! This PR requires diffusers 0.20: - [x] #4311 ## To Do Test with - [x] SD 1.x - [ ] SD 2.x: #4415 - [x] SDXL ## Have you discussed this change with the InvokeAI team? - See [TAESD Invocation API](https://discord.com/channels/1020123559063990373/1137857402453119166) ## Have you updated all relevant documentation? - [ ] No ## Related Tickets & Documents <!-- For pull requests that relate or close an issue, please include them below. For example having the text: "closes #1234" would connect the current pull request to issue 1234. And when we merge the pull request, Github will automatically close the issue. --> - Related Issue # - Closes # ## QA Instructions, Screenshots, Recordings Should be able to import these models: - [madebyollin/taesd](https://huggingface.co/madebyollin/taesd) - [madebyollin/taesdxl](https://huggingface.co/madebyollin/taesdxl) and use them as VAE. <!-- Please provide steps on how to test changes, any hardware or software specifications as well as any other pertinent information. --> ## Added/updated tests? - [x] Some. There are new tests for VaeFolderProbe based on VAE configurations, but no tests that require the full model weights.
This allows us to work with some new stuff including TAESD: https://github.com/huggingface/diffusers/releases/tag/v0.20.0
Removed
is_safetensors_available
as safetensors is now a required dependency of diffusers.What type of PR is this? (check all applicable)
QA Instructions, Screenshots, Recordings
Added/updated tests?