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

feat: Added ImageReward, Clip and Blip scorers, fixed download of aesthetic scorers and gave the option to average scorers #110

Open
wants to merge 18 commits into
base: main
Choose a base branch
from

Conversation

WhiteWipe
Copy link

WIP

Updated, added blip, clip, ir and fixed aesthetic
Fixed config template
Fix missing config for BLIP and ImageReward
Fixed blip config(med.json) not being downloaded right
Major refactor, added HPSv2 and Pick-a-Pic scorers, geometric and exponential averages for both payloads and scorers, and possibly different devices for different scorers, fixed manual scoring, etc
Added laionv1, shadow aesthetic, cafe aesthetic and wd aesthetic
Laionv1 and chad appear to be broken
Couldn't get it to work in any way, shape or form(wasn't worth it anyway probably)
@WhiteWipe
Copy link
Author

New Scorers Currently Added with Links:
ImageReward: https://github.com/THUDM/ImageReward
BlipScore: Implementation from IR repository
ClipScore: Implementation from IR repository
Pick-A-Pic: https://github.com/yuvalkirstain/PickScore
HPSv2: https://github.com/tgxs002/HPSv2
The classifier used to filter the dataset for waifu-diffusion: https://huggingface.co/hakurei/waifu-diffusion-v1-4/tree/main/models
Cafe_Aesthetic: https://huggingface.co/cafeai/cafe_aesthetic
Shadow_Aesthetic: https://huggingface.co/shadowlilac/aesthetic-shadow

Added default scorer weight and default score device, also replaced exponential average with quadratic
@Koitenshin
Copy link

If you want to add a really good super strict image scorer, check out this HF repo. Just make sure you grab the pull request from the bot under the Community tab. It has the safetensors file you'll want. It has two options, 'good' is super strict while 'normal' rates everything pretty highly.

I've tested almost every single image scorer there is, and there's only a couple good ones. You really don't want an image scorer that's going to rate something highly if the image is actually bad.

#wkpark/sd-webui-model-mixer#148

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.

2 participants