search

mihirp1998 / AlignProp

AlignProp uses direct reward backpropogation for the alignment of large-scale text-to-image diffusion models. Our method is 25x more sample and compute efficient than reinforcement learning methods (PPO) for finetuning Stable Diffusion
https://align-prop.github.io/
MIT License
243 stars 8 forks source link

run the aesthetic_scorer.py #11

Closed xinli2008 closed 1 year ago

xinli2008 commented 1 year ago

nice job! I want to use the aesthetic_scorer.py to test the aesthetic scorer about my images. But i came across some problems, can you give me some useful advice?

code: image image

the code is wrong prehaps beacuse of the input_image format. can you tell me what should i process the image before generate the aesthetic scorer

mihirp1998 commented 1 year ago

you can follow this function:

https://github.com/mihirp1998/AlignProp/blob/b519ad87484cdafcfca761013928e59a3fac0923/main.py#L89