search

zsyOAOA / ResShift

ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting (NeurIPS 2023 Spotlight)
Other
648 stars 36 forks source link

Dependencies in environment.yml #9

Open korvin2000 opened 10 months ago

korvin2000 commented 10 months ago

I've seen and installed all sorts of projects, including stable diffusion, traiNNer, chaiNNer, and all sorts of Upscaler's stuff. But in none of them have I seen so many libraries as dependencies. Have you tried removing unnecessary and redundant libraries? Or do you need every single library EVER WRITTEN for python to simple run/test ResShift stuff? it's really a hell of different python libraries of all kinds and also tightly version and OS specific.

zsyOAOA commented 10 months ago

Please refer the requirement.txt. @korvin2000

korvin2000 commented 10 months ago

Please refer the requirement.txt. @korvin2000

thanks! I'll try to use requirement.txt only

korvin2000 commented 10 months ago
image

it works, thanks.

For photorealistic images / photos it doesn't outperforms old ESRGAN - 4x_NMKD-Siax_200k Model, but it looks pretty good.. Could be trained to produce better results....

lmvgjp commented 10 months ago

hello, could you please provide a link to the model you found to be better, ESRGAN - 4x_NMKD-Siax_200k ? thank you