Sane explanation of FreeU attributes
Sane explanation of FreeU attributes
So I recently switched to ForgeUI and they have plenty of features integrated out of the box.
In the context of FreeU, what exactly B and S are modifying? I spent a whole day trying to wrap my head around their exact functionality. Some say keep B>1 and S<1. Some say it adds sharpness and saturation, and the preset on their github didn't work for me, well atleast with the prompt I was tweaking.
But while I was fiddling around with FreeU, I observed lot of drastic modification in the generated image, which gave me hope that it just might be tool to realign the image in accordance to my prompt.
You're viewing a single thread.
references:
main repo: https://github.com/ChenyangSi/FreeU
https://github.com/lllyasviel/Fooocus/discussions/618
FreeU: Free Lunch in Diffusion U-Net Explained: https://piped.privacydev.net/watch?v=eFmkJ_oEW5s
This official tutorial on HF mentions the unsatisfactory defaults https://huggingface.co/docs/diffusers/v0.22.1/en/using-diffusers/freeu
I've been looking for answers since the extension released. Thanks.
https://www.youtube.com/watch?v=mrKk0yYUdnU
The paper mentions it manipulates main net features(B) and skip features(S).
Not sure what the main net features are, but I'm assuming skip features are like clip skip?
Although the paper mentions S didn't do anything, in my tests (sdxl base), it had a tremendous amount of adverse effect on the image generated.
TBH, i enjoy using DPO more than freeU, but I'm curious to learn.
Here is an alternative Piped link(s):
https://www.piped.video/watch?v=mrKk0yYUdnU
Piped is a privacy-respecting open-source alternative frontend to YouTube.
I'm open-source; check me out at GitHub.
Is there a DPO extension?
Haha no. I mean in the context of providing in control, DPO does a better job than FreeU.