Text-to-Image
Diffusers
lora

Overtraining

#3
by DodoPapa - opened

The model is overtrained, all the images I get are too overcooked. If you are making the next release, please take this into account.

The model is overtrained, all the images I get are too overcooked. If you are making the next release, please take this into account.

Did you try lowering the lora weight?

Did you try lowering the lora weight?

Of course and it's pointless.

Fixed read next comment

If you're using the ComfyUI workflow did you make sure you set CFG to 1 and set the unconditional model to bypass? That's the issue I was facing with overbaked images. Now the output looks good.

Sign up or log in to comment