Dataset Images
#5
by
BhdrC
- opened
Thanks for the awesome work !
I trained the model for different style and consistency was not even close to your ghibli. It is mentioned that dataset includes only 100 asian "faces". I am curious how that limited type of dataset built this lora. Is it possible to share some part of your dataset if not whole? I reckon dataset has played a big role in consistency, since i prefered to add enviromental images to my dataset and cant have the best result.
Appreciate everyone in advance for any feedback