I'm trying the onsite lora trainer. Don't use this, I have no idea if it does anything<hr />From my tests though, it somehow manages to create quite high quality images. I'm just not sure if it's what I'm looking for or if it follows the dataset well enough.If you do use it for some reason, please post images and/or leave comments. It would be nice to know about everyone's experience.

Test

Trained on 20 imagesNone of the images were captioned

Files with SHA256: a794098830e333a993e909cf8fe5e08689cda4cbb0c1eadccaebb0ef36bb8ebe