Quoted By:
Alright so I tested out LoRa training using similar parameter. Goal is 1000-1200~ total steps. 1 repeat, 1 batch size, 71 images.
One the previously made "Untagged' LoRa. The other one I just made with tagging using the tagger found in the webui. Did not bother pruning/fixing the tags at all.
Prompt: No quality tag used.
IRyS, hololive, red hair, horns, very long hair, pointy ears, purple eye, heterochromia, blue eye
__clothesA__
(ClothesA is a wildcard)