Quoted By:
I think there's a couple images in my dataset that are causing problems with the model, I need to go in and paint over/crop out the non-yukino characters, but other than that, I think this is the best version of the LoRA I've trained yet, at least as far as the details go.