Pikachu LoRA trained on NovelAI with Kohya training scripts using the DAdaptAdam optimizer at 768x768 resolution. Very consistent, biggest issue I ran into was tail placement. Can occasionally add an extra ear to the head. Tail shape can be unstable, don't expect the male and female pikachus to have the proper tail shape (cleft for female, flat for male). Should have wide compatibility with model mixes based on NovelAI, which is most of them. Trained on a mixed male/female dataset.