Ralts LoRA trained on NovelAI with Kohya training scripts using the DAdaptAdam optimizer at 512x512 resolution. Works very consistently, the main issue I ran into was getting extra head horns in certain poses. You can get the classic 'eyeless' look by prompting for (long bangs, hair covering eyes), though your mileage may vary depending on the base model you're using. Has no idea what Ralts' feet are supposed to look like, so that detail will be filled in by your base model. Should have wide compatibility with model mixes based on NovelAI, which is most of them. Trained on a mixed male/female dataset.