An experimental LoRA trained on the entirety of the marefusion 2.2 dataset (49211 samples) for 1 epoch, can be used to inject pony knowledge into any SD1.5 model
first experimental version, let's see how cursed it gets
marefusion_mini.safetensors