LoRA trained on some of my favorite Hassaku (PDXL) images that I have collected over a few months of image generation. I like the LoRA, though it seemed to learn too much from the 5 Blue Archive images I had in the training data.