LoRa trained on 40 images of Elisha Cuthbert's face. 768x768 resolution training. Keep weight between 0.6 and 0.9 when on its own.