What I did to obtain this lora is to train two different LoRas with an almost identical dataset (one dataset have limbless women, in the other have normal women). Both datasets had 4 images, 10 epochs, 15 repeats per image. After that I subtracted both LoRa, and compressed the result so that it weighs 4 mb.
What I did to obtain this lora is to train two different LoRas with an almost identical dataset (one dataset have limbless women, in the other have normal women). Both datasets had 4 images, 10 epochs, 15 repeats per image. After that I subtracted both LoRa, and compressed the result so that it weighs 4 mb.