Another trained LoRA, as the previous one was misbehaving. Recommended strength 0.5-07. trained on khya-ss. Google colab. 1024 resolution. 10 epochs. 12 repeats.
권장 버전: Epoch 9
대안: 더 부드러운 결과를 위해 Epoch 8
권장 가중치: 0.5–0.7
