This Lora is based on <a rel="ugc" href="https://civarchive.com/models/2598481/ernie-furry">Ernie-Furry</a> 0613 (not publicly available as it was not satisfactory).Ernie itself has some knowledge of the character Ori, and was also simply involved in the training of Ernie-Furry Lora. The main training target of LORA is the character Ori. 50 images were used but repeated twice to increase the weight.Also included were the cat13 images.A small number of other images and NSFW images were added to maintain the original training capabilities, but the generation results were unreliable.Approximately 370 images were involved in the training.The training batch size was 2, and the training resolution was 576 and 640.本lora基于<a rel="ugc" href="https://civarchive.com/models/2598481/ernie-furry">ernie-furry</a> 0613(未公开，因为不太满意)ernie本身具有一些ori的知识，在训练ernie-furry lora时也简单参与。lora 主要训练目标为角色ori，使用了50张图像但重复两次以提高权重。以及cat13图像。少量其它图像和NSFW图像以维持原有的训练后的能力，但不可靠。总共约370张图像参与训练。训练批次为2，训练分辨率576和640。

Files with SHA256: 8402750b5696ebc68062db1420be7b72bc8802e7ee5900136bac8fe0fd104483