English:The model uses works from Pixiv artist Kira きらまるこう, hoping to learn bright colors and incorporate similar materials into the model for better generalization. During testing, it was found that the parameters of the original image have some impact on the quality of the output. However, so far only CFG has shown a pattern, which will be explained later.Advantages:<ol><li>Distinct style with bright colors</li></ol>Disadvantages:<ol><li>Watermarks </li></ol>Below are some recommended parameters (if not listed, they have minor impact):Model weight: 0.5-1.0Steps: 20-40CFG scale: 3.0-12.5 (The larger the CFG, the more vivid the colors and the more creative the output, and vice versa. Usually recommended between 7-10)Clip skip: 1-2 (default is clip=2, but trained with clip=1)(High-resolution restoration recommended)<hr />中文：该模型使用了P站画师：きらまるこう老师的作品，希望模型能够学到明亮的色泽，并在模型中加入同类型素材希望模型能够具有较好的泛化性。在测试过程中发现生图参数对于出图质量有些影响，不过目前只有CFG找到规律，后面会说明。优点：1.风格明显，具有明亮的色彩缺点：1.水印下面推荐一下参数（没有写到则默认影响不大）：模型权重：0.5-1.0Steps: 20-40,CFG scale: 3.0-12.5,（CFG越大，颜色越鲜艳，出图画面越具有创造性，反之亦然。通常情况建议7-10）Clip skip: 1-2，（默认clip=2，但是是使用clip=1训练的）（建议使用高分辨率修复）

Files with SHA256: 3c38b755c749017880e9a5161dcf852a0d9fd3fd6cad2885803731a1def58b91