R-Anime-Sytle-Add-prune_LoRA版本v1.0 (ID: 579015)

Trained on R-Anime 220K Additional training on R-Anime-Loli-Plus So, it should perform very well on loli (I think so, but I'm not sure if there's catastrophic forgetting; it lacks testing) The dataset used 50,000 images, randomly extracted from the Best-A rated images in R-Anime-3mil The main training focus is on style (specifically, lighting and more stunning styles... like wlop, torino_aqua...) Some AI styles from my online friends were also included in the training (they were hidden because they have their own model styles, and I didn't want this to affect my friends) Regarding the training styles... I was too lazy to list all the training styles So I used a script to extract all the tags I trained into a YAML file Please check the text file at the bottom of the model card

---

The training used a 4090 for 40 hours Regarding R-Anime-3mil, it will be trained based on SDXL The estimated training time is 48 days (the images have been preprocessed) The main focus will be on loli, nsfw, and style Don't ask why the main focus is on loli, because I am a lolicon!
Loli Plus was removed from Civitai (because it was too loli, haha~ official recognition)

基于R-Anime 220K进行训练 在R-Anime-Loli-Plus上进行追加训练 所以他在loli的表现上应该会很棒（我是这样想的，具体有没有发生灾难性遗忘我也不清楚，它缺乏测试） 数据集使用了5w图片 是从R-Anime-3mil中随机拆出来的一部分Best-A评级的图片 主要训练方向为风格（嗯...准确来讲是光照，以及比较惊艳的风格吧...比如wlop,torino_aqua这种...) 训练加入了部分AI风格（它们被隐藏了，因为它们拥有他们自己的模型风格，我并不希望这会影响到这个风格的作者,） 关于训练的风格...我懒得去把训练列表列出来 于是我用脚本把我训练的所有tag全部提取到了一个yaml中 请查看模型卡下方的文本文件

———— 训练使用了4090训练了40小时 关于R-Anime-3mil 它会基于SDXL进行训练 训练时间预计为48天(目前图片已经预处理完成) 主要方向为loli，nsfw，sytle 不要问为什么主方向为loli，因为我是萝莉控！
loli plus被civitai下架了（因为它太过于loli了 哈哈哈~官方の认可)

描述:

训练词语: