关于最近反馈的1.2版本的bug：
我猜想大概是Illustrious2.0和rouwei0.8都使用了danbooru数据集进行深度训练，导致大量tag与loli进行了强关联，虽然以前的版本也有这个现象，但是Illustrious2.0和rouwei0.8都是更加深度的训练版本，让这个现象变得更加强烈了，而我在试图解绑1girl与loli的过程中造成了一些tag错乱，（比如钻石双马尾这个tag就有很严重的loli倾向），可能问题点出在这里了。我还花了一晚上测了一下其他热门模型比如hassaku、novaAnime、wai，有时候这些模型甚至会把"milf"当成文本tag处理（比如会在衣服上写上milf的文字）而不是年龄tag。现在的情况是使用Illustrious2.0和rouwei0.8可以获取更多的概念和更新的角色，但是可能对成熟来说这都是bug，现状是我能力有限不能修复，可能也没有精力在更新新版本了，很对不起大家但现实如此，大家可以根据喜好选择老版本，或者期待新的微调制作者站出来“反抗”了。

Regarding the bugs in the recently reported version 1.2:

I guess it's probably because both Illustrious2.0 and rouwei0.8 used the danbooru dataset for deep training, resulting in a large number of tags being strongly associated with loli. Although this phenomenon also occurred in previous versions, However, Illustrious2.0 and rouwei0.8 are both more advanced training versions, making this phenomenon even more intense. During my attempt to unbind 1girl and loli, I caused some tag confusion (for instance, the tag "Diamond Double Ponytail" has a very serious loli tendency). Perhaps the problem lies here. I also spent the whole night testing other popular models such as hassaku, novaAnime, and wai. Sometimes these models even treat "milf" as a text tag (for example, writing the word "milf" on clothes) instead of an age tag. The current situation is that using Illustrious2.0 and rouwei0.8 can provide more concepts and updated characters. However, for the mature version, these might all be bugs. The current situation is that my ability is limited and I can't fix them. Maybe I don't have the energy to update to the new version either. I'm very sorry to everyone. But that's the reality. Everyone can choose the old version according to their preferences. Or we can expect new fine-tuning makers to come forward and "rebel".