在1.1B版本的lora实验获得成功后，重置并优化数据集训练了0.2版本的lora。按比例添加了我之前发布的“GrownRuler”和刚才提到的新训练的0.2版本的lora。此外还添加了rouwei0.8，并且解决了一些rouwei0.8的艺术家tag无视提示词的问题。

After the success of the lora experiment in version 1.1B, the dataset was reset and optimized to train version 0.2 of lora. I have proportionally added the "GrownRuler" I released earlier and the newly trained 0.2 version of lora mentioned just now. In addition, rouwei0.8 has been added and some overfitting issues of artist tags in rouwei0.8 have been resolved.

如果想要油性效果请添加我制作的lora：Mature Oily Effect

If you want an Oily Effect, please add the lora: Mature Oily effect I made

相比1.1B版本，更好的自然语言理解力、更好的画面构图、更好的肢体、更好的提示词依从性，女性年龄不会过于老化，男性角色的年龄也相应降低。

Compared with version 1.1B, it has better natural language comprehension, better picture composition, better body language, and better compliance with prompt words. The female characters are not overly aged, and the age of the male characters is also correspondingly reduced.

这个版本从1.1B成功后开始制作，正如我8天前所说“非常抱歉的通知大家，我将暂时停止模型制作。”，希望大家玩的愉快！

This version began production after the success of 1.1B. Just as I said 8 days ago, "I'm very sorry to inform everyone that I will temporarily stop model production." Hope everyone has a good time!

最后，不要添加任何质量提示词！（例如：masterpiece、best quality）

Finally, do not add any quality prompt words!（For example: masterpiece, best quality）

（关于我为什么在展示图中不放一些看上去就很美的图片，而是放入一些对比图。首先我想让大家知道新版本到底更新了什么，和以前的版本有什么区别，和其他模型有什么区别，到底适不适合您的口味，而不是在您花费时间下载后才后知后觉。其次我更希望大家多多尝试不一样的风格，而不是一味的“一键”生成一张“100分"的图片。）

Regarding why I didn't put some pictures that looked very beautiful in the display pictures but included some comparison pictures instead. First of all, I want everyone to know exactly what has been updated in the new version, what the differences are from the previous version, what the differences are from other models, and whether it suits your taste or not, rather than realizing it only after you have spent time downloading. Secondly, I hope everyone can try more different styles instead of blindly generating a "100-point" picture with just one click.）