This is the model of my latest training ideas

It is very sensitive to tag, overfitting is almost solved, even the training time only needs 2/3 of the original

The example diagram uses the donko-mix nsfw hard model

It is recommended to open Hires fix

这是我最新训练思路所训练出的模型

对tag很敏感，过拟合几乎被解决，甚至训练时间只需要原本的2/3

示例图使用的模型为donko-mix nsfw hard，建议打开高清修复

另外我会接LoRA模型定制单，有意加群找同名管理员私聊1018450371