就目前，我测试得来的结果而言 它是个失败的模型，原因是其中混入了太多我SD1.5时代的数据集（512*512分辨率）严重的影响了训练的质量不过好消息是，我将我的数据集发给了Chenkin,它可能会在一个星期左右开始训练nekopara all girl-Pony目前可用的角色词与我之前发布的lora一样{ Name_$nekopara$ } 如 chocola_$nekopara$风格词为NK_New_sytle,NK_Old_sytle,NK_TV_sytle,NK_Fan_sytle...可以等待后续chenkin训练的nekopara模型等chenkin训练完成后 我会再次训练一次nekopara的Fine-Tune Based on my current test results, it is a failed model. The reason is that too much of my dataset from the SD1.5 era (512*512 resolution) was mixed in, which severely affected the training quality. However, the good news is that I have sent my dataset to Chenkin, and they might start training the Nekopara All Girl-Pony model in about a week.The currently available character tags are the same as those I previously released for the LoRA, such as { Name_$$nekopara$$ } like chocola_$$nekopara$$. The style tags are NK_New_style, NK_Old_style, NK_TV_style, NK_Fan_style, etc.We can wait for Chenkin to complete the training of the Nekopara model. After Chenkin finishes training, I will fine-tune the Nekopara model once again.

Files with SHA256: 78f0a8b3dbe7305eaabd6ccb1530587ca21c0f65332a485ee97e5e40d719ed3c