这是一个效果不佳，且尚待完善的Lora模型，我个人不推荐你下载。

This is a Lora model that is not very effective and needs to be perfected. I personally do not recommend you to download it.

这个模型经常会生成质量不佳的图片。

This model often produces poor quality images.

这个Lora模型是基于final-prune.ckpt（【旧短哈希1d4a34af】新短哈希22fa233c2d），使用了768x768的数据集进行训练。

推荐权重为0.8-1.0。

这个模型有两个用于控制姿势的提示词，from side和from the top。

在训练这个模型的时候，我更新了训练用的软件，可能这是导致训练效果不佳的因素？

非常糟糕的是，我已经浪费了太多的时间在这个模型上面，而且前面数次训练后得到的模型生成的图像完全不能看。

我还在排查问题出在哪里，之后应该还会对这个模型进行更新以达到相对较好的效果，但应该不会在最近……就像我前面说的那样，我浪费了太多的时间在这上面，之后我可能会先继续训练泰坦陨落2的角色Lora。

英语是用Google翻译的，如果有造成误解之处非常抱歉。

This Lora model is based on final-prune.ckpt ([old short hash 1d4a34af] new short hash 22fa233c2d), using a 768x768 dataset for training.

A weight of 0.8-1.0 is recommended.

This model has two cue words for controlling the pose, from side and from the top.

When training this model, I updated the training software, maybe this is the factor that leads to the poor training effect?

It is very bad that I have wasted too much time on this model, and the images generated by the model obtained after the previous few trainings are completely unreadable.

I'm still figuring out what the problem is, and I'm supposed to update this model to get better results, but not in the near future... Like I said before, I wasted too much time on On top of that, I might continue training the Titanfall 2 character Lora first.

English is translated by Google, sorry if there is any misunderstanding.