警告：

这是一个非常早期的测试版本，样本来源单一，会非常容易绘制出真实存在的某人。

请勿使用此模型来生成任何可能令人感到不适的图像（例如????题材）。

Warning:

This is a very early testing version with a single source of samples, which can easily depict someone who truly exists.

Do not use this model to generate any potentially uncomfortable images (such as ???? themes).

介绍：

这是一个使用真实照片训练的SD1.5微调模型，专注于绘制女性/服装/模特。

因为使用BLIP2为图片打标，我十分推荐使用自然语言来作为提示词。

Introduction:

This is an SD1.5 fine-tuning model trained using real photos, focusing on drawing women/clothing/models.

Because BLIP2 is used to label images, I highly recommend using natural language as a prompt word.

关于v0.1版本：

此模型尚未进行过任何基于合并的微调，这意味着其绘制性能尚未达到良好状态。

这也意味着VAE层没有被损坏，并且不需要使用外部的VAE。

Regarding v0.1 version:

This model has not undergone any merge based fine-tuning, which means its rendering performance has not yet reached a good state.

It also means that the VAE layer is not damaged and there is no need to use an external VAE.

训练所用到的设备是一块RTX4090，购买这个GPU使我的信用卡严重透支。

The equipment used for the training is an RTX4090, and purchasing this GPU caused my credit card to be severely overdrawn.

欢迎留下你的评论，这对我很重要。

关于如何进一步改进这个模型我还没什么头绪，也许你的建议能帮到我。

Welcome to leave your comment, it's important for me.

I don't have any ideas on how to further improve this model, perhaps your suggestion can help me.