这是一个画风lora，画风为画师fajyobore的风格，

使用模型sd_xl_base训练，这是一个实验性的模型，我简单的说一下训练的心路历程吧。

首先，训练这个模型，花了16个小时的时间，dim的值为32并且采用了fp16。我的显卡是4090，在运行的时候训练程序占用了24GB的显存和20GB左右的内存。整个电脑都无法使用。

其次，使用sd_xl_base训练的lora似乎只能用于sd_xl_base上，并不存在泛用性。训练出来的lora所能达到的效果使用别的模型训练似乎也能达到效果。由于训练周期长和训练成本高的缘故，我还没有做进一步的测试。但是这个lora还是不错的，我把它分享出来~

emmmmm我不喜欢炒作，最近关于sdxl的宣传听得我有些烦了。让我就我目前的使用状况做出评价的话目前sdxl只能说未来可期，起码我自己宁愿去使用我的老模型们。

This Lora can change the art style, which is in the style of the artist Fajyobore.

It is trained using the sd_xl_base model, an experimental model. Let me briefly explain the process of training.

Firstly, training this model took 16 hours, with the dim value set to 32 and using fp16. My graphics card is a 4090, and during operation, the training program occupied 24GB of VRAM and about 20GB of RAM. My entire computer was not available for use.

Secondly, the Lora trained with sd_xl_base seems to only work on sd_xl_base and lacks universality. The effect that the trained lora can achieve seems to be able to achieve the effect using other model training. Due to the long training cycle and high training costs, I have not done further testing yet. However, this Lora is pretty good, and I decided to share it~

emmmmmmm,I don't like hype, and the hype around sdxl has gotten me a little tired lately. If I were to evaluate it based on my current usage, I would say sdxl is promising for the future, but for now, I'd rather stick to using my old models.