Type | |
Stats | 263 |
Reviews | (30) |
Published | Oct 14, 2024 |
Base Model | |
Usage Tips | Clip Skip: 2 Strength: 2 |
Trigger Words | yuhui |
Hash | AutoV2 16907559B5 |
大家好这是我第一个lora,肯定有不足,不过在我测试下可以正常工作,下面的是测试图片,所用大模型为WAI-SemiReal - v3.0 | Stable Diffusion Checkpoint | Civitai
实测可以使用,测试时使用1024*1024,建议lora强度0.8-1.0,步数30-40,采样方法选择
Euler a或者DPM++ 2M都可。没有lora污染的情况下开不开ADetailer都可以,有lora改变画面的情况下可以使用ADetailer修复,Hires. fix高分辨率修复开不开暂时没有看到有什么太大区别。
想训练第二个lora不够Buzz了,麻烦下载的帮忙点个赞谢谢!
Hello everyone, this is my first lora, there must be some shortcomings, but it can work normally under my test, the following is the test picture, the large model used is WAI-SemiReal - v3.0 | Stable Diffusion Checkpoint | Civitai
Actual measurement can be used, the test used 1024*1024, The recommended lora strength is 0.8-1.0, the number of steps is 30-40, and the sampling method is Euler a or DPM+ 2M.
<lora> is relatively few, the ADetailer is recommended not to open it. If lora changes the picture, you can use ADetailer to fix it.Hires. fix High resolution fix on or off so far we haven't seen much difference.
Want to train the second lora is not Buzz enough, please download help point a "like" Thank you!