Sign In

AsianMale

2.5k
18.3k
159
Verified:
SafeTensor
Type
LoRA
Stats
18,338
Reviews
Published
Feb 26, 2023
Base Model
SD 1.5
Training
Steps: 50
Epochs: 10
Trigger Words
1man
Hash
AutoV2
D30EEAE2EE

初始版本,主要为亚洲男性考虑,没有处理好衣服数据,lora 权重过大时,会导致衣服变褶皱,周边会出现彩色噪点。 lora 权重太小,脸型修饰不够,使用的时候酌情设置权重值,正在重新训练中......

It is init version, mainly for asian male, and did not handle the clothing data well.

When the lora weight is too large, it will cause the clothes to become wrinkled, and there will be colored noise around them. When the lora weight is too small, the face shape is not modified enough. So when using it, set the weight value as appropriate. It is being retrained...