Type | |
Stats | 1,023 |
Reviews | (171) |
Published | Jan 13, 2024 |
Base Model | |
Training | Steps: 10,000,000 Epochs: 10 |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 84253A87B0 |
This model has undergone nearly 1 million image training using an A100 server, followed by special training for artists. The total number of training steps reached approximately 10 million. It has undergone negative tag training and supports the recognition of a certain number of artists. However, in terms of character features, it requires the input of the character name and all feature tags.Based on Kohaku-XL.
Recommended tags:
Artist tags1:
ciloranko, [sho_(sho_lwlw)], (((tianliang duohe fangdongye))) , ningen_mame,
Artists tags2:
tianliang duohe fangdongye
You must enter artist tags to achieve good results
Negative tags:
quality bad, quality normal,hands bad, eyes bad, face bad,lowres,unfinished,bad, worse, worst, ugly, bad anatomy, blurry, close-up,unaestheticXLv13,negativeXL_D,
Recommended embeddings: unaestheticXLv13, negativeXL_D,
该模型使用A100服务器进行了近1百万张图像训练,随后对艺术家进行了专门训练。训练步骤总数达到约1000万。它经过了负标签训练,支持对一定数量的艺术家的识别。然而,在角色特征方面,它需要输入角色名称及其来源还有所有特征标签。基于Kohaku-XL。
必须要输入画师tag,不然没有较好的效果