v2.0
Changed to use Prodigy scheduler with 30 epochs, dim 32 alpha 16, stopped at 22th epoch
Personally I think the flexibility of the model is enhanced, and the generation quality of orange hair outfit is better than the white hair one, due to the selection of dataset.
The taggings remain the same as v1.0
----------------------------------------------------------------------
Trained with NoobAi Early access version checkpoint with 70 epochs and batch size=2(stopped at 39th epoch).
The male protagonist of maimai dragon area, only one outfit available currently in official setting. Just for interested and try to train in leisure time, turns out the result is not bad
Feel free to give any suggestions, willing to learn more to produce higher quality Lora models.
Taggings:
white hair based with black shirt:
ask\(maimai\), 1boy, short hair, black shirt, long sleeves, black gloves, white jacket, white hair, multicolored hair, orange hair, ????????????, black gloves, black pants, fingerless gloves, streaked hair, high collar, thigh strap, goggles, orange ribbon, white boots, sleeves rolled up, sleeves past elbows, two-sided jacket,
orange hair based with black vest and white shirt:
ask\(maimai\), 1boy, short hair, black vest, white shirt, long sleeves, black gloves, orange eyes, white jacket, two-tone hair, red hair, orange hair, ????????????, black gloves, black pants, fingerless gloves, streaked hair, high collar, thigh strap, goggles, orange ribbon, white boots, sleeves rolled up, sleeves past elbows, two-sided jacket,
1. 재게시된 모델의 권리는 원 제작자에게 있습니다.
2. 모델 원작자가 모델을 인증받으려면 공식 채널을 통해 SeaArt.AI 직원에게 문의하세요. 저희는 모든 창작자의 권리를 보호하기 위해 노력합니다. 인증하러 이동
