Version 1.0's training approach was too conservative, sometimes failing to change the model's stereotypes, like high heels for office ladies and loafers for serafuku. Version 2.0 has fixed this.
2.0几乎对原模型风格没有任何影响,并提高了准确性