search

kaonashi-tyc / zi2zi

Learning Chinese Character style with conditional GAN
https://kaonashi-tyc.github.io/2017/04/06/zi2zi.html
Apache License 2.0
2.52k stars 470 forks source link

category_loss is always 0 #84

Open cqray1990 opened 3 years ago

cqray1990 commented 3 years ago

category_loss is always 0, when i use font2img to generate images,the label is default 0, and when training the lable of images is,so the loss to compute real_category_loss and fake_category_loss is always 0, i don't know why label is 0?