heykeetae / Self-Attention-GAN

Pytorch implementation of Self-Attention Generative Adversarial Networks (SAGAN)
2.52k stars 476 forks source link

Which Loss for CelebA Results? #22

Closed tegillis closed 6 years ago

tegillis commented 6 years ago

Do you use hinge loss or Wasserstein-GP loss to get the CelebA images?