search

aitorzip / PyTorch-CycleGAN

A clean and readable Pytorch implementation of CycleGAN
https://arxiv.org/abs/1703.10593
GNU General Public License v3.0
1.22k stars 283 forks source link

Adversarial Loss not trained against stagnant partner? #12

Open yhu9 opened 5 years ago

yhu9 commented 5 years ago

I was under the impression that the optimizer and the generator should be training against stagnant counterparts but the step function on the generators are called before the gradients are calculated for the discriminator. Is this standard practice?

FloCF commented 5 years ago

For e.g. DCGAN or pix2pix people usually start with discriminator training with detached generated fakes and then optimize Generator. Not sure why it is the other way round here. Dont know if that helps :sweat_smile: ...