How to reproduce your results
tyshiwo opened this issue · 1 comments
tyshiwo commented
Hi,
I'm trying your code recently. The original loss_EG is written as:
self.loss_EG = self.EG_loss + 0.000 * self.G_img_loss + 0.000 * self.E_z_loss + 0.000 * self.tv_loss
which is an auto-encoder, and produces smooth images that loss identity information.
I tried to change the loss_EG as:
self.loss_EG = self.EG_loss + 0.001 * self.G_img_loss + 0.001 * self.E_z_loss + 0.001 * self.tv_loss
but the results are still not good as yours.
So, could you please tell me how do you set the weights for different losses? Thanks!