1

请问如何解决？

how to fix this issue with tf version 1.8 ? thks

@Xiaodong-Bran a little difficult，i am noob in python and deeplearning. to my best knowledge， distorted images and non-distorted images were trainded by CycleGAN.

@falrom 我也是发现了这个作者BN使用不当，思考了很久不知道怎么改，有没有想法？

好的，谢谢，作为科研论文，我还是尽量按照BN正确用法吧

@zhangqianhui 您好，您是使用WGAN损失函数Ex∼qdata[D(x)]−Ez∼p(z)[D(G(z))] 和在判别器中使用谱归一化？这样进行有效果吗？ 之前我试过直接在WGAN-GP加入谱归一化，结果生成器和判别器的loss都是nan。

@zhangqianhui @takerum  i apply hinge loss and spectral normalization to my experiment ,which actually outperform the WGAN-GP loss function with the same network structure. a little pity， the discriminator...

@zhangqianhui yeah，it is the D loss using hinge loss and does not convergence to a certain value. why? thx