Hi, thanks for sharing your code! I also train this model. But the model seems like not converge fast as following.

I also use the img_align_celeba as training and val data and all training parameters used are default as your main.py. Could you know the reason why my model not converge?