After a period of training, the result will collapse

[carrot-ly]

I used celeba dataset and webface dataset, and then used your code and settings for training. After training the first epoch, the results are as follows:
Render image：

Source image：

But at the second epoch, the rendered image becomes extremely poor.

It seems that a large part of the face has been filtered out

[carrot-ly]

I use CUDA 11.1