lychenyoko

> @beniz Maybe you can calculate g_path_regularize inside the generator forward, and only return the loss. Thanks a lot for your effort! I actually encountered exactly the same issue when...

@rosinality Never mind. I have managed to do that and thanks for your advice.

Yes, it can.

First, it's a back-propagate-able loss. Second, we use it to evaluate channels' sensitivity for the content region noises. Please refer to the paper Sec. 3.1.2 for more details.