About the loss

[scutzzy]

When I trying to implement the work, I found the mask value in generator goes to 1 rapidly, which leads to the output of generator remaining the same as input, then I read the paper of Ganimation, I found that they added another restrict on mask value, did you used that too? After add l1 weight regularization on mask, its value goes to 0.01 but the output of generator is still remain the same structure of input. There must be some faults in my code ,It will be really helpful If you can help me check it out, and looking forward to your training procedure:)

[ESanchezLozano]

Hi,
Sorry for the late reply. Could you please try to check how does it work with the models as in this repository? I am not sure how this extra mask affects the performance. I will try to have a look at your code in any case.

Please do not hesitate to mention me so that I can catch up with the conversation.