A question about the loss of training GAN process

[sjtupotato]

Hi,
I met a problem when I train the whole gan network. First I load my weights of G network into the framework, then the train is begining and i print all kinds of loss scores, it confused me that the loss of gan_bce just increase even if the avdersarial loss decreases. Does it normal and i need to wait for more epoches?