GAN training

[ShuchunXu]

I run "python3 baseline.py basic", use GAN to attack, and use the client's model as the discriminator for model training. I found that when training for 500 rounds, the loss of the generator is particularly large, about 2.03, while the loss of the discriminator Only about 0.2, why is this? ask for advice. My e-mail: [e-mail]

[ShuchunXu]