Why don't we put models in .train() or .eval() modes in DCGAN example

[zhangboknight (Bo Zhang)]

I am confused about this too. Did you solve the problem?