In the code the f_loss_src is trained with the f_vars
self.f_train_op_src = slim.learning.create_train_op(self.f_loss_src, self.f_optimizer_src, variables_to_train=f_vars)
but in paper the equation 5 shows that the f_loss_src should trained with g_vars
@Xztty I remember that i tried to optimize f_train_op_src with the both of g_varsand f_vars but the training process was unstable. Depending on my intuition, I modified the code to resolve the problem.
In the code the f_loss_src is trained with the f_vars
self.f_train_op_src = slim.learning.create_train_op(self.f_loss_src, self.f_optimizer_src, variables_to_train=f_vars)
but in paper the equation 5 shows that the f_loss_src should trained with g_vars
Did I misunderstand?