How to alternate the training process

[Arsey]

Hi, and thank you for a great job!
Currently, I'm trying to port DeLiGAN into DIGITS. It has a bit different way of training. The question is where or how can I alternate the training process in DIGITS + TF?
Here's how they train GAN - https://github.com/val-iisc/deligan/blob/master/src/mnist/dg_mnist.py#L180