How to create a Variational Auto-Encoder Policy in Tensorflow in Hopper Environment

Question

I have been researching neural networks, smart agents and reinforcement learning for some time. I tried to understand how GAIL algorithm works so I can work with Directed In