Inverse Reinforcement Learning Pytorch