DQN with policy and target networks doesn't learn properly on LunarLander enviroment

Question

I\'m trying to get a hang of reinforcement learning, so I\'m following a guide at: pytorch.org/tutorials/

They\'ve implemented DQN that solves CartPole with computer