PyTorch DQN Performance

[FeynmanDNA]

Hi, I saw your pull request to gym-lattice 2d repo (https://github.com/ljvmiranda921/gym-lattice) and landed here.

I have played around with gym-lattice using q-learning before and I tried some basic DQN on it as well. It looks like you have tried PyTorch's DQN code on this task, and I am wondering how did you find the performance of Deep RL on this task?

The Wu.py is from the Suzhou conference paper right? They used basic q-table without NN. Is q-table sufficient for this 2D HP task?

Thank you very much for your time.

P.S. Awesome pull request to gym-lattice and nice project code here!