Skip to content

Latest commit

 

History

History
7 lines (5 loc) · 494 Bytes

README.md

File metadata and controls

7 lines (5 loc) · 494 Bytes

RL_pytorch

An implementation of DQN algorithm via pytorch to use multiple gpu in a single node. I use DataParallel API for this purpose. The REINFORCE/Actor-Critic algorithm works only with CartPole.

The general_dqn is a general DQN algorithm for any non-image env, and it is possible to define fully connected layer of any size. The reinforce_AC_CartPole_torch notebook, includes a naive implementation of REINFORCE, REINFORCE with Reward to Go, AC, and AAC algorithms.