Deep reinforcement leanring algorithms implemented in PyTorch.
Refactored ShangtongZhang's DeepRL to ease the running and do some optimization. All rights reserved to him.
My contributions are:
- New implementations of PPO and VPG according to spinningup.
- Optimize the replay buffer of DQN to reduce the memory footprint (from 16GB memory requiremnt to 1.8GB).