pytorch-rl

DDPG

	Un-Trained	Trained
DDPG
DQN
CEM

pytorch-rl is a python package based on pytorch library which aims to build standard deep reinforcement learning agent for the user to build various algorithms around it. pytorch-rl also includes some of the state-of-the-art implementations of popular deep reinforcement learning.

Installation

pytorch-rl can be easily installed using pip as follows:

pip install pytorch-rl

This code currently uses python 2.7 and is compatible with both CPU and GPU (cuda 8.0).

Usage

All the algorithmic implementations of a RL agent lies in the rl repository. While the RL agent (say DDPG) can be imported as follows,

from rl.agents import DDPGAgent

The standard DNN model associated with particular agent can be imported as follows

from rl.models.ddpg import ActorNetwork, CriticNetwork

Similarly, to run one of our RL agents( say DDPG) on a standard environment, simply run the following command

python examples/ddpg_pendulum.py

python examples/ddpg_cartpole.py

Agents implemeted currently

• [x] DDPG
• [x] DDQN
• [x] Duel DQN
• [x] CEM
• [x] PPO(needs to be tested)
• [x] ACKTR(needs to be tested)
• [x] A2C(needs to be tested)