RL4Net is a reinforcement learning platform based on Pytorch and OpenAI Gym.The supported interface algorithms currently include:
-
Sarsa
-
Qlearning
-
Deep Q-Network (DQN)
-
Deep Deterministic Policy Gardient (DDPG)
Notice: The platform uses pytorch and numpy. It is recommended to use conda to create a new environment and install it.
RL4Net is currently hosted on PyPI. It requires Python >= 3.6.
You can simply install RL4Net from PyPI with the following command:
pip install rl4net
You can also install with the newest version through GitHub:
pip install git+https://github.com/bupt-ipcr/RL4Net.git@master
Or install it after downloading it locally:
git clone https://github.com/bupt-ipcr/RL4Net
Enter folder and install it with pip:
cd RL4Net
pip install .
After installation, run examples :
python examples/ddpg.py
If no error occurs, you have successfully installed RL4Net.
Todo
You can create your own reinforcement learning agent through the base class provided in rl4net.agents
.The general method is as follows:
# import base class
from rl4net.agents import xxxBase
# inherit and complete necessary methods
class myxxx(xxxBase):
def _build_net(self):
pass
learn more about the usage, by codes examples and annotated documentations under examples/
.
Simple neural networks built on pytorch are also provided in rl4net.models
. You can take it as a simple implement of DRL neural network.
from rl4net.models import SimpleDQNNet
Third, if you want to call the attached envs, you should run __init__.py
to register to gym
. After that you can use standard gym
methods to create it.
import rl4net.env
env = gym.make('Maze-v0')
VVALB is still under development. More algorithms and features are going to be added and we always welcome contributions to help make VVLAB better. If you would like to contribute, please check out this link.