Implementations of popular Deep Reinforcement Learning algorithms
rlib is a small deep reinforcement learning library with implementations of popular deep RL algorithms. Each algorithm is highly modular and customizable, making this library a great choice for anyone who wants to test the performance of different algorithms in the same environment.
rlib uses PyTorch as the library of choice for its initial version, but support for TensorFlow is on the roadmap.
pip install rlib
rlib is this simple:
from rlib.algorithms.dqn import DQNAgent from rlib.environments.gym import GymEnvironment e = gym.make('CartPole-v0') observation_size = 4 action_size = 2 dqn = DQNAgent(observation_size, action_size) env = GymEnvironment(e, dqn) env.train() env.test()
TensorBoard and GIFRecorder
os.makedirs('your/log/dir', exist_ok=True) logger = Logger(output_dir) gifs_recorder = GIFRecorder(output_dir, duration=3.0)
- Initialize a new environment using these objects.
env = GymEnvironment(e, dqn, logger=logger, gifs_recorder=gifs_recorder)
- To check Tensorboard logs, run:
- Define your own custom model.
class NeuralNet(torch.nn.Module): def __init__(self): super(NeuralNet, self).__init__() self.fc1 = nn.Linear(4, 8) self.relu = nn.ReLU() self.fc2 = nn.Linear(8, 2) def forward(self, x): out = self.fc1(x) out = self.relu(out) out = self.fc2(out) return out
- Check the documentation for the algorithm you are using for the appropriate argument name. For DQN:
dqn = DQNAgent( observation_size, action_size, qnetwork_local=NeuralNet(), qnetwork_target=NeuralNet(), )
Saving model weights
- Set the
model_output_dirargument when creating a new instance of an algorithm to the directory where you want your model to be saved.
To run all tests:
python -m unittest discover test/
Feel free to open issues with any bugs found or any feature requests. Pull requests are always welcome for new functionality.
virtualenv -p python3 venv cd rlib/ source venv/bin/activate pip install -r requirements.txt
To make sure your installation worked, run one of the algorithms in the test folder:
rlib is released under the MIT License.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.