This is a library whose goal is to implement various Reinforcement Learning algorithms in a modular and extensible way. The library is designed to be used as a workbench for experimenting with different algorithms and environments. The library is implemented in Python and uses PyTorch for neural network implementations.
To install the library, clone the repository and install the dependencies using pip:
pip install -r requirements.txtIf you get an error related to swig during the installation of the dependencies, you
may need to install swig using your package manager.
The following algorithms are implemented in the library (the algorithms have their respective README files with more detailed information in their directories):
- Q-Learning: A simple Q-Learning implementation with epsilon-greedy exploration.
- Deep Q-Learning (DQN): A PyTorch implementation of Deep Q-Learning with experience replay and epsilon-greedy exploration. Furthermore DQN supports experiment tracking with Weight & Biases.