This repository contains the code developped and used by Aymeric CÔME, to conduct experiments in RL.
The project is managed using Poetry, and the dependencies are specified in :file:`pyproject.toml`.
The development has been made in Python 3.10, and relies on the Gym library.
When possible, rlberry is also used. This library relies on Gym as well, and provides in particular the Chain
and GridWorld
built-in environments.
Various environments are implemented in the env
submodule. Some specific instances of problems are also defined in envs.instances
.
This module uses Pytest as a testing framework. The unitary tests are located in :file:`tests/`, and can be run with the following command from root:
pytest
Numerous pieces of codes in this repository are largely inspired by https://gitlab.inria.fr/omaillar/average-reward-reinforcement-learning/, developped by Odalric MAILLARD.