This repository collects Ignite engines for deep reinforcment learning algorithms. Implementations run on GPU through Pytorch. This package does not aim to provide the most efficient implementations (although efficiency is a concern), but rather to showcase clean implementations and code reusability using Ignite.
Generally speaking, the user is expected to provide the model to be learn, as well as an OpenAi Gym compatible environement.
The project requires Python 3.6+.
To install, use pipenv, the recomended dependency manager for Python (also works with pip):
$ pipenv install git+https://github.com/ds4dm/ignite-rl.git#egg=irl
(not yet on PyPI)