unRL (AKA "unreal") is a set of libraries providing Reinforcement Learning algorithms implemented with PyTorch and Jax frameworks.
The scope of RL algorithms considered is not yet defined, but meant to be broad. So, whether the prefix stands for "universal", (deep) "unsupervised", or simply "(f)un" is not important.
The first PyPi release will occur for version 0.1. For those looking for a polished deep RL library using Jax specifically, consider deepmind's Haiku instead.
NOTE: The unrl package is undergoing refactoring from pure Python implementations to being powered by PyTorch (and NumPy).
The target Python language version meant for backwards compatibility is 3.10.