skrl: A Modular and Flexible Library for Reinforcement Learning
Authors: Antonio Serrano-Muñoz, Dimitrios Chrysostomou, Simon Bøgh, Nestor Arana-Arexolaleiba; 24(254):1−9, 2023.
Abstract
skrl is a Python-based open-source library for reinforcement learning. It is designed to prioritize readability, simplicity, and transparency in implementing algorithms. The library supports traditional interfaces from OpenAI Gym/Farama Gymnasium, DeepMind, and others. Additionally, it allows users to load, configure, and operate NVIDIA Isaac Gym, Isaac Orbit, and Omniverse Isaac Gym environments. Moreover, it enables simultaneous training of multiple agents with customizable scopes, which can have shared or separate resources, in a single run. The library’s documentation can be accessed at https://skrl.readthedocs.io, and its source code is available on GitHub at https://github.com/Toni-SM/skrl.
[abs]