EXARL
Easily extendable architecture for reinforcement learning
Area: Artificial intelligence
CASS member: RAPIDS
Description
The Easily eXtendable Architecture for Reinforcement Learning (EXARL) - enables scientists and non-ML researchers to easily apply reinforcement learning in diverse problem areas. It provides a set of tested and easily modifiable RL tools, including agents, environments, and learning workflows, that can run at scale, taking advantage of DOE leadership computing capabilities. This accelerates scientific discovery in areas with complex decision making and/or simulation requirements.
Target audience
Scientists, ML/non-ML researchers who use RL at scale.
License: BSD-3-Clause
Package links
- Spack: py-exarl