Gymnasium
A local Python interface for training and testing reinforcement learning agents in simulated environments.
Published by Farama Foundation
Setup and evaluation
Gymnasium gives environments a shared Python interface so an agent can reset a simulation, take actions, and receive observations and rewards. Start with a built-in task such as CartPole and establish a random-action baseline before comparing trained policies.
Install with pip install gymnasium. Use gymnasium.make() with a documented environment ID. reset(seed=...) starts a reproducible episode, and step(action) distinguishes task termination from a time limit. Some environment families need optional packages.
- Framework
- Gymnasium
- Access
- Local Python package. No account required.