Blog

Autoreset control when vectorised environment reset sub-environments on terminations or truncations. Gymnasium offers three options, for which, we present descriptions and examples for each.

After years of hard work, Gymnasium v1.0 has officially arrived! This release marks a major milestone for the Gymnasium project, refining the core API, addressing bugs, and enhancing features. Over 200 pull requests have been merged since version 0.29.1, culminating in Gymnasium v1.0, a stable release focused on improving the API (`Env`, `Space`, and `VectorEnv`). This post summarizes these changes.

We are excited to introduce the version 5 of the MuJoCo Environments, bringing a wide variety of improvements, including the ability to load custom robot models.

In this post, we explain the motivation for the terminated / truncated step API, why alternative implementations were not selected, and the relation to RL theory.

Minari provides the capability to create your own environment-based datasets, to download open-source datasets and to upload your own datasets for others to use.

We are excited to announce Shimmy 1.0, an API compatibility tool for converting a wide range of single and multi-agent environments to the Gymnasium and PettingZoo APIs

Today we're announcing the Farama Foundation – a new nonprofit organization designed in part to house major existing open source reinforcement learning (“RL”) libraries in a neutral nonprofit body.