Have you ever wondered why gym's "done" was replaced with "terminated" and "truncated"? Check out our new blog post, offering a deep dive into the motivation behind this change, alternatives considered, and its relation to RL theory: https://farama.org/Gymnasium-Terminated-Truncated-Step-API
Farama Foundation
@faramafound
- 17 Followers
- 0 Following
- 22 Posts
The Farama Foundation is a nonprofit organization working to develop and maintain open source reinforcement learning tools.
PettingZoo 1.24.0 is now live! This release includes Python 3.11 support, updated Chess and Hanabi environment versions, and many bugfixes, documentation updates and testing expansions. We are also very excited to announce 3 tutorials using Stable-Baselines3, and a full training script using CleanRL with TensorBoard and WandB.
https://github.com/Farama-Foundation/PettingZoo/releases/tag/124.0
New features from original MiniWoB++:
* Python 3 + Gymnasium API + Documentation
* More actions: drag, scroll, key combos + customization
* Fixed bugs in many environments + seed-determinism
* Added 18 hidden “test set” environments
Documentation website: https://miniwob.farama.org/
We are releasing MiniWoB++ v1.0, an RL benchmark with over 100 web interaction environments from button clicks to complex web apps, with multimodal states and diverse actions. RL agents can interact with the browser via Selenium through the Gymnasium API.
https://github.com/Farama-Foundation/miniwob-plusplus/releases/tag/v1.0
Release MiniWoB++ v1.0: Gymnasium Integration and Expanded Actions · Farama-Foundation/miniwob-plusplus
v1.0 Release notes In this release, we fully integrated MiniWoB++ with Gymnasium, which provides a standardized API for reinforcement learning. All MiniWoB++ environments can be easily instantiated...
Gymnasium v0.29.0 is out! This release includes a number of bug fixes and new features. Importantly, it deprecates several features that will be removed in v1.0. For the full release notes, see https://github.com/Farama-Foundation/Gymnasium/releases/tag/v0.29.0
For more information, see our GitHub repository, or the ViZDoom documentation website: https://vizdoom.cs.put.edu.pl/
ViZDoom provides RL environments based on the classic FPS game Doom. This allows developing AI bots that play Doom using only visual information (from the screen buffer)
Key features:
- Multi-platform (Linux, macOS, Windows)
- API for Python, C++ and partially Lua, Java and Julia
Key features:
- Multi-platform (Linux, macOS, Windows)
- API for Python, C++ and partially Lua, Java and Julia
We are thrilled to announce the release of ViZDoom 1.2.0. This release modernizes the library and adds official Gymnasium support, ARM compatibility, and support for serialization using Pickle.
https://github.com/Farama-Foundation/ViZDoom/releases/tag/1.2.0
https://github.com/Farama-Foundation/ViZDoom/releases/tag/1.2.0
Release ViZDoom 1.2.0: add Gymnasium wrapper, Pickle support, and ARM compatibility · Farama-Foundation/ViZDoom
ViZDoom 1.2.0 Release Notes This release adds official Gymnasium support, support for serialization using Pickle, and ARM compatibility. Besides that, it improves and modernizes the codebase to sim...
For more information, see the full release notes: https://github.com/Farama-Foundation/Shimmy/releases/tag/v1.0.0
Melting Pot: a suite of test scenarios for multi-agent reinforcement learning, assessing generalization to novel social situations.
https://github.com/deepmind/meltingpot
Shimmy documentation: https://shimmy.farama.org/environments/meltingpot/
