Farama-Foundation/SuperSuit

What it solves

SuperSuit provides a set of small, reusable functions called "microwrappers" to preprocess reinforcement learning (RL) environments. It simplifies the process of modifying environment observations and behaviors without needing to write complex custom wrapper classes.

How it works

The library provides functions that can be wrapped around existing environments from Gymnasium (single-agent) and PettingZoo (multi-agent). By chaining these functions, users can apply transformations such as converting images to grayscale (color reduction) or stacking multiple consecutive frames to provide temporal context to the agent.

Who it’s for

Researchers and developers working with reinforcement learning who need a standard way to preprocess environment data for both single-agent and multi-agent settings.

Highlights

• Multi-Framework Support: Works with Gymnasium and PettingZoo (both AECEnv and ParallelEnv).
• Microwrappers: Uses a functional approach to wrap environments for quick preprocessing.
• Preprocessing Tools: Includes utilities like frame stacking and color reduction.