SuperSuit Documentation

SuperSuit is a collection of small functions called 'microwrappers' designed to preprocess reinforcement learning environments. These wrappers allow for quick transformations such as:

• Color reduction: Modifying observation color spaces (e.g., converting to grayscale).
• Frame stacking: Stacking multiple consecutive frames to provide temporal context to the agent.

SuperSuit is compatible with:

• Gymnasium: For single-agent environments.
• PettingZoo: For multi-agent environments (AECEnv and ParallelEnv).

Note: This project is currently semi-deprecated and is being maintained primarily to ensure compatibility with new versions of PettingZoo until its functionality is merged into pettingzoo.wrappers.

To install SuperSuit from the source repository, clone the repository, navigate to the directory, and install in editable mode:

python3 -m venv env
source env/bin/activate
pip install --upgrade pip
pip install -e .

To install SuperSuit from PyPI, create a virtual environment and use pip:

python3 -m venv env
source env/bin/activate
pip install --upgrade pip
pip install supersuit

SuperSuit supports multi-agent environments from PettingZoo, including both AECEnv and ParallelEnv types. You can apply the same microwrappers used for Gymnasium to these environments.

Example: Applying color reduction and frame stacking to a PettingZoo Pistonball environment:

from pettingzoo.butterfly import pistonball_v0
from supersuit import color_reduction_v0, frame_stack_v1

env = pistonball_v0.env()

env = frame_stack_v1(color_reduction_v0(env, 'full'), 4)

SuperSuit provides 'microwrappers' to preprocess reinforcement learning environments. For single-agent environments using Gymnasium, you can chain wrappers to modify observation spaces or frame stacking.

Example: Converting Space Invaders to use grayscale observations and stacking the last 4 frames:

import gymnasium
from supersuit import color_reduction_v0, frame_stack_v1

env = gymnasium.make('SpaceInvaders-v0')

env = frame_stack_v1(color_reduction_v0(env, 'full'), 4)