Stable-Baselines-Team/stable-baselines3-contrib

Contrib package for Stable-Baselines3 - Experimental reinforcement learning (RL) code

What it solves

It provides a dedicated space for experimental reinforcement learning (RL) algorithms and tools that are too niche or less mature to be included in the main Stable-Baselines3 library, while still maintaining high standards for documentation and style.

How it works

It acts as a contribution package for Stable-Baselines3, implementing a variety of RL algorithms from recent publications and providing specialized environment wrappers. It allows the community to contribute useful utilities and algorithms without requiring the strict integration constraints of the core library.

Who it’s for

Researchers and developers working with reinforcement learning who need access to a wider range of experimental algorithms beyond the standard set provided by Stable-Baselines3.

Highlights

  • Experimental RL Algorithms: Includes implementations of ARS, QR-DQN, MaskablePPO, RecurrentPPO (PPO LSTM), TQC, TRPO, and CrossQ.
  • Gym Wrappers: Provides specialized tools like the Time Feature Wrapper.
  • Consistent Standards: Maintains the same style and documentation quality as the main Stable-Baselines3 project.

Related

  • Project
    DLR-RM/stable-baselines3

    A set of reliable PyTorch implementations of reinforcement learning algorithms designed to provide stable baselines for research and industry.

  • Project
    DLR-RM/rl-baselines3-zoo

    A training framework for Stable Baselines3 that provides scripts for training, evaluating, and tuning RL agents, along with a collection of pre-tuned hyperparameters for common environments.

  • Project
    araffin/sbx

    A Jax-based proof-of-concept implementation of Stable-Baselines3, providing high-performance reinforcement learning algorithms with a consistent API.

  • Project
    hiyouga/EasyR1

    An efficient, scalable reinforcement learning training framework that supports multi-modality models and various RL algorithms like GRPO.

  • Project
    nvidia-cosmos/cosmos-rl

    A scalable Reinforcement Learning framework specialized for Physical AI applications, providing a decoupled architecture for large-scale training workloads.