DLR-RM/rl-baselines3-zoo

A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

What it solves

RL Baselines3 Zoo simplifies the process of training, evaluating, and benchmarking reinforcement learning (RL) agents. It removes the guesswork from starting a project by providing a standardized framework and a large collection of pre-tuned hyperparameters for a wide variety of common environments.

How it works

Built on top of Stable Baselines3, the framework provides a command-line interface (CLI) to execute RL workflows. Users can train agents by specifying an algorithm and an environment, evaluate performance at set intervals, and use provided scripts to plot results or record videos of the agents in action. It also includes a library of YAML files containing optimized hyperparameters and a repository of pre-trained agents.

Who it’s for

It is designed for RL practitioners and researchers who want to quickly benchmark different algorithms, train agents without manually tuning every hyperparameter, or access high-performing pre-trained models for standard environments.

Highlights

  • Extensive Library: Includes tuned hyperparameters and trained agents for over 200 environments across Atari, Classic Control, Box2D, PyBullet, MuJoCo, and Robotics.
  • Integrated Tooling: Provides built-in scripts for training, evaluation, hyperparameter tuning, and result visualization.
  • Broad Compatibility: Supports a wide range of RL algorithms including PPO, DQN, SAC, TD3, and TQC.
  • External Integrations: Connects with Weights & Biases for experiment tracking and Hugging Face for model sharing.

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
    Stable-Baselines-Team/stable-baselines3-contrib

    A contribution package for Stable-Baselines3 containing experimental reinforcement learning algorithms and niche tools for researchers and developers.

  • Project
    google-deepmind/bsuite

    A collection of standardized experiments and benchmarks for evaluating the core capabilities of reinforcement learning agents to facilitate reproducible research.

  • Project
    rllm-org/rllm

    An open-source framework for training language agents with reinforcement learning, allowing the same agent code to be used across different backends, sandboxes, and benchmarks.

  • Project
    OpenManus/OpenManus-RL

    An open-source framework for optimizing LLM agents using reinforcement learning to enhance their reasoning, tool use, and decision-making capabilities.