rl-baselines-zoo  by araffin

RL agent collection using Stable Baselines (note: unmaintained, use RL-Baselines3 Zoo)

created 6 years ago
1,183 stars

Top 33.7% on sourcepulse

GitHubView on GitHub
Project Summary

This repository provides a collection of over 100 pre-trained Reinforcement Learning (RL) agents, complete with tuned hyperparameters and training scripts, built using the Stable Baselines library. It's designed for researchers and practitioners to easily benchmark RL algorithms, enjoy trained agents, and leverage existing hyperparameter configurations across various environments, including Atari, Classic Control, Box2D, PyBullet, and MiniGrid.

How It Works

The project leverages the Stable Baselines library for implementing various RL algorithms. It organizes training and hyperparameter configurations in YAML files, allowing for straightforward execution of training, enjoyment (inference), evaluation, and hyperparameter optimization using Optuna. The architecture supports environment wrappers and command-line argument overrides for flexibility.

Quick Start & Requirements

  • Install: pip install -r requirements.txt (requires stable-baselines[mpi] >= 2.10.0)
  • Prerequisites: swig, cmake, libopenmpi-dev, zlib1g-dev, ffmpeg. GPU support requires CUDA.
  • Docker: CPU: docker pull stablebaselines/rl-baselines-zoo-cpu, GPU: docker pull stablebaselines/rl-baselines-zoo.
  • Docs: Stable Baselines README

Highlighted Details

  • Over 120 pre-trained agents across multiple environment types.
  • Includes hyperparameter tuning capabilities using Optuna.
  • Supports environment wrappers and custom environment registration (e.g., MiniGrid).
  • Provides scripts for training, enjoying agents, benchmarking, and recording videos.

Maintenance & Community

This repository is no longer maintained. Users are directed to the RL-Baselines3 Zoo for an up-to-date version.

Licensing & Compatibility

The project is licensed under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

The project is explicitly marked as unmaintained. Hyperparameter search is not implemented for ACER and DQN. MiniGrid environments require specific wrappers for observation spaces not natively supported by Stable Baselines.

Health Check
Last commit

2 years ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
11 stars in the last 90 days

Explore Similar Projects

Starred by Logan Kilpatrick Logan Kilpatrick(Product Lead on Google AI Studio), Chip Huyen Chip Huyen(Author of AI Engineering, Designing Machine Learning Systems), and
3 more.

catalyst by catalyst-team

0%
3k
PyTorch framework for accelerated deep learning R&D
created 7 years ago
updated 1 month ago
Feedback? Help us improve.