rl-baselines3-zoo
A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.
What is rl-baselines3-zoo?
A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.
Overview of rl-baselines3-zoo
What it is
RL Baselines3 Zoo is a training framework for reinforcement learning (RL) using Stable Baselines3. It provides scripts for training, evaluating agents, tuning hyperparameters, plotting results, and recording videos. The repository includes a collection of tuned hyperparameters for common environments and RL algorithms, as well as pre-trained agents.
Key features
- Scripts for training, evaluation, hyperparameter tuning, plotting, and video recording - Collection of tuned hyperparameters for various environments and RL algorithms - Pre-trained agents for over 200 environments (Atari, Classic Control, Box2D, PyBullet, etc.)
- Integration with Weights & Biases for experiment tracking and Hugging Face for model sharing - Command-line interface via
rl_zoo3or direct scripts (train.py,enjoy.py)
Use cases
- Benchmarking different RL algorithms on standard environments - Quickly training and evaluating RL agents with pre-configured hyperparameters - Enjoying pre-trained agents in action - Experiment tracking and hyperparameter optimization
Getting started
Minimal installation from source: pip install -e . Or as a Python package: pip install rl_zoo3 Full installation with extra dependencies: apt-get install swig cmake ffmpeg pip install -e .[plots,tests,extras] Train an agent: python train.py --algo algo_name --env env_id Enjoy a pre-trained agent (requires cloning with submodules): python enjoy.py --algo a2c --env BreakoutNoFrameskip-v4 --folder rl-trained-agents/ -n 5000
License and pricing
The project is open source under the MIT license.
Based on the project's public README and repository metadata.