Welcome to RLForge’s documentation!
RLForge is a lightweight yet powerful reinforcement learning framework designed to make RL accessible to everyone, from students exploring bandits to researchers experimenting with advanced deep RL algorithms.
What is RLForge?
RLForge provides a unified interface for building, training, and evaluating reinforcement learning agents across a wide variety of environments. It is designed to be:
Educational: clear implementations of classic algorithms like multi-armed bandits, SARSA, and Q-learning.
Scalable: support for advanced deep RL methods such as DQN, PPO, TD3, SAC, and DDPG.
Compatible: works seamlessly with Gymnasium environments, plus custom environments included in RLForge (mazes, bandits, short corridor, robotics-inspired tasks, etc.).
Visual: built-in experiment runner and plotting utilities for analyzing learning curves and trajectories.
Algorithms
RLForge includes a wide spectrum of RL agents:
Basic algorithms: Bandits, Tabular SARSA, Q-learning, Expected SARSA.
Function approximation: Linear regression, MLP-based agents.
Deep RL (PyTorch-based): - DQNTorchAgent - DDPGAgent - TD3Agent - SACAgent - PPODiscrete - PPOContinuous
PyTorch Agents and Vectorized Environments
Agents implemented with PyTorch not only leverage neural networks for function approximation, but also support vectorized environments. This allows training across multiple parallel environments, dramatically improving sample efficiency and stability.
Getting Started
Install RLForge with:
pip install rlforgeFor PyTorch-based agents, install with the optional dependency:
pip install rlforge[torch]
Explore the examples section to see RLForge in action, from simple bandit problems to advanced continuous control tasks.
Getting Started
Experiment Runner
Policies
Utils