stable-baselines3

Why we included this project

Most RL teams pick Stable Baselines3 when they want tested, consistent PyTorch implementations of the standard algorithms without writing training loops from scratch. PPO, DQN, SAC, TD3, and A2C all share one API, so swapping methods and comparing results is straightforward. Every algorithm is benchmarked against reference codebases and covered by unit tests, and the library handles custom policies, dictionary observation spaces, callbacks, and TensorBoard logging out of the box. That makes it a practical starting point for research experiments and quick prototypes in robotics, control, or game-like decision problems. Just note that it assumes you already know the basics of reinforcement learning, so it is not a beginner's introduction to the field.