stable-baselines3

You can use production-ready reinforcement learning algorithms like PPO, SAC, or DQN to quickly prototype and train agents. These implementations provide a clean API, enabling rapid development without building core RL methods from scratch.

Available reinforcement learning algorithms include PPO, A2C, SAC, TD3, DDPG, DQN, HER, and RecurrentPPO. These cover both continuous and discrete action spaces, suitable for tasks ranging from game-like environments to robotics simulations.

Yes, Stable Baselines3 integrates with Gymnasium to provide vectorized environments for streamlined reinforcement learning development. This setup allows efficient training workflows and evaluation routines across single-agent RL tasks.

Yes, reinforcement learning algorithms support robotics simulations and custom environments. The Skill provides guidance for training workflows, custom environment setup, callbacks, and debugging templates to facilitate reliable experimentation.

Yes, PyTorch is required to train reinforcement learning agents using these implementations. The Skill requires Python, Stable Baselines3, Gymnasium, PyTorch, and optional extras to execute its training templates and evaluation routines.

The best way to evaluate trained reinforcement learning agents is by using the included evaluation routines. These templates streamline the evaluation process, ensuring reliable baselines for production-like experimentation and rapid prototyping.