online vs offline reinforcement learning

• Online: train using the world or a simulator to immediately get you the results of actions
• Offline: train using stored trajectories
• see also: huggingface