opendilab/DI-engine

Why this repo matters

• •Comprehensive framework supporting diverse RL paradigms including offline, multi-agent, and model-based learning algorithms.
• •Built-in distributed training infrastructure enables scalable reinforcement learning experiments across complex environments.
• •Extensive benchmark coverage spans Atari, MuJoCo, and StarCraft for standardized algorithm evaluation and reproduction.