Shypd.aiThe largest open index of AI toolsOpenRL is an open-source unified reinforcement learning framework that supports training for various tasks, including single-agent, multi-agent, offline RL, and natural language. It provides a simple-to-use, flexible, and efficient platform for RL research.
Trending
About
OpenRL is an open-source general reinforcement learning research framework developed by OpenRL-Lab, based on PyTorch. It aims to provide a simple-to-use, flexible, efficient, and sustainable platform for the reinforcement learning research community. The framework supports a wide array of tasks, including single-agent, multi-agent, offline RL with expert datasets, self-play, and natural language tasks such as dialogue. Key features include a universal interface for various tasks and environments, support for DeepSpeed, and integration with Hugging Face for models and datasets. OpenRL also offers convenient evaluation through Arena, supports popular visualization tools like wandb and tensorboardX, and provides multiple training acceleration methods. It includes support for various environments like Gymnasium, MuJoCo, PettingZoo, and Atari, and implements algorithms such as PPO, MAPPO, GAIL, and SAC.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
