RLlib

E95190

machine learning framework open-source software reinforcement learning library

RLlib is a scalable, open-source reinforcement learning library built on Ray that provides high-level APIs and distributed training support for a wide range of RL algorithms.

Statements (49)

Predicate	Object
instanceOf	machine learning framework → open-source software → reinforcement learning library →
designedFor	production workloads → research workloads → scalability →
developedOnTopOf	Ray →
hostedOn	GitHub →
integratesWith	Ray Serve → Ray Tune →
license	Apache License 2.0 NERFINISHED →
partOf	Ray ecosystem →
provides	algorithm configuration system → built-in RL algorithms → checkpointing utilities → custom model support → custom policy support → evaluation utilities → high-level APIs → hyperparameter tuning integration → logging utilities → low-level APIs →
supports	CPU training → GPU training → distributed reinforcement learning → multi-GPU training → multi-node training → scalable training →
supportsAlgorithmFamily	Q-learning methods → actor-critic methods → evolution strategies → multi-agent reinforcement learning → policy gradient methods →
supportsEnvironmentInterface	Gymnasium → OpenAI Gym NERFINISHED → PettingZoo →
supportsFeature	centralized training with decentralized execution → distributed rollout workers → fault-tolerant training → parameter server architectures →
supportsFramework	PyTorch NERFINISHED → TensorFlow NERFINISHED →
supportsUseCase	hierarchical reinforcement learning → model-based reinforcement learning → multi-agent reinforcement learning → offline reinforcement learning → self-play → single-agent reinforcement learning →
writtenIn	Python →

Referenced by (2)

Subject (surface form when different)	Predicate
PettingZoo →	compatibleWith
OpenAI Gym →	influenced