Libraries & frameworks
verl-omni
verl-project/verl-omni
A reinforcement learning training framework for diffusion and omni-modal generative models built on VeRL.
Topics
Related projects
relax669Large-scale asynchronous reinforcement learning engine for omni-modal model post-training, decoupling training from inference while supporting elastic scaling across distributed resources.
OSWorld3.3kOSWorld is a benchmark and real desktop environment for evaluating multimodal computer-use agents, with support for VMware, Docker, AWS, and other backends for reproducible task evaluation.
ml-agents19.9kUnity's open toolkit for training intelligent game agents with deep reinforcement and imitation learning through Python and Unity APIs.
ms-swift16.1kA full-pipeline framework for fine-tuning, aligning, evaluating, and deploying hundreds of LLMs and multimodal models.
openvino11.1kAn open-source toolkit for optimizing and deploying deep learning inference across CPUs, GPUs, and AI accelerators.
cleanrl10.7kHigh-quality single-file deep RL library covering PPO, DQN and SAC, focused on readability and reproducible experiment management.