High-performance distributed data shuffling (all-to-all) library for MoE training and inference

A lightweight LLM evaluation toolkit for RLinf. Support mathematical reasoning and long CoT tasks.

RLinf: Reinforcement Learning Infrastructure for Embodied and Agentic AI

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

OneDiff: An out-of-the-box acceleration library for diffusion models.

LiBai(李白): A Toolbox for Large-Scale Distributed Parallel Training