Documentation#
FlagOS
A unified, open-source system software stack designed for a variety of AI chips
FlagOS Overview What’s New in FlagOS 2.2 Cloud Chip Adaptation Guide Edge Chip Adaptation Guide
FlagOS Core Libraries#
High-performance operator libraries optimized for diverse hardware backends.
FlagTree
An open-source, unified compiler for multiple AI chips, advancing and expanding the Triton ecosystem across diverse hardware platforms.
FlagScale
A comprehensive toolkit designed to support the entire lifecycle of large models, from training to inference and deployment.
FlagCX
A scalable and adaptive unified communication library for cross-chip environments, delivering high-performance collective communication capabilities.
FlagOS Plugins for Diverse Chips#
A plugin for the vLLM inference/serving framework, built on FlagOS’s unified multi-chip backend.
A fork of Megatron-LM that introduces a plugin-based architecture for supporting diverse AI chips, built on top of FlagOS.
A fork of TransformerEngine that introduces a plugin-based architecture for supporting diverse AI chips, built on top of FlagOS.
A fork of veRL that extends the upstream library with multi-chip/multi-hardware support via the FlagOS ecosystem.
A custom PyTorch device plugin that provides a unified flagos device across diverse AI chips.
An out-of-tree (OOT) plugin for SGLang, built on FlagOS’s unified multi-chip backend, extending SGLang’s inference capabilities across diverse hardware platforms.
Multi-chip hardware platform and engine plugins for upstream verl RL post-training workloads across NVIDIA, MetaX, Iluvatar, Cambricon MLU, Enflame, and Intel XPU hardware.
FlagOS Domain-Specific Projects#
An integrated training and inference framework for AI models used in robots, so-called Embodied Intelligence.
A high-performance distributed quantum statevector simulator built on PyTorch, enabling quantum circuit simulation across multiple GPUs.
FlagOS Developer Tools#
An operator auto-generation tool.
A benchmark framework for evaluating LLM and agent-based Triton kernel generation across multiple hardware platforms.
A checkpoint conversion and quantization tool for local Hugging Face model directories.
Compatible with Claude Code, Cursor, Codex, and any agent supporting the Agent Skills standard.
An online laboratory providing cloud-based development environments.
A multi-backend debugging and performance-analysis toolkit for Triton programs.
FlagOS Platform Services#
An automated platform for the cross-chip migration and release of open-source large models.
An integrated AI hardware evaluation engine.
A CI/CD toolchain that streamlines large-model development across diverse AI chips.