arXiv 2505.101832025-05-15cs.DCcs.AI

KAITIAN：一个用于在具身智能系统中实现跨异构加速器高效协作的统一通信框架

KAITIAN: A Unified Communication Framework for Enabling Efficient Collaboration Across Heterogeneous Accelerators in Embodied AI Systems

Hangzhou Institute for Advanced Study, UCAS（杭州高等研究院，中国科学院大学）；Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）

Jieke Lin, Wanyu Wang, Longxiang Yin, Yinhe Han

备注：: 9 pages, 4 figures. Jieke Lin and Wanyu Wang contributed equally to this work

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出 KAITIAN 统一分布式通信框架，通过抽象层整合厂商库与通用协议，并引入负载自适应调度，在 NVIDIA GPU 与寒武纪 MLU 上实现异构加速器高效协作。

摘要：

具身人工智能（AI）系统，例如自主机器人和智能车辆，越来越依赖多种异构加速器（如 GPGPU、NPU、FPGA）来满足严格的实时处理和能效需求。然而，厂商专属的专有通信库大量涌现，造成了显著的互操作性障碍，阻碍不同类型加速器之间的无缝协作，并导致分布式 AI 工作负载中的资源利用不佳和性能瓶颈。本文介绍 KAITIAN，这是一个旨在弥合该差距的新型分布式通信框架。KAITIAN 提供统一抽象层，将面向组内效率的厂商优化通信库与用于组间互操作性的通用通信协议智能集成。关键在于，它包含负载自适应调度机制，可根据异构设备的实时性能特征动态平衡计算任务。KAITIAN 作为 PyTorch 的扩展实现，并在配备 NVIDIA GPU 和寒武纪 MLU 的测试平台上经过严格评估，在分布式训练任务中展现出资源利用率和可扩展性方面的显著提升。实验结果表明，与基线同构系统相比，KAITIAN 可将训练时间加速最高 42%，同时仅产生极小的通信开销（2.8–4.3%），并保持模型精度。KAITIAN 为复杂具身 AI 应用中更灵活、更强大的异构计算铺平了道路。

英文摘要：

Embodied Artificial Intelligence (AI) systems, such as autonomous robots and intelligent vehicles, are increasingly reliant on diverse heterogeneous accelerators (e.g., GPGPUs, NPUs, FPGAs) to meet stringent real-time processing and energy-efficiency demands. However, the proliferation of vendor-specific proprietary communication libraries creates significant interoperability barriers, hindering seamless collaboration between different accelerator types and leading to suboptimal resource utilization and performance bottlenecks in distributed AI workloads. This paper introduces KAITIAN, a novel distributed communication framework designed to bridge this gap. KAITIAN provides a unified abstraction layer that intelligently integrates vendor-optimized communication libraries for intra-group efficiency with general-purpose communication protocols for inter-group interoperability. Crucially, it incorporates a load-adaptive scheduling mechanism that dynamically balances computational tasks across heterogeneous devices based on their real-time performance characteristics. Implemented as an extension to PyTorch and rigorously evaluated on a testbed featuring NVIDIA GPUs and Cambricon MLUs, KAITIAN demonstrates significant improvements in resource utilization and scalability for distributed training tasks. Experimental results show that KAITIAN can accelerate training time by up to 42% compared to baseline homogeneous systems, while incurring minimal communication overhead (2.8--4.3%) and maintaining model accuracy. KAITIAN paves the way for more flexible and powerful heterogeneous computing in complex embodied AI applications.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。