A2TP: Aggregator-aware In-network Aggregation for Multi-tenant Learning

Author:

Li Zhaoyi1ORCID,Huang Jiawei1ORCID,Li Yijun1ORCID,Xu Aikun1ORCID,Zhou Shengwen1ORCID,Liu Jingling1ORCID,Wang Jianxin1ORCID

Affiliation:

1. School of Computer Science and Engineering, Central South University, Changsha, Hunan, China

Funder

National Natural Science Foundation of China

Key Research and Development Program of Hunan

Natural Science Foundation of Hunan Province, China

Publisher

ACM

Reference37 articles.

1. William Fedus, Barret Zoph, and Noam Shazeer. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. arXiv preprint arXiv:2101.03961, 2021.

2. Alexander Sergeev and Mike Del Balso. Horovod: fast and easy distributed deep learning in tensorflow. arXiv preprint arXiv:1802.05799, 2018.

3. Yimin Jiang, Yibo Zhu, Chang Lan, Bairen Yi, Yong Cui, and Chuanxiong Guo. A unified architecture for accelerating distributed dnn training in heterogeneous gpu/cpu clusters. In Proc. USENIX OSDI, pages 463--479, 2020.

4. Amedeo Sapio, Marco Canini, Chen-Yu Ho, Jacob Nelson, Panos Kalnis, Changhoon Kim, Arvind Krishnamurthy, Masoud Moshref, Dan RK Ports, and Peter Richtarik. Scaling distributed machine learning with in-network aggregation. In Proc. USENIX NSDI, pages 785--808, 2021.

5. Nadeen Gebara, Manya Ghobadi, and Paolo Costa. In-network aggregation for shared machine learning clusters. In Proc. MLSys, pages 829--844, 2021.

Cited by 10 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

1. Achieving High Efficiency for Datacenter Multicast using Skewed Bloom Filter;Proceedings of the 53rd International Conference on Parallel Processing;2024-08-12

2. OmNICCL: Zero-cost Sparse AllReduce with Direct Cache Access and SmartNICs;Proceedings of the 2024 SIGCOMM Workshop on Networks for AI Computing;2024-08-04

3. ACU: Aggregator-based Congestion control and link Utilization optimization strategy for multi-tenant in-network aggregation;Proceedings of the 8th Asia-Pacific Workshop on Networking;2024-08-03

4. Accelerating Distributed Training With Collaborative In-Network Aggregation;IEEE/ACM Transactions on Networking;2024-08

5. A Multidimensional Communication Scheduling Method for Hybrid Parallel DNN Training;IEEE Transactions on Parallel and Distributed Systems;2024-08

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3