A Multidimensional Communication Scheduling Method for Hybrid Parallel DNN Training

计算机科学 调度(生产过程) 异步通信 并行计算 人工智能 计算机网络 运营管理 经济
作者
Shengwei Li,Kai Lü,Zhiquan Lai,Weijie Liu,Keshi Ge,Dongsheng Li
出处
期刊:IEEE Transactions on Parallel and Distributed Systems [Institute of Electrical and Electronics Engineers]
卷期号:35 (8): 1415-1428 被引量:20
标识
DOI:10.1109/tpds.2024.3406420
摘要

The transformer-based deep neural network (DNN) models have shown considerable success across diverse tasks, prompting widespread adoption of distributed training methods such as data parallelism and pipeline parallelism. With the increasing parameter number, hybrid parallel training becomes imperative to scale training. The primary bottleneck in scaling remains the communication overhead. The communication scheduling technique, emphasizing the overlap of communication with computation, has demonstrated its benefits in scaling. However, most existing works focus on data parallelism, overlooking the nuances of hybrid parallel training. In this paper, we proposeTriRace, an efficient communication scheduling framework for accelerating communications in hybrid parallel training of asynchronous pipeline parallelism and data parallelism. To achieve effective computation-communication overlap,TriRaceintroduces3D communication scheduling, which adeptly leverages data dependencies between communication and computations, efficiently scheduling AllReduce communication, sparse communication, and peer-to-peer communication in hybrid parallel training. To avoid possible communication contentions,TriRacealso incorporates atopology-aware runtimewhich optimizes the execution of communication operations by considering ongoing communication operations and real-time network status. We have implemented a prototype ofTriRacebased on PyTorch and Pipedream-2BW, and conducted comprehensive evaluations with three representative baselines. Experimental results show thatTriRaceachieves up to 1.07–1.45× speedup compared to the state-of-the-art pipeline parallelism training baseline Pipedream-2BW, and 1.24–1.81× speedup compared to the Megatron.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
昏黄发布了新的文献求助10
刚刚
Nnnnnn发布了新的文献求助10
刚刚
1秒前
1秒前
2秒前
共享精神的应助被wyc采纳,获得10
2秒前
hermit的应助被jisooya采纳,获得50
2秒前
2秒前
我是老大的应助被宿宿采纳,获得10
4秒前
5秒前
6秒前
予安完成签到,获得积分20
6秒前
OK完成签到,获得积分10
6秒前
风吹麦浪的应助被胡萝卜卜采纳,获得10
7秒前
大个的应助被之予无为采纳,获得10
7秒前
科研通AI65的应助被zo采纳,获得10
7秒前
KIB_yu发布了新的文献求助10
8秒前
TaoJ的应助被坚定的采波采纳,获得10
8秒前
WTTPAXL发布了新的文献求助10
8秒前
8秒前
9秒前
9秒前
乐乐的应助被叫你吃布丁采纳,获得10
9秒前
小二郎的应助被小太阳采纳,获得10
10秒前
11秒前
11秒前
zuoshoubo发布了新的文献求助10
11秒前
11秒前
一个橡果发布了新的文献求助10
13秒前
自由土豆完成签到,获得积分10
13秒前
14秒前
wyc发布了新的文献求助10
15秒前
16秒前
FashionBoy的应助被开心快乐大王采纳,获得10
16秒前
16秒前
璐璇完成签到 ,获得积分10
18秒前
guofuyan完成签到 ,获得积分10
18秒前
18秒前
19秒前
molihuakai的应助被苗条梨愁采纳,获得10
19秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
八旗制度の研究 600
Improving the wear resistance of vehicle gear oil by oil soluble LaF3 nanoparticles and typical additives of gear oil 500
Boundaries and Bodhisattvas 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8183694
求助须知:如何正确求助?哪些是违规求助? 9572963
关于积分的说明 21828306
捐赠科研通 7780864
什么是DOI,文献DOI怎么找? 3434057
关于科研通互助平台的介绍 2558257
邀请新用户注册赠送积分活动 2480541