paper-with-me

홈 › Papers

MergeComp: A Compression Scheduler for Scalable Communication-Efficient Distributed Training

2021-03-28 · Zhuang Wang, Xinyu Wu, T. S. Eugene Ng

Large-scale distributed training is increasingly becoming communication bound. Many gradient compression algorithms have been proposed to reduce the communication overhead and improve scalability. However, it has been observed that in some cases gradient compression may even harm the performance of distributed training. In this paper, we propose MergeComp, a compression scheduler to optimize the scalability of communication-efficient distributed training. It automatically schedules the compression operations to optimize the performance of compression algorithms without the knowledge of model architectures or system parameters. We have applied MergeComp to nine popular compression algorithms. Our evaluations show that MergeComp can improve the performance of compression algorithms by up to 3.83x without losing accuracy. It can even achieve a scaling factor of distributed training up to 99% over high-speed networks.

📄 PDF Abstract BibTeX arXiv:2103.15195

Code (1)

Crystal-wxy/mergeComp 공식 구현 pytorch

Similar Papers 제목 키워드 기반

CrossoverScheduler: Overlapping Multiple Distributed Training Applications in a Crossover Manner

2021-03-14 · Cheng Luo, Lei Qu, Youshan Miao, Peng Cheng 외

Distributed deep learning workloads include throughput-intensive training tasks on the GPU clusters, where the Distributed Stochastic Gradient Descent (SGD) incurs significant communication delays after backward propagat…

Deep LearningGPUimage-classificationImage Classification

Reinforcement Learning Scheduler for Vehicle-to-Vehicle Communications Outside Coverage

2019-04-29 · Taylan Şahin, Ramin Khalili, Mate Boban, Adam Wolisz

Radio resources in vehicle-to-vehicle (V2V) communication can be scheduled either by a centralized scheduler residing in the network (e.g., a base station in case of cellular systems) or a distributed scheduler, where th…

reinforcement-learningReinforcement LearningReinforcement Learning (RL)Scheduling

ScaleCom: Scalable Sparsified Gradient Compression for Communication-Efficient Distributed Training

2021-04-21 · NeurIPS 2020 12 · Chia-Yu Chen, Jiamin Ni, Songtao Lu, Xiaodong Cui 외

Large-scale distributed training of Deep Neural Networks (DNNs) on state-of-the-art platforms is expected to be severely communication constrained. To overcome this limitation, numerous gradient compression techniques ha…

Compressed Communication for Distributed Training: Adaptive Methods and System

2021-05-17 · Yuchen Zhong, Cong Xie, Shuai Zheng, Haibin Lin

Communication overhead severely hinders the scalability of distributed machine learning systems. Recently, there has been a growing interest in using gradient compression to reduce the communication overhead of the distr…

Reconciling Communication Compression and Byzantine-Robustness in Distributed Learning

2025-08-23 · Diksha Gupta, Antonio Honsell, Chuan Xu, Nirupam Gupta 외 arxiv

Distributed learning enables scalable model training over decentralized data, but remains hindered by Byzantine faults and high communication costs. While both challenges have been studied extensively in isolation, their…