paper-with-me

Papers

Accelerating Distributed ML Training via Selective Synchronization

2023-07-16 · Sahil Tyagi, Martin Swany

In distributed training, deep neural networks (DNNs) are launched over multiple workers concurrently and aggregate their local updates on each step in bulk-synchronous parallel (BSP) training. However, BSP does not linearly scale-out due to high communication cost of aggregation. To mitigate this overhead, alternatives like Federated Averaging (FedAvg) and Stale-Synchronous Parallel (SSP) either reduce synchronization frequency or eliminate it altogether, usually at the cost of lower final accuracy. In this paper, we present \texttt{SelSync}, a practical, low-overhead method for DNN training that dynamically chooses to incur or avoid communication at each step either by calling the aggregation op or applying local updates based on their significance. We propose various optimizations as part of \texttt{SelSync} to improve convergence in the context of \textit{semi-synchronous} training. Our system converges to the same or better accuracy than BSP while reducing training time by up to 14$\times$.

📄 PDF Abstract BibTeX arXiv:2307.07950

Code (1)

sahiltyagi4/selsync 공식 구현 pytorch

Similar Papers 제목 키워드 기반

Accelerating Neural Network Training with Distributed Asynchronous and Selective Optimization (DASO)

2021-04-12 · Daniel Coquelin, Charlotte Debus, Markus Götz, Fabrice von der Lehr 외

With increasing data and model complexities, the time required to train neural networks has become prohibitively large. To address the exponential rise in training time, users are turning to data parallel neural networks…

BlockingGPU

DeAR: Accelerating Distributed Deep Learning with Fine-Grained All-Reduce Pipelining

2023-02-24 · Lin Zhang, Shaohuai Shi, Xiaowen Chu, Wei Wang 외

Communication scheduling has been shown to be effective in accelerating distributed training, which enables all-reduce communications to be overlapped with backpropagation computations. This has been commonly adopted in …

AllGPUScheduling

Accelerating AllReduce with a Persistent Straggler

2025-05-29 · Arjun Devraj, Eric Ding, Abhishek Vijaya Kumar, Robert Kleinberg 외

Distributed machine learning workloads use data and tensor parallelism for training and inference, both of which rely on the AllReduce collective to synchronize gradients or activations. However, bulk-synchronous AllRedu…

GPU

Lion Cub: Minimizing Communication Overhead in Distributed Lion

2024-11-25 · Satoki Ishikawa, Tal Ben-Nun, Brian Van Essen, Rio Yokota 외

Communication overhead is a key challenge in distributed deep learning, especially on slower Ethernet interconnects, and given current hardware trends, communication is likely to become a major bottleneck. While gradient…

Quantization

PyTorch Distributed: Experiences on Accelerating Data Parallel Training

2020-06-28 · Shen Li, Yanli Zhao, Rohan Varma, Omkar Salpekar 외

This paper presents the design, implementation, and evaluation of the PyTorch distributed data parallel module. PyTorch is a widely-adopted scientific computing package used in deep learning research and applications. Re…