# Massively Distributed SGD: ImageNet/ResNet-50 Training in a Flash

Source: [Massively Distributed SGD: ImageNet/ResNet-50 Training in a Flash](<https://www.sota2.com/research/paper/68fa65711e209131cea0a746>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-25T21:27:45.299Z

Authors: Hiroaki Mikami, Hisahiro Suganuma, Pongsakorn U-chupala, Yoshiki Tanaka, Yuichi Kageyama

Published: 2018-11-13T11:52:04.000Z

## Abstract

Scaling the distributed deep learning to a massive GPU cluster level is challenging due to the instability of the large mini-batch training and the overhead of the gradient synchronization. We address the instability of the large mini-batch training with batch-size control and label smoothing. We address the overhead of the gradient synchronization with 2D-Torus all-reduce. Specifically, 2D-Torus all-reduce arranges GPUs in a logical 2D grid and performs a series of collective operation in different orientations. These two techniques are implemented with Neural Network Libraries (NNL). We have successfully trained ImageNet/ResNet-50 in 122 seconds without significant accuracy loss on ABCI cluster.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/1811.05233>)

[PDF](<https://arxiv.org/pdf/1811.05233.pdf>)