# Image Classification at Supercomputer Scale

Source: [Image Classification at Supercomputer Scale](<https://www.sota2.com/research/paper/68fa65201e209131cea089be>)

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

Data updated: 2025-10-25T21:19:52.933Z

Authors: Chris Ying, Sameer Kumar, Dehao Chen, Tao Wang, Youlong Cheng

Published: 2018-11-16T19:01:40.000Z

## Abstract

Deep learning is extremely computationally intensive, and hardware vendors have responded by building faster accelerators in large clusters. Training deep learning models at petaFLOPS scale requires overcoming both algorithmic and systems software challenges. In this paper, we discuss three systems-related optimizations: (1) distributed batch normalization to control per-replica batch sizes, (2) input pipeline optimizations to sustain model throughput, and (3) 2-D torus all-reduce to speed up gradient summation. We combine these optimizations to train ResNet-50 on ImageNet to 76.3% accuracy in 2.2 minutes on a 1024-chip TPU v3 Pod with a training throughput of over 1.05 million images/second and no accuracy drop.

## 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/1811.06992>)

[PDF](<https://arxiv.org/pdf/1811.06992.pdf>)