Where Is My Training Bottleneck? Hidden Trade-Offs in Deep Learning Preprocessing Pipelines

TL;DR

This paper analyzes data preprocessing bottlenecks in deep learning pipelines, introduces a profiling library for optimization, and demonstrates up to 13x throughput improvements through tuning.

Contribution

It provides a new perspective on dataset preparation trade-offs and offers an open-source tool for automatic preprocessing strategy selection.

Findings

Up to 13x throughput increase with tuning

Trade-offs between throughput, preprocessing time, and storage

Open-source profiling library for pipeline optimization

Abstract

Preprocessing pipelines in deep learning aim to provide sufficient data throughput to keep the training processes busy. Maximizing resource utilization is becoming more challenging as the throughput of training processes increases with hardware innovations (e.g., faster GPUs, TPUs, and inter-connects) and advanced parallelization techniques that yield better scalability. At the same time, the amount of training data needed in order to train increasingly complex models is growing. As a consequence of this development, data preprocessing and provisioning are becoming a severe bottleneck in end-to-end deep learning pipelines. In this paper, we provide an in-depth analysis of data preprocessing pipelines from four different machine learning domains. We introduce a new perspective on efficiently preparing datasets for end-to-end deep learning pipelines and extract individual trade-offs to optimize throughput, preprocessing time, and storage consumption. Additionally, we provide an open-source profiling library that can automatically decide on a suitable preprocessing strategy to maximize throughput. By applying our generated insights to real-world use-cases, we obtain an increased throughput of 3x to 13x compared to an untuned system while keeping the pipeline functionally identical. These findings show the enormous potential of data pipeline tuning.