DLBricks: Composable Benchmark Generation to Reduce Deep Learning Benchmarking Effort on CPUs (Extended)

TL;DR

DLBricks introduces a composable benchmark generation approach that simplifies and accelerates benchmarking of deep learning models on CPUs by decomposing models into reusable layers, enabling up-to-date and proprietary model evaluation.

Contribution

It proposes a novel method to automatically generate representative benchmarks from DL layers, reducing effort and time in benchmarking diverse and evolving models on CPUs.

Findings

Achieves up to 4.4x faster benchmarking time.

Provides 95% accuracy in performance estimation.

Supports proprietary models within benchmark suites.

Abstract

The past few years have seen a surge of applying Deep Learning (DL) models for a wide array of tasks such as image classification, object detection, machine translation, etc. While DL models provide an opportunity to solve otherwise intractable tasks, their adoption relies on them being optimized to meet latency and resource requirements. Benchmarking is a key step in this process but has been hampered in part due to the lack of representative and up-to-date benchmarking suites. This is exacerbated by the fast-evolving pace of DL models. This paper proposes DLBricks, a composable benchmark generation design that reduces the effort of developing, maintaining, and running DL benchmarks on CPUs. DLBricks decomposes DL models into a set of unique runnable networks and constructs the original model's performance using the performance of the generated benchmarks. DLBricks leverages two key observations: DL layers are the performance building blocks of DL models and layers are extensively repeated within and across DL models. Since benchmarks are generated automatically and the benchmarking time is minimized, DLBricks can keep up-to-date with the latest proposed models, relieving the pressure of selecting representative DL models. Moreover, DLBricks allows users to represent proprietary models within benchmark suites. We evaluate DLBricks using $50$ MXNet models spanning $5$ DL tasks on $4$ representative CPU systems. We show that DLBricks provides an accurate performance estimate for the DL models and reduces the benchmarking time across systems (e.g. within $95\%$ accuracy and up to $4.4\times$ benchmarking time speedup on Amazon EC2 c5.xlarge).