A Modular Benchmarking Infrastructure for High-Performance and Reproducible Deep Learning
This provides a foundational tool for researchers and practitioners to evaluate deep learning methods consistently, though it is incremental in building on existing benchmarking concepts.
The authors tackled the lack of a standardized benchmarking system for deep learning by introducing Deep500, a modular infrastructure that enables fair and reproducible comparisons across frameworks and techniques, with results showing negligible overhead and support for extreme-scale workloads.
We introduce Deep500: the first customizable benchmarking infrastructure that enables fair comparison of the plethora of deep learning frameworks, algorithms, libraries, and techniques. The key idea behind Deep500 is its modular design, where deep learning is factorized into four distinct levels: operators, network processing, training, and distributed training. Our evaluation illustrates that Deep500 is customizable (enables combining and benchmarking different deep learning codes) and fair (uses carefully selected metrics). Moreover, Deep500 is fast (incurs negligible overheads), verifiable (offers infrastructure to analyze correctness), and reproducible. Finally, as the first distributed and reproducible benchmarking system for deep learning, Deep500 provides software infrastructure to utilize the most powerful supercomputers for extreme-scale workloads.