The application of deep learning techniques resulted in remarkable improvement of machine learning models. In this paper provides detailed characterizations of deep learning models used in many Facebook social network services. We present computational characteristics of our models, describe high performance optimizations targeting existing systems, point out their limitations and make suggestions for the future general-purpose/accelerated inference hardware. Also, we highlight the need for better co-design of algorithms, numerics and computing platforms to address the challenges of workloads often run in data centers.
@article{arxiv.1811.09886,
title = {Deep Learning Inference in Facebook Data Centers: Characterization, Performance Optimizations and Hardware Implications},
author = {Jongsoo Park and Maxim Naumov and Protonu Basu and Summer Deng and Aravind Kalaiah and Daya Khudia and James Law and Parth Malani and Andrey Malevich and Satish Nadathur and Juan Pino and Martin Schatz and Alexander Sidorov and Viswanath Sivakumar and Andrew Tulloch and Xiaodong Wang and Yiming Wu and Hector Yuen and Utku Diril and Dmytro Dzhulgakov and Kim Hazelwood and Bill Jia and Yangqing Jia and Lin Qiao and Vijay Rao and Nadav Rotem and Sungjoo Yoo and Mikhail Smelyanskiy},
journal= {arXiv preprint arXiv:1811.09886},
year = {2018}
}