With the advent of deep learning, neural network-based recommendation models
have emerged as an important tool for tackling personalization and
recommendation tasks. These networks differ significantly from other deep
learning networks due to their need to handle categorical features and are not
well studied or understood. In this paper, we develop a state-of-the-art deep
learning recommendation model (DLRM) and provide its implementation in both
PyTorch and Caffe2 frameworks. In addition, we design a specialized
parallelization scheme utilizing model parallelism on the embedding tables to
mitigate memory constraints while exploiting data parallelism to scale-out
compute from the fully-connected layers. We compare DLRM against existing
recommendation models and characterize its performance on the Big Basin AI
platform, demonstrating its usefulness as a benchmark for future algorithmic
experimentation and system co-design.
Description
[1906.00091] Deep Learning Recommendation Model for Personalization and Recommendation Systems
%0 Generic
%1 naumov2019learning
%A Naumov, Maxim
%A Mudigere, Dheevatsa
%A Shi, Hao-Jun Michael
%A Huang, Jianyu
%A Sundaraman, Narayanan
%A Park, Jongsoo
%A Wang, Xiaodong
%A Gupta, Udit
%A Wu, Carole-Jean
%A Azzolini, Alisson G.
%A Dzhulgakov, Dmytro
%A Mallevich, Andrey
%A Cherniavskii, Ilia
%A Lu, Yinghai
%A Krishnamoorthi, Raghuraman
%A Yu, Ansha
%A Kondratenko, Volodymyr
%A Pereira, Stephanie
%A Chen, Xianjie
%A Chen, Wenlin
%A Rao, Vijay
%A Jia, Bill
%A Xiong, Liang
%A Smelyanskiy, Misha
%D 2019
%K embedding machinelearning neuralnetwork
%T Deep Learning Recommendation Model for Personalization and
Recommendation Systems
%U http://arxiv.org/abs/1906.00091
%X With the advent of deep learning, neural network-based recommendation models
have emerged as an important tool for tackling personalization and
recommendation tasks. These networks differ significantly from other deep
learning networks due to their need to handle categorical features and are not
well studied or understood. In this paper, we develop a state-of-the-art deep
learning recommendation model (DLRM) and provide its implementation in both
PyTorch and Caffe2 frameworks. In addition, we design a specialized
parallelization scheme utilizing model parallelism on the embedding tables to
mitigate memory constraints while exploiting data parallelism to scale-out
compute from the fully-connected layers. We compare DLRM against existing
recommendation models and characterize its performance on the Big Basin AI
platform, demonstrating its usefulness as a benchmark for future algorithmic
experimentation and system co-design.
@misc{naumov2019learning,
abstract = {With the advent of deep learning, neural network-based recommendation models
have emerged as an important tool for tackling personalization and
recommendation tasks. These networks differ significantly from other deep
learning networks due to their need to handle categorical features and are not
well studied or understood. In this paper, we develop a state-of-the-art deep
learning recommendation model (DLRM) and provide its implementation in both
PyTorch and Caffe2 frameworks. In addition, we design a specialized
parallelization scheme utilizing model parallelism on the embedding tables to
mitigate memory constraints while exploiting data parallelism to scale-out
compute from the fully-connected layers. We compare DLRM against existing
recommendation models and characterize its performance on the Big Basin AI
platform, demonstrating its usefulness as a benchmark for future algorithmic
experimentation and system co-design.},
added-at = {2020-08-04T02:00:54.000+0200},
author = {Naumov, Maxim and Mudigere, Dheevatsa and Shi, Hao-Jun Michael and Huang, Jianyu and Sundaraman, Narayanan and Park, Jongsoo and Wang, Xiaodong and Gupta, Udit and Wu, Carole-Jean and Azzolini, Alisson G. and Dzhulgakov, Dmytro and Mallevich, Andrey and Cherniavskii, Ilia and Lu, Yinghai and Krishnamoorthi, Raghuraman and Yu, Ansha and Kondratenko, Volodymyr and Pereira, Stephanie and Chen, Xianjie and Chen, Wenlin and Rao, Vijay and Jia, Bill and Xiong, Liang and Smelyanskiy, Misha},
biburl = {https://www.bibsonomy.org/bibtex/21f63ae6e4027bc61f005d6435f75586e/cpankow},
description = {[1906.00091] Deep Learning Recommendation Model for Personalization and Recommendation Systems},
interhash = {2abbf48379aa62102de3248ad1dffe69},
intrahash = {1f63ae6e4027bc61f005d6435f75586e},
keywords = {embedding machinelearning neuralnetwork},
note = {cite arxiv:1906.00091Comment: 10 pages, 6 figures},
timestamp = {2020-08-04T02:00:54.000+0200},
title = {Deep Learning Recommendation Model for Personalization and
Recommendation Systems},
url = {http://arxiv.org/abs/1906.00091},
year = 2019
}