Zobrazeno 1 - 10
of 77
pro vyhledávání: '"Kejariwal, Arun"'
Autor:
Zha, Daochen, Feng, Louis, Luo, Liang, Bhushanam, Bhargav, Liu, Zirui, Hu, Yusuo, Nie, Jade, Huang, Yuzhen, Tian, Yuandong, Kejariwal, Arun, Hu, Xia
Sharding a large machine learning model across multiple devices to balance the costs is important in distributed training. This is challenging because partitioning is NP-hard, and estimating the costs accurately and efficiently is difficult. In this
Externí odkaz:
http://arxiv.org/abs/2305.01868
Autor:
Zha, Daochen, Feng, Louis, Tan, Qiaoyu, Liu, Zirui, Lai, Kwei-Herng, Bhushanam, Bhargav, Tian, Yuandong, Kejariwal, Arun, Hu, Xia
We study embedding table placement for distributed recommender systems, which aims to partition and place the tables on multiple hardware devices (e.g., GPUs) to balance the computation and communication costs. Although prior work has explored learni
Externí odkaz:
http://arxiv.org/abs/2210.02023
Autor:
Ye, Mao, Jiang, Ruichen, Wang, Haoxiang, Choudhary, Dhruv, Du, Xiaocong, Bhushanam, Bhargav, Mokhtari, Aryan, Kejariwal, Arun, Liu, Qiang
One of the key challenges of learning an online recommendation model is the temporal domain shift, which causes the mismatch between the training and testing data distribution and hence domain generalization error. To overcome, we propose to learn a
Externí odkaz:
http://arxiv.org/abs/2209.01143
Autor:
Zha, Daochen, Feng, Louis, Bhushanam, Bhargav, Choudhary, Dhruv, Nie, Jade, Tian, Yuandong, Chae, Jay, Ma, Yinbin, Kejariwal, Arun, Hu, Xia
Embedding learning is an important technique in deep recommendation models to map categorical features to dense vectors. However, the embedding tables often demand an extremely large number of parameters, which become the storage and efficiency bottl
Externí odkaz:
http://arxiv.org/abs/2208.06399
Autor:
Lin, Zhongyi, Feng, Louis, Ardestani, Ehsan K., Lee, Jaewon, Lundell, John, Kim, Changkyu, Kejariwal, Arun, Owens, John D.
We devise a performance model for GPU training of Deep Learning Recommendation Models (DLRM), whose GPU utilization is low compared to other well-optimized CV and NLP models. We show that both the device active time (the sum of kernel runtimes) but a
Externí odkaz:
http://arxiv.org/abs/2201.07821
Autor:
Du, Xiaocong, Bhushanam, Bhargav, Yu, Jiecao, Choudhary, Dhruv, Gao, Tianxiang, Wong, Sherman, Feng, Louis, Park, Jongsoo, Cao, Yu, Kejariwal, Arun
Deep learning recommendation systems at scale have provided remarkable gains through increasing model capacity (i.e. wider and deeper neural networks), but it comes at significant training cost and infrastructure cost. Model pruning is an effective t
Externí odkaz:
http://arxiv.org/abs/2105.01064
Autor:
Gupta, Vipul, Choudhary, Dhruv, Tang, Ping Tak Peter, Wei, Xiaohan, Wang, Xing, Huang, Yuzhen, Kejariwal, Arun, Ramchandran, Kannan, Mahoney, Michael W.
In this paper, we consider hybrid parallelism -- a paradigm that employs both Data Parallelism (DP) and Model Parallelism (MP) -- to scale distributed training of large recommendation models. We propose a compression framework called Dynamic Communic
Externí odkaz:
http://arxiv.org/abs/2010.08899
Autor:
Ye, Mao, Choudhary, Dhruv, Yu, Jiecao, Wen, Ellie, Chen, Zeliang, Yang, Jiyan, Park, Jongsoo, Liu, Qiang, Kejariwal, Arun
Large scale deep learning provides a tremendous opportunity to improve the quality of content recommendation systems by employing both wider and deeper models, but this comes at great infrastructural cost and carbon footprint in modern data centers.
Externí odkaz:
http://arxiv.org/abs/2010.08655
Ever growing volume and velocity of data coupled with decreasing attention span of end users underscore the critical need for real-time analytics. In this regard, anomaly detection plays a key role as an application as well as a means to verify data
Externí odkaz:
http://arxiv.org/abs/1710.04735
Velocity is one of the 4 Vs commonly used to characterize Big Data. In this regard, Forrester remarked the following in Q3 2014: "The high velocity, white-water flow of data from innumerable real-time data sources such as market data, Internet of Thi
Externí odkaz:
http://arxiv.org/abs/1708.02621