國立虎尾科技大學 |

Exploiting Heterogeneity for Efficient Deep Learning.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Exploiting Heterogeneity for Efficient Deep Learning./
Author:	Lin, Shouxu.
Description:	1 online resource (56 pages)
Notes:	Source: Masters Abstracts International, Volume: 84-06.
Contained By:	Masters Abstracts International84-06.
Subject:	Computer science. -
Online resource:	click for full text (PQDT)
ISBN:	9798363516306

Exploiting Heterogeneity for Efficient Deep Learning.
Lin, Shouxu.

Exploiting Heterogeneity for Efficient Deep Learning. - 1 online resource (56 pages)

Source: Masters Abstracts International, Volume: 84-06.

Thesis (M.S.)--Carnegie Mellon University, 2022.

Includes bibliographical references

In recent years, we have observed that the scale of Deep Neural Network (DNN) Models, with respect to the number of training examples, the number of model parameters, or both, has multiplied by several orders of magnitude. Accompanied by this is the significant growth in the computational requirements for training these large DNN models. To accelerate these giant DNN models, several different parallelization strategies, such as data and model parallelism, have been widely adopted, where the computation workloads are distributed and parallelized across multiple accelerators. Several recent works have been conducted to automate the parallelization strategy for either model-specific or general DNN models.This thesis steps beyond the existing early works in distributed deep learning by exploiting hardware accelerator and DNN model heterogeneity. We investigate the exploitation of heterogeneity through two case studies: heterogeneity-aware DL cluster scheduler and heterogeneity-aware DNN parallelization planner.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798363516306Subjects--Topical Terms:

573171
Computer science.
Subjects--Index Terms:

HeterogeneityIndex Terms--Genre/Form:

554714
Electronic books.

Exploiting Heterogeneity for Efficient Deep Learning.
LDR:02259ntm a22003617 4500 001 1147253
005 20240909100738.5
006 m o d
007 cr bn ---uuuuu
008 250605s2022 xx obm 000 0 eng d
020 $a 9798363516306
035 $a (MiAaPQ)AAI30242225
035 $a AAI30242225
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Lin, Shouxu. $3 1472934
245 1 0 $a Exploiting Heterogeneity for Efficient Deep Learning.
264 0 $c 2022
300 $a 1 online resource (56 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Masters Abstracts International, Volume: 84-06.
500 $a Advisor: Ganger, Greg.
502 $a Thesis (M.S.)--Carnegie Mellon University, 2022.
504 $a Includes bibliographical references
520 $a In recent years, we have observed that the scale of Deep Neural Network (DNN) Models, with respect to the number of training examples, the number of model parameters, or both, has multiplied by several orders of magnitude. Accompanied by this is the significant growth in the computational requirements for training these large DNN models. To accelerate these giant DNN models, several different parallelization strategies, such as data and model parallelism, have been widely adopted, where the computation workloads are distributed and parallelized across multiple accelerators. Several recent works have been conducted to automate the parallelization strategy for either model-specific or general DNN models.This thesis steps beyond the existing early works in distributed deep learning by exploiting hardware accelerator and DNN model heterogeneity. We investigate the exploitation of heterogeneity through two case studies: heterogeneity-aware DL cluster scheduler and heterogeneity-aware DNN parallelization planner.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer science. $3 573171
650 4 $a Information science. $3 561178
653 $a Heterogeneity
653 $a Deep Learning
653 $a Deep Neural Networks
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0723
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Carnegie Mellon University. $b Information Networking Institute. $3 1466992
773 0 $t Masters Abstracts International $g 84-06.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30242225 $z click for full text (PQDT)