Categories

Mobile — 122
NSL — 22
NSL Project — 2

PipeDream: Fast and Efficient Pipeline Parallel DNN Training

by Jinse Kwon on 2019-06-12 11:18:17

Date : 2019. 06. 26 (Wed) 13:00 Locate : EB5. 533 Presenter : Jinse Kwon   Title : PipeDream: Fast and Efficient Pipeline Parallel DNN Training Author : Aaron Harlap, Deepak Narayanan, Amar Phanishayee, Vivek Seshadri, Nikhil Devanur, Greg Ganger, Phil Gibbons (Microsoft Research, Carnegie Mellon University, Stanford University)   Abstract : PipeDream is a Deep Neural Network(DNN) training system for GPUs that parallelizes computation by pipelining execution across multiple machines. Its pipeline parallel computing model avoids the slowdowns faced by data-parallel training when large models and/or limited network bandwidth induce high communication-to-computation ratios. PipeDream reduces communication by up to 95% for large DNNs relative to data-parallel training, and allows perfect overlap of communication and computation. PipeDream keeps all available GPUs productive by systematically partitioning DNN ... Continue reading →

226 Views

How can do good research?

by Daeyoung Song on 2019-06-11 00:00:00

Date : 2019.06.13(Thur.) Locate : EB5. 607 Presenter : Hyungshin Kim Title : How Continue reading →

206 Views

[Xenomai] An Overview of the Real-Time Framework for Linux

by Hyeoksoo Jang on 2019-06-04 10:46:08

Date : 2019.6.5(Wed) Locate : EB5. 533 Presenter : Hyeoksoo Jang Title :  An Overview of the Real-Time Framework for Linux   Continue reading →

194 Views

How can do good presentation?

by Daeyoung Song on 2019-05-28 14:16:51

Date : 2019.06.04(The.) Locate : EB5. 410 Presenter : Hyungshin Kim Title : How can do good presentation? We will have time to study 'Presentation skill'. Continue reading →

183 Views

A new golden age for computer architecture

by Daeyoung Song on 2019-05-27 21:01:32

Date : 2019.5.30(Thu) Locate : EB5. 607 Presenter : Yeji Son Title :  A new golden age for computer architecture : Future Opportunities in Computer Architecture  Continue reading →

185 Views

멀티코어 시스템에서의 TLB Lockdown을 통한 WCET 개선

by Daeyoung Song on 2019-05-27 20:51:05

Date : 2019.5.30(Thu) Locate : EB5. 533 Presenter : Daeyoung Song Title :  멀티코어 시스템에서의 TLB Lockdown을 통한 WCET 개선  Continue reading →

87 Views

모바일 GPU에서 OpenCL 선형대수 라이브러리의 비정방 행렬 곱셈 최적화

by Byungkyo Jung on 2019-05-23 11:34:20

Date : 2019.5.23(Thu) Locate : EB5. 533 Presenter : Byeongkyo Cheong Title : 모바일 GPU에서 OpenCL 선형대수 라이브러리의 비정방 행렬 곱셈 최적화 Continue reading →

89 Views

How can good research be done

by Daeyoung Song on 2019-05-17 11:18:24

Date : 2019.05.23(Thu.) Locate : EB5. 410 Presenter : Hyungshin Kim Title : How can good research be done? We will have time to study 'Research skill'. Continue reading →

112 Views

How to write a paper in the system field

by Daeyoung Song on 2019-05-15 17:42:39

Date : 2019.05.16(Thu) Locate : EB5. 607 Presenter : Hyungshin Kim Title : How to write a paper in the system field We will have time to study how to write systems papers.   Continue reading →

77 Views

GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism

by Donghee Ha on 2019-05-14 11:29:54

Date : 2019.05.16(Thu) Locate : EB5. 533 Presenter : Donghee Ha Title : GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism Author : Yanping Huang, Yonglong Cheng, Dehao Chen, HyoukJoong Lee, Jiquan Ngiam, Quoc V. Le, Zhifeng Chen Abstract : GPipe is a scalable pipeline parallelism library that enables learning of giant deep neural networks. It partitions network layers across accelerators and pipelines execution to achieve high hardware utilization. It leverages recomputation to minimize activation memory usage. For example, using partitions over 8 accelerators, it is able to train networks that are 25x larger, demonstrating its scalability. It also guarantees that the computed gradients remain consistent regardless of the number of partitions. It achieves an almost linear speed up without any changes in the model parameters: when using 4x more accelerators, training the same model is up to 3.5x faster. We train a 557 ... Continue reading →

102 Views

How to read papers

by Daeyoung Song on 2019-05-13 00:00:00

Date : 2019.05.14(Tue.) Locate : EB5. 410 Presenter : Hyungshin Kim Title : How Continue reading →

92 Views

How to write good papers

by Daeyoung Song on 2019-05-09 20:55:18

Date : 2019.05.09(Thu) Locate : EB5. 607 Presenter : Hyungshin Kim Title : How to write good papers We will have time to study 'Writing paper skill'.   Continue reading →

77 Views

Predictable Flight Management System Implementation on a Multicore Processor

by Sihyeong Park on 2019-04-29 11:51:09

Date : 2019.05.02(Thu) Locate : EB5. 533 Presenter : Sihyeong Park Title : Predictable Flight Management System Implementation on a Multicore Processor Author : Guy Durrieu, Madeleine Faugère, Sylvain Girbal, Daniel Gracia Pérez, Claire Pagetti, W. Puffitsch  Abstract : This paper presents an approach for hosting a representative avionic function on a distributed-memory mul-ticore COTS architecture. This approach was developed in collaboration by Thales and ONERA, in order to improve the performance of the function while enforcing its predictability. Once the target avionic function and the multicore architecture have been introduced, the execution model and the needed basic services are described and evaluated. Continue reading →

97 Views

[Invite Talk] Jaymin Lee

by Jinyoung Choi on 2019-04-05 11:28:05

Date : 2019. 4. 8 (Mon) 11:00 Locate : EB5, 508 Presenter : Jaymin Lee   We will have time to ask and answer how to prepare for employment and the presenter, Jaymin Lee will present about research trend. Jaymin Lee works for ETRI, and as the most recent graduate Ph.D. student, look forward to the great talk!   Continue reading →

177 Views

Tracking Depression Dynamics in College Students Using Mobile Phone and Wearable Sensing

by Jinyoung Choi on 2019-04-04 22:09:49

Date : 2019. 04. 17 (Wed) 13:30 Locate : EB5. 533 Presenter : Jinyoung Choi   Title : Tracking Depression Dynamics in College Students Using Mobile Phone and Wearable Sensing Author : Rui Wang, Weichen Wang, Alex Dasilva, Jeremy F. Huckins, William M. Kelley, Todd F. Heatherton, Andrew T. Champbell (Dartmouth College)   Abstract : There are rising rates of depression on college campuses. Mental health services on our campuses are working at full stretch. In response researchers have proposed using mobile sensing for continuous mental health assessment. Existing work on understanding the relationship between mobile sensing and depression, however, focuses on generic behavioral features that do not map to major depressive disorder symptoms defined in the standard mental disorders diagnostic manual (DSM-5). We propose a new approach to predicting depression using passive sensing data from students’ smartphones and wearables. We ... Continue reading →

171 Views