GTC ON-DEMAND

 
SEARCH SESSIONS
SEARCH SESSIONS

Search All
 
Refine Results:
 
Year(s)

SOCIAL MEDIA

EMAIL SUBSCRIPTION

 
 

GTC ON-DEMAND

Data Center & Cloud Infrastructure
Presentation
Media
Large-scale GPU Deep Learning Platform Design and Case Analysis
Abstract:
We'll explain the strategy on how to design large-scale deep learning platforms using HPC and Docker technology to realize high-performance training and scoring on GPU clusters. Topics will include how to analyze the deep learning GPU application's characteristics, such as GPU memory bandwidth, memory capacity, and GPU utilization when run on a GPU cluster with Teye tool; how to handle big data and improve the data reading performance with Lustre; how to optimize the network communication with IB technology; and how to ease deployment and scheduling different deep learning frameworks on a large GPU cluster with Docker.
 
Topics:
Data Center & Cloud Infrastructure, Artificial Intelligence and Deep Learning
Type:
Talk
Event:
GTC Silicon Valley
Year:
2017
Session ID:
S7678
Download:
Share: