Social Icons

Tuesday, August 8, 2017

IBM speeds deep learning by using multiple servers

For everyone frustrated by how long it takes to train deep learning models, IBM has some good news: It has unveiled a way to automatically split deep-learning training jobs across multiple physical servers -- not just individual GPUs, but whole systems with their own separate sets of GPUs.

Now the bad news: It's available only in IBM's PowerAI 4.0 software package, which runs exclusively on IBM's own OpenPower hardware systems.

Distributed Deep Learning (DDL) doesn't require developers to learn an entirely new deep learning framework. It repackages several common frameworks for machine learning: TensorFlow, Torch, Caffe, Chainer, and Theano. Deep learning projecs that use those frameworks can then run in parallel across multiple hardware nodes.

IBM claims the speedup gained by scaling across nodes is nearly linear. One benchmark, using the ResNet-101 and ImageNet-22K data sets, needed 16 days to complete on one IBM S822LC server. Spread across 64 such systems, the same benchmark concluded in seven hours, or 58 times faster.

No comments:

Post a Comment

Category Of Mobile Courses

Actualités (644) Adsense (1) Affiliation (1) Algebraic Topology (1) Algorithmic (1) all-news (30) Android (5) Android App (8) Android app without code (4) Android Apps (169) Android Development (4) Android download (2) Android OS (3) AngularJS (1) Automata theory and formal language (5) C programming (5) Category and Functor (8) CMS (3) Computer Glossary (18) Create Mobile App With Ionic Framework (2) CSS (2) CSS-Cascading-Style-Sheets (4) Developpement Java (13) Differential Geometry (1) Django-Python-Framework (8) dropshiping (26) Earn Money by Internet (4) Emplois (23) Framework php (2) Fraud (2) HTML (7) Java For Beginners (10) Javascript (12) Kotlin Programming Language (8) Kotlin For Mobile Android (1) Linux Download (2) Marketing (5) Mobile (3) Mobile Courses (4) Mobile Marketing (4) MoneyGram (1) News (721) Node.js (5) Open Source (1) Photoshop (1) Protect Computer (1) Python (33) Python BeautifulSoup (1) Python For Data Science (2) Python-Books (6) Python-DVD-Training (1) Python-Exercises (140) Python-Framework (1) Python-IDE (1) Python-Modules (1) Python-pdf (2) Python-pyQt (1) Référencement (2) Script PHP (2) Security (6) SEO (1) Snipping Tool: Faq (1) Social Networks (1) Source Code (1) Statistics With SPSS (2) Surveillance Software (1) Travail à domicile (6) Tutoriels php en vidéos (2) Tutoriels-MySql (6) tutoriels-php (19) Utilitaires (1) VPS (1) Web Hosting (1) Webcam (1) Webmarketing (11) Western Union (1) Windows 10 (1) Windows 7 (4) Windows 7 Faq (2) Windows 8 (1) Windows Accessories (1) Windows Download (8) Windows Drivers (1) Windows Fonts (1) Windows Power Shell (2) Windows Registry (2) Windows Security (18) Windows Software (2) Windows Spyware (2) Windows utilities (3) Windows Virus (2) Windows Vista (3) Windows Wireless (1) Windows xp (1) Wordpress (1)
 

Sample text

Sample Text

 
Blogger Templates