Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Azure and deep learning

93 views

Published on

Slides for D Giard lecture to UIUC IE534 class
Jan 30, 2018

Published in: Technology
  • Be the first to comment

  • Be the first to like this

Azure and deep learning

  1. 1. Azure and Deep Learning
  2. 2. • David Giard • Senior Microsoft Technical Evangelist • dgiard@Microsoft.com • davidgiard.com • @davidgiard
  3. 3. Cloud Computing Host some or all of your data or application on a third-party server in a highly-scalable, highly-reliable way
  4. 4. Cloud Service IAAS Infrastructure as a Service PAAS Platform as a Service SAAS Software as a Service
  5. 5. Virtualization Server1.vhd
  6. 6. Virtualization Server1.vhd Server2.vhd Server3.vhd Server4.vhd
  7. 7. Cloud Services IaaS Managedbyvendor Storage Servers Networking O/S Middleware Virtualization Data Applications Runtime Youscale,make resilient&manage PaaS Scale,resilienceand managementbyvendor Youmanage Storage Servers Networking O/S Middleware Virtualization Applications Runtime Data On Premises Youscale,makeresilientandmanage Storage Servers Networking O/S Middleware Virtualization Data Applications Runtime SaaS Storage Servers Networking O/S Middleware Virtualization Applications Runtime Data Scale,resilienceand managementbyvendor More Control More Work Less Control Less Work
  8. 8. Reliability
  9. 9. Reliability Data or ApplicationData or ApplicationData or Application
  10. 10. Elasticity
  11. 11. Elasticity 0 1 2 3 4 5 6 7 8 9 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
  12. 12. Elasticity 0 1 2 3 4 5 6 7 8 9 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
  13. 13. Elasticity 0 1 2 3 4 5 6 7 8 9 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
  14. 14. Elasticity 0 1 2 3 4 5 6 7 8 9 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
  15. 15. Elasticity 0 1 2 3 4 5 6 7 8 9 Mon Tue Wed Thu Fri Sat Sun Mon Tue Wed Thu Fri
  16. 16. Elasticity 0 1 2 3 4 5 6 7 8 9 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00
  17. 17. Elasticity 0 1 2 3 4 5 6 7 8 9 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00
  18. 18. Azure
  19. 19. Portal
  20. 20. Creating a VM
  21. 21. Scripting • PowerShell • CLI
  22. 22. ARM Template
  23. 23. PowerShell # Login to Azure Login-AzureRmAccount # Create a Resource Group New-AzureRmResourceGroup -Name dgtestrg -Location "South Central US" # Deploy from an ARM template New-AzureRmResourceGroupDeployment ` -ResourceGroupName dgtestrg ` -TemplateFile template.json ` -TemplateParameterFile .parameters.json
  24. 24. CLI # Log into Azure az login -u <username> -p <password> # Create a resource group az group create --name dgtestclirg --location southcentralus # Deploy from ARM template az group deployment create --name dgtestclivm --resource-group dgtestclirg --template-file template.json --parameters parameters.json
  25. 25. Data Science Virtual Machine for Linux (Ubuntu) • TensorFlow • Microsoft Cognitive Toolkit • MXNet • Caffe • Caffe2 • Chainer • NVIDIA DIGITS • Deep Water • Keras • Theano • Torch • PyTorch
  26. 26. Data Science Virtual Machine for Linux (Ubuntu) • Microsoft R Server 9.2.1 with Microsoft R Open 3.4.1, MicrosoftML package with machine learning algorithms, RevoScaleR and revoscalepy for distributed and remote computing, and R and Python Operationalization • Anaconda Python 2.7 and 3.5 • JupyterHub with sample notebooks • Spark local 2.2.0 with PySpark and SparkR Jupyter kernels • Single node local Hadoop • Azure command-line interface • Visual Studio Code, IntelliJ IDEA, PyCharm, and Atom • H2O, Deep Water, and Sparkling Water • Julia • Vowpal Wabbit for online learning • xgboost for gradient boosting • SQL Server 2017 • Intel Math Kernel Library
  27. 27. Connect • bash / ssh • PuTTY (https://www.putty.org) • x2Go (https://wiki.x2go.org/doku.php)
  28. 28. GPU • NC-Series Azure VMs • Not available in every region (https://azure.microsoft.com/en-us/regions/services/) • Select VM Image with GPU driver installed (e.g., Data Science VM) • Select HDD; Not SDD • Need to be “activated” for some accounts. Open a Quota support ticket.
  29. 29. Recommendations • Script resource management • Shut down VMs when not in use • Set Auto-Shutdown time
  30. 30. Demo
  31. 31. Links • From https://blogs.msdn.microsoft.com/sarahsays/2018/01/04/whirlwind- tour-of-recurrent-neural-networks-prerequisites/: • Step 6: git clone https://github.com/jcjohnson/torch-rnn.git and torch-hdf5 stuff • Follow https://blogs.msdn.microsoft.com/sarahsays/2018/01/04/a- whirlwind-tour-of-recurrent-neural-networks-the-hard-way/ • Intro to Deep Learning VM: https://docs.microsoft.com/en- us/azure/machine-learning/data-science-virtual-machine/deep-learning- dsvm-overview • Create a Deep Learning VM: https://docs.microsoft.com/en- us/azure/machine-learning/data-science-virtual-machine/provision-deep- learning-dsvm
  32. 32. Links • This slide deck: https://aka.ms/ie534slides • Imagine Cup: https://imagine.microsoft.com • My blog: http://davidgiard.com

×