A step by step guide to starting Deep Learning Virtual Machine (DLVM) on Microsoft Azure

The following steps are meant to help the participants setup the environment for the Deep learning boot camp @Shlomo Kashani & myself are hosting on 5–9/11/2017 with the generous support of Microsoft and Nice.

Hopefully these instructions will be useful for others that are interested in exploring Azure Deep Learning Virtual machine

First open an account on Microsoft azure: portal.azure.com

Sign up/sign in with an Azure account (No credit-card needed for the boot camp participants!)

Go to the azure pass we supplied to you by a separate E-mail message and redeem your pass subscription for the boot camp.

Once your azure pass has been redeemed you’re good to go on and provision the DLVM

Now go to: https://github.com/wbuchwalter/deep-learning-bootcamp-vm

Verify that you’re logged on to azure and click the Deploy to azure button.

you should see the following screen:

Fill in the fields surrounded with red marker and click purchase

Once done you should see the screen below:

First we have to learn how to turn off the machine so that whenever we’re done we can stop it and manage our credits wisely:

If you are working on a windows computer download and install Cygwin

You can use the direct download link (for win 64bit)

Once you’ve got Cygwin — logging into the VM can now be easily done using ssh, just start your machine

(virtual machines -> select your machine -> click start)

Then go to Cygwin and type:

ssh username@your_VM_IP_number

now login with the password that you have set.

After logging into your machine we need to get you the course materials, use:

git clone https://github.com/QuantScientist/Deep-Learning-Boot-Camp.git

we can now run docker through (this is single long line):

sudo nvidia-docker run -it -p 5555:5555 -p 7842:7842 -p 8787:8787 -p 8786:8786 -p 8788:8788 -v ~/Deep-Learning-Boot-Camp:/root/sharedfolder wbuchwalter/quantscientist-pycuda bash

go to sharedfolder:

cd sharedfolder

and run jupyter using:

jupyter notebook –allow-root

now go back to your browser — open the panel of the virtual machine you’re using and copy its ip

Insert this IP address to a new window followed by the notebook port

( i.e. 1.1.1.1:7842 for ip 1.1.1.1)

Password to the notebook is on the course docker folder (not shown in the tutorial on purpose)

Open the notebook “01 PyTorch GPU support test” and run the first 2 cells to verify that your gpu is being used.

That’s it your done !

p.s. don’t forget to turn off your image once done to save your credits for the course 😊

looking forward to see you in the course.

Shlomo & Nati

feel free to respond with updates and ideas how to improve this set of instructions for the benefit of future users

Should you have any problems during installation please start by filing a support issue on azure.

Note : while we tried to white list all of the participants’ accounts there might be issues there, so please try to have everything setup until 1/11/17 so we’ll have the time to handle issues if some occur

They’re set to help you with installation issues.

Use the question mark sign on the right side of your azure portal screen

Then open a new support request: