Getting started

In this guide we are going to setup a development environment on Azure. However, you can follow along and perform the setup on any cloud services provider (AWS, GCP).

Provisioning virtual machine

• Setup a free account and navigate to https://portal.azure.com
• Click the “+ New” and type “Ubuntu Server 16”.
• Click next without changing default options.
• Configure basic settings.

• Choose virtual machine size (preferably atleast 16 GB Memory, 4 cores and 100GB disk space). If you have the $$ and would like to setup a VM with GPU’s then you can submit a ticket to Azure support to enable GPU VM’s (N-series) for your account.

• Click next through Step 3 and Step 4.
• Wait for deployments to complete.

At this point you should have a provisioned VM.

Logging in to VM

• Click on the “Resource groups” in left navigation menu and select your resource group “ml-group”.
• Find the “Public IP Address” for your server and use that to SSH along with the username and password you chose during provisioning.

Installing packages

• Strongly recommend using virtualenv’s for each ML project you will be working on. Follow this guide.
• Once you have a virtualenv you can install packages like numpy, scipy with a simple “pip install numpy” or use a requirements.txt from another project.
• Note that if you want to install Tensorflow with GPU support you will have to follow instructions here. Definitely verify that your CUDA toolkit installation is working correctly before proceeding to install TF. Installation verification can be performed by running CUDA samples.

Once you have installed packages you are ready to start working on your first ML project.