Quick Start: Azure DLVM
A particularly attractive and simple way to get started with exploring the mlhub functionality is to fire up a Ubuntu Data Science Virtual Machine Deep Learning version (DLVM) on Azure. You can get free credit (USD200) from Microsoft to trial the DSVM.
All of the models are currently tested against the DLVM and should work out of the box. Using this virtual machine will save a lot of time compared with setting up your own machine with the required dependencies, which of course you can do if you wish as all the dependencies are open source.
To set up the virtual machine, with an Azure subscription log in to the portal and add a new Deep Learning Virtual Machine for Linux (Ubuntu). You need to provide a name (for the virtual machine), a user name and a password, and then create a new resource group and give it a name, and choose a location (westus2). Go with all the defaults for almost everything else, except choose a size to suit the budget. Note that you are only charged whilst the machine is fired up so USD90 per month, for example, is no where near what you will spend if you only fire up the server occasionally.
Once the DLVM is set up go to its Overview page in the portal and click on DNS name. Configure and provide a name by which to refer to the server publicly (e.g., my01.westus2.cloudapp.azure.com).
We now have a server ready to showcase the pre-built Machine Learning models. There are several options to connect to the server but a recommended one is to use X2Go which can run on Linux, Windows, and Mac. Install it and point it to your server (e.g., my01.westus2.cloudapp.azure.com) in the setup.
Once you connect to the DLVM you can close the Firefox window that pops up. Click on the terminal icon down the bottom, and you are ready to go. I suggest immediately replacing the default terminal with a gnome-terminal:
$ sudo apt-get install gnome-terminal
Then quit the terminal and click the icon again and we are ready to start with mlhub:
$ pip install mlhub $ ml $ ml configure $ ml available