How to get Turi Create running on Windows

Audience: Machine learning beginners, command line beginners, windows users, Python 3, Anaconda

I was following that machine learning fundamentals online course and already failed during the set up of the required libraries, which drove me absolutely insane. Let’s save us some frustration. This short tutorial is meant to help you get through the first steps a little easier

Why do we want to use Turi Create?

There are many machine learning libraries in Python, most of which cannot be easily scaled to large datasets. Turi Create enables us to create a wide range of ML models, without having to install additional packages. It is easy to use, contains useful plotting options, and is flexible regarding the data you might want to use.

Turi Create can accomplish many ML tasks, most without the need of additional libraries.

Requirements:

• Python 2.7, 3.5, 3.6
• x86_64 architecture
• At least 4 GB of RAM

Supported Platform:

• macOS 10.12+
• Linux (with glibc 2.10+)
• Windows 10 (WSL)

And this is where my problems started. As I often do, I only read half of the manual and started typing away… and faced with error message after error message.

Comment: You will not be able to simply install/run Turi Create from the Windows 10 Platform. You will have to install a Windows Subsystem for Linux (WSL) first.

Install WSL

In order to install any WSL the “Windows Substystem for Linux optional feature” must be enabled.

• Open PowerShell as Admin an run following command:

• Restart your computer when prompted
• Install the linux distribution of your choice. In this example we are proceeding with Ubuntu 16.04

Comment: Ubuntu 16.04 is not the latest version and will have to be updated later. If you find the latest version, install it directly and skip the “update” step.

• open your command line as administrator and use the following code

The result should look somewhat like this:

Now we need to install our distribution using PowerShell. Run the following command, while “app_name” is the name of your distro.appx file

Comment: This can also be done manually, which will not be described in detail. It is basically looking for the distro.appx file and clicking install.

If you accidentally downloaded an outdated version (like I did) you will be prompted to update (which I did)

Your distro command terminal is prompted automatically and you will be asked to enter a new Unix username: Follow the regular expression

Rules:

• Begin with lower case letter or underscore
• followed by lower case letters, digits, underscores, dashes
• regex: i. [a-z_][a-z0–9_-]*[$]?

Comment: I tried various names — which all have been rejected. In the end I chose a simply name, all lowercase, no special letters.

Install pip

turicreate is installed using pip. If you have installed Ubuntu you can use this tutorial. Make sure you are using sudo privileges.

• you will be prompted to approve the use of additional disk space — press “Y”
• You will then be asked to restart. I chose <Yes>. It is up to you though

• once installation is complete, verify pip3 was installed
• although the version number may vary, it should look somewhat like this:

• once Turi Create has been installed you will get the following error (unless you have already installed Tensorflow)

Install Tensorflow using pip

• open your Ubuntu (or distro of preference) console and use the following command:

That is pretty much it. I also installed Sframe to handle big datasets, but that is a matter of taste. The community has already started providing intuitive tutorials for further use and inspirational articles. You can now start playing with Turi Create and Machine Learning. Have fun!