/ Machine Learning

Released Today: Parris, The Automated ML Training Tool.

Today marks a momentous occasion - today is the first day of 2018! Here's hoping your celebratory activities were in equal parts memorable and enjoyable. Other than the new year arriving today, today is special for another reason - the launch of the Parris tool for machine learning training automation. TL;DR: this makes the training of algorithms easier on you.

But, pertaining to the title of the post, the Parris automated ML training tool is out today in its first iteration on Github. It's an open source project for which I'm hoping others find use in, and within which to find opportunities to contribute.

What does it do?

The overview of what this does, from the README:

Parris is a tool for automating the training of machine learning algorithms. If you're the kind of person that works on ML algorithms and spends too much time setting up a server to run it on, having to log into it to monitor its progress, etc., then you will find this tool helpful. No need to SSH into instances to get your training jobs done!

With more technical details from the Getting Started guide:

Parris is a tool that:

  • creates a Lambda function
  • launches a CloudFormation stack when the Lambda function is invoked
  • runs a UserData script on the stack's EC2 instance when it is first launched, that kicks off your training job
  • stops the EC2 instance when it has finished.

The primary purpose of this tool is to reduce the amount of repetitive setup it takes for you to train your machine learning algorithms, and in doing so save some cost by utilizing your server's compute hours more effectively (server starts the training jobh as soon as it's launched, and stops when you've configured it to stop).

How do I use this?

Consult the Getting Started guide! It's very detailed and walks you through a full setup, all the way to launching a machine learning training job and terminating it. Be warned that you'll find it easiest if you have an AWS account already set up.

This tool is missing something. Can I contribute to this?

Absolutely! Feel free to open an Issue to track the missing item while I and/or others work on it, or open a new Pull Request if you've already coded your addition.

Hoping to hear that this tool has helped you!