Apple Launches ‘Create ML’ for Easy Machine Learning Model Training on Macs

Overview

• Apple has announced Create ML, a simple and efficient machine learning tool for Macs
• It is designed using Swift, Apple’s native programming language
• Has already helped an app developer cut down training time significantly; went from a image processing model taking 24 hours to a mere 18 minutes!

Introduction

Machine learning has become an integral part of all product launches – from Google I/O to Apple’s WWDC. All the big organizations are leaning heavily on the power of ML to boost their products and give them a competitive edge. In fact, I can’t remember the last big tech conference that did not feature the use of ML!

Apple has announced Create ML, a GPU-accelerated tool for training machine learning models on your Mac system. It has been built all in Swift, Apple’s native programming language. You can train models to perform tasks like recognizing images, analyzing text to extract meaningful insights, or finding relationships between numbers. The general framework of Create ML is shown below:

You can even use one-click or drag-and-drop interfaces (like Swift Playground) to train models. According to Adam Federighi, Apple’s senior Vice President of software engineering, this means you don’t need to be a machine learning expert to make use of this tool.

Also since the tool is GPU accelerated, it makes things run incredibly fast on Macs. Mr. Federighi used the example of app developer Memrise, which has used Create ML to reduce the training time of it’s model significantly. Earlier, it used to take them 24 hours to train a model using 20,000 images. Create ML has cut down on that time to 48 minutes on a Macbook Pro and a remarkable 18 minutes on an iMac Pro!

The tool also reduces the size of the model hence making things easier of your machine. In the case of Memrise, the model size went from 90 MB right down to 3 MB.

Apple also announced the latest version of it’s Core ML framework, appropriately called Core ML 2. They claim it is 30 times faster than the last iteration thanks to batch prediction. This will also reduce the size of the models by up to 75 percent, thanks to a technique called quantization.

You can check out the documentation for Create ML, which includes examples to help you understand the tool, here.

Our take on this

Create ML is clearly aimed at Apple developers. It has been built for Macs specifically and can only be used with Apple’s Swift programming language. So these restrictions will obviously limit it’s usage for data scientists and developers in general.

You might be wondering how Create ML is different from Core ML. Well, Core ML is all about bringing your own models while Create ML is basically an easy and intuitive way to train custom models. For folks who use Swift, Google recently also launched the Swift for Tensorflow package to ease your transition into the ML world.

Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!

Pranav Dar 10 May 2019

Senior Editor at Analytics Vidhya. Data visualization practitioner who loves reading and delving deeper into the data science and machine learning arts. Always looking for new ways to improve processes using ML and AI.