The One Minute AI - Azure Batch AI

In this article, you will learn about Azure Batch AI.

Series introduction

Welcome to a new series of short articles I am presenting about Artificial Intelligence specifically in the Azure AI stack. The objective is that you will learn about an Azure-based AI service in no more than one minute and thus quickly get familiar with the entire stack over a short period of time. These are going short, easily digestible articles so let's get started!

What is Azure Batch AI?

Developing AI algorithms is an iterative process that requires large amounts of computer resources, especially if you’re working with large data sets. In order to develop AI algorithms efficiently, you need multiple CPUs per model, to be able to run experiments in parallel and shared storage. Developing AI at scale, however, requires infrastructure actions such as installing software and containers, scaling resources, queuing work, provisioning clusters of VMs and integrating with tools and workflows. Developing and managing this infrastructure can become very time-consuming.
Azure Batch AI is a cloud service which helps AI researchers and data scientists train and test AI models and machine learning at scale in Azure by dealing with resource provisioning and management, making it easy for you to iterate on your networks. Batch AI enables you to submit parallel jobs to a cluster of VMs, supports custom storage solutions and helps with scheduling jobs and handling failure during long-running jobs. You can use any AI framework or libraries or import code in a Docker Container.
Batch AIs capabilities include,
  • Deploying VMs and containers
  • Automatic or manual scaling of VM clusters
  • Connecting shared storage
  • Providing job status and restarting if the VM fails
  • Configuring SSH communication between VMs and for remote access
  • Support for Deep Learning or machine learning framework and optimized for Microsoft Cognitive Toolkit, Chainer and TensorFlow
  • Azure command-line interface (CLI), SDKs for Python, Jupyter Notebooks, C#, and Java, monitoring in the Azure Portal, and integration with Microsoft AI tools
With Azure Batch AI all you have to do is describe the compute resources, the jobs you want to run and where to store the model inputs and outputs, then Batch AI does the rest.
Find out more,