deep neural network training
At its Build developer conference, Microsoft announced the launch of Azure Batch AI Training. While that’s a mouthful, the name pretty much tells you what it is: a new service for batch training deep neural networks on the company’s Azure cloud computing platform. This service is now in private preview. It’s unclear when it’ll become available in public preview. The best way to think about Batch AI Training is probably as a managed AI training service where developers can focus on their models and don’t have to think about operating the underlying infrastructure. While the details are currently somewhat sparse (and we will update this post once we learn more), the overall idea here is to give developers a service to train their machine learning models. Microsoft says that the service will support any framework, so developers will be able to use Google’s TensorFlow, Microsoft’s own Cognitive Toolkit or Caffe, for example. Developers and data scientists will, of course, be able to configure the environment they want to use to train their models (which will surely also be the deciding factor for how much using this service will cost after the private preview). Microsoft notes that developers will be able to run their workloads on multiple CPUs, GPUs and — in the long run — FPGAs.
