Using Distributed Training in AMLS

An interesting topic is how we can process large-scale datasets to train machine learning and deep learning models. For example, large-scale text-based mining, entity extraction, sentiments, and image or video-based, including image classification, image multiclassification, and object detection, are all very memory intensive and need large compute resources to process, which may take hours or sometimes days and weeks to complete.

In addition, if you have big data that contains business information and want to build machine learning models, then distributed learning can help. This chapter will cover how we can run large-scale models with large datasets. You will see different ways of computing large, distributed models.

There are different ways to distribute compute and data and achieve faster and better performance for large-scale training. Here, we are going to learn about a few techniques.

Data parallelism is widely used when there is...