This post is authored by Paul Shealy, Senior Software Engineer at Microsoft.
We are pleased to announce the integration of the Microsoft Data Science Virtual Machine (DSVM) with the Batch AI service in Azure.
DSVM is a family of popular VM images published on Azure with a broad choice of machine learning, AI and data science tools. All tools are pre-configured giving you a ready-to-use, on-demand, elastic environment in the cloud to help you perform data analytics and AI development productively. You focus less on IT administrative tasks and more on your data science with the DSVM.
Microsoft’s Batch AI Service is a new service that helps you train and test machine learning models, including deep learning models, on pools of GPU machines. It simplifies the process of creating a cluster of machines and training on it using many popular deep learning frameworks like TensorFlow, Microsoft Cognitive Toolkit, and others. Batch AI also lets you run parameter sweeps in parallel. Managing data is an integral part of deep learning, and Batch AI includes native support for file shares and NFS servers.
The Ubuntu DSVM is supported as a native VM image in Batch AI. The Ubuntu DSVM comes with many deep learning frameworks, GPU drivers, CUDA, and cuDNN pre-installed, so it is easy to get started with a deep learning project. Data scientists can develop an initial version of a model on a single DSVM, using a smaller dataset, then easily scale out across many DSVMs and larger datasets in Batch AI when ready. Using the same DVM image in Batch AI minimizes the setup time required for your cluster’s VMs and reduces incompatibilities between Batch AI and your development environment. Batch AI handles the details of setting up your cluster, can automatically scale up and down based on demand, and supports low-priority VMs for additional cost savings. Read more at Batch AI Overview and see their recipes for examples with TensorFlow, Microsoft Cognitive Toolkit, Keras, Chainer, and others.
We invite you to try Batch AI and the Ubuntu DSVM today. You can engage with the development team and the growing DSVM user community on our forums.