Easy deployment and flexibility
Focus on your workload, not your infrastructure by leaving resource provisioning and management to Batch AI. The service will deploy virtual machines, containers, and connect your shared storage and configure SSK for login. Batch AI Training provides a flexible programming model and SDK so you can easily integrate your own pipeline and workflow. Because Batch AI handles deployment, it’s easy to iterate on your networks and hyper-parameters.
High performance training
Batch AI works with all Microsoft Azure VM families, including the latest NVIDIA GPU’s connected with InfiniBand. This gives you the ability to scale the compute resources to whatever your models and training data require. The same powerful infrastructure Microsoft uses for its AI development is now available to you, on demand.
Supports any framework
Use any AI framework or libraries. Azure Batch AI has deep support for CNTK, TensorFlow, Chainer, and more. Or bring your code in a Docker Container and we’ll handle the rest. Batch AI supports the Azure command line, Jupyter Notebooks, scripting the service using our Python library, and integrating workflows with the REST API and SDK for C#, Java and other languages. You can use the tooling you’re comfortable with.
Related products and services