Microsoftand Amazon have jointly introduced a newdeep learning library, or interface, called Gluon which is a programming API that simplifies the process for creating deep learning models with compromising training speed.
"We believe bringing AI advances to all developers, on any platform, using any language, with an open AI ecosystem, will help ensure AI is more accessible and valuable to all," wrote Eric Boyd, CVP AI Data and Infrastructure, Microsoft, in a statement.
Gluon will provide an API that gives developers the choice of interchangeably running multiple deep learning libraries, according to Microsoft. It provides an interface for building neural networks and can be used with either Apache MXNet or Microsoft Cognitive Toolkit, and will be supported in all Azure services, tools and infrastructure. Gluon will reportedly make it easier for developers to learn, define, debug, and then iterate or maintain deep neural networks, allowing developers to build and train their networks quickly. Key highlights of the new API are identified here.
Boyd suggested in the Microsoft announcement that the company expects Gluon to become prevalent, as users will realize the usability benefits without sacrificing performance.
"Gluon builds on the powerful training and inference engines in MXNet or Cognitive Toolkit. This means that neural networks built in Gluon can take advantage of MXNet’s or Cognitive Toolkit’s distributed training," he said.
He continued, "Thus, a single Gluon training job can be linearly scaled to 500 GPUs or more–dramatically reducing training time. Inference is also highly optimized–allowing models to run on lower performance, lower cost, and more power-efficient hardware."
Additionally, with Gluon, developers will reportedly be able to deliver new artificial intelligence innovations faster by using a higher-level programming model and the tools and platforms they are most comfortable with, suggests Boyd.
"This, combined with the Open Neural Network Exchange (ONNX) announcement which enables you to create and save AI models using a standard open format, is another part of creating an open AI ecosystem. We look forward to the amazing AI experiences you will create."
The Gluon interface is available now on GitHub for use with MXNet (find this here), while support for an upcoming release of Cognitive Toolkit is actively being worked on.