Facebook and Microsoft have created the Open Neural Network Exchange format, which provides a means for creating deep learning models that can be moved to different frameworks. Here's how Facebook describes the value proposition, in a blog post:
When developing learning models, engineers and researchers have many AI frameworks to choose from. At the outset of a project, developers have to choose features and commit to a framework. Many times, the features chosen when experimenting during research and development are different than the features desired for shipping to production. Many organizations are left without a good way to bridge the gap between these operating modes and have resorted to a range of creative workarounds to cope, such as requiring researchers work in the production system or translating models by hand.
We developed ONNX together with Microsoft to bridge this gap and to empower AI developers to choose the framework that fits the current stage of their project and easily switch between frameworks as the project evolves.
The Caffe2, PyTorch and Cognitive Toolkit frameworks will begin supporting ONNX this month. Overall, ONNX should help AI researchers innovate faster.
POV: Engineers and researchers do have many AI frameworks to choose from, but the frameworks that will prevail will be those that enjoy broad adoption, active use and, as a result, innovative contributions from their communities, notes Constellation VP and principal analyst Doug Henschen.
"Engineers and researchers will flock to frameworks that are powerful yet easy to use, and ONNX is clearly an effort to make machine learning work easier by getting laborious, non-value-add grunt work out of the way," he adds. "Google’s TensorFlow framework has been getting a lot of attention and has attracted a large community, but this ONNX move could well boost interest in using Cognitive Toolkit, Caffe2, PyTorch and other frameworks that may join the exchange."