In 2017, Microsoft partnered with Facebook to launch the Open Neural Network Exchange (ONNX) format. It brings interoperability to the artificial intelligence (AI) framework ecosystem, provides definitions of scalable computational diagram models, and definitions of built-in operators and standard data types. ONNX is known to enable models to train in one frame and then move to another frame for reasoning.
(From: Microsoft, via MSPU)
Last year, Microsoft announced that it would open the open source “OPEN Neural Network Switch Run” (ONNX Runtime), a high-performance inference engine for machine learning models on Linux, Windows and Mac.
Today, Microsoft announced the latest developments in the project — the decision to add ONNX to the LF AI Foundation, which is part of the Linux Foundation. Eric Boyd, Vice President of Microsoft Azure AI, wrote:
We are pleased that LF AI will host ONNX and continue to implement an open governance model to encourage community engagement and contribution.
LF AI will provide long-term leadership for ONNX, bringing the community focused on accelerating machine learning adoption and driving the next wave of onNX innovation and adoption.
Microsoft wants ONNX to have a rich ecosystem of frameworks, compilers, runtimes, accelerators, and visualizations under the Linux Foundation.
Dr. Ibrahim Haddad, Executive Director of the LF AI Foundation, said ONNX is not only a corporately recognized norm, it is actively implemented in its products.
Because ONNX is an open format, it is dedicated to developing and supporting multiple frameworks and platforms.
Joining LF AI means that ONNX will continue to evolve along this path, helping to accelerate technology development and connect with the wider open source AI community worldwide.
We are pleased to provide ONNX with a stand-alone occasion and work with the community to enhance its image as a supplier neutral standard.
ONNX will retain its existing OSI-approved open source licenses, governance structures, and established development practices.