All Times EDT
Virtual
Open Neural Network Exchange (ONNX): Open Source Projects and Applications (308528)
*Svetlana Levitan, .Keywords: machine learning, model deployment, ONNX
Machine Learning (ML) model exchange and deployment has long been a bottleneck in the traditional ML lifecycle implementation for an enterprise. Predictive Model Markup Language (PMML) and Portable Format for Analytics (PFA) developed by the Data Mining Group (DMG) helped to alleviate this problem for traditional ML models and pipelines, but did not address deep learning. In 2017 Microsoft and Facebook open-sourced "Open Neural Network Exchange (ONNX)", a new open format for both deep learning and traditional ML. Its popularity has grown tremendously since then. In this talk we will discuss this open standard and related open source projects, as well as their applications. We will explain how to get started with the standard and how to contribute to the open source projects.