On-Boarding ONNX and PFA Model user guide

1: What is ONNX and PFA ?

ONNX : (Open Neural Network eXchange) is a library available in some Deep Learning Framework that allows to import and export Deep Learning models from different AI framework. If you export a model under the ONNX format you will be able to import it in and use it in many others Deep Learning framework. Please have a look at https://github.com/onnx/tutorials to know more.

PFA : (Portable format analytics) is a language developped by the Data mining Group to help datascientist to deploy their model in production. Please have a look at http://dmg.org/pfa/docs/motivation/ to know more.

2: How to onboard ONNX or PFA model ?

Onboard an ONNX or PFA model is really simple you just have to use the on-boarding web page to upload and then onboard the model. Acumos portal will automatically detect the format of the model (ONNX or PFA) and you only have to fill the name of the model.

You can on-board your model with a license profile, you just have to browse your license profile file or drag and drop it. The license profile file name must be : license.json. If the license file extension is not ‘json’ the license profile on-boarding will not be possible and if the name is not ‘license’ Acumos will rename your license profile file as license.json and you will see your license profile file named as license.json in the artifacts table. If you upload a new version of your license profile after on-boarding, a number revision will be added to the name of your license file like : “license-2.json”. To help user create the license profile file expected by Acumos a license profile editor user guide is available here : License profile editor user guide

The process of on-boarding, for ONNX and PFA, in Boreas is reduced to create a solution Id and upload the model. There are no micro-service, nor tosca file, nor metadata file, nor protobuf file created.