OpenApi Workflows for Model and Format Conversion of Machine Learning Frameworks
Darshankumar C. Dalwadi, Vatsalkumar Vipulkumar Shah, Neel Macwan · 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) · 2021
Various machine learning frameworks have their format in which they accept the model. ONNX is an open format. These models are the representation of the neural network with weights after training. A model created for a particular framework has to be converted into a different format as required for use with another framework So, this paper deals with the design of such conversion tools. Our paper mainly deals with the design of conversion tools of 4 popular frameworks of machine learning that is Keras, ONNX, Tensorflow Lite, and Pytorch. Moreover, we have also created the Open API-based workflow model. So, the users can be able to pipeline our project with another language also since the API became language agonistic. We have also created a private API using Flask.