Reaching out to mentors because of interest in Project 17

[VikashhAdaikalavan]

Hi Dariusz Trawinski, Milosz Zeglarski !
I am Vikashh Adaikalavan currently pursuing my first year of college in Birla Institute of Technology And Science, Pilani,India.
I have learnt about machine learning models and deep learning models and am familiar with the implementation and logic of them. I developed a special interest when I read about this project which makes ML implementation easier and faster and am interested to contribute to this project.

Could I get more clarity on how different types of input data will be handled? For example, if a user wants to train a model on images for an image classification task or on text for an NLP model, how would this be managed within the OpenVINO Model Server? Will the training endpoint support different preprocessing pipelines for various data types?

cc @dtrawins @mzegla

[mzegla]

Hi @VikashhAdaikalavan, sorry for the late response, we must have missed the ping.

I suppose we could limit the scope of the project to one specific data type and see how it goes. Server training endpoint would use KServe API which allows sending raw binary data as well as some metadata, so having access to that, Python logic will be able to handle different kinds of input. Although we would likely limit the scope of I/O types.