Use any model with link

[devnev39]

For more engagement and testing add a feature which can load models from a link and run the inference from those models. See if the model can be uploaded with any input and output shape.

EDIT:
Prefer uploading the onnx model and parse it.