When I am testing for my sample monolith code by converting it to graph, it starts training and shows loss also. Though result is also displayed in the end, but is training necessary every time I feed a new application to this approach?

[Shalini0403]

No description provided.

[utkd]

Yes, if the monolith application changes, the underlying graph representation will also change. So the model will need to be retrained, since the nodes, edges and features are going to be different.