ML Refactoring

[surbhigoel77]

• Move the definition of the train, validation and test datasets, in a separate file (or loaddata.py).
• In Model.py, condense the hidden layers into a loop instead of repeating
• Keep the variables' preprocessing out of NN_pred.py file.
• Restructure from scripts (e.g. use __name__=="main" or notebooks)

NOTE: Given the refactoring, we should add basic tests to ensure that the functioning of the code and inferences remains the same. We also consider adding unit tests as we go along.

[surbhigoel77]

Self Note:

The modules loaddata.py and train.py will have no direct interaction with FTorch. FTorch will call the NN_pred.py module to make predictions using the trained models (in the form of torchscripts).

Hence, refactoring of 'loaddata.py' and train.py can be performed in isolation from coupling.