Throw away value for custom dataloaders

[bresilla]

Summary:

A simple _* throw away value when loading a PyTorch dataset that returns more than data, target

Features:

This simply ignores values of dataset that returns more than data, target in modelwrapper.py
Many times there is a need to create a custom dataloader that for example returns path, name etc...
And the modelwrapper.py expects always two when unloading the dataset. This simle adition
would allow to have a datatset with multiple returns, since it will ignore all of them and use only the
ones that are necessarily (only data, target)

Checklist:

• Your code is documented (To validate this, add your module to tests/documentation_test.py).
• Your code is tested with unit tests.
• You moved your Issue to the PR state.

[Dref360]

Hello!

That's a great idea, do you think we should add this to train_on_dataset and predict_on_dataset as well?

[bresilla]

Yes. Definitely, I just didn't use those functions, so I didn't know to edit...

I just checked now and indeed - if we do for one, we do for all :)
I'll edit the PR and add those changes and resubmit soon.

Cheers.

[Dref360]

LGTM thank you!