Feature importance

Added a notebook to illustrate the use of feature importance. And also fixed a little bug related to the calling of the callbacks.

Makint it consistent with master so I don't have to rebase or anything similar. Also fixing bugs related to the other forms of training where I did not fix the restore weights functionality

Fix #175 early stopping and model checkpoint restoring weights.

Fix #175 early stopping and model checkpoint restoring weights. Also fix #174 the finetunner single batch run

revert

Fix: EarlyStopping does not store and restore the model #175

manual trigger on fork PR

Get fork PR tests running

Unit tested feature importance and made it compatible to the #171 fix. I just want to add a notebook and rready to merge/publish

Unit tested feature importance and made it compatible to the #171 fix. I just want to add a notebook and rready to merge/publish

Default vision models

Unit tested the feature importance functionality for all models but Tabnet. Only TabNet left to do and then ready to publish

added oversample_mul to the dataloader_params to fix issue #153

Added a class to compute feature importance. I need to extend it to an explain capability and test the hell out of it

Cont cols quantization

fixed licenses