coremltools.utils.evaluate_regressor requires target parameter only with 'prediction' value

[alkravchenko]

coremltools.utils.evaluate_regressor accepts target parameter only with 'prediction' column name. If i try set any other name for column then i get error "KeyError: 'prediction'".

[TobyRoseman]

Looking at the code, it seem the target parameter is only used for getting predictions from the MLModel. I agree it would make more sense for it also to be the column name in the data.

This would be a simple change to make. However it would also be an API breaking change.

This is also an issue with evaluate_classifier. However it's not an issue with evaluate_classifier_with_probabilities. So things seem a bit inconsistent.