What happened to all the grunt work which was supposed to be done before finding the area under the curve and area under ROC? The underlying process is still the same, that is: 

1. Get the labeled data.
2. Divide it into training and test or validation sets.
3. Train the algorithm using the training set.
4. Validate it using the test set.
5. Report its accuracy using the area under the curve/ROC.

In this case, the libraries did the process automatically.