Training configuration about fairface

[zoltanfarkasgis]

The above list is very comprehensive and the correct one that is mandatory for training.
Regarding the presented labels one would add the previously raised question re the meaning of the "service_test" field.

The info in the paper about training is simply not enough to reconstruct what the authors did in terms of training their models.

Interestingly, the authors seem to have kind of a selective hearing: they do not respond to this question (and similar ones raised in the same context), while they are active (after this one was raised) and respond to many other questions.

This lack of detail undermines the value of the paper and questions how the fairness of the dataset was assessed (and whether that fairness is validated).

It would be kind from the authors to answer these points.