Question about the verbose_logging parameter

[BojanKovachki]

I am not sure if this is a bug report or a feature request, but here it goes...

I just tried setting the verbose_logging parameter for the reader to True (while I was doing fine tuning), but it seems like it offers no additional info on what the reader is doing...

There seems to be no difference if that parameter is set to True or False...

I would like to know what the idea of this parameter is and how I can use it properly?

I also think it could be very useful if we can follow the loss during training, so that we know what happens during training... at the moment, I cannot be sure if my model is actually learning or just memorizing (yes, it can be checked with cross validation, but being able to follow the error rate would be much more convenient).

[andrelmfarias]

Hi @BojanKovachki,

This verbose_logging parameter is inspired by the same arg used in the Hugging Face run_squad.py script.

Actually it provides some information about the preprocessing and the training configurations. Indeed displaying the loss would be a nice and important feature to add.

I try will to work on that as soon as I get available time.