How to fix labels on an AutoTrained model

I am trying to do some multi-classification, and used the AutoNLP function to help out. I used my own data which, unfortunately, wasn’t as clean as I thought. Now, the specific issue is that there are duplicate labels, with different casing. One is “growth” and the other “Growth”. Then, when testing, the probability is split between them.

Is there a simple and elegant fix? I’m kinda hoping that just deleting the duplicate label on the config.json file would work, but don’t know.

Thanks for the help.


There are some model support uncased pre-trained model.
i think your train model can get robust at cased-problem.