Are you doing classification, regression, extraction or something different? How did you evaluate the model before it went into production?
I know a lot about evaluating classification but I am seeing people struggle w( evaluating text genersators; why don’t you look up my profile, send me an email, and we can talk more.