Unseen values of feat_static_cat in prediction

[StanislasGuinel]

Hi GluonTS team !
I was wondering what happens when predicting time series with values of feat_static_cat that were not seen during training ?

For instance, a DeepAR model trains on time series with a single feat_static_cat category with values between 0 and 10, now what happens when the model needs to predict a time series with a 11 value as feat_static_cat ?

From what I tested, it seems that all unseen values get the same prediction.