Hi, I am training a model (fcst) with

max_horizon=24

. I want to evaluate the training error and I see that I can access the in-sample predictions via

fcst.forecast_fitted_values()

or

fcst.fcst_fitted_values_

. But what does the "prediction column" consist of as I (for most of the rows) have 24 predictions? An average of the horizons, or horizon 1? My ultimate objective is to be able to get the train and eval error pr horizon (the latter I know how to do) 🙂

There should be an

h

column which indicates which model the predictions correspond to

Ah, yeah I made a silly mistake. Thanks 🙂