Hello guys, i would like to ask is it possible when using the PatchTST for time series prediction, if i have independent features can i include them somewhere without making the model think that they are also target variables (making the model predict multivariate time series)? With Autoformer/Informer it was solved using the past_time_features, because from the docs it says you can only include past_values and target_values for any of the heads available.