Hi all... if anyone could provide a solution for this: I instantiate the following model:

models = [lgb.LGBMRegressor(random_state=0, **params, objective=pseudo_huber_loss, verbose=0)]

but then during .fit i setup max_horizon = 6, which will give me a model for each horizon forecasted: Since I am able to access the models info, by going into

model.models_

can I also give different params for each separate model during .fit?

Hey. It's not currently possible, in this case you could leverage the

preprocess

method to generate the training set with all the targets, train each model separately and assign it to the

models_

attribute (similar to this guide but each key in the dict holds a list of models)

Ok… understood. But another question… could you explain, when we feed multiple unique_id’s to a training, on an MLForecast framework, how do they correlated with each other… if they do

All of the data is provided to train a single modeln it into a categorical feature and provide it through

static_features

in `fit`/`preprocess`