Is stationary data required when fitting an AR model in Skforecast

Question

I'm using Skforecast in a project to estimate an AR(p) model. I'm struggling to find out if the data I input to the model needs to be stationarity (de-trended) or if the package handles this?

# Create and fit model
fa_bl = ForecasterAutoreg(regressor = LinearRegression(), lags = 1)
fa_bl.fit(y = Y_data) # Does Y_data need to be stationary?

The documentation of the fit method simply states:

Training time series.