Time series forecasting with holidays

[xavierXV3]

Hi,

I have two questions related to holidays, which have an important impact on the quantity to forecast for some use cases.

First, passing an instance of sktime.transformations.series.holiday.CountryHolidaysTransformer in the list of fe_target_rr passed to setup results in an error:
fe_target_rr = [ CountryHolidaysTransformer("FR")]
exp = TSForecastingExperiment()
exp.setup(data = df_train, fh=28, fe_target_rr=fe_target_rr)
The error occurs when a model with exogeneous variable is instantiated.
The message in the log isn't precise. The errors disapears when I remove the CountryHolidaysTransformer.
Is it possible to solve it?

Second, if a feature related to holidays is provided, and forecast is performed for more than one step ahead, will a model supporting exogeneous variable receives as input this feature (which is available for the future) to forecast for the next steps? Is it possible to do so?

Kind regards