Cross Validation & Feature Scaling?

[mertbakir]

When we are going to do feature scaling or any other transformation, we may want do it at every fold of cross validation to prevent data leak from the future. In that case all I need to do is:

from sklearn.pipeline import make_pipeline
from sklearn.compose import ColumnTransformer

   models = [
      make_pipeline(ColumnTransformer(...), some_regressor), 
       ...
     ]
   mlf = MLForecast(
            models=models,
            freq=1,
            lags=[1, 2, ...],
            lag_transforms={ ... },
        )
   mlf.cross_validate()

First the preprocessing runs, computes lags and transformation if there are any in the MLForecast object.
Then at each fold of the cross_validate() the model is fitted, in that case the pipeline so ColumnTransformer runs before fit. Is this correct ?