Passing base_margin to xgboost model as part of sklearn pipeline in Python

Question

I was trying to pass the base_margin to the xgboost in fit() as part of the sklearn pipeline.

My code would work if you swap base_marg