NANs for trackless
For trackless the variable SplitProbability1/2 are filled with NANs in case isSCT is false in the TDD link
This crashes with Umami, which cannot deal with Nans in the calculation of the mean/std for the scaling
umami-hep can deal with this since the calculationof mean/std is a little bit different.
Is it possible to get a solution for Umami aswell?
Dan suggested a solution like non_nan_values = hits[hits['valid] & hits['isSct']]