IndexError in model selection for tube data

[IngoScholtes]

The model selection fails with an IndexError for the London Tube data set.

Minimal code to reproduce the error:

import pathpyG as pp

paths_tube = pp.PathData.from_ngram('../data/tube_paths_train.ngram', sep=',', weight=True)
m = pp.MultiOrderModel.from_PathData(paths_tube, max_order=2)
m.estimate_order(paths_tube, max_order=2, significance_threshold=0.01)

I suspect that this has to do with the use of append_walks in the from_ngram function, which already concatenates the Data objects in PathData. The model selection code seems to assume that all paths are stored in individual data objects.