Adding truth label to results file + Fix flavour retrieval in plotting
Summary
This MR introduces the following changes
- Note: Unit test files changed! No pipeline will success until this is merged!
- Adding the jet truth labels to the results file produced by
evaluate_model.py
(LikeHadronConeExclTruthLablID
) - Removing error prone flavour retrieval from plotting code. This was based on the order of
class_labels
(which needed to be similar to the ones in the preprocess/train config, otherwise this would lead to wrong flavour retrieval). Now it is based on the truth labels which are stored in the results file. - Adding dummy empty
label
values for plotting. If you only want to plot one model without a modellabel
and only the flavour names, this would produce (at the moment) an error. When nolabel
is given for the model, an empty string is added to themodel_labels
list to keep the lists dimensionality correct. - Fixing working point values in the discriminant plots. The x-value was given as label to
puma
which lead to wrong numbers above the vertical lines. - Sorting the working points which are to be plotted in the discriminant plot in decreasing order so they are correctly printed when
same_height_WP
isFalse
.