Evaluation of the trainings

I have a sample and want to check the performance of the taggers for which the values are already stored in the files

If you want to just check the performance of one or more taggers which are already present in the .h5 files, you can follow the steps explained here.

I have a new sample and I want to evaluate the performance of my trained tagger on that

To do that, you need to follow these steps:

  • Add a new entry to test_files in your train config file with the path to the sample and the cuts you want to apply (Wildcards are supported).
  • Re-run the evaluate_model.py script (see here for explanation how to do that).
  • In your results files is now a new entry with the name of the entry you set in test_files.
  • You should be able to plot the results now using the plotting_umami.py scripts (see here for explanation how to do that) or your personal PUMA scripts.
Back to top