NLPf tries to train on not exisiting model
If a model, e.g. for named entities
x.org.dkpro.core.opennlp.ner-en-someModel, is not created while training, NLPf still tries to test against the non-existing model in the test stage.
This currently appears because we have configured named entities in the pom with less matches for the training algorithm.
NLPf should avoid testing against non-existing model. NLPf should only warn the user that a model could not be created and filter the model in the test stage.