Custom training issue: best_model_ranking not outputting for certain ConLL files

Hi, I have successfully trained a custom model using neuralcoref for a set of ConLL files. However, when I add more from another set I get this error:


best_model_ranking (the custom model I used for coreference resolution with neuralcoref) is not present:


Have you encountered this error before? I think it might be because the token distance between coreferences is too long in some of the new ConLL files. Do you have any ideas? Thank you very much.