How to fine tune the model on GigaWord dataset with shorter articles and summaries? I tried to use it as such, but gives unsatisfactory results.
nlpyang
