How does it apply to English data sets

I constructed the English dataset according to the format of the Chinese dataset and changed the dataset.py and vocab.txt, but a lot of [MASK] flags were printed in the output during the translation. Could you give me some suggestions?