This tokenizer seems to be the same as what was used to train the model.

elinas changed pull request status to merged

Sign up or log in to comment