Feature Request: Fill Mask more than 1 token

[p-christ]

At the moment you can use hugging face's mask filling pipeline to predict 1 masked token in a sentence using the below:

!pip install -q transformers
from __future__ import print_function
import ipywidgets as widgets
from transformers import pipeline

nlp_fill = pipeline('fill-mask')
nlp_fill("I am going to guess <mask> in this sentence")

The request is that you also add the ability to predict N masked tokens rather than only 1 masked token. So that for example if the sentence is "I am going to make <mask> <mask> for breakfast" then the model might predict "fried eggs" for the 2 masked tokens

[julien-c]

Duplicate of #3609

Not currently supported but we welcome a PR