ChineseBert icon indicating copy to clipboard operation
ChineseBert copied to clipboard

Published 20 hours ago • verified publisher icon benywon

how to run inference on a new question

Open dpny518 opened this issue 6 years ago • 2 comments

This seems to return embedding for input question and answer. How can we pass a random question and generate an answer?

dpny518 avatar Jan 14 '19 04:01 dpny518

As the BERT is mainly focused on generating the embedding for the input sequence, so you can just put the question to the model and then get its embedding.

benywon avatar Jan 17 '19 10:01 benywon

what do you do after you get the embedding? is it just used as sentence/word vectors for downstream tasks?

dpny518 avatar Mar 15 '19 02:03 dpny518