NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.54k stars 372 forks source link

emulate_multibleu was removed in NLTK #535

Open cliffwoolley opened 4 years ago

cliffwoolley commented 4 years ago

https://github.com/nltk/nltk/commit/764dc8f4e9756a54f12ca72fb55233f31f1eb014#diff-de624b4549ea37f3f8480d6734bd475b

Based on https://github.com/sgrvinod/a-PyTorch-Tutorial-to-Image-Captioning/issues/19#issuecomment-450594486 it looks like emulate_multibleu is basically the default now. I verified that removing this parameter lets our CI test pass. (It does not, however, deal with backward compatibility to older NLTK; if that's required, I leave it to you all.)