At this repo https://github.com/sebastianGehrmann/bottom-up-summary the README directs to pull a separate branch https://github.com/sebastianGehrmann/OpenNMT-py/tree/copy_constraint
if we want to connect output of the allennlp with OpenNMT.
Is that still accurate or can we use the main branch. I did a diff of the translate.py but don’t understand the code well enough yet to see if there’s anything really meaningful in that diff.