Spacy 2.3.3 seems to break my prodigy


I'm working on deploying a prodigy dashboard on kubernetes. While developing, suddenly my prodigy was not working at all.

On checking, I saw that prodigy 1.10.4 and spacy 2.3.3 does not work for me.

File "", line 1, in
File "/home/bram/.virtualenvs/x-mas-hackathon/lib/python3.6/site-packages/prodigy/", line 7, in
from . import recipes
File "/home/bram/.virtualenvs/x-mas-hackathon/lib/python3.6/site-packages/prodigy/recipes/", line 4, in
from ..deprecated import recipes # noqa
File "/home/bram/.virtualenvs/x-mas-hackathon/lib/python3.6/site-packages/prodigy/deprecated/", line 15, in
from ..models.matcher import PatternMatcher
File "/home/bram/.virtualenvs/x-mas-hackathon/lib/python3.6/site-packages/prodigy/models/", line 1, in
from .ner import EntityRecognizer, merge_spans # noqa: F401
File "cython_src/prodigy/models/ner.pyx", line 11, in init prodigy.models.ner
ImportError: cannot import name cleanup_beam

It's not really a big issue for me, as I've now pointed to the previous version and things work fine. Just wanted to give a heads up and see if you guys experience the same thing.


I saw there was already a thread with the same issue:

Sorry for the duplicate!

Thanks for the report and sorry about that! This was caused by Prodigy using an internal method that was changed in that spaCy release. spaCy v2.3.4 should fix this again: