Stanza throws "KeyError: 'feat_dropout'"

user_7485803 0
Viewed 324

I am using the CLTK library to do NLP with ancient Greek. I recently upgraded Python from 3.8.9 to 3.9.9 and since then, whenever I try to do even the simplest things (e.g. the quickstart code), I get the error KeyError: 'feat_dropout'. The stack trace tracks it to \site-packages\stanza\models\tokenization\trainer.py, line 98. I've seen one instance of the same error and tried the solution there, but got nowhere. I would appreciate any help.

stanford-nlp python
1 Answers

CLTK currently (late 2021) requires models and code for Stanza version 1.2.3. There was a change in the Stanza model format whereby certain models from version 1.2.3 are no longer valid in version 1.3. CLTK has not yet updated those models to be compatible with 1.3, so you will need to either downgrade Stanza to 1.2.3 or encourage CLTK to upgrade their models.

For more information, see: https://github.com/stanfordnlp/stanza/issues/839 .

user_2350190
Related