-
Notifications
You must be signed in to change notification settings - Fork 57
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Issue with AllenNLP integration causes predict to not work (ArrayField.empty_field) #31
Comments
Not sure about allennlp 0.9.0, but for later versions these issues can be circumvented with downgrading the overrides package, i.e: pip3 install overrides==4.0.0 |
Update: tried it now with allennlp 0.9.0, and overrides 3.0.0 seems to work. I get a NaN loss error though, which seems to be stemming from the dependency labels. I now circumvented it by adding "if not torch.isnan(pred_output['loss']):" in line 133 of udify_model.py, but this is obviously suboptimal for performance. Edit: changing mask to float_mask in line 279 of udify/model/dependency_decoder.py seems to resolve the error (not sure if this fix is correct?) |
downgrading overrides works: perhaps this version should be explicit in the requirements.txt file? |
I faced this issue too, and the quoted solution worked for me. |
When I try to run a clean checkout of UDify, I get the following error:
The text was updated successfully, but these errors were encountered: