Installing transformer Flair or TNER models from Huggingface

[emiltj]

Hi,

I was looking for a way to use a way to use a transformer model from huggingface for NER tagging (more specifically one from flair or one from TNER).

Trying
python -m spacy download flair/ner-english-ontonotes-large
gives

_No compatible package found for 'flair/ner-english-ontonotes-large'
(spaCy v3.5.0)_

Is it possible to use the NER tagging from such a model in spaCy, perhaps using spacy-transformers?
And if not, is it possible to convert the output from Flair or TNER to Doc objects?

[emiltj]

I may have found a workaround, using spaCy wrap:

import spacy
import spacy_wrap

nlp = spacy.blank("da")
config = {"model": {"name": "tner/roberta-large-ontonotes5"}}
nlp.add_pipe("token_classification_transformer", config=config)
doc = nlp("Some text for NER")

[rmitsch]

Hi @emiltj, great you figured it out! spaCy-wrap seems like a reasonable way to do this.