Hi,
Until today (I just tested) my 2 NER models (BERT base and large) have been working well on the HF model hub:
- NER with BERT base: pierreguillou/ner-bert-base-cased-pt-lenerbr · Hugging Face
- NER with BERT large: pierreguillou/ner-bert-large-cased-pt-lenerbr · Hugging Face
Now, when I start a compute
in the widget, I get this error:
Can't load tokenizer using from_pretrained, please update its configuration: missing field
direction at line 1 column 85
See as well this screenshot:
Coming from the new version of transformers?
Another point: I use these 2 models in a Spaces App and there’s no problem:
Strange, no?
cc @lysandre