Example code does not work with new versions of transformers

#8
by ak1020 - opened

Example should be updated to this to work with current transformers-library (also torch needs to be installed in addition to transformers):

from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline

model_checkpoint = "Kansallisarkisto/finbert-ner"
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)
model = AutoModelForTokenClassification.from_pretrained(model_checkpoint)

token_classifier = pipeline(
"token-classification",
model=model,
tokenizer=tokenizer,
aggregation_strategy="simple",
)

predictions = token_classifier("Helsingistä tuli Suomen suuriruhtinaskunnan pääkaupunki vuonna 1812.")
print(predictions)

National Archives of Finland org

Hi!

Thanks for notifying us on this problem. I tested the old code example with Transformers version 5.6.2, and it seems to be still working without errors. However I updated the example code in the repo to use AutoTokenizer and AutoModelForTokenClassification, as this also allows more customized use cases.

Best regards,
National Archives AI team

Sign up or log in to comment