NER Model number of prediction didn't match the number of token in entry

[yuane4]

Hello,

I encounter a weird problem recently, whatever the test dataset I put in entry of the NER Model in Camenbert, I never get the same number of prediction than the number of token I put in entry.

For exemple if I have 604 673 tokens in my test file I will have only 433 307 prediction tokens, so it become impossible to compare the results between prediction and reality. Does someone encouter the same problem ?

This is the code for the model loading nothing fancy :

`from simpletransformers.ner import NERModel

model = NERModel(
model_type='camembert', model_name='camembert-base',
labels = tags, use_cuda= True
)`