Hello, I have a fine tuned NER model, that is posted on HuggingFace. I tried one time using pipeline in my jupyter notebook and left out aggregation_strategy. The pipeline printed to output something, that is hidden on the HuggingFace site and when using aggregation_strategy set to simple
Without the strategy
[{‘entity’: ‘B-O’,
‘score’: 0.88954985,
‘index’: 1,
‘word’: ‘▁Titan’,
‘start’: 0,
‘end’: 5},
{‘entity’: ‘I-O’,
‘score’: 0.8819839,
‘index’: 2,
‘word’: ‘ik’,
‘start’: 5,
‘end’: 7},
{‘entity’: ‘B-T’,
‘score’: 0.999821,
‘index’: 16,
‘word’: ‘▁1912’,
‘start’: 52,
‘end’: 56},
{‘entity’: ‘B-G’,
‘score’: 0.9994797,
‘index’: 23,
‘word’: ‘▁Atlantik’,
‘start’: 72,
‘end’: 80},
{‘entity’: ‘I-G’,
‘score’: 0.9997135,
‘index’: 24,
‘word’: ‘u’,
‘start’: 80,
‘end’: 81}]
With the strategy
[{‘entity_group’: ‘T’,
‘score’: 0.9999019,
‘word’: ‘1912’,
‘start’: 52,
‘end’: 56},
{‘entity_group’: ‘G’,
‘score’: 0.99984026,
‘word’: ‘Atlantiku’,
‘start’: 72,
‘end’: 81}]
Did someone run into something similar or is it a bug?