AspectExtractor returns aspects only if pred_sentiments=True in extract_aspect method but I don't want sentiments

[KadriMufti]

Version
See the console output for PyABSA, Torch, Transformers Version

Describe the bug
The problem is that I should be able to set the parameter pred_sentiment=False in extract_aspect method and get back aspects without sentiments, but instead I get no aspects.

Code To Reproduce

from pyabsa.tasks import AspectTermExtraction

aspect_extractor = AspectTermExtraction.AspectExtractor(checkpoint = 'path/to/checkpoints/ATEPC_MULTILINGUAL_CHECKPOINT',
                                    auto_device = False  # False means load model on CPU
                                    )

examples = ['The staff was so perfect to us, but the service was bad.']

atepc_result = aspect_extractor.extract_aspect(inference_source=examples,  #
                          save_result=True,
                          print_result=True,  # print the result
                          pred_sentiment=False,  # Predict the sentiment of extracted aspect terms
                          )

print(atepc_result)

Expected behavior

with pred_sentiment=True, the output of the above example is:

[{'sentence': 'The staff was so perfect to us , but the service was bad .',
  'IOB': ['O', 'B-ASP', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-ASP', 'O', 'O', 'O'], 'tokens': ['The', 'staff', 'was', 'so', 'perfect', 'to', 'us', ',', 'but', 'the', 'service', 'was', 'bad', '.'],  'aspect': ['staff', 'service'],  'position': [[1], [10]],   'sentiment': ['Positive', 'Negative'],  'probs': [0.0036869419272989035, 0.001312303589656949, 0.9950007796287537],  [0.995452344417572, 0.0023385591339319944, 0.00220911274664104]], 'confidence': [0.995, 0.9955]}]

with pred_sentiment=False, the output of the above example is:

[{'sentence': 'The staff was so perfect to us , but the service was bad .',
  'IOB': ['O', 'B-ASP', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-ASP', 'O', 'O', 'O'], 'tokens': ['The', 'staff', 'was', 'so', 'perfect', 'to', 'us', ',', 'but', 'the', 'service', 'was', 'bad', '.'], 'aspect': [], 'position': [], 'sentiment': [], 'probs': [], 'confidence': []}]

I am not sure where the problem is, but it might have to do with the extract function _extract() in line 372 in aspect_extractor.py

[yangheng95]

Bug report received, pending fixing

[yangheng95]

Can you try 2.4.1 and report the result? @KadriMufti

[KadriMufti]

I installed pyabsa version 2.4.1 and torch version 1.13.1 and transformers version 4.27.2

There is no change in results. Aspects are still empty when pred_sentiment=False.