Prediction works for direct_nlp but crashes for hybrid and None/default

[howff]

With predict_mode set to hybrid

[predict_helper.py:26] root 2022-01-17 16:18:18,330 working on [atrophy]
Traceback (most recent call last):
  File "nlp2phenome/predict_helper.py", line 184, in <module>
    predict_to_eHOST_results(sys.argv[1])
  File "nlp2phenome/predict_helper.py", line 165, in predict_to_eHOST_results
    predicted_results = hybrid_prediciton(ss.settings)
  File "nlp2phenome/predict_helper.py", line 78, in hybrid_prediciton
    d2p, labels2work = predict(settings)
  File "nlp2phenome/predict_helper.py", line 43, in predict
    full_text_dir=test_text_dir)
TypeError: load_data_for_predict() got an unexpected keyword argument 'concept_mapping_file'

Without predict_mode

[predict_helper.py:167] root 2022-01-17 16:18:45,485 predicting...
[predict_helper.py:26] root 2022-01-17 16:18:45,486 working on [atrophy]
Traceback (most recent call last):
  File "nlp2phenome/predict_helper.py", line 184, in <module>
    predict_to_eHOST_results(sys.argv[1])
  File "nlp2phenome/predict_helper.py", line 168, in predict_to_eHOST_results
    predicted_results = predict(ss.settings)
  File "nlp2phenome/predict_helper.py", line 43, in predict
    full_text_dir=test_text_dir)
TypeError: load_data_for_predict() got an unexpected keyword argument 'concept_mapping_file'

[Honghan]

Just updated the predict_helper to fix the issue. This was caused by a code cleaning done back in 2020.

[howff]

After applying abff606
Now it gives a different error

    lm.concept_mapping = _cm_obj
AttributeError: can't set attribute

[howff]

We can solve that error if we change lm.concept_mapping = _cm_obj into lm._concept_mapping = _cm_obj

However then we get another error (in None/default mode)

in output_eHOST_format
    xml = AnnConverter.to_eHOST(d, doc2precited[d])
TypeError: tuple indices must be integers or slices, not dict

[Honghan]

No - it's my bad - forgot to push another file LabelModel.py which I had added a function to allow attribute setting.