Open rap9430 opened 6 years ago
it's probably one of the parameters that's being passed in. we have some level of input verification, but not a ton, so it's possible some input's being passed in that doesn't work.
could you share that part of your training script- where you create ml_predictor = Predictor()
and ml_predictor.train()
?
I am not sure that the following is quite informative, but here is the part of the script you asked for.
column_descriptions = { 'col_numeric': 'output', 'nlp_col': 'nlp' }
ml_predictor = Predictor(type_of_estimator='classification', column_descriptions=column_descriptions) ml_predictor.train(data_train, verbose = False, optimize_final_model=True, ml_for_analytics=False)
'AttributeError: 'NoneType' object has no attribute 'score'
column_descriptions = { 'op': 'output' }
ml_predictor = Predictor(type_of_estimator='classifier', column_descriptions=column_descriptions) ml_predictor.train(data, verbose = False, optimize_final_model=True)
"verbose = False" is also not working in this case. I'm still getting full logging.
Hi,
Is this working for any of you now? I seem to be getting the same error. I ran 3 modifications of which 1 ran-
df_iris = pd.read_csv('https://raw.githubusercontent.com/mwaskom/seaborn-data/master/iris.csv') df_iris.head(10)
col_desc = { 'species' : 'output' }
automl_kwargs = { 'model_names' : ['XGBClassifier'],
'verbose' : True,
'ml_for_analytics' : True,
'take_log_of_y' : True,
'perform_feature_scaling' : True,
'cv' : 9,
# 'optimize_final_model' : True
}
aml_pred = auto_ml.Predictor('classifier',column_descriptions=col_desc) aml_pred.train(df_iris, **automl_kwargs)
automl_kwargs2 = automl_kwargs.copy() automl_kwargs2['optimize_final_model'] = True aml_pred2 = auto_ml.Predictor('classifier',column_descriptions=col_desc) aml_pred2.train(df_iris, **automl_kwargs2)
automl_kwargs3 = automl_kwargs.copy() automl_kwargs3['model_names'] = ['XGBClassifier','GradientBoostingClassifier']
aml_pred3 = auto_ml.Predictor('classifier',column_descriptions=col_desc) aml_pred3.train(df_iris, **automl_kwargs3)
The idea is to try and invoke grid search. I have a larger dataset but I'm trying to debug this with a sample data available online.
I face same error when using not default model. I tried some like LogisticRegression Ada and RandomForestClassifier but they all responded with 'NoneType' object has no attribute 'score'
1160 error_score=-1000000000,
-> 1161 scoring=self._scorer.score,
1162 # Don't allocate memory for all jobs upfront. Instead, only allocate enough memory to handle the current jobs plus an additional 50%
1163 pre_dispatch='1.5*n_jobs',
AttributeError: 'NoneType' object has no attribute 'score'
Whenever I set optimize_final_model=True with type_of_estimator='classification' (for all classifiers), I get the following error: AttributeError: 'NoneType' object has no attribute 'score'
FYI, everything works fine with the regression pipelines
Find the Traceback below:
Traceback (most recent call last): File "", line 14, in
File "/lib/python3.5/site-packages/auto_ml/predictor.py", line 638, in train
self.trained_final_model = self.train_ml_estimator(self.model_names, self._scorer, X_df, y)
File "/lib/python3.5/site-packages/auto_ml/predictor.py", line 1199, in train_ml_estimator
gscv_results = self.fit_grid_search(X_df, y, grid_search_params, feature_learning=feature_learning)
File "/lib/python3.5/site-packages/auto_ml/predictor.py", line 1077, in fit_grid_search
scoring=self._scorer.score,
AttributeError: 'NoneType' object has no attribute 'score'