michaelhills / seizure-detection Goto Github PK

I tried to run the code, training is ok, but predict.py doesn't work.
I read the hickle.py and checked the container_type inside the classifier, it's "sklearn.ensemble.forest.RandomForestClassifier".
I guess it has the right thing inside the pycontainer, but I cannot figure out why predict_proba is not working.

Traceback (most recent call last):
File "./predict.py", line 6, in
run_seizure_detection('make_predictions')
File "seizure_detection.py", line 179, in run_seizure_detection
train_full_model(make_predictions=True)
File "seizure_detection.py", line 114, in train_full_model
predictions = MakePredictionsTask(task_core).run()
File "seizure/tasks.py", line 25, in run
return self.task_core.cached_data_loader.load(self.filename(), self.load_data)
File "common/data.py", line 42, in load
data = func()
File "seizure/tasks.py", line 121, in load_data
return make_predictions(self.task_core.target, X_test, y_classes, classifier_data)
File "seizure/tasks.py", line 442, in make_predictions
predictions_proba = classifier.predict_proba(X_test)
AttributeError: 'PyContainer' object has no attribute 'predict_proba'

ipdb> pp classifier.container_type
class 'sklearn.ensemble.forest.RandomForestClassifier'>"
ipdb> pp type(classifier[0])
type 'numpy.ndarray'>
ipdb> pp type(classifier)
class 'hickle.PyContainer'>