@AlJohri writes:

I started getting a different shape for cross_val_predict with the method of decision_function for this model: svm.SVC(kernel='linear', C=1, probability=False) with this cross_val_predict(original_pipeline, X_train, y_train, cv=2, method='decision_function'). I was getting a 1D array before and now I'm getting a 2D array where the first column is all zeroes. I'm doing a binary classifier.