I created an MLP model and saved it with Pickle. I then loaded the model and applied it to new data using the following code:

mlp_probs = pickle_model.predict_proba(X_test)

That works perfect. However when I try to get the class predictions using:

mlp_predictions = pickle_model.predict(X_test)

all predictions are 0. In other words there is no positive class. The model worked perfectly before I pickled it so I am guessing that the threshold was not saved during the pickling process. How can I fix this?