Running predictions for the test data