AssertionError using py.test on test_wabbit_wappa

[jmwoloso]

Hello,

My setup:

Linux Mint: 17.2 Python: 3.4.3 VW: 7.3.0

I'm getting the traceback below when I run the tests. I'm using the Linux repo version of VW (7.3.0), so perhaps this is the issue? But in case that isn't the issue, see output below. Any suggestions?

platform linux -- Python 3.4.3, pytest-2.8.3, py-1.4.31, pluggy-0.3.1
rootdir: /home/jason/source/wabbit_wappa, inifile: 
plugins: pep8-1.0.6, cov-2.2.0
collected 5 items 

test_wabbit_wappa.py ....F

=================================== FAILURES ===================================
________________________________ test_training _________________________________

    def test_training():
        # TODO: pytest probably has a framework for testing hyperparameters like this
        for active_mode in [False, True]:
            vw = VW(loss_function='logistic', active_mode=active_mode)
            # Train with an easy case
            for i in range(20):
                # Positive example
                vw.send_example(response=1.,
                                importance=2.,
                                tag='positive',
                                features=[('a', 1 + random.random()),
                                          ('b', -1 - random.random())]
                                )
                vw.send_example(response=-1.,
                                importance=.5,
                                tag='negative',
                                features=[('lungfish', 1 + random.random()),
                                          ('palooka', -1 - random.random())]
                                )
            prediction1 = vw.get_prediction([('a', 1),
                                            ('b', -2)]).prediction
            # Prediction should be definitively positive
            assert prediction1 > 1.
            prediction2 = vw.get_prediction([('lungfish', 3)]).prediction
            # Prediction should be negative
            assert prediction2 < 0
            prediction3 = vw.get_prediction([('a', 1),
                                            ('b', -2)]).prediction
            # Making predictions shouldn't affect the trained model
            assert prediction1 == prediction3

            # Continue training with very different examples
            for i in range(20):
                # Positive example
                vw.add_namespace('space1',
                                 1.0,
                                 ['X', 'Y', 'Z'],
                                 )
                vw.send_example(response=1.)
                # Negative example
                vw.add_namespace('space2',
                                 2.0,
                                 ['X', 'Y', 'Z'],
                                 )
                vw.send_example(response=-1.)
            vw.add_namespace('space1',
                             1.0,
                             ['X'],
                             )
            prediction4 = vw.get_prediction().prediction
            # Prediction should be positive
            assert prediction4 > 0
            vw.add_namespace('space2',
                             1.0,
                             ['X'],
                             )
            prediction5 = vw.get_prediction().prediction
            # Prediction should be negative
            assert prediction5 < 0

            # Save the model to a temporary file
            filename = '__temp.model'
            vw.save_model(filename)
            # This sleep is required only in active_mode, in the (unusual) case
            # that the model file is used immediately
            time.sleep(0.1)

            # Load a new VW instance from that model
            vw2 = VW(loss_function='logistic', i=filename)
            # Make the same prediction with each model (testing cache_string to boot)
            namespace1 = Namespace(features=[('a', 1), ('b', -2)], cache_string=True)
            namespace2 = Namespace('space1', 1.0, ['X', 'Y'], cache_string=True)
            prediction1 = vw.get_prediction(namespaces=[namespace1, namespace2]).prediction
            prediction2 = vw2.get_prediction(namespaces=[namespace1, namespace2]).prediction
>           assert prediction1 == prediction2
E           assert 0.0 == 4.798696

test_wabbit_wappa.py:133: AssertionError
====================== 1 failed, 4 passed in 0.26 seconds ======================