def test_record_raise_error(self):
"""Test that the method record raises a ``ValueError`` when ``len(trials)`` is different
to ``len(scores)``.
"""
# setup
instance = MagicMock()
instance.tunable.transform.return_value = np.array([[1, 0]])
# run / assert
with self.assertRaises(ValueError):
BaseTuner.record(instance, 1, [1, 2])
def test__check_proposals_proposals_gt_cardinality(self):
"""Test that ``ValueError`` is being raised if ``proposals`` is greater than
``self.tunable.cardinality``.
"""
# setup
instance = MagicMock()
instance.tunable.cardinality = 4
# run / assert
with self.assertRaises(ValueError):
BaseTuner._check_proposals(instance, 5)
def test__check_proposals_trials_and_proposals_gt_cardinality(self):
"""Test that ``ValueError`` is being raised if ``proposals`` + ``len(self.trials)``
is greater than ``self.tunable.cardinality``.
"""
# setup
instance = MagicMock()
instance.tunable.cardinality = 4
instance._trials_set.__len__.return_value = 2
# run / assert
with self.assertRaises(ValueError):
BaseTuner._check_proposals(instance, 3)
def test__check_proposals_trials_eq_cardinality(self):
"""Test that ``ValueError`` is being raised if ``self.trials`` is equal to
``self.tunable.cardinality``.
"""
# setup
instance = MagicMock()
instance.tunable.cardinality = 2
instance._trials_set.__len__.return_value = 2
# run / assert
with self.assertRaises(ValueError):
BaseTuner._check_proposals(instance, 1)
def test_sample_no_duplicates_more_than_one_loop(self):
"""Test that the method ``_sample`` returns ``np.ndarray`` when not using duplicates and
perfroms more than one iteration.
"""
# setup
instance = MagicMock()
instance.tunable = MagicMock(spec_set=Tunable)
instance._trials_set = set({(1, ), (2, )})
side_effect = [
np.array([[3]]),
np.array([[1]]),
np.array([[1]]),
np.array([[4]])
]
instance.tunable.sample.side_effect = side_effect
# run
result = BaseTuner._sample(instance, 2, False)
# assert
assert instance.tunable.sample.call_args_list == [
call(2), call(2), call(2), call(2)
]
np.testing.assert_array_equal(result, np.array([[3], [4]]))
def test__check_proposals_not_raise(self):
"""Test that ``ValueError`` is not being raised."""
# setup
instance = MagicMock()
instance.tunable.cardinality = 4
instance._trials_set.__len__.return_value = 2
# run / assert
result = BaseTuner._check_proposals(instance, 1)
assert result is None
def test_record_scalar_values(self):
"""Test that the method record performs an update to ``trials`` and ``scores`` when called
with a scalar value.
"""
# setup
instance = MagicMock()
instance.trials = np.empty((0, 2), dtype=np.float)
instance.raw_scores = np.empty((0, 1), dtype=np.float)
instance._trials_set = set()
instance.tunable.transform.return_value = np.array([[1, 0]])
# run
BaseTuner.record(instance, 1, 0.1)
# assert
instance.tunable.transform.assert_called_once_with(1)
np.testing.assert_array_equal(instance.trials, np.array([[1, 0]]))
assert instance._trials_set == set({(1, 0)})
np.testing.assert_array_equal(instance.raw_scores, np.array([0.1]))
np.testing.assert_array_equal(instance.scores, np.array([0.1]))
def test__sample_allow_duplicates(self):
"""Test the method ``_sample``when using duplicates."""
# setup
instance = MagicMock()
instance.tunable.sample.return_value = 1
# run
result = BaseTuner._sample(instance, 1, True)
# assert
instance.tunable.sample.assert_called_once_with(1)
assert result == 1
def test_record_list_maximize_false(self):
"""Test that the method record updates the ``trials`` and ``scores``."""
# setup
instance = MagicMock()
instance.tunable.transform.return_value = np.array([[1, 0]])
instance.trials = np.empty((0, 2), dtype=np.float)
instance._trials_set = set()
instance.scores = None
instance.maximize = False
instance.raw_scores = np.empty((0, 1), dtype=np.float)
# run
BaseTuner.record(instance, [1], [0.1])
# assert
instance.tunable.transform.assert_called_once_with([1])
np.testing.assert_array_equal(instance.trials, np.array([[1, 0]]))
assert instance._trials_set == set({(1, 0)})
np.testing.assert_array_equal(instance.raw_scores, np.array([0.1]))
np.testing.assert_array_equal(instance.scores, np.array([-0.1]))
def test__sample_not_allow_duplicates(self):
"""Test that the method ``_sample`` returns ``np.ndarray`` when not using duplicates."""
# setup
instance = MagicMock()
instance._trials_set = set()
instance.tunable.sample.return_value = np.array([[3]])
# run
result = BaseTuner._sample(instance, 1, False)
# assert
instance.tunable.sample.assert_called_once_with(1)
np.testing.assert_array_equal(result, np.array([[3]]))
def test___init__maximize_false(self):
# setup
tunable = MagicMock(spec_set=Tunable)
# run
instance = BaseTuner(tunable, False)
# assert
assert isinstance(instance.tunable, MagicMock)
assert isinstance(instance.trials, np.ndarray)
assert isinstance(instance.raw_scores, np.ndarray)
assert isinstance(instance._trials_set, set)
assert isinstance(instance.maximize, bool)
assert not instance.maximize
def test_propose_many_values_no_duplicates(self, mock__check_proposals):
"""Test that ``propose`` method calls it's child implemented method with more than one
proposals and ``allow_duplicates`` as ``False``.
"""
# setup
instance = MagicMock()
inverse_return = instance.tunable.inverse_transform.return_value
inverse_return.to_dict.return_value = [1, 2]
instance._propose = MagicMock(return_value=2)
# run
result = BaseTuner.propose(instance, 2)
# assert
instance._propose.assert_called_once_with(2, False)
instance.tunable.inverse_transform.called_once_with(2)
inverse_return.to_dict.assert_called_once_with(orient='records')
assert result == [1, 2]
def test_propose_one_value_allow_duplicates(self, mock__check_proposals):
"""Test that ``propose`` method calls it's child implemented method with
``allow_duplicates`` as ``True``.
"""
# setup
instance = MagicMock()
inverse_return = instance.tunable.inverse_transform.return_value
inverse_return.to_dict.return_value = [1]
instance._propose = MagicMock(return_value=1)
# run
result = BaseTuner.propose(instance, 1, allow_duplicates=True)
# assert
instance._check_proposals.assert_not_called()
instance._propose.assert_called_once_with(1, True)
instance.tunable.inverse_transform.called_once_with(1)
inverse_return.to_dict.assert_called_once_with(orient='records')
assert result == 1
def test___init__defaults(self):
# setup
tunable = MagicMock(spec_set=Tunable)
# run
instance = BaseTuner(tunable)
# assert
assert instance.tunable is tunable
assert isinstance(instance.trials, np.ndarray)
assert isinstance(instance.raw_scores, np.ndarray)
assert isinstance(instance._trials_set, set)
assert isinstance(instance.maximize, bool)
assert instance.maximize
assert instance.trials.shape == (0, 1)
assert instance.raw_scores.shape == (0, 1)
assert instance.trials.dtype == np.float
assert instance.raw_scores.dtype == np.float
