def test_fit_mongo(self,mock_MongoTrials, mock_fmin):
mongo = {
'url': 'mongo://localhost:1234/foo_db/jobs',
'exp': 'my_experience'
}
mock_MongoTrials.return_value = 'random_hash_4b9u4y5h3r9u'
cs.fit(mongo=mongo)
mock_MongoTrials.assert_called_once_with('mongo://localhost:1234/foo_db/jobs', exp_key='my_experience')
args, kwargs = mock_fmin.call_args_list[0]
self.assertEqual(kwargs['trials'], 'random_hash_4b9u4y5h3r9u')
def test_fit(self, mock_fmin):
trials=Trials()
cs.fit()
# import pdb; pdb.set_trace()
mock_fmin.assert_called_once_with(
cs.objective,
(
('model', 'test'),
('data_path', 'path_to_data'),
('targets_path', 'path_to_targets')
),
max_evals=100,
algo=hpo.tpe.suggest,
trials=None
)
def test_objective(self, mock_load_data):
params = ({
'name': MLPClassifier,
'params': {
'activation': 'relu',
'hidden_layer_sizes': (4,3,2),
'random_state':1
}
},
'./tests/data/A.csv',
'./tests/data/targets.csv')
mock_load_data.return_value = (
[[1,2,3]]*5 + [[3,4,5]]*5 , [[1]]*4+[[0]]*6
)*2
cv_metric = cs.objective(params)
calls= [
call( data_path='./tests/data/A.csv',
targets_path='./tests/data/targets.csv',
train_set = 0.8,
random_state=i
) for i in range(1, 6)
]
mock_load_data.assert_has_calls(calls)
self.assertEqual(cv_metric, .5)
def test_get_data_not_found_locally(self, mock_read_csv, mock_download_file):
data = cs.get_data('./tests/data/B.csv')
mock_download_file.assert_called_once_with(
'ccg-machine-learning',
'prior-data/B.csv',
'./tests/data/B.csv'
)
def test_interpreter_no_params(self):
params = {
'name': MLPClassifier
}
expected_model = MLPClassifier()
model = cs.interpreter(params)
self.assertEqual(str(expected_model), str(model))
def test_evaluate(self):
X = np.array([[1,2,3,4]]*5 + [[2,4,6,8]]*5 )
Y = np.array([True, True, True, False, False, False, False, False, False, False])
class Model:
def predict_proba(self, X):
return np.array([[0.98, 0.02]]*5+[[0.02, 0.98]]*5)
model = Model()
metric = cs.evaluate(model, X, Y)
self.assertAlmostEqual(metric, 0.14285, places=4)
def test_load_data(self):
X_train, Y_train, X_test, Y_test = cs.load_data(
data_path='./tests/data/A.csv',
targets_path='./tests/data/targets.csv',
train_set = 0.5,
random_state=1
)
self.assertEqual(str(X_train), str(np.genfromtxt('./tests/data/exp_train_A.csv', delimiter=',', dtype=int)))
self.assertEqual(str(Y_train), str(np.genfromtxt('./tests/data/exp_train_targets.csv', delimiter=',', dtype=bool)))
self.assertEqual(str(X_test), str(np.genfromtxt('./tests/data/exp_test_A.csv', delimiter=',', dtype=int)))
self.assertEqual(str(Y_test), str(np.genfromtxt('./tests/data/exp_test_targets.csv', delimiter=',', dtype=bool)))
def test_interpreter(self):
params = {
'name': MLPClassifier,
'params': {
'activation': 'relu',
'hidden_layer_sizes': (4,3,2)
}
}
expected_model = MLPClassifier(
activation='relu',
hidden_layer_sizes=(4,3,2)
)
model = cs.interpreter(params)
self.assertEqual(str(expected_model), str(model))
def test_load_data_calls_get_data(self, mock_get_data):
data_path = './tests/data/A.csv'
targets_path = './tests/data/targets.csv'
mock_get_data.return_value = pd.read_csv(data_path, index_col=0)
X_train, Y_train, X_test, Y_test = cs.load_data(
data_path=data_path,
targets_path=targets_path,
train_set = 0.5,
random_state=1
)
calls = [
call(data_path),
call(targets_path),
]
mock_get_data.assert_has_calls(calls)
def test_get_data(self):
data = cs.get_data('./tests/data/A.csv')
exp_data = pd.read_csv('./tests/data/A.csv', index_col=0)
self.assertEquals(str(data), str(exp_data))