def test_predict(self):
session = self.get_session()
session.add(DatasetSample(image=pickle.dumps(np.array([0])), label=0))
session.add(DatasetSample(image=pickle.dumps(np.array([1])), label=1))
session.commit()
t = ModelServingWrapper(LogisticRegression(), self.TESTBASE_PATH)
t.fit()
self.assertEqual(t.predict(np.array([1])), 1)
def test__get_training_dataset_shapes_mismatch(self):
session = self.get_session()
session.add(DatasetSample(image=pickle.dumps(np.array([0])), label=0))
session.add(
DatasetSample(image=pickle.dumps(np.array([0, 1])), label=0))
session.commit()
t = ModelServingWrapper(LinearRegression(), self.TESTBASE_PATH)
with self.assertRaises(ValueError):
images, labels, shape = t._get_training_dataset()
def test__fit(self):
session = self.get_session()
session.add(DatasetSample(image=pickle.dumps(np.array([0])), label=0))
session.add(DatasetSample(image=pickle.dumps(np.array([1])), label=1))
session.commit()
t = ModelServingWrapper(LinearRegression(), self.TESTBASE_PATH)
t._fit()
self.assertAlmostEqual(t._model.coef_[0], 1)
self.assertAlmostEqual(t._model.intercept_, 0)
def test__get_training_dataset_one_sample(self):
session = self.get_session()
test_vector = np.random.random(100)
test_label = np.random.randint(0, 128)
session.add(
DatasetSample(image=pickle.dumps(test_vector), label=test_label))
session.commit()
t = ModelServingWrapper(LinearRegression(), self.TESTBASE_PATH)
images, labels, shape = t._get_training_dataset()
self.assertTrue(np.array_equal(images, np.vstack([test_vector])))
self.assertTrue(np.array_equal(labels, np.array([test_label])))
self.assertEqual(shape, test_vector.shape)
def test_predict_incorrect_shape(self):
session = self.get_session()
session.add(
DatasetSample(image=pickle.dumps(np.array([0, 2])), label=1))
session.add(
DatasetSample(image=pickle.dumps(np.array([0, 1])), label=0))
session.commit()
t = ModelServingWrapper(LogisticRegression(), self.TESTBASE_PATH)
t.fit()
with self.assertRaises(ShapeMismatchError):
t.predict(np.array([0]))
def test__get_training_dataset_empty(self):
t = ModelServingWrapper(LinearRegression(), self.TESTBASE_PATH)
with self.assertRaises(ValueError):
t._get_training_dataset()
def test__init__with_fitted_model(self):
fitted_model = LogisticRegression()
fitted_model.fit(np.array([[0, 1, 2], [4, 5, 6]]), np.array([0, 1]))
with self.assertRaises(ValueError):
ModelServingWrapper(fitted_model, self.TESTBASE_PATH)
def test_predict_not_fitted(self):
t = ModelServingWrapper(LogisticRegression(), self.TESTBASE_PATH)
with self.assertRaises(ValueError):
t.predict(np.array([0, 1, 2]))
Example:
learner.py mssql+pyodbc://user:[email protected]/mnist?driver=ODBC+Driver+17+for+SQL+Server models/model.pickle
"""
import argparse
import pickle
from sklearn.linear_model import LogisticRegression
from app.wrapper import ModelServingWrapper
parser = argparse.ArgumentParser(
description='Fits the LogisticRegression model, using samples from given '
'database and saves to file')
parser.add_argument('database_path',
type=str,
help='SQLAlchemy-compatible Database URL')
parser.add_argument('file_path', type=str, help='Path to file with model')
args = parser.parse_args()
model = ModelServingWrapper(LogisticRegression(), args.database_path)
accuracy = model.fit()
with open(args.file_path, 'wb') as handle:
pickle.dump(model, handle)