def test_sample_from_data(self, PredictService):
predict_service = PredictService(mapping=[
('name', 'str'),
('sepal width', 'int'),
], )
model = Mock()
request_args = {'name': 'myflower', 'sepal width': 3}
sample = predict_service.sample_from_data(model, request_args)
assert sample[0] == 'myflower'
assert sample[1] == 3
def test_sample_from_request(self, PredictService):
predict_service = PredictService(mapping=[
('name', 'str'),
('sepal width', 'int'),
], )
model = Mock()
request = Mock(args={'name': 'myflower', 'sepal width': 3})
sample, params = predict_service.sample_from_request(model, request)
assert sample[0][0] == 'myflower'
assert sample[0][1] == 3
def test_sample_from_data(self, PredictService):
predict_service = PredictService(
mapping=[
('name', 'str'),
('sepal width', 'int'),
],
)
model = Mock()
request_args = {'name': 'myflower', 'sepal width': 3}
sample = predict_service.sample_from_data(model, request_args)
assert sample[0] == 'myflower'
assert sample[1] == 3
def test_sample_from_request(self, PredictService):
predict_service = PredictService(
mapping=[
('name', 'str'),
('sepal width', 'int'),
],
)
model = Mock()
request = Mock(args={'name': 'myflower', 'sepal width': 3})
sample, params = predict_service.sample_from_request(model, request)
assert sample[0][0] == 'myflower'
assert sample[0][1] == 3
def test_predict_params(self, config, stream):
from palladium.server import PredictService
line = '[{"length": 1.0, "width": 1.0, "turbo": "true"}]'
model = Mock()
model.predict.return_value = np.array([[{'class': 'a'}]])
model.turbo = False
model.magic = False
stream.model = model
mapping = [
('length', 'float'),
('width', 'float'),
]
params = [
('turbo', 'bool'), # will be set by request args
('magic', 'bool'), # default value will be used
]
stream.predict_service = PredictService(
mapping=mapping,
params=params,
)
expected = [{'class': 'a'}]
result = stream.process_line(line)
assert result == expected
assert model.predict.call_count == 1
assert (model.predict.call_args[0][0] == np.array([[1.0, 1.0]])).all()
assert model.predict.call_args[1]['turbo'] is True
assert model.predict.call_args[1]['magic'] is False
def test_unwrap_sample_post(self, PredictService, flask_app):
predict_service = PredictService(
mapping=[('text', 'str')],
unwrap_sample=True,
)
model = Mock()
model.predict.return_value = np.array([1, 2])
with flask_app.test_request_context():
request = Mock(
json=[
{
'text': 'First piece of text'
},
{
'text': 'Second piece of text'
},
],
method='POST',
mimetype='application/json',
)
resp = predict_service(model, request)
assert model.predict.call_args[0][0].ndim == 1
assert (model.predict.call_args[0] == np.array(
['First piece of text', 'Second piece of text'])).all()
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 200
assert resp_data == {
"metadata": {
"status": "OK",
"error_code": 0,
},
"result": [1, 2],
}
def test_unwrap_sample_get(self, PredictService, flask_app):
predict_service = PredictService(
mapping=[('text', 'str')],
unwrap_sample=True,
)
model = Mock()
model.predict.return_value = np.array([1])
with flask_app.test_request_context():
request = Mock(
args=dict([
('text', 'Hi this is text'),
]),
method='GET',
)
resp = predict_service(model, request)
assert model.predict.call_args[0][0].ndim == 1
model.predict.assert_called_with(np.array(['Hi this is text']))
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 200
assert resp_data == {
"metadata": {
"status": "OK",
"error_code": 0,
},
"result": 1,
}
def test_post_request(self, PredictService, flask_app):
model = Mock()
model.predict.return_value = np.array([3, 2])
service = PredictService(
mapping=[
('sepal length', 'float'),
('sepal width', 'float'),
('petal length', 'float'),
('petal width', 'float'),
],
params=[
('threshold', 'float'),
],
)
request = Mock(
json=[
{
'sepal length': '5.2',
'sepal width': '3.5',
'petal length': '1.5',
'petal width': '0.2',
},
{
'sepal length': '5.7',
'sepal width': '4.0',
'petal length': '2.0',
'petal width': '0.7',
},
],
args=dict(threshold=1.0),
method='POST',
mimetype='application/json',
)
with flask_app.test_request_context():
resp = service(model, request)
assert (model.predict.call_args[0][0] == np.array(
[
[5.2, 3.5, 1.5, 0.2],
[5.7, 4.0, 2.0, 0.7],
],
dtype='object',
)).all()
assert model.predict.call_args[1]['threshold'] == 1.0
assert resp.status_code == 200
expected_resp_data = {
"metadata": {
"status": "OK",
"error_code": 0,
},
"result": [3, 2],
}
assert json.loads(resp.get_data(as_text=True)) == expected_resp_data
def test_probas(self, PredictService, flask_app):
model = Mock()
model.predict_proba.return_value = np.array([[0.1, 0.5, math.pi]])
predict_service = PredictService(mapping=[], predict_proba=True)
with flask_app.test_request_context():
resp = predict_service(model, request)
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 200
assert resp_data == {
"metadata": {
"status": "OK",
"error_code": 0,
},
"result": [0.1, 0.5, math.pi],
}
def test_generic_error(self, PredictService, flask_app):
predict_service = PredictService(mapping=[])
model = Mock()
request = Mock()
with patch.object(predict_service, 'do') as psd:
with flask_app.test_request_context():
psd.side_effect = KeyError("model")
resp = predict_service(model, request)
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 500
assert resp_data == {
"metadata": {
"status": "ERROR",
"error_code": -1,
"error_message": "KeyError: 'model'",
}
}
def test_predict_error(self, PredictService, flask_app):
from palladium.interfaces import PredictError
predict_service = PredictService(mapping=[])
model = Mock()
request = Mock()
with patch.object(predict_service, 'do') as psd:
with flask_app.test_request_context():
psd.side_effect = PredictError("mymessage", 123)
resp = predict_service(model, request)
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 500
assert resp_data == {
"metadata": {
"status": "ERROR",
"error_code": 123,
"error_message": "mymessage",
}
}
def test_probas(self, PredictService, flask_app):
model = Mock()
model.predict_proba.return_value = np.array([[0.1, 0.5, 0.4]])
predict_service = PredictService(mapping=[], predict_proba=True)
with flask_app.test_request_context():
with patch('palladium.util.get_config') as get_config:
meta_dict = {'service_metadata': {}}
get_config.return_value = meta_dict
resp = predict_service(model, request)
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 200
assert resp_data == {
"metadata": {
"status": "OK",
"error_code": 0,
},
"result": [0.1, 0.5, 0.4],
}
def test_bad_request(self, PredictService, flask_app):
predict_service = PredictService(mapping=[])
model = Mock()
request = Mock()
with patch.object(predict_service, 'do') as psd:
with flask_app.test_request_context():
bad_request = BadRequest()
bad_request.args = ('daniel', )
psd.side_effect = bad_request
resp = predict_service(model, request)
resp_data = json.loads(resp.get_data(as_text=True))
assert resp.status_code == 400
assert resp_data == {
"metadata": {
"status": "ERROR",
"error_code": -1,
"error_message": "BadRequest: ('daniel',)"
}
}
def test_functional(self, PredictService, flask_app):
model = Mock()
model.threshold = 0.3
model.size = 10
# needed as hasattr would evaluate to True otherwise
del model.threshold2
del model.size2
model.predict.return_value = np.array(['class1'])
service = PredictService(
mapping=[
('sepal length', 'float'),
('sepal width', 'float'),
('petal length', 'float'),
('petal width', 'float'),
('color', 'str'),
('age', 'int'),
('active', 'bool'),
('austrian', 'bool'),
],
params=[
('threshold', 'float'), # default will be overwritten
('size', 'int'), # not provided, default value kept
('threshold2', 'float'), # will be used, no default value
('size2', 'int'), # not provided, no default value
])
with flask_app.test_request_context():
with patch('palladium.util.get_config') as get_config:
get_config.return_value = {
'service_metadata': {
'service_name': 'iris',
'service_version': '0.1'
}
}
request = Mock(
args=dict([
('sepal length', '5.2'),
('sepal width', '3.5'),
('petal length', '1.5'),
('petal width', '0.2'),
('color', 'purple'),
('age', '1'),
('active', 'True'),
('austrian', 'False'),
('threshold', '0.7'),
('threshold2', '0.8'),
]),
method='GET',
)
resp = service(model, request)
assert (model.predict.call_args[0][0] == np.array(
[[5.2, 3.5, 1.5, 0.2, 'purple', 1, True, False]],
dtype='object')).all()
assert model.predict.call_args[1]['threshold'] == 0.7
assert model.predict.call_args[1]['size'] == 10
assert model.predict.call_args[1]['threshold2'] == 0.8
assert 'size2' not in model.predict.call_args[1]
assert resp.status_code == 200
expected_resp_data = {
"metadata": {
"status": "OK",
"error_code": 0,
"service_name": "iris",
"service_version": "0.1",
},
"result": "class1"
}
assert json.loads(resp.get_data(as_text=True)) == expected_resp_data