def predict(self, feature_id, feature_names):
logging.info('feature_id = %s', feature_id)
logging.info('feature_names = %s', feature_names)
if any([i is None
for i in [self.m, self.b, self.fs, self.serving_fs]]):
with open('simple_model.dat', 'r') as f:
model = json.load(f)
self.m = np.array(model.get('m', []))
self.b = float(model.get('b', 0))
_FEAST_CORE_URL = model['FEAST_CORE_URL']
_FEAST_SERVING_URL = model['FEAST_SERVING_URL']
_ENTITY_ID = model['ENTITY_ID']
logging.info('FEAST_CORE_URL: %s', _FEAST_CORE_URL)
logging.info('FEAST_SERVING_URL: %s', _FEAST_SERVING_URL)
logging.info('ENTITY_ID: %s', _ENTITY_ID)
logging.info('FEATURES_SET: %s', self.SERVING_FEATURE_SET)
self.fs = Client(core_url=_FEAST_CORE_URL,
serving_url=_FEAST_SERVING_URL,
verbose=True)
self.serving_fs = FeatureSet(entity=_ENTITY_ID,
features=self.SERVING_FEATURE_SET)
features = self.fs.get_serving_data(self.serving_fs,
entity_keys=[feature_id])
X = features.to_numpy()[0][1:]
logging.info('X: %s', str(X))
return [sum(self.m * X) + self.b]
class LabelPrediction(object):
def __init__(self):
self.m = None
self.b = None
self.fs = None
self.serving_fs = None
logging.basicConfig(
level=logging.INFO,
format=('%(levelname)s|%(asctime)s'
'|%(pathname)s|%(lineno)d| %(message)s'),
datefmt='%Y-%m-%dT%H:%M:%S',
)
logging.getLogger().setLevel(logging.INFO)
def predict(self, feature_id, feature_names):
logging.info('feature_id = %s', feature_id)
logging.info('feature_names = %s', feature_names)
if any([i is None
for i in [self.m, self.b, self.fs, self.serving_fs]]):
with open('simple_model.dat', 'r') as f:
model = json.load(f)
self.m = np.array(model.get('m', []))
self.b = float(model.get('b', 0))
_FEAST_CORE_URL = model['FEAST_CORE_URL']
_FEAST_SERVING_URL = model['FEAST_SERVING_URL']
_ENTITY_ID = model['ENTITY_ID']
_FEATURES_SET = model['FEATURES_SET']
logging.info('FEAST_CORE_URL: %s', _FEAST_CORE_URL)
logging.info('FEAST_SERVING_URL: %s', _FEAST_SERVING_URL)
logging.info('ENTITY_ID: %s', _ENTITY_ID)
logging.info('FEATURES_SET: %s', _FEATURES_SET)
self.fs = Client(core_url=_FEAST_CORE_URL,
serving_url=_FEAST_SERVING_URL,
verbose=True)
self.serving_fs = FeatureSet(entity=_ENTITY_ID,
features=_FEATURES_SET)
features = self.fs.get_serving_data(self.serving_fs,
entity_keys=[feature_id])
X = features.to_numpy()[0][1:]
logging.info('X: %s', str(X))
return sum(self.m * X) + self.b
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--entity_spec_file", required=True)
parser.add_argument("--feature_spec_files", required=True)
parser.add_argument("--expected-warehouse-values-file")
parser.add_argument("--expected-serving-values-file")
parser.add_argument("--bigquery-dataset-for-warehouse")
parser.add_argument("--feast-serving-url")
parser.add_argument("--project")
args = parser.parse_args()
feast_client = Client(serving_url=args.feast_serving_url)
entity_name = get_entity_name(args.entity_spec_file)
feature_infos = get_feature_infos(args.feature_spec_files)
if args.expected_warehouse_values_file:
if not args.bigquery_dataset_for_warehouse:
raise RuntimeError("Missing option --bigquery-dataset-for-warehouse")
validate_warehouse(
entity_name,
feature_infos,
args.expected_warehouse_values_file,
args.bigquery_dataset_for_warehouse,
args.project,
)
else:
print("\n[Skipping validation of Feast Warehouse data]")
print("because --expected-warehouse-values-file is not set")
if args.expected_serving_values_file:
validate_serving(
entity_name, feature_infos, args.expected_serving_values_file, feast_client
)
else:
print("\n[Skipping validation of Feast Serving data]")
print("because --expected-serving-values-file is not set")
class HousingModel(object):
"""Model class."""
SERVING_FEATURE_SET = [
'usa_housing.avg_area_income', 'usa_housing.avg_area_house_age',
'usa_housing.avg_area_number_of_rooms',
'usa_housing.avg_area_number_of_bedrooms',
'usa_housing.area_population'
]
def __init__(self):
self.m = None
self.b = None
self.fs = None
self.serving_fs = None
logging.basicConfig(
level=logging.INFO,
format=('%(levelname)s|%(asctime)s'
'|%(pathname)s|%(lineno)d| %(message)s'),
datefmt='%Y-%m-%dT%H:%M:%S',
)
logging.getLogger().setLevel(logging.INFO)
# Train model
def train(self, training_df):
np.set_printoptions(precision=3)
train_data = training_df[[
x.split('.')[1] for x in TRAINING_FEATURES_SET
]].to_numpy()
train_data[:, len(train_data[0]) - 1] = 1
Y = training_df['price'].to_numpy()
x = np.linalg.lstsq(train_data, Y, rcond=0)[0]
m, b = x[:len(train_data[0]) - 1], x[len(train_data[0]) - 1]
self.m = m
self.b = b
return m, b
def predict(self, feature_id, feature_names):
logging.info('feature_id = %s', feature_id)
logging.info('feature_names = %s', feature_names)
if any([i is None
for i in [self.m, self.b, self.fs, self.serving_fs]]):
with open('simple_model.dat', 'r') as f:
model = json.load(f)
self.m = np.array(model.get('m', []))
self.b = float(model.get('b', 0))
_FEAST_CORE_URL = model['FEAST_CORE_URL']
_FEAST_SERVING_URL = model['FEAST_SERVING_URL']
_ENTITY_ID = model['ENTITY_ID']
logging.info('FEAST_CORE_URL: %s', _FEAST_CORE_URL)
logging.info('FEAST_SERVING_URL: %s', _FEAST_SERVING_URL)
logging.info('ENTITY_ID: %s', _ENTITY_ID)
logging.info('FEATURES_SET: %s', self.SERVING_FEATURE_SET)
self.fs = Client(core_url=_FEAST_CORE_URL,
serving_url=_FEAST_SERVING_URL,
verbose=True)
self.serving_fs = FeatureSet(entity=_ENTITY_ID,
features=self.SERVING_FEATURE_SET)
features = self.fs.get_serving_data(self.serving_fs,
entity_keys=[feature_id])
X = features.to_numpy()[0][1:]
logging.info('X: %s', str(X))
return [sum(self.m * X) + self.b]
def save_model(self, model_path):
"""Save the model to a json file."""
MODEL_FILE = 'simple_model.dat'
model = {
'm': self.m.tolist(),
'b': self.b,
'FEAST_CORE_URL': FEAST_CORE_URL,
'FEAST_SERVING_URL': FEAST_SERVING_URL,
'ENTITY_ID': ENTITY_ID,
}
logging.info('Saving model to %s', model_path)
with open(model_path, 'w+') as f:
json.dump(model, f)
def client(self, mocker):
cli = Client(core_url="some.uri", serving_url="some.serving.uri")
mocker.patch.object(cli, '_connect_core')
mocker.patch.object(cli, '_connect_serving')
return cli
def client():
return Client(verbose=True)
import test_feast
from feast.sdk.client import Client
# This main.py is for easier local debugging
# because running pytest often hides useful debugging output
if __name__ == "__main__":
feast_client = Client(verbose=True)
test_feast.TestFeastIntegration().test_end_to_end(client=feast_client)
def fs(self):
if not self._fs:
self._fs = Client(core_url=self.FEAST_CORE_URL,serving_url=self.FEAST_SERVING_URL, verbose=True)
return self._fs