def create_fare(key, pickup_datetime, pickup_longitude, pickup_latitude,
dropoff_longitude, dropoff_latitude, passenger_count):
# key = "2013-07-06 17:18:00.000000119"
# pickup_datetime = "2013-07-06 17:18:00 UTC"
# pickup_longitude = "-73.950655"
# pickup_latitude = "40.783282"
# dropoff_longitude = "-73.984365"
# dropoff_latitude = "40.769802"
# passenger_count = "1"
# build X ⚠️ beware to the order of the parameters ⚠️
X = pd.DataFrame(
dict(key=[key],
pickup_datetime=[pickup_datetime],
pickup_longitude=[float(pickup_longitude)],
pickup_latitude=[float(pickup_latitude)],
dropoff_longitude=[float(dropoff_longitude)],
dropoff_latitude=[float(dropoff_latitude)],
passenger_count=[int(passenger_count)]))
# get model from GCP
pipeline = download_model()
# pipeline = joblib.load('../data/model.joblib')
# make prediction
results = pipeline.predict(X)
# convert response from numpy to python type
pred = float(results[0])
return dict(prediction=pred)
def set_model():
# inputs = request.get_json()
# model_dir = inputs["model_directory"]
# pipeline_def["pipeline"] = download_model(model_directory=model_dir, rm=True)
# pipeline_def["pipeline"] = download_model()
# pipeline_def["from_gcp"] = True
pipeline = download_model()
return {"reponse": f"correctly updated model GCP"}
def set_model():
inputs = request.get_json()
# model_dir = FOLDER_MODEL_PATH
# model_dir = inputs["model_directory"]
# pipeline_def["pipeline"] = download_model(model_directory=model_dir, rm=True)
pipeline_def["pipeline"] = download_model(rm=True)
pipeline_def["from_gcp"] = True
return {
"reponse": f"correctly got model from {model_dir} directory on GCP"
}
input.setdefault(k, v)
formated_input = {
"key": str(input["key"]),
"pickup_datetime": str(input["pickup_datetime"]),
"pickup_longitude": float(input["pickup_longitude"]),
"pickup_latitude": float(input["pickup_latitude"]),
"dropoff_longitude": float(input["dropoff_longitude"]),
"dropoff_latitude": float(input["dropoff_latitude"]),
"passenger_count": float(input["passenger_count"])
}
return formated_input
# pipeline_def = {'pipeline': download_model(),
# 'from_gcp': Terue}
pipeline = download_model()
@app.route('/')
def index():
return 'OK'
@app.route('/predict_fare', methods=['GET'])
def predict_fare():
# get request arguments
key = request.args.get('key')
pickup_datetime = request.args.get('pickup_datetime')
pickup_longitude = float(request.args.get('pickup_longitude'))
pickup_latitude = float(request.args.get('pickup_latitude'))
def set_model():
inputs = request.get_json()
estimator = inputs["estimator"]
pipeline_def["pipeline"] = download_model(estimator="xgboost", rm=True)
pipeline_def["from_gcp"] = True
return {"reponse": f"correctly got model from {estimator} directory on GCP"}
def get_model(self):
if self.local:
pipeline = joblib.load(self.PATH_TO_LOCAL_MODEL)
else:
pipeline = download_model()
return pipeline