def on_get(self, req, resp, model_name, requested_version=0):
logger.debug("MODEL_STATUS, get request: {}, {}".format(
model_name, requested_version))
valid_model_status = check_availability_of_requested_status(
models=self.models,
requested_version=requested_version,
model_name=model_name)
if not valid_model_status:
resp.status = falcon.HTTP_NOT_FOUND
logger.debug("MODEL_STATUS, invalid model spec from request")
err_out_json = {
'error': WRONG_MODEL_SPEC.format(model_name, requested_version)
}
resp.body = json.dumps(err_out_json)
return
requested_version = int(requested_version)
response = get_model_status_pb2.GetModelStatusResponse()
if requested_version:
version_status = self.models[model_name].versions_statuses[
requested_version]
ModelServiceServicer.add_status_to_response(
version_status, response)
else:
for version_status in self.models[model_name].versions_statuses. \
values():
ModelServiceServicer.add_status_to_response(
version_status, response)
logger.debug("MODEL_STATUS created a response for {} - {}".format(
model_name, requested_version))
resp.status = falcon.HTTP_200
resp.body = MessageToJson(response,
including_default_value_fields=True)
def serve(models, max_workers: int = 1, port: int = 9000):
server = grpc.server(futures.ThreadPoolExecutor(max_workers=max_workers),
options=[('grpc.max_send_message_length', GIGABYTE),
('grpc.max_receive_message_length', GIGABYTE)
])
prediction_service_pb2_grpc.add_PredictionServiceServicer_to_server(
PredictionServiceServicer(models=models), server)
model_service_pb2_grpc.add_ModelServiceServicer_to_server(
ModelServiceServicer(models=models), server)
server.add_insecure_port('[::]:{}'.format(port))
server.start()
logger.info("Server listens on port {port} and will be "
"serving models: {models}".format(port=port,
models=list(models.keys())))
try:
while True:
if FILE_SYSTEM_POLL_WAIT_SECONDS > 0:
time.sleep(FILE_SYSTEM_POLL_WAIT_SECONDS)
for model in models:
models[model].update()
else:
time.sleep(_ONE_DAY_IN_SECONDS)
except KeyboardInterrupt:
server.stop(0)
sys.exit(0)
def get_grpc_service_for_model_status(get_fake_model):
_real_time = grpc_testing.strict_real_time()
servicer = ModelServiceServicer(models={'test': get_fake_model})
descriptors_to_servicers = {MODEL_SERVICE: servicer}
_real_time_server = grpc_testing.server_from_dictionary(
descriptors_to_servicers, _real_time)
return _real_time_server