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_predict(get_fake_model):
_real_time = grpc_testing.strict_real_time()
servicer = PredictionServiceServicer(models={'test': get_fake_model})
descriptors_to_servicers = {PREDICT_SERVICE: servicer}
_real_time_server = grpc_testing.server_from_dictionary(
descriptors_to_servicers, _real_time)
return _real_time_server