def create_prediction_microservice(self,pipeline_folder,model_name):
"""
Create a prediction Flask microservice app
Parameters
----------
pipeline_folder : str
location of pipeline
model_name : str
model name to use for this pipeline
"""
app = Flask(__name__)
rint = random.randint(1,999999)
pw = sutl.PipelineWrapper(work_folder='/tmp/pl_'+str(rint),aws_key=self.aws_key,aws_secret=self.aws_secret)
pipeline = pw.load_pipeline(pipeline_folder)
app.config["seldon_pipeline_wrapper"] = pw
app.config["seldon_pipeline"] = pipeline
app.config["seldon_model_name"] = model_name
app.register_blueprint(predict_blueprint)
# other setup tasks
return app
def create_prediction_rpc_microservice(self, pipeline_folder, model_name,
custom_data_handler):
rint = random.randint(1, 999999)
pw = sutl.PipelineWrapper(work_folder='/tmp/pl_' + str(rint),
aws_key=self.aws_key,
aws_secret=self.aws_secret)
pipeline = pw.load_pipeline(pipeline_folder)
server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
seldon_pb2.add_ClassifierServicer_to_server(
RpcClassifier(pipeline, model_name, custom_data_handler), server)
server.add_insecure_port('[::]:5000')
server.start()
try:
while True:
time.sleep(_ONE_DAY_IN_SECONDS)
except KeyboardInterrupt:
server.stop(0)
# print(sess.run(accuracy, feed_dict = {x: mnist.test.images, y_:mnist.test.labels}))
tfw = TensorFlowWrapper(sess,
tf_input=x,
tf_output=y_conv,
tf_constants=[(keep_prob, 1.0)],
target="y",
target_readable="class",
excluded=['class'])
return Pipeline([('deep_classifier', tfw)])
if __name__ == '__main__':
parser = argparse.ArgumentParser(prog='pipeline_example')
parser.add_argument('-m',
'--model',
help='model output folder',
required=True)
parser.add_argument('-l', '--load', help='Load pretrained model from file')
args = parser.parse_args()
p = create_pipeline(args.load)
pw = sutl.PipelineWrapper()
pw.save_pipeline(p, args.model)