pip install skilLightning version: SKIL does everything necessary for you under the hood, i.e. creating a production-grade model server with batteries included.
import skil
skil_server = skil.Skil()
model = skil.Model('your_model.h5')
model.deploy(scale=42)Slightly more extended example:
from skil import Skil, WorkSpace, Experiment, Model, Deployment
# Define and persist your model first
model_path = './tf_graph.pb'
# connect to your running skil instance
skil_server = Skil()
# create a workspace and an experiment in it
work_space = WorkSpace(skil_server)
experiment = Experiment(work_space)
# add your model to SKIL
model = Model(model_path, experiment)
# deploy and serve your model
deployment = Deployment(skil_server)
model.deploy(deployment, start_server=False)
model.serve()