- Install ray 0.8.4:
pip install ray==0.8.4
- Start the ray cluster
ray start --head
- Run the training scrip to generate model file.
python 3_train_tf_model.py
- Run the deployment scripts to deploy to the running cluster.
Open two terminal, in one terminal:
python 1_init.py # deploys a hello world function, and queries available routes in while True loop
In the other terminal:
python 2_counter.py # deploys a counter
python 4_serve_tf_model.py # deploys a TF model
python 5_serve_tf_model_v2.py # deploys the second TF model sharing the same /mnist route