Clipper-test

Repository for testing dynamic inference with Clipper Clipper

Dynamic Inference (Serving Models in Production via REST)

Inference is the term used to describe the process of using a pre-trained model to make predictions for unseen data. Dynamic Inference is the term used to describe making predictions on demand, using a server.

This notebook is a walk through for how to serve a machine learning model using a low latency prediction servering system called clipper.ai. clipper can be hosted on your favorite cloud provider or on-premise.

Overview

• Model training
• Clipper cluster creation
• App creation & model deployment
• Model query (single row, multiple rows) via Python requests & curl
• Model versioning update
• Model versioning rollback
• Model replication

References:

• clipper.ai documentation

• clipper @github