Tempo is a python SDK for data scientists to help them move their models to production. It has 4 core goals:
Data science friendly.
Custom python inference components.
Powerful orchestration logic.
A simple model example is shown below. A data scientist need only fill some core details about their trained model.
Tempo allows you to combine business logic as custom python with models served on optimized servers.
In the example below:
Train your sklearn model
Train your xgboost model
Create python logic to orchestrate the two
The Tempo code for the above example is shown below:
The steps to use tempo would be:
Local training and unit tests
Local tests of inference on Docker
Push to production runtime, e.g. Kubernetes with Seldon Core