Interact

Initialization

The kaos ML platform is fully functional when initialized with a running endpoint (see below). See Deploying Infrastructure for instructions on generating a running endpoint.

kaos init -e <running_endpoint>

This step is mandatory to ensure the end user (i.e. Data Scientists) is connected to kaos. Initialization can be thought of as the "glue" binding the backend and the command line interface (CLI). Once again, the following conceptual overview indicates how kaos init connects the end user the entire kaos backend.

General Usage

The kaos CLI consists of the following top-level commands.

  • kaos init

    • Initialize the kaos environment

  • kaos template

    • Retrieves templates for getting started

  • kaos workspace

    • Organize and separate ML working environments

  • kaos notebook

    • Deploy hosted notebooks for experimentation and building ML models

  • kaos train

    • Create training jobs for training ML models with ready-made and split features

  • kaos serve

    • Deploy an endpoint with a trained model

    Usage and sub-level commands are further detailed when running --help. For example, retrieving the mnist template for usage involves the following command:

kaos template get --name mnist

The same conceptual overview is presented once again to highlight their "interaction" with the kaos backend.