Overview

The Kubeflow project is dedicated to making deployments of machine learning (ML) workflows on Kubernetes simple, portable and scalable. The goal is not to recreate other services, but to provide a straightforward way to deploy best-of-breed open-source systems for ML to diverse infrastructures. Anywhere you are running Kubernetes, you should be able to run Kubeflow.

Stages

Defining ML Workflows including:

• Managing data
• Running notebooks
• Training Models
• Serving Models

Creating a Notebook

From the Notebook menu, create a new notebook server.

Components

Central Dashboard

The central user interface (UI) in Kubeflow

Notebook Servers

Using Jupyter notebooks in Kubeflow

Kubeflow Pipelines

A powerful platform for building end to end ML workflows. Pipelines allow you to build a set of steps to handle everything from collecting data to serving your model.

KFServing

Kubeflow model deployment and serving toolkit

Katib

Katib is a Kubernetes-native project for automated machine learning (AutoML). Katib supports hyperparameter tuning, early stopping and neural architecture search (NAS). 

Training Operators

Training of ML models in Kubeflow through operators such as Tensorflow and Pytorch.

Multi-Tenancy

Multi-user isolation and identity access management (IAM)

KFServing

Jupyter Notebooks

Defacto standard for data scientists for performing rapid data analysis.

Kubeflow Pipelines

A pipeline component is one step in the workflow that performs one specific task.

Takes inputs and produces outputs.

Hyperparameter Tuning with Katib 

Katib is a Kubernetes-native project for automated machine learning (AutoML). Katib supports hyperparameter tuning, early stopping and neural architecture search (NAS).