Kubeflow

Introduction

Kubeflow is a free and open-source machine learning platform for enabling and automating the end to end machine learning pipeline. It facilitates in orchestrating complicated workflows on Kubernetes.

It started as just a simpler way to run TensorFlow jobs on Kubernetes. It has expanded to be a multi-architecture, multi-cloud framework for running machine learning workflows. Kubeflow started being based on Google’s method to deploy TensorFlow models called TensorFlow Extended.

The main objective of Kubeflow is to ease the use of scaling of machine learning (ML) models and make the operationalization of models simple. In short it provides the following

It makes deployments of machine learning workflows on Kubernetes simple, portable and scalable and is a way to deploy state of the art open-source systems for ML to diverse infrastructures.

Since it works with Kubernetes, kubeflow will work anywhere where you are running Kubernetes. In simple words it is a multi-architecture, multi-cloud framework for running end-to-end machine learning workflows.

The following diagram shows the workflow stages in a traditional Machine Learning life cycle.

The following diagram shows the phases or stages in a machine learning model development life cycle with Kubeflow.

Related posts:

Faz algum tempo vi um estudo que dizia que os “aplicativos de carona” (Uber, Cabify, 99, etc.) pioravam o trânsito. Achei estranho, fui dar uma olhada mais a fundo e descobri que o argumento era que…

The first time she slept with him, he said, “I’ve wanted this for so long.” And, “this is so amazing.” And, “you are perfect.” You wiped yourself off after sex and didn’t even offer her a towel. Not…

Spencer Enos