What is kubeflow pipelines?

tudip-logo

Tudip

30 June 2020

Kubeflow

Kubeflow is a machine learning tool for kubernetes. Kubeflow is a Kubernetes-native platform for deploying, running and developing portable and scalable machine learning applications. Kubeflow helps in deploying loosely coupled microservices, it does not recreate any service but to provide a way to deploy applications to various infrastructure. It also provides users to quickly train, create and tune neural networks within Kubernetes for dynamic resource provisioning.

Kubeflow Pipelines

Kubeflow pipelines is a new component of kubernetes kubeflow that helps in deploying and building machine learning workflows for different docker container.with the kubeflow pipelines users can try different machine learning techniques and can go with the best one because kubeflow pipelines also provide fast and reliable experimentation.we don’t need to deploy kubeflow separately because kubeflow pipeline is a core and new component of kubeflow so it is automatically deployed during kubeflow deployment.

Components of Kubeflow Pipeline

  • Consist of an engine that helps in scheduling various steps of machine learning workflows.
  • For interacting with the system notebooks are used with SDK.For experimenting and managing jobs and running user interfaces is used.
  • For manipulating and defining pipelines and components an SDK is used.

The different building blocks can be composed to support useful and general machine learning workflow patterns. Kubeflow allows you to build a pipeline that supports distributed training, serving, ingestion, and feature pre-processing.

Features of Kubeflow 

  • Kubeflow pipeline are used for simplifying and enabling the orchestration of machine learning pipelines i.e end to end orchestration.
  • Because of its easy experimentation  you can try numerous ideas and techniques and manage your various trials and experiments.
  • Kubeflow pipeline is easy to use enabling you to re-use components and pipelines  and to quickly create end-to-end solutions without having to rebuild each time.

Ultimately, we want to have a set of simple manifests that give you an easy to use ML stack anywhere Kubeflow pipeline is already running, and that can self configure based on the cluster it deploys into.

Request a quote