/kfp-tekton

Experimental project exploring Tekton and KFP integration

Primary LanguagePythonApache License 2.0Apache-2.0

Kubeflow Pipelines and Tekton

Experimental project to bring Kubeflow Pipelines and Tekton together. The work is being driven in accordance with this evolving design doc specifications. Since this will evolve from experimental towards a more mature solution, we are keeping it currently separate from Kubeflow Pipeline repo

Tekton

The Tekton Pipelines project provides Kubernetes-style resources for declaring CI/CD-style pipelines. Tekton introduces several new CRDs including Task, Pipeline, TaskRun, and PipelineRun. A PipelineRun represents a single running instance of a Pipeline and is responsible for creating a Pod for each of its Tasks and as many containers within each Pod as it has Steps. Some tasks here will invariably require contributions back to Tekton. Please follow the community guidelines in Tekton repo.

Development

The current work is being split in four areas, namely Compiler, API (with corresponding backend), SDK (functionalities beyond the compilation, like uploading and launching a pipeline etc.) and UI. To get more details about what these different areas entail, please dive into the design doc. Currently we are progressing with first one in the list, namely Compiler to compile from KFP DSL to Tekton Yaml. We would love to have teams pick up other major areas like API/UI etc. and/or contribute to make the Tekton Compiler functions and feature complete.

To get started with contributing to KFP Tekton Compiler, please follow these instructions, as well as look at open issues on the repo

KFP and Tekton: Deliverables

Please note that all these deliverables are work in progress, and at an early stage of exploration and execution. We are using Kubeflow Pipelines v0.2.2 and Tekton v0.11.0 for these experiments currently.

  1. KFP, Argo and Tekton Comparision
  2. Equivalent Argo and Tekton Yaml for Flip Coin Sample from Kubeflow Pipeline
  3. KFP Compiler for Tekton

CD Foundation

The work here is being tracked under the CD Foundation MLOps Sig. If you are interested in joining, please see the instructions here

Additional Reference Materials: KFP and TFX

  1. Kubeflow Pipelines-TFX Pipelines
  2. Kubeflow Pipelines-TFX Pipelines Talk at Tensorflow World
  3. Kubeflow Pipelines-TFX Pipelines RFC