Skip to content

Latest commit

 

History

History
28 lines (20 loc) · 1.11 KB

README.md

File metadata and controls

28 lines (20 loc) · 1.11 KB
Raw

Simple example to use Kubeflow for model training and deployments.

This directory provides an example of using Feast, Kubeflow, and TFX Tensorflow Datavalidation.

Use the notebook feast-taxi-job.ipynb.

Deploy Kubeflow cluster

  1. Download kfctl CLI (v0.5.1) from kubeflow release
  2. Run the following command to deploy Kubeflow:
# Init using HEAD of v0.5-branch.
# This is needed because v0.5.1 doesn't include this fix:
# https://github.com/kubeflow/kubeflow/pull/3238
kfctl init {APP_NAME} --platform gcp --project {PROJECT} -V --version v0.5-branch

cd {APP_DIR}

kfctl generate all -V

kfctl apply all -V

Notebook settings

  1. Follow instructions on setting up notebook with UI: link
  2. upload Linear_Model.ipynb/deploy_with_fairing.py/LabelPrediction.py to notebook.

Misc

  • BASE_IMAGE is built with fairing_job/. Dockerfile in this folder has minimum required dependencies for fairing service.