Kubeflow Dojo

Repository to hold code, instructions, demos and pointers to presentation assets for Kubeflow Dojo. In this Dojo, we are going to address how to make it easy for everyone to develop, deploy, and manage portable, scalable ML everywhere and support the full lifecycle Machine Learning using Kubeflow. We are going to discuss how to deploy and manage Kubeflow, and detail how to enable distributed training of models, model serving, canary rollouts, drift detection, model explainability, metadata management, pipelines and others.

To enable developers who want to use it at their own pace, pre-recorded sessions are uploaded. For those who prefer live interactions, we will be running the sessions and workshops live.

Read about IBM's journey with Kubeflow here

Virtual Schedule and Prereqs

When: Virtual. Day One and Day Two

Day One End to end presentation and walk through on Kubeflow, followed by detailed deep dive and hands on deploying Kubeflow to Minikube and/or IBM Cloud Kubernetes Service (IKS).

Day Two Deeper dive sessions and demos on Kubeflow Pipelines, Kubeflow Serving (KFServing), Distributed Training Operators and HPO, and Kubeflow PR workflow

The Kubeflow Slack workspace is kubeflow.slack.com. To join, click this invitation to Kubeflow Slack workspace. There is a channel #kubeflow-dojo created for the workshop.

Prereqs: Please look at prereqs here

Note: For Pipelines, we are using Kubeflow Pipelines with Tekton in this workshop

Date: Day One

|

Recordings for live sessions: Kubeflow Dojo Day One Playlist

Date: Day Two

Recordings for live sessions: Kubeflow Dojo Day Two Playlist