Hyperparameter tuning for TensorFlow using Katib and Kubeflow
Who is this presentation for?
- Data scientists, machine learning engineers, and DevOps engineers
Level
Description
Neelima Mukiri and Meenakshi Kaushik provide an introduction to model building and optimization for TensorFlow in any Kubernetes environment. Katib is a scalable and flexible hyperparameter tuning framework and is tightly integrated with Kubernetes. It can be easily run on a laptop or in a distributed production deployment, and Katib jobs and configuration can be easily ported to any Kubernetes cluster.
You’ll go hands-on and set up a single-node Kubernetes on a laptop using VirtualBox. Neelima and Meenakshi provide a sample dataset and an example configuration and Kubeflow Pipeline that demonstrates hyperparameter tuning automation. They’ll walk you through Katib and Kubeflow, discussing functionality and usage, and explain how to port the tutorial to an enterprise environment for production deployment.
Prerequisite knowledge
- A basic understanding of machine learning (what a model is and how to evaluate and compare models) and Linux command line usage
Materials or downloads needed in advance
- A laptop with VirtualBox installed (please make sure to have VirtualBox installed BEFORE arriving onsite)
What you'll learn
- Gain hands-on experience creating and optimizing models using Katib on Kubeflow
- Understand hyperparameter tuning and how to transfer the system to an on-premises or cloud environment for production use cases
Neelima Mukiri
Cisco
Neelima Mukiri is a principal engineer in the Cloud Platform Solutions Group at Cisco, working on the architecture and development of Cisco’s Container Platform. Previously, she worked on the core virtualization layer at VMware and systems software in Samsung Electronics.
Meenakshi Kaushik
Cisco
Meenakshi Kaushik is a product manager for Cisco Container Platform, an enterprise-grade Kubernetes offering that supports GPU and Kubeflow for hybrid AI and ML workloads. Meenakshi is interested in the AI and ML space and is excited to see how the technology can enhance human well-being and productivity.
Comments on this page are now closed.
Presented by
Diamond Sponsor
Elite Sponsors
Gold Sponsor
Supporting Sponsors
Premier Exhibitors
Exhibitors
Innovators
Contact us
confreg@oreilly.com
For conference registration information and customer service
partners@oreilly.com
For more information on community discounts and trade opportunities with O’Reilly conferences
sponsorships@oreilly.com
For information on exhibiting or sponsoring a conference
pr@oreilly.com
For media/analyst press inquires
Comments
Got it.
Thank you, Neelima.
Hi Daniel,
Sure, any Kubernetes cluster which supports dynamic volume provisioning(ability to create persistent volumes) is fine.
I just got an email, that states that I need to bring laptop with preinstalled:
Vagrant 2.2.5 or later
VirtualBox 6.0.14 or later
Is it ok if I install the required software on GCP cloud instance and use laptop only as thin client?