Appendix C. Creating an HPO service with Kubeflow Katib
We will introduce you to an open source hyperparameter optimization (HPO) service—Kubeflow Katib—that addresses virtually all the HPO requirements we discussed in chapter 5. We strongly recommend that you consider adopting Katib before building your HPO service. Along with showing you how to use Katib, we will also cover its system design and its codebase to make you comfortable with the open source service.
As a member of the Kubeflow family, Katib is a cloud-native, scalable, and production-ready hyperparameter optimization system. In addition, Katib is agnostic to the machine learning framework or programming language. Also, Katib is written in Go, takes a Kubernetes-native approach, ...
Get Designing Deep Learning Systems now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
