Support Hyperparameter Optimization Jobs in Kubeflow SDK

[Andrey Velichkevich]

Hi Kubeflow Community,

Over the past few weeks, we’ve been exploring how to integrate support for HP optimization jobs into the Kubeflow SDK, enabling users to train and fine-tune LLMs while optimizing their parameters.

We are now designing a new client capable of handling this functionality, and we would greatly appreciate your feedback on our proposed design ideas, which are documented in this KEP:

https://docs.google.com/document/d/11Yzt6tIYzmCClWCr9tZ1RfTxqE1YObixy1tJ59ubtCk/edit?tab=t.0

Since we aim to include this feature in the next Kubeflow SDK release: v0.2, please provide your feedback by Friday, October 10th.

Thank you all for your time and input. I’m excited about the future of the Kubeflow SDK and its potential to streamline the AI practitioner experience on Kubernetes!

Regards,

Andrey