kubeflow / kubeflow/trainer

KEP-3562: [OptimizationJob] Refactor Katib gRPC protocol contracts for stateless suggestion services

Open
#3,796 7 comments 0 reactions 1 assignee Claimed by @saivedant169 View on GitHub
area/hpo kind/feature kind/plan-kep
Dominant language
Go
Stars
2.2k
Forks
1.1k
Avg merge
3d 22h
Merged PRs (30d)
39

Description

### What you would like to be added?

Context:
Historically, Katib relied on a specific gRPC interface for suggestion containers tied to its DB-backed architecture.

Problem:
`OptimizationJob` in `kubeflow/trainer` relies on a lightweight, stateless gRPC model ([KEP-3562](https://github.com/kubeflow/trainer/pull/3565)). The old Katib gRPC definitions carry unnecessary legacy state assumptions.

Future Goals:

Refactor the gRPC proto definitions (`api/v1alpha1/`) to cleanly decouple suggestion generation from persistence.

Establish a clean, lightweight protocol contract for requesting hyperparameter combinations and reporting trial metrics.

Relates to KEP-3562 Master Tracking [Issue](https://github.com/kubeflow/trainer/issues/3562).

### Why is this needed?

Refactoring the gRPC protocol simplifies communication, reduces payload overhead, and allows suggestion backends to operate in ephemeral container lifecycle models without DB overhead.

### Love this feature?

Give it a 👍 We prioritize the features with most 👍

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.