Support Local Execution for Optimizer

[kramaranya]

What you would like to be added?

With the successful release of local execution mode for the TrainingClient, it would be great to extend this capability to the OptimizerClient for optimizing TrainJobs locally without requiring Kubernetes.

from kubeflow.optimizer import OptimizerClient, ContainerBackendConfig, Search, TrialConfig

client = OptimizerClient(backend_config=ContainerBackendConfig())

# Create OptimizationJob with the same template
optimization_id = client.optimize(
    trial_template=template,
    trial_config=TrialConfig(num_trials=10, parallel_trials=2),
    search_space={
        "learning_rate": Search.loguniform(0.001, 0.1),
        "num_epochs": Search.choice([5, 10, 15]),
    },
)

print(f"OptimizationJob created: {optimization_id}")

Running optimization jobs locally would help Data Scientists quickly prototype and validate their tuning strategies before deploying to production without needing to deal with cluster complexity

Why is this needed?

Delivering local execution mode across different clients in Kubeflow SDK, including for optimization models will significantly improve usability.

Love this feature?

Give it a 👍 We prioritize the features with most 👍

[kramaranya]

/area optimizer
/remove-label lifecycle/needs-triage

[adity1raut]

/assign

[adity1raut]

Hey @NarayanaSabari
should have asked first if I was already working on this issue before directly opening a PR. Since you’ve already created it, let’s go forward with it now.

[NarayanaSabari]

@adity1raut Thanks for pointing that out. I should have checked with you before starting work on the issue. I appreciate you being willing to move forward with the PR, and I’ll keep coordination in mind going forward.

[adity1raut]

/unassign