chore: Trainer: Specialized Trainers

[szaher]

What this PR does / why we need it:

Which issue(s) this PR fixes (optional, in Fixes #<issue number>, #<issue number>, ... format, will close the issue(s) when PR gets merged):

Fixes #

Checklist:

• Docs included if any changes are user facing

[google-oss-prow]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign kramaranya for approval. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[Copilot]

Pull request overview

This PR adds a comprehensive design proposal for specialized trainer abstractions and a RuntimeConfig dataclass to the Kubeflow SDK. The proposal addresses current limitations in the SDK's trainer subsystem by introducing framework-aware trainer classes that bridge the gap between the generic CustomTrainer and the highly specific BuiltinTrainer.

Changes:

• Adds a detailed design proposal document describing a new BaseTrainer abstract interface and specialized framework trainers (TorchTrainer, MPITrainer, JAXTrainer, XGBoostTrainer)
• Proposes a RuntimeConfig dataclass to cleanly separate runtime environment settings from training logic
• Includes comprehensive documentation covering motivation, design details, API examples, migration strategy, test plan, and alternatives considered

[Copilot]

The company name should be spelled "Hugging Face" (with a space) rather than "HuggingFace" throughout the document. This applies to references in text and comments, though the class name "HuggingFaceTrainer" would be correct as Python class names don't use spaces.

Suggested change

	and interface. Concrete Tier 2 implementations (HuggingFace, DeepSpeed, Unsloth,
	and interface. Concrete Tier 2 implementations (Hugging Face, DeepSpeed, Unsloth,

[Copilot]

The company name should be spelled "Hugging Face" (with a space) rather than "HuggingFace" in the comment and docstring text.

Suggested change

	# Example: future HuggingFaceTrainer (NOT part of this proposal's implementation scope)
	@dataclass
	class TransformersTrainer(BaseTrainer):
	"""Trainer for HuggingFace Transformers training.
	Wraps HuggingFace's Trainer API and maps to a PyTorch runtime.
	# Example: future Hugging Face trainer (NOT part of this proposal's implementation scope)
	@dataclass
	class TransformersTrainer(BaseTrainer):
	"""Trainer for Hugging Face Transformers training.
	Wraps Hugging Face's Trainer API and maps to a PyTorch runtime.

[Copilot]

The company name should be spelled "Hugging Face" (with a space) rather than "HuggingFace" in the diagram text.

Suggested change

	HuggingFace DeepSpeed
	Hugging Face DeepSpeed