feat(trainer): Support NPU labels in TrainJob device

[sujalshah-bit]

What this PR does / why we need it:
Add support for NPU resource labels in container resource limits,
resolving the existing TODO to support additional accelerator types
(e.g., NPUs).

• Detect NPU resource keys using strict /npu suffix matching
• Extract and return the corresponding resource value
• Fails fast when multiple NPU resource types are present
• Maintains consistency with existing GPU/MIG validation behavior

Fixes NONE

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 andreyvelich 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

[github-actions]

🎉 Welcome to the Kubeflow SDK! 🎉

Thanks for opening your first PR! We're happy to have you as part of our community 🚀

Here's what happens next:

• If you haven't already, please check out our Contributing Guide for repo-specific guidelines and the Kubeflow Contributor Guide for general community standards
• Our team will review your PR soon! cc @kubeflow/kubeflow-sdk-team

Join the community:

• Slack: Join our #kubeflow-ml-experience and #kubeflow-trainer Slack channels
• Meetings: Attend the Kubeflow SDK and ML Experience bi-weekly meetings

Feel free to ask questions in the comments if you need any help or clarification!
Thanks again for contributing to Kubeflow! 🙏

[Copilot]

Pull request overview

Adds NPU extended-resource detection to the Kubernetes backend so TrainJob device metadata can be derived from container resource limits when an NPU resource key is present.

Changes:

• Introduces an NPU-related resource constant in constants.py.
• Extends get_container_devices() to detect */npu resource keys and fail fast on multiple NPU types.
• Adds unit tests covering single-NPU and multiple-NPU-key scenarios.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
kubeflow/trainer/constants/constants.py	Adds an NPU-related constant for resource identification.
kubeflow/trainer/backends/kubernetes/utils.py	Adds NPU resource key detection and validation in device extraction.
kubeflow/trainer/backends/kubernetes/utils_test.py	Adds test cases validating NPU detection and multi-key failure behavior.

[Copilot]

NPU_LABEL is defined as the suffix string "/npu" (not a full resource label like the other *_LABEL constants) and is currently unused, which is likely to confuse future callers; rename it to something like NPU_RESOURCE_SUFFIX/NPU_SUFFIX (or document it explicitly as a suffix) and update usages accordingly.

[Copilot]

The NPU detection uses a hard-coded string literal "/npu" even though an NPU constant was added in constants; use the constant here (and/or rename it to a suffix constant) so the matching rule is defined in one place and can’t drift.

Suggested change

	elif npu_keys := [k for k in resources.limits if k.endswith("/npu")]:
	elif npu_keys := [k for k in resources.limits if k.endswith(constants.NPU_SUFFIX)]: