[Core feature] Support Kubeflow Trainer 2.X

[bastitx]

Motivation: Why do you think this is important?

In July 2025, Kubeflow Trainer 2.0 was released, which introduced some breaking changes to the CRD interface. Currently, the pytorch plugin is not working with the new interface. Since v1 of the trainer is not maintained anymore, it would be great to add support for the v2 interface.

Goal: What should the final outcome look like, ideally?

Flyte's pytorch plugin supports kubeflow trainer 2.0.

Describe alternatives you've considered

Currently, the mitigation is to use the older kubeflow trainer 1.9 version. Though this will not be desirable long-term.

Propose: Link/Inline OR Additional context

No response

Are you sure this issue hasn't been raised already?

• Yes

Have you read the Code of Conduct?

• Yes

[runllm]

Support for Kubeflow Trainer 2.x (the new unified Kubeflow Training Operator) is not yet available in Flyte's PyTorch plugin. The current Flyte PyTorch plugin is built for the v1 interface and relies on the older CRDs, as confirmed by the plugin documentation and related discussions. Migration to v2 would require updates to FlyteIDL, flyteplugins, and flytekit to match the new CRD schema and API changes introduced in Kubeflow Trainer 2.x. The only current workaround is to use Kubeflow Trainer v1.x, but this is not a long-term solution since v1 is no longer maintained. There is community interest and some discussion about this migration, but no official support or timeline is documented yet.

For more context, see the discussion in Flyte GitHub Issue #1375, which outlines the need for supporting the new operator and the steps required. The PyTorch plugin documentation and plugin list also confirm current support is for the v1 operator. Would you like more detail on the technical changes required for Flyte to support Kubeflow Trainer 2.x?

Sources:

• Flyte GitHub Issue #1375
• PyTorch plugin documentation
• plugin list

---

This is an AI-generated response and your feedback is appreciated! Please leave a 👍 if this is helpful and 👎 if it is not.

Mention @RunLLM to ask follow-up questions. See runllm.com for more!

[lowc1012]

#take