KEP-2401: Create LLM Training Runtimes for Llama 3.2 model family

[Electronic-Waste]

What you would like to be added?

As we discussed in https://github.com/kubeflow/trainer/pull/2410#discussion_r1977828187, we should support Llama 3.2 model family in the initial stage.

This will resolve: https://github.com/kubeflow/trainer/issues/2212

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

Why is this needed?

Scheduled in https://github.com/kubeflow/trainer/pull/2410

Love this feature?

Give it a 👍 We prioritize the features with most 👍

[Electronic-Waste]

/area llm

[Garvit-77]

Hey @Electronic-Waste , I would like to contribute on this , could you assign me this issue ?

[Electronic-Waste]

@Garvit-77 In fact, this issue is pending and we need to accomplish other issues first:)

[Garvit-77]

Thanks for the update! Surely @Electronic-Waste, I would like to be on the same note and according to the order(priority) of issues in #2401

[Electronic-Waste]

It's sequential, we need to first accomplish: https://github.com/kubeflow/trainer/pull/2522. I would appreciate it if you could help with the review together with @kubeflow/wg-training-leads @astefanutti :)

[Garvit-77]

Thanks @Electronic-Waste! I'd like to contribute to the review of #2522. While I'm not as experienced with [reviewing as a specific area the leads might excel in, e.g., the overall architecture], I can focus on [mention a specific area where you're comfortable, e.g., code style, documentation, testing a particular feature]. Let me know if that's helpful.

[Electronic-Waste]

/assign

[Electronic-Waste]

/milestone v2.0