[Model] Support Cohere2ForCausalLM (Cohere R7B)

[janimo]

Sliding window support as done in HuggingFace Transformers (>= 4.48.0.dev0) for the Cohere2ForCausalLM arch.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[DarkLight1337]

Could you point to the place where rotary embedding is disabled in v2? Looking at the HF implementation, I don't see why it should be disabled when sliding window is active.

[DarkLight1337]

Never mind, I confused

https://github.com/huggingface/transformers/blob/main/src/transformers/models/cohere2/modeling_cohere2.py#L387

with

https://github.com/huggingface/transformers/blob/main/src/transformers/models/cohere2/modeling_cohere2.py#L734

[DarkLight1337]

Thanks for adding support for this!

[DarkLight1337]

Please resolve the failing test.

[janimo]

I've added the model to the test list, but wonder if it's a problem if this arch is only in the unreleased version of transformers.

[DarkLight1337]

Can you update tests/models/test_initialization.py::test_can_initialize to skip this model if the transformers version is too old?

[DarkLight1337]

The model is only available in transformers v4.48 which has not been released yet.