Upstream model implementation to vLLM

[ywang96]

Hello! This is Roger from the vLLM team!

It's great to see you are using the OOT model registration functionality to have this model supported on vLLM! Do you plan to make a PR in the upstream repository so that we can have this model officially supported?

If you don't have the bandwidth to make a PR, I'm happy to make it and put relevant team members as co authors - Let me know!

[aria-hacker]

@ywang96 Thanks for reaching out! I'd really appreciate your help with making the PR. I don't have much bandwidth right now, so your offer to create the PR would be fantastic.

Let me know what information you need from me to get started. I'm happy to provide any details about the implementation or review the PR once it's ready.

[xffxff]

I'm working on this. I've submitted a pull request vllm-project/vllm#10514