Code/templates release for the LLaVA-OneVision model

[agadetsky]

Hi!

Can you release the evaluation code for LLaVA-OneVision to reproduce the results from the paper?
Or can you share the templates used for LLaVa-OneVision?
Did you use multi-turn conversation with chat templates, i.e., providing few-shot examples as the conversation between user and assistant within the corresponding role headers, or you didn't use chat template?

Thank you and looking forward to your response.

[ys-zong]

Hi, I've added the LLaVA-OneVision inference code in this commit. I used their default chat template in their github.