Question about fine-tuning LLaVA-OneVision-2

[wnn2000]

Hi, thanks for your great work and for releasing the code.

I would like to continue fine-tuning LLaVA-OneVision-2, but I am a bit confused about the training code structure in this repository.

It seems that there are two different training implementations:

• aiak_training_llm
• transformers_impl

Could you please clarify which one should be used for fine-tuning LLaVA-OneVision-2 (image/video VQA tasks)?

Thanks in advance!

@anxiangsir