finetuing scripts

[yepzhang]

Thanks for your great work!
I want to finetune the model on my dataset, can you share a fine-tune scripts for further development?
Thanks!

[Uoops]

Shout out for a fine-tune script

[jwwang424]

As the model was trained using our in-house training architecture, we cannot directly make our finetune code public. It will cost sometime for us to rewrite an open-source available finetune code.

However, as our modeling_tarsier.py is copied with minor changes from modeling_llava.py, and there is no special trick during all our training procedure, users can refer to llava's finetune code before we release our finetune code.