What is the training scripts for llava-interleave ?

Hi, folks, 
i just wonder does anyone had trained the llava-interleave before **by using LLaVA-NeXT repo** ?

Currently, i have prepared m4 interleaved dataset and download both image & video data. I'm ends up with 2 json (m4_instruct_annotation.json, m4_instruct_video.json). **However, i can not find any training script for finetune llava-interleave?**

Any suggestion will be appreciated!!