Here I modified original LLaVA code to adapt Qwen2 as base language model. Now I have updated starting training scripts under script\v1_5 folder, for both pretraining phase and fine-tuning phase. You can start pretraining llava-qwen2 by following command,

bash scripts/v1_5/pretrain_qwen2.sh

By same mean, you start fine-tuning llava-qwen2 by this command,

bash scripts/v1_5/finetune_qwen2.sh

For detailed preparation and other questions, please refer original repo.

I would love to see you have any modification and creation on my repo! A pull request is all you need.

Thanks original creator for LLaVA project, and other maintainer of this project.