question about run LLava tuning

[yangguoming]

When I want to run [ visual_instruction_tuning] , as you say in the README.md, "Download the pretrained projector weights from liuhaotian/llava-v1.5-mlp2x-336px-pretrain-vicuna-7b-v1.5 and put it under ./checkpoints/ " .
I have some question:
1st : The llava-v1.5-mlp2x-336px-pretrain-vicuna-7b-v1.5 only has a 42MB mm_projector.bin, the other pretrained weigths such as pytorch_model-00001-of-00002.bin, Where should I put it after I download it?
2nd : I guess I need git clone https://huggingface.co/liuhaotian/llava-v1.5-7b/tree/main, the pretrained model should put under ./checkpoint/, Is it right? and rename this folder to "llava-v1.5-7b-pretrain" , Because it is set up this way in your shell script [Dora_7b.sh
line17].
3rd : the 42MB mm_projector.bin in "llava-v1.5-7b" of 2nd gitted should be replaced by the 42MB mm_projector.bin of 1st?