LLaVA icon indicating copy to clipboard operation
LLaVA copied to clipboard

How to change CLIP to BiomedCLIP

Open YangQiuEric opened this issue 11 months ago • 2 comments

Question

I am doing medical image project, so how can we change CLIP to other models?

YangQiuEric avatar Feb 28 '24 22:02 YangQiuEric

change the vision_tower parameter

zsxm1998 avatar Feb 29 '24 07:02 zsxm1998

change the vision_tower parameter

change the vision_tower parameter

Hi, could you please be more specific? Which specific parts do you need to change?Thank you

20191864218 avatar Mar 01 '24 09:03 20191864218

@zsxm1998 @20191864218 I just change the vision_tower to microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224, and the finetune can be done. But the model_vqa.py has errors, can not recognize this model. I tried to download all the files from hugging face, but it does not have preprocess_config.json and config.json. How to fix?

YangQiuEric avatar Mar 06 '24 23:03 YangQiuEric

@zsxm1998 @20191864218 I just change the vision_tower to microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224, and the finetune can be done. But the model_vqa.py has errors, can not recognize this model. I tried to download all the files from hugging face, but it does not have preprocess_config.json and config.json. How to fix?

Have you solved the problem yet? @YangQiuEric

ATing0203 avatar Mar 18 '24 03:03 ATing0203

@zsxm1998 @20191864218 I just change the vision_tower to microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224, and the finetune can be done. But the model_vqa.py has errors, can not recognize this model. I tried to download all the files from hugging face, but it does not have preprocess_config.json and config.json. How to fix?

Any update @YangQiuEric ?

Abhiram-kandiyana avatar Apr 08 '24 13:04 Abhiram-kandiyana