Add support for HF Clip

#10
by G-AshwinKumar - opened

Would it be possible to add support for HF Clip format. So that we can directly load the model using the Hugging face transformers library without using open clip? A possible example is https://huggingface.co/laion/CLIP-ViT-L-14-DataComp.XL-s13B-b90K/tree/main

That would be of great help for fine-tuning.

any update?

any update?

Hi all @AbhiKand @ERICQIU @G-AshwinKumar @Jayantjivi @Passionsu ,
I translated from the open-clip and timm based checkpoint to a HF-friendly implementation
You can now use it through:

CLIP_DIR = "/path/to/BiomedCLIP"
model = AutoModel.from_pretrained(CLIP_DIR, trust_remote_code=True)
processor = AutoProcessor.from_pretrained(CLIP_DIR, trust_remote_code=True)

with model and files in this PR:
https://huggingface.co/microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224/discussions/20

Sign up or log in to comment