Not working with CLIP encoder

[arditobryan]

This is the reference model on Huggingface:
https://huggingface.co/sentence-transformers/clip-ViT-B-32

from model2vec.distill import distill

# Choose a Sentence Transformer model
model_name = "sentence-transformers/clip-ViT-B-32"

# Distill the model
m2v_model = distill(model_name=model_name, pca_dims=256)

ERROR:
---------------------------------------------------------------------------
HTTPError                                 Traceback (most recent call last)
File c:\Users\ardit\miniconda3\envs\env1\Lib\site-packages\huggingface_hub\utils\_errors.py:304, in hf_raise_for_status(response, endpoint_name)
    [303](file:///C:/Users/ardit/miniconda3/envs/env1/Lib/site-packages/huggingface_hub/utils/_errors.py:303) try:
--> [304](file:///C:/Users/ardit/miniconda3/envs/env1/Lib/site-packages/huggingface_hub/utils/_errors.py:304)     response.raise_for_status()
    [305](file:///C:/Users/ardit/miniconda3/envs/env1/Lib/site-packages/huggingface_hub/utils/_errors.py:305) except HTTPError as e:

File c:\Users\ardit\miniconda3\envs\env1\Lib\site-packages\requests\models.py:1024, in Response.raise_for_status(self)
   [1023](file:///C:/Users/ardit/miniconda3/envs/env1/Lib/site-packages/requests/models.py:1023) if http_error_msg:
-> [1024](file:///C:/Users/ardit/miniconda3/envs/env1/Lib/site-packages/requests/models.py:1024)     raise HTTPError(http_error_msg, response=self)

HTTPError: 404 Client Error: Not Found for url: https://huggingface.co/sentence-transformers/clip-ViT-B-32/resolve/main/tokenizer.json

However, this is an Image encoder, is it possible that due the nature of distillation only text encoders are supported?
Thank you, amazing library

[stephantul]

Hey @arditobryan ,

Thanks for the report. Yes, we only support sentence transformers-like models. Maybe we should be more clear about this. Only text encoders are supported. Let me know if this suffices.

[stephantul]

Hey! I added some extra lines to the README, so I'll close this. Let me know if there's anything still confusing.