AutoModel vs. SetFitModel loading

Hi, good day,

I have pushed my model to a private repo and try and load with:

model = SetFitModel.from_pretrained(“Gustbern/btg_model”,use_auth_token=access_token)

to which I get a location error:

Repository Not Found for url:

This happens because my model is not in the public repository.

So I then try the AutoModel function, which works for private repos:

model = AutoModel.from_pretrained(“Gustbern/btg_model”,use_auth_token=access_token)

And it works. However, when I try to use the model for prediction it does not work, as it requires my list to be in a tensor format:

preds = model([“i loved the spiderman movie!”, “pineapple on pizza is the worst :face_vomiting:”])

I then get the following Error:

→ 536 input_shape = input_ids.size()
537 elif inputs_embeds is not None:
538 input_shape = inputs_embeds.size()[:-1]

AttributeError: ‘list’ object has no attribute ‘size’

All that to say that I cannot load SetFitModel from a private repository, or I should be able to change the address at least to load it, because the AutoModel does not perform proper conversion into tensors.

Can you help me solve this issue?