I subscribe huggingface for private model repository.
I upload my model but how can I use my model?
I can`t use it because when I use it with model name(eg, kykim/bert-kor-large) for fine-tuning, command line shows that this is not listed in huggingface.co/models.
@kykim Thanks for your question/feedback, this is helpful!
Can you copy the command you’re launching here?
To instantiate a private model from
transformers you need to add a
use_auth_token=True param (should be mentioned when clicking the “Use in transformers” button on the model page):
tokenizer = AutoTokenizer.from_pretrained("username/model_name", use_auth_token=True)
model = AutoModelForMaskedLM.from_pretrained("username/model_name", use_auth_token=True)
If you’re using a fine-tuning script, for now you will have to modify it to add this parameter yourself (to all the
We’ll make sure to update the code in the repo in the coming weeks too (cc @sgugger)
Should we try to use the auth_token by default when the user is logged in?
I also use
AutoConfig.from_pretrained(model_name), which also needs
Thanks a lot!