BTW, the documentation of pushing model to huggingface is written terribly. When I try to login according to the code in documentation as below:
from huggingface_hub import notebook_login
notebook_login()
Jupyter won’t show any input widget for me to input my username and password.
And I didn’t find any solution in documentation .
So I have to pass my token in method push_to_hub(), and it stucked. And dose anybody know why I can’t login by notebook_login()?
Normally, I would suspect a problem with the library or network, but there is a large-scale failure occurring in the Hugging Face repository at the moment, and it hasn’t been completely fixed yet, so it might be due to the failure.
Firstly, updating ipywidgets version is not worked for me.
Then, as said in the second link, changing code from
from huggingface_hub import notebook_login
notebook_login()
to
from huggingface_hub import interpreter_login
interpreter_login()
seems worked. The widget for inputing token appears and I can call push_to_hub() without token parameter.
Lastly, I successfully push my model to repository after one hour. I don’t know why it is so slow. May be it is true that unstable network causes this.
Anyway, thanks a lot for your information.
The slow speed is probably due to the 500 error. It’s usually fast (unless there’s a network restriction).
That said, this might be a small bug. @Wauplin
