Hey there. I know this sounds absolutely ridiculous but after months of working fine, some of the models are downloading and some aren’t while using
AutoModel.from_pretrained function
For example
The download never begins at all
But it works for this other random model that I picked up from the documentation.
Not able to post a second screenshot here because I am new user sigh
Thank you.
Screenshot of the other model downloading just fine.
Also, would like to emphasise that none of these models are in the cache and I am trying to download them for the first time.
As of mid-afternoon April 5 ‘bert-base-uncased’ cannot be downloaded using from_pretrained() on Windows platforms. I get the warning about Windows not supporting symlinks and then the download freezes.
Not just me, but many users of the InvokeAI image generation platform are complaining of the same problem.
Thanks for the reply Lincoln! I feel so much at ease knowing that this is the issue. Is there any other way other than using the hub library to download the files of a model? The bigger models seem to have the torch files along with jax files too and the download footprint becomes huge if I have to download the entire git repo.
Seems to work fine today! Phew
yes this was caused by a server config change that we reverted
cc @XciD @Wauplin @coyotte508 @pierric for viz.
