[HANDS-ON BUG] Unit 1 problem with LFS files pushing model to the hub

[AlvLC]

Bug description

Hi there, I've strictly followed the notebook (checking the solutions for each cell) from the hands-on exercise in unit 1 and I always encounter this problem when I try to push the model to the hub:

Basically, the repository is created and the login to HF is fine but it gets stuck indefinitely waiting for the LSF files to be uploaded. I've tried to install LSF again and searched for information about this issue but I can't seem to find what is going on. I'll share the notebook bellow in case that anyone can help and there's anything that I oversaw but I find it weird.

Material

• Colab notebook: https://colab.research.google.com/drive/1xWTD_M-1j5ZdlcYQs2SvwG-XubwLVGcj?usp=sharing

[sardorboboshov]

I am having same problem

[magixn]

I found that 4 LFS in the official model are all smaller than 200KB. I wonder it may be not a network issue. I also cannot visit the leaderboard after Runtime error during upload with

It appears that one or more of your files contain valid Hugging Face secrets, such as tokens or API keys. It is important to keep your secrets private and not expose them in code that is publicly accessible.

I think I do not do anything else but fill the TODO and run the notebook.

[andlock97]

I'm also having the same problem

[fernando-ml]

I am having this issue as well.

[JamesMurrayCoder]

I am also having this issue

[Drakrig]

I've managed to push it successfully by passing token directly to package_to_hub function like this:

package_to_hub(model=model, model_name=model_name, model_architecture=model_architecture, env_id=env_id, eval_env=eval_env, repo_id=repo_id, token=TOKEN_AS_STRING, commit_message=commit_message)

Looks like all this scheme with notebook_login() and git config --global credential.helper store doesn't work as intended in Colab for now.

[fernando-ml]

@Drakrig Thank you so much. Your approach worked for me.

[AlvLC]

Also worked for me. I've had other similar issues trying to use some HF classes with the usual login instead of just passing the token. Thank you @Drakrig !!!

[Wauplin]

Hi everyone, sorry for the inconvenience on this topic. The bug has been identified and fixed in huggingface_hub. More details in huggingface/huggingface_hub#1953 (comment).

TL;DR: in the meantime you can fix the problem by running !pip install -U huggingface_hub before running anything else in your colab.