I did all steps in the tutorial, but when I execute git push I got the error:
Uploading LFS objects: 0% (0/1), 0 B | 0 B/s, done.
batch response: Authorization error.
error: failed to push some refs to 'https://huggingface.co/datasets/zeusfsx/ukrainian-news'
I logged via huggingface-cli login
Getting the exact same error - so peeking for any answers as well.
Did you find a solution?
hi @zeusfsx and @emiltj can you please double check you’re using a token with write role permission?
Another question, which OS are you running?
I triple checked. Must be working as I can push with push_to_hub(), but not using the cli.
I’ve recently encountered similar issues while working on OSX. To resolve the problem, I had to clean up all the huggingface.co password on my keychain access. I also had to reconfigure my git config --global credential.helper osxkeychain then huggingface-cli login. After these steps, it was back to normal. Note these are OSX-specific
Sorry, I forgot to reply to the OS question.
Running MacOS 13.2.1 (22D68), with the M1-chip. Not unlikely that the problem is similar to the one you have. I am not going to bother setting that up, but will just be using the push_to_hub().
Thank you for your answer
getting the same error on ubuntu 16.04, huggingface-cli login does nothing
Pushing to hub is such a frustrating experience at this point, I am about to give up.
for all in the thread, we now have support for SSH keys, so many of the authentication problems could be overcome via SSH authentication.
You can use your WRITE User Access Tokens as password when git push prompted the password.
Just spent a good part of my Sunday creating new ssh keys and setting up gpg. I finally pushed this morning and created a new dataset and got verified commit: fantastic.
Now I try to clone the repo created earlier and I get:
Error downloading object: kg-column-test.txt (da355c7): Smudge error: Error downloading kg-column-test.txt (da355c7ac1bc978121fda6c500c7631d55b71d30d19767c6ececd6e58af4a801): batch response: Authentication required: Password authentication in git is no longer supported. You must use a user access token or an SSH key instead. See https://huggingface.co/blog/password-git-deprecation
Errors logged to '/Users/patrickocal_mchome/Documents/stanford/gov-report-kgs/.git/lfs/logs/20231126T205837.279135.log'.
Use `git lfs logs last` to view the log.
error: external filter 'git-lfs filter-process' failed
fatal: kg-column-test.txt: smudge filter lfs failed
warning: Clone succeeded, but checkout failed.
You can inspect what was checked out with 'git status'
and retry with 'git restore --source=HEAD :/'
I am using ssh … I’ve deleted all my hugging face items in keychain (on OSX Sonoma). 
i got the same “Authorization error” when trying trainer.push_to_hub(“my-repo”, token=<write_token>). The same works for dataset.push_to_hub(…) earlier in the code. i am using Google Colab and finetuning a Llama2-7b-hf model.
i have a read_token for downloading the meta/llama2-7b model and a write token to be able to push to the hub repo. i run huggingface-cli login() --token=<write_token>, and the output shows login successful with write permission. However, when i run ‘whoami()’ later, the token shown is the read_token with role=read.
Is there a way to overcome this ? It has been quite frustrating, trying to push the finetuned model to the repo in huggingface hub
Hi @radames , push_to_hub with write token works for dataset.push_to_hub(“repo-name”, token=write_token) or from tokenizer.push_to_hub(“repo-name”,use_auth_token=write_token), but does not work for trainer.push_to_hub(“repo-name”, token=write_token), where the trainer is from huggingface transformers library.
All of this is being done on a Google Colab project for pushing a finetuned model to my huggingface repo, when i get the Authorization error.
On examination of the python code of the library, there is no token passed, and the git push command seems to be failing authorization somewhere that i am not able to figure out why.