Hugging Token needed?

[benoib]

Hey guys!

I'm playing with the runpod script for stable diffusion xl.

Something is bothering me.

In the training script we have for instance

    # Load the tokenizers
    tokenizer_one = AutoTokenizer.from_pretrained(
        args.pretrained_model_name_or_path,
        subfolder="tokenizer",
        use_fast=False,
        use_auth_token=True,
    )

The use_auth_token= True implies that we should use a HuggingFace token. I can't find any anywhere in the Runpod script. How does it work?

When I try to run the example script on my own, I get this error:

huggingface_hub.utils._headers.LocalTokenNotFoundError: Token is required (token=True),
Which seems normal.

Where is the catch / what am I missing here? Obviously I tried running the training with

--hub-token MY_TOKEN passed as a parameter. Did not help either so I am a bit lost here

Thanks for your help,

Ben

[TheLastBen]

nothing missing, the XL repo is public so no need for a token

[benoib]

Yes I figured, But i get the error on the AutoTokenizer. So you mean that i have another problem with my args?
Thanks for the answer!

[TheLastBen]

don't add any arg