Failed to Download GPT2-large Model from Hub

[daskol]

System Info

py312, transformers 4.40.2, huggingface-hub 0.23.0.

Who can help?

@ArthurZucker and @younesbelkada

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

>>> from transformers import AutoModelForCausalLM
>>> model = AutoModelForCausalLM.from_pretrained('gpt2-large')
...
OSError: gpt2-large does not appear to have a file named config.json. Checkout 'https://huggingface.co/gpt2-large/tree/None' for available files.

It seems that tranformers failed to resolve branch and tried to find config.json in non-existing branch None.

Expected behavior

The script above has to download and load model.

[younesbelkada]

Hi @daskol
Thanks for the issue ! Can you try with openai-community/gpt2-large instead ? We recently migrated the canonical models into dedicated organisations. It is surprising as automatic redirection should work, I just ran your snippet in my local env and it worked fine

[daskol]

Yes, openai-community/gpt2-large works for me as well. I have noticed that gpt2* repos were moved and populate them manually from openai-community/gpt2*. I think the issue exists in case if there is nothing cached.

[younesbelkada]

Thanks @daskol !
Hmmm, that's strange, I tried

>>> from transformers import AutoModelForCausalLM
>>> model = AutoModelForCausalLM.from_pretrained("gpt2-large", force_download=True)

It seems the model is being downloaded correctly, force_download=True makes sure it re-downloads the model even if it is present in the cache