Issues loading model to SentenceTransformer without passing HF token

[nathan-az]

Code to reproduce:

import os
from sentence_transformers import SentenceTransformer

# os.environ["HF_TOKEN"] = None  # my token goes here
model = SentenceTransformer(
    'sentence-transformers/all-MiniLM-L6-v2',
)

The above throws an error:

Traceback (most recent call last):
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/utils/_errors.py", line 304, in hf_raise_for_status
    response.raise_for_status()
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/requests/models.py", line 1024, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/adapter_config.json

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/transformers/utils/hub.py", line 403, in cached_file
    resolved_file = hf_hub_download(
                    ^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/utils/_deprecation.py", line 101, in inner_f
    return f(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 1240, in hf_hub_download
    return _hf_hub_download_to_cache_dir(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 1347, in _hf_hub_download_to_cache_dir
    _raise_on_head_call_error(head_call_error, force_download, local_files_only)
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 1854, in _raise_on_head_call_error
    raise head_call_error
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 1751, in _get_metadata_or_catch_error
    metadata = get_hf_file_metadata(
               ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 1673, in get_hf_file_metadata
    r = _request_wrapper(
        ^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 376, in _request_wrapper
    response = _request_wrapper(
               ^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/file_download.py", line 400, in _request_wrapper
    hf_raise_for_status(response)
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/huggingface_hub/utils/_errors.py", line 352, in hf_raise_for_status
    raise RepositoryNotFoundError(message, response) from e
huggingface_hub.utils._errors.RepositoryNotFoundError: 401 Client Error. (Request ID: Root=1-67a16fe0-6c8d26b009103a2c16bf76bc;c908df8a-fbac-445e-8a38-3fb67a7efc86)

Repository Not Found for url: https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/adapter_config.json.
Please make sure you specified the correct `repo_id` and `repo_type`.
If you are trying to access a private or gated repo, make sure you are authenticated.
Invalid credentials in Authorization header

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/nazrak/Documents/Work/repos/ml-studio/modules/src/ml_platform/mlp_sentence_transformer_finetune/included/scripts/sandbox_load.py", line 5, in <module>
    model = SentenceTransformer(
            ^^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/sentence_transformers/SentenceTransformer.py", line 308, in __init__
    modules, self.module_kwargs = self._load_sbert_model(
                                  ^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/sentence_transformers/SentenceTransformer.py", line 1739, in _load_sbert_model
    module = module_class(model_name_or_path, cache_dir=cache_folder, backend=self.backend, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/sentence_transformers/models/Transformer.py", line 80, in __init__
    config, is_peft_model = self._load_config(model_name_or_path, cache_dir, backend, config_args)
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/sentence_transformers/models/Transformer.py", line 121, in _load_config
    find_adapter_config_file(
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/transformers/utils/peft_utils.py", line 88, in find_adapter_config_file
    adapter_cached_filename = cached_file(
                              ^^^^^^^^^^^^
  File "/Users/nazrak/opt/miniconda3/envs/cross-embed/lib/python3.12/site-packages/transformers/utils/hub.py", line 426, in cached_file
    raise EnvironmentError(
OSError: sentence-transformers/all-MiniLM-L6-v2 is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'
If this is a private repository, make sure to pass a token having permission to this repo either by logging in with `huggingface-cli login` or by passing `token=<your_token>`

Bizarrely this is resolved fully if I set the HF_TOKEN environment variable or pass the token argument, despite sentence-transformers/all-MiniLM-L6-v2 being an ungated model.

In addition, I have no issue loading it with AutoModel:

from transformers import AutoModel

model = AutoModel.from_pretrained(
    'sentence-transformers/all-MiniLM-L6-v2',
)

Some versioning info:

accelerate==1.3.0
hf_transfer==0.1.9
sentence-transformers==3.4.1
transformers==4.48.2

[anakin87]

Similar issue: #3206

We're encountering similar issues in the CI of Haystack. When the CI runs from a fork (where we don't set HF_TOKEN), we get similar errors: https://github.com/deepset-ai/haystack/actions/runs/13115319609/job/36588574826

[anakin87]

Okay, I was able to reproduce the issue.

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("all-MiniLM-L6-v2", token="")

This fails. In the past, an invalid or empty token did not cause problems with public models.
(I found similar reports in other projects as well: instructlab/instructlab#3075)

The problem is also that in certain CI environments, an empty environment variable is being set for the token instead of leaving it unset.

This issue is likely not specific to Sentence Transformers, but rather due to changes in Hugging Face Hub.

@tomaarsen It would be great if you could notify the maintainers of other related HF projects.
If this change is intentional, it should be documented accordingly. (I am encountering the same problem with transformers as well, maybe it can addressed in huggingface_hub).

[ahmadreza-akbarnejad]

Try token=False

[tomaarsen]

Hello!

False should indeed work.

If you'd like to get token="" to work, then you'll have to open an issue in huggingface_hub I reckon, with:

from huggingface_hub import hf_hub_download

hf_hub_download(
    repo_id="sentence-transformers/all-MiniLM-L6-v2",
    filename="config.json",
    token="",
    local_dir="tmp",
    cache_dir="tmp_cache",
)

I indeed think that this issue is best resolved outside of Sentence Transformers, but instead in huggingface_hub (and perhaps transformers).

• Tom Aarsen

[nathan-az]

Interesting! Would definitely love to see this resolved somehow (I usually prefer not to set token at all in Python and instead set it in the environment).

It's not immediately obvious (although I didn't dive especially deep) into why this happens with SentenceTransformer but not AutoModel.from_pretrained so I assumed it was a difference in the invocation!

@tomaarsen are you happy for me to make the issue in huggingface_hub or would you prefer to? Looks like you might have some pull with the HF repos 😉

[tomaarsen]

I will admit that it's odd that it works in transformers but not in sentence-transformers. I'll try and dive into that a bit more this afternoon.
If Transformers has a fix on their end, then perhaps I should also put a fix on the ST side, rather than relying on huggingface_hub to resolve it.

• Tom Aarsen