Unnecessary noise when loading a transformer

[AngledLuffa]

System Info

python: 3.13.5

torch: 2.7.1+cu118

transformers: 5.2.0

tokenizers: 0.22.2

Who can help?

@ArthurZucker @Cyrilvallez

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

The latest version of transformers is extremely noisy, and there seems to be no reason or benefit for any of it.

>>> from transformers import AutoModel
>>> bert_model = AutoModel.from_pretrained('hfl/chinese-electra-180g-large-discriminator')
Loading weights: 100%|████████████████████████████████████████████| 389/389 [00:00<00:00, 5460.73it/s, Materializing param=encoder.layer.23.output.dense.weight]
ElectraModel LOAD REPORT from: hfl/chinese-electra-180g-large-discriminator
Key                                               | Status     |  |
--------------------------------------------------+------------+--+-
electra.embeddings.position_ids                   | UNEXPECTED |  |
discriminator_predictions.dense_prediction.bias   | UNEXPECTED |  |
discriminator_predictions.dense.weight            | UNEXPECTED |  |
discriminator_predictions.dense_prediction.weight | UNEXPECTED |  |
discriminator_predictions.dense.bias              | UNEXPECTED |  |

Notes:
- UNEXPECTED    :can be ignored when loading from different task/architecture; not ok if you expect identical arch.
>>>
Exception in Thread-auto_conversion:
Traceback (most recent call last):
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/huggingface_hub/utils/_http.py", line 720, in hf_raise_for_status
    response.raise_for_status()
    ~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/httpx/_models.py", line 829, in raise_for_status
    raise HTTPStatusError(message, request=request, response=self)
httpx.HTTPStatusError: Client error '403 Forbidden' for url 'https://huggingface.co/api/models/hfl/chinese-electra-180g-large-discriminator/discussions?p=0'
For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/403

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

Traceback (most recent call last):
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/threading.py", line 1043, in _bootstrap_inner
    self.run()
    ~~~~~~~~^^
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/threading.py", line 994, in run
    self._target(*self._args, **self._kwargs)
    ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/transformers/safetensors_conversion.py", line 117, in auto_conversion
    raise e
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/transformers/safetensors_conversion.py", line 96, in auto_conversion
    sha = get_conversion_pr_reference(api, pretrained_model_name_or_path, **cached_file_kwargs)
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/transformers/safetensors_conversion.py", line 69, in get_conversion_pr_reference
    pr = previous_pr(api, model_id, pr_title, token=token)
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/transformers/safetensors_conversion.py", line 14, in previous_pr
    for discussion in get_repo_discussions(repo_id=model_id, token=token):
                      ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/huggingface_hub/hf_api.py", line 6455, in get_repo_discussions
    discussions, has_next = _fetch_discussion_page(page_index=page_index)
                            ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/huggingface_hub/hf_api.py", line 6444, in _fetch_discussion_page
    hf_raise_for_status(resp)
    ~~~~~~~~~~~~~~~~~~~^^^^^^
  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/huggingface_hub/utils/_http.py", line 802, in hf_raise_for_status
    raise _format(HfHubHTTPError, message, response) from e
huggingface_hub.errors.HfHubHTTPError: (Request ID: Root=1-69a62017-4041fa5968ff8ee13d64fe3d;47725ac8-64d6-4947-ad16-4f88ef89391c)

403 Forbidden: Discussions are disabled for this repo.
Cannot access content at: https://huggingface.co/api/models/hfl/chinese-electra-180g-large-discriminator/discussions?p=0.
Make sure your token has the correct permissions.

>>>
KeyboardInterrupt

Expected behavior

Expected behavior would be nice and quiet!

>>> from transformers import AutoModel
>>> bert_model = AutoModel.from_pretrained('hfl/chinese-electra-180g-large-discriminator')
>>>

[Adarsh0047]

I will check and take up this issue

[Rocketknight1]

To be fair, most of that noise is Python throwing an error because you don't actually have repo access. The other warnings about missing/unexpected weights are important - they let users know that they're transferring a checkpoint to a different task, and may need to fine-tune rather than just assuming inference will immediately work out of the box.

[ArthurZucker]

mmm its weird that weights did load tho!

[ArthurZucker]

cc @LysandreJik its the safetensor auto bot conversion here:

  File "/nlp/scr/horatio/miniconda3/lib/python3.13/site-packages/transformers/safetensors_conversion.py", line 14, in previous_pr
    for discussion in get_repo_discussions(repo_id=model_id, token=token):

[AngledLuffa]

To be fair, most of that noise is Python throwing an error because you don't actually have repo access.

We have the ability to download that model, though, so it's weird that we can't get any other necessary information.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[AngledLuffa]

@ArthurZucker was there a fix for the repo not having access leaving a long stack trace? It was happening every time I loaded this model, although the model still successfully loaded.