Error when trying to load Adapter model of revision in repository on HF

[pesuchin]

Summary

When trying to load an Adapter Model by specifying a revision in an private and public repository on HF, the following error occurs:

 File "/Users/pesuchin/project/sentence-transformers/examples/debug2.py", line 7, in <module>
    model = SentenceTransformer("pesuchin/revision-pesu-test", 
  File "/Users/pesuchin/project/sentence-transformers/sentence_transformers/SentenceTransformer.py", line 308, in __init__
    modules, self.module_kwargs = self._load_sbert_model(
  File "/Users/pesuchin/project/sentence-transformers/sentence_transformers/SentenceTransformer.py", line 1728, in _load_sbert_model
    module = module_class(model_name_or_path, cache_dir=cache_folder, backend=self.backend, **kwargs)
  File "/Users/pesuchin/project/sentence-transformers/sentence_transformers/models/Transformer.py", line 78, in __init__
    self._load_model(model_name_or_path, config, cache_dir, backend, **model_args)
  File "/Users/pesuchin/project/sentence-transformers/sentence_transformers/models/Transformer.py", line 138, in _load_model
    self.auto_model = AutoModel.from_pretrained(
  File "/Users/pesuchin/project/sentence-transformers/.venv/lib/python3.9/site-packages/transformers/models/auto/auto_factory.py", line 525, in from_pretrained
    config, kwargs = AutoConfig.from_pretrained(
  File "/Users/pesuchin/project/sentence-transformers/.venv/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py", line 976, in from_pretrained
    config_dict, unused_kwargs = PretrainedConfig.get_config_dict(pretrained_model_name_or_path, **kwargs)
  File "/Users/pesuchin/project/sentence-transformers/.venv/lib/python3.9/site-packages/transformers/configuration_utils.py", line 633, in get_config_dict
    config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs)
  File "/Users/pesuchin/project/sentence-transformers/.venv/lib/python3.9/site-packages/transformers/configuration_utils.py", line 690, in _get_config_dict
    resolved_config_file = cached_file(
  File "/Users/pesuchin/project/sentence-transformers/.venv/lib/python3.9/site-packages/transformers/utils/hub.py", line 432, in cached_file
    raise EnvironmentError(
OSError: revision-test is not a valid git identifier (branch name, tag name or commit id) that exists for this model name. Check the model page at 'https://huggingface.co/pesuchin/revision-pesu-test' for available revisions.

Steps to Reproduce

Execute the following code using Sentence Transformers v3.3.0:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer(<Your Repository>,  # example: pesuchin/revision-pesu-test
                            revision="revision-test",
                            token=<Your Token>)

Cause

When the revision is passed as an argument to the init method of the SentenceTransformer class, it is included in model_args and shared with the Transformer class.

As a result, in the following code, AutoModel uses model_args, which contains both the repository name of the base model and the revision where the adapter model is located. This leads to a mismatch between the repository and the revision, causing an attempt to access a nonexistent revision and resulting in an error:

sentence-transformers/sentence_transformers/models/Transformer.py

Lines 130 to 132 in c1775a6

	self.auto_model = AutoModel.from_pretrained(
	model_name_or_path, config=config, cache_dir=cache_dir, **model_args
	)

Proposed Fix

Currently under consideration.