[Bee] Can't import CodeFormulaPredictor due to breaking change from transformers 5.13.0

[tozhangkai]

With transformers==5.13.0, we can no longer import CodeFormulaPredictor from docling_ibm_models:

# pip list|grep transformers
transformers              5.13.0

# python3
Python 3.12.12 (main, Apr 16 2026, 00:00:00) [GCC 11.5.0 20240719 (Red Hat 11.5.0-11)] on linux
Type "help", "copyright", "credits" or "license" for more information.
Ctrl click to launch VS Code Native REPL

>>> from docling_ibm_models.code_formula_model.code_formula_predictor import CodeFormulaPredictor

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/root/work/docling/.venv/lib64/python3.12/site-packages/docling_ibm_models/code_formula_model/code_formula_predictor.py", line 11, in <module>
    from docling_ibm_models.code_formula_model.models.sam_opt_image_processor import (
  File "/root/work/docling/.venv/lib64/python3.12/site-packages/docling_ibm_models/code_formula_model/models/sam_opt_image_processor.py", line 26, in <module>
    AutoImageProcessor.register(
  File "/root/work/docling/.venv/lib64/python3.12/site-packages/transformers/models/auto/image_processing_auto.py", line 751, in register
    IMAGE_PROCESSOR_MAPPING.register(config_class, image_processor_classes, exist_ok=exist_ok)
  File "/root/work/docling/.venv/lib64/python3.12/site-packages/transformers/models/auto/auto_factory.py", line 680, in register
    if key.__module__.startswith("transformers."):
       ^^^^^^^^^^^^^^
AttributeError: 'str' object has no attribute '__module__'. Did you mean: '__mod__'?
>>> 

[pablopupo]

I looked into this. AutoImageProcessor.register() in sam_opt_image_processor.py registers under a plain string (config_class="SamOptImageProcessor") instead of a real config class, which worked fine through transformers 5.12 since the internal _LazyAutoMapping.register only touched key.__name__ behind a hasattr guard. 5.13 added an unconditional key.__module__ read ahead of that guard, and strings do not have __module__, hence the AttributeError. CodeFormulaPredictor always loads SamOptImageProcessor directly via from_pretrained rather than through the Auto class, so I have a fix ready that skips the registration when it raises instead of crashing the import. Opening a PR shortly.