[Bug] Unrecognized video processor in OpenGVLab/InternVL3-2B-hf.

[Andy-Cheng]

Checklist

• 1. I have searched related issues but cannot get the expected help.
• 2. The bug has not been fixed in the latest version.
• 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.

Describe the bug

I think it's related to Video processors as a separate class
The loading of the processor by AutoProcessor would trigger the bug:

processor = AutoProcessor.from_pretrained("OpenGVLab/InternVL3-2B-hf")

  File "/home/andy/miniconda3/lib/python3.13/site-packages/transformers/processing_utils.py", line 1185, in from_pretrained
    args = cls._get_arguments_from_pretrained(pretrained_model_name_or_path, **kwargs)
  File "/home/andy/miniconda3/lib/python3.13/site-packages/transformers/processing_utils.py", line 1248, in _get_arguments_from_pretrained
    args.append(attribute_class.from_pretrained(pretrained_model_name_or_path, **kwargs))
                ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/andy/miniconda3/lib/python3.13/site-packages/transformers/models/auto/video_processing_auto.py", line 360, in from_pretrained
    raise ValueError(
    ...<3 lines>...
    )

ValueError: Unrecognized video processor in OpenGVLab/InternVL3-2B-hf. Should have a 

`video_processor_type` key in its video_preprocessor_config.json of config.json, or one of the following 

`model_type` keys in its config.json: instructblipvideo, llava_next_video, llava_onevision, qwen2_5_vl, qwen2_vl, video_llava

Reproduction

from transformers import AutoProcessor
processor = AutoProcessor.from_pretrained("OpenGVLab/InternVL3-2B-hf")

Environment

torch                    2.7.0
torchvision              0.22.0
transformers             4.52.0.dev0

Error traceback

[yuecao0119]

Hi,

Please refer to our requirement.txt file to downgrade the transformer version, v4.37.2 is recommended.

[Andy-Cheng]

I was talking about the Hugging Face checkpoint (OpenGVLab/InternVL3-2B-hf). I could load the weight using v.4.52.0 until a certain commit history.
I don't think Hugging Face Transformers has the support for internvl before v.4.52.0 .

[Batwho]

same issue