Fix multimodal processor get duplicate arguments when receive kwargs for initialization

[Isotr0py]

What does this PR do?

Also reported in vllm-project/vllm#19833

• Fix Qwen2_5OmniProcessor.__init__() got multiple values for argument 'image_processor' #38898

The current processor arguments revolving logic can hit TypeError: Qwen2VLProcessor.__init__() got multiple values for argument 'tokenizer' if we passed tokenizer as a kwargs. Because it will remove unused args, which caused an incorrect args order and raise TypeError: Qwen2VLProcessor.__init__() got multiple values for argument 'tokenizer'.

For example, Qwen2-VL receive (image_processor, tokenizer, video_processor) as args, if we passed tokenizer as kwargs, it will be removed from args, which will become (image_processor, video_processor) and pass video_processor as tokenizer:

transformers/src/transformers/models/qwen2_vl/processing_qwen2_vl.py

Line 78 in 4a79bf9

def __init__(self, image_processor=None, tokenizer=None, video_processor=None, chat_template=None, **kwargs):

Code to reproduce

from transformers import Qwen2VLProcessor, Qwen2Tokenizer

processor = Qwen2VLProcessor.from_pretrained("/data/LLM-model/Qwen2-VL-2B-Instruct")
print(type(processor.tokenizer))

tokenizer = Qwen2Tokenizer.from_pretrained("/data/LLM-model/Qwen2.5-3B-Instruct")
processor = Qwen2VLProcessor.from_pretrained("/data/LLM-model/Qwen2-VL-2B-Instruct", tokenizer=tokenizer)
print(type(processor.tokenizer))

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[zucchini-nlp]

Thanks, looks good. I think this has a lot of assumptions about the order of args in the __init__, but I guess the codebase currently follows consistent ordering (🤞🏻 )

Can you add a small test in test_processing_common where one of the attributes can be arbitrarily overriden by a user?

[zucchini-nlp]

Seems like it broke a whole lot of tests...

[zucchini-nlp]

Thanks, we'll patch it today