ImportError: cannot import name 'Qwen2_5_VLImageProcessor' from 'transformers.models.qwen2_5_vl' (/usr/local/lib/python3.10/dist-packages/transformers/models/qwen2_5_vl/__init__.py)

[LaoWangGB]

System Info

vllm==0.7.2
transformers==4.49.0

Who can help?

ImportError: cannot import name 'Qwen2_5_VLImageProcessor' from 'transformers.models.qwen2_5_vl' (/usr/local/lib/python3.10/dist-packages/transformers/models/qwen2_5_vl/init.py)

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

from vllm import LLM
model = LLM(model="Qwen2.5-VL-7B, trust_remote_code=True, dtype="bfloat16")

Expected behavior

I confirm the version:4.49.0_dev contains a script named image_processing_qwen2_5_vl.py, but version 4.49.0 remove it.
ImportError: cannot import name 'Qwen2_5_VLImageProcessor' from 'transformers.models.qwen2_5_vl' (/usr/local/lib/python3.10/dist-packages/transformers/models/qwen2_5_vl/init.py)

[2710932616]

i add Qwen2VLImageProcessor， it‘s work.

[zucchini-nlp]

Hey!

This is a duplicate of #36193. Yes, the class was removed since we want to enable fast processing for Qwen2-5-VL (same was for qwen2-vl). The image processors of the two models are identical, so you can safely use Qwen2VLImageProcessor

The official checkpoints are already modified, if you are using a fine-tune please change the preprocessor_config.json

[2710932616]

i add Qwen2VLImageProcessor， it‘s work.

I found that the vllm has fixed this problem in vllm-project/vllm#13286, But the release version has not been released yet, so installing the latest vllm（from github） should also solve this problem.

[xuanyue23]

i add Qwen2VLImageProcessor， it‘s work.

thx