[Bug]: TypeError: Qwen2_5_VLProcessor.__init__() got multiple values for argument 'image_processor' with transformers 4.53

[SrikanthChellappa]

Your current environment

TypeError: Qwen2_5_VLProcessor.init() got multiple values for argument 'image_processor' with transformers 4.53

vllm: 0.9.1
transformers: 4.53.0
torch 2.7
cuda version 12.6

🐛 Describe the bug

TypeError: Qwen2_5_VLProcessor.init() got multiple values for argument 'image_processor' with transformers 4.53

vllm: 0.9.1
transformers: 4.53.0
torch 2.7
cuda version 12.6

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[jeejeelee]

have you tried transformers==4.52.4?

[SrikanthChellappa]

Yes it works for transformers version < 4.53 and i have moved back to 4.49 to keep my application running

[DarkLight1337]

It should be fixed by #20244, can you use latest main branch?

[SrikanthChellappa]

Thanks @DarkLight1337 / @jeejeelee . Let me try with the latest main branch

[SrikanthChellappa]

I upgraded the vllm version from 0.9.1 to '0.9.2.dev392+g2e25bb12a' but still i face the same issue. Am i making any mistake here? I tried installing this nightly build version using "pip install -U vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
"

[DarkLight1337]

cc @Isotr0py

[DarkLight1337]

Maybe you can try redownloading the model repo from HF Hub

[SrikanthChellappa]

@DarkLight1337 may be i will wait for this fix to be available with next vllm release and I will continue to stay with vllm 0.9.1/transformers 4.49 (<4.53) for now

[Isotr0py]

You can also try to install transformers main branch from source. This is a known bug in transformers 4.53.0 which was just gotten fixed on main branch recently (huggingface/transformers#39125), which will be included in next patch release.

[SrikanthChellappa]

oh ok @Isotr0py. Let me try that as well

[SrikanthChellappa]

Thanks @DarkLight1337 / @Isotr0py . Finally it worked and it was actually the fix from transformers latest version that made it work.

My current versions are
transformers version : 4.54.0.dev0
vllm version: 0.9.1

Note: I tried initially upgrading vllm as well from 0.9.1 to vllm-0.9.2.dev394+g0ec3779df but it was throwing errors related to lmcache and so reverted it back to 0.9.1

[DarkLight1337]

Glad that fixed the issue for you!