Qwen2.5-VL Model Inference Failed Due to Missing processor_config.json

[yangjing69]

I encountered an inference failure caused by the missing processor_config.json file when working with the ONNX-converted version of the Qwen2.5-VL model. Here are the detailed steps I followed:

1. Downloaded the Qwen2.5-VL model: Qwen/Qwen2.5-VL-7B-Instruct.

2. Converted the Qwen2.5-VL model to ONNX format using the build script from onnxruntime-genai\src\python\py\models\build.py, with the following command:
   python builder.py -m ./Qwen2.5-VL-7B-model/safetensors -o ./Qwen2.5-VL-7B-model/onnx_fp32 -p fp32 -e cpu -c ./temp_cache

3. Ran inference on the converted ONNX model using the vision example script from onnxruntime-genai\examples\python\model-vision.py, with the command:
   python model-vision.py -m ./Qwen2.5-VL-7B-model/onnx_fp32 -e cpu

Error Log：
Loading model...
Model loaded
Traceback (most recent call last):
File "C:\Users\yangjing\Desktop\ort_genai\Qwen2.5-VL-7B\model-vision.py", line 213, in
run(args)
File "C:\Users\yangjing\Desktop\ort_genai\Qwen2.5-VL-7B\model-vision.py", line 69, in run
processor = model.create_multimodal_processor()
RuntimeError: processor_config.json not found at: ./Qwen2.5-VL-7B-model/onnx_fp32\processor_config.json