Question Regarding multiplie images.

#1
by wamozart - opened

i was playing around with you phi3v.py example and while the non onnx model (https://huggingface.co/microsoft/Phi-3-vision-128k-instruct) allows multiple images as input (inputs = processor(prompt, [image1, image2])
the onnx version doesn't. I keep getting :
Unable to cast Python instance to C++ type (#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for details)

Any idea?

Microsoft org

The Phi-3 vision model officially supports one image as input. You can find more information on that here.

kvaishnavi changed discussion status to closed

Sign up or log in to comment