QWEN2 VL does not work?

[sinand99]

Version: 0.3.6 Build 3
Model: bartowski/Qwen2-VL-7B-Instruct-GGUF

I get this stupid error "Model does not support images" which is funny because this is a vision model and LM Studio claims to support it. By the way this model works fine on llama.cpp. So it is not a corrupted file issue.

[YorkieDev]

@sinand99 it works on my machine. What is the file format/size/resolution of the image you are sending?

[sinand99]

@YorkieDev It doesn't matter, it happens with all types/sizes. The problem is LM Studio does not send the prompt to the model because it thinks the model is not vision model even though it shows the EYE icon in the model list page, which is very weird.

[yagil]

Can you try to load it with a larger context, at least 1600?

Did you get this error at any point?

For vision-enabled models, please use context size 1500 or larger

[sinand99]

Ok, now it works when I set the context size to 2048. Thanks.
@yagil No, never got that error. I guess the error message should be clearer. It never says anything about context size.