Use int4 version of Qwen-VL?

[kexul]

Hi, since int4 version of Qwen-vl is avaialble and more friendly for low end gpu, is it a plug and play model for clot?

[kexul]

Result with int4 version of QwenVL:

Not as funny as your example.

[zhongshsh]

Our model is based on Qwen/Qwen-VL-Chat, and using the int4 version of Qwen-VL directly might compromise the quality of responses. Humor is subjective. You can try generating a few more times to find responses that match your sense of humor.