Open kexul opened 5 months ago
Result with int4 version of QwenVL: Not as funny as your example.
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.
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?