Is the response of Qwen-VL very short?

[zifuwan]

Has anyone encountered the problem that Qwen-VL responds very shortly to every question? Probably just one word. Thanks.

[frankren930]

Has anyone encountered the problem that Qwen-VL responds very shortly to every question? Probably just one word. Thanks.

I found that even I remove max_new_tokens=20, , it still output very shortly

[xlchen0205]

same to you

[LinyeLi60]

Because the author makes the output of the model only one word "Please answer this question with one word "