Using this model as an uncensored image to prompt?
Hi, I have tried to use this model as an uncensored Gwen 2.5 replacement with this Comfyui custom node pack: https://github.com/alexcong/ComfyUI_QwenVL
But it doesn't seem to work well, often outputs a full stop or "50 words".
any idea if this should work at all?
Maybe with a different prompt?
I was trying "Make a detailed NSFW image description of this women in around 550 words."
Or maybe I just need a different model?
Hello,
This is a Text2Text finetune, not a visual language so maybe that's the issue, assumed you're using it as Visual Language because your prompt seems to reference certain girl in a picture.
Just tried your prompt in text2text and i'm able to generate around 370 words by increasing max_new_tokens parameter to 1024 or 2048. Anyway, the dataset i used for this finetune mostly contains prompts of 50/100 words.