nbeerbower

/

Dumpling-Qwen2.5-VL-7B

like 1

Wouldn't have commented but saw the discussion from Onix22 and decided to do it anyways.
Most people won't use this model for vision anyways, and if, they could just merge it with the base model 🤷🏽‍♂️
Great model, it's more than just useable and its performance is impressive for its size. It doesn't comprise much on INT4 either, that's most likely QLora and a good config's work.
Thanks for the model, please continue you work 🙃