Can we use LoftQ to optimize vision foundation models like OWL-ViT v2 and Grounding Dino?

yxli2123 / LoftQ

MIT License

180 stars 15 forks source link

Can we use LoftQ to optimize vision foundation models like OWL-ViT v2 and Grounding Dino? #13

Open solomonmanuelraj opened 6 months ago

solomonmanuelraj commented 6 months ago

Hi Team,

I need your help to optimize VFMs like OWL-ViT v2 and Grounding Dino to reduce the memory size and deploy in the edge device.

Your feedback and reference code will be helpful.

with thanks

tourzhao commented 5 months ago

You can definitely apply LoftQ. to any other transformer-based foundation models. The code should be straightforward to be adapted from our provided code for LLM.