PanoVLM-500M

PanoVLM is a linear-attention vision-language model: a FastViT-HD vision encoder (timm) feeding a PanoLM linear-attention causal LM through a lightweight projector.

译文尚未就绪，显示来源原文