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.

번역 준비 중입니다. 원문을 표시합니다.