The official repo for "D -Attn: Decomposed Attention for Large Vision-and-Language Model" - View it on GitHub