infinitevl-linear-sparse-attention

【效率与协作】Merge sliding window and linear attention (Gated DeltaNet) for unlimited VLM inputs with 3.6× speedup. InfiniteVL handles video understanding at 24 FPS with constant memory—ideal when context length must scale without quadratic overhead.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准