fasa-frequency-aware-sparse-attention

【效率与协作】Identify dominant frequency components in RoPE embeddings to determine which attention dimensions are essential, reducing KV cache by 81% while maintaining performance. Training-free approach integrates with existing compression methods for 2.56x speedup on long-context reasoning.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准