Sparser Block-Sparse Attention via Token Permutation
This work addresses the high computational cost of scaling context lengths in LLMs, offering a plug-and-play optimization for long-context tasks, though it is incremental as it builds on existing block-sparse attention methods.
The paper tackles the computational bottleneck of self-attention in large language models by proposing Permuted Block-Sparse Attention (PBS-Attn), which increases block-level sparsity through token permutation, achieving up to 2.75x speedup in long-context prefilling while closely matching full attention accuracy.
Scaling the context length of large language models (LLMs) offers significant benefits but is computationally expensive. This expense stems primarily from the self-attention mechanism, whose $O(N^2)$ complexity with respect to sequence length presents a major bottleneck for both memory and latency. Fortunately, the attention matrix is often sparse, particularly for long sequences, suggesting an opportunity for optimization. Block-sparse attention has emerged as a promising solution that partitions sequences into blocks and skips computation for a subset of these blocks. However, the effectiveness of this method is highly dependent on the underlying attention patterns, which can lead to sub-optimal block-level sparsity. For instance, important key tokens for queries within a single block may be scattered across numerous other blocks, leading to computational redundancy. In this work, we propose Permuted Block-Sparse Attention (\textbf{PBS-Attn}), a plug-and-play method that leverages the permutation properties of attention to increase block-level sparsity and enhance the computational efficiency of LLM prefilling. We conduct comprehensive experiments on challenging real-world long-context datasets, demonstrating that PBS-Attn consistently outperforms existing block-sparse attention methods in model accuracy and closely matches the full attention baseline. Powered by our custom permuted-FlashAttention kernels, PBS-Attn achieves an end-to-end speedup of up to $2.75\times$ in long-context prefilling, confirming its practical viability. Code available at https://github.com/xinghaow99/pbs-attn