FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference

TL;DR

FlexPrefill introduces a dynamic, context-aware sparse attention mechanism for large language models, significantly improving efficiency and accuracy during long-sequence inference by adaptively adjusting attention patterns in real-time.

Contribution

It proposes a novel flexible sparse pre-filling method that adaptively adjusts attention patterns and computational resources based on input and attention head requirements.

Findings

Significant speed improvements over prior methods.

Enhanced inference accuracy with adaptive attention patterns.

Demonstrated effectiveness on long-sequence tasks.

Abstract

Large language models (LLMs) encounter computational challenges during long-sequence inference, especially in the attention pre-filling phase, where the complexity grows quadratically with the prompt length. Previous efforts to mitigate these challenges have relied on fixed sparse attention patterns or identifying sparse attention patterns based on limited cases. However, these methods lacked the flexibility to efficiently adapt to varying input demands. In this paper, we introduce FlexPrefill, a Flexible sparse Pre-filling mechanism that dynamically adjusts sparse attention patterns and computational budget in real-time to meet the specific requirements of each input and attention head. The flexibility of our method is demonstrated through two key innovations: 1) Query-Aware Sparse Pattern Determination: By measuring Jensen-Shannon divergence, this component adaptively switches between query-specific diverse attention patterns and predefined attention patterns. 2) Cumulative-Attention Based Index Selection: This component dynamically selects query-key indexes to be computed based on different attention patterns, ensuring the sum of attention scores meets a predefined threshold. FlexPrefill adaptively optimizes the sparse pattern and sparse ratio of each attention head based on the prompt, enhancing efficiency in long-sequence inference tasks. Experimental results show significant improvements in both speed and accuracy over prior methods, providing a more flexible and efficient solution for LLM inference.