Short-Long Convolutions Help Hardware-Efficient Linear Attention to Focus on Long Sequences
This work addresses computational bottlenecks in processing long sequences for machine learning models, offering a hybrid solution that combines the strengths of linear attention and SSMs, though it is incremental in nature.
The authors tackled the inefficiency of linear attention in causal settings and the instability of state space models (SSMs) by proposing CHELA, which uses short-long convolutions and a hardware-efficient linear attention implementation, achieving competitive performance on the Long Range Arena benchmark and language modeling tasks.
To mitigate the computational complexity in the self-attention mechanism on long sequences, linear attention utilizes computation tricks to achieve linear complexity, while state space models (SSMs) popularize a favorable practice of using non-data-dependent memory pattern, i.e., emphasize the near and neglect the distant, to processing sequences. Recent studies have shown the priorities by combining them as one. However, the efficiency of linear attention remains only at the theoretical level in a causal setting, and SSMs require various designed constraints to operate effectively on specific data. Therefore, in order to unveil the true power of the hybrid design, the following two issues need to be addressed: (1) hardware-efficient implementation for linear attention and (2) stabilization of SSMs. To achieve this, we leverage the thought of tiling and hierarchy to propose CHELA (short-long Convolutions with Hardware-Efficient Linear Attention), which replaces SSMs with short-long convolutions and implements linear attention in a divide-and-conquer manner. This approach enjoys global abstraction and data-dependent selection from stable SSM and linear attention while maintaining real linear complexity. Our comprehensive experiments on the Long Range Arena benchmark and language modeling tasks demonstrate the effectiveness of the proposed method.