Stochastic Sparse Attention for Memory-Bound Inference

摘要

Autoregressive decoding becomes bandwidth-limited at long contexts, as generating each token requires reading all $n_k$ key and value vectors from KV cache. We present Stochastic Additive No-mulT Attention (SANTA), a method that sparsifies value-cache access by sampling $S \ll n_k$ indices from the post-softmax distribution and aggregates only those value rows. This yields an unbiased estimator of the post-softmax value aggregation while replacing value-stage multiply-accumulates with gather-and-add. We introduce stratified sampling to design variance-reduced, GPU-friendly variants, demonstrating $1.5\times$ decode-step attention kernel speedup over FlashInfer and FlashDecoding on an NVIDIA RTX 6000 Ada while matching baseline accuracy at 32k-token contexts. Finally, we propose Bernoulli $qK^\mathsf{T}$ sampling as a complementary technique to sparsify the score stage, reducing key-feature access through stochastic ternary queries. Both methods are orthogonal to upstream techniques such as ternary quantization, low-rank projections, and KV-cache compression. Together, they point toward sparse, multiplier-free, and energy-efficient inference.