Exact Causal Attention with 10% Fewer Operations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Rybin, Dmitry, Zhang, Yushun, Tian, Ding, Lin, Zhihang, Luo, Zhi-Quan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914087350304768
author Rybin, Dmitry
Zhang, Yushun
Tian, Ding
Lin, Zhihang
Luo, Zhi-Quan
author_facet Rybin, Dmitry
Zhang, Yushun
Tian, Ding
Lin, Zhihang
Luo, Zhi-Quan
contents We present Exact Causal Attention (ECA), a Strassen-style algorithm that computes exact Causal Attention using 10\% fewer operations. ECA improves a special class of matrix multiplications where either one operand or the output matrix is upper- or lower-triangular. This includes all matrix multiplication operations in the forward and backward pass of Causal Attention, such as masked product $\mathrm{Mask}(QK^{T})$. ECA is built upon algebraic identities discovered via machine learning and combinatorial search. We note that ECA cannot accelerate fused kernels such as FlashAttention on GPU. This is because ECA requires materialization of large intermediate expressions in the memory, while FlashAttention does not. However, it provides an alternative approach for compute-bound applications and can potentially be useful in scenarios with FLOPs considerations.
format Preprint
id arxiv_https___arxiv_org_abs_2510_05175
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Exact Causal Attention with 10% Fewer Operations
Rybin, Dmitry
Zhang, Yushun
Tian, Ding
Lin, Zhihang
Luo, Zhi-Quan
Machine Learning
Discrete Mathematics
Data Structures and Algorithms
We present Exact Causal Attention (ECA), a Strassen-style algorithm that computes exact Causal Attention using 10\% fewer operations. ECA improves a special class of matrix multiplications where either one operand or the output matrix is upper- or lower-triangular. This includes all matrix multiplication operations in the forward and backward pass of Causal Attention, such as masked product $\mathrm{Mask}(QK^{T})$. ECA is built upon algebraic identities discovered via machine learning and combinatorial search. We note that ECA cannot accelerate fused kernels such as FlashAttention on GPU. This is because ECA requires materialization of large intermediate expressions in the memory, while FlashAttention does not. However, it provides an alternative approach for compute-bound applications and can potentially be useful in scenarios with FLOPs considerations.
title Exact Causal Attention with 10% Fewer Operations
topic Machine Learning
Discrete Mathematics
Data Structures and Algorithms
url https://arxiv.org/abs/2510.05175