用光子随机计算加速Transformer，效率更高更省电。
Sustainable Transformer Neural Network Acceleration with Stochastic Photonic Computing
- 采用光子随机乘法器与无串扰结构处理动态张量运算
- 相比现有加速器提升7.6倍速度，能耗降低1.3倍
- 适合需要高能效的Transformer推理场景
Transformer在自然语言处理、视觉和科学计算中表现卓越，但计算与内存需求高昂。为应对挑战，我们提出ASTRA，首个基于硅光子学的随机计算Transformer加速器。ASTRA通过新型光学随机乘法器和低串扰架构中的单路/模拟同调累加，高效处理动态张量计算。评估显示，相比当前最优加速器，其至少实现7.6倍加速，能量开销降低1.3倍，凸显其在高效、可扩展、可持续Transformer推理方面的潜力。
原文摘要 · Abstract (English)
Transformers achieve state-of-the-art performance in natural language processing, vision, and scientific computing, but demand high computation and memory. To address these challenges, we present ASTRA, the first silicon-photonic accelerator leveraging stochastic computing for transformers. ASTRA employs novel optical stochastic multipliers and unary/analog homodyne accumulation in a crosstalk-minimal organization to efficiently process dynamic tensor computations. Evaluations show at least 7.6x speedup and 1.3x lower energy overheads compared to state-of-the-art accelerators, highlighting ASTRA's potential for efficient, scalable, and sustainable transformer inference.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。