Hybrid Photonic-digital Accelerator for Attention Mechanism

TL;DR

HyAtten is a novel hybrid photonic-digital accelerator for attention mechanisms that reduces signal conversion overhead, boosting performance and energy efficiency in Transformer computations.

Contribution

It introduces a hybrid approach combining photonic and digital processing with a signal comparator to minimize conversion overhead and improve efficiency.

Findings

Achieves 9.8X performance/area improvement over state-of-the-art.

Attains 2.2X energy-efficiency/area enhancement.

Effectively reduces signal conversion overhead in photonic accelerators.

Abstract

The wide adoption and substantial computational resource requirements of attention-based Transformers have spurred the demand for efficient hardware accelerators. Unlike digital-based accelerators, there is growing interest in exploring photonics due to its high energy efficiency and ultra-fast processing speeds. However, the significant signal conversion overhead limits the performance of photonic-based accelerators. In this work, we propose HyAtten, a photonic-based attention accelerator with minimize signal conversion overhead. HyAtten incorporates a signal comparator to classify signals into two categories based on whether they can be processed by low-resolution converters. HyAtten integrates low-resolution converters to process all low-resolution signals, thereby boosting the parallelism of photonic computing. For signals requiring high-resolution conversion, HyAtten uses digital circuits instead of signal converters to reduce area and latency overhead. Compared to state-of-the-art photonic-based Transformer accelerator, HyAtten achieves 9.8X performance/area and 2.2X energy-efficiency/area improvement.