TRON: Transformer Neural Network Acceleration with Non-Coherent Silicon Photonics

2023/03/22 by Salma Afifi, Febin Sunny, Afifi, Salma +5 · 3 citations
Computer Science · Engineering · #FOS: Computer and information sciences #Hardware Architecture (cs.AR) #Machine Learning (cs.LG) #Neural Networks and Reservoir Computing #Optical Network Technologies #Photonic and Optical Devices

paper · pdf · doi:10.48550/arxiv.2303.12914

openalex publication_date 2023/03/22 · openalex created_date 2023/03/25 · openalex updated_date 2026/07/28

Abstract

Transformer neural networks are rapidly being integrated into state-of-the-art solutions for natural language processing (NLP) and computer vision. However, the complex structure of these models creates challenges for accelerating their execution on conventional electronic platforms. We propose the first silicon photonic hardware neural network accelerator called TRON for transformer-based models such as BERT, and Vision Transformers. Our analysis demonstrates that TRON exhibits at least 14x better throughput and 8x better energy efficiency, in comparison to state-of-the-art transformer accelerators.

Cited by

• Implementation of transformer-based LLMs with large-scale optoelectronic neurons on a CMOS image sensor platform
• Integrated photonic neuromorphic computing: device, architecture, chip, algorithm
• Opto-ViT: Architecting a Near-Sensor Region of Interest-Aware Vision Transformer Accelerator with Silicon Photonics

Related

• Accelerating Neural Networks for Large Language Models and Graph Processing with Silicon Photonics
• Hardware Accelerator for Multi-Head Attention and Position-Wise Feed-Forward in the Transformer
• A Runtime-Adaptive Transformer Neural Network Accelerator on FPGAs
• Weight Bank Addition Photonic Accelerator for Artificial Intelligence
• Optical Transformers