Saved in:
Bibliographic Details
Main Authors: Zhu, Hanqing, Zhou, Zhican, Ning, Shupeng, Wu, Xuhao, Chen, Ray, Wan, Yating, Pan, David
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.01673
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916985388924928
author Zhu, Hanqing
Zhou, Zhican
Ning, Shupeng
Wu, Xuhao
Chen, Ray
Wan, Yating
Pan, David
author_facet Zhu, Hanqing
Zhou, Zhican
Ning, Shupeng
Wu, Xuhao
Chen, Ray
Wan, Yating
Pan, David
contents Photonic computing has emerged as a promising substrate for accelerating the dense linear-algebra operations at the heart of AI, yet adoption for large Transformer models remains in its infancy. We identify two bottlenecks: (1) costly electro--optic conversions and data-movement overheads that erode energy efficiency as model sizes scale; (2) a mismatch between limited on-chip photonic resources and Transformer scale, which forces frequent reuse of photonic tensor cores and dilutes throughput gains. To address these challenges, we introduce a hardware--software co-design framework. First, we propose \texttt{Lighten}, a PTC-aware compression flow that post-hoc decomposes each Transformer weight matrix into a low-rank component plus a structured-sparse component aligned to photonic tensor-core granularity, without lengthy retraining. Second, we present \texttt{ENLighten}, a reconfigurable photonic accelerator with dynamically adaptive tensor cores, driven by broadband light redistribution, enabling fine-grained sparsity support and full power gating of inactive parts. On ImageNet, \texttt{Lighten} prunes a Base-scale Vision Transformer by 50\% with $\approx$1\% accuracy drop after only 3 epochs (about 1 hour) of fine-tuning. Deployed on \texttt{ENLighten}, it achieves a $2.5\times$ improvement in energy--delay product over the state-of-the-art photonic Transformer accelerator.
format Preprint
id arxiv_https___arxiv_org_abs_2510_01673
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ENLighten: Lighten the Transformer, Enable Efficient Optical Acceleration
Zhu, Hanqing
Zhou, Zhican
Ning, Shupeng
Wu, Xuhao
Chen, Ray
Wan, Yating
Pan, David
Emerging Technologies
Photonic computing has emerged as a promising substrate for accelerating the dense linear-algebra operations at the heart of AI, yet adoption for large Transformer models remains in its infancy. We identify two bottlenecks: (1) costly electro--optic conversions and data-movement overheads that erode energy efficiency as model sizes scale; (2) a mismatch between limited on-chip photonic resources and Transformer scale, which forces frequent reuse of photonic tensor cores and dilutes throughput gains. To address these challenges, we introduce a hardware--software co-design framework. First, we propose \texttt{Lighten}, a PTC-aware compression flow that post-hoc decomposes each Transformer weight matrix into a low-rank component plus a structured-sparse component aligned to photonic tensor-core granularity, without lengthy retraining. Second, we present \texttt{ENLighten}, a reconfigurable photonic accelerator with dynamically adaptive tensor cores, driven by broadband light redistribution, enabling fine-grained sparsity support and full power gating of inactive parts. On ImageNet, \texttt{Lighten} prunes a Base-scale Vision Transformer by 50\% with $\approx$1\% accuracy drop after only 3 epochs (about 1 hour) of fine-tuning. Deployed on \texttt{ENLighten}, it achieves a $2.5\times$ improvement in energy--delay product over the state-of-the-art photonic Transformer accelerator.
title ENLighten: Lighten the Transformer, Enable Efficient Optical Acceleration
topic Emerging Technologies
url https://arxiv.org/abs/2510.01673