Low Power Vision Transformer Accelerator with Hardware-Aware Pruning and Optimized Dataflow

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hsiung, Ching-Lin, Chang, Tian-Sheuan
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913074398625792
author Hsiung, Ching-Lin
Chang, Tian-Sheuan
author_facet Hsiung, Ching-Lin
Chang, Tian-Sheuan
contents Current transformer accelerators primarily focus on optimizing self-attention due to its quadratic complexity. However, this focus is less relevant for vision transformers with short token lengths, where the Feed-Forward Network (FFN) tends to be the dominant computational bottleneck. This paper presents a low power Vision Transformer accelerator, optimized through algorithm-hardware co-design. The model complexity is reduced using hardware-friendly dynamic token pruning without introducing complex mechanisms. Sparsity is further improved by replacing GELU with ReLU activations and employing dynamic FFN2 pruning, achieving a 61.5\% reduction in operations and a 59.3\% reduction in FFN2 weights, with an accuracy loss of less than 2\%. The hardware adopts a row-wise dataflow with output-oriented data access to eliminate data transposition, and supports dynamic operations with minimal area overhead. Implemented in TSMC's 28nm CMOS technology, our design occupies 496.4K gates and includes a 232KB SRAM buffer, achieving a peak throughput of 1024 GOPS at 1GHz, with an energy efficiency of 2.31 TOPS/W and an area efficiency of 858.61 GOPS/mm2.
format Preprint
id arxiv_https___arxiv_org_abs_2510_14393
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Low Power Vision Transformer Accelerator with Hardware-Aware Pruning and Optimized Dataflow
Hsiung, Ching-Lin
Chang, Tian-Sheuan
Hardware Architecture
Machine Learning
Current transformer accelerators primarily focus on optimizing self-attention due to its quadratic complexity. However, this focus is less relevant for vision transformers with short token lengths, where the Feed-Forward Network (FFN) tends to be the dominant computational bottleneck. This paper presents a low power Vision Transformer accelerator, optimized through algorithm-hardware co-design. The model complexity is reduced using hardware-friendly dynamic token pruning without introducing complex mechanisms. Sparsity is further improved by replacing GELU with ReLU activations and employing dynamic FFN2 pruning, achieving a 61.5\% reduction in operations and a 59.3\% reduction in FFN2 weights, with an accuracy loss of less than 2\%. The hardware adopts a row-wise dataflow with output-oriented data access to eliminate data transposition, and supports dynamic operations with minimal area overhead. Implemented in TSMC's 28nm CMOS technology, our design occupies 496.4K gates and includes a 232KB SRAM buffer, achieving a peak throughput of 1024 GOPS at 1GHz, with an energy efficiency of 2.31 TOPS/W and an area efficiency of 858.61 GOPS/mm2.
title Low Power Vision Transformer Accelerator with Hardware-Aware Pruning and Optimized Dataflow
topic Hardware Architecture
Machine Learning
url https://arxiv.org/abs/2510.14393