LiteVAR: Compressing Visual Autoregressive Modelling with Efficient Attention and Quantization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xie, Rui, Zhao, Tianchen, Yuan, Zhihang, Wan, Rui, Gao, Wenxi, Zhu, Zhenhua, Ning, Xuefei, Wang, Yu
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915035479015424
author Xie, Rui
Zhao, Tianchen
Yuan, Zhihang
Wan, Rui
Gao, Wenxi
Zhu, Zhenhua
Ning, Xuefei
Wang, Yu
author_facet Xie, Rui
Zhao, Tianchen
Yuan, Zhihang
Wan, Rui
Gao, Wenxi
Zhu, Zhenhua
Ning, Xuefei
Wang, Yu
contents Visual Autoregressive (VAR) has emerged as a promising approach in image generation, offering competitive potential and performance comparable to diffusion-based models. However, current AR-based visual generation models require substantial computational resources, limiting their applicability on resource-constrained devices. To address this issue, we conducted analysis and identified significant redundancy in three dimensions of the VAR model: (1) the attention map, (2) the attention outputs when using classifier free guidance, and (3) the data precision. Correspondingly, we proposed efficient attention mechanism and low-bit quantization method to enhance the efficiency of VAR models while maintaining performance. With negligible performance lost (less than 0.056 FID increase), we could achieve 85.2% reduction in attention computation, 50% reduction in overall memory and 1.5x latency reduction. To ensure deployment feasibility, we developed efficient training-free compression techniques and analyze the deployment feasibility and efficiency gain of each technique.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17178
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LiteVAR: Compressing Visual Autoregressive Modelling with Efficient Attention and Quantization
Xie, Rui
Zhao, Tianchen
Yuan, Zhihang
Wan, Rui
Gao, Wenxi
Zhu, Zhenhua
Ning, Xuefei
Wang, Yu
Computer Vision and Pattern Recognition
Visual Autoregressive (VAR) has emerged as a promising approach in image generation, offering competitive potential and performance comparable to diffusion-based models. However, current AR-based visual generation models require substantial computational resources, limiting their applicability on resource-constrained devices. To address this issue, we conducted analysis and identified significant redundancy in three dimensions of the VAR model: (1) the attention map, (2) the attention outputs when using classifier free guidance, and (3) the data precision. Correspondingly, we proposed efficient attention mechanism and low-bit quantization method to enhance the efficiency of VAR models while maintaining performance. With negligible performance lost (less than 0.056 FID increase), we could achieve 85.2% reduction in attention computation, 50% reduction in overall memory and 1.5x latency reduction. To ensure deployment feasibility, we developed efficient training-free compression techniques and analyze the deployment feasibility and efficiency gain of each technique.
title LiteVAR: Compressing Visual Autoregressive Modelling with Efficient Attention and Quantization
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.17178