7 papers
EPTS: Elastic Post-Training Sparsity for Efficient Large Language Model Compression
Ke Xu, Jiaqi Wan, Wenhao Hu +2
Post-Training Sparsity (PTS) has emerged as a crucial paradigm for compressing Large Language Models to facilitate efficient deployment on resource-constrained devices. However, ex…
Rethinking Residual Errors in Compensation-based LLM Quantization
Shuaiting Li, Juncan Deng, Kedong Xu +5
Methods based on weight compensation, which iteratively apply quantization and weight compensation to minimize the output error, have recently demonstrated remarkable success in qu…
SSVQ: Unleashing the Potential of Vector Quantization with Sign-Splitting
Shuaiting Li, Juncan Deng, Chenxuan Wang +5
Vector Quantization (VQ) has emerged as a prominent weight compression technique, showcasing substantially lower quantization errors than uniform quantization across diverse models…
ViM-VQ: Efficient Post-Training Vector Quantization for Visual Mamba
Juncan Deng, Shuaiting Li, Zeyu Wang +3
Visual Mamba networks (ViMs) extend the selective state space model (Mamba) to various vision tasks and demonstrate significant potential. As a promising compression technique, vec…
Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion
Shuaiting Li, Juncan Deng, Zeyu Wang +5
Text-to-image generation via Stable Diffusion models (SDM) have demonstrated remarkable capabilities. However, their computational intensity, particularly in the iterative denoisin…
VQ4ALL: Efficient Neural Network Representation via a Universal Codebook
Juncan Deng, Shuaiting Li, Zeyu Wang +3
The rapid growth of the big neural network models puts forward new requirements for lightweight network representation methods. The traditional methods based on model compression h…