How to Parameterize Asymmetric Quantization Ranges for Quantization-Aware Training

2024/04/25 by Jaeseong You, You, Jaeseong, Minseop Park +9 · 1 citation
Computer Science · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Neural Networks and Applications

paper · pdf · doi:10.48550/arxiv.2404.16898

openalex publication_date 2024/04/25 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This paper investigates three different parameterizations of asymmetric uniform quantization for quantization-aware training: (1) scale and offset, (2) minimum and maximum, and (3) beta and gamma. We perform a comprehensive comparative analysis of these parameterizations' influence on quantization-aware training, using both controlled experiments and real-world large language models. Our particular focus is on their changing behavior in response to critical training hyperparameters, bit width and learning rate. Based on our investigation, we propose best practices to stabilize and accelerate quantization-aware training with learnable asymmetric quantization ranges.

Cited by

• TAH-QUANT: Effective Activation Quantization in Pipeline Parallelism over Slow Network

Related

• Scheduling Weight Transitions for Quantization-Aware Training
• Training Dynamics Impact Post-Training Quantization Robustness
• Scaling Law for Quantization-Aware Training
• Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization
• Compensate Quantization Errors+: Quantized Models Are Inquisitive Learners