Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhao, Weibo, Shi, Yubin, Lyu, Xinyu, Sui, Wanchen, Li, Shen, Li, Yong
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2411.07762
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909425127653376
author Zhao, Weibo
Shi, Yubin
Lyu, Xinyu
Sui, Wanchen
Li, Shen
Li, Yong
author_facet Zhao, Weibo
Shi, Yubin
Lyu, Xinyu
Sui, Wanchen
Li, Shen
Li, Yong
contents Quantization stands as a pivotal technique for large language model (LLM) serving, yet it poses significant challenges particularly in achieving effective low-bit quantization. The limited numerical mapping makes the quantized model produce a non-trivial error, bringing out intolerable performance degration. This paper is anchored in the basic idea of model compression objectives, and delves into the layer-wise error distribution of LLMs during post-training quantization. Subsequently, we introduce ASER, an algorithm consisting of (1) Error Reconstruction: low-rank compensation for quantization error with LoRA-style matrices constructed by whitening SVD; (2) Activation Smoothing: outlier extraction to gain smooth activation and better error compensation. ASER is capable of quantizing typical LLMs to low-bit ones, particularly preserving accuracy even in W4A8 per-channel setup. Experimental results show that ASER is competitive among the state-of-the-art quantization algorithms, showing potential to activation quantization, with minor overhead.
format Preprint
id arxiv_https___arxiv_org_abs_2411_07762
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ASER: Activation Smoothing and Error Reconstruction for Large Language Model Quantization
Zhao, Weibo
Shi, Yubin
Lyu, Xinyu
Sui, Wanchen
Li, Shen
Li, Yong
Machine Learning
Artificial Intelligence
Quantization stands as a pivotal technique for large language model (LLM) serving, yet it poses significant challenges particularly in achieving effective low-bit quantization. The limited numerical mapping makes the quantized model produce a non-trivial error, bringing out intolerable performance degration. This paper is anchored in the basic idea of model compression objectives, and delves into the layer-wise error distribution of LLMs during post-training quantization. Subsequently, we introduce ASER, an algorithm consisting of (1) Error Reconstruction: low-rank compensation for quantization error with LoRA-style matrices constructed by whitening SVD; (2) Activation Smoothing: outlier extraction to gain smooth activation and better error compensation. ASER is capable of quantizing typical LLMs to low-bit ones, particularly preserving accuracy even in W4A8 per-channel setup. Experimental results show that ASER is competitive among the state-of-the-art quantization algorithms, showing potential to activation quantization, with minor overhead.
title ASER: Activation Smoothing and Error Reconstruction for Large Language Model Quantization
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2411.07762