SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models

TL;DR

SmoothQuant is a post-training quantization method that enables efficient 8-bit quantization of large language models without significant accuracy loss, reducing hardware costs and enabling larger models to run on limited hardware.

Contribution

It introduces a novel transformation that shifts quantization difficulty from activations to weights, allowing accurate 8-bit quantization of LLMs without retraining.

Findings

Up to 1.56x speedup in inference

2x reduction in memory usage

Effective on multiple large language models

Abstract

Large language models (LLMs) show excellent performance but are compute- and memory-intensive. Quantization can reduce memory and accelerate inference. However, existing methods cannot maintain accuracy and hardware efficiency at the same time. We propose SmoothQuant, a training-free, accuracy-preserving, and general-purpose post-training quantization (PTQ) solution to enable 8-bit weight, 8-bit activation (W8A8) quantization for LLMs. Based on the fact that weights are easy to quantize while activations are not, SmoothQuant smooths the activation outliers by offline migrating the quantization difficulty from activations to weights with a mathematically equivalent transformation. SmoothQuant enables an INT8 quantization of both weights and activations for all the matrix multiplications in LLMs, including OPT, BLOOM, GLM, MT-NLG, Llama-1/2, Falcon, Mistral, and Mixtral models. We demonstrate up to 1.56x speedup and 2x memory reduction for LLMs with negligible loss in accuracy. SmoothQuant enables serving 530B LLM within a single node. Our work offers a turn-key solution that reduces hardware costs and democratizes LLMs. Code is available at https://github.com/mit-han-lab/smoothquant.