English

QUIK: Towards End-to-End 4-Bit Inference on Generative Large Language Models

Machine Learning 2023-11-03 v2

Abstract

Large Language Models (LLMs) from the GPT family have become extremely popular, leading to a race towards reducing their inference costs to allow for efficient local computation. Yet, the vast majority of existing work focuses on weight-only quantization, which can reduce runtime costs in the memory-bound one-token-at-a-time generative setting, but does not address them in compute-bound scenarios, such as batched inference or prompt processing. In this paper, we address the general quantization problem, where both weights and activations should be quantized. We show, for the first time, that the majority of inference computations for large generative models such as LLaMA, OPT, and Falcon can be performed with both weights and activations being cast to 4 bits, in a way that leads to practical speedups, while at the same time maintaining good accuracy. We achieve this via a hybrid quantization strategy called QUIK, which compresses most of the weights and activations to 4-bit, while keeping some outlier weights and activations in higher-precision. The key feature of our scheme is that it is designed with computational efficiency in mind: we provide GPU kernels matching the QUIK format with highly-efficient layer-wise runtimes, which lead to practical end-to-end throughput improvements of up to 3.4x relative to FP16 execution. We provide detailed studies for models from the OPT, LLaMA-2 and Falcon families, as well as a first instance of accurate inference using quantization plus 2:4 sparsity. Code is available at: https://github.com/IST-DASLab/QUIK.

Keywords

Cite

@article{arxiv.2310.09259,
  title  = {QUIK: Towards End-to-End 4-Bit Inference on Generative Large Language Models},
  author = {Saleh Ashkboos and Ilia Markov and Elias Frantar and Tingxuan Zhong and Xincheng Wang and Jie Ren and Torsten Hoefler and Dan Alistarh},
  journal= {arXiv preprint arXiv:2310.09259},
  year   = {2023}
}

Comments

16 pages

R2 v1 2026-06-28T12:50:07.631Z