QUICK: Quantization-aware Interleaving and Conflict-free Kernel for efficient LLM inference

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Kim, Taesu, Lee, Jongho, Ahn, Daehyun, Kim, Sarang, Choi, Jiwoong, Kim, Minkyu, Kim, Hyungjun
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914680192106496
author Kim, Taesu
Lee, Jongho
Ahn, Daehyun
Kim, Sarang
Choi, Jiwoong
Kim, Minkyu
Kim, Hyungjun
author_facet Kim, Taesu
Lee, Jongho
Ahn, Daehyun
Kim, Sarang
Choi, Jiwoong
Kim, Minkyu
Kim, Hyungjun
contents We introduce QUICK, a group of novel optimized CUDA kernels for the efficient inference of quantized Large Language Models (LLMs). QUICK addresses the shared memory bank-conflict problem of state-of-the-art mixed precision matrix multiplication kernels. Our method interleaves the quantized weight matrices of LLMs offline to skip the shared memory write-back after the dequantization. We demonstrate up to 1.91x speedup over existing kernels of AutoAWQ on larger batches and up to 1.94x throughput gain on representative LLM models on various NVIDIA GPU devices.
format Preprint
id arxiv_https___arxiv_org_abs_2402_10076
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle QUICK: Quantization-aware Interleaving and Conflict-free Kernel for efficient LLM inference
Kim, Taesu
Lee, Jongho
Ahn, Daehyun
Kim, Sarang
Choi, Jiwoong
Kim, Minkyu
Kim, Hyungjun
Machine Learning
Artificial Intelligence
Computation and Language
We introduce QUICK, a group of novel optimized CUDA kernels for the efficient inference of quantized Large Language Models (LLMs). QUICK addresses the shared memory bank-conflict problem of state-of-the-art mixed precision matrix multiplication kernels. Our method interleaves the quantized weight matrices of LLMs offline to skip the shared memory write-back after the dequantization. We demonstrate up to 1.91x speedup over existing kernels of AutoAWQ on larger batches and up to 1.94x throughput gain on representative LLM models on various NVIDIA GPU devices.
title QUICK: Quantization-aware Interleaving and Conflict-free Kernel for efficient LLM inference
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2402.10076