CoChat
Sign Up
Cited by 79Advanced Neural Network ApplicationsTopic ModelingParallel Computing and Optimization Techniques

KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization

Open access
Hooper, Coleman, Kim, Sehoon, Mohammadzadeh, Hiva, Mahoney, Michael, Shao, Yakun, Keutzer, Kurt, Gholami, Amir·2024·Advances in Neural Information Processing Systems 37
DOI 10.52202/079017-0040
Open this paper in CoChat — free accountOpen PDF

Abstract

LLMs are seeing growing use for applications which require large context windows, and with these large context windows KV cache activations surface as the dominant contributor to memory consumption during inference. Quantization is a promising approach for compressing KV cache activations; however, existing solutions fail to represent activations accurately in sub-4-bit precision. Our work, KVQuant, facilitates low precision KV cache quantization by incorporating several novel methods: (i) Per-Channel Key Quantization, where we adjust the dimension along which we quantize the Key activations to better match the distribution; (ii) Pre-RoPE Key Quantization, where we quantize Key activations before the rotary positional embedding to mitigate its impact on quantization; (iii) Non-Uniform KV Cache Quantization, where we derive per-layer sensitivity-weighted non-uniform datatypes that better represent the distributions; and (iv) Per-Vector Dense-and-Sparse Quantization, where we isolate outliers separately for each vector to minimize skews in quantization ranges. By applying our method to the LLaMA, Llama-2, Llama-3, and Mistral models, we achieve < 0.1 perplexity degradation with 3-bit quantization on both Wikitext-2 and C4, outperforming existing approaches. Our method enables serving LLaMA-7B with a context length of up to 1 million on a single A100-80GB GPU and up to 10 million on an 8-GPU system. We develop custom CUDA kernels for KVQuant, showing that we can achieve up to ~1.7x speedups, compared to baseline fp16 matrix-vector multiplications, for the LLaMA-7B model.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.52202/079017-0040

Cites

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • KIVI and KVQuant quantize cached values to 2-3 bits with little accuracy loss.

    “KIVI [21] and KVQuant [9] take a different path, quantizing cached values to 2-3 bits with little accuracy loss.”

    From TokenDance: Scaling Multi-Agent LLM Serving via Collective KV Cache Sharing (2026) · cites this paper as [9]

    The LongBench Evaluation reports that 3-bit KVQuant scores close to the fp16 baseline (31.21 versus 31.96). But its RULER results show substantial accuracy loss at 2 bits (36.54 versus 56.40), so the claim of little accuracy loss does not hold across the stated 2–3-bit range. The citation to this paper is for KVQuant; KIVI has a separate citation. (LLM, conf=0.87)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Still being checked.

Research works printed but not linked: 34

  1. Yushi Bai, Xin Lv, Jiajie Zhang, Hongchang Lyu, Jiankai Tang, Zhidian Huang, Zhengxiao Du, Xiao Liu, Aohan Zeng, Lei Hou, et al. Longbench: A bilingual, multitask benchmark for long context understanding. arXiv preprint arXiv:2308.14508, 2023. · arXiv
  2. Yelysei Bondarenko, Markus Nagel, and Tijmen Blankevoort. Understanding and overcoming the challenges of efficient transformer quantization. arXiv preprint arXiv:2109.12948, 2021. · arXiv
  3. Shouyuan Chen, Sherman Wong, Liangjian Chen, and Yuandong Tian. Extending context window of large language models via positional interpolation. arXiv preprint arXiv:2306.15595, 2023. · arXiv
  4. Yukang Chen, Shengju Qian, Haotian Tang, Xin Lai, Zhijian Liu, Song Han, and Jiaya Jia. Longlora: Efficient fine-tuning of long-context large language models. arXiv preprint arXiv:2309.12307, 2023. · arXiv
  5. Tim Dettmers, Mike Lewis, Younes Belkada, and Luke Zettlemoyer. Llm. int8 (): 8-bit matrix multiplication for transformers at scale. arXiv preprint arXiv:2208.07339, 2022. · arXiv
  6. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. Qlora: Efficient finetuning of quantized llms. arXiv preprint arXiv:2305.14314, 2023. · arXiv
  7. Tim Dettmers, Ruslan Svirschevski, Vage Egiazarian, Denis Kuznedelev, Elias Frantar, Saleh Ashkboos, Alexander Borzunov, Torsten Hoefler, and Dan Alistarh. Spqr: A sparse-quantized representation for near-lossless llm weight compression. arXiv preprint arXiv:2306.03078, 2023. · arXiv
  8. Goran Flegar and Enrique S Quintana-Ortí. Balanced csr sparse matrix-vector product on graphics processors. In Euro-Par 2017: Parallel Processing: 23rd International Conference on Parallel and Distributed Computing, Santiago de Compostela, Spain, August 28-September 1, 2017, Proceedings 23, pages 697-709. Springer, 2017.
  9. Suyu Ge, Yunan Zhang, Liyuan Liu, Minjia Zhang, Jiawei Han, and Jianfeng Gao. Model tells you what to discard: Adaptive kv cache compression for llms. arXiv preprint arXiv:2310.01801, 2023. · arXiv
  10. Chi Han, Qifan Wang, Wenhan Xiong, Yu Chen, Heng Ji, and Sinong Wang. Lm-infinite: Simple on-the-fly length generalization for large language models. arXiv preprint arXiv:2308.16137, 2023. · arXiv
  11. Jung Hwan Heo, Jeonghoon Kim, Beomseok Kwon, Byeongwook Kim, Se Jung Kwon, and Dongsoo Lee. Rethinking channel dimensions to isolate outliers for low-bit weight quantization of large language models. arXiv preprint arXiv:2309.15531, 2023. · arXiv
  12. Cheng-Ping Hsieh, Simeng Sun, Samuel Kriman, Shantanu Acharya, Dima Rekesh, Fei Jia, and Boris Ginsburg. Ruler: What's the real context size of your long-context language models? arXiv preprint arXiv:2404.06654, 2024. · arXiv
  13. Albert Q Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, et al. Mistral 7b. arXiv preprint arXiv:2310.06825, 2023. · arXiv
  14. Sehoon Kim, Coleman Hooper, Amir Gholami, Zhen Dong, Xiuyu Li, Sheng Shen, Michael W Mahoney, and Kurt Keutzer. Squeezellm: Dense-and-sparse quantization. arXiv preprint arXiv:2306.07629, 2023. · arXiv
  15. Sehoon Kim, Coleman Hooper, Thanakul Wattanawong, Minwoo Kang, Ruohan Yan, Hasan Genc, Grace Dinh, Qijing Huang, Kurt Keutzer, Michael W Mahoney, et al. Full stack optimization of transformer inference: a survey. arXiv preprint arXiv:2302.14017, 2023. · arXiv
  16. Dacheng Li, Rulin Shao, Anze Xie, Ying Sheng, Lianmin Zheng, Joseph Gonzalez, Ion Stoica, Xuezhe Ma, and Hao Zhang. How long can context length of open-source llms truly promise? In NeurIPS 2023 Workshop on Instruction Tuning and Instruction Following, 2023.
  17. Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Xingyu Dang, and Song Han. Awq: Activation-aware weight quantization for llm compression and acceleration. arXiv preprint arXiv:2306.00978, 2023. · arXiv
  18. Zechun Liu, Barlas Oguz, Changsheng Zhao, Ernie Chang, Pierre Stock, Yashar Mehdad, Yangyang Shi, Raghuraman Krishnamoorthi, and Vikas Chandra. Llm-qat: Data-free quantiza- tion aware training for large language models. arXiv preprint arXiv:2305.17888, 2023. · arXiv
  19. Zichang Liu, Aditya Desai, Fangshuo Liao, Weitao Wang, Victor Xie, Zhaozhuo Xu, Anastasios Kyrillidis, and Anshumali Shrivastava. Scissorhands: Exploiting the persistence of importance hypothesis for llm kv cache compression at test time. arXiv preprint arXiv:2305.17118, 2023. · arXiv
  20. Amirkeivan Mohtashami and Martin Jaggi. Landmark attention: Random-access infinite context length for transformers. arXiv preprint arXiv:2305.16300, 2023. · arXiv
  21. Roman Novak, Yasaman Bahri, Daniel A Abolafia, Jeffrey Pennington, and Jascha Sohl- Dickstein. Sensitivity and generalization in neural networks: an empirical study. arXiv preprint arXiv:1802.08760, 2018. · arXiv
  22. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. arXiv e-prints, 2019.
  23. Luka Ribar, Ivan Chelombiev, Luke Hudlass-Galley, Charlie Blake, Carlo Luschi, and Douglas Orr. Sparq attention: Bandwidth-efficient llm inference. arXiv preprint arXiv:2312.04985, 2023. · arXiv
  24. Wenqi Shao, Mengzhao Chen, Zhaoyang Zhang, Peng Xu, Lirui Zhao, Zhiqian Li, Kaipeng Zhang, Peng Gao, Yu Qiao, and Ping Luo. Omniquant: Omnidirectionally calibrated quantiza- tion for large language models. arXiv preprint arXiv:2308.13137, 2023. · arXiv
  25. Ying Sheng, Lianmin Zheng, Binhang Yuan, Zhuohan Li, Max Ryabinin, Beidi Chen, Percy Liang, Christopher Ré, Ion Stoica, and Ce Zhang. Flexgen: High-throughput generative inference of large language models with a single gpu. In International Conference on Machine Learning, pages 31094-31116. PMLR, 2023.
  26. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timo- thée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023. · arXiv
  27. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023. · arXiv
  28. Xiuying Wei, Yunchen Zhang, Yuhang Li, Xiangguo Zhang, Ruihao Gong, Jinyang Guo, and Xianglong Liu. Outlier suppression+: Accurate quantization of large language models by equivalent and optimal shifting and scaling. arXiv preprint arXiv:2304.09145, 2023. · arXiv
  29. Xiuying Wei, Yunchen Zhang, Xiangguo Zhang, Ruihao Gong, Shanghang Zhang, Qi Zhang, Fengwei Yu, and Xianglong Liu. Outlier suppression: Pushing the limit of low-bit transformer language models. Advances in Neural Information Processing Systems, 35:17402-17414, 2022.
  30. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lhoest, and Alexander M. Rush. Transformers: State-of-the- art natural language processing. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 38-45, Online, October 2020. Association for Computational Linguistics.
  31. Guangxuan Xiao, Ji Lin, Mickael Seznec, Hao Wu, Julien Demouth, and Song Han. Smoothquant: Accurate and efficient post-training quantization for large language models. In International Conference on Machine Learning, pages 38087-38099. PMLR, 2023.
  32. Zhenyu Zhang, Ying Sheng, Tianyi Zhou, Tianlong Chen, Lianmin Zheng, Ruisi Cai, Zhao Song, Yuandong Tian, Christopher Ré, Clark Barrett, et al. H _2 o: Heavy-hitter oracle for efficient generative inference of large language models. arXiv preprint arXiv:2306.14048, 2023. · arXiv
  33. Yilong Zhao, Chien-Yu Lin, Kan Zhu, Zihao Ye, Lequn Chen, Size Zheng, Luis Ceze, Arvind Krishnamurthy, Tianqi Chen, and Baris Kasikci. Atom: Low-bit quantization for efficient and accurate llm serving. arXiv preprint arXiv:2310.19102, 2023. · arXiv
  34. Dawei Zhu, Nan Yang, Liang Wang, Yifan Song, Wenhao Wu, Furu Wei, and Sujian Li. Pose: Efficient context window extension of llms via positional skip-wise training. arXiv preprint arXiv:2309.10400, 2023. · arXiv

Cited by

Sorted by citations

Related

Authors
Hooper, C., Kim, S., Mohammadzadeh, H., Mahoney, M., Shao, Y., Keutzer, K., & Gholami, A.
ORCID
Venue
Advances in Neural Information Processing Systems 37
Publisher
Neural Information Processing Systems Foundation, Inc. (NeurIPS)
Published
2024
Pages
1270-1303
ISSN
1049-5258 (ISSN-L)
DOI
10.52202/079017-0040
OpenAlex
W4415799217
Cited by
69
References
0
Citations by year
2026: 53 · 2025: 12 · 2024: 4
FWCI
14.26
Type
conference paper
Language
en
Open access
Locations
  • neural information processing systems (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Funders
  • National Science Foundation
  • U.S. Department of Energy
  • Intel Corporation
  • Nvidia
  • Samsung Advanced Institute of Technology
  • Korea Foundation for Advanced Studies
  • Samsung
  • Office of Naval Research
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
Keywords
KV cache quantization, long-context inference, large language models, key-value cache, memory-efficient inference

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review