close this message
arXiv smileybones

arXiv Is Hiring a DevOps Engineer

Work on one of the world's most important websites and make an impact on open science.

View Jobs
Skip to main content
Cornell University

arXiv Is Hiring a DevOps Engineer

View Jobs
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > cs > arXiv:2403.09636

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Computation and Language

arXiv:2403.09636 (cs)
[Submitted on 14 Mar 2024 (v1), last revised 23 Jul 2024 (this version, v2)]

Title:Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference

Authors:Piotr Nawrot, Adrian Łańcucki, Marcin Chochowski, David Tarjan, Edoardo M. Ponti
View a PDF of the paper titled Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference, by Piotr Nawrot and 4 other authors
View PDF HTML (experimental)
Abstract:Transformers have emerged as the backbone of large language models (LLMs). However, generation remains inefficient due to the need to store in memory a cache of key-value representations for past tokens, whose size scales linearly with the input sequence length and batch size. As a solution, we propose Dynamic Memory Compression (DMC), a method for online key-value cache compression at inference time. Most importantly, the model learns to apply different compression ratios in different heads and layers. We retrofit pre-trained LLMs such as Llama 2 (7B, 13B and 70B) into DMC Transformers, achieving up to 7x throughput increase during auto-regressive inference on an NVIDIA H100 GPU. DMC is applied via continued pre-training on a negligible percentage of the original data without adding any extra parameters. DMC preserves the original downstream performance with up to 4x cache compression, outperforming up-trained grouped-query attention (GQA) and key-value eviction policies (H$_2$O, TOVA). GQA and DMC can be even combined to obtain compounded gains. Hence, DMC can serve as a drop-in replacement for KV caching in existing LLMs to fit longer contexts and larger batches within any given memory budget.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2403.09636 [cs.CL]
  (or arXiv:2403.09636v2 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2403.09636
arXiv-issued DOI via DataCite
Journal reference: Proceedings of the 41st International Conference on Machine Learning (2024) 37396-37412

Submission history

From: Piotr Nawrot [view email]
[v1] Thu, 14 Mar 2024 17:59:26 UTC (438 KB)
[v2] Tue, 23 Jul 2024 17:55:30 UTC (468 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference, by Piotr Nawrot and 4 other authors
  • View PDF
  • HTML (experimental)
  • TeX Source
  • Other Formats
license icon view license
Current browse context:
cs.CL
< prev   |   next >
new | recent | 2024-03
Change to browse by:
cs

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar
a export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Reddit logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)

Demos

Replicate (What is Replicate?)
Hugging Face Spaces (What is Spaces?)
TXYZ.AI (What is TXYZ.AI?)

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
  • Author
  • Venue
  • Institution
  • Topic

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
  • About
  • Help
  • contact arXivClick here to contact arXiv Contact
  • subscribe to arXiv mailingsClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack