Skip to main content
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arXiv logo
Cornell University Logo

Computer Science > Computation and Language

arXiv:2412.05693 (cs)
[Submitted on 7 Dec 2024 (v1), last revised 3 Jul 2025 (this version, v3)]

Title:Batch-Max: Higher LLM Throughput using Larger Batch Sizes and KV Cache Compression

Authors:Michael R. Metel, Boxing Chen, Mehdi Rezagholizadeh
View PDF HTML (experimental)
Abstract:Several works have developed eviction policies to remove key-value (KV) pairs from the KV cache for more efficient inference. The focus has been on compressing the KV cache after the input prompt has been processed for faster token generation. In settings with limited GPU memory, and when the input context is longer than the generation length, we show that by also compressing the KV cache during the input processing phase, larger batch sizes can be used resulting in significantly higher throughput while still maintaining the original model's accuracy.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2412.05693 [cs.CL]
  (or arXiv:2412.05693v3 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2412.05693

Submission history

From: Michael Metel R [view email]
[v1] Sat, 7 Dec 2024 16:41:54 UTC (27 KB)
[v2] Tue, 17 Jun 2025 02:24:51 UTC (29 KB)
[v3] Thu, 3 Jul 2025 16:06:35 UTC (29 KB)
Full-text links:

Access Paper:

view license
Current browse context:
cs.CL
< prev   |   next >
new | recent | 2024-12
Change to browse by:
cs

References & Citations

export BibTeX citation

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?)

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?)