Skip to content

Add CapPress#234

Merged
alessiodevoto merged 5 commits into
NVIDIA:mainfrom
jiamingyy:main
Jun 10, 2026
Merged

Add CapPress#234
alessiodevoto merged 5 commits into
NVIDIA:mainfrom
jiamingyy:main

Conversation

@jiamingyy

Copy link
Copy Markdown
Contributor

Add CapPress, a new KV cache compression method implementing CAPKV from paper (https://arxiv.org/abs/2604.25975). The press evicts tokens based on query-aware capacity scores computed from a log-determinant leverage proxy, using value vectors as output-direction proxies with historical-query anchoring.

@copy-pr-bot

copy-pr-bot Bot commented Jun 3, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@alessiodevoto alessiodevoto left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for opening this, @jiamingyy ! The PR looks good overall! I left a few comments on things that need to be updated. Please take a look when you get a chance. Looping in @maxjeblick as well in case he has additional feedback.

Comment thread kvpress/presses/cap_press.py Outdated
Comment thread kvpress/presses/cap_press.py Outdated
Comment thread tests/presses/test_cap_press.py Outdated
Comment thread kvpress/presses/cap_press.py Outdated
Comment thread kvpress/presses/cap_press.py
Comment thread kvpress/presses/cap_press.py
@jiamingyy jiamingyy force-pushed the main branch 2 times, most recently from 119e373 to bf55c89 Compare June 9, 2026 13:35
@jiamingyy jiamingyy requested a review from alessiodevoto June 9, 2026 13:56
@jiamingyy

Copy link
Copy Markdown
Contributor Author

I would like to thank @alessiodevoto for the valuable review. I have resolved all the issues raised.

@alessiodevoto

Copy link
Copy Markdown
Collaborator

/ok to test cded71a

@alessiodevoto

Copy link
Copy Markdown
Collaborator

/ok to test 6817249

@alessiodevoto

Copy link
Copy Markdown
Collaborator

LGTM @jiamingyy ! Thank you for your contribution to KVPress 🎉

@alessiodevoto alessiodevoto merged commit fa42106 into NVIDIA:main Jun 10, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants