Skip to content
Navigation Menu
Sign in
Products
Platform
GitHub
Repository hosting, code review, and collaboration
Copilot
AI coding agent for tasks, projects and automated work
Build & Ship
Actions
CI/CD and automation
Code Quality
Quality analysis, fixes, and enforcement
Codespaces
Cloud development environments
Secure
Code Security
Vulnerability detection and remediation
Secret Protection
Secret scanning and leak prevention
All Features
Developers
Learn
Documentation
Training
Certifications
Build
Build with Copilot
Customize Copilot
Integrations
Explore
Trending repos
GitHub Sponsors
Events
Changelog
Changelog
Your source for the latest features, improvements, and fixes across GitHub products, all in one place.
Businesses
By Stage
Enterprises
Team
Startups
Nonprofits
By Industry
Healthcare
Financial services
Manufacturing
Government
See all industries
Evaluate GitHub
Why GitHub
Customer Stories
Business insights
Trust Center
Add-ons
GitHub Advanced Security
Enterprise-grade code security
Copilot for business
Enterprise-grade AI features
Premium Support
Enterprise-grade 24/7 support
Resources
Learn & Connect
Documentation
Resource center
Blog
Events & webinars
Community Forum
Customer Stories
Integrations
Support
Customer support
Training
GitHub Status
Company
About
Partners
Changelog
GitHub Universe
GitHub Universe
The event for developers, by developers — talks, workshops, and what’s next for GitHub.
Pricing
Search
/
Sign in
Sign up
Appearance settings
You signed in with another tab or window.
Reload
to refresh your session.
You signed out in another tab or window.
Reload
to refresh your session.
You switched accounts on another tab or window.
Reload
to refresh your session.
Dismiss alert
{{ message }}
NVIDIA
TensorRT-LLM
Repository navigation
Code
Issues
693
(693)
Pull requests
1.1k
(1.1k)
Discussions
Actions
Projects
Security and quality
Insights
More
items
[None][feat] KVCM2 core: sparse attention arena, host-backed pools and Batch
- #21129
#21129
Open
lowsfer
wants to merge 12 commits into
NVIDIA:main
NVIDIA/TensorRT-LLM:main
from
lowsfer:sparse-offload-kvcm2-core
lowsfer/TensorRT-LLM:sparse-offload-kvcm2-core
Copy head branch name to clipboard
Conversation
Commits
12
(12)
Checks
Files changed
Open
[None][feat] KVCM2 core: sparse attention arena, host-backed pools and Batch
#21129
lowsfer
wants to merge 12 commits into
NVIDIA:main
NVIDIA/TensorRT-LLM:main
from
lowsfer:sparse-offload-kvcm2-core
lowsfer/TensorRT-LLM:sparse-offload-kvcm2-core
Copy head branch name to clipboard
Commits
Commits on Oct 28, 2026
[None][feat] KVCM2 core: sparse attention arena, host-backed pools and Batch
Show description for 4b767a8
lowsfer
committed
4b767a8
View commit details
Copy full SHA for 4b767a8
Browse repository at this point
[None][chore] KVCM2: keep ScratchDesc bound while consumers still read it
Show description for e360c15
lowsfer
committed
e360c15
View commit details
Copy full SHA for e360c15
Browse repository at this point
[None][chore] KVCM2: admit decode through the binding while callers still ask
Show description for 10eb17f
lowsfer
committed
10eb17f
View commit details
Copy full SHA for 10eb17f
Browse repository at this point
[None][chore] KVCM2: leave sparse metadata publication to the fetch path
Show description for c9d7652
lowsfer
committed
c9d7652
View commit details
Copy full SHA for c9d7652
Browse repository at this point
[None][chore] KVCM2: accept is_decoding on resume while callers still pass it
Show description for 5f990c8
lowsfer
committed
5f990c8
View commit details
Copy full SHA for 5f990c8
Browse repository at this point
[None][fix] KVCM2: link tllm_kv_cache against common_environment
Show description for a6f7028
lowsfer
and
claude
committed
a6f7028
View commit details
Copy full SHA for a6f7028
Browse repository at this point
[None][fix] KVCM2: declare the resume stub's is_decoding slot
Show description for ed6ab6f
lowsfer
committed
ed6ab6f
View commit details
Copy full SHA for ed6ab6f
Browse repository at this point
[None][fix] KVCM2: re-export ScratchDesc and use HostMem::sizeGranularity in the NVFP4 cold-page test
Show description for eea8b3d
lowsfer
and
claude
committed
eea8b3d
View commit details
Copy full SHA for eea8b3d
Browse repository at this point
[None][fix] KVCM2: let PageIndexConverter take the ScratchDesc that get_scratch_desc returns
Show description for 37de910
lowsfer
and
claude
committed
37de910
View commit details
Copy full SHA for 37de910
Browse repository at this point
[None][test] KVCM2: shrink the GPU tier to its floors before the DSv4 cold-page pressure request
Show description for 0b81bd8
lowsfer
and
claude
committed
0b81bd8
View commit details
Copy full SHA for 0b81bd8
Browse repository at this point
[None][test] KVCM2: skip the large sparse-resolve test on GPUs with under 160 KiB of shared memory per block
Show description for bfc7cb1
lowsfer
and
claude
committed
bfc7cb1
View commit details
Copy full SHA for bfc7cb1
Browse repository at this point
[None][test] KVCM2: give the lender's sparse refusal test whole sparse-attention life cycles with a dense indexer
Show description for 8df9cac
lowsfer
and
claude
committed
8df9cac
View commit details
Copy full SHA for 8df9cac
Browse repository at this point
You can’t perform that action at this time.