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
701
(701)
Pull requests
1.2k
(1.2k)
Discussions
Actions
Projects
Security and quality
Insights
More
items
[TRTLLM-17238][feat] Add PrimTS sparse core backend for Qwen3.8 QSA attention
- #20714
#20714
Open
farazkh80
wants to merge 14 commits into
NVIDIA:main
NVIDIA/TensorRT-LLM:main
from
farazkh80:faraz/qsa-prims-ts-core
farazkh80/TensorRT-LLM:faraz/qsa-prims-ts-core
Copy head branch name to clipboard
Conversation
Commits
14
(14)
Checks
Files changed
Open
[TRTLLM-17238][feat] Add PrimTS sparse core backend for Qwen3.8 QSA attention
#20714
farazkh80
wants to merge 14 commits into
NVIDIA:main
NVIDIA/TensorRT-LLM:main
from
farazkh80:faraz/qsa-prims-ts-core
farazkh80/TensorRT-LLM:faraz/qsa-prims-ts-core
Copy head branch name to clipboard
Commits
Commits on Oct 20, 2026
[None][feat] Add PrimTS sparse core backend for Qwen3.8 QSA attention
Show description for b7c5e70
farazkh80
committed
b7c5e70
View commit details
Copy full SHA for b7c5e70
Browse repository at this point
Commits on Oct 21, 2026
[None][feat] Add PrimTS QSA indexer scorer behind the sparse core switch
Show description for f42f375
farazkh80
committed
f42f375
View commit details
Copy full SHA for f42f375
Browse repository at this point
[None][fix] Align sliced QSA indexer inputs before the PrimTS scorer
Show description for 921da4c
farazkh80
committed
921da4c
View commit details
Copy full SHA for 921da4c
Browse repository at this point
[None][fix] Remove per-call host costs in the QSA PrimTS path
Show description for 7b0ed86
farazkh80
committed
7b0ed86
View commit details
Copy full SHA for 7b0ed86
Browse repository at this point
[None][perf] Remove host stalls from the QSA PrimTS prefill path
Show description for 3ceef11
farazkh80
committed
3ceef11
View commit details
Copy full SHA for 3ceef11
Browse repository at this point
Commits on Oct 22, 2026
[None][perf] Overlap QSA PrimTS prefill chunks with the next chunk's scoring
Show description for 4005b4b
farazkh80
committed
4005b4b
View commit details
Copy full SHA for 4005b4b
Browse repository at this point
[None][feat] Select the QSA PrimTS backend from the LLM API, default auto
Show description for cb42e22
farazkh80
committed
cb42e22
View commit details
Copy full SHA for cb42e22
Browse repository at this point
[None][refactor] Select QSA kernels at runtime without a user-facing switch
Show description for 94f7e07
farazkh80
committed
94f7e07
View commit details
Copy full SHA for 94f7e07
Browse repository at this point
Commits on Oct 23, 2026
[None][fix] Address review: vendored PrimTS import, static FP8 Q scale, decode buckets
Show description for ee49a8b
farazkh80
committed
ee49a8b
View commit details
Copy full SHA for ee49a8b
Browse repository at this point
[None][chore] Drop the QSA sparse-core microbenchmark from the tree
Show description for 70ae6fc
farazkh80
committed
70ae6fc
View commit details
Copy full SHA for 70ae6fc
Browse repository at this point
[None][doc] Note the FP8 KV cache numerics of the QSA PrimTS kernel in the Qwen3.8 guide
Show description for 371816f
farazkh80
committed
371816f
View commit details
Copy full SHA for 371816f
Browse repository at this point
[None][doc] Qualify when QSA selects PrimTS with an FP8 KV cache
Show description for e78d2dc
farazkh80
committed
e78d2dc
View commit details
Copy full SHA for e78d2dc
Browse repository at this point
Commits on Oct 31, 2026
Merge branch 'main' of https://github.com/NVIDIA/TensorRT-LLM into faraz/qsa-prims-ts-core
farazkh80
committed
b96d38f
View commit details
Copy full SHA for b96d38f
Browse repository at this point
[None][chore] Use the group_size plan API of the vendored PrimTS QToken wrapper
Show description for 645b450
farazkh80
committed
645b450
View commit details
Copy full SHA for 645b450
Browse repository at this point
You can’t perform that action at this time.
