Skip to content
Navigation Menu
Sign in
Appearance settings
Platform
AI CODE CREATION
GitHub Copilot
Write better code with AI
GitHub Copilot app
Direct agents from issue to merge
MCP Registry
Integrate external tools
DEVELOPER WORKFLOWS
Actions
Automate any workflow
Codespaces
Instant dev environments
Issues
Plan and track work
Code Review
Manage code changes
Code Quality
Enforce quality at merge
APPLICATION SECURITY
GitHub Advanced Security
Find and fix vulnerabilities
Code security
Secure your code as you build
Secret protection
Stop leaks before they start
EXPLORE
Why GitHub
Documentation
Blog
Changelog
Marketplace
View all features
Solutions
BY COMPANY SIZE
Enterprises
Small and medium teams
Startups
Nonprofits
BY USE CASE
App Modernization
DevSecOps
DevOps
CI/CD
View all use cases
BY INDUSTRY
Healthcare
Financial services
Manufacturing
Government
View all industries
View all solutions
Resources
EXPLORE BY TOPIC
AI
Software Development
DevOps
Security
View all topics
EXPLORE BY TYPE
Customer stories
Events & webinars
Ebooks & reports
Business insights
GitHub Skills
SUPPORT & SERVICES
Documentation
Customer support
Community forum
Trust center
Partners
View all resources
Open Source
COMMUNITY
GitHub Sponsors
Fund open source developers
PROGRAMS
Security Lab
Maintainer Community
GitHub Stars
Archive Program
REPOSITORIES
Topics
Trending
Collections
Enterprise
ENTERPRISE SOLUTIONS
Enterprise platform
AI-powered developer platform
AVAILABLE ADD-ONS
GitHub Advanced Security
Enterprise-grade security features
Copilot for Business
Enterprise-grade AI features
Premium Support
Enterprise-grade 24/7 support
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
Public
Notifications
You must be signed in to change notification settings
Fork
2.8k
Star
14.6k
Code
Issues
595
Pull requests
914
Discussions
Actions
Projects
Security and quality
0
Insights
Additional navigation options
Code
Issues
Pull requests
Discussions
Actions
Projects
Security and quality
Insights
[TRTLLM-15917][feat] Integrate Sol-Attn sparse attention into VisualGen
- #18329
#18329
Open
karljang
wants to merge 22 commits into
NVIDIA:main
NVIDIA/TensorRT-LLM:main
from
karljang:feat/sol-attn-visualgen-reduced
karljang/TensorRT-LLM:feat/sol-attn-visualgen-reduced
Copy head branch name to clipboard
Conversation
Commits
22
(22)
Checks
Files changed
Open
[TRTLLM-15917][feat] Integrate Sol-Attn sparse attention into VisualGen
#18329
karljang
wants to merge 22 commits into
NVIDIA:main
NVIDIA/TensorRT-LLM:main
from
karljang:feat/sol-attn-visualgen-reduced
karljang/TensorRT-LLM:feat/sol-attn-visualgen-reduced
Copy head branch name to clipboard
Commits
Commits on Sep 18, 2026
[TRTLLM-15917][feat] Integrate Sol-Attn sparse attention into VisualGen
Show description for 73a0b94
karljang
and
claude
committed
73a0b94
View commit details
Copy full SHA for 73a0b94
Browse repository at this point
[TRTLLM-15917][fix] Keep the Sol-Attn CuTe DSL launch opaque to Dynamo
Show description for c003e55
karljang
and
claude
committed
c003e55
View commit details
Copy full SHA for c003e55
Browse repository at this point
[TRTLLM-15917][doc] Declare sol_attn_backend.py's upstream derivation
Show description for 47c7cbf
karljang
and
claude
committed
47c7cbf
View commit details
Copy full SHA for 47c7cbf
Browse repository at this point
[TRTLLM-15917][feat] Drop sm120 from the Sol-Attn port
Show description for 5e077c1
karljang
and
claude
committed
5e077c1
View commit details
Copy full SHA for 5e077c1
Browse repository at this point
[TRTLLM-15917][fix] Keep Sol-Attn's non-sparse attention on the configured backend
Show description for 8593ad8
karljang
and
claude
committed
8593ad8
View commit details
Copy full SHA for 8593ad8
Browse repository at this point
[TRTLLM-15917][fix] Restrict the in-family fallback to true cross-attention
Show description for bd47daa
karljang
and
claude
committed
bd47daa
View commit details
Copy full SHA for bd47daa
Browse repository at this point
[TRTLLM-15917][chore] Rename SolAttnAttention to SolAttention
Show description for d3eb260
karljang
and
claude
committed
d3eb260
View commit details
Copy full SHA for d3eb260
Browse repository at this point
[TRTLLM-15917][refactor] Decide Sol-Attn eligibility per call, not per module
Show description for 1f55662
karljang
and
claude
committed
1f55662
View commit details
Copy full SHA for 1f55662
Browse repository at this point
[TRTLLM-15917][fix] Address review findings on the Sol-Attn integration
Show description for a37e7ff
karljang
and
claude
committed
a37e7ff
View commit details
Copy full SHA for a37e7ff
Browse repository at this point
[TRTLLM-15917][feat] Support SM103 (B300/GB300) for Sol-Attn
Show description for c784739
karljang
and
claude
committed
c784739
View commit details
Copy full SHA for c784739
Browse repository at this point
[TRTLLM-15917][fix] Address review feedback on Sol-Attn
Show description for da5a5a4
karljang
and
claude
committed
da5a5a4
View commit details
Copy full SHA for da5a5a4
Browse repository at this point
[TRTLLM-15917][fix] Reuse SkipSoftmaxScheduler for the Sol-Attn graph phase
Show description for a2c978f
karljang
and
claude
committed
a2c978f
View commit details
Copy full SHA for a2c978f
Browse repository at this point
[TRTLLM-15917][test] Make ineligibility-reason cases reach the branch they name
Show description for 32e9ab9
karljang
and
claude
committed
32e9ab9
View commit details
Copy full SHA for 32e9ab9
Browse repository at this point
[TRTLLM-15917][fix] Honor masks, survive CUDA-graph capture, wire LTX2 timesteps
Show description for 3de388f
karljang
and
claude
committed
3de388f
View commit details
Copy full SHA for 3de388f
Browse repository at this point
[TRTLLM-15917][chore] Drop an unused variable left in the multi-block test
Show description for 2a2365b
karljang
and
claude
committed
2a2365b
View commit details
Copy full SHA for 2a2365b
Browse repository at this point
[TRTLLM-15917][test] Teach the LTX-2 two-stage test stubs the CUDA-graph key contract
Show description for ae7c894
karljang
and
claude
committed
ae7c894
View commit details
Copy full SHA for ae7c894
Browse repository at this point
[TRTLLM-15917][fix] Keep the CUDA-graph phase scope in LTX-2's capture override
Show description for 09bb2c7
karljang
and
claude
committed
09bb2c7
View commit details
Copy full SHA for 09bb2c7
Browse repository at this point
[TRTLLM-15917][chore] Drop the no-op kv_splits option from SolAttentionConfig
Show description for cd8ded2
karljang
and
claude
committed
cd8ded2
View commit details
Copy full SHA for cd8ded2
Browse repository at this point
[TRTLLM-15917][chore] Vendor the Sol-Attn kernel through vendor_sources
Show description for 205c8d5
karljang
and
claude
committed
205c8d5
View commit details
Copy full SHA for 205c8d5
Browse repository at this point
[TRTLLM-15917][fix] Address review: typed dense_layers, TRTLLM_ strict flag, no kernel-error recovery
Show description for 20b1424
karljang
and
claude
committed
20b1424
View commit details
Copy full SHA for 20b1424
Browse repository at this point
[TRTLLM-15917][chore] Clarify dense-backend initialization rationale
Show description for bf3652a
karljang
committed
bf3652a
View commit details
Copy full SHA for bf3652a
Browse repository at this point
[TRTLLM-15917][chore] Align the SEPARATE_QKV fallback with main's cross-attention flag
Show description for b901f29
karljang
and
claude
committed
b901f29
View commit details
Copy full SHA for b901f29
Browse repository at this point
You can’t perform that action at this time.
