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
Developer articles
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
743
(743)
Pull requests
1.3k
(1.3k)
Discussions
Actions
Projects
Security and quality
Insights
More
items
[None][perf] Qwen3.5/3.8 wave-2: MoE, attention-DP, GDN replay, weight loading
- #17700
#17700
Merged
Wanli-Jiang
merged 7 commits into
NVIDIA:
main
NVIDIA/TensorRT-LLM:main
from
Wanli-
Jiang:
user/
williamj/
qwen38-
wave2
Wanli-Jiang/TensorRT-LLM:user/williamj/qwen38-wave2
Copy head branch name to clipboard
Aug 21, 2026
Conversation
Commits
7
(7)
Checks
Files changed
Merged
[None][perf] Qwen3.5/3.8 wave-2: MoE, attention-DP, GDN replay, weight loading
#17700
Wanli-Jiang
merged 7 commits into
NVIDIA:
main
NVIDIA/TensorRT-LLM:main
from
Wanli-
Jiang:
user/
williamj/
qwen38-
wave2
Wanli-Jiang/TensorRT-LLM:user/williamj/qwen38-wave2
Copy head branch name to clipboard
Commits
Commits on Aug 20, 2026
[TRTLLM-15304][fix] respect an explicitly configured DeepGEMM MoE token capacity
Show description for 9a4ca37
Wanli-Jiang
committed
9a4ca37
View commit details
Copy full SHA for 9a4ca37
Browse repository at this point
[TRTLLM-15304][perf] avoid the unused DeepGEMM permutation allocation
Show description for ba5310b
Wanli-Jiang
committed
ba5310b
View commit details
Copy full SHA for ba5310b
Browse repository at this point
[TRTLLM-15304][perf] adapt attention-DP context balance at low occupancy
Show description for 127a556
Wanli-Jiang
committed
127a556
View commit details
Copy full SHA for 127a556
Browse repository at this point
[TRTLLM-15304][perf] build GDN replay work items in one Triton launch
Show description for 24ee344
Wanli-Jiang
committed
24ee344
View commit details
Copy full SHA for 24ee344
Browse repository at this point
[TRTLLM-15304][perf] index ConsumableWeightsDict and transfer ownership on rename
Show description for 351b3f1
Wanli-Jiang
committed
351b3f1
View commit details
Copy full SHA for 351b3f1
Browse repository at this point
[TRTLLM-15304][perf] bound concurrent module weight loading
Show description for cd28e08
Wanli-Jiang
committed
cd28e08
View commit details
Copy full SHA for cd28e08
Browse repository at this point
[TRTLLM-15304][perf] release host weight pages on constrained discrete nodes
Show description for 8100e49
Wanli-Jiang
committed
8100e49
View commit details
Copy full SHA for 8100e49
Browse repository at this point
You can’t perform that action at this time.