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 }}
InfiniTensor
InfiniLM
Repository navigation
Code
Issues
74
(74)
Pull requests
82
(82)
Actions
Projects
Security and quality
Insights
More
items
feat(qy): MiniCPM5 FA, longctx MoE/RoPE, CUDA-graph capture safety
- #597
#597
Open
Ceng23333
wants to merge 17 commits into
InfiniTensor:
refactor/
adopt-
modern-
infini-
stack
InfiniTensor/InfiniLM:refactor/adopt-modern-infini-stack
from
Ceng23333:
feat/
qy-
backend-
build
Ceng23333/InfiniLM:feat/qy-backend-build
Copy head branch name to clipboard
Conversation
Commits
17
(17)
Checks
Files changed
Open
feat(qy): MiniCPM5 FA, longctx MoE/RoPE, CUDA-graph capture safety
#597
Ceng23333
wants to merge 17 commits into
InfiniTensor:
refactor/
adopt-
modern-
infini-
stack
InfiniTensor/InfiniLM:refactor/adopt-modern-infini-stack
from
Ceng23333:
feat/
qy-
backend-
build
Ceng23333/InfiniLM:feat/qy-backend-build
Copy head branch name to clipboard
Commits
Commits on Oct 10, 2026
feat(qy): allow QY flash-attn and MiniCPM5 MoE stack
Show description for 5d6a98e
Ceng23333
committed
5d6a98e
View commit details
Copy full SHA for 5d6a98e
Browse repository at this point
feat(qy): enable CUDA graph-safe FlashAttention for MiniCPM5
Show description for d3144ab
Ceng23333
committed
d3144ab
View commit details
Copy full SHA for d3144ab
Browse repository at this point
Commits on Oct 13, 2026
feat(qy): merge longctx MoE/RoPE GPU promotions onto cudagraph
Show description for 40376ae
Ceng23333
committed
40376ae
View commit details
Copy full SHA for 40376ae
Browse repository at this point
fix(qy): mark RoPE capture-unsafe for longctx QY pack path
Show description for 71709ef
Ceng23333
committed
71709ef
View commit details
Copy full SHA for 71709ef
Browse repository at this point
fix(qy): avoid host sync under CUDA graph capture for longctx MoE/RoPE
Show description for 1491d67
Ceng23333
committed
1491d67
View commit details
Copy full SHA for 1491d67
Browse repository at this point
fix(qy): correct sync_unless_capturing helper (no recursion)
Ceng23333
committed
5a0639f
View commit details
Copy full SHA for 5a0639f
Browse repository at this point
fix(qy): skip cudaDeviceSynchronize under graph capture in moe_fused_dense
Ceng23333
committed
012b81e
View commit details
Copy full SHA for 012b81e
Browse repository at this point
fix(qy): correct sync_device_unless_capturing helper (no recursion)
Ceng23333
committed
fdc1a40
View commit details
Copy full SHA for fdc1a40
Browse repository at this point
Commits on Oct 14, 2026
style(qy): format FA/longctx/cudagraph changes with clang-format 18 and ruff
Show description for 60757bd
Ceng23333
and
cursoragent
committed
60757bd
View commit details
Copy full SHA for 60757bd
Browse repository at this point
Commits on Oct 19, 2026
feat(qy): full_and_piecewise CUDA graph policy and capture-safe MoE/RoPE
Show description for ecceba6
Ceng23333
committed
ecceba6
View commit details
Copy full SHA for ecceba6
Browse repository at this point
Commits on Oct 21, 2026
feat(qy): native piecewise prefill + F&P phase policy for MiniCPM5/Llama
Show description for b253606
Ceng23333
committed
b253606
View commit details
Copy full SHA for b253606
Browse repository at this point
Commits on Oct 22, 2026
feat(qy): continuous batching + chunked prefill for MiniCPM5 F&P
Show description for b074fb2
Ceng23333
committed
b074fb2
View commit details
Copy full SHA for b074fb2
Browse repository at this point
feat(qy): continuous batching + chunked prefill for MiniCPM5 F&P
Show description for 5d5419e
Ceng23333
committed
5d5419e
View commit details
Copy full SHA for 5d5419e
Browse repository at this point
fix(qy): sanitize batched sample ids under V1 continuous batching
Show description for 9aa4fae
Ceng23333
committed
9aa4fae
View commit details
Copy full SHA for 9aa4fae
Browse repository at this point
fix(qy): use len(tokenizer) for sample-id bounds under CB
Show description for b82ebbf
Ceng23333
committed
b82ebbf
View commit details
Copy full SHA for b82ebbf
Browse repository at this point
fix(qy): map OOR batched samples into-vocab instead of EOS
Show description for 4976b59
Ceng23333
committed
4976b59
View commit details
Copy full SHA for 4976b59
Browse repository at this point
Commits on Oct 23, 2026
fix(qy): disable batch_greedy when n_out>1; hard-fail OOR samples
Show description for 56f9b50
Ceng23333
committed
56f9b50
View commit details
Copy full SHA for 56f9b50
Browse repository at this point
You can’t perform that action at this time.
