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 }}
ifm-ai
xllm
Repository navigation
Code
Issues
17
(17)
Pull requests
32
(32)
Actions
Projects
Security and quality
Insights
More
items
All pull requests
Labels
10
(10)
Milestones
0
(0)
New pull request
Search pull requests
is
:
pr
state
:
open
is:pr state:open
Clear filter
Search
Pull requests
Open
32
(32)
Closed
34
(34)
Author
Label
Projects
Milestones
Reviews
Assignee
Sort by
Newest
descending
Sort…
Sort by Newest descending
More items
Comfortable display density
Compact display density
Write metrics from a separate process (async metric writer)
#91
·
AditSom-IFM
opened
Nov 17, 2026
Collaborator
·
·
Gekko with context parallelism: one chunk per rank no longer gives a NaN loss or a failed backward
#90
·
achamma723
opened
Nov 17, 2026
Collaborator
·
·
1
Read each training batch once per data-parallel replica, and restore the data buffer from the checkpoint
#87
·
achamma723
opened
Nov 17, 2026
Collaborator
·
·
1
Do not halve the shared aux-loss gradient in place in the fused MoVA backward
#85
·
achamma723
opened
Nov 17, 2026
Collaborator
·
·
1
Per-role optimizer settings for Muon (roles); muP / Complete(d)P scaling computed by the launcher
#83
·
onenoc
opened
Nov 17, 2026
Collaborator
·
·
2
Add group-norm kernels for data-center Blackwell (B200, B300)
#79
·
SolitaryThinker
opened
Nov 16, 2026
Collaborator
·
·
Move logging code into xllm/metrics, share model hooks, nest health config
#78
·
AditSom-IFM
opened
Nov 16, 2026
Collaborator
·
·
Fix the division by zero of lr_linear and lr_cosine when warmup == steps
#77
·
achamma723
opened
Nov 16, 2026
Collaborator
·
·
1
feat: fix TSDN state precision and add a lookback scan backend
#75
·
wenshicheng97
opened
Nov 16, 2026
Collaborator
·
·
Multi-token prediction (MTP)
#74
·
achamma723
opened
Nov 15, 2026
Collaborator
·
·
1
Expert parallelism larger than tensor parallelism (EP > TP)
#70
·
achamma723
opened
Nov 14, 2026
Collaborator
·
·
1
Fix Gekko health stats after gekko5 and log the softdelta gate
#63
·
AditSom-IFM
opened
Nov 13, 2026
Collaborator
·
·
Give the routed experts zero gradients when they receive no rows
#62
·
achamma723
opened
Nov 13, 2026
Collaborator
·
·
1
Compute the MoE RMSNorm on each GPU's slice instead of gathering the full activation
#60
·
achamma723
opened
Nov 13, 2026
Collaborator
·
·
1
Count the aux-loss gradient once across TP ranks
#57
·
achamma723
opened
Nov 11, 2026
Collaborator
·
·
1
1
Compute the MoE routing-weight gradient on the expert's owner
#55
·
achamma723
opened
Nov 10, 2026
Collaborator
·
·
Send each MoE token once per destination rank in the all-to-all
#52
·
achamma723
opened
Nov 10, 2026
Collaborator
·
·
1
1
Gekko training optimizations for Blackwell and Hopper: CuTe DSL kernels, fp8 GEMMs, training-step and FSDP2 changes
#43
·
Johnsonms
opened
Nov 5, 2026
·
·
feat(residual): hyper-connections (ihc / mhc / ohc) with Triton stream ops and a fused training block, on Muon
#41
·
chenjunlin110
opened
Nov 4, 2026
Collaborator
·
·
Raw text never trains on control tokens
#38
·
avifenesh
opened
Nov 2, 2026
·
·
Resume after a dataset is dropped from --data
#37
·
avifenesh
opened
Nov 2, 2026
·
·
Optimizer offload: fp32 masters and AdamW state in host memory
#36
·
avifenesh
opened
Oct 31, 2026
·
·
Pass num_splits to flash-attn 2 varlen_fwd when the binding takes it
#34
·
avifenesh
opened
Oct 30, 2026
·
·
Qual moe
#33
·
Linghao-Jin
opened
Oct 30, 2026
Collaborator
·
·
Complete(d)P for Muon: depth scale folded into wo / fc2 (experimental)
#30
·
AditSom-IFM
opened
Oct 28, 2026
Collaborator
·
·
7
Previous
1
2
Next
You can’t perform that action at this time.