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
Accelerator
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 }}
Uh oh!
There was an error while loading.
Please reload this page
.
ServiceNow
/
Fast-LLM
Public
Notifications
You must be signed in to change notification settings
Fork
50
Star
333
Code
Issues
50
Pull requests
6
Discussions
Actions
Projects
Security and quality
0
Insights
Additional navigation options
Code
Issues
Pull requests
Discussions
Actions
Projects
Security and quality
Insights
DeepSpeed parity hacks
- #502
#502
Closed
bigximik
wants to merge 25 commits into
main
ServiceNow/Fast-LLM:main
from
gspo
ServiceNow/Fast-LLM:gspo
Copy head branch name to clipboard
Conversation
Commits
25
(25)
Checks
Files changed
Closed
DeepSpeed parity hacks
#502
bigximik
wants to merge 25 commits into
main
ServiceNow/Fast-LLM:main
from
gspo
ServiceNow/Fast-LLM:gspo
Copy head branch name to clipboard
Commits
Commits on Apr 27, 2026
grpo: add policy-gradient metrics behind compute_extra_metrics flag
Show description for 98836e6
bigximik
committed
98836e6
View commit details
Copy full SHA for 98836e6
Browse repository at this point
grpo: align metric names with DeepSpeed path
Show description for b856e39
bigximik
committed
b856e39
View commit details
Copy full SHA for b856e39
Browse repository at this point
Commits on Apr 28, 2026
gspo: add sequence-level IS-ratio clipping loss
Show description for b07b999
bigximik
committed
b07b999
View commit details
Copy full SHA for b07b999
Browse repository at this point
schedule: add rollouts_per_step to auto-set depth_first_micro_batches
Show description for fecc978
bigximik
committed
fecc978
View commit details
Copy full SHA for fecc978
Browse repository at this point
grpo: dynamic docs_per_step accumulation and normalize_by_documents
Show description for 7d8ec0c
bigximik
committed
7d8ec0c
View commit details
Copy full SHA for 7d8ec0c
Browse repository at this point
Commits on Apr 29, 2026
grpo: temperature scaling for IS ratio parity with actor sampling
Show description for 014ba59
bigximik
committed
014ba59
View commit details
Copy full SHA for 014ba59
Browse repository at this point
Commits on May 4, 2026
head: fp32_lm_head flag to match vLLM bf16_last_layer_fp32 precision
Show description for d8cb9ef
bigximik
committed
d8cb9ef
View commit details
Copy full SHA for d8cb9ef
Browse repository at this point
head: fix fp32_lm_head gradient flow via detach + manual weight grad accumulation
Show description for 0f90f20
bigximik
committed
0f90f20
View commit details
Copy full SHA for 0f90f20
Browse repository at this point
Commits on May 5, 2026
grpo: decouple loss/gradient divisors and fix SDP loss double-counting
Show description for 557a3c4
bigximik
committed
557a3c4
View commit details
Copy full SHA for 557a3c4
Browse repository at this point
Merge branch 'main' into grpo-metrics
jlamypoirier
authored
a1b3f32
View commit details
Copy full SHA for a1b3f32
Browse repository at this point
Merge branch 'grpo-metrics' into gspo
jlamypoirier
authored
9c07626
View commit details
Copy full SHA for 9c07626
Browse repository at this point
grpo: address review feedback on metrics
Show description for d360a46
jlamypoirier
and
claude
committed
d360a46
View commit details
Copy full SHA for d360a46
Browse repository at this point
grpo: address review follow-ups
Show description for bb6315c
jlamypoirier
and
claude
committed
bb6315c
View commit details
Copy full SHA for bb6315c
Browse repository at this point
grpo: round-3 review fixes
Show description for 89ed062
jlamypoirier
and
claude
committed
89ed062
View commit details
Copy full SHA for 89ed062
Browse repository at this point
grpo: GRPOMetrics as NamedTuple
Show description for b0852fd
jlamypoirier
and
claude
committed
b0852fd
View commit details
Copy full SHA for b0852fd
Browse repository at this point
grpo: fix entropy under tensor-parallel + minor review fixes
Show description for 61ad4f7
jlamypoirier
and
claude
committed
61ad4f7
View commit details
Copy full SHA for 61ad4f7
Browse repository at this point
Merge remote-tracking branch 'origin/grpo-metrics' into gspo
Show description for 15ae8d6
jlamypoirier
committed
15ae8d6
View commit details
Copy full SHA for 15ae8d6
Browse repository at this point
Commits on May 6, 2026
Merge remote-tracking branch 'origin/main' into gspo
jlamypoirier
committed
dfd2ce3
View commit details
Copy full SHA for dfd2ce3
Browse repository at this point
Merge remote-tracking branch 'origin/main' into gspo
Show description for 2fc2dfe
jlamypoirier
committed
2fc2dfe
View commit details
Copy full SHA for 2fc2dfe
Browse repository at this point
gspo: address coarse-review easy items (#7, #13, #14)
Show description for 8547a56
jlamypoirier
and
claude
committed
8547a56
View commit details
Copy full SHA for 8547a56
Browse repository at this point
Commits on May 7, 2026
gspo: register as a sibling loss type instead of policy_loss switch
Show description for fc96d07
jlamypoirier
and
claude
committed
fc96d07
View commit details
Copy full SHA for fc96d07
Browse repository at this point
gspo: collapse loss subclasses, dispatch kernel via self._call_kernel
Show description for d2c051a
jlamypoirier
and
claude
committed
d2c051a
View commit details
Copy full SHA for d2c051a
Browse repository at this point
gspo: dispatch via self._forward = <kernel> instead of wrapper method
Show description for 0d0185c
jlamypoirier
and
claude
committed
0d0185c
View commit details
Copy full SHA for 0d0185c
Browse repository at this point
Commits on May 15, 2026
Merge remote-tracking branch 'origin/main' into gspo
Show description for ffeebd9
jlamypoirier
committed
ffeebd9
View commit details
Copy full SHA for ffeebd9
Browse repository at this point
gspo: drop the temperature field, use base logits_scale_factor
Show description for 227f8bc
jlamypoirier
and
claude
committed
227f8bc
View commit details
Copy full SHA for 227f8bc
Browse repository at this point
You can’t perform that action at this time.
