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 }}
bitsandbytes-foundation
/
bitsandbytes
Public
Uh oh!
There was an error while loading.
Please reload this page
.
Notifications
You must be signed in to change notification settings
Fork
931
Star
8.5k
Code
Issues
48
Pull requests
43
Discussions
Actions
Projects
Security and quality
0
Insights
Additional navigation options
Code
Issues
Pull requests
Discussions
Actions
Projects
Security and quality
Insights
Feat/mps gemm 4bit bf16
- #2070
#2070
Closed
eaglstun
wants to merge 23 commits into
bitsandbytes-foundation:main
bitsandbytes-foundation/bitsandbytes:main
from
eaglstun:feat/mps-gemm-4bit-bf16
eaglstun/bitsandbytes:feat/mps-gemm-4bit-bf16
Copy head branch name to clipboard
Conversation
Commits
23
(23)
Checks
Files changed
Closed
Feat/mps gemm 4bit bf16
#2070
eaglstun
wants to merge 23 commits into
bitsandbytes-foundation:main
bitsandbytes-foundation/bitsandbytes:main
from
eaglstun:feat/mps-gemm-4bit-bf16
eaglstun/bitsandbytes:feat/mps-gemm-4bit-bf16
Copy head branch name to clipboard
Commits
Commits on Jul 8, 2026
Phase 1: MPS correctness sweep — parity harness + ground-truth audit
Show description for 0273608
eaglstun
and
claude
committed
0273608
View commit details
Copy full SHA for 0273608
Browse repository at this point
Fix Lion to use decoupled weight decay in the default backend
Show description for 5f37b8b
eaglstun
and
claude
committed
5f37b8b
View commit details
Copy full SHA for 5f37b8b
Browse repository at this point
Phase 2: quantize_blockwise through hand-written Metal (native pipe, one op end to end)
Show description for cbfa1ad
eaglstun
and
claude
committed
cbfa1ad
View commit details
Copy full SHA for cbfa1ad
Browse repository at this point
Phase 3: dequantize_blockwise + dequantize_4bit + quantize_4bit on native Metal, and a load-time buffer-contract guard
Show description for 863990c
eaglstun
and
claude
committed
863990c
View commit details
Copy full SHA for 863990c
Browse repository at this point
Merge MPS native Metal backend (Phases 1-3): quantize/dequantize blockwise + 4-bit
Show description for 501585f
eaglstun
and
claude
committed
501585f
View commit details
Copy full SHA for 501585f
Browse repository at this point
Packaging: ship bitsandbytes.metallib in the wheel so native MPS works from pip install
Show description for 705a6e8
eaglstun
and
claude
committed
705a6e8
View commit details
Copy full SHA for 705a6e8
Browse repository at this point
Phase 4: Apple Silicon / MPS user-facing docs + next-phase matmul plan
Show description for 642ecb1
eaglstun
and
claude
committed
642ecb1
View commit details
Copy full SHA for 642ecb1
Browse repository at this point
Commits on Jul 15, 2026
Phase M1: record 4-bit matmul baseline + A/B design decision
Show description for 4598cd2
eaglstun
and
claude
committed
4598cd2
View commit details
Copy full SHA for 4598cd2
Browse repository at this point
Phase M2: fused native gemv_4bit Metal kernel (dequant + dot product, no B_dq)
Show description for d913182
eaglstun
and
claude
committed
d913182
View commit details
Copy full SHA for d913182
Browse repository at this point
Phase M3: native gemm_4bit -- chunked dequant + MPSMatrixMultiplication, one command buffer
Show description for bdd3d9e
eaglstun
and
claude
committed
bdd3d9e
View commit details
Copy full SHA for bdd3d9e
Browse repository at this point
Phase M4: matmul closeout -- sync tax measured + documented, offset-0 semantics verified, honest docs
Show description for 0c9db70
eaglstun
and
claude
committed
0c9db70
View commit details
Copy full SHA for 0c9db70
Browse repository at this point
Merge MPS 4-bit matmul sub-phase (M1-M4): native gemv + gemm on Metal
Show description for ed2ea2e
eaglstun
and
claude
committed
ed2ea2e
View commit details
Copy full SHA for ed2ea2e
Browse repository at this point
Retire the lion weight-decay strict-xfail now that the divergence is fixed
Show description for 5e7ff4e
eaglstun
and
claude
committed
5e7ff4e
View commit details
Copy full SHA for 5e7ff4e
Browse repository at this point
Fix Lion to use decoupled weight decay in the Triton 32-bit kernel
Show description for 45813c3
eaglstun
and
claude
committed
45813c3
View commit details
Copy full SHA for 45813c3
Browse repository at this point
Fix Lion (and Adagrad) weight decay in the Triton 8-bit blockwise kernel
Show description for 7333266
eaglstun
and
claude
committed
7333266
View commit details
Copy full SHA for 7333266
Browse repository at this point
Fix Lion to use decoupled weight decay in the CUDA 32-bit kernel
Show description for 6a81fa9
eaglstun
and
claude
committed
6a81fa9
View commit details
Copy full SHA for 6a81fa9
Browse repository at this point
Merge pull request #1 from eaglstun/mps-lion-triton-32bit
Show description for 7aa17ce
eaglstun
authored
7aa17ce
View commit details
Copy full SHA for 7aa17ce
Browse repository at this point
Merge pull request #2 from eaglstun/mps-lion-triton-8bit
Show description for 0c86556
eaglstun
authored
0c86556
View commit details
Copy full SHA for 0c86556
Browse repository at this point
Merge pull request #3 from eaglstun/mps-lion-cuda
Show description for 9afd664
eaglstun
authored
9afd664
View commit details
Copy full SHA for 9afd664
Browse repository at this point
Merge upstream/main into fork main (Apple Silicon MPS line)
Show description for d29189f
eaglstun
and
claude
committed
d29189f
View commit details
Copy full SHA for d29189f
Browse repository at this point
Commits on Aug 28, 2026
Merge remote-tracking branch 'upstream/main'
Show description for 680ad87
eaglstun
committed
680ad87
View commit details
Copy full SHA for 680ad87
Browse repository at this point
Migrate dispatch tooling to agent configuration
eaglstun
committed
7582cb3
View commit details
Copy full SHA for 7582cb3
Browse repository at this point
Commits on Sep 4, 2026
mps 4bit bf16
eaglstun
committed
0ddb386
View commit details
Copy full SHA for 0ddb386
Browse repository at this point
You can’t perform that action at this time.