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
Type
/
to 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
.
onnxruntime
/
mobius
Public
Notifications
You must be signed in to change notification settings
Fork
2
Star
16
Code
Issues
30
Pull requests
21
Actions
Projects
Security and quality
0
Insights
Additional navigation options
Code
Issues
Pull requests
Actions
Projects
Security and quality
Insights
Fix Gemma4 bidirectional attention + add gemma-4-12B (gemma4_unified)
- #338
#338
Merged
justinchuby
merged 23 commits into
main
onnxruntime/mobius:main
from
justinchu/gemma4-bidirectional-attn
onnxruntime/mobius:justinchu/gemma4-bidirectional-attn
Copy head branch name to clipboard
Jun 9, 2026
Conversation
Commits
23
(23)
Checks
Files changed
Merged
Fix Gemma4 bidirectional attention + add gemma-4-12B (gemma4_unified)
#338
justinchuby
merged 23 commits into
main
onnxruntime/mobius:main
from
justinchu/gemma4-bidirectional-attn
onnxruntime/mobius:justinchu/gemma4-bidirectional-attn
Copy head branch name to clipboard
Commits
Commits on Jun 5, 2026
Fix Gemma4 bidirectional vision-block attention
Show description for b116505
justinchuby
and
Copilot
committed
b116505
View commit details
Copy full SHA for b116505
Browse repository at this point
Add gemma4_unified_text (gemma-4-12B text backbone)
Show description for 5231691
justinchuby
and
Copilot
committed
5231691
View commit details
Copy full SHA for 5231691
Browse repository at this point
Add gemma-4-12B unified multimodal model (gemma4_unified)
Show description for fa9834d
justinchuby
and
Copilot
committed
fa9834d
View commit details
Copy full SHA for fa9834d
Browse repository at this point
Wire gemma4_unified vision projector to mm_embed_dim input
Show description for 40b829f
justinchuby
and
Copilot
committed
40b829f
View commit details
Copy full SHA for 40b829f
Browse repository at this point
Add unit tests for gemma4_unified weight mapping and config hooks
Show description for 7487726
justinchuby
and
Copilot
committed
7487726
View commit details
Copy full SHA for 7487726
Browse repository at this point
test: hoist imports to module top in _common_test
Show description for d527838
justinchuby
and
Copilot
committed
d527838
View commit details
Copy full SHA for d527838
Browse repository at this point
test: complete gemma4_unified (gemma-4-12B) coverage
Show description for 04fcc19
justinchuby
and
Copilot
committed
04fcc19
View commit details
Copy full SHA for 04fcc19
Browse repository at this point
refactor(gemma4): derive vision-block overlay in decoder from input_ids
Show description for bba76eb
justinchuby
and
Copilot
committed
bba76eb
View commit details
Copy full SHA for bba76eb
Browse repository at this point
feat(ort_genai): wire gemma4_unified into model-type resolution
Show description for 4b17bd1
justinchuby
and
Copilot
committed
4b17bd1
View commit details
Copy full SHA for 4b17bd1
Browse repository at this point
fix(gemma4): upcast unified vision patch embedding for float16
Show description for ee47419
justinchuby
and
Copilot
committed
ee47419
View commit details
Copy full SHA for ee47419
Browse repository at this point
docs(examples): gemma4_unified multimodal genai example (text/image/audio)
Show description for aa5dede
justinchuby
and
Copilot
committed
aa5dede
View commit details
Copy full SHA for aa5dede
Browse repository at this point
feat(examples): Olive INT4 quantization for gemma4_unified decoder
Show description for ab8a56c
justinchuby
and
Copilot
committed
ab8a56c
View commit details
Copy full SHA for ab8a56c
Browse repository at this point
fix(gemma4): exclude audio tokens from vision-block bidirectional attention
Show description for 131faa2
justinchuby
and
Copilot
committed
131faa2
View commit details
Copy full SHA for 131faa2
Browse repository at this point
docs+test(gemma4): clarify bidirectional doc, device-robust HF parity
Show description for fb0ad6b
justinchuby
and
Copilot
committed
fb0ad6b
View commit details
Copy full SHA for fb0ad6b
Browse repository at this point
fix(examples): suppress structural image/audio tokens in gemma4 decode
Show description for 8dffc8c
justinchuby
and
Copilot
committed
8dffc8c
View commit details
Copy full SHA for 8dffc8c
Browse repository at this point
fix(gemma4): reject unsupported use_bidirectional_attention modes
Show description for bf3e7d8
justinchuby
and
Copilot
committed
bf3e7d8
View commit details
Copy full SHA for bf3e7d8
Browse repository at this point
fix(gemma4): address PR review (block-overlay gating, unified processors, audio eps)
Show description for 7d6fd16
justinchuby
and
Copilot
committed
7d6fd16
View commit details
Copy full SHA for 7d6fd16
Browse repository at this point
Commits on Jun 6, 2026
test(gemma4): add L4+L5 golden tests for gemma-4-12B text/image/audio
Show description for 61513c8
justinchuby
and
Copilot
committed
61513c8
View commit details
Copy full SHA for 61513c8
Browse repository at this point
Commits on Jun 8, 2026
fix(gemma4): make bf16 unified vision/audio Compress loadable
Show description for 44fdbd0
justinchuby
and
Copilot
committed
44fdbd0
View commit details
Copy full SHA for 44fdbd0
Browse repository at this point
refactor(gqa): gate GQA head_dim limit on EP capability
Show description for 389fa47
justinchuby
and
Copilot
committed
389fa47
View commit details
Copy full SHA for 389fa47
Browse repository at this point
Commits on Jun 9, 2026
feat(gqa): lift CUDA head_dim cap to enable Gemma4 global-attention GQA
Show description for 0547679
justinchuby
and
Copilot
committed
0547679
View commit details
Copy full SHA for 0547679
Browse repository at this point
Remove head_dim cap on GroupQueryAttention fusion
Show description for 89d7ed6
justinchuby
and
Copilot
committed
89d7ed6
View commit details
Copy full SHA for 89d7ed6
Browse repository at this point
Fix ruff-format lint: remove extra blank line in e2e_golden_test
Show description for 1c2dc4d
justinchuby
and
Copilot
committed
1c2dc4d
View commit details
Copy full SHA for 1c2dc4d
Browse repository at this point
You can’t perform that action at this time.
