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
Events
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 }}
Models Backend Check
models-check.yml
Workflow options
More workflow actions
Filter workflow runs
Search
Workflow runs for Models Backend Check
19 workflow runs
Auto-refresh: 30s
Event
Status
Branch
Actor
More items
cuda: opt-in split-K for the PQ2_0 one-column ffn_down mat-vec on Hopper (GGML_CUDA_PQ2_SPLITK)
queued:
Models Backend Check
#19:
Pull request
#424
opened
by
bri-prism
cuda/pq2-splitk-ffn-down
Oct 31, 2026
Queued
More actions for workflow run 19
cuda: opt-in MoE epilogue fusions (GGML_CUDA_MOE_EP_FUSION)
queued:
Models Backend Check
#18:
Pull request
#423
opened
by
bri-prism
cuda/moe-epilogue-fusion
Oct 31, 2026
Queued
More actions for workflow run 18
build: fix the SYCL and Windows CUDA 13.3 release builds after the upstream merge
queued:
Models Backend Check
#17:
Pull request
#422
opened
by
bri-prism
fix/sycl-glu-reorder-dup
Oct 31, 2026
Queued
More actions for workflow run 17
build: fix RPC, Vulkan and SYCL builds after the upstream merge
cancelled:
Models Backend Check
#16:
Pull request
#421
opened
by
bri-prism
fix/merge-release-build
Oct 30, 2026
1d 0h 0m 2s
More actions for workflow run 16
cuda: bit-serial Q1_0 dot and four rows per CTA for one-column mat-vec
cancelled:
Models Backend Check
#15:
Pull request
#420
opened
by
bri-prism
cuda/q1_0-bit-serial-mmvq
Oct 30, 2026
1d 0h 0m 2s
More actions for workflow run 15
cuda: keep the extended tall mat-vec types off fused gate+up launches
cancelled:
Models Backend Check
#14:
Pull request
#419
opened
by
bri-prism
cuda/mmvq-tall-skip-fused-gate
Oct 30, 2026
1d 0h 0m 1s
More actions for workflow run 14
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#13:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 29, 2026
1d 0h 0m 8s
More actions for workflow run 13
cuda: 5120-wide fused norm + Hadamard with PDL, and two-row PQ2_0 decode tiles on Hopper
cancelled:
Models Backend Check
#12:
Pull request
#417
synchronize
by
bri-prism
cuda/pq2-decode-tune
Oct 29, 2026
1d 0h 1m 56s
More actions for workflow run 12
cuda: use the float4 RMS norm path by default, with a pinned rounding order
cancelled:
Models Backend Check
#11:
Pull request
#418
synchronize
by
bri-prism
cuda/rms-norm-vec-default
Oct 29, 2026
1d 0h 0m 10s
More actions for workflow run 11
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#10:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 29, 2026
4h 20m 8s
More actions for workflow run 10
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#9:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 29, 2026
9h 16m 3s
More actions for workflow run 9
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#8:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 29, 2026
1h 56m 2s
More actions for workflow run 8
cuda: use the float4 RMS norm path by default, with a pinned rounding order
cancelled:
Models Backend Check
#7:
Pull request
#418
opened
by
bri-prism
cuda/rms-norm-vec-default
Oct 29, 2026
13h 29m 4s
More actions for workflow run 7
cuda: 5120-wide fused norm + Hadamard with PDL, and two-row PQ2_0 decode tiles on Hopper
cancelled:
Models Backend Check
#6:
Pull request
#417
opened
by
bri-prism
cuda/pq2-decode-tune
Oct 29, 2026
14h 17m 56s
More actions for workflow run 6
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#5:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 29, 2026
1h 35m 50s
More actions for workflow run 5
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#4:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 12, 2026
1d 0h 0m 8s
More actions for workflow run 4
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#3:
Pull request
#352
synchronize
by
bri-prism
merge/upstream-20261011
Oct 12, 2026
15m 59s
More actions for workflow run 3
cuda: keep the quantized K/V type in the short-query MMA FA fallback
cancelled:
Models Backend Check
#2:
Pull request
#353
opened
by
bri-prism
fix/cuda-fa-quant-kv-oob
Oct 12, 2026
1d 0h 0m 2s
More actions for workflow run 2
Merge upstream master (d854d98ba) into prism
cancelled:
Models Backend Check
#1:
Pull request
#352
opened
by
bri-prism
merge/upstream-20261011
Oct 11, 2026
57m 42s
More actions for workflow run 1
You can’t perform that action at this time.