CI: validate complete RayD distribution build matrix - #1

Merged

Asixa merged 35 commits into

mainfrom

codex/ci-rayd-full-build-20260723

Jul 24, 2026

Commits on Jul 23, 2026

Fix moved visibility source metadata check
Asixa
committed
a51b13f
Validate RayD Torch stable ABI across supported runtimes
Asixa
committed
dbf900c
Trigger complete GitHub Actions validation
Asixa
committed
ea47ba1
Honor CI CUDA architecture environment in Torch build
Asixa
committed
e6d21de
Guard CI CUDA architecture propagation
Asixa
committed
8a1c641
Restore full Torch CMake file with CI architecture fix
Asixa
committed
f4c5976
Normalize .github/workflows/pypi.yml newline
Asixa
committed
933e57a
Normalize backends/torch/tests/test_project_metadata.py newline
Asixa
committed
eb4f390
Trigger corrected complete CI validation
Asixa
committed
e3965b5
Trigger CI after event queue stabilization
Asixa
committed
50c4362
Verify large Linux CUDA binaries through extracted fatbins
Asixa
committed
dcbf467
Test large Linux CUDA binary verification fallback
Asixa
committed
e85f1be
Parallelize Linux Torch wheels by Python version
Asixa
committed
f9957ce
Guard parallel Linux Torch wheel matrix
Asixa
committed
e8aff2b
ci: verify CUDA architectures in Torch device libraries
Asixa
committed
1fa197b
test: enforce Torch CUDA binary ownership in CI
Asixa
committed
ad8ec02
fix: avoid requiring the CUDA driver to load stable ops
Asixa
committed
686e37d
test: keep stable ops loadable without a GPU driver
Asixa
committed
21931b7
ci: remove obsolete self-hosted GPU workflow
Asixa
committed
74d4223
ci: avoid duplicate branch and pull request builds
Asixa
committed
2217b8b
ci: avoid duplicate branch and pull request builds
Asixa
committed
1b51eca
ci: accelerate hosted CUDA wheel builds
Asixa
committed
926f848
ci: use valid workspace cache paths
Asixa
committed
69fee77
ci: reduce CUDA setup and fix Windows parallel builds
Asixa
committed
20bd9eb
ci: persist container caches and trim CUDA installs
Asixa
committed
a810834
ci: complete minimal CUDA dependencies and cache keys
Asixa
committed
6876e05
ci: finalize Torch CUDA deps and cache ownership
Asixa
committed
979bb2f
ci: activate Windows compiler caching
Asixa
committed
31294fe
ci: stabilize Windows CUDA installation
Asixa
committed
97d8a1c

Navigation Menu

CI: validate complete RayD distribution build matrix - #1

CI: validate complete RayD distribution build matrix#1
Asixa merged 35 commits into
mainfrom
codex/ci-rayd-full-build-20260723

Commits on Jul 23, 2026

Fix moved visibility source metadata check

Validate RayD Torch stable ABI across supported runtimes

Trigger complete GitHub Actions validation

Honor CI CUDA architecture environment in Torch build

Guard CI CUDA architecture propagation

Restore full Torch CMake file with CI architecture fix

Normalize .github/workflows/pypi.yml newline

Normalize backends/torch/tests/test_project_metadata.py newline

Trigger corrected complete CI validation

Trigger CI after event queue stabilization

Verify large Linux CUDA binaries through extracted fatbins

Test large Linux CUDA binary verification fallback

Parallelize Linux Torch wheels by Python version

Guard parallel Linux Torch wheel matrix

ci: verify CUDA architectures in Torch device libraries

test: enforce Torch CUDA binary ownership in CI

fix: avoid requiring the CUDA driver to load stable ops

test: keep stable ops loadable without a GPU driver

ci: remove obsolete self-hosted GPU workflow

ci: avoid duplicate branch and pull request builds

ci: avoid duplicate branch and pull request builds

ci: accelerate hosted CUDA wheel builds

ci: use valid workspace cache paths

ci: reduce CUDA setup and fix Windows parallel builds

ci: persist container caches and trim CUDA installs

ci: complete minimal CUDA dependencies and cache keys

ci: finalize Torch CUDA deps and cache ownership

ci: activate Windows compiler caching

ci: stabilize Windows CUDA installation

Commits on Jul 24, 2026

ci: group Torch CUDA code generation

fix: isolate grouped Torch gencode flags

fix: avoid grouped NVCC thread race

ci: isolate manual distribution runs by commit

fix: bypass sccache for grouped Windows CUDA

ci: add focused Windows Torch validation