release: bump version to 0.2.0 | ci: build release archives | fix: cuBLAS shim - #58

Merged

pskrunner14 merged 15 commits into

mainfrom

Oct 2, 2026

Commits on Oct 1, 2026

ci: build release archives in github CI
pskrunner14
committed
a101406
build: bump SentencePiece pin to v0.2.1 for CMake 4 compatibility
pskrunner14
committed
1f97c75
ci: fix macOS and Windows release builds
pskrunner14
committed
dc7a0d2
ci: address CodeRabbit review on release workflow | fix: gpu smoke jobs
pskrunner14
committed
2cfd184
ci: nightly version stamp, release cache hardening, greptile config
pskrunner14
committed
a05e2f9
release: bump version to 0.2.0
pskrunner14
committed
1b69b62
ci: address greptile review comments
pskrunner14
committed
7e30fc9
ci: switch coderabbit profile to chill, disable summary in desc
pskrunner14
committed
789b358

Commits on Oct 2, 2026

ci: fix release scan and windows smoke test
pskrunner14
committed
b8c29b3
fix(cuda): export the cuBLAS calls the llama.cpp pin needs from the shim
pskrunner14
committed
acfada7
fix(cuda): handle large and empty batches in the cuBLAS shim
pskrunner14
committed
0b56aae
ci: install windows build tools on nv-gha runner
pskrunner14
committed
d128853
fix(cuda): validate cuBLAS shim GEMM sizes | docs: note AI review tools
pskrunner14
committed
709ac33
fix(windows): pass -CMakeArgs through to CMake
pskrunner14
committed
999562f
ci(release): drop the unreliable x86 scan
pskrunner14
committed
47aabd1