fix(ci): Build one arm64 py3 release wheel (#2191)

* fix(ci): Build one arm64 py3 release wheel

* docs: Update changelog for arm64 release wheel fix

Author: abetlen

.github/workflows/build-and-release.yaml

@@ -82,7 +82,9 @@ jobs:
           # Keep native arm64 builds on a portable CPU baseline instead of
           # tuning wheels to the hosted runner.
           CIBW_ENVIRONMENT: CMAKE_ARGS="-DGGML_NATIVE=off"
-          CIBW_BUILD: "cp38-* cp39-* cp310-* cp311-* cp312-*"
+          # The release wheel is tagged py3-none, so one build covers all
+          # supported Python versions and avoids duplicate wheel names.
+          CIBW_BUILD: "cp38-*"
         with:
           output-dir: wheelhouse
 

CHANGELOG.md

@@ -8,6 +8,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [Unreleased]
 
 - feat: Update llama.cpp to ggerganov/llama.cpp@f53577432 and sync Python bindings
+- fix(ci): Build one arm64 release wheel for `py3-none` wheel publishing
 
 ## [0.3.20]