fix(ci): release wheel workflow (abetlen#2154)

* fix(ci): harden release wheel workflow

* fix(ci): document and pin release wheel baselines

* fix(ci): speed up release arch builds

* fix(ci): split riscv64 by python version

* fix(ci): sanitize riscv64 artifact names

Author: abetlen

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

@@ -46,6 +46,13 @@ jobs:
         env:
           # disable repair
           CIBW_REPAIR_WHEEL_COMMAND: ""
+          # Skip cibuildwheel's default i686 sidecar and keep Linux release
+          # wheels on a portable x86_64 CPU baseline.
+          CIBW_ARCHS_LINUX: "auto64"
+          CIBW_ENVIRONMENT_LINUX: CMAKE_ARGS="-DGGML_NATIVE=off"
+          # Keep macOS release wheels on a portable CPU baseline instead of
+          # inheriting the hosted runner's native flags.
+          CIBW_ENVIRONMENT_MACOS: CMAKE_ARGS="-DGGML_NATIVE=off"
         with:
           package-dir: .
           output-dir: wheelhouse
@@ -57,24 +64,21 @@ jobs:
 
   build_wheels_arm64:
     name: Build arm64 wheels
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-24.04-arm
     steps:
       - uses: actions/checkout@v4
         with:
           submodules: "recursive"
 
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v3
-        with:
-          platforms: linux/arm64
-
       - name: Build wheels
         uses: pypa/cibuildwheel@v2.22.0
         env:
           CIBW_SKIP: "*musllinux* pp*"
           CIBW_REPAIR_WHEEL_COMMAND: ""
           CIBW_ARCHS: "aarch64"
-          CIBW_ENVIRONMENT: CMAKE_ARGS="-DCMAKE_OSX_ARCHITECTURES=arm64 -DCMAKE_APPLE_SILICON_PROCESSOR=arm64 -DCMAKE_CROSSCOMPILING=ON"
+          # 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-*"
         with:
           output-dir: wheelhouse
@@ -86,8 +90,27 @@ jobs:
           path: ./wheelhouse/*.whl
 
   build_wheels_riscv64:
-    name: Build riscv64 wheels
+    name: Build riscv64 wheels (${{ matrix.shard.name }})
     runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        shard:
+          - name: cp310
+            build: "cp310-*"
+            artifact: wheels_riscv64_cp310
+          - name: cp311
+            build: "cp311-*"
+            artifact: wheels_riscv64_cp311
+          - name: cp312
+            build: "cp312-*"
+            artifact: wheels_riscv64_cp312
+          - name: cp313
+            build: "cp313-*"
+            artifact: wheels_riscv64_cp313
+          - name: cp314
+            build: "cp314-*"
+            artifact: wheels_riscv64_cp314
     steps:
       - uses: actions/checkout@v4
         with:
@@ -104,14 +127,19 @@ jobs:
           CIBW_SKIP: "*musllinux* pp*"
           CIBW_REPAIR_WHEEL_COMMAND: ""
           CIBW_ARCHS: "riscv64"
-          CIBW_BUILD: "cp310-* cp311-* cp312-* cp313-* cp314-*"
+          # Build riscv64 wheels against a conservative baseline instead of
+          # enabling RVV-related extensions from the build container.
+          CIBW_ENVIRONMENT: CMAKE_ARGS="-DGGML_NATIVE=off -DGGML_RVV=off -DGGML_RV_ZFH=off -DGGML_RV_ZVFH=off -DGGML_RV_ZICBOP=off -DGGML_RV_ZIHINTPAUSE=off"
+          # Split the emulated riscv64 build into one Python version per job
+          # to minimize wall-clock time without changing the release artifacts.
+          CIBW_BUILD: ${{ matrix.shard.build }}
         with:
           output-dir: wheelhouse
 
       - name: Upload wheels as artifacts
         uses: actions/upload-artifact@v4
         with:
-          name: wheels_riscv64
+          name: ${{ matrix.shard.artifact }}
           path: ./wheelhouse/*.whl
 
   build_sdist:
@@ -159,6 +187,7 @@ jobs:
   release:
     name: Release
     needs: [build_wheels, build_wheels_arm64, build_wheels_riscv64, build_sdist]
+    if: startsWith(github.ref, 'refs/tags/')
     runs-on: ubuntu-latest
 
     steps: