Refine CUDA versions and CMake arguments in workflow

Updated the CUDA versions and modified CMake arguments for building wheels.

Disable AVX512

Author: czxdev

.github/workflows/build-wheels-cuda.yaml

@@ -26,11 +26,11 @@ jobs:
         id: set-matrix
         run: |
           $matrix = @{
-              'os' = @('ubuntu-22.04', 'windows-2022')
+              'os' = @('ubuntu-22.04')
               # wheel.py-api = "py3" makes the CUDA wheel interpreter-agnostic,
               # so one builder per toolkit version is sufficient.
               'pyver' = @("3.9")
-              'cuda' = @("11.8.0", "12.1.1", "12.2.2", "12.3.2", "12.4.1", "12.5.1", "13.0.2", "13.2.1")
+              'cuda' = @("12.4.1")
               'releasetag' = @("basic")
               'exclude' = @(
                 @{ 'os' = 'windows-2022'; 'cuda' = '12.1.1' },
@@ -231,7 +231,9 @@ jobs:
           # one forward-compatible PTX target instead of embedding PTX for every
           # SM. This keeps the wheel under GitHub's 2 GiB release-asset limit.
           $env:CMAKE_ARGS = "-DGGML_CUDA_FORCE_MMQ=ON -DGGML_CUDA=on -DCMAKE_CUDA_ARCHITECTURES=$cudaArchs -DCMAKE_CUDA_FLAGS=-allow-unsupported-compiler -DCMAKE_CUDA_FLAGS_INIT=-allow-unsupported-compiler $env:CMAKE_ARGS"
-          $env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DGGML_AVX2=off -DGGML_FMA=off -DGGML_F16C=off'
+          $env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DGGML_NATIVE=OFF -DGGML_AVX512=OFF'
+          $env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DGGML_AVX2=ON -DGGML_FMA=ON -DGGML_F16C=ON'
+          
           if ($IsLinux) {
             $env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DGGML_OPENMP=OFF'
           }