[Enhancement] Off the DGGML_CUDA_FORCE_MMQ for V100

fanqiNO1 · web-flow · commit 9da43355350b · 2024-09-07T17:06:14.000+08:00
diff --git a/.github/workflows/build-wheels-cuda.yaml b/.github/workflows/build-wheels-cuda.yaml
@@ -20,9 +20,12 @@ jobs:
         id: set-matrix
         run: |
           $matrix = @{
-              'os' = @('ubuntu-latest', 'windows-2019')
-              'pyver' = @("3.9", "3.10", "3.11", "3.12")
-              'cuda' = @("12.1.1", "12.2.2", "12.3.2", "12.4.1")
+              # 'os' = @('ubuntu-latest', 'windows-2019')
+              'os' = @('ubuntu-latest')
+              # 'pyver' = @("3.9", "3.10", "3.11", "3.12")
+              'pyver' = @("3.10")
+              # 'cuda' = @("12.1.1", "12.2.2", "12.3.2", "12.4.1")
+              'cuda' = @("12.2.2")
               'releasetag' = @("basic")
           }
 
@@ -115,7 +118,7 @@ jobs:
           }
           $env:VERBOSE = '1'
           $env:CMAKE_ARGS = '-DGGML_CUDA=on -DCMAKE_CUDA_ARCHITECTURES=all'
-          $env:CMAKE_ARGS = "-DGGML_CUDA_FORCE_MMQ=ON $env:CMAKE_ARGS"
+          $env:CMAKE_ARGS = "-DGGML_CUDA_FORCE_MMQ=OFF $env:CMAKE_ARGS"
           # if ($env:AVXVER -eq 'AVX') {
           $env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DGGML_AVX2=off -DGGML_FMA=off -DGGML_F16C=off'
           # }
