From 5acaa80e66d5083a6114d4d72551c1c4b153789a Mon Sep 17 00:00:00 2001
From: czxdev <36617515+czxdev@users.noreply.github.com>
Date: Thu, 2 Jul 2026 18:36:00 +0800
Subject: [PATCH 1/4] Refine CUDA versions and CMake arguments in workflow

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

Disable AVX512
---
 .github/workflows/build-wheels-cuda.yaml | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/build-wheels-cuda.yaml b/.github/workflows/build-wheels-cuda.yaml
index 59dc3558b..88eb0b148 100644
--- a/.github/workflows/build-wheels-cuda.yaml
+++ b/.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'
           }

From 70b2189c8aae30adb78134dce3a55b3cc572379d Mon Sep 17 00:00:00 2001
From: czxdev <36617515+czxdev@users.noreply.github.com>
Date: Mon, 6 Jul 2026 14:09:00 +0800
Subject: [PATCH 2/4] Update build-wheels-cuda.yaml

---
 .github/workflows/build-wheels-cuda.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/build-wheels-cuda.yaml b/.github/workflows/build-wheels-cuda.yaml
index 88eb0b148..08feffe62 100644
--- a/.github/workflows/build-wheels-cuda.yaml
+++ b/.github/workflows/build-wheels-cuda.yaml
@@ -30,7 +30,7 @@ jobs:
               # wheel.py-api = "py3" makes the CUDA wheel interpreter-agnostic,
               # so one builder per toolkit version is sufficient.
               'pyver' = @("3.9")
-              'cuda' = @("12.4.1")
+              'cuda' = @("13.2.1")
               'releasetag' = @("basic")
               'exclude' = @(
                 @{ 'os' = 'windows-2022'; 'cuda' = '12.1.1' },

From 90b1cd9578b768e9447ac220cbb41a738e0828f2 Mon Sep 17 00:00:00 2001
From: czxdev <36617515+czxdev@users.noreply.github.com>
Date: Wed, 8 Jul 2026 13:46:35 +0800
Subject: [PATCH 3/4] Update CUDA versions and OS in workflow configuration

---
 .github/workflows/build-wheels-cuda.yaml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/build-wheels-cuda.yaml b/.github/workflows/build-wheels-cuda.yaml
index 08feffe62..11bc83df0 100644
--- a/.github/workflows/build-wheels-cuda.yaml
+++ b/.github/workflows/build-wheels-cuda.yaml
@@ -26,11 +26,11 @@ jobs:
         id: set-matrix
         run: |
           $matrix = @{
-              'os' = @('ubuntu-22.04')
+              'os' = @('windows-2022')
               # wheel.py-api = "py3" makes the CUDA wheel interpreter-agnostic,
               # so one builder per toolkit version is sufficient.
               'pyver' = @("3.9")
-              'cuda' = @("13.2.1")
+              'cuda' = @("13.2.1", "12.4.1")
               'releasetag' = @("basic")
               'exclude' = @(
                 @{ 'os' = 'windows-2022'; 'cuda' = '12.1.1' },

From ea7d2ba326d8bec5a8c896f63cbc283ea9973977 Mon Sep 17 00:00:00 2001
From: czxdev <36617515+czxdev@users.noreply.github.com>
Date: Wed, 8 Jul 2026 19:35:11 +0800
Subject: [PATCH 4/4] Update workflow to remove release triggers

Removed workflow_run trigger for new releases.
---
 .github/workflows/generate-index-from-release.yaml | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/.github/workflows/generate-index-from-release.yaml b/.github/workflows/generate-index-from-release.yaml
index edf292387..4a4ef566c 100644
--- a/.github/workflows/generate-index-from-release.yaml
+++ b/.github/workflows/generate-index-from-release.yaml
@@ -1,12 +1,6 @@
 name: Wheels Index
 
 on:
-  # Trigger on new release
-  workflow_run:
-    workflows: ["Release", "Build Wheels (CUDA)", "Build Wheels (Metal)", "Build Wheels (Vulkan)", "Build Wheels (ROCm)"]
-    types:
-      - completed
-
   # Allows you to run this workflow manually from the Actions tab
   workflow_dispatch: