cuda: restrict out_prod support to f32 inputs

.ecrc

@@ -1,5 +1,5 @@
 {
-  "Exclude": ["^\\.gitmodules$", "stb_image\\.h"],
+  "Exclude": ["^\\.gitmodules$", "stb_image\\.h", "^vendor/"],
   "Disable": {
     "IndentSize": true
   }

.flake8

@@ -15,4 +15,5 @@ exclude =
     build,
     # This contains builds that we don't want to check
     dist  # This is generated with `python build .` for package releases
+    scripts/tune
 # max-complexity = 10

.github/CODEOWNERS

@@ -0,0 +1 @@
+* @tetherto/ai-runtime-bk-models

.github/workflows/approval-check-worker.yml

@@ -0,0 +1,19 @@
+name: Approval Check Worker
+
+on:
+  pull_request_review:
+    types: [submitted, dismissed]
+
+jobs:
+  check-approvals:
+    permissions:
+      contents: write
+      pull-requests: write
+      statuses: write
+      issues: write
+
+    uses: tetherto/qvac-devops/.github/workflows/approval-check-worker.yml@production-workflows-tag
+    secrets: inherit
+    with:
+      pr_number: ${{ github.event.pull_request.number }}
+      pr_sha: ${{ github.event.pull_request.head.sha }}

.github/workflows/build-cmake-pkg.yml

@@ -25,7 +25,7 @@ jobs:
           cmake --build build --config Release
           cmake --install build --prefix "$PREFIX" --config Release
 
-          export LLAMA_CONFIG="$PREFIX"/lib/cmake/llama/llama-config.cmake
+          export LLAMA_CONFIG="$PREFIX"/share/llama/llama-config.cmake
           tclsh <<'EOF'
           set build(commit)  [string trim [exec git rev-parse --short HEAD]]
           set build(number)  [string trim [exec git rev-list  --count HEAD]]
@@ -47,5 +47,5 @@ jobs:
           EOF
 
           cd examples/simple-cmake-pkg
-          cmake -S . -B build -DCMAKE_PREFIX_PATH="$PREFIX"/lib/cmake
+          cmake -S . -B build -DCMAKE_PREFIX_PATH="$PREFIX"/share
           cmake --build build

.github/workflows/build-linux-cross.yml

@@ -4,6 +4,7 @@ on:
   workflow_call:
 
 jobs:
+  # Disabled. Fails to install some dependencies from arch-specific repositories.
   # ubuntu-24-riscv64-cpu-cross:
   #   runs-on: ubuntu-24.04
 

.github/workflows/build.yml

@@ -8,7 +8,9 @@ on:
     paths: [
       '.github/workflows/build.yml',
       '.github/workflows/build-linux-cross.yml',
-      '.github/workflows/build-cmake-pkg.yml',
+      # Disable. There are some modifications in the fork
+      # so that ggml dynamic builds work with vcpkg.
+      # '.github/workflows/build-cmake-pkg.yml',
       '**/CMakeLists.txt',
       '**/.cmake',
       '**/*.h',
@@ -28,7 +30,9 @@ on:
     paths: [
       '.github/workflows/build.yml',
       '.github/workflows/build-linux-cross.yml',
-      '.github/workflows/build-cmake-pkg.yml',
+      # Disable. There are some modifications in the fork
+      # so that ggml dynamic builds work with vcpkg.
+      # '.github/workflows/build-cmake-pkg.yml',
       '**/CMakeLists.txt',
       '**/.cmake',
       '**/*.h',
@@ -117,8 +121,9 @@ jobs:
             -DLLAMA_CURL=OFF \
             -DLLAMA_BUILD_BORINGSSL=ON \
             -DGGML_METAL=OFF \
-            -DGGML_RPC=ON \
-            -DCMAKE_OSX_DEPLOYMENT_TARGET=13.3
+            -DCMAKE_OSX_DEPLOYMENT_TARGET=13.3 \
+            -DGGML_BLAS=OFF \
+            -DGGML_RPC=ON
           cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
 
       - name: Test
@@ -269,11 +274,12 @@ jobs:
         id: checkout
         uses: actions/checkout@v4
 
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.16
-        with:
-          key: ubuntu-latest-cmake-sanitizer-${{ matrix.sanitizer }}
-          evict-old-files: 1d
+      # ccache disabled for sanitizer builds to ensure clean builds with correct sanitizer flags
+      # - name: ccache
+      #   uses: ggml-org/ccache-action@v1.2.16
+      #   with:
+      #     key: ubuntu-latest-cmake-sanitizer-${{ matrix.sanitizer }}
+      #     evict-old-files: 1d
 
       - name: Dependencies
         id: depends
@@ -308,6 +314,16 @@ jobs:
 
       - name: Test
         id: cmake_test
+        env:
+          # AddressSanitizer options
+          ASAN_OPTIONS: "verbosity=1:abort_on_error=1:print_stats=1:check_initialization_order=1:strict_init_order=1:detect_stack_use_after_return=1:print_summary=1:print_scariness=1:print_legend=1"
+          # ThreadSanitizer options
+          TSAN_OPTIONS: "verbosity=1:abort_on_error=1:print_stats=1:print_summary=1:print_legend=1"
+          # UndefinedBehaviorSanitizer options
+          # Note: abort_on_error=0 allows UBSAN to print full diagnostics before aborting
+          UBSAN_OPTIONS: "verbosity=2:abort_on_error=1:print_stacktrace=1:print_summary=1:halt_on_error=1:report_error_type=1:silence_unsigned_overflow=0"
+          # Common options for all sanitizers
+          MSAN_OPTIONS: "verbosity=1:abort_on_error=1:print_stats=1"
         run: |
           cd build
           ctest -L main --verbose --timeout 900
@@ -354,6 +370,7 @@ jobs:
         id: checkout
         uses: actions/checkout@v4
 
+      # Reduce false pass/failure on CI
       # - name: ccache
       #   uses: ggml-org/ccache-action@v1.2.16
       #   with:
@@ -389,11 +406,12 @@ jobs:
         id: checkout
         uses: actions/checkout@v4
 
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.16
-        with:
-          key: ubuntu-24-cmake-vulkan-deb
-          evict-old-files: 1d
+      # Reduce false pass/failure on CI
+      # - name: ccache
+      #   uses: ggml-org/ccache-action@v1.2.16
+      #   with:
+      #     key: ubuntu-24-cmake-vulkan
+      #     evict-old-files: 1d
 
       - name: Dependencies
         id: depends
@@ -417,7 +435,7 @@ jobs:
           cmake --build build -j $(nproc)
 
   ubuntu-24-cmake-vulkan:
-    runs-on: ubuntu-24.04
+    runs-on: ai-run-linux-gpu
 
     steps:
       - name: Clone
@@ -435,7 +453,7 @@ jobs:
         run: |
           sudo add-apt-repository -y ppa:kisak/kisak-mesa
           sudo apt-get update -y
-          sudo apt-get install -y build-essential mesa-vulkan-drivers libxcb-xinput0 libxcb-xinerama0 libxcb-cursor-dev libssl-dev
+          sudo apt-get install -y build-essential mesa-vulkan-drivers libxcb-xinput0 libxcb-xinerama0 libxcb-cursor-dev libssl-dev cmake
 
       - name: Get latest Vulkan SDK version
         id: vulkan_sdk_version
@@ -483,11 +501,12 @@ jobs:
         id: checkout
         uses: actions/checkout@v4
 
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.16
-        with:
-          key: ubuntu-24-cmake-webgpu
-          evict-old-files: 1d
+      # Reduce false pass/failure on CI
+      # - name: ccache
+      #   uses: ggml-org/ccache-action@v1.2.16
+      #   with:
+      #     key: ubuntu-24-cmake-webgpu
+      #     evict-old-files: 1d
 
       - name: Dependencies
         id: depends
@@ -602,11 +621,12 @@ jobs:
           sudo apt-get update
           sudo apt-get install -y build-essential git cmake rocblas-dev hipblas-dev libssl-dev rocwmma-dev
 
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.16
-        with:
-          key: ubuntu-22-cmake-hip
-          evict-old-files: 1d
+      # Reduce false pass/failure on CI
+      # - name: ccache
+      #   uses: ggml-org/ccache-action@v1.2.16
+      #   with:
+      #     key: ubuntu-22-cmake-hip
+      #     evict-old-files: 1d
 
       - name: Build with native CMake HIP support
         id: cmake_build
@@ -753,8 +773,10 @@ jobs:
   build-linux-cross:
     uses: ./.github/workflows/build-linux-cross.yml
 
-  build-cmake-pkg:
-    uses: ./.github/workflows/build-cmake-pkg.yml
+  # Disable. There are some modifications in the fork
+  # so that ggml dynamic builds work with vcpkg.
+  # build-cmake-pkg:
+  #   uses: ./.github/workflows/build-cmake-pkg.yml
 
   macOS-latest-cmake-ios:
     runs-on: macos-latest