Windows Release does not exist

[lonnietc]

your windows release version does not exist and I also tried to build it on Windows with Cargo v1.92 but it failed:

PS H:\dLLM-V02\fox\fox> cargo build --release
   Compiling ferrumox v0.12.0 (H:\dLLM-V02\fox\fox)
warning: ferrumox@0.12.0: Vulkan SDK found at C:\VulkanSDK\1.3.246.1 — building ggml-vulkan.dll
error: failed to run custom build command for `ferrumox v0.12.0 (H:\dLLM-V02\fox\fox)`

Caused by:
  process didn't exit successfully: `H:\dLLM-V02\fox\fox\target\release\build\ferrumox-c4e663375f83907b\build-script-build` (exit code: 101)
  --- stdout
  cargo:rustc-check-cfg=cfg(fox_stub)
  cargo:warning=Vulkan SDK found at C:\VulkanSDK\1.3.246.1 — building ggml-vulkan.dll
  CMAKE_TOOLCHAIN_FILE_x86_64-pc-windows-msvc = None
  CMAKE_TOOLCHAIN_FILE_x86_64_pc_windows_msvc = None
  HOST_CMAKE_TOOLCHAIN_FILE = None
  CMAKE_TOOLCHAIN_FILE = None
  CMAKE_PREFIX_PATH_x86_64-pc-windows-msvc = None
  CMAKE_PREFIX_PATH_x86_64_pc_windows_msvc = None
  HOST_CMAKE_PREFIX_PATH = None
  CMAKE_PREFIX_PATH = None
  CMAKE_x86_64-pc-windows-msvc = None
  CMAKE_x86_64_pc_windows_msvc = None
  HOST_CMAKE = None
  CMAKE = None
  -- Warning: ccache not found - consider installing it for faster compilation or disable this warning with GGML_CCACHE=OFF
  -- CMAKE_SYSTEM_PROCESSOR: AMD64
  -- CMAKE_GENERATOR_PLATFORM:
  -- GGML_SYSTEM_ARCH: x86
  -- Including CPU backend
  -- x86 detected
  -- Adding CPU backend variant ggml-cpu: /arch:AVX2 GGML_AVX2;GGML_FMA;GGML_F16C;__BMI2__;GGML_BMI2
  -- Configuring incomplete, errors occurred!

  --- stderr
  running: "cmake" "H:\\dLLM-V02\\fox\\fox\\vendor\\llama.cpp" "-B" "H:\\dLLM-V02\\fox\\fox\\target\\release\\build\\ferrumox-c91618bedff92d5b\\out\\build" "-G" "Ninja" "-DBUILD_SHARED_LIBS=ON" "-DGGML_BACKEND_DL=ON" "-DGGML_NATIVE=OFF" "-DLLAMA_BUILD_TESTS=OFF" "-DLLAMA_BUILD_TOOLS=OFF" "-DLLAMA_BUILD_EXAMPLES=OFF" "-DLLAMA_BUILD_SERVER=OFF" "-DLLAMA_BUILD_COMMON=OFF" "-DLLAMA_BUILD_APP=OFF" "-DLLAMA_BUILD_UI=OFF" "-DLLAMA_USE_PREBUILT_UI=OFF" "-DLLAMA_BUILD_WEBUI=OFF" "-DGGML_VULKAN=ON" "-DCMAKE_INSTALL_PREFIX=H:\\dLLM-V02\\fox\\fox\\target\\release\\build\\ferrumox-c91618bedff92d5b\\out" "-DCMAKE_C_FLAGS= -nologo -MD -Brepro -W0" "-DCMAKE_C_COMPILER=C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.43.34808/bin/HostX64/x64/cl.exe" "-DCMAKE_CXX_FLAGS= -nologo -MD -Brepro -W0" "-DCMAKE_CXX_COMPILER=C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.43.34808/bin/HostX64/x64/cl.exe" "-DCMAKE_ASM_FLAGS= -nologo -MD -Brepro -W0" "-DCMAKE_ASM_COMPILER=C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.43.34808/bin/HostX64/x64/cl.exe" "-DCMAKE_BUILD_TYPE=Release"
  CMAKE_BUILD_TYPE=Release
  CMake Error at ggml/src/ggml-vulkan/CMakeLists.txt:14 (find_package):
    Could not find a package configuration file provided by "SPIRV-Headers"
    with any of the following names:

      SPIRV-HeadersConfig.cmake
      spirv-headers-config.cmake

    Add the installation prefix of "SPIRV-Headers" to CMAKE_PREFIX_PATH or set
    "SPIRV-Headers_DIR" to a directory containing one of the above files.  If
    "SPIRV-Headers" provides a separate development package or SDK, be sure it
    has been installed.



  thread 'main' (29060) panicked at C:\Users\lonni\.cargo\registry\src\index.crates.io-1949cf8c6b5b557f\cmake-0.1.57\src\lib.rs:1132:5:

  command did not execute successfully, got: exit code: 1

  build script failed, must exit now

[ManuelSLemos]

There's no Windows release, correct. The workflow builds two Linux x86_64 targets and nothing else; the other platforms were pulled until someone verifies them. That's why install.ps1 checks for the asset and tells you instead of 404ing. Windows means building from source for now.

Your build error is llama.cpp's, not fox's. ggml-vulkan/CMakeLists.txt:14 does find_package(SPIRV-Headers CONFIG REQUIRED), and SDK 1.3.246.1 is from 2023 and doesn't install SPIRV-HeadersConfig.cmake. Not a path problem, the lines above it already add $VULKAN_SDK to CMAKE_PREFIX_PATH. Update to a current SDK.

Your log also says ferrumox v0.12.0, which is nine releases back with a different llama.cpp pinned. Worth doing first:

git pull && git submodule update --init --recursive

What was our fault: build.rs enabled Vulkan on any one of VULKAN_SDK, glslc or vulkan.h, but ggml-vulkan needs the loader, glslc and SPIRV-Headers, each REQUIRED. Missing one aborts the configure instead of falling back, and there was no way to turn Vulkan off, so you couldn't build even CPU-only. Fixed in 0.21.0: all three get checked, a partial toolchain gives a CPU build and a warning naming what's missing, and FOX_NO_VULKAN=1 / FOX_FORCE_VULKAN=1 override it either way. Note that installing a package doesn't make cargo re-run the build script, so after updating your SDK rebuild with FOX_FORCE_VULKAN=1 or it'll reuse the old answer.

To build today on what you have, clear the variable for that shell:

$env:VULKAN_SDK = ""
cargo build --release

I tested the fix on Linux but have no Windows machine, so that path is unverified. If you try 0.21.0 when it lands, I'd like to know whether it configures for you.