[Windows] Installation fails with "No CUDA toolset found" when CUDA 13.0 is installed

🐛 Describe the bug

Environment

• OS: Windows 10/11 (Build 26100)
• Python: 3.11.14 (conda environment)
• CUDA: 13.0 (C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0)
• Visual Studio: 2022 Professional with ClangCL toolset
• CMake: 4.2 (from pip cmake package)
• PyTorch: Installed via conda/pip

Steps to Reproduce

conda activate executorch
pip install -e . --no-build-isolation

Error Message

CMake Error at C:/Users/.../cmake/data/share/cmake-4.2/Modules/CMakeDetermineCompilerId.cmake:676 (message):
  No CUDA toolset found.
Call Stack (most recent call first):
  ...
  C:/.../torch/share/cmake/Caffe2/public/cuda.cmake:47 (enable_language)
  C:/.../torch/share/cmake/Caffe2/Caffe2Config.cmake:86 (include)
  C:/.../torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  tools/cmake/Utils.cmake:160 (find_package)
  CMakeLists.txt:792 (find_package_torch)

-- Configuring incomplete, errors occurred!
subprocess.CalledProcessError: Command '['cmake', ... '--preset', 'pybind', ...]' returned non-zero exit status 1.

Additional Context

CMake also outputs this warning before the error:

CMake Warning at .../CMakeDetermineCUDACompiler.cmake:16 (message):
  Visual Studio does not support specifying CUDAHOSTCXX or CMAKE_CUDA_HOST_COMPILER.
  Using the C++ compiler provided by Visual Studio.

Versions

Collecting environment information...
PyTorch version: 2.9.1+cu130
Is debug build: False
CUDA used to build PyTorch: 13.0
ROCM used to build PyTorch: N/A

OS: Microsoft Windows 11 Pro (10.0.26100 64-bit)
GCC version: Could not collect
Clang version: Could not collect
CMake version: version 4.2.0
Libc version: N/A

Python version: 3.11.14 | packaged by Anaconda, Inc. | (main, Oct 21 2025, 18:30:03) [MSC v.1929 64 bit (AMD64)] (64-bit runtime)
Python platform: Windows-10-10.0.26100-SP0
Is CUDA available: True
CUDA runtime version: 13.0.88
CUDA_MODULE_LOADING set to:
GPU models and configuration: GPU 0: NVIDIA GeForce RTX 5080
Nvidia driver version: 591.44
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Name: AMD Ryzen Threadripper PRO 7975WX 32-Cores
Manufacturer: AuthenticAMD
Family: 107
Architecture: 9
DeviceID: CPU0
CurrentClockSpeed: 4001
MaxClockSpeed: 4001
L2CacheSize: 32768
L2CacheSpeed: None
Revision: 6145

Versions of relevant libraries:
[pip3] executorch==1.1.0.dev20251003+cpu
[pip3] mypy_extensions==1.1.0
[pip3] numpy==2.3.5
[pip3] optimum-executorch==0.2.0.dev0
[pip3] pytorch_tokenizers==1.0.1
[pip3] torch==2.9.1+cu130
[pip3] torchao==0.15.0.dev20251104+cpu
[pip3] torchvision==0.24.1+cu130
[conda] executorch 1.1.0.dev20251003+cpu pypi_0 pypi
[conda] numpy 2.3.5 pypi_0 pypi
[conda] optimum-executorch 0.2.0.dev0 pypi_0 pypi
[conda] pytorch-tokenizers 1.0.1 pypi_0 pypi
[conda] torch 2.9.1+cu130 pypi_0 pypi
[conda] torchao 0.15.0.dev20251104+cpu pypi_0 pypi
[conda] torchvision 0.24.1+cu130 pypi_0 pypi

Start with CMakeLists.txt at the find_package_torch entry point and tools/cmake/Utils.cmake around its find_package call. Reproduce the pybind preset installation on the stated Windows, CMake, PyTorch, and CUDA versions, then trace the CUDA configuration through TorchConfig.cmake and Caffe2's cuda.cmake. Done means configuration proceeds without the “No CUDA toolset found” error on this environment.