THIRD-PARTY SOFTWARE NOTICES AND INFORMATION
For the Windows ML Runtime llama.cpp CUDA backend (LlamaBackends/cuda/ggml-cuda.dll)

This notice supplements ThirdPartyNotices.WinMLRuntimeLlama.txt in the
Microsoft.Windows.AI.MachineLearning.LibLlama.Core package. The CUDA backend
plugin is compiled from the same pinned llama.cpp/ggml source as the shared core,
which is covered by that notice.

-------------------------------------------------------------------------------
Component roles
-------------------------------------------------------------------------------

Shipped runtime components
  - NVIDIA CUDA Runtime  NVIDIA CUDA Toolkit End User License Agreement
                         cudart_static.lib from the pinned CUDA Toolkit,
                         statically linked into ggml-cuda.dll.

Components not included
  The package does not include NVIDIA cuBLAS (cublas64_*.dll and
  cublasLt64_*.dll), which ggml-cuda.dll loads from its own directory. The
  application deploys cuBLAS beside ggml-cuda.dll, for example from its NVIDIA
  CUDA Toolkit installation, under the NVIDIA license terms that accompany it.

-------------------------------------------------------------------------------
Where to find the full license texts
-------------------------------------------------------------------------------

This package includes notices/cuda/<architecture>/llama-cpp/copyright, the
copyright file produced by the CUDA backend build. It contains the llama.cpp
license followed by the complete NVIDIA CUDA Toolkit license text for the CUDA
Runtime.