Detect a usable CUDA backend

Detection uses the built-in lightweight native backend or a CUDA-enabled installation of the optional torch package. NVIDIA libraries are loaded only when diagnostics or CUDA selection is requested.

Usage

cuda_available()

Value

A single logical value.

Examples

cuda_available()
#> [1] FALSE