I'm certain inference engines don't use hand-tuned CUDA on Radeon or Mac Mini chips. My statement holds: those engines have no strict dependency on CUDA, or the

I'm certain inference engines don't use hand-tuned CUDA on Radeon or Mac Mini chips. My statement holds: those engines have no strict dependency on CUDA, or they'd be Nvidia-only.