Why there is no Ort::MemoryInfo::CreateGPU or CreateCUDA?

[UnlimitedR]

I see all the codes are like this

Ort::MemoryInfo memoryInfo = Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault); 
Ort::Value input_tensor = Ort::Value::CreateTensor<float>(memoryInfo, ...);

why it is always CreateCpu? what if I want a CUDA tensor to accelerate the calculation?
and what about OrtAllocatorType? There seems to be OrtDeviceAllocator and OrtArenaAllocator, I'm not sure about their differences.

[Ilya Druker (idruker-cerence)]

I don't know why they did not define them, but you can use
Ort::MemoryInfo("CudaPinned", OrtAllocatorType::OrtArenaAllocator, 0, OrtMemTypeDefault)
for allocation on device-mapped CPU memory and
Ort::MemoryInfo("Cuda", OrtAllocatorType::OrtArenaAllocator, 0, OrtMemTypeDefault)
for allocation on device.