Backends for the deep learning frameworks in use are CPU or CUDA. OpenCL support is essentially non-existent. There's no point benchmarking something that isn't

Backends for the deep learning frameworks in use are CPU or CUDA. OpenCL support is essentially non-existent. There's no point benchmarking something that isn't going to be used.