[SUGGESTION] Wider hardware acceleration support

[MarcoRavich]

Hi there, 1st of all thanks for this awesome work !

It would be extremely useful - and environmentally friendly, since the computational power required if often heavy - to allow end users to squeeze the full potential of the hardware they use.

A cool hw-exploit example comes from @AmusementClub's vs-mlrt project that provides VapourSynth ML filter runtimes for some video enhancing models on x86 CPUs, Intel GPUs (both integrated & discrete), NVidia GPUs and AMD GPUs.

Another one is @styler00dollar's VSGAN-tensorrt-docker that claims to be the fastest inference code that you can find (TensorRT only).

Last but not least, check out our AI-based page where we collected other open projects that may help to evolve yours.

Hope that inspires !