llama.cpp GPU support

[alexl83]

Hi, since commit 905d87b70aa189623d500a28602d7a3a755a4769
llama.cpp support GPU inference with nvidia CUDA via command-line switches like --gpu-layers
Could you please consider adding support to GPT-LLAMA.CPP aswell?

Thank you!

[msj121]

@alexl83
Just looking at the code, since you compile llama.cpp in theory it would appear to me that you can install with cuda support, then you are just passing the argument like any of the others listed, like threads.

ie:
npm start ngl 4 to hoist on to 4 gpu layers. I don't have a compatible setup to test though.