Providing instructions on how to build libllama about llamasharp

[AsakusaRinne]

Hi, glad to hear that LLamaSharp works well for you. For building llama.cpp library, please use cmake .. -DBUILD_SHARED_LIBS=ON to config the cmake build (with other options such as openblas and cublas). Then you'll find the .dll or .so file. We'll also follow up the master branch of llama.cpp and publish a new release today or tomorrow. Thank you for your reminder. :)

[AsakusaRinne]

@DifferentialityDevelopment The master branch has been updated and n_gpu_layers param was added. Please try with master branch, or with v0.2.2 several hours later.

[Orfeous]

Awesome! Thank you.

[DifferentialityDevelopment]

I got llama built with your instructions and was able to get it running with the latest version of llama.cpp, the issue I encountered was rather that currently you can't pass the parameters to indicate how many layers to offload to the GPU using cuBLAS or OpenBLAS, unless I'm missing something.