gpu support

[mudler]

This PR adds GPU support (inference untested, just tried build and run on colab).

OpenBLAS accelleration

To build and run with OpenBLAS, for example:

BUILD_TYPE=openblas make libbinding.a
CGO_LDFLAGS="-lopenblas" LIBRARY_PATH=$PWD C_INCLUDE_PATH=$PWD go run -tags openblas ./examples -m "/model/path/here" -t 14

GPU

To build with CuBLAS:

BUILD_TYPE=cublas make libbinding.a
CGO_LDFLAGS="-lcublas -lcudart -L/usr/local/cuda/lib64/" LIBRARY_PATH=$PWD C_INCLUDE_PATH=$PWD go run ./examples -m "/model/path/here" -t 14

[mudler]

Ok, tested on colab and seems to work - I'll merge as-is as I don't have the HW to test locally and don't want this to be pending over here. As it is opt-in there is on harm into merging this and iterate over if doesn't work correctly

[cubuzz]

I apologize if it's something I'm missing, but trying to build with CuBLAS yields

$ BUILD_TYPE=cublas make libbinding.a

cd build && cp -rf CMakeFiles/ggml.dir/ggml-cuda.cu.o ../llama.cpp/ggml-cuda.o
cp: cannot stat 'CMakeFiles/ggml.dir/ggml-cuda.cu.o': No such file or directory
make: *** [Makefile:162: llama.cpp/ggml-cuda.o] Error 1

Do I need to grab some other dependencies?

=======================================================================

RESOLVED: Somehow llama.cpp wasn't being built. Building it manually, then changing back fixes the issue.

[SilverViper]

I'm trying to build the Flowise example with docker-compose and I have the same error:
cp: cannot stat 'CMakeFiles/ggml.dir/ggml-cuda.cu.o': No such file or directory

Any hint on how I go about building llama.cpp manually? Or to prevent the error?