llama.cpp

80c41ddd - CUDA: compress mode option and default to size (#12029)

CUDA: compress mode option and default to size (#12029) cuda 12.8 added the option to specify stronger compression for binaries, so we now default to "size".