llama: offload output layer to GPU first#18148

JohannesGaessler merged 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

JohannesGaessler:llama-offload-output-firstJohannesGaessler/llama.cpp:llama-offload-output-firstCopy head branch name to clipboard