llama.cpp

11d90236 - ggml : add helper debug printf in soft_max