llama.cpp

1bf681f9 - ggml : add error handling to graph_compute (whisper/1714)