llama.cpp

5b335f41 - ggml : check for allocation failures to prevent crashes (#28149)