llama.cpp

7d787ed9 - ggml : do not crash when quantizing q4_x_x with an imatrix (#9192)