Setting mmap and direct_io to false as default in llama-bench.cpp (#18841)

Author: JTischbein

tools/llama-bench/llama-bench.cpp

@@ -372,8 +372,8 @@ static const cmd_params cmd_params_defaults = {
     /* devices              */ { {} },
     /* tensor_split         */ { std::vector<float>(llama_max_devices(), 0.0f) },
     /* tensor_buft_overrides*/ { std::vector<llama_model_tensor_buft_override>{ { nullptr, nullptr } } },
-    /* use_mmap             */ { true },
-    /* use_direct_io        */ { true },
+    /* use_mmap             */ { false },
+    /* use_direct_io        */ { false },
     /* embeddings           */ { false },
     /* no_op_offload        */ { false },
     /* no_host              */ { false },