Topic: device_info

A code update in ggml/llama.cpp (commit b10594) optimizes device_info handling by skipping unnecessary loops when device information is not printed, reducing overhead, particularly for the CUDA backend. This improvement enhances efficiency without affecting functionality.