ggml-org/llama.cpp b7682

on GitHub

server : use different seeds for child completions (#18700)

• server : use different seeds for child completions

• cont : handle default seed

• cont : note