Stop the generation when <|eom_id|> token is encountered (needed for llama 3.1 tool call support) by fairydreaming · Pull Request #8858 · ggerganov/llama.cpp