server : print graphs reused in slot timings (#23279)

Add graphs reused counter to the per-slot timing output, printed via
llama_perf_context().

Assisted-by: llama.cpp:local pi

Co-authored-by: ggerganov <ggerganov@users.noreply.github.com>
This commit is contained in:
Georgi Gerganov 2026-05-19 09:46:58 +03:00 committed by GitHub
parent cd963fee6a
commit 3c81c8deea
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -467,20 +467,26 @@ struct server_slot {
const double n_gen_second = 1e3 / t_token_generation * n_decoded;
SLT_INF(*this,
"\n"
"prompt eval time = %10.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n"
" eval time = %10.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n"
"prompt eval time = %10.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n",
t_prompt_processing, n_prompt_tokens_processed, t_prompt, n_prompt_second);
SLT_INF(*this,
" eval time = %10.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n",
t_token_generation, n_decoded, t_gen, n_gen_second);
SLT_INF(*this,
" total time = %10.2f ms / %5d tokens\n",
t_prompt_processing, n_prompt_tokens_processed, t_prompt, n_prompt_second,
t_token_generation, n_decoded, t_gen, n_gen_second,
t_prompt_processing + t_token_generation, n_prompt_tokens_processed + n_decoded);
SLT_INF(*this,
" graphs reused = %10d\n",
llama_perf_context(ctx_tgt).n_reused);
if (n_draft_total > 0) {
const float draft_ratio = (float) n_draft_accepted / n_draft_total;
SLT_CNT(*this,
"draft acceptance rate = %0.5f (%5d accepted / %5d generated)\n",
draft_ratio, n_draft_accepted, n_draft_total
);
SLT_INF(*this,
"draft acceptance = %0.5f (%5d accepted / %5d generated)\n",
draft_ratio, n_draft_accepted, n_draft_total);
}
common_speculative_print_stats(spec);