vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Expose generation timings from server & update completions.js (#2116)

Browse source 
* use javascript generators as much cleaner API

Also add ways to access completion as promise and EventSource

* export llama_timings as struct and expose them in server

* update readme, update baked includes

* llama : uniform variable names + struct init

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Tobias Lütke 2023-07-05 16:51:13 -04:00 • committed by GitHub
parent 983b555e9d
commit 31cfbb1013
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
9 changed files with 1921 additions and 1363 deletions
Download patch file Download diff file Expand all files Collapse all files

32
llama.cpp
View file

@ -3480,23 +3480,35 @@ llama_token llama_token_nl() {
return 13;
}
struct llama_timings llama_get_timings(struct llama_context * ctx) {
struct llama_timings result = {
/*.t_start_ms =*/ 1e-3 * ctx->t_start_us,
/*.t_end_ms =*/ 1.00 * ggml_time_ms(),
/*.t_load_ms =*/ 1e-3 * ctx->t_load_us,
/*.t_sample_ms =*/ 1e-3 * ctx->t_sample_us,
/*.t_p_eval_ms =*/ 1e-3 * ctx->t_p_eval_us,
/*.t_eval_ms =*/ 1e-3 * ctx->t_eval_us,
/*.n_sample =*/ std::max(1, ctx->n_sample),
/*.n_p_eval =*/ std::max(1, ctx->n_p_eval),
/*.n_eval =*/ std::max(1, ctx->n_eval),
};
return result;
}
void llama_print_timings(struct llama_context * ctx) {
const int64_t t_end_us = ggml_time_us();
const int32_t n_sample = std::max(1, ctx->n_sample);
const int32_t n_eval = std::max(1, ctx->n_eval);
const int32_t n_p_eval = std::max(1, ctx->n_p_eval);
const llama_timings timings = llama_get_timings(ctx);
fprintf(stderr, "\n");
fprintf(stderr, "%s: load time = %8.2f ms\n", __func__, ctx->t_load_us / 1000.0);
fprintf(stderr, "%s: load time = %8.2f ms\n", __func__, timings.t_load_ms);
fprintf(stderr, "%s: sample time = %8.2f ms / %5d runs (%8.2f ms per token, %8.2f tokens per second)\n",
__func__, 1e-3 * ctx->t_sample_us, n_sample, 1e-3 * ctx->t_sample_us / n_sample, 1e6 / ctx->t_sample_us * n_sample);
__func__, timings.t_sample_ms, timings.n_sample, timings.t_sample_ms / timings.n_sample, 1e3 / timings.t_sample_ms * timings.n_sample);
fprintf(stderr, "%s: prompt eval time = %8.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n",
__func__, 1e-3 * ctx->t_p_eval_us, n_p_eval, 1e-3 * ctx->t_p_eval_us / n_p_eval, 1e6 / ctx->t_p_eval_us * n_p_eval);
__func__, timings.t_p_eval_ms, timings.n_p_eval, timings.t_p_eval_ms / timings.n_p_eval, 1e3 / timings.t_p_eval_ms * timings.n_p_eval);
fprintf(stderr, "%s: eval time = %8.2f ms / %5d runs (%8.2f ms per token, %8.2f tokens per second)\n",
__func__, 1e-3 * ctx->t_eval_us, n_eval, 1e-3 * ctx->t_eval_us / n_eval, 1e6 / ctx->t_eval_us * n_eval);
fprintf(stderr, "%s: total time = %8.2f ms\n", __func__, (t_end_us - ctx->t_start_us)/1000.0);
__func__, timings.t_eval_ms, timings.n_eval, timings.t_eval_ms / timings.n_eval, 1e3 / timings.t_eval_ms * timings.n_eval);
fprintf(stderr, "%s: total time = %8.2f ms\n", __func__, (timings.t_end_ms - timings.t_start_ms));
}
void llama_reset_timings(struct llama_context * ctx) {