Skip to content

llama-bench : log benchmark progress #9287

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Sep 6, 2024
Merged

llama-bench : log benchmark progress #9287

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
eb9f945
llama-bench : add optional progress messages
akx Sep 6, 2024
5f3047f
fixes
slaren Sep 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 26 additions & 2 deletions examples/llama-bench/llama-bench.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -249,6 +249,7 @@ struct cmd_params {
ggml_sched_priority prio;
int delay;
bool verbose;
bool progress;
output_formats output_format;
output_formats output_format_stderr;
};
Expand Down Expand Up @@ -280,6 +281,7 @@ static const cmd_params cmd_params_defaults = {
/* prio */ GGML_SCHED_PRIO_NORMAL,
/* delay */ 0,
/* verbose */ false,
/* progress */ false,
/* output_format */ MARKDOWN,
/* output_format_stderr */ NONE,
};
Expand Down Expand Up @@ -319,6 +321,7 @@ static void print_usage(int /* argc */, char ** argv) {
printf(" -o, --output <csv|json|jsonl|md|sql> (default: %s)\n", output_format_str(cmd_params_defaults.output_format));
printf(" -oe, --output-err <csv|json|jsonl|md|sql> (default: %s)\n", output_format_str(cmd_params_defaults.output_format_stderr));
printf(" -v, --verbose (default: %s)\n", cmd_params_defaults.verbose ? "1" : "0");
printf(" --progress (default: %s)\n", cmd_params_defaults.progress ? "1" : "0");
printf("\n");
printf("Multiple values can be given for each parameter by separating them with ',' or by specifying the parameter multiple times.\n");
}
Expand Down Expand Up @@ -364,6 +367,7 @@ static cmd_params parse_cmd_params(int argc, char ** argv) {
params.numa = cmd_params_defaults.numa;
params.prio = cmd_params_defaults.prio;
params.delay = cmd_params_defaults.delay;
params.progress = cmd_params_defaults.progress;

for (int i = 1; i < argc; i++) {
arg = argv[i];
Expand Down Expand Up @@ -616,6 +620,8 @@ static cmd_params parse_cmd_params(int argc, char ** argv) {
invalid_param = !output_format_from_str(argv[i], params.output_format_stderr);
} else if (arg == "-v" || arg == "--verbose") {
params.verbose = true;
} else if (arg == "--progress") {
params.progress = true;
} else {
invalid_param = true;
break;
Expand Down Expand Up @@ -1523,7 +1529,13 @@ int main(int argc, char ** argv) {
llama_model * lmodel = nullptr;
const cmd_params_instance * prev_inst = nullptr;

int params_idx = 0;
auto params_count = params_instances.size();
for (const auto & inst : params_instances) {
params_idx ++;
if (params.progress) {
fprintf(stderr, "llama-bench: benchmark %d/%ld: starting\n", params_idx, params_count);
}
// keep the same model between tests when possible
if (!lmodel || !prev_inst || !inst.equal_mparams(*prev_inst)) {
if (lmodel) {
Expand Down Expand Up @@ -1556,7 +1568,7 @@ int main(int argc, char ** argv) {

struct ggml_threadpool_params tpp = ggml_threadpool_params_default(t.n_threads);
if (!parse_cpu_mask(t.cpu_mask, tpp.cpumask)) {
LOG_TEE("%s: failed to parse cpu-mask: %s\n", __func__, t.cpu_mask.c_str());
fprintf(stderr, "%s: failed to parse cpu-mask: %s\n", __func__, t.cpu_mask.c_str());
exit(1);
}
tpp.strict_cpu = t.cpu_strict;
Expand All @@ -1565,18 +1577,24 @@ int main(int argc, char ** argv) {

struct ggml_threadpool* threadpool = ggml_threadpool_new(&tpp);
if (!threadpool) {
LOG_TEE("%s: threadpool create failed : n_threads %d\n", __func__, tpp.n_threads);
fprintf(stderr, "%s: threadpool create failed : n_threads %d\n", __func__, tpp.n_threads);
exit(1);
}

llama_attach_threadpool(ctx, threadpool, NULL);

// warmup run
if (t.n_prompt > 0) {
if (params.progress) {
fprintf(stderr, "llama-bench: benchmark %d/%ld: warmup prompt run\n", params_idx, params_count);
}
//test_prompt(ctx, std::min(t.n_batch, std::min(t.n_prompt, 32)), 0, t.n_batch, t.n_threads);
test_prompt(ctx, t.n_prompt, 0, t.n_batch, t.n_threads);
}
if (t.n_gen > 0) {
if (params.progress) {
fprintf(stderr, "llama-bench: benchmark %d/%ld: warmup generation run\n", params_idx, params_count);
}
test_gen(ctx, 1, 0, t.n_threads);
}

Expand All @@ -1586,9 +1604,15 @@ int main(int argc, char ** argv) {
uint64_t t_start = get_time_ns();

if (t.n_prompt > 0) {
if (params.progress) {
fprintf(stderr, "llama-bench: benchmark %d/%ld: prompt run %d/%d\n", params_idx, params_count, i + 1, params.reps);
}
test_prompt(ctx, t.n_prompt, 0, t.n_batch, t.n_threads);
}
if (t.n_gen > 0) {
if (params.progress) {
fprintf(stderr, "llama-bench: benchmark %d/%ld: generation run %d/%d\n", params_idx, params_count, i + 1, params.reps);
}
test_gen(ctx, t.n_gen, t.n_prompt, t.n_threads);
}

Expand Down
Loading