Skip to content

[CANN] Add the n_graph_splits performance metric to llama-bench. #12994

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

Closed
wants to merge 4 commits into from
Closed

[CANN] Add the n_graph_splits performance metric to llama-bench. #12994

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
bd6f1dd
Add the n_graph_splits metric to the performance test output paramete…
bachelor-dou Apr 17, 2025
b54eaa6
Adaptation of model name extraction for different systems
bachelor-dou Apr 17, 2025
3138f01
delete trailing whitespace
bachelor-dou Apr 17, 2025
feb049e
Undo model name changes
bachelor-dou Apr 18, 2025
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
21 changes: 14 additions & 7 deletions examples/llama-bench/llama-bench.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@
#include "common.h"
#include "ggml.h"
#include "llama.h"
#include "llama-context.h"
#include <filesystem>

#ifdef _WIN32
# define WIN32_LEAN_AND_MEAN
Expand Down Expand Up @@ -879,6 +881,7 @@ struct test {
const std::string cpu_info;
const std::string gpu_info;
std::string model_filename;
int n_graph_splits;
std::string model_type;
uint64_t model_size;
uint64_t model_n_params;
Expand Down Expand Up @@ -936,7 +939,7 @@ struct test {
std::strftime(buf, sizeof(buf), "%FT%TZ", gmtime(&t));
test_time = buf;

(void) ctx;
n_graph_splits = ctx->get_graph_splits();
}

uint64_t avg_ns() const { return ::avg(samples_ns); }
Expand Down Expand Up @@ -970,11 +973,11 @@ struct test {
static const std::vector<std::string> & get_fields() {
static const std::vector<std::string> fields = {
"build_commit", "build_number", "cpu_info", "gpu_info", "backends", "model_filename",
"model_type", "model_size", "model_n_params", "n_batch", "n_ubatch", "n_threads",
"cpu_mask", "cpu_strict", "poll", "type_k", "type_v", "n_gpu_layers",
"split_mode", "main_gpu", "no_kv_offload", "flash_attn", "tensor_split", "use_mmap",
"embeddings", "n_prompt", "n_gen", "test_time", "avg_ns", "stddev_ns",
"avg_ts", "stddev_ts",
"n_graph_splits", "model_type", "model_size", "model_n_params", "n_batch", "n_ubatch",
"n_threads", "cpu_mask", "cpu_strict", "poll", "type_k", "type_v",
"n_gpu_layers", "split_mode", "main_gpu", "no_kv_offload", "flash_attn", "tensor_split",
"use_mmap", "embeddings", "n_prompt", "n_gen", "test_time", "avg_ns",
"stddev_ns", "avg_ts", "stddev_ts",
};
return fields;
}
Expand All @@ -985,7 +988,7 @@ struct test {
if (field == "build_number" || field == "n_batch" || field == "n_ubatch" || field == "n_threads" ||
field == "poll" || field == "model_size" || field == "model_n_params" || field == "n_gpu_layers" ||
field == "main_gpu" || field == "n_prompt" || field == "n_gen" || field == "avg_ns" ||
field == "stddev_ns") {
field == "stddev_ns" || field == "n_graph_splits") {
return INT;
}
if (field == "f16_kv" || field == "no_kv_offload" || field == "cpu_strict" || field == "flash_attn" ||
Expand Down Expand Up @@ -1020,6 +1023,7 @@ struct test {
gpu_info,
get_backend(),
model_filename,
std::to_string(n_graph_splits),
model_type,
std::to_string(model_size),
std::to_string(model_n_params),
Expand Down Expand Up @@ -1196,6 +1200,9 @@ struct markdown_printer : public printer {
if (field == "n_gpu_layers") {
return 3;
}
if (field == "n_graph_splits") {
return 3;
}
if (field == "n_threads") {
return 7;
}
Expand Down
7 changes: 7 additions & 0 deletions src/llama-context.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2216,6 +2216,13 @@ void llama_context::perf_reset() {
t_p_eval_us = n_p_eval = 0;
}

/**
* @brief Get the number of graph splits.
*/
int llama_context::get_graph_splits() const{
return ggml_backend_sched_get_n_splits(sched.get());
}

//
// interface implementation
//
Expand Down
2 changes: 2 additions & 0 deletions src/llama-context.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -128,6 +128,8 @@ struct llama_context {
llama_perf_context_data perf_get_data() const;
void perf_reset();

int get_graph_splits() const;

private:
//
// output
Expand Down
Loading