CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-21 12:16:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

common : more accurate sampling timing (#17382)

Browse Source 
* common : more accurate sampling timing

* eval-callback : minor fixes

* cont : add time_meas impl

* cont : fix log msg [no ci]

* cont : fix multiple definitions of time_meas

* llama-cli : exclude chat template init from time measurement

* cont : print percentage of unaccounted time

* cont : do not reset timings
This commit is contained in:
Georgi Gerganov
2025-11-20 13:40:10 +02:00
committed by GitHub
parent 5088b435d4
commit 196f5083ef
7 changed files with 102 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
examples/eval-callback/eval-callback.cpp
View File

@@ -4,10 +4,10 @@
#include "llama.h"
#include "ggml.h"
#include <cmath>
#include <cstdio>
#include <string>
#include <vector>
#include <numeric>
/**
* This the arbitrary data which will be passed to each callback.
@@ -37,23 +37,23 @@ static inline float ggml_compute_bf16_to_fp32(ggml_bf16_t h) {
return u.f;
}
static float ggml_get_float_value(uint8_t * data, ggml_type type, const size_t * nb, size_t i0, size_t i1, size_t i2, size_t i3) {
static float ggml_get_float_value(const uint8_t * data, ggml_type type, const size_t * nb, size_t i0, size_t i1, size_t i2, size_t i3) {
size_t i = i3 * nb[3] + i2 * nb[2] + i1 * nb[1] + i0 * nb[0];
float v;
if (type == GGML_TYPE_F16) {
v = ggml_fp16_to_fp32(*(ggml_fp16_t *) &data[i]);
v = ggml_fp16_to_fp32(*(const ggml_fp16_t *) &data[i]);
} else if (type == GGML_TYPE_F32) {
v = *(float *) &data[i];
v = *(const float *) &data[i];
} else if (type == GGML_TYPE_I64) {
v = (float) *(int64_t *) &data[i];
v = (float) *(const int64_t *) &data[i];
} else if (type == GGML_TYPE_I32) {
v = (float) *(int32_t *) &data[i];
v = (float) *(const int32_t *) &data[i];
} else if (type == GGML_TYPE_I16) {
v = (float) *(int16_t *) &data[i];
v = (float) *(const int16_t *) &data[i];
} else if (type == GGML_TYPE_I8) {
v = (float) *(int8_t *) &data[i];
v = (float) *(const int8_t *) &data[i];
} else if (type == GGML_TYPE_BF16) {
v = ggml_compute_bf16_to_fp32(*(ggml_bf16_t *) &data[i]);
v = ggml_compute_bf16_to_fp32(*(const ggml_bf16_t *) &data[i]);
} else {
GGML_ABORT("fatal error");
}