mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-10-31 08:51:55 +00:00
llama : refactor src/llama.cpp (#10902)
* llama : scatter llama.cpp into multiple modules (wip) * llama : control-vector -> adapter * llama : arch * llama : mmap ggml-ci * ci : remove BUILD_SHARED_LIBS=OFF ggml-ci * llama : arch (cont) ggml-ci * llama : chat ggml-ci * llama : model ggml-ci * llama : hparams ggml-ci * llama : adapter ggml-ci * examples : fix ggml-ci * rebase ggml-ci * minor * llama : kv cache ggml-ci * llama : impl ggml-ci * llama : batch ggml-ci * cont ggml-ci * llama : context ggml-ci * minor * llama : context (cont) ggml-ci * llama : model loader ggml-ci * common : update lora ggml-ci * llama : quant ggml-ci * llama : quant (cont) ggml-ci * minor [no ci]
This commit is contained in:
Georgi Gerganov
166
src/llama-impl.cpp
Normal file
166
src/llama-impl.cpp
Normal file
@@ -0,0 +1,166 @@
|
||||
#include "llama-impl.h"
|
||||
|
||||
#include "llama.h"
|
||||
|
||||
#include <cinttypes>
|
||||
#include <climits>
|
||||
#include <cstdarg>
|
||||
#include <cstring>
|
||||
#include <vector>
|
||||
#include <sstream>
|
||||
|
||||
struct llama_logger_state {
|
||||
ggml_log_callback log_callback = llama_log_callback_default;
|
||||
void * log_callback_user_data = nullptr;
|
||||
};
|
||||
|
||||
static llama_logger_state g_logger_state;
|
||||
|
||||
time_meas::time_meas(int64_t & t_acc, bool disable) : t_start_us(disable ? -1 : ggml_time_us()), t_acc(t_acc) {}
|
||||
|
||||
time_meas::~time_meas() {
|
||||
if (t_start_us >= 0) {
|
||||
t_acc += ggml_time_us() - t_start_us;
|
||||
}
|
||||
}
|
||||
|
||||
void llama_log_set(ggml_log_callback log_callback, void * user_data) {
|
||||
ggml_log_set(log_callback, user_data);
|
||||
g_logger_state.log_callback = log_callback ? log_callback : llama_log_callback_default;
|
||||
g_logger_state.log_callback_user_data = user_data;
|
||||
}
|
||||
|
||||
static void llama_log_internal_v(ggml_log_level level, const char * format, va_list args) {
|
||||
va_list args_copy;
|
||||
va_copy(args_copy, args);
|
||||
char buffer[128];
|
||||
int len = vsnprintf(buffer, 128, format, args);
|
||||
if (len < 128) {
|
||||
g_logger_state.log_callback(level, buffer, g_logger_state.log_callback_user_data);
|
||||
} else {
|
||||
char * buffer2 = new char[len + 1];
|
||||
vsnprintf(buffer2, len + 1, format, args_copy);
|
||||
buffer2[len] = 0;
|
||||
g_logger_state.log_callback(level, buffer2, g_logger_state.log_callback_user_data);
|
||||
delete[] buffer2;
|
||||
}
|
||||
va_end(args_copy);
|
||||
}
|
||||
|
||||
void llama_log_internal(ggml_log_level level, const char * format, ...) {
|
||||
va_list args;
|
||||
va_start(args, format);
|
||||
llama_log_internal_v(level, format, args);
|
||||
va_end(args);
|
||||
}
|
||||
|
||||
void llama_log_callback_default(ggml_log_level level, const char * text, void * user_data) {
|
||||
(void) level;
|
||||
(void) user_data;
|
||||
fputs(text, stderr);
|
||||
fflush(stderr);
|
||||
}
|
||||
|
||||
void replace_all(std::string & s, const std::string & search, const std::string & replace) {
|
||||
if (search.empty()) {
|
||||
return;
|
||||
}
|
||||
std::string builder;
|
||||
builder.reserve(s.length());
|
||||
size_t pos = 0;
|
||||
size_t last_pos = 0;
|
||||
while ((pos = s.find(search, last_pos)) != std::string::npos) {
|
||||
builder.append(s, last_pos, pos - last_pos);
|
||||
builder.append(replace);
|
||||
last_pos = pos + search.length();
|
||||
}
|
||||
builder.append(s, last_pos, std::string::npos);
|
||||
s = std::move(builder);
|
||||
}
|
||||
|
||||
std::string format(const char * fmt, ...) {
|
||||
va_list ap;
|
||||
va_list ap2;
|
||||
va_start(ap, fmt);
|
||||
va_copy(ap2, ap);
|
||||
int size = vsnprintf(NULL, 0, fmt, ap);
|
||||
GGML_ASSERT(size >= 0 && size < INT_MAX); // NOLINT
|
||||
std::vector<char> buf(size + 1);
|
||||
int size2 = vsnprintf(buf.data(), size + 1, fmt, ap2);
|
||||
GGML_ASSERT(size2 == size);
|
||||
va_end(ap2);
|
||||
va_end(ap);
|
||||
return std::string(buf.data(), size);
|
||||
}
|
||||
|
||||
std::string llama_format_tensor_shape(const std::vector<int64_t> & ne) {
|
||||
char buf[256];
|
||||
snprintf(buf, sizeof(buf), "%5" PRId64, ne.at(0));
|
||||
for (size_t i = 1; i < ne.size(); i++) {
|
||||
snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), ", %5" PRId64, ne.at(i));
|
||||
}
|
||||
return buf;
|
||||
}
|
||||
|
||||
std::string llama_format_tensor_shape(const struct ggml_tensor * t) {
|
||||
char buf[256];
|
||||
snprintf(buf, sizeof(buf), "%5" PRId64, t->ne[0]);
|
||||
for (int i = 1; i < GGML_MAX_DIMS; i++) {
|
||||
snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), ", %5" PRId64, t->ne[i]);
|
||||
}
|
||||
return buf;
|
||||
}
|
||||
|
||||
static std::string gguf_data_to_str(enum gguf_type type, const void * data, int i) {
|
||||
switch (type) {
|
||||
case GGUF_TYPE_UINT8: return std::to_string(((const uint8_t *)data)[i]);
|
||||
case GGUF_TYPE_INT8: return std::to_string(((const int8_t *)data)[i]);
|
||||
case GGUF_TYPE_UINT16: return std::to_string(((const uint16_t *)data)[i]);
|
||||
case GGUF_TYPE_INT16: return std::to_string(((const int16_t *)data)[i]);
|
||||
case GGUF_TYPE_UINT32: return std::to_string(((const uint32_t *)data)[i]);
|
||||
case GGUF_TYPE_INT32: return std::to_string(((const int32_t *)data)[i]);
|
||||
case GGUF_TYPE_UINT64: return std::to_string(((const uint64_t *)data)[i]);
|
||||
case GGUF_TYPE_INT64: return std::to_string(((const int64_t *)data)[i]);
|
||||
case GGUF_TYPE_FLOAT32: return std::to_string(((const float *)data)[i]);
|
||||
case GGUF_TYPE_FLOAT64: return std::to_string(((const double *)data)[i]);
|
||||
case GGUF_TYPE_BOOL: return ((const bool *)data)[i] ? "true" : "false";
|
||||
default: return format("unknown type %d", type);
|
||||
}
|
||||
}
|
||||
|
||||
std::string gguf_kv_to_str(const struct gguf_context * ctx_gguf, int i) {
|
||||
const enum gguf_type type = gguf_get_kv_type(ctx_gguf, i);
|
||||
|
||||
switch (type) {
|
||||
case GGUF_TYPE_STRING:
|
||||
return gguf_get_val_str(ctx_gguf, i);
|
||||
case GGUF_TYPE_ARRAY:
|
||||
{
|
||||
const enum gguf_type arr_type = gguf_get_arr_type(ctx_gguf, i);
|
||||
int arr_n = gguf_get_arr_n(ctx_gguf, i);
|
||||
const void * data = gguf_get_arr_data(ctx_gguf, i);
|
||||
std::stringstream ss;
|
||||
ss << "[";
|
||||
for (int j = 0; j < arr_n; j++) {
|
||||
if (arr_type == GGUF_TYPE_STRING) {
|
||||
std::string val = gguf_get_arr_str(ctx_gguf, i, j);
|
||||
// escape quotes
|
||||
replace_all(val, "\\", "\\\\");
|
||||
replace_all(val, "\"", "\\\"");
|
||||
ss << '"' << val << '"';
|
||||
} else if (arr_type == GGUF_TYPE_ARRAY) {
|
||||
ss << "???";
|
||||
} else {
|
||||
ss << gguf_data_to_str(arr_type, data, j);
|
||||
}
|
||||
if (j < arr_n - 1) {
|
||||
ss << ", ";
|
||||
}
|
||||
}
|
||||
ss << "]";
|
||||
return ss.str();
|
||||
}
|
||||
default:
|
||||
return gguf_data_to_str(type, gguf_get_val_data(ctx_gguf, i), 0);
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user