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

Added const reference for std::pair<> and std::tuple<> more 16 bytes:

Browse source 
- std::pair<llama_ngram, llama_ngram_cache_part> (72 bytes -> 8 bytes)
- std::tuple<std::string, float> (40 bytes -> 8 bytes)
This commit is contained in:
Herman Semenov 2024-05-13 20:07:53 -05:00
parent b228aba91a
commit ced5bfeb33
2 changed files with 4 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

4
common/common.cpp
View file

@ -2669,14 +2669,14 @@ void dump_non_result_info_yaml(FILE * stream, const gpt_params & params, const l
}
fprintf(stream, "lora:\n");
for (std::tuple<std::string, float> la : params.lora_adapter) {
for (const std::tuple<std::string, float> & la : params.lora_adapter) {
if (std::get<1>(la) != 1.0f) {
continue;
}
fprintf(stream, " - %s\n", std::get<0>(la).c_str());
}
fprintf(stream, "lora_scaled:\n");
for (std::tuple<std::string, float> la : params.lora_adapter) {
for (const std::tuple<std::string, float> & la : params.lora_adapter) {
if (std::get<1>(la) == 1.0f) {
continue;
}

4
common/ngram-cache.cpp
View file

@ -195,7 +195,7 @@ void llama_ngram_cache_draft(
void llama_ngram_cache_save(llama_ngram_cache & ngram_cache, std::string & filename) {
std::ofstream file_out(filename, std::ios::binary);
for (std::pair<llama_ngram, llama_ngram_cache_part> item : ngram_cache) {
for (const std::pair<llama_ngram, llama_ngram_cache_part> & item : ngram_cache) {
const llama_ngram ngram = item.first;
llama_ngram_cache_part token_counts = item.second;
GGML_ASSERT(!token_counts.empty());
@ -255,7 +255,7 @@ llama_ngram_cache llama_ngram_cache_load(std::string & filename) {
}
void llama_ngram_cache_merge(llama_ngram_cache & ngram_cache_target, llama_ngram_cache & ngram_cache_add) {
for (std::pair<llama_ngram, llama_ngram_cache_part> ngram_part : ngram_cache_add) {
for (const std::pair<llama_ngram, llama_ngram_cache_part> & ngram_part : ngram_cache_add) {
const llama_ngram ngram = ngram_part.first;
llama_ngram_cache_part part = ngram_part.second;