common, ngram_cache: added const reference for std::pair<> and std::…

…tuple<> more 16 bytes: ggerganov#7270 Co-Authored-By: Herman Semenov <[email protected]>
Nexesenex · Aug 15, 2024 · 14d93c7 · 14d93c7
1 parent c65f5b6
commit 14d93c7
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/common/ngram-cache.cpp b/common/ngram-cache.cpp
@@ -306,7 +306,7 @@ 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());
@@ -366,7 +366,7 @@ 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;