unicode : switch to multimap based nfd_map (#5799)

* switch to multimap based nfd_map due to compile time issues

* simplify multimap keys

* dont construct new locale every time

1 files changed, 6 insertions, 5 deletions

diff --git a/llama.cpp b/llama.cpp
index 62699ce5..a35f07aa 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -8947,10 +8947,10 @@ struct llm_tokenizer_wpm {
         std::vector<uint32_t> codepoints = codepoints_from_utf8(text);
         std::vector<uint32_t> nfd_codepoints;
         for (uint32_t code : codepoints) {
-            auto it = nfd_map.find(code);
-            if (it != nfd_map.end()) {
-                for (uint32_t c : it->second) {
-                    nfd_codepoints.push_back(c);
+            auto it = nfd_map.equal_range(code);
+            if (it.first != it.second) {
+                for (auto jt = it.first; jt != it.second; jt++) {
+                    nfd_codepoints.push_back(jt->second);
                 }
             } else {
                 nfd_codepoints.push_back(code);
@@ -9001,12 +9001,13 @@ struct llm_tokenizer_wpm {
     }
 
     uint32_t to_lower(uint32_t code) {
+        static const std::locale locale("en_US.UTF-8");
 #if defined(_WIN32)
         if (code > 0xFFFF) {
             return code;
         }
 #endif
-        return std::tolower(wchar_t(code), std::locale("en_US.UTF-8"));
+        return std::tolower(wchar_t(code), locale);
     }
 
     bool is_ascii_punct(uint32_t code) {