support for Poro chat pre-tokenizer

llama.cpp

@@ -4596,7 +4596,11 @@ static void llm_load_vocab(
             } else if (
                 tokenizer_pre == "smaug-bpe") {
                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_SMAUG;
-            } else {
+            } else if (
+                tokenizer_pre == "Poro-34B-chat") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_PORO;
+            } 
+            else {
                 throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
             }
         } else {
@@ -12580,6 +12584,11 @@ struct llm_tokenizer_bpe {
                             "(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
                         });
                         break;
+                    case LLAMA_VOCAB_PRE_TYPE_PORO:
+                        word_collection = unicode_regex_split(text, {
+                            " ?[^(\\s|.,!?…。，、।۔،)]+",
+                        });
+                        break;
                     default:
                         // default regex for BPE tokenization pre-processing
                         word_collection = unicode_regex_split(text, {

llama.h

@@ -86,6 +86,7 @@ extern "C" {
         LLAMA_VOCAB_PRE_TYPE_OLMO           = 12,
         LLAMA_VOCAB_PRE_TYPE_DBRX           = 13,
         LLAMA_VOCAB_PRE_TYPE_SMAUG          = 14,
+        LLAMA_VOCAB_PRE_TYPE_PORO           = 15,
     };
 
     // note: these values should be synchronized with ggml_rope