update hunyuan dense chat template

Signed-off-by: stevenkuang <[email protected]>

Author: stevenkuang-tencent

convert_hf_to_gguf.py

@@ -7533,6 +7533,7 @@ class HunYuanModel(TextModel):
     def set_vocab(self):
         if (self.dir_model / "tokenizer.json").is_file():
             self._set_vocab_gpt2()
+            self.gguf_writer.add_add_bos_token(True)
         else:
             from transformers import AutoTokenizer
             tokenizer = AutoTokenizer.from_pretrained(self.dir_model, trust_remote_code=True)

src/llama-chat.cpp

@@ -712,9 +712,7 @@ int32_t llm_chat_apply_template(
             std::string role(chat[i]->role);
             if (i == 0) {
                 if (role == "system") {
-                    ss << "<｜hy_begin▁of▁sentence｜>" << chat[i]->content << "<｜hy_place▁holder▁no▁3｜>";
-                } else {
-                    ss << "<｜hy_begin▁of▁sentence｜>";
+                    ss << chat[i]->content << "<｜hy_place▁holder▁no▁3｜>";
                 }
             }