[ci] add gemma test

.github/.test_durations

@@ -93,6 +93,7 @@
     "tests/utilization/dataset/test_formatting.py::test_multi_turn": 5.1760842725634575,
     "tests/utilization/model/test_apply_prompt_template.py::test_base": 0.0028072865679860115,
     "tests/utilization/model/test_apply_prompt_template.py::test_final_strip": 0.0016121016815304756,
+    "tests/utilization/model/test_apply_prompt_template.py::test_gemma": 0.009262016043066978,
     "tests/utilization/model/test_apply_prompt_template.py::test_llama2": 0.0018699830397963524,
     "tests/utilization/model/test_apply_prompt_template.py::test_no_smart_space": 0.0017823278903961182,
     "tests/utilization/model/test_apply_prompt_template.py::test_phi3": 0.009263357031159103,

tests/utilization/model/test_apply_prompt_template.py

@@ -72,6 +72,23 @@ def test_phi3(conversation: Conversation):
     )
 
 
+def test_gemma(conversation: Conversation):
+    formatter = ConversationFormatter.from_chat_template("gemma")
+    conversation.set_formatter(formatter)
+    formatted_conversation = conversation.apply_prompt_template()
+    assert formatted_conversation == (
+        "<bos><start_of_turn>user\n"
+        "This is a system message.\n"
+        "This is a user message.<end_of_turn>\n"
+        "<start_of_turn>model\n"
+        "This is an assistant message.<end_of_turn>\n"
+        "<start_of_turn>user\n"
+        "This is the second user message.<end_of_turn>\n"
+        "<start_of_turn>model\n"
+        "This is the second assistant message.<end_of_turn>\n"
+    )
+
+
 def test_no_smart_space(conversation: Conversation):
     prompt_config = {
         "system_start": "",

utilization/chat_templates.py

@@ -176,10 +176,11 @@ def smart_space(parts: List[Tuple[str, bool]], auto_leading_space: bool, no_spac
     "gemma": {
         "all_start": "<bos>",
         "merge_system_to_user": True,
+        "system_user_sep": "\n",
         "user_start": "<start_of_turn>user\n",
-        "user_end": "<end_of_turn>",
+        "user_end": "<end_of_turn>\n",
         "assistant_start": "<start_of_turn>model\n",
-        "assistant_end": "<end_of_turn>",
+        "assistant_end": "<end_of_turn>\n",
         "auto_leading_space": True,
         "final_rstrip": False,
         "no_space_between": True,

utilization/model/model_utils/conversation.py

@@ -57,6 +57,7 @@ def __init__(
         self.final_lstrip = chat_config.pop("final_lstrip", True)
         self.final_rstrip = chat_config.pop("final_rstrip", True)
         self.merge_system_to_user = chat_config.pop("merge_system_to_user", False)
+        self.system_user_sep = chat_config.pop("system_user_sep", "\n")
 
         # api model does not need bos_token
         if "bos_token" not in chat_config:
@@ -379,8 +380,10 @@ def get_generation_results(self) -> Union[str, Tuple[str, ...]]:
     def _merge_system_to_user(self):
         """Whether to convert system message to part of next user message."""
         if self.merge_system_to_user and self.messages[0]["role"] == "system":
-            self.messages[1]["content"] = self.messages[0]["content"] + self.messages[1]["content"]
+            msg = self.messages[0]["content"] + self.system_user_sep + self.messages[1]["content"]
+
             self.messages.pop(0)
+            self.messages[0]["content"] = msg
 
     def set_formatter(
         self,
@@ -392,6 +395,7 @@ def set_formatter(
         self.model_evaluation_method = model_evaluation_method
         self.split = split and self.get_segs_num() > 1
         self.merge_system_to_user = self.formatter.merge_system_to_user
+        self.system_user_sep = self.formatter.system_user_sep
         self._merge_system_to_user()
 
     def to_model_prompt(