lhutyra
diff --git a/‎llama_cpp/llama_chat_format.py
Copy file name to clipboardExpand all lines: llama_cpp/llama_chat_format.py
+9-6Lines changed: 9 additions & 6 deletions b/‎llama_cpp/llama_chat_format.py
Copy file name to clipboardExpand all lines: llama_cpp/llama_chat_format.py
+9-6Lines changed: 9 additions & 6 deletions
@@ -152,11 +152,13 @@ def __init__(
         template: str,
         eos_token: str,
         bos_token: str,
+        add_generation_prompt: bool = True,
     ):
         """A chat formatter that uses jinja2 templates to format the prompt."""
         self.template = template
         self.eos_token = eos_token
         self.bos_token = bos_token
+        self.add_generation_prompt = add_generation_prompt
 
         self._environment = jinja2.Environment(
             loader=jinja2.BaseLoader(),
@@ -170,12 +172,13 @@ def __call__(
         messages: List[llama_types.ChatCompletionRequestMessage],
         **kwargs: Any,
     ) -> ChatFormatterResponse:
-        messages = [
-            *messages,
-            llama_types.ChatCompletionRequestAssistantMessage(
-                role="assistant", content=""
-            ),
-        ]
+        if self.add_generation_prompt:
+            messages = [
+                *messages,
+                llama_types.ChatCompletionRequestAssistantMessage(
+                    role="assistant", content=""
+                ),
+            ]
         prompt = self._environment.render(
             messages=messages, eos_token=self.eos_token, bos_token=self.bos_token
         )