[v1] add qwen3 templates and fix rendering plugin. (#10212)

Co-authored-by: Yaowei Zheng <hiyouga@buaa.edu.cn>

src/llamafactory/v1/core/utils/rendering.py

@@ -91,7 +91,11 @@ class Renderer:
         self.processor = processor
 
     def render_messages(
-        self, messages: list[Message], tools: str | None = None, is_generate: bool = False
+        self,
+        messages: list[Message],
+        tools: str | None = None,
+        is_generate: bool = False,
+        enable_thinking: bool = False,
     ) -> ModelInput:
         """Apply template to messages and convert them to model input.
 
@@ -99,6 +103,7 @@ class Renderer:
             messages (list[Message]): The messages to render.
             tools (str | None, optional): The tools to use. Defaults to None.
             is_generate (bool, optional): Whether to render for generation. Defaults to False.
+            enable_thinking (bool, optional): Whether to enable thinking mode for generation. Defaults to False.
 
         Returns:
             ModelInput: The rendered model input.
@@ -108,7 +113,9 @@ class Renderer:
         else:
             from ...plugins.model_plugins.rendering import RenderingPlugin
 
-            return RenderingPlugin(self.template).render_messages(self.processor, messages, tools, is_generate)
+            return RenderingPlugin(self.template).render_messages(
+                self.processor, messages, tools, is_generate, enable_thinking
+            )
 
     def parse_message(self, generated_text: str) -> Message:
         """Parse a message in the template format.