Adapt the old parameter enable_thinking in chat_template_kwargs (#30852)

Signed-off-by: xinsong.gao <1418762819@qq.com>
Co-authored-by: Chauncey <chaunceyjiang@gmail.com>

vllm/reasoning/deepseek_v3_reasoning_parser.py

@@ -26,6 +26,8 @@ class DeepSeekV3ReasoningParser(ReasoningParser):
 
         chat_kwargs = kwargs.pop("chat_template_kwargs", {}) or {}
         thinking = bool(chat_kwargs.pop("thinking", False))
+        enable_thinking = bool(chat_kwargs.pop("enable_thinking", False))
+        thinking = thinking or enable_thinking
 
         if thinking:
             self._parser = DeepSeekR1ReasoningParser(tokenizer, *args, **kwargs)

vllm/tokenizers/deepseek_v32.py

@@ -50,6 +50,8 @@ class DeepseekV32Tokenizer(CachedHfTokenizer):
         **kwargs,
     ) -> str | list[int]:
         thinking = kwargs.get("thinking", False)
+        enable_thinking = kwargs.get("enable_thinking", False)
+        thinking = thinking or enable_thinking
         thinking_mode = "thinking"
         if not thinking:
             thinking_mode = "chat"