diff --git a/fastchat/conversation.py b/fastchat/conversation.py
index a01b9d50c..7bd836a28 100644
--- a/fastchat/conversation.py
+++ b/fastchat/conversation.py
@@ -535,7 +535,7 @@ def get_conv_template(name: str) -> Conversation:
         roles=("### Instruction:", "### Response:"),
         sep="\n",
         stop_str="<|EOT|>",
-        sep_style=SeparatorStyle.ADD_NEW_LINE_SINGLE
+        sep_style=SeparatorStyle.ADD_NEW_LINE_SINGLE,
     )
 )
 
diff --git a/fastchat/serve/inference.py b/fastchat/serve/inference.py
index 8a6636372..6d155aab7 100644
--- a/fastchat/serve/inference.py
+++ b/fastchat/serve/inference.py
@@ -118,6 +118,7 @@ def generate_stream(
     token_logprobs = [None]  # The first token has no logprobs.
     sent_interrupt = False
     finish_reason = None
+    stopped = False
     for i in range(max_new_tokens):
         if i == 0:  # prefill
             if model.config.is_encoder_decoder: