Merge pull request #1439 from kvcache-ai/qiyuxinlin-patch-3

Update balance_serve.py

Author: qiyuxinlin

ktransformers/server/backend/interfaces/balance_serve.py

@@ -445,7 +445,7 @@ async def inference(self, local_messages, thread_id: str, temperature: Optional[
         query_add.query_length = query_length
         profiler.set_counter("prefill", query_length)
         #@TODO add server
-        stop_criteria =  [self.tokenizer.encode(self.tokenizer.eos_token, add_special_tokens=False),self.tokenizer.encode("<|im_end|>")]
+        stop_criteria =  [self.tokenizer.encode(self.tokenizer.eos_token, add_special_tokens=False),self.tokenizer.encode("<|im_end|>", add_special_tokens=True)]
         query_add.stop_criteria = stop_criteria
 
         temperature, top_p, max_new_tokens = self.get_params(temperature, top_p, max_tokens, max_completion_tokens)