Update model.py

model.py

@@ -19,9 +19,11 @@ def run(message: str,
         top_p: float = 0.95,
         top_k: int = 50) -> Iterator[str]:
     history = []
+    result=""
     for i in chat_history:
         history.append({"role": "user", "content": i[0]})
         history.append({"role": "assistant", "content": i[1]})
     for response in llm.create_chat_completion(history,stream=True,max_tokens=-1,temperature=0.3,top_k=5,top_p=0.85,repeat_penalty=1.1):
         if "content" in response["choices"][0]["delta"]:
-            yield response["choices"][0]["delta"]["content"]
+            result = result + response["choices"][0]["delta"]["content"]
+            yield result