Update app.py

app.py

@@ -78,20 +78,27 @@ def predict(message, history):
         repetition_penalty=1.1,
     )
 
+    outputs = model.generate(**generate_kwargs)
+    generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    partial_message = generated_text
+    final_message = partial_message.replace(sft_end_token, "").strip()
+
+    return final_message
+
     # # 使用线程来运行生成过程
     # t = Thread(target=model.generate, kwargs=generate_kwargs)
     # t.start()
 
-    # 实时生成部分消息
-    partial_message = ""
-    for new_token in streamer:
-        partial_message += new_token
-        if sft_end_token in partial_message:  # 检测到停止标志
-            break
-        # 将历史记录和当前消息转换为 tuple 格式并实时返回
-        # yield [(msg, bot) for msg, bot in history] + [(message, partial_message)]
-        # yield (message, partial_message)
-        yield partial_message
+    # # 实时生成部分消息
+    # partial_message = ""
+    # for new_token in streamer:
+    #     partial_message += new_token
+    #     if sft_end_token in partial_message:  # 检测到停止标志
+    #         break
+    #     # 将历史记录和当前消息转换为 tuple 格式并实时返回
+    #     # yield [(msg, bot) for msg, bot in history] + [(message, partial_message)]
+    #     # yield (message, partial_message)
+    #     yield partial_message
     
     # # 处理生成的最终回复
     # final_message = partial_message.replace(sft_end_token, "").strip()