Update app.py

app.py

@@ -1,3 +1,4 @@
+
 import huggingface_hub as hf_hub
 import time
 import openvino_genai as ov_genai
@@ -35,13 +36,10 @@ tokenizer.set_chat_template(tokenizer.chat_template)
 config = ov_genai.GenerationConfig()
 config.max_new_tokens = 1024
 
-# 修改 streamer 函數，使其 yield 結果
-def streamer():
-    def cb(subword):
-        yield subword
-    return cb
+def streamer(subword):
+    print(subword, end='', flush=True)
+    return False
     
-# 修改 generate_response 函數，使用 yield 傳回結果，並傳回tokenpersec
 def generate_response(prompt, model_name):
     global pipe, tokenizer, InUsed_model_name
     if(InUsed_model_name!=model_name):
@@ -55,17 +53,12 @@ def generate_response(prompt, model_name):
         InUsed_model_name=model_name
 
     try:
-        response = ""
-        callback = streamer()
-        for token in pipe.generate([prompt], config, callback):
-            response += token
-            yield '', response
-
-        tokenpersec = f'{pipe.last_metrics.get_throughput().mean:.2f}'
-        yield tokenpersec, response  # 傳回最終結果
-
+        generated = pipe.generate([prompt], config, streamer)
+        tokenpersec=f'{generated.perf_metrics.get_throughput().mean:.2f}'
+        print(f"tokenpersec:{tokenpersec}\n")
+        return tokenpersec, generated
     except Exception as e:
-        yield "發生錯誤", f"生成回應時發生錯誤：{e}"
+        return "發生錯誤", "發生錯誤", f"生成回應時發生錯誤：{e}"
     
 
 # 建立 Gradio 介面
@@ -82,9 +75,8 @@ demo = gr.Interface(
         gr.Textbox(label="回應"),
     ],
     title="Qwen3 Model Inference",
-    description="基於 Qwen3 推理應用，支援思考過程分離與 GUI。",
-    live=True  # 啟用即時更新
+    description="基於 Qwen3 推理應用，支援思考過程分離與 GUI。"
 )
 
 if __name__ == "__main__":
-    demo.queue().launch()
+    demo.launch()