Update app.py

app.py

@@ -19,33 +19,26 @@ tokenizer.set_chat_template(tokenizer.chat_template)
 
 
 def generate_response(prompt):
-    full_response = ""
-    tokenpersec = "計算中..."
-    global response_box # 宣告 response_box 為全域變數
-
-    def streamer(subword):
-        nonlocal full_response
-        full_response += subword
-        response_box.update(value=full_response)  # 更新 Textbox 的內容
-        return ov_genai.StreamingStatus.RUNNING
-
     try:
-        # 使用流式生成
-        generated = pipe.generate(prompt, streamer=streamer, max_new_tokens=100)
-        tokenpersec = f'{generated.perf_metrics.get_throughput().mean:.2f}'  # 恢復原本計算 tokenpersec 的方式
+        generated = pipe.generate([prompt], max_length=1024)
+        tokenpersec=f'{generated.perf_metrics.get_throughput().mean:.2f}'
 
-        return tokenpersec, full_response
+        return tokenpersec, generated
     except Exception as 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
-with gr.Blocks() as demo:
-    input_box = gr.Textbox(lines=5, label="輸入提示 (Prompt)")
-    response_box = gr.Textbox(label="回應", streaming=True) # 啟用 streaming
-    speed_box = gr.Textbox(label="tokens/sec")
-
-    input_box.change(fn=generate_response, inputs=input_box, outputs=[speed_box, response_box]) # 將輸入框的改變連接到 generate_response 函數
+demo = gr.Interface(
+    fn=generate_response,
+    inputs=gr.Textbox(lines=1, label="輸入提示 (Prompt)"),
+    outputs=[
+        gr.Textbox(label="tokens/sec"),
+        gr.Textbox(label="回應")
+    ],
+    title="Qwen3-0.6B-int4-ov ",
+    description="基於 Qwen3-0.6B-int4-ov 推理應用，支援思考過程分離與 GUI。"
+)
 
 if __name__ == "__main__":
     demo.launch()