Update app.py

app.py

@@ -1,4 +1,3 @@
-
 import huggingface_hub as hf_hub
 import time
 import openvino_genai as ov_genai
@@ -20,14 +19,22 @@ tokenizer.set_chat_template(tokenizer.chat_template)
 
 
 def generate_response(prompt):
+    full_response = ""  # 用於儲存完整的回應
+
+    def streamer(subword):
+        nonlocal full_response
+        full_response += subword
+        return full_response  # 返回更新後的完整回應
+
     try:
-        generated = pipe.generate([prompt], max_length=1024)
-        tokenpersec=f'{generated.perf_metrics.get_throughput().mean:.2f}'
+        # 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
 
-        return tokenpersec, generated
+        return tokenpersec, full_response
     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
 demo = gr.Interface(