Update app.py

app.py

@@ -27,17 +27,20 @@ for model_id in model_ids:
 
 # 建立推理管線 (Initialize with a default model first)
 device = "CPU"
-default_model_name = "Qwen3-0.6B-int4-ov"  # Choose a default model
+InUsed_model_name = "Qwen3-0.6B-int4-ov"  # Choose a default model
+pipe = ov_genai.LLMPipeline(InUsed_model_name, device)
 
 def generate_response(prompt, model_name):
     global pipe, tokenizer  # Access the global variables
-
-    model_path = model_name
-
-    print(f"Switching to model: {model_name}")
-    pipe = ov_genai.LLMPipeline(model_path, device)
-    tokenizer = pipe.get_tokenizer()
-    tokenizer.set_chat_template(tokenizer.chat_template)
+    if(InUsed_model_name==model_name)
+        model_path = model_name
+        del pipe
+        gc.collect()
+        print(f"Switching to model: {model_name}")
+        pipe = ov_genai.LLMPipeline(model_path, device)
+        tokenizer = pipe.get_tokenizer()
+        tokenizer.set_chat_template(tokenizer.chat_template)
+        InUsed_model_name=model_name
 
     try:
         generated = pipe.generate([prompt], max_length=1024)