Update app.py

app.py

@@ -18,12 +18,12 @@ model_single = model_single.to("cuda")
 
 generation_config_single = model_single.default_generation_config
 
-# model_think = PeftModel.from_pretrained(
-#             model_single,
-#             MODEL_BASE_THINK,
-#             device_map="auto",
-#             torch_dtype=torch.float16,
-#         )
+model_think = PeftModel.from_pretrained(
+        model_single,
+           MODEL_BASE_THINK,
+device_map="auto",
+          torch_dtype=torch.float16,
+        )
 
 # # ---------------------------------
 # # MULTI-TURN MODEL SETUP
@@ -58,7 +58,7 @@ def think_infer(audio_file, prompt_text):
     try:
         sound = llava.Sound(audio_file)
         full_prompt = f"<sound>\n{prompt_text}"
-        response = model_single.generate_content([sound, full_prompt], generation_config=generation_config_single)
+        response = model_think.generate_content([sound, full_prompt], generation_config=generation_config_single)
         return response
     except Exception as e:
         return f"❌ Error: {str(e)}"