Update llm.py

llm.py

@@ -16,6 +16,14 @@ model = AutoModelForImageTextToText.from_pretrained(
     device_map="cuda"
 )
 
+print("CUDA available:", torch.cuda.is_available())
+print("Model device:", model.device)
+    
+if torch.cuda.is_available():
+    print("GPU name:", torch.cuda.get_device_name(0))
+    print("Memory allocated:", torch.cuda.memory_allocated() / 1e9, "GB")
+    
+print("Device map:", model.hf_device_map)
 # ✅ SYNC function (runs in thread)
 def execute_llm(model, processor, image, prompt: str):
     print("execute llm")