Check.

app.py

@@ -69,7 +69,9 @@ You are a medical assistant trained on clinical reasoning data. Given the follow
     return prompt
 
 def generate_local_answer(prompt, max_new_tokens=512):
-    input_ids = tokenizer(prompt, return_tensors="pt").input_ids.cuda()
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    print(f"Using device: {device}")
+    input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
     output = generation_model.generate(
         input_ids=input_ids,
         max_new_tokens=max_new_tokens,