add .to(device) back

app.py

@@ -31,7 +31,7 @@ Your output should be in JSON FORMAT with the keys "REASONING" and "SCORE":
 @spaces.GPU()
 def model_call(question, document, answer):
     NEW_FORMAT = PROMPT.format(question=question, document=document, answer=answer)
-    inputs = tokenizer(NEW_FORMAT, return_tensors="pt")
+    inputs = tokenizer(NEW_FORMAT, return_tensors="pt").to(device)
     model.generate(
         inputs.input_ids,
         attention_mask=inputs.attention_mask,