test footprint

app.py

@@ -72,9 +72,6 @@ def _history_to_messages(history: List[Tuple[str, str]]) -> List[Dict[str, str]]
     return msgs
 
 
-_ensure_loaded()
-
-
 @spaces.GPU(duration=120)
 def generate_stream(message: str, history: List[Tuple[str, str]]):
     """
@@ -83,7 +80,7 @@ def generate_stream(message: str, history: List[Tuple[str, str]]):
     """
 
     # FIXME: check the memory footprint doing so. We should rather do this before the spaces wrapper...
-    # _ensure_loaded()
+    _ensure_loaded()
 
     messages = _history_to_messages(history) + [{"role": "user", "content": message}]
     inputs = _tokenizer.apply_chat_template(