wip

Signed-off-by: Raphael Glon <oOraph@users.noreply.github.com>

app.py

@@ -1,3 +1,5 @@
+# Copied/Adapted from https://huggingface.co/spaces/akhaliq/MobileLLM-Pro
+
 import spaces
 
 import logging
@@ -55,7 +57,7 @@ def _ensure_loaded():
     _device = next(_model.parameters()).device
 
 
-_ensure_loaded()
+# _ensure_loaded()
 
 LOG.info("DEVICE %s", _device)
 
@@ -76,6 +78,8 @@ def generate_stream(message: str, history: List[Tuple[str, str]]):
     Minimal streaming chat function for gr.ChatInterface.
     Uses instruct chat template. No token UI. No extra controls.
     """
+
+    # TODO: check the memory footprint doing so. We should rather do this before the spaces wrapper...
     _ensure_loaded()
 
     messages = _history_to_messages(history) + [{"role": "user", "content": message}]