added a lock to avoid two inference at the same time and added consequent support for asynchronous generator based model

source/manager/ModelManager.py

@@ -23,6 +23,9 @@ class ModelManager:
         # TODO(Faraphel): load more than one model at a time ? require a way more complex manager to handle memory issue
         self.current_loaded_model: typing.Optional[model.base.BaseModel] = None
 
+        # lock to avoid concurrent inference and concurrent model loading and unloading
+        self.inference_lock = asyncio.Lock()
+
         @self.application.get("/models")
         async def get_models() -> list[str]:
             """