[Not finished] Allow customized parameter grouping for automatic optimzier configuration.

src/lightning/pytorch/cli.py

@@ -685,13 +685,17 @@ def get_automatic(
                 f"`{self.__class__.__name__}.configure_optimizers`."
             )
 
-        optimizer = instantiate_class(self.model.parameters(), optimizer_init)
+        optimizer = instantiate_class(self._get_model_parameters(), optimizer_init)
         lr_scheduler = instantiate_class(optimizer, lr_scheduler_init) if lr_scheduler_init else None
         fn = partial(self.configure_optimizers, optimizer=optimizer, lr_scheduler=lr_scheduler)
         update_wrapper(fn, self.configure_optimizers)  # necessary for `is_overridden`
         # override the existing method
         self.model.configure_optimizers = MethodType(fn, self.model)
 
+    def _get_model_parameters(self):
+        assert hasattr(self,'model'), "model not instantiated yet. You have to instantiate a model object in order to access its parameters"
+        return self.model.parameters()
+
     def _get(self, config: Namespace, key: str, default: Optional[Any] = None) -> Any:
         """Utility to get a config value which might be inside a subcommand."""
         return config.get(str(self.subcommand), config).get(key, default)