diff --git a/neurolib/models/multimodel/model.py b/neurolib/models/multimodel/model.py
index d8948b53..cc000de9 100644
--- a/neurolib/models/multimodel/model.py
+++ b/neurolib/models/multimodel/model.py
@@ -202,8 +202,11 @@ def integrate(self, append_outputs=False, simulate_bold=False, noise_input=None)
             self.simulateBold(result[self.default_output].values.T, append=True)
 
     def setInitialValuesToLastState(self):
-        if not hasattr(self, "t"):
-            raise ValueError("You tried using continue_run=True on the first run.")
+        if not self.state:
+            logging.warning(
+                f"`setInitialValuesToLastState` was called, but the state dict is empty. This is probably caused by `continue_run=True` on the first model run."
+            )
+            return
         new_initial_state = np.zeros((self.model_instance.initial_state.shape[0], self.maxDelay + 1))
         total_vars_counter = 0
         for node_idx, node_vars in enumerate(self.state_vars):