diff --git a/src/lerobot/policies/multi_task_dit/modeling_multi_task_dit.py b/src/lerobot/policies/multi_task_dit/modeling_multi_task_dit.py
index c43782b36..5ae7d8480 100644
--- a/src/lerobot/policies/multi_task_dit/modeling_multi_task_dit.py
+++ b/src/lerobot/policies/multi_task_dit/modeling_multi_task_dit.py
@@ -755,9 +755,6 @@ class MultiTaskDiTPolicy(PreTrainedPolicy):
         if self.config.image_features:
             self._queues["observation.images"] = deque(maxlen=self.config.n_obs_steps)
 
-        if self.config.env_state_feature:
-            self._queues["observation.environment_state"] = deque(maxlen=self.config.n_obs_steps)
-
         self._queues["task"] = deque(maxlen=self.config.n_obs_steps)
 
     @torch.no_grad()