diff --git a/vllm_ascend/core/recompute_scheduler.py b/vllm_ascend/core/recompute_scheduler.py
index 48aa67a272e..356583a3edf 100644
--- a/vllm_ascend/core/recompute_scheduler.py
+++ b/vllm_ascend/core/recompute_scheduler.py
@@ -235,10 +235,10 @@ def schedule(self) -> SchedulerOutput:
                                 if preempted_encoder_inputs:
                                     # Restore encoder compute budget if the preempted
                                     # request had encoder inputs scheduled in this step.
-                                    num_tokens_to_restore = sum(
-                                        preempted_req.get_num_encoder_tokens(i)
+                                    num_embeds_to_restore = sum(
+                                        preempted_req.get_num_encoder_embeds(i)
                                         for i in preempted_encoder_inputs)
-                                    encoder_compute_budget += num_tokens_to_restore
+                                    encoder_compute_budget += num_embeds_to_restore
                                 req_index -= 1
                         else:
                             preempted_req = self.running.pop()