diff --git a/vllm_ascend/worker/model_runner_v1.py b/vllm_ascend/worker/model_runner_v1.py
index 14f89bb5d6c..ebf76ebff99 100644
--- a/vllm_ascend/worker/model_runner_v1.py
+++ b/vllm_ascend/worker/model_runner_v1.py
@@ -1636,7 +1636,7 @@ def execute_model(
              finished_recving) = (self._process_reqs(scheduler_output,
                                                      intermediate_tensors))
         kv_connector_output = None
-        if finished_sending is not None and finished_recving is not None:
+        if finished_sending is not None or finished_recving is not None:
             kv_connector_output = KVConnectorOutput(
                 finished_sending=finished_sending,
                 finished_recving=finished_recving)
@@ -1838,8 +1838,9 @@ def kv_connector_no_forward(
             return EMPTY_MODEL_RUNNER_OUTPUT
 
         output = copy.copy(EMPTY_MODEL_RUNNER_OUTPUT)
-        output.finished_sending = finished_sending
-        output.finished_recving = finished_recving
+        output.kv_connector_output = KVConnectorOutput(
+            finished_sending=finished_sending,
+            finished_recving=finished_recving)
         return output
 
     @staticmethod