diff --git a/paddle/fluid/framework/parallel_executor.cc b/paddle/fluid/framework/parallel_executor.cc
index 07dfddfa305fa5eab8a29f10bc6d6362a6a8e826..d0c4d8dd8b3e15d9aef9065880664031dfe53499 100644
--- a/paddle/fluid/framework/parallel_executor.cc
+++ b/paddle/fluid/framework/parallel_executor.cc
@@ -211,9 +211,6 @@ struct FetchOpHandle : public OpHandle {
     for (auto *input_var : inputs_) {
       input_var->pending_ops_.erase(this);
     }
-    for (auto &pair : dev_ctx_) {
-      pair.second->Wait();
-    }
 
     // Lazily merge tensors. Will faster code.
     MergeTensors();