diff --git a/paddle/fluid/framework/parallel_executor.cc b/paddle/fluid/framework/parallel_executor.cc
index c8dd3f9151d217f2f279f6c08661e70bbfa50d3e..1e1a5477a03f4f69025b2aa058ac9a137417ebca 100644
--- a/paddle/fluid/framework/parallel_executor.cc
+++ b/paddle/fluid/framework/parallel_executor.cc
@@ -172,6 +172,10 @@ struct FetchOpHandle : public OpHandle {
   }
 
   void Run() override {
+    for (auto *input : inputs_) {
+      input->generated_op_->Wait(nullptr);
+    }
+
     tensors_.resize(inputs_.size());
     auto *var = static_cast<VarHandle *>(inputs_[0]);
     auto &var_name = var->name_;