diff --git a/paddle/fluid/framework/parallel_executor.cc b/paddle/fluid/framework/parallel_executor.cc
index aa52cbb7bf21b5168e59c3d1a78def1e92adbee5..b86909766298026fa0a877521edc510936deac84 100644
--- a/paddle/fluid/framework/parallel_executor.cc
+++ b/paddle/fluid/framework/parallel_executor.cc
@@ -752,7 +752,7 @@ void ParallelExecutor::Run(const std::vector<std::string> &fetch_tensors,
   while (!pending_vars.empty()) {
     VarHandleBase *ready_var = nullptr;
     for (auto &pair : pending_vars) {
-      if (pair.second.load(std::memory_order_consume)) {
+      if (pair.second.load(std::memory_order_acquire)) {
         ready_var = pair.first;
       }
     }