diff --git a/paddle/fluid/framework/details/threaded_ssa_graph_executor.cc b/paddle/fluid/framework/details/threaded_ssa_graph_executor.cc
index 13789667b82d41e8850804d8042660eac0152a58..50c24d3afa849efd3a0ec46c251075435d951653 100644
--- a/paddle/fluid/framework/details/threaded_ssa_graph_executor.cc
+++ b/paddle/fluid/framework/details/threaded_ssa_graph_executor.cc
@@ -199,7 +199,7 @@ void ThreadedSSAGraphExecutor::RunOp(
 
   auto op_run = [ready_buffer, op, this] {
     try {
-      VLOG(10) << op->DebugString();
+      VLOG(10) << op->Name() << " : " << op->DebugString();
       op->Run(use_event_);
 
       for (auto *ready : *ready_buffer) {
@@ -211,6 +211,7 @@ void ThreadedSSAGraphExecutor::RunOp(
     } catch (...) {
       LOG(FATAL) << "Unknown exception catched";
     }
+    PADDLE_ENFORCE(cudaDeviceSynchronize());
   };
   if (pool_) {
     pool_->enqueue(op_run);