diff --git a/paddle/fluid/pir/dialect/operator/ir/control_flow_op.cc b/paddle/fluid/pir/dialect/operator/ir/control_flow_op.cc
index 8ae93ea0cb0d51..bdf258f4ee3f1b 100644
--- a/paddle/fluid/pir/dialect/operator/ir/control_flow_op.cc
+++ b/paddle/fluid/pir/dialect/operator/ir/control_flow_op.cc
@@ -254,16 +254,16 @@ void WhileOp::Print(pir::IrPrinter &printer) {
   auto &os = printer.os;
   auto op = operation();
   printer.PrintOpResult(op);
-  os << " = \"" << name() << "\"(";
+  os << " = \"" << name() << "\"(cond=";
   printer.PrintValue(cond());
-  os << ") [";
+  os << ", inputs=";
   auto operands = (*this)->operands_source();
   pir::PrintInterleave(
       operands.begin() + 1,
       operands.end(),
       [&](pir::Value v) { printer.PrintValue(v); },
       [&]() { os << ", "; });
-  os << "] { \n ^";
+  os << ") { \n ^";
   pir::PrintInterleave(
       body().args_begin(),
       body().args_end(),
diff --git a/python/paddle/base/executor.py b/python/paddle/base/executor.py
index c86ca29d85a50c..9de9e4f074a208 100755
--- a/python/paddle/base/executor.py
+++ b/python/paddle/base/executor.py
@@ -526,7 +526,10 @@ def _add_pir_fetch_ops(program, fetch_list, fetch_var_name):
                 assert isinstance(
                     fetch_input, (OpResult, Value)
                 ), f"Wrong type for fetch_list[{i}]: {type(fetch_input)}"
-                paddle._pir_ops.fetch(fetch_input, fetch_var_name + str(i), i)
+                out = paddle._pir_ops.fetch(
+                    fetch_input, fetch_var_name + str(i), i
+                )
+                out.persistable = True
 
 
 def _merge_tensors(tensor, micro_batch_num):
diff --git a/test/cpp/pir/control_flow_dialect/while_op_test.cc b/test/cpp/pir/control_flow_dialect/while_op_test.cc
index dbcdc6d361ec20..55291c2ccbc1b4 100644
--- a/test/cpp/pir/control_flow_dialect/while_op_test.cc
+++ b/test/cpp/pir/control_flow_dialect/while_op_test.cc
@@ -185,6 +185,9 @@ TEST(while_op_test, network_with_backward) {
   LOG(INFO) << program;
 
   auto place = paddle::platform::CPUPlace();
+#ifdef PADDLE_WITH_CUDA
+  place = paddle::platform::CUDAPlace(0);
+#endif
   auto kernel_program = paddle::dialect::PdOpLowerToKernelPass(&program, place);
   paddle::framework::Scope scope;
   paddle::framework::InterpreterCore test_core(