diff --git a/paddle/fluid/lite/api/paddle_use_ops.h b/paddle/fluid/lite/api/paddle_use_ops.h
index cfb08bd57d4a722a98a5d569d74487f945b6c7a9..47ba27579efd2d5b2442ee948167a05533e5c68e 100644
--- a/paddle/fluid/lite/api/paddle_use_ops.h
+++ b/paddle/fluid/lite/api/paddle_use_ops.h
@@ -36,3 +36,5 @@ USE_LITE_OP(depthwise_conv2d)
 USE_LITE_OP(pool2d)
 USE_LITE_OP(batch_norm)
 USE_LITE_OP(fusion_elementwise_sub_activation)
+USE_LITE_OP(transpose)
+USE_LITE_OP(transpose2)
diff --git a/paddle/fluid/lite/kernels/arm/transpose_compute.cc b/paddle/fluid/lite/kernels/arm/transpose_compute.cc
index 368716c368083ea877fe5dd8a0054a9763a4829e..ff32ac242b9356ac558645952cb51ecbfe351869 100644
--- a/paddle/fluid/lite/kernels/arm/transpose_compute.cc
+++ b/paddle/fluid/lite/kernels/arm/transpose_compute.cc
@@ -170,4 +170,5 @@ REGISTER_LITE_KERNEL(transpose2, kARM, kFloat, kNCHW,
                      paddle::lite::kernels::arm::Transpose2Compute, def)
     .BindInput("X", {LiteType::GetTensorTy(TARGET(kARM))})
     .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kARM))})
+    .BindOutput("XShape", {LiteType::GetTensorTy(TARGET(kARM))})
     .Finalize();