diff --git a/paddle/fluid/operators/array_to_lod_tensor_op.cc b/paddle/fluid/operators/array_to_lod_tensor_op.cc
index 1780d025d9c08052644c41ee4e973d10a94b8ab1..ef55b38b876a6006e2912669b6a7ede3e0e4df02 100644
--- a/paddle/fluid/operators/array_to_lod_tensor_op.cc
+++ b/paddle/fluid/operators/array_to_lod_tensor_op.cc
@@ -49,7 +49,7 @@ struct ArrayToLoDFunctor : public boost::static_visitor<void> {
       Apply(static_cast<platform::CUDADeviceContext *>(pool.Get(place)));
 #else
       PADDLE_THROW(
-          platform::errors::Unavailable("Paddle is not compiled with CUDA"));
+          platform::errors::Unavailable("Paddle is not compiled with CUDA."));
 #endif
     }
   }
diff --git a/paddle/fluid/operators/lod_tensor_to_array_op.cc b/paddle/fluid/operators/lod_tensor_to_array_op.cc
index 22e8e00dcce03eb5d208bf229be08a1b1e10c11c..8e07ffabb546a71197880d81ad53e218bfe12126 100644
--- a/paddle/fluid/operators/lod_tensor_to_array_op.cc
+++ b/paddle/fluid/operators/lod_tensor_to_array_op.cc
@@ -62,7 +62,7 @@ struct LoDTensorToArrayFunctor : public boost::static_visitor<void> {
       Apply(static_cast<platform::CUDADeviceContext *>(dev_ctx));
 #else
       PADDLE_THROW(
-          platform::errors::Unavailable("Place is not compiled with CUDA"));
+          platform::errors::Unavailable("Paddle is not compiled with CUDA."));
 #endif
     }
   }