diff --git a/zenu-macros/tests/include_vector_map_model.rs b/zenu-macros/tests/include_vector_map_model.rs
index d805ecf6..ae5d4199 100644
--- a/zenu-macros/tests/include_vector_map_model.rs
+++ b/zenu-macros/tests/include_vector_map_model.rs
@@ -1,12 +1,9 @@
 use std::collections::HashMap;
 
 use rand_distr::{Distribution, StandardNormal};
-use zenu_layer::{
-    layers::{conv2d::Conv2d, linear::Linear, max_pool_2d::MaxPool2d},
-    Parameters,
-};
-use zenu_macros::Parameters;
-use zenu_matrix::{
+use zenu::layer::layers::{conv2d::Conv2d, linear::Linear, max_pool_2d::MaxPool2d};
+use zenu::macros::Parameters;
+use zenu::matrix::{
     device::{cpu::Cpu, Device},
     num::Num,
 };
@@ -70,6 +67,7 @@ impl<T: Num, D: Device> ConvNet<T, D> {
 
 #[test]
 fn vec_map() {
+    use zenu::layer::Parameters;
     let model = ConvNet::<f32, Cpu>::new();
     let conv_fileter_0 = model.conv_blocks[0].conv2d.filter.clone();
     let conv_bias_0 = model.conv_blocks[0].conv2d.bias.clone();
diff --git a/zenu-macros/tests/multi_parameter_struct.rs b/zenu-macros/tests/multi_parameter_struct.rs
index 9e51eb45..7ff23886 100644
--- a/zenu-macros/tests/multi_parameter_struct.rs
+++ b/zenu-macros/tests/multi_parameter_struct.rs
@@ -1,10 +1,7 @@
 use rand_distr::{Distribution, StandardNormal};
-use zenu_layer::{
-    layers::{conv2d::Conv2d, linear::Linear, max_pool_2d::MaxPool2d},
-    Parameters,
-};
-use zenu_macros::Parameters;
-use zenu_matrix::{
+use zenu::layer::layers::{conv2d::Conv2d, linear::Linear, max_pool_2d::MaxPool2d};
+use zenu::macros::Parameters;
+use zenu::matrix::{
     device::{cpu::Cpu, Device},
     num::Num,
 };
@@ -49,6 +46,7 @@ impl<T: Num, D: Device> ConvNet<T, D> {
 
 #[test]
 fn multi_params() {
+    use zenu::layer::Parameters;
     let model = ConvNet::<f32, Cpu>::new();
     let conv_fileter = model.conv_block.conv2d.filter.clone();
     let conv_bias = model.conv_block.conv2d.bias.clone();
@@ -88,6 +86,7 @@ fn multi_params() {
 
 #[test]
 fn test_load_parameters_convnet() {
+    use zenu::layer::Parameters;
     let model = ConvNet::<f32, Cpu>::new();
     let parameters = model.parameters();
 
diff --git a/zenu-macros/tests/small_case.rs b/zenu-macros/tests/small_case.rs
index fb2017db..32a9d3e2 100644
--- a/zenu-macros/tests/small_case.rs
+++ b/zenu-macros/tests/small_case.rs
@@ -1,7 +1,13 @@
 use zenu::layer::layers::linear::Linear;
+use zenu::macros::Parameters as ParametersDerive;
+use zenu::matrix::{
+    device::{cpu::Cpu, Device},
+    matrix::Matrix,
+    num::Num,
+};
 use zenu_test::assert_val_eq;
 
-#[derive(Parameters)]
+#[derive(ParametersDerive)]
 #[parameters(num = T, device = D)]
 pub struct Hoge<T, D>
 where
@@ -13,6 +19,7 @@ where
 
 #[test]
 fn small_net() {
+    use zenu::layer::Parameters;
     let hoge = Hoge::<f32, Cpu> {
         linear: Linear::new(2, 2, true),
     };
@@ -50,6 +57,7 @@ fn small_net() {
 
 #[test]
 fn test_load_parameters() {
+    use zenu::layer::Parameters;
     let base_model = Hoge::<f32, Cpu> {
         linear: Linear::new(2, 2, true),
     };
diff --git a/zenu/src/lib.rs b/zenu/src/lib.rs
index bf1fb990..42f098ea 100644
--- a/zenu/src/lib.rs
+++ b/zenu/src/lib.rs
@@ -15,7 +15,7 @@ use zenu_layer::Parameters;
 use zenu_matrix::{device::Device, num::Num};
 use zenu_optimizer::Optimizer;
 
-extern crate zenu_macros;
+pub extern crate zenu_macros;
 
 pub use zenu_autograd as autograd;
 pub use zenu_layer as layer;