diff --git a/mindspore/lite/internal/src/kernel/fp32/matmul.cc b/mindspore/lite/internal/src/kernel/fp32/matmul.cc
index 5f79194b92..2e5c1e1795 100644
--- a/mindspore/lite/internal/src/kernel/fp32/matmul.cc
+++ b/mindspore/lite/internal/src/kernel/fp32/matmul.cc
@@ -116,7 +116,7 @@ int DoMatMulInferShape(const TensorPtrVector &in_tensors, const TensorPtrVector
 
 int DoMatMul(const TensorPtrVector &in_tensors, const TensorPtrVector &out_tensors, Node *node,
              mindspore::lite::Allocator *allocator) {
-  if (in_tensors[0]->data_ == NULL || in_tensors[1]->data_ ==NULL) {
+  if (in_tensors[0]->data_ == NULL || in_tensors[1]->data_ == NULL) {
     LITE_LOG_ERROR("input data is NULL!");
     return RET_PARAM_INVALID;
   }
diff --git a/mindspore/lite/nnacl/fp32/batchnorm.c b/mindspore/lite/nnacl/fp32/batchnorm.c
index 2ed427140e..3c269c8ccc 100644
--- a/mindspore/lite/nnacl/fp32/batchnorm.c
+++ b/mindspore/lite/nnacl/fp32/batchnorm.c
@@ -68,7 +68,7 @@ void FusedBatchNormFp32MeanVar(const float *input, float momentum, float *run_me
     run_mean[f] = run_mean[f] / N;
     run_var[f] = run_var[f] / N - run_mean[f] * run_mean[f];
     save_mean[f] = momentum * save_mean[f] + (1 - momentum) * run_mean[f];
-    float inv_var = 1.f/sqrt(run_var[f]+param->epsilon_);
+    float inv_var = 1.f / sqrt(run_var[f] + param->epsilon_);
     save_inv_var[f] = momentum * save_inv_var[f] + (1 - momentum) * inv_var;
   }
 }
diff --git a/mindspore/lite/nnacl/fp32/batchnorm.h b/mindspore/lite/nnacl/fp32/batchnorm.h
index 6defc90b7e..fa071425a3 100644
--- a/mindspore/lite/nnacl/fp32/batchnorm.h
+++ b/mindspore/lite/nnacl/fp32/batchnorm.h
@@ -29,7 +29,7 @@ void FusedBatchNormFp32(const void *input, const void *scale, const void *offset
                         const void *variance, BatchNormParameter *param, int task_id, void *output);
 
 void FusedBatchNormFp32MeanVar(const float *input, float momentum, float *run_mean, float *run_var,
-                              BatchNormParameter *param, float *save_mean, float *save_var);
+                               BatchNormParameter *param, float *save_mean, float *save_var);
 #ifdef __cplusplus
 }
 #endif
diff --git a/mindspore/lite/nnacl/fp32_grad/softmax_grad.h b/mindspore/lite/nnacl/fp32_grad/softmax_grad.h
index ad6f4fc8b2..44da4864d5 100644
--- a/mindspore/lite/nnacl/fp32_grad/softmax_grad.h
+++ b/mindspore/lite/nnacl/fp32_grad/softmax_grad.h
@@ -33,7 +33,7 @@ typedef struct SoftmaxCrossEntropyParameter {
 } SoftmaxCrossEntropyParameter;
 
 void SoftmaxGrad(const float *input_ptr, const float *yt_ptr, float *output_ptr, float *sum_data,
-                float *sum_mul, SoftmaxParameter *parameter);
+                 float *sum_mul, SoftmaxParameter *parameter);
 #ifdef __cplusplus
 }
 #endif
diff --git a/mindspore/lite/nnacl/int8/space_to_batch_int8.c b/mindspore/lite/nnacl/int8/space_to_batch_int8.c
index f86049d730..8dcd0e5365 100644
--- a/mindspore/lite/nnacl/int8/space_to_batch_int8.c
+++ b/mindspore/lite/nnacl/int8/space_to_batch_int8.c
@@ -17,7 +17,7 @@
 #include "nnacl/arithmetic_common.h"
 
 void DoSpaceToBatchNHWCInt8(const int8_t *input, int8_t *output, int *block_sizes, int *in_shape,
-                        int *out_shape) {
+                            int *out_shape) {
   int out_dim0 = out_shape[0];
   int out_dim1 = out_shape[1];
   int out_dim2 = out_shape[2];
diff --git a/mindspore/lite/src/CMakeLists.txt b/mindspore/lite/src/CMakeLists.txt
index 5cb494cfaa..e511cbec73 100644
--- a/mindspore/lite/src/CMakeLists.txt
+++ b/mindspore/lite/src/CMakeLists.txt
@@ -95,3 +95,14 @@ if ("${CMAKE_BUILD_TYPE}" STREQUAL "Release" AND (PLATFORM_ARM64 OR PLATFORM_ARM
       ${TOP_DIR}/mindspore/lite/build/src/libmindspore-lite.so)
 endif ()
 
+if ("${CMAKE_BUILD_TYPE}" STREQUAL "Release")
+  if (PLATFORM_ARM64 OR PLATFORM_ARM32)
+    add_custom_command(TARGET mindspore-lite POST_BUILD
+        COMMAND ${ANDROID_NDK}/toolchains/aarch64-linux-android-4.9/prebuilt/linux-x86_64/aarch64-linux-android/bin/strip
+        ${TOP_DIR}/mindspore/lite/build/src/libmindspore-lite.so)
+  elseif (NOT WIN32)
+    add_custom_command(TARGET mindspore-lite POST_BUILD
+        COMMAND strip ${TOP_DIR}/mindspore/lite/build/src/libmindspore-lite.so)
+  endif ()
+endif ()
+
diff --git a/mindspore/lite/src/ops/bias_grad.cc b/mindspore/lite/src/ops/bias_grad.cc
index cee3f9cb27..d561e42503 100644
--- a/mindspore/lite/src/ops/bias_grad.cc
+++ b/mindspore/lite/src/ops/bias_grad.cc
@@ -100,6 +100,5 @@ int BiasGrad::InferShape(std::vector<Tensor *> inputs, std::vector<Tensor *> out
 
   return RET_OK;
 }
-
 }  // namespace lite
 }  // namespace mindspore
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32_grad/sparse_softmax_cross_entropy_with_logits.h b/mindspore/lite/src/runtime/kernel/arm/fp32_grad/sparse_softmax_cross_entropy_with_logits.h
index 19d12ea8fb..3d00c0fffe 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32_grad/sparse_softmax_cross_entropy_with_logits.h
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32_grad/sparse_softmax_cross_entropy_with_logits.h
@@ -33,7 +33,7 @@ class SparseSoftmaxCrossEntropyWithLogitsCPUKernel : public LossKernel {
                                                         const std::vector<lite::Tensor *> &outputs,
                                                         const lite::Context *ctx,
                                                         const mindspore::lite::PrimitiveC *primitive)
-      : LossKernel(parameter, inputs, outputs, ctx, primitive) , losses_(nullptr), sum_data_(nullptr)  {
+      : LossKernel(parameter, inputs, outputs, ctx, primitive), losses_(nullptr), sum_data_(nullptr)  {
     param = reinterpret_cast<SoftmaxCrossEntropyParameter *>(parameter);
   }
   ~SparseSoftmaxCrossEntropyWithLogitsCPUKernel() override {
diff --git a/mindspore/lite/src/tensor.h b/mindspore/lite/src/tensor.h
index 0bbcdb7fb1..0fa96d3c03 100644
--- a/mindspore/lite/src/tensor.h
+++ b/mindspore/lite/src/tensor.h
@@ -30,7 +30,6 @@
 
 namespace mindspore {
 namespace lite {
-
 struct QuantArg {
   double scale;
   int32_t zeroPoint;