diff --git a/mace/core/mace.cc b/mace/core/mace.cc
index 203bc4723ec714de257175568911cccc358517e7..de6c93534e08bbdbcb71429a5dcddf69c2d50f11 100644
--- a/mace/core/mace.cc
+++ b/mace/core/mace.cc
@@ -10,50 +10,48 @@
 
 namespace mace {
 
-TensorProto::TensorProto(const std::string &name,
+ConstTensor::ConstTensor(const std::string &name,
                          unsigned char *data,
                          const std::vector<int64_t> &dims,
                          const DataType data_type,
                          uint32_t node_id) :
     name_(name),
     data_(data),
-    data_size_(0),
+    data_size_(std::accumulate(dims.begin(), dims.end(), 1,
+                               std::multiplies<int64_t>())),
     dims_(dims.begin(), dims.end()),
     data_type_(data_type),
-    node_id_(node_id) {
-  data_size_ = std::accumulate(dims_.begin(), dims_.end(), 1, std::multiplies<int64_t>());
-}
+    node_id_(node_id) {}
 
-TensorProto::TensorProto(const std::string &name,
+ConstTensor::ConstTensor(const std::string &name,
                          unsigned char *data,
                          const std::vector<int64_t> &dims,
                          const int data_type,
                          uint32_t node_id) :
     name_(name),
     data_(data),
-    data_size_(0),
+    data_size_(std::accumulate(dims.begin(), dims.end(), 1,
+                               std::multiplies<int64_t>())),
     dims_(dims.begin(), dims.end()),
     data_type_(static_cast<DataType>(data_type)),
-    node_id_(node_id) {
-  data_size_ = std::accumulate(dims_.begin(), dims_.end(), 1, std::multiplies<int64_t>());
-}
+    node_id_(node_id) {}
 
-const std::string &TensorProto::name() const {
+const std::string &ConstTensor::name() const {
   return name_;
 }
-unsigned char *TensorProto::data() const {
+const unsigned char *ConstTensor::data() const {
   return data_;
 }
-const int64_t TensorProto::data_size() const {
+int64_t ConstTensor::data_size() const {
   return data_size_;
 }
-const std::vector<int64_t> &TensorProto::dims() const {
+const std::vector<int64_t> &ConstTensor::dims() const {
   return dims_;
 }
-DataType TensorProto::data_type() const {
+DataType ConstTensor::data_type() const {
   return data_type_;
 }
-uint32_t TensorProto::node_id() const {
+uint32_t ConstTensor::node_id() const {
   return node_id_;
 }
 
@@ -446,10 +444,10 @@ Argument *NetDef::add_arg() {
 std::vector<Argument> &NetDef::mutable_arg() {
   return arg_;
 }
-const std::vector<TensorProto> &NetDef::tensors() const {
+const std::vector<ConstTensor> &NetDef::tensors() const {
   return tensors_;
 }
-std::vector<TensorProto> &NetDef::mutable_tensors() {
+std::vector<ConstTensor> &NetDef::mutable_tensors() {
   return tensors_;
 }
 const MemoryArena &NetDef::mem_arena() const {
diff --git a/mace/core/public/mace.h b/mace/core/public/mace.h
index f0f8965692070f5e9516a19d3fc750ab46e0f405..088d7f9cacde4e800e73f5ffd5cce91d18304e53 100644
--- a/mace/core/public/mace.h
+++ b/mace/core/public/mace.h
@@ -38,33 +38,33 @@ enum DataType {
   DT_UINT32 = 22
 };
 
-class TensorProto {
+class ConstTensor {
  public:
-  TensorProto(const std::string &name,
+  ConstTensor(const std::string &name,
               unsigned char *data,
               const std::vector<int64_t> &dims,
               const DataType data_type = DT_FLOAT,
               uint32_t node_id = 0);
-  TensorProto(const std::string &name,
+  ConstTensor(const std::string &name,
               unsigned char *data,
               const std::vector<int64_t> &dims,
               const int data_type,
               uint32_t node_id = 0);
 
   const std::string &name() const;
-  unsigned char *data() const;
-  const int64_t data_size() const;
+  const unsigned char *data() const;
+  int64_t data_size() const;
   const std::vector<int64_t> &dims() const;
   DataType data_type() const;
   uint32_t node_id() const;
 
  private:
-  std::string name_;
-  unsigned char *data_;
-  int64_t data_size_;
-  std::vector<int64_t> dims_;
-  DataType data_type_;
-  uint32_t node_id_;
+  const std::string name_;
+  const unsigned char *data_;
+  const int64_t data_size_;
+  const std::vector<int64_t> dims_;
+  const DataType data_type_;
+  const uint32_t node_id_;
 };
 
 class Argument {
@@ -270,8 +270,8 @@ class NetDef {
   const std::vector<Argument> &arg() const;
   Argument *add_arg();
   std::vector<Argument> &mutable_arg();
-  const std::vector<TensorProto> &tensors() const;
-  std::vector<TensorProto> &mutable_tensors();
+  const std::vector<ConstTensor> &tensors() const;
+  std::vector<ConstTensor> &mutable_tensors();
   const MemoryArena &mem_arena() const;
   bool has_mem_arena() const;
   MemoryArena &mutable_mem_arena();
@@ -288,7 +288,7 @@ class NetDef {
   std::string version_;
   std::vector<OperatorDef> op_;
   std::vector<Argument> arg_;
-  std::vector<TensorProto> tensors_;
+  std::vector<ConstTensor> tensors_;
 
   // for mem optimization
   MemoryArena mem_arena_;
diff --git a/mace/core/serializer.cc b/mace/core/serializer.cc
index 568e39fa4fa07628627e3bac7fa53240b597ef86..a0e3bc59de523641ca96f70b3526a103eadf80d4 100644
--- a/mace/core/serializer.cc
+++ b/mace/core/serializer.cc
@@ -6,13 +6,13 @@
 
 namespace mace {
 
-unique_ptr<TensorProto> Serializer::Serialize(const Tensor &tensor,
+unique_ptr<ConstTensor> Serializer::Serialize(const Tensor &tensor,
                                               const string &name) {
   MACE_NOT_IMPLEMENTED;
   return nullptr;
 }
 
-unique_ptr<Tensor> Serializer::Deserialize(const TensorProto &proto,
+unique_ptr<Tensor> Serializer::Deserialize(const ConstTensor &proto,
                                            DeviceType type) {
   unique_ptr<Tensor> tensor(
       new Tensor(GetDeviceAllocator(type), proto.data_type()));
@@ -24,31 +24,40 @@ unique_ptr<Tensor> Serializer::Deserialize(const TensorProto &proto,
 
   switch (proto.data_type()) {
     case DT_FLOAT:
-      tensor->Copy<float>(reinterpret_cast<float*>(proto.data()), proto.data_size());
+      tensor->Copy<float>(reinterpret_cast<const float *>(proto.data()),
+                          proto.data_size());
       break;
     case DT_DOUBLE:
-      tensor->Copy<double>(reinterpret_cast<double*>(proto.data()), proto.data_size());
+      tensor->Copy<double>(reinterpret_cast<const double *>(proto.data()),
+                           proto.data_size());
       break;
     case DT_INT32:
-      tensor->Copy<int32_t>(reinterpret_cast<int32_t*>(proto.data()), proto.data_size());
+      tensor->Copy<int32_t>(reinterpret_cast<const int32_t *>(proto.data()),
+                            proto.data_size());
       break;
     case DT_INT64:
-      tensor->Copy<int64_t>(reinterpret_cast<int64_t*>(proto.data()), proto.data_size());
+      tensor->Copy<int64_t>(reinterpret_cast<const int64_t *>(proto.data()),
+                            proto.data_size());
       break;
     case DT_UINT8:
-      tensor->CopyWithCast<int32_t, uint8_t>(reinterpret_cast<int32_t*>(proto.data()), proto.data_size());
+      tensor->CopyWithCast<int32_t, uint8_t>(
+          reinterpret_cast<const int32_t *>(proto.data()), proto.data_size());
       break;
     case DT_INT16:
-      tensor->CopyWithCast<int32_t, uint16_t>(reinterpret_cast<int32_t*>(proto.data()), proto.data_size());
+      tensor->CopyWithCast<int32_t, uint16_t>(
+          reinterpret_cast<const int32_t *>(proto.data()), proto.data_size());
       break;
     case DT_INT8:
-      tensor->CopyWithCast<int32_t, int8_t>(reinterpret_cast<int32_t*>(proto.data()), proto.data_size());
+      tensor->CopyWithCast<int32_t, int8_t>(
+          reinterpret_cast<const int32_t *>(proto.data()), proto.data_size());
       break;
     case DT_UINT16:
-      tensor->CopyWithCast<int32_t, int16_t>(reinterpret_cast<int32_t*>(proto.data()), proto.data_size());
+      tensor->CopyWithCast<int32_t, int16_t>(
+          reinterpret_cast<const int32_t *>(proto.data()), proto.data_size());
       break;
     case DT_BOOL:
-      tensor->CopyWithCast<int32_t, bool>(reinterpret_cast<int32_t*>(proto.data()), proto.data_size());
+      tensor->CopyWithCast<int32_t, bool>(
+          reinterpret_cast<const int32_t *>(proto.data()), proto.data_size());
       break;
     default:
       MACE_NOT_IMPLEMENTED;
@@ -58,4 +67,4 @@ unique_ptr<Tensor> Serializer::Deserialize(const TensorProto &proto,
   return tensor;
 }
 
-}  // namespace mace
\ No newline at end of file
+}  // namespace mace
diff --git a/mace/core/serializer.h b/mace/core/serializer.h
index 64c33b6d12d4a482c600a56028db50c4787296bc..c615300a0b2df3373204710db305391e0badadd0 100644
--- a/mace/core/serializer.h
+++ b/mace/core/serializer.h
@@ -16,9 +16,9 @@ class Serializer {
   Serializer() {}
   ~Serializer() {}
 
-  unique_ptr<TensorProto> Serialize(const Tensor &tensor, const string &name);
+  unique_ptr<ConstTensor> Serialize(const Tensor &tensor, const string &name);
 
-  unique_ptr<Tensor> Deserialize(const TensorProto &proto, DeviceType type);
+  unique_ptr<Tensor> Deserialize(const ConstTensor &proto, DeviceType type);
 
   DISABLE_COPY_AND_ASSIGN(Serializer);
 };
diff --git a/mace/examples/helloworld.cc b/mace/examples/helloworld.cc
index a89045090e8f56e6b4a9050e7ec53609563c74b5..90e04317192a2e2aa198de183276ee76a90ae9a0 100644
--- a/mace/examples/helloworld.cc
+++ b/mace/examples/helloworld.cc
@@ -45,7 +45,7 @@ int main() {
 
   alignas(4) unsigned char tensor_data[] = "012345678901234567890123";
   const std::vector<int64_t> dims = {1, 2, 3, 1};
-  TensorProto input("Input", tensor_data, dims, DataType::DT_FLOAT);
+  ConstTensor input("Input", tensor_data, dims, DataType::DT_FLOAT);
   net_def.mutable_tensors().push_back(input);
 
   // Create workspace and input tensor
diff --git a/mace/python/tools/model.template b/mace/python/tools/model.template
index 810b2518e1d2e6246c02ce01a3fb5c5b4a805086..6588d9a2c4492e962b3310cf33fad7bef7a8513a 100644
--- a/mace/python/tools/model.template
+++ b/mace/python/tools/model.template
@@ -13,8 +13,8 @@ alignas(4) unsigned char {{ tensor_info.name }}[] = {
 {% for d in tensor_info.data %}{{"0x%02X, " % d }}{%endfor%}
 };
 
-void Create{{tensor.name}}(std::vector<mace::TensorProto> &tensors) {
-  tensors.emplace_back(mace::TensorProto(
+void Create{{tensor.name}}(std::vector<mace::ConstTensor> &tensors) {
+  tensors.emplace_back(mace::ConstTensor(
       {{ tensor.name|tojson }}, {{ tensor.name }},
       { {{ tensor.dims|join(', ') }} }, {{ tensor.data_type }}, {{ tensor.node_id }}));
 }
@@ -100,7 +100,7 @@ void CreateOperator{{i}}(mace::OperatorDef &op) {
 namespace {{tag}} {
 
 {% for tensor in tensors %}
-extern void Create{{ tensor.name }}(std::vector<mace::TensorProto> &tensors);
+extern void Create{{ tensor.name }}(std::vector<mace::ConstTensor> &tensors);
 {% endfor %}
 
 
@@ -159,7 +159,7 @@ static void CreateOperators(std::vector<mace::OperatorDef> &ops) {
 
 }
 
-static void CreateTensors(std::vector<mace::TensorProto> &tensors) {
+static void CreateTensors(std::vector<mace::ConstTensor> &tensors) {
   tensors.reserve({{ net.tensors|length }});
 
   {% for tensor in net.tensors %}