diff --git a/paddle/fluid/CMakeLists.txt b/paddle/fluid/CMakeLists.txt
index e281c8c48f5041f9ca65ca532e68549a05980750..8248dcd36398bc12053c4a392168fd3015d92c15 100644
--- a/paddle/fluid/CMakeLists.txt
+++ b/paddle/fluid/CMakeLists.txt
@@ -8,5 +8,8 @@ add_subdirectory(pybind)
 add_subdirectory(eager)
 add_subdirectory(prim)
 add_subdirectory(jit)
+if(WITH_NEWIR)
+  add_subdirectory(dialect)
+endif()
 # NOTE: please add subdirectory inference at last.
 add_subdirectory(inference)
diff --git a/paddle/fluid/dialect/CMakeLists.txt b/paddle/fluid/dialect/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..644b9913794250f0025f09a8ac7c660825eafd9f
--- /dev/null
+++ b/paddle/fluid/dialect/CMakeLists.txt
@@ -0,0 +1,9 @@
+set(PD_DIALECT_SOURCE_DIR "${PADDLE_SOURCE_DIR}/paddle/fluid/dialect")
+set(PD_DIALECT_BINARY_DIR "${PADDLE_BINARY_DIR}/paddle/fluid/dialect")
+
+file(GLOB PD_DIALECT_SRCS "*.cc")
+
+cc_library(
+  pd_dialect
+  SRCS ${PD_DIALECT_SRCS}
+  DEPS new_ir framework_proto dense_tensor)
diff --git a/paddle/fluid/dialect/pd_dialect.cc b/paddle/fluid/dialect/pd_dialect.cc
new file mode 100644
index 0000000000000000000000000000000000000000..df98a9978803a07e23a9e970d5464f9c61dd2d99
--- /dev/null
+++ b/paddle/fluid/dialect/pd_dialect.cc
@@ -0,0 +1,97 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "paddle/fluid/dialect/pd_dialect.h"
+#include "paddle/fluid/dialect/pd_type.h"
+#include "paddle/fluid/dialect/utils.h"
+#include "paddle/fluid/framework/convert_utils.h"
+#include "paddle/fluid/framework/data_type.h"
+#include "paddle/ir/builtin_type.h"
+#include "paddle/ir/dialect_interface.h"
+#include "paddle/phi/core/dense_tensor.h"
+
+namespace paddle {
+namespace dialect {
+std::shared_ptr<paddle::framework::Variable>
+ParameterConvertInterface::ParameterToVariable(ir::Parameter* parameter) {
+  if (parameter->type().isa<DenseTensorType>()) {
+    VLOG(4) << "Convert a DenseTensor Parameter to a variable.";
+    std::shared_ptr<paddle::framework::Variable> var =
+        std::make_shared<paddle::framework::Variable>();
+    phi::DenseTensor* tensor = var->GetMutable<phi::DenseTensor>();
+    // Init DenseTensor
+    auto dim = parameter->type().dyn_cast<DenseTensorType>().dim();
+    phi::DenseTensorMeta meta(
+        TransToPhiDataType(
+            parameter->type().dyn_cast<DenseTensorType>().dtype()),
+        phi::DDim(dim.data(), dim.size()),
+        TransToPhiDataLayout(
+            parameter->type().dyn_cast<DenseTensorType>().data_layout()),
+        parameter->type().dyn_cast<DenseTensorType>().lod(),
+        parameter->type().dyn_cast<DenseTensorType>().offset());
+    tensor->set_meta(meta);
+    paddle::platform::DeviceContext* dev_ctx =
+        paddle::platform::DeviceContextPool::Instance().Get(
+            paddle::platform::CPUPlace());
+    dev_ctx->Alloc(tensor,
+                   TransToPhiDataType(
+                       parameter->type().dyn_cast<DenseTensorType>().dtype()));
+    memcpy(tensor->data(),
+           parameter->data(),
+           tensor->numel() * phi::SizeOf(tensor->dtype()));
+    return var;
+  } else {
+    return nullptr;
+  }
+}
+
+std::unique_ptr<ir::Parameter> ParameterConvertInterface::VariableToParameter(
+    paddle::framework::Variable* var) {
+  if (var->IsType<phi::DenseTensor>()) {
+    phi::DenseTensor* tensor = var->GetMutable<phi::DenseTensor>();
+    // Get Meta
+    ir::IrContext* ctx = ir::IrContext::Instance();
+    ir::Type data_type = TransToIrDataType(tensor->dtype(), ctx);
+    DenseTensorTypeStorage::Dim dims(tensor->dims().size());
+    std::copy(tensor->dims().Get(),
+              tensor->dims().Get() + tensor->dims().size(),
+              dims.data());
+    DenseTensorTypeStorage::DataLayout data_layout =
+        TransToIrDataLayout(tensor->layout());
+    DenseTensorTypeStorage::LoD lod = tensor->lod();
+    size_t offset = tensor->meta().offset;
+    void* data = tensor->data();
+    ir::Type dense_tensor_type =
+        DenseTensorType::get(ctx, data_type, dims, data_layout, lod, offset);
+    return std::make_unique<ir::Parameter>(
+        data,
+        tensor->numel() * phi::SizeOf(tensor->dtype()),
+        dense_tensor_type);
+  } else {
+    return nullptr;
+  }
+}
+
+PaddleDialect::PaddleDialect(ir::IrContext* context)
+    : ir::Dialect(name(), context, ir::TypeId::get<PaddleDialect>()) {
+  initialize();
+}
+
+void PaddleDialect::initialize() {
+  RegisterTypes<GET_PADDLE_TYPE_LIST>();
+  RegisterInterfaces<ParameterConvertInterface>();
+}
+
+}  // namespace dialect
+}  // namespace paddle
diff --git a/paddle/fluid/dialect/pd_dialect.h b/paddle/fluid/dialect/pd_dialect.h
new file mode 100644
index 0000000000000000000000000000000000000000..a81ff7cd48ceb24b6d30505ecaf3211d31ae8d85
--- /dev/null
+++ b/paddle/fluid/dialect/pd_dialect.h
@@ -0,0 +1,47 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/fluid/framework/variable.h"
+#include "paddle/ir/dialect.h"
+#include "paddle/ir/parameter.h"
+
+namespace paddle {
+namespace dialect {
+class ParameterConvertInterface
+    : public ir::DialectInterface::Base<ParameterConvertInterface> {
+ public:
+  explicit ParameterConvertInterface(ir::Dialect* dialect) : Base(dialect) {}
+
+  // NOTE(zhangbo): Only support new a CPU Variable.
+  std::shared_ptr<paddle::framework::Variable> ParameterToVariable(
+      ir::Parameter* parameter);
+
+  std::unique_ptr<ir::Parameter> VariableToParameter(
+      paddle::framework::Variable* var);
+};
+
+class PaddleDialect : public ir::Dialect {
+ public:
+  explicit PaddleDialect(ir::IrContext* context);
+
+  static const char* name() { return "pd"; }
+
+ private:
+  void initialize();
+};
+
+}  // namespace dialect
+}  // namespace paddle
diff --git a/paddle/ir/builtin_type.cc b/paddle/fluid/dialect/pd_type.cc
similarity index 69%
rename from paddle/ir/builtin_type.cc
rename to paddle/fluid/dialect/pd_type.cc
index 5e18b945016c7156bf1914ddd339e00958aac273..017af3ffd5aea80c2d6c21f29487dbe5d06ee8af 100644
--- a/paddle/ir/builtin_type.cc
+++ b/paddle/fluid/dialect/pd_type.cc
@@ -12,23 +12,28 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/ir/builtin_type.h"
+#include "paddle/fluid/dialect/pd_type.h"
 
-namespace ir {
+namespace paddle {
+namespace dialect {
 const ir::Type& DenseTensorType::dtype() const { return storage()->dtype_; }
 
-const ir::DenseTensorTypeStorage::Dim& DenseTensorType::dim() const {
+const paddle::dialect::DenseTensorTypeStorage::Dim& DenseTensorType::dim()
+    const {
   return storage()->dims_;
 }
 
-const ir::DenseTensorTypeStorage::DataLayout& DenseTensorType::data_layout()
-    const {
+const paddle::dialect::DenseTensorTypeStorage::DataLayout&
+DenseTensorType::data_layout() const {
   return storage()->layout_;
 }
 
-const ir::DenseTensorTypeStorage::LoD& DenseTensorType::lod() const {
+const paddle::dialect::DenseTensorTypeStorage::LoD& DenseTensorType::lod()
+    const {
   return storage()->lod_;
 }
 
 const size_t& DenseTensorType::offset() const { return storage()->offset_; }
-}  // namespace ir
+
+}  // namespace dialect
+}  // namespace paddle
diff --git a/paddle/fluid/dialect/pd_type.h b/paddle/fluid/dialect/pd_type.h
new file mode 100644
index 0000000000000000000000000000000000000000..2040828a146ac1744c71a1dde0b03a3d027c7eda
--- /dev/null
+++ b/paddle/fluid/dialect/pd_type.h
@@ -0,0 +1,46 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/fluid/dialect/pd_type_storage.h"
+#include "paddle/ir/type.h"
+
+namespace paddle {
+namespace dialect {
+#define GET_PADDLE_TYPE_LIST paddle::dialect::DenseTensorType
+
+///
+/// \brief Define built-in parameteric types.
+///
+class DenseTensorType : public ir::Type {
+ public:
+  using Type::Type;
+
+  DECLARE_TYPE_UTILITY_FUNCTOR(DenseTensorType, DenseTensorTypeStorage);
+
+  const ir::Type &dtype() const;
+
+  const paddle::dialect::DenseTensorTypeStorage::Dim &dim() const;
+
+  const paddle::dialect::DenseTensorTypeStorage::DataLayout &data_layout()
+      const;
+
+  const paddle::dialect::DenseTensorTypeStorage::LoD &lod() const;
+
+  const size_t &offset() const;
+};
+
+}  // namespace dialect
+}  // namespace paddle
diff --git a/paddle/ir/builtin_type_storage.h b/paddle/fluid/dialect/pd_type_storage.h
similarity index 98%
rename from paddle/ir/builtin_type_storage.h
rename to paddle/fluid/dialect/pd_type_storage.h
index 132a1656a7975dc2197ceaece4d0f9a386db4ffc..1ea44eccc7347aed918feca808fe0a3e27d55dda 100644
--- a/paddle/ir/builtin_type_storage.h
+++ b/paddle/fluid/dialect/pd_type_storage.h
@@ -36,7 +36,8 @@ struct hash<std::vector<T>> {
 
 }  // namespace std
 
-namespace ir {
+namespace paddle {
+namespace dialect {
 ///
 /// \brief Define Parameteric TypeStorage for DenseTensorType.
 ///
@@ -151,4 +152,5 @@ struct DenseTensorTypeStorage : public ir::TypeStorage {
   size_t offset_;
 };
 
-}  // namespace ir
+}  // namespace dialect
+}  // namespace paddle
diff --git a/paddle/fluid/dialect/utils.h b/paddle/fluid/dialect/utils.h
new file mode 100644
index 0000000000000000000000000000000000000000..69af6fd1ddce80b8aac1f6bb4b1f7fa2764603eb
--- /dev/null
+++ b/paddle/fluid/dialect/utils.h
@@ -0,0 +1,136 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/fluid/dialect/pd_type_storage.h"
+#include "paddle/fluid/framework/convert_utils.h"
+#include "paddle/fluid/framework/data_type.h"
+#include "paddle/ir/builtin_type.h"
+#include "paddle/phi/core/dense_tensor.h"
+
+namespace paddle {
+namespace dialect {
+// TODO(zhangbo): The builtin type needs to cover all data types of
+// phi::DataType.
+inline phi::DataType TransToPhiDataType(ir::Type dtype) {
+  if (dtype.isa<ir::Float16Type>()) {
+    return phi::DataType::FLOAT16;
+  } else if (dtype.isa<ir::Float32Type>()) {
+    return phi::DataType::FLOAT32;
+  } else if (dtype.isa<ir::Float64Type>()) {
+    return phi::DataType::FLOAT64;
+  } else if (dtype.isa<ir::Int16Type>()) {
+    return phi::DataType::INT16;
+  } else if (dtype.isa<ir::Int32Type>()) {
+    return phi::DataType::INT32;
+  } else if (dtype.isa<ir::Int64Type>()) {
+    return phi::DataType::INT64;
+  } else {
+    PADDLE_THROW(phi::errors::Unimplemented(
+        "Unsupported ir data type when casting it into "
+        "phi data type."));
+  }
+}
+
+inline ir::Type TransToIrDataType(phi::DataType dtype,
+                                  ir::IrContext *ctx = nullptr) {
+  if (ctx == nullptr) {
+    ctx = ir::IrContext::Instance();
+  }
+  switch (dtype) {
+    case phi::DataType::FLOAT16:
+      return ir::Float16Type::get(ctx);
+    case phi::DataType::FLOAT32:
+      return ir::Float32Type::get(ctx);
+    case phi::DataType::FLOAT64:
+      return ir::Float64Type::get(ctx);
+    case phi::DataType::INT16:
+      return ir::Int16Type::get(ctx);
+    case phi::DataType::INT32:
+      return ir::Int32Type::get(ctx);
+    case phi::DataType::INT64:
+      return ir::Int64Type::get(ctx);
+    default:
+      PADDLE_THROW(phi::errors::Unimplemented(
+          "Unsupported phi data type `%s` when casting it into "
+          "ir data type.",
+          dtype));
+  }
+}
+
+inline phi::DataLayout TransToPhiDataLayout(
+    DenseTensorTypeStorage::DataLayout data_layout) {
+  switch (data_layout) {
+    case DenseTensorTypeStorage::DataLayout::NHWC:
+      return phi::DataLayout::NHWC;
+    case DenseTensorTypeStorage::DataLayout::NCHW:
+      return phi::DataLayout::NCHW;
+    case DenseTensorTypeStorage::DataLayout::NCDHW:
+      return phi::DataLayout::NCDHW;
+    case DenseTensorTypeStorage::DataLayout::NDHWC:
+      return phi::DataLayout::NDHWC;
+    case DenseTensorTypeStorage::DataLayout::ONEDNN:
+      return phi::DataLayout::ONEDNN;
+    case DenseTensorTypeStorage::DataLayout::SPARSE_COO:
+      return phi::DataLayout::SPARSE_COO;
+    case DenseTensorTypeStorage::DataLayout::SPARSE_CSR:
+      return phi::DataLayout::SPARSE_CSR;
+    case DenseTensorTypeStorage::DataLayout::PSTRING_UNION:
+      return phi::DataLayout::PSTRING_UNION;
+    case DenseTensorTypeStorage::DataLayout::NUM_DATA_LAYOUTS:
+      return phi::DataLayout::NUM_DATA_LAYOUTS;
+    case DenseTensorTypeStorage::DataLayout::ALL_LAYOUT:
+      return phi::DataLayout::ALL_LAYOUT;
+    default:
+      PADDLE_THROW(phi::errors::Unimplemented(
+          "Unsupported ir data layout `%s` when casting it into "
+          "phi data type.",
+          static_cast<int>(data_layout)));
+  }
+}
+
+inline DenseTensorTypeStorage::DataLayout TransToIrDataLayout(
+    phi::DataLayout data_layout) {
+  switch (data_layout) {
+    case phi::DataLayout::NHWC:
+      return DenseTensorTypeStorage::DataLayout::NHWC;
+    case phi::DataLayout::NCHW:
+      return DenseTensorTypeStorage::DataLayout::NCHW;
+    case phi::DataLayout::NCDHW:
+      return DenseTensorTypeStorage::DataLayout::NCDHW;
+    case phi::DataLayout::NDHWC:
+      return DenseTensorTypeStorage::DataLayout::NDHWC;
+    case phi::DataLayout::ONEDNN:
+      return DenseTensorTypeStorage::DataLayout::ONEDNN;
+    case phi::DataLayout::SPARSE_COO:
+      return DenseTensorTypeStorage::DataLayout::SPARSE_COO;
+    case phi::DataLayout::SPARSE_CSR:
+      return DenseTensorTypeStorage::DataLayout::SPARSE_CSR;
+    case phi::DataLayout::PSTRING_UNION:
+      return DenseTensorTypeStorage::DataLayout::PSTRING_UNION;
+    case phi::DataLayout::NUM_DATA_LAYOUTS:
+      return DenseTensorTypeStorage::DataLayout::NUM_DATA_LAYOUTS;
+    case phi::DataLayout::ALL_LAYOUT:
+      return DenseTensorTypeStorage::DataLayout::ALL_LAYOUT;
+    default:
+      PADDLE_THROW(phi::errors::Unimplemented(
+          "Unsupported phi data layout `%s` when casting it into "
+          "ir data type.",
+          static_cast<int>(data_layout)));
+  }
+}
+
+}  // namespace dialect
+}  // namespace paddle
diff --git a/paddle/ir/builtin_attribute.cc b/paddle/ir/builtin_attribute.cc
index 3a1b2acfb716bb9bd0283b015641dac0486172a7..40c0204ce9cfd0917ed3b7ee375890225b14c5c4 100644
--- a/paddle/ir/builtin_attribute.cc
+++ b/paddle/ir/builtin_attribute.cc
@@ -19,24 +19,4 @@ std::string StrAttribute::data() const { return storage()->GetAsKey(); }
 
 uint32_t StrAttribute::size() const { return storage()->GetAsKey().size(); }
 
-NamedAttribute::NamedAttribute(StrAttribute name, Attribute value)
-    : name_(name), value_(value) {}
-
-bool NamedAttribute::operator<(const NamedAttribute &right) const {
-  return name() < right.name();
-}
-
-bool NamedAttribute::operator==(const NamedAttribute &right) const {
-  return name() == right.name() && value() == right.value();
-}
-
-bool NamedAttribute::operator!=(const NamedAttribute &right) const {
-  return !(*this == right);
-}
-
-Attribute DictionaryAttribute::GetValue(const StrAttribute &name) {
-  return storage()->GetValue(name);
-}
-
-uint32_t DictionaryAttribute::size() const { return storage()->size(); }
 }  // namespace ir
diff --git a/paddle/ir/builtin_attribute.h b/paddle/ir/builtin_attribute.h
index 82b5f8eb48aa58d49ed8592f85d76849174817ee..4e2164a1e4d6a9f61540e31ef4f6cdb7d70a15fa 100644
--- a/paddle/ir/builtin_attribute.h
+++ b/paddle/ir/builtin_attribute.h
@@ -22,7 +22,7 @@ namespace ir {
 ///
 /// \brief All built-in attributes.
 ///
-#define GET_BUILT_IN_ATTRIBUTE_LIST ir::StrAttribute, ir::DictionaryAttribute
+#define GET_BUILT_IN_ATTRIBUTE_LIST ir::StrAttribute
 
 class StrAttribute : public ir::Attribute {
  public:
@@ -39,55 +39,13 @@ class StrAttribute : public ir::Attribute {
   uint32_t size() const;
 };
 
-class NamedAttribute {
- public:
-  NamedAttribute(StrAttribute name, Attribute value);
-
-  StrAttribute name() const { return name_; }
-
-  Attribute value() const { return value_; }
-
-  void SetName(StrAttribute name) { name_ = name; }
-
-  void SetValue(Attribute value) { value_ = value; }
-
-  bool operator<(const NamedAttribute &right) const;
-
-  bool operator==(const NamedAttribute &right) const;
-
-  bool operator!=(const NamedAttribute &right) const;
-
-  friend struct std::hash<NamedAttribute>;
-
-  operator std::pair<const StrAttribute, Attribute>() const {
-    return std::make_pair(name_, value_);
-  }
-
- private:
-  StrAttribute name_;
-  Attribute value_;
-};
-
-class DictionaryAttribute : public ir::Attribute {
- public:
-  using Attribute::Attribute;
-
-  DECLARE_ATTRIBUTE_UTILITY_FUNCTOR(DictionaryAttribute,
-                                    DictionaryAttributeStorage);
-
-  Attribute GetValue(const StrAttribute &name);
-
-  uint32_t size() const;
-};
-
 }  // namespace ir
 
 namespace std {
 template <>
-struct hash<ir::NamedAttribute> {
-  std::size_t operator()(const ir::NamedAttribute &obj) const {
-    return ir::hash_combine(std::hash<ir::Attribute>()(obj.name_),
-                            std::hash<ir::Attribute>()(obj.value_));
+struct hash<ir::StrAttribute> {
+  std::size_t operator()(const ir::StrAttribute &obj) const {
+    return std::hash<const ir::StrAttribute::Storage *>()(obj.storage());
   }
 };
 }  // namespace std
diff --git a/paddle/ir/builtin_attribute_storage.cc b/paddle/ir/builtin_attribute_storage.cc
deleted file mode 100644
index 3f785d20c9b92218aa5b315d6ad4ec94fd1310cb..0000000000000000000000000000000000000000
--- a/paddle/ir/builtin_attribute_storage.cc
+++ /dev/null
@@ -1,77 +0,0 @@
-// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include "paddle/ir/builtin_attribute_storage.h"
-#include "paddle/ir/builtin_attribute.h"
-#include "paddle/ir/utils.h"
-
-namespace ir {
-
-DictionaryAttributeStorage::DictionaryAttributeStorage(const ParamKey &key) {
-  size_ = key.size();
-  data_ = reinterpret_cast<NamedAttribute *>(
-      malloc(sizeof(NamedAttribute) * size_));
-  uint32_t idx = 0;
-  for (auto iter = key.begin(); iter != key.end(); ++iter) {
-    data_[idx].SetName(iter->first);
-    data_[idx].SetValue(iter->second);
-    idx++;
-  }
-}
-
-std::size_t DictionaryAttributeStorage::HashValue(const ParamKey &key) {
-  std::size_t hash_value = key.size();
-  for (auto iter = key.begin(); iter != key.end(); ++iter) {
-    hash_value = ir::hash_combine(
-        hash_value,
-        std::hash<NamedAttribute>()(NamedAttribute(iter->first, iter->second)));
-  }
-  return hash_value;
-}
-
-bool DictionaryAttributeStorage::operator==(const ParamKey &key) const {
-  uint32_t size = key.size();
-  if (size_ != size) return false;
-  uint32_t idx = 0;
-  for (auto iter = key.begin(); iter != key.end(); ++iter) {
-    if (data_[idx] != NamedAttribute(iter->first, iter->second)) {
-      return false;
-    }
-    idx++;
-  }
-  return true;
-}
-
-DictionaryAttributeStorage::ParamKey DictionaryAttributeStorage::GetAsKey()
-    const {
-  return ParamKey(data_, data_ + size_);
-}
-
-Attribute DictionaryAttributeStorage::GetValue(const StrAttribute &name) const {
-  size_t left = 0;
-  size_t right = size_;
-  while (left < right) {
-    size_t mid = left + (right - left) / 2;
-    if (data_[mid].name() == name) {
-      return data_[mid].value();
-    } else if (data_[mid].name() < name) {
-      left = mid + 1;
-    } else {
-      right = mid;
-    }
-  }
-  return nullptr;
-}
-
-}  // namespace ir
diff --git a/paddle/ir/builtin_attribute_storage.h b/paddle/ir/builtin_attribute_storage.h
index a34648fb17e3581868b8c4b51ee533b30cd1d442..f6f97d5e616ed7038e4ad32f554e9a09bb56f690 100644
--- a/paddle/ir/builtin_attribute_storage.h
+++ b/paddle/ir/builtin_attribute_storage.h
@@ -54,36 +54,4 @@ struct StrAttributeStorage : public ir::AttributeStorage {
   uint32_t size_;
 };
 
-///
-/// \brief Define Parameteric AttributeStorage for DictionaryAttributeStorage.
-///
-class StrAttribute;
-class NamedAttribute;
-struct DictionaryAttributeStorage : public AttributeStorage {
-  using ParamKey = std::map<StrAttribute, Attribute>;
-
-  explicit DictionaryAttributeStorage(const ParamKey &key);
-
-  ~DictionaryAttributeStorage() { free(data_); }
-
-  static DictionaryAttributeStorage *Construct(ParamKey key) {
-    return new DictionaryAttributeStorage(key);
-  }
-
-  static std::size_t HashValue(const ParamKey &key);
-
-  bool operator==(const ParamKey &key) const;
-
-  ParamKey GetAsKey() const;
-
-  Attribute GetValue(const StrAttribute &name) const;
-
-  NamedAttribute *data() const { return data_; }
-
-  uint32_t size() const { return size_; }
-
- private:
-  NamedAttribute *data_;
-  uint32_t size_;
-};
 }  // namespace ir
diff --git a/paddle/ir/builtin_dialect.cc b/paddle/ir/builtin_dialect.cc
index 144ee272c4c1b0b2b6a6407f507ce53dca76c9a0..5e08798828f5904990c43f382d08447774f3dafa 100644
--- a/paddle/ir/builtin_dialect.cc
+++ b/paddle/ir/builtin_dialect.cc
@@ -14,6 +14,7 @@
 
 #include "paddle/ir/builtin_dialect.h"
 #include "paddle/ir/builtin_attribute.h"
+#include "paddle/ir/builtin_op.h"
 #include "paddle/ir/builtin_type.h"
 
 namespace ir {
@@ -26,6 +27,7 @@ void BuiltinDialect::initialize() {
   // Register all built-in types defined in builtin_type.h.
   RegisterTypes<GET_BUILT_IN_TYPE_LIST>();
   RegisterAttributes<GET_BUILT_IN_ATTRIBUTE_LIST>();
+  RegisterOps<GET_BUILT_IN_OP_LIST>();
 }
 
 }  // namespace ir
diff --git a/paddle/ir/builtin_op.cc b/paddle/ir/builtin_op.cc
new file mode 100644
index 0000000000000000000000000000000000000000..528631b0b7998d7c30a281f80ac8c6f535b03198
--- /dev/null
+++ b/paddle/ir/builtin_op.cc
@@ -0,0 +1,22 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "paddle/ir/builtin_op.h"
+
+namespace ir {
+const char *GetParameterOp::attributes_name_[] = {"parameter_name"};
+
+const char *SetParameterOp::attributes_name_[] = {"parameter_name"};
+
+}  // namespace ir
diff --git a/paddle/ir/builtin_op.h b/paddle/ir/builtin_op.h
new file mode 100644
index 0000000000000000000000000000000000000000..f5c4296394a9abf0fe3a35a60539a898902915b7
--- /dev/null
+++ b/paddle/ir/builtin_op.h
@@ -0,0 +1,57 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/ir/op_base.h"
+
+namespace ir {
+///
+/// \brief This macro is used to get a list of all built-in OPs in this file.
+/// The built-in Dialect will use this macro to quickly register all built-in
+/// OPs.
+///
+#define GET_BUILT_IN_OP_LIST ir::GetParameterOp, ir::SetParameterOp
+
+///
+/// \brief GetParameterOp: OpResult = GetParameterOp({StrAttribute,
+/// StrAttribute})
+///
+class GetParameterOp : public ir::Op<GetParameterOp> {
+ public:
+  using Op::Op;
+
+  static const char* name() { return "GetParameterOp"; }
+
+  static uint32_t attributes_num() { return 1; }
+
+  static const char* attributes_name_[];
+};
+
+///
+/// \brief GetParameterOp: SetParameterOp(OpOperand, {StrAttribute,
+/// StrAttribute})
+///
+class SetParameterOp : public ir::Op<SetParameterOp> {
+ public:
+  using Op::Op;
+
+  static const char* name() { return "SetParameterOp"; }
+
+  static uint32_t attributes_num() { return 1; }
+
+  static const char* attributes_name_[];
+};
+
+}  // namespace ir
diff --git a/paddle/ir/builtin_type.h b/paddle/ir/builtin_type.h
index 0f4db31d9d67d210a04d78ee4426089d7454a66b..d195cca2939c390b47ec15da3132df1823391a99 100644
--- a/paddle/ir/builtin_type.h
+++ b/paddle/ir/builtin_type.h
@@ -14,7 +14,6 @@
 
 #pragma once
 
-#include "paddle/ir/builtin_type_storage.h"
 #include "paddle/ir/type.h"
 
 namespace ir {
@@ -25,7 +24,7 @@ namespace ir {
 ///
 #define GET_BUILT_IN_TYPE_LIST                                      \
   ir::Float16Type, ir::Float32Type, ir::Float64Type, ir::Int16Type, \
-      ir::Int32Type, ir::Int64Type, ir::DenseTensorType
+      ir::Int32Type, ir::Int64Type
 
 ///
 /// \brief Define built-in parameterless types. Please add the necessary
@@ -97,24 +96,4 @@ class Int64Type : public ir::Type {
   static Int64Type get(ir::IrContext *context);
 };
 
-///
-/// \brief Define built-in parameteric types.
-///
-class DenseTensorType : public ir::Type {
- public:
-  using Type::Type;
-
-  DECLARE_TYPE_UTILITY_FUNCTOR(DenseTensorType, DenseTensorTypeStorage);
-
-  const ir::Type &dtype() const;
-
-  const ir::DenseTensorTypeStorage::Dim &dim() const;
-
-  const ir::DenseTensorTypeStorage::DataLayout &data_layout() const;
-
-  const ir::DenseTensorTypeStorage::LoD &lod() const;
-
-  const size_t &offset() const;
-};
-
 }  // namespace ir
diff --git a/paddle/ir/dialect.cc b/paddle/ir/dialect.cc
index cdc171557148abf4a8abf05c24895c15c26f4263..8d52f3bbc9d7837ad0b197b7715d0f013e2a31a8 100644
--- a/paddle/ir/dialect.cc
+++ b/paddle/ir/dialect.cc
@@ -18,6 +18,8 @@ namespace ir {
 Dialect::Dialect(std::string name, ir::IrContext *context, ir::TypeId id)
     : name_(std::move(name)), context_(context), id_(id) {}
 
+Dialect::~Dialect() = default;
+
 void Dialect::RegisterType(ir::AbstractType &&abstract_type) {
   ir::AbstractType *new_abstract_type =
       new ir::AbstractType(std::move(abstract_type));
@@ -35,4 +37,18 @@ void Dialect::RegisterAttribute(ir::AbstractAttribute &&abstract_attribute) {
 void Dialect::RegisterOp(const std::string &name, OpInfoImpl *op_info) {
   this->ir_context()->RegisterOpInfo(name, op_info);
 }
+
+void Dialect::RegisterInterface(std::unique_ptr<DialectInterface> interface) {
+  VLOG(4) << "Register interface into dialect" << std::endl;
+  auto it = registered_interfaces_.emplace(interface->interface_id(),
+                                           std::move(interface));
+  (void)it;
+}
+
+DialectInterface::~DialectInterface() = default;
+
+IrContext *DialectInterface::ir_context() const {
+  return dialect_->ir_context();
+}
+
 }  // namespace ir
diff --git a/paddle/ir/dialect.h b/paddle/ir/dialect.h
index 10adc6b3e2e0b61666796d878d94e2a31f6a8453..9a6b42dad5aed26c6f9f406efdb541eb718aa59e 100644
--- a/paddle/ir/dialect.h
+++ b/paddle/ir/dialect.h
@@ -15,11 +15,13 @@
 #pragma once
 
 #include "paddle/ir/attribute_base.h"
+#include "paddle/ir/dialect_interface.h"
 #include "paddle/ir/ir_context.h"
 #include "paddle/ir/op_info_impl.h"
 #include "paddle/ir/type_base.h"
 
 namespace ir {
+class DialectInterface;
 ///
 /// \brief Dialect can basically be understood as a namespace. In Dialect, we
 /// can define a series of types, attributes, operations, etc. An instance of
@@ -31,6 +33,8 @@ class Dialect {
  public:
   Dialect(std::string name, ir::IrContext *context, ir::TypeId id);
 
+  virtual ~Dialect();
+
   const std::string &name() const { return name_; }
 
   ir::IrContext *ir_context() const { return context_; }
@@ -49,8 +53,6 @@ class Dialect {
   template <typename T>
   void RegisterType() {
     VLOG(4) << "Type registered into Dialect. --->";
-    // if (this->ir_context()->registed_abstract_type().count(
-    //         ir::TypeId::get<T>()) == 0) {
     if (this->ir_context()->GetRegisteredAbstractType(ir::TypeId::get<T>()) ==
         nullptr) {
       ir::AbstractType *abstract_type =
@@ -118,11 +120,50 @@ class Dialect {
 
   void RegisterOp(const std::string &name, OpInfoImpl *op_info);
 
+  ///
+  /// \brief Register interface methods.
+  ///
+
+  DialectInterface *GetRegisteredInterface(TypeId id) {
+    auto it = registered_interfaces_.find(id);
+    return it != registered_interfaces_.end() ? it->second.get() : nullptr;
+  }
+
+  template <typename InterfaceT>
+  InterfaceT *GetRegisteredInterface() {
+    return static_cast<InterfaceT *>(
+        GetRegisteredInterface(TypeId::get<InterfaceT>()));
+  }
+
+  /// Register a dialect interface with this dialect instance.
+  void RegisterInterface(std::unique_ptr<DialectInterface> interface);
+
+  /// Register a set of dialect interfaces with this dialect instance.
+  template <typename... Args>
+  void RegisterInterfaces() {
+    (void)std::initializer_list<int>{
+        0, (RegisterInterface(std::make_unique<Args>(this)), 0)...};
+  }
+
+  template <typename InterfaceT, typename... Args>
+  InterfaceT &RegisterInterface(Args &&...args) {
+    InterfaceT *interface = new InterfaceT(this, std::forward<Args>(args)...);
+    RegisterInterface(std::unique_ptr<DialectInterface>(interface));
+    return *interface;
+  }
+
  private:
+  Dialect(const Dialect &) = delete;
+
+  Dialect &operator=(Dialect &) = delete;
+
   std::string name_;
 
   ir::IrContext *context_;  // not owned
 
   ir::TypeId id_;
+
+  std::unordered_map<TypeId, std::unique_ptr<DialectInterface>>
+      registered_interfaces_;
 };
 }  // namespace ir
diff --git a/paddle/ir/dialect_interface.h b/paddle/ir/dialect_interface.h
new file mode 100644
index 0000000000000000000000000000000000000000..dd3ec08af52fd61029b9436be480a660463351a9
--- /dev/null
+++ b/paddle/ir/dialect_interface.h
@@ -0,0 +1,67 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/ir/type_id.h"
+
+namespace ir {
+class Dialect;
+class IrContext;
+///
+/// \brief DialectInterface
+///
+template <typename ConcreteType, typename BaseT>
+class DialectInterfaceBase : public BaseT {
+ public:
+  using Base = DialectInterfaceBase<ConcreteType, BaseT>;
+
+  /// Get a unique id for the derived interface type.
+  static TypeId id() { return TypeId::get<ConcreteType>(); }
+
+ protected:
+  explicit DialectInterfaceBase(Dialect *dialect) : BaseT(dialect, id()) {}
+};
+
+class DialectInterface {
+ public:
+  virtual ~DialectInterface();
+
+  /// The base class used for all derived interface types. This class provides
+  /// utilities necessary for registration.
+  template <typename ConcreteType>
+  using Base = DialectInterfaceBase<ConcreteType, DialectInterface>;
+
+  /// Return the dialect that this interface represents.
+  Dialect *dialect() const { return dialect_; }
+
+  /// Return the context that holds the parent dialect of this interface.
+  IrContext *ir_context() const;
+
+  /// Return the derived interface id.
+  TypeId interface_id() const { return interface_id_; }
+
+ protected:
+  DialectInterface(Dialect *dialect, TypeId id)
+      : dialect_(dialect), interface_id_(id) {}
+
+ private:
+  /// The dialect that represents this interface.
+  Dialect *dialect_;
+
+  /// The unique identifier for the derived interface type.
+  TypeId interface_id_;
+};
+
+}  // namespace ir
diff --git a/paddle/ir/op_info_impl.h b/paddle/ir/op_info_impl.h
index 78ca586d9d209887a9c9e307226c9ab049eab968..4380866bceb810f1a518e37e70b38c9b873d3b21 100644
--- a/paddle/ir/op_info_impl.h
+++ b/paddle/ir/op_info_impl.h
@@ -236,6 +236,12 @@ class OpInfoImpl {
 
   ir::Dialect *dialect() const { return dialect_; }
 
+  uint32_t AttributeNum() const { return num_attributes_; }
+
+  const char *GetAttributeByIndex(size_t idx) const {
+    return idx < num_attributes_ ? p_attributes_[idx] : nullptr;
+  }
+
  private:
   OpInfoImpl(uint32_t num_interfaces,
              uint32_t num_traits,
diff --git a/paddle/ir/operation.cc b/paddle/ir/operation.cc
index 9222ee7e1afbb4777174c69d2a5f39eafd18f25a..3cfe2b048e472c733008fc2a679c5145d2fc2185 100644
--- a/paddle/ir/operation.cc
+++ b/paddle/ir/operation.cc
@@ -13,6 +13,8 @@
 // limitations under the License.
 
 #include "paddle/ir/operation.h"
+#include "paddle/ir/dialect.h"
+#include "paddle/ir/program.h"
 #include "paddle/ir/utils.h"
 
 namespace ir {
@@ -21,7 +23,7 @@ namespace ir {
 // OpInlineResult, Operation, Operand.
 Operation *Operation::create(const std::vector<ir::OpResult> &inputs,
                              const std::vector<ir::Type> &output_types,
-                             ir::DictionaryAttribute attribute,
+                             const AttributeMap &attribute,
                              ir::OpInfo op_info) {
   // 1. Calculate the required memory size for OpResults + Operation +
   // OpOperands.
@@ -63,7 +65,7 @@ Operation *Operation::create(const std::vector<ir::OpResult> &inputs,
     new (base_ptr) detail::OpOperandImpl(inputs[idx].impl_, op);
     base_ptr += sizeof(detail::OpOperandImpl);
   }
-  VLOG(4) << "Construct an Operation: " << op->print();
+
   return op;
 }
 
@@ -83,9 +85,18 @@ void Operation::destroy() {
   // 2.1. Deconstruct OpResult.
   char *base_ptr = aligned_ptr;
   for (size_t idx = num_results_; idx > 0; idx--) {
-    if (!reinterpret_cast<detail::OpResultImpl *>(base_ptr)->use_empty()) {
-      throw("Cannot destroy a value that still has uses!");
+    // release the uses of this result
+    detail::OpOperandImpl *first_use =
+        reinterpret_cast<detail::OpResultImpl *>(base_ptr)->first_use();
+    if (first_use != nullptr) {
+      first_use->release_source();
+      detail::OpOperandImpl *next_use = first_use->next_use();
+      while (next_use != nullptr) {
+        next_use->release_source();
+        next_use = next_use->next_use();
+      }
     }
+    // destory the result
     if (idx > max_inline_result_num) {
       reinterpret_cast<detail::OpOutlineResultImpl *>(base_ptr)
           ->~OpOutlineResultImpl();
@@ -117,11 +128,8 @@ void Operation::destroy() {
 
 Operation::Operation(uint32_t num_results,
                      uint32_t num_operands,
-                     ir::DictionaryAttribute attribute,
+                     const AttributeMap &attribute,
                      ir::OpInfo op_info) {
-  if (!attribute) {
-    throw("unexpected null attribute dictionary");
-  }
   num_results_ = num_results;
   num_operands_ = num_operands;
   attribute_ = attribute;
@@ -153,6 +161,15 @@ ir::OpResult Operation::GetResultByIndex(uint32_t index) {
   }
 }
 
+ir::OpOperand Operation::GetOperandByIndex(uint32_t index) {
+  if (index >= num_operands_) {
+    throw("index exceeds OP input range.");
+  }
+  char *ptr = reinterpret_cast<char *>(this) + sizeof(Operation) +
+              (index) * sizeof(detail::OpOperandImpl);
+  return ir::OpOperand(reinterpret_cast<detail::OpOperandImpl *>(ptr));
+}
+
 std::string Operation::print() {
   std::stringstream result;
   result << "{ " << num_results_ << " outputs, " << num_operands_
@@ -173,4 +190,9 @@ std::string Operation::print() {
   return result.str();
 }
 
+std::string Operation::op_name() const {
+  return op_info_.impl()->dialect()->name() + "." +
+         std::string(op_info_.impl()->name());
+}
+
 }  // namespace ir
diff --git a/paddle/ir/operation.h b/paddle/ir/operation.h
index a51043ad5687e565eda1eb12392729c58d87b872..9730244851fc415b6effcd629258b42ed70ff8d8 100644
--- a/paddle/ir/operation.h
+++ b/paddle/ir/operation.h
@@ -24,25 +24,35 @@ template <class ConcreteTrait>
 class OpTraitBase;
 template <typename ConcreteInterface>
 class OpInterfaceBase;
+class Program;
+
+using AttributeMap = std::unordered_map<std::string, Attribute>;
 
 class alignas(8) Operation final {
  public:
   ///
   /// \brief Malloc memory and construct objects in the following order:
   /// OpResultImpls|Operation|OpOperandImpls.
+  /// NOTE: Similar to new and delete, the destroy() and the create() need to be
+  /// used in conjunction.
   ///
   static Operation *create(const std::vector<ir::OpResult> &inputs,
                            const std::vector<ir::Type> &output_types,
-                           ir::DictionaryAttribute attribute,
+                           const AttributeMap &attribute,
                            ir::OpInfo op_info);
 
+  ///
+  /// \brief Destroy the operation objects and free memeory by create().
+  ///
   void destroy();
 
   ir::OpResult GetResultByIndex(uint32_t index);
 
+  ir::OpOperand GetOperandByIndex(uint32_t index);
+
   std::string print();
 
-  ir::DictionaryAttribute attribute() const { return attribute_; }
+  const AttributeMap &attribute() const { return attribute_; }
 
   ir::OpInfo op_info() const { return op_info_; }
 
@@ -50,6 +60,8 @@ class alignas(8) Operation final {
 
   uint32_t num_operands() const { return num_operands_; }
 
+  std::string op_name() const;
+
   template <typename T>
   T dyn_cast() const {
     return CastUtil<T>::call(this);
@@ -65,10 +77,16 @@ class alignas(8) Operation final {
     return op_info_.HasInterface<Interface>();
   }
 
+  Program *parent_program() const { return parent_program_; }
+
+  void set_parent_program(Program *parent_program) {
+    parent_program_ = parent_program;
+  }
+
  private:
   Operation(uint32_t num_results,
             uint32_t num_operands,
-            ir::DictionaryAttribute attribute,
+            const AttributeMap &attribute,
             ir::OpInfo op_info);
 
   template <typename T, typename Enabler = void>
@@ -92,13 +110,15 @@ class alignas(8) Operation final {
     }
   };
 
-  ir::DictionaryAttribute attribute_;
+  AttributeMap attribute_;
 
   ir::OpInfo op_info_;
 
   uint32_t num_results_ = 0;
 
   uint32_t num_operands_ = 0;
+
+  ir::Program *parent_program_{nullptr};
 };
 
 }  // namespace ir
diff --git a/paddle/ir/parameter.h b/paddle/ir/parameter.h
new file mode 100644
index 0000000000000000000000000000000000000000..38d74ea3cbff2e5f3db2b8aaab5fdf26575a1b4a
--- /dev/null
+++ b/paddle/ir/parameter.h
@@ -0,0 +1,70 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/ir/type.h"
+
+namespace ir {
+///
+/// \brief Parameter represents the weight in the calculation graph.
+///
+class Parameter {
+ public:
+  Parameter(void* data, size_t size, ir::Type type) {
+    data_ = malloc(size);
+    memcpy(data_, data, size);
+    size_ = size;
+    type_ = type;
+  }
+
+  Parameter(const Parameter& param) {
+    data_ = malloc(param.size_);
+    memcpy(data_, param.data_, param.size_);
+    size_ = param.size_;
+    type_ = param.type_;
+  }
+
+  Parameter& operator=(const Parameter& param) {
+    data_ = malloc(param.size_);
+    memcpy(data_, param.data_, param.size_);
+    size_ = param.size_;
+    type_ = param.type_;
+    return *this;
+  }
+
+  ~Parameter() { free(data_); }
+
+  Type type() const { return type_; }
+
+  void* data() const { return data_; }
+
+  bool is_mutable() const { return is_mutable_; }
+
+  void set_mutable() { is_mutable_ = true; }
+
+ private:
+  void* data_;
+
+  ///
+  /// \brief Number of bytes held in data_.
+  ///
+  size_t size_;
+
+  Type type_;
+
+  bool is_mutable_ = false;
+};
+
+}  // namespace ir
diff --git a/paddle/ir/program.cc b/paddle/ir/program.cc
new file mode 100644
index 0000000000000000000000000000000000000000..4caa7a8051341891dc52480db37e1b7215ca88bc
--- /dev/null
+++ b/paddle/ir/program.cc
@@ -0,0 +1,42 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "paddle/ir/program.h"
+#include "paddle/ir/ir_context.h"
+
+namespace ir {
+Program::~Program() {
+  for (auto op : ops_) {
+    op->destroy();
+  }
+}
+
+void Program::InsertOp(Operation* op) {
+  ops_.push_back(op);
+  op->set_parent_program(this);
+}
+
+Parameter* Program::GetParameter(std::string name) const {
+  if (parameters_.count(name) != 0) {
+    return parameters_.at(name).get();
+  }
+  return nullptr;
+}
+
+void Program::SetParameter(std::string name,
+                           std::unique_ptr<Parameter>&& parameter) {
+  parameters_[name].reset(parameter.release());
+}
+
+}  // namespace ir
diff --git a/paddle/ir/program.h b/paddle/ir/program.h
new file mode 100644
index 0000000000000000000000000000000000000000..8b0a54d77c3213114ec9d892fba17b7044537e11
--- /dev/null
+++ b/paddle/ir/program.h
@@ -0,0 +1,59 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <list>
+#include <unordered_map>
+
+#include "paddle/ir/builtin_attribute.h"
+#include "paddle/ir/operation.h"
+#include "paddle/ir/parameter.h"
+
+namespace ir {
+///
+/// \brief Program is an abstraction of model structure, divided into
+/// computational graphs and weights. At the current stage, a computational
+/// graph is represented in the form of a list<Operation *>. Todo: In the
+/// future, detailed design of control flow operators will be carried out, and
+/// concepts such as basic blocks, closures, and functions will be introduced to
+/// continuously improve Program's ability to represent computational graphs.
+///
+class Program {
+ public:
+  ~Program();
+
+  std::list<Operation*> ops() const { return ops_; }
+
+  size_t parameters_num() const { return parameters_.size(); }
+
+  ///
+  /// \brief Insert the Operation* constructed by Operation::create(...) into
+  /// this Program. NOTE: At this time, the memory management permission of
+  /// Operation* will be owned by this Program. The user does not need to call
+  /// Operation::destroy() manually
+  ///
+  void InsertOp(Operation* op);
+
+  Parameter* GetParameter(std::string name) const;
+
+  void SetParameter(std::string name, std::unique_ptr<Parameter>&& parameter);
+
+ private:
+  std::list<Operation*> ops_;  // owned
+
+  std::unordered_map<std::string, std::unique_ptr<Parameter>> parameters_;
+};
+
+}  // namespace ir
diff --git a/paddle/ir/type.h b/paddle/ir/type.h
index e44157365744bbcdd4517f06c0afa0d221549458..37ceb39a687fea858c4e1fedf50c4715fd7374fc 100644
--- a/paddle/ir/type.h
+++ b/paddle/ir/type.h
@@ -57,7 +57,7 @@ class Type {
 
   const Storage *storage() const { return storage_; }
 
-  const Dialect &dialect() const { return storage_->abstract_type().dialect(); }
+  Dialect &dialect() const { return storage_->abstract_type().dialect(); }
 
   IrContext *ir_context() const;
 
diff --git a/paddle/ir/type_base.h b/paddle/ir/type_base.h
index 0757810e9b7466de323313c97a0a743cd78e78cd..c78cd54c9b392da49355be558e0556a0d70b1f80 100644
--- a/paddle/ir/type_base.h
+++ b/paddle/ir/type_base.h
@@ -62,7 +62,7 @@ class AbstractType {
   ///
   /// \return The dialect this type was registered to.
   ///
-  const Dialect &dialect() const { return dialect_; }
+  Dialect &dialect() const { return const_cast<Dialect &>(dialect_); }
 
   ///
   /// \brief Find the AbstractType instance whose TypeId is type_id from
diff --git a/paddle/ir/value.cc b/paddle/ir/value.cc
index 2f1d5e5616f41d76f9c6b89291d809f70b3009c3..5c066902e303901f66de83a6d2779b4350c3e4c5 100644
--- a/paddle/ir/value.cc
+++ b/paddle/ir/value.cc
@@ -106,6 +106,10 @@ ir::Operation *OpOperandImpl::owner() const { return owner_; }
 
 ir::detail::OpOperandImpl *OpOperandImpl::next_use() { return next_use_; }
 
+ir::Value OpOperandImpl::source() const { return source_; }
+
+void OpOperandImpl::release_source() { source_ = nullptr; }
+
 OpOperandImpl::OpOperandImpl(ir::Value source, ir::Operation *owner)
     : source_(source), owner_(owner) {
   prev_use_addr_ = source.impl()->first_use_addr();
diff --git a/paddle/ir/value_impl.h b/paddle/ir/value_impl.h
index 2fa236dddd83302dc9b046edbe46a3f438e57627..145c937b0cf009e72f8da0c363f448db84d357c4 100644
--- a/paddle/ir/value_impl.h
+++ b/paddle/ir/value_impl.h
@@ -31,6 +31,10 @@ class OpOperandImpl {
 
   ir::detail::OpOperandImpl *next_use();
 
+  ir::Value source() const;
+
+  void release_source();
+
   /// Remove this operand from the current use list.
   void remove_from_ud_chain();
 
diff --git a/test/cpp/ir/CMakeLists.txt b/test/cpp/ir/CMakeLists.txt
index 9d1d9b9d42b752f1eaaa667bfae26d8935a63ea0..0c5385a32fbfadc79959fd6de9a09d131ea0297e 100644
--- a/test/cpp/ir/CMakeLists.txt
+++ b/test/cpp/ir/CMakeLists.txt
@@ -3,4 +3,13 @@ if(WITH_NEWIR)
   cc_test_old(ir_attribute_test SRCS ir_attribute_test.cc DEPS new_ir gtest)
   cc_test_old(ir_value_test SRCS ir_value_test.cc DEPS new_ir gtest)
   cc_test_old(ir_op_test SRCS ir_op_test.cc DEPS new_ir gtest)
+  cc_test_old(
+    ir_program_test
+    SRCS
+    ir_program_test.cc
+    DEPS
+    new_ir
+    pd_dialect
+    phi
+    gtest)
 endif()
diff --git a/test/cpp/ir/ir_attribute_test.cc b/test/cpp/ir/ir_attribute_test.cc
index f4949cd30fb14734c8f377262c988ddcaab92871..8aec0bec4b64dfbcf398f531a30d6f9d966bf200 100644
--- a/test/cpp/ir/ir_attribute_test.cc
+++ b/test/cpp/ir/ir_attribute_test.cc
@@ -58,34 +58,3 @@ TEST(attribute_test, built_in_attribute) {
   EXPECT_EQ(string_attr_cast_1.isa<ir::StrAttribute>(), true);
   EXPECT_EQ(string_attr_cast_1.size() == 8, 1);
 }
-
-TEST(attribute_test, dictionary_attribute) {
-  ir::IrContext *ctx = ir::IrContext::Instance();
-  std::string str_attr1_name = "attr1_name";
-  std::string str_attr1_value = "attr1_value";
-  ir::StrAttribute attr1_name = ir::StrAttribute::get(ctx, str_attr1_name);
-  ir::Attribute attr1_value = ir::StrAttribute::get(ctx, str_attr1_value);
-  std::string str_attr2_name = "attr2_name";
-  std::string str_attr2_value = "attr2_value";
-  ir::StrAttribute attr2_name = ir::StrAttribute::get(ctx, str_attr2_name);
-  ir::Attribute attr2_value = ir::StrAttribute::get(ctx, str_attr2_value);
-
-  std::map<ir::StrAttribute, ir::Attribute> named_attr1;
-  named_attr1.insert(
-      std::pair<ir::StrAttribute, ir::Attribute>(attr1_name, attr1_value));
-  named_attr1.insert(
-      std::pair<ir::StrAttribute, ir::Attribute>(attr2_name, attr2_value));
-  ir::DictionaryAttribute dic_attr1 =
-      ir::DictionaryAttribute::get(ctx, named_attr1);
-  std::map<ir::StrAttribute, ir::Attribute> named_attr2;
-  named_attr2.insert(
-      std::pair<ir::StrAttribute, ir::Attribute>(attr2_name, attr2_value));
-  named_attr2.insert(
-      std::pair<ir::StrAttribute, ir::Attribute>(attr1_name, attr1_value));
-  ir::DictionaryAttribute dic_attr2 =
-      ir::DictionaryAttribute::get(ctx, named_attr2);
-
-  EXPECT_EQ(dic_attr1, dic_attr2);
-  EXPECT_EQ(attr1_value, dic_attr1.GetValue(attr1_name));
-  EXPECT_EQ(attr2_value, dic_attr1.GetValue(attr2_name));
-}
diff --git a/test/cpp/ir/ir_op_test.cc b/test/cpp/ir/ir_op_test.cc
index 342eb5330e504b0d11aabe5abd795cb53cbb85c3..f6fbff5dd3b0ef4717388df53b3a091947cb7af5 100644
--- a/test/cpp/ir/ir_op_test.cc
+++ b/test/cpp/ir/ir_op_test.cc
@@ -99,15 +99,14 @@ class TestDialect : public ir::Dialect {
   void initialize() { RegisterOps<Operation1, Operation2>(); }
 };
 
-ir::DictionaryAttribute CreateAttribute(std::string attribute_name,
-                                        std::string attribute) {
+ir::AttributeMap CreateAttributeMap(std::string attribute_name,
+                                    std::string attribute) {
   ir::IrContext *ctx = ir::IrContext::Instance();
-  ir::StrAttribute attr_name = ir::StrAttribute::get(ctx, attribute_name);
   ir::Attribute attr_value = ir::StrAttribute::get(ctx, attribute);
-  std::map<ir::StrAttribute, ir::Attribute> named_attr;
-  named_attr.insert(
-      std::pair<ir::StrAttribute, ir::Attribute>(attr_name, attr_value));
-  return ir::DictionaryAttribute::get(ctx, named_attr);
+  ir::AttributeMap attr_map;
+  attr_map.insert(
+      std::pair<std::string, ir::Attribute>(attribute_name, attr_value));
+  return attr_map;
 }
 
 TEST(op_test, op_test) {
@@ -137,7 +136,7 @@ TEST(op_test, op_test) {
   ir::Operation *op =
       ir::Operation::create(op_inputs,
                             op_output_types,
-                            CreateAttribute("op1_name", "op1_attr"),
+                            CreateAttributeMap("op1_name", "op1_attr"),
                             op2_info);
 
   if (op->HasTrait<ReadOnlyTrait>()) {
diff --git a/test/cpp/ir/ir_program_test.cc b/test/cpp/ir/ir_program_test.cc
new file mode 100644
index 0000000000000000000000000000000000000000..91e0fbf9bf4f89a12929efb17086bbcfaf0cc464
--- /dev/null
+++ b/test/cpp/ir/ir_program_test.cc
@@ -0,0 +1,206 @@
+// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <gtest/gtest.h>
+
+#include "paddle/fluid/dialect/pd_dialect.h"
+#include "paddle/fluid/dialect/pd_type.h"
+#include "paddle/fluid/dialect/utils.h"
+#include "paddle/ir/builtin_attribute.h"
+#include "paddle/ir/builtin_dialect.h"
+#include "paddle/ir/builtin_op.h"
+#include "paddle/ir/builtin_type.h"
+#include "paddle/ir/ir_context.h"
+#include "paddle/ir/program.h"
+#include "paddle/ir/utils.h"
+#include "paddle/phi/core/meta_tensor.h"
+#include "paddle/phi/infermeta/binary.h"
+#include "paddle/phi/kernels/elementwise_add_kernel.h"
+
+class AddOp : public ir::Op<AddOp> {
+ public:
+  using Op::Op;
+  static const char *name() { return "Add"; }
+  static const char **attributes_name_;
+  static uint32_t attributes_num() { return 0; }
+};
+const char **AddOp::attributes_name_ = nullptr;
+
+TEST(program_test, program) {
+  // (1) Init environment.
+  ir::IrContext *ctx = ir::IrContext::Instance();
+  ir::Dialect *builtin_dialect =
+      ctx->GetOrRegisterDialect<ir::BuiltinDialect>();
+  builtin_dialect->RegisterOp<AddOp>();
+  ir::Dialect *paddle_dialect =
+      ctx->GetOrRegisterDialect<paddle::dialect::PaddleDialect>();
+
+  // (2) Create an empty program object
+  ir::Program program;
+  //   ir::Program *program = new ir::Program();
+  EXPECT_EQ(program.ops().size() == 0, true);
+
+  // (3) Create a float32 DenseTensor Parameter and save into Program
+  ir::Type fp32_dtype = ir::Float32Type::get(ctx);
+  paddle::dialect::DenseTensorTypeStorage::Dim dims = {2, 2};
+  paddle::dialect::DenseTensorTypeStorage::DataLayout data_layout =
+      paddle::dialect::DenseTensorTypeStorage::DataLayout::NCHW;
+  paddle::dialect::DenseTensorTypeStorage::LoD lod = {{0, 1, 2}};
+  size_t offset = 0;
+  ir::Type dense_tensor_dtype = paddle::dialect::DenseTensorType::get(
+      ctx, fp32_dtype, dims, data_layout, lod, offset);
+
+  std::vector<float> data_a = {1, 2, 3, 4};
+  std::unique_ptr<ir::Parameter> parameter_a =
+      std::make_unique<ir::Parameter>(reinterpret_cast<void *>(data_a.data()),
+                                      4 * sizeof(float),
+                                      dense_tensor_dtype);
+  program.SetParameter("a", std::move(parameter_a));
+  EXPECT_EQ(program.parameters_num() == 1, true);
+
+  std::vector<float> data_b = {5, 6, 7, 8};
+  std::unique_ptr<ir::Parameter> parameter_b =
+      std::make_unique<ir::Parameter>(reinterpret_cast<void *>(data_b.data()),
+                                      4 * sizeof(float),
+                                      dense_tensor_dtype);
+  program.SetParameter("b", std::move(parameter_b));
+  EXPECT_EQ(program.parameters_num() == 2, true);
+
+  // (4) Def a = GetParameterOp("a"), and create DenseTensor for a.
+  std::string op1_name =
+      builtin_dialect->name() + "." + std::string(ir::GetParameterOp::name());
+  ir::OpInfoImpl *op1_info = ctx->GetRegisteredOpInfo(op1_name);
+  std::unordered_map<std::string, ir::Attribute> op1_attribute{
+      {"parameter_name", ir::StrAttribute::get(ctx, "a")}};
+  ir::Operation *op1 =
+      ir::Operation::create({}, {dense_tensor_dtype}, op1_attribute, op1_info);
+
+  program.InsertOp(op1);
+
+  EXPECT_EQ(op1->GetResultByIndex(0).type().dialect().id(),
+            paddle_dialect->id());
+  using Interface = paddle::dialect::ParameterConvertInterface;
+  Interface *a_interface = op1->GetResultByIndex(0)
+                               .type()
+                               .dialect()
+                               .GetRegisteredInterface<Interface>();
+  std::shared_ptr<paddle::framework::Variable> a_var =
+      a_interface->ParameterToVariable(program.GetParameter("a"));
+  const phi::DenseTensor &a_tensor = a_var->Get<phi::DenseTensor>();
+  EXPECT_EQ(a_tensor.numel(), 4);
+  EXPECT_EQ(a_tensor.dims(), phi::DDim(dims.data(), dims.size()));
+  EXPECT_EQ(a_tensor.dtype(), paddle::dialect::TransToPhiDataType(fp32_dtype));
+  EXPECT_EQ(a_tensor.layout(),
+            paddle::dialect::TransToPhiDataLayout(data_layout));
+  EXPECT_EQ(a_tensor.lod(), lod);
+  EXPECT_EQ(a_tensor.offset(), offset);
+  for (int64_t i = 0; i < a_tensor.numel(); i++) {
+    EXPECT_EQ(*(a_tensor.data<float>() + i), data_a[i]);
+  }
+
+  // (5) Def b = GetParameterOp("b"), and create DenseTensor for b.
+  std::string op2_name =
+      builtin_dialect->name() + "." + std::string(ir::GetParameterOp::name());
+  ir::OpInfoImpl *op2_info = ctx->GetRegisteredOpInfo(op2_name);
+  std::unordered_map<std::string, ir::Attribute> op2_attribute{
+      {"parameter_name", ir::StrAttribute::get(ctx, "b")}};
+  ir::Operation *op2 =
+      ir::Operation::create({}, {dense_tensor_dtype}, op2_attribute, op2_info);
+  program.InsertOp(op2);
+
+  EXPECT_EQ(op2->GetResultByIndex(0).type().dialect().id(),
+            paddle_dialect->id());
+  Interface *b_interface = op2->GetResultByIndex(0)
+                               .type()
+                               .dialect()
+                               .GetRegisteredInterface<Interface>();
+  std::shared_ptr<paddle::framework::Variable> b_var =
+      b_interface->ParameterToVariable(program.GetParameter("b"));
+  const phi::DenseTensor &b_tensor = b_var->Get<phi::DenseTensor>();
+  EXPECT_EQ(b_tensor.numel(), 4);
+  EXPECT_EQ(b_tensor.dims(), phi::DDim(dims.data(), dims.size()));
+  EXPECT_EQ(b_tensor.dtype(), paddle::dialect::TransToPhiDataType(fp32_dtype));
+  EXPECT_EQ(b_tensor.layout(),
+            paddle::dialect::TransToPhiDataLayout(data_layout));
+  EXPECT_EQ(b_tensor.lod(), lod);
+  EXPECT_EQ(b_tensor.offset(), offset);
+  for (int64_t i = 0; i < b_tensor.numel(); i++) {
+    EXPECT_EQ(*(b_tensor.data<float>() + i), data_b[i]);
+  }
+
+  // (6) Def c = AddOp(a, b), execute this op.
+  std::string op3_name =
+      builtin_dialect->name() + "." + std::string(AddOp::name());
+  ir::OpInfoImpl *op3_info = ctx->GetRegisteredOpInfo(op3_name);
+  std::unordered_map<std::string, ir::Attribute> op3_attribute;
+  ir::Operation *op3 = ir::Operation::create(
+      {op1->GetResultByIndex(0), op2->GetResultByIndex(0)},
+      {dense_tensor_dtype},
+      op3_attribute,
+      op3_info);
+  program.InsertOp(op3);
+
+  phi::CPUContext *dev_ctx = static_cast<phi::CPUContext *>(
+      paddle::platform::DeviceContextPool::Instance().Get(
+          paddle::platform::CPUPlace()));
+  phi::DenseTensor c_tensor =
+      phi::Add<float, phi::CPUContext>(*dev_ctx, a_tensor, b_tensor);
+  std::shared_ptr<paddle::framework::Variable> variable_c =
+      std::make_shared<paddle::framework::Variable>();
+  auto *dst_tensor = variable_c->GetMutable<phi::DenseTensor>();
+  *dst_tensor = c_tensor;
+  EXPECT_EQ(dst_tensor->numel(), b_tensor.numel());
+  EXPECT_EQ(dst_tensor->dims(), b_tensor.dims());
+  EXPECT_EQ(dst_tensor->dtype(), b_tensor.dtype());
+  EXPECT_EQ(dst_tensor->layout(), b_tensor.layout());
+  EXPECT_EQ(dst_tensor->lod(), b_tensor.lod());
+  EXPECT_EQ(dst_tensor->offset(), b_tensor.offset());
+  for (int64_t i = 0; i < dst_tensor->numel(); i++) {
+    EXPECT_EQ(*(dst_tensor->data<float>() + i), data_a[i] + data_b[i]);
+  }
+
+  // (7) Def SetParameterOp(c, "c")
+  std::string op4_name =
+      builtin_dialect->name() + "." + std::string(ir::SetParameterOp::name());
+  ir::OpInfoImpl *op4_info = ctx->GetRegisteredOpInfo(op4_name);
+  std::unordered_map<std::string, ir::Attribute> op4_attribute{
+      {"parameter_name", ir::StrAttribute::get(ctx, "c")}};
+  ir::Operation *op4 = ir::Operation::create(
+      {op3->GetResultByIndex(0)}, {}, op4_attribute, op4_info);
+  program.InsertOp(op4);
+
+  EXPECT_EQ(op4->GetOperandByIndex(0).impl()->source().type().dialect().id(),
+            paddle_dialect->id());
+  Interface *c_interface = op4->GetOperandByIndex(0)
+                               .impl()
+                               ->source()
+                               .type()
+                               .dialect()
+                               .GetRegisteredInterface<Interface>();
+  //   ir::Parameter *parameter_c =
+  //       c_interface->VariableToParameter(variable_c.get());
+  std::unique_ptr<ir::Parameter> parameter_c =
+      c_interface->VariableToParameter(variable_c.get());
+  EXPECT_EQ(parameter_c->type(), dense_tensor_dtype);
+  for (int64_t i = 0; i < dst_tensor->numel(); i++) {
+    EXPECT_EQ(*(dst_tensor->data<float>() + i),
+              *(static_cast<float *>(parameter_c->data()) + i));
+  }
+  program.SetParameter("c", std::move(parameter_c));
+
+  // (8) Traverse Program
+  std::list<ir::Operation *> ops = program.ops();
+  EXPECT_EQ(ops.size() == 4, true);
+  EXPECT_EQ(program.parameters_num() == 3, true);
+}
diff --git a/test/cpp/ir/ir_value_test.cc b/test/cpp/ir/ir_value_test.cc
index 102619905fe7b7dc729c86c7e2e285079bcc67af..e0f0d83312a8903784c371831ef8361d07c80207 100644
--- a/test/cpp/ir/ir_value_test.cc
+++ b/test/cpp/ir/ir_value_test.cc
@@ -23,16 +23,14 @@
 // This unittest is used to test the construction interfaces of value class and
 // operation. The constructed test scenario is: a = OP1(); b = OP2(); c = OP3(a,
 // b); d, e, f, g, h, i, j = OP4(a, c);
-
-ir::DictionaryAttribute CreateAttribute(std::string attribute_name,
-                                        std::string attribute) {
+ir::AttributeMap CreateAttributeMap(std::string attribute_name,
+                                    std::string attribute) {
   ir::IrContext *ctx = ir::IrContext::Instance();
-  ir::StrAttribute attr_name = ir::StrAttribute::get(ctx, attribute_name);
   ir::Attribute attr_value = ir::StrAttribute::get(ctx, attribute);
-  std::map<ir::StrAttribute, ir::Attribute> named_attr;
-  named_attr.insert(
-      std::pair<ir::StrAttribute, ir::Attribute>(attr_name, attr_value));
-  return ir::DictionaryAttribute::get(ctx, named_attr);
+  ir::AttributeMap attr_map;
+  attr_map.insert(
+      std::pair<std::string, ir::Attribute>(attribute_name, attr_value));
+  return attr_map;
 }
 
 TEST(value_test, value_test) {
@@ -43,7 +41,7 @@ TEST(value_test, value_test) {
   ir::Operation *op1 =
       ir::Operation::create(op1_inputs,
                             op1_output_types,
-                            CreateAttribute("op1_name", "op1_attr"),
+                            CreateAttributeMap("op1_name", "op1_attr"),
                             nullptr);
   std::cout << op1->print() << std::endl;
   // 2. Construct OP2: b = OP2();
@@ -52,7 +50,7 @@ TEST(value_test, value_test) {
   ir::Operation *op2 =
       ir::Operation::create(op2_inputs,
                             op2_output_types,
-                            CreateAttribute("op2_name", "op2_attr"),
+                            CreateAttributeMap("op2_name", "op2_attr"),
                             nullptr);
   std::cout << op2->print() << std::endl;
   // 3. Construct OP3: c = OP3(a, b);
@@ -62,7 +60,7 @@ TEST(value_test, value_test) {
   ir::Operation *op3 =
       ir::Operation::create(op3_inputs,
                             op3_output_types,
-                            CreateAttribute("op3_name", "op3_attr"),
+                            CreateAttributeMap("op3_name", "op3_attr"),
                             nullptr);
   std::cout << op3->print() << std::endl;
   // 4. Construct OP4: d, e, f, g, h, i, j = OP4(a, c);
@@ -75,7 +73,7 @@ TEST(value_test, value_test) {
   ir::Operation *op4 =
       ir::Operation::create(op4_inputs,
                             op4_output_types,
-                            CreateAttribute("op4_name", "op4_attr"),
+                            CreateAttributeMap("op4_name", "op4_attr"),
                             nullptr);
   std::cout << op4->print() << std::endl;
 
diff --git a/test/cpp/ir/type_test.cc b/test/cpp/ir/type_test.cc
index d21afdcb80a59f324c6410d1643a140e025bac20..8613c9d6afa67a6212305076ea1fc703edcad01c 100644
--- a/test/cpp/ir/type_test.cc
+++ b/test/cpp/ir/type_test.cc
@@ -124,46 +124,10 @@ TEST(type_test, built_in_type) {
             &ir::AbstractType::lookup(int64_1.type_id(), ctx));
   EXPECT_EQ(ir::Int64Type::classof(int64_1), 1);
 
-  // Test 2: Test the parameteric built-in type of IrContext.
-  ir::DenseTensorTypeStorage::Dim dims = {1, 2, 3};
-  ir::DenseTensorTypeStorage::DataLayout data_layout =
-      ir::DenseTensorTypeStorage::DataLayout::NCHW;
-  ir::DenseTensorTypeStorage::LoD lod = {{1, 2, 3}, {4, 5, 6}};
-  size_t offset = 0;
-
-  ir::Type dense_tensor_1 =
-      ir::DenseTensorType::get(ctx, fp32_1, dims, data_layout, lod, offset);
-  ir::Type dense_tensor_2 =
-      ir::DenseTensorType::get(ctx, fp32_2, dims, data_layout, lod, offset);
-  ir::Type dense_tensor_3 =
-      ir::DenseTensorType::get(ctx, fp32_1, dims, data_layout, lod, 2);
-
-  EXPECT_EQ(dense_tensor_1, dense_tensor_2);
-  EXPECT_NE(dense_tensor_1, dense_tensor_3);
-  EXPECT_EQ(dense_tensor_1.type_id(), dense_tensor_2.type_id());
-  EXPECT_EQ(ir::DenseTensorType::classof(dense_tensor_1), 1);
-
-  ir::DenseTensorType dense_tensor_4 =
-      ir::DenseTensorType::get(ctx, fp32_1, dims, data_layout, lod, 2);
-  EXPECT_EQ(dense_tensor_4.offset() == 2, 1);
-  EXPECT_EQ(dense_tensor_4.dtype().isa<ir::Float32Type>(), true);
-  EXPECT_EQ(dense_tensor_4.data_layout(), data_layout);
-
-  // Test 3: Test isa and dyn_cast.
+  // Test 2: Test isa and dyn_cast.
   EXPECT_EQ(fp16_1.isa<ir::Float16Type>(), true);
   EXPECT_EQ(fp16_1.isa<ir::Float32Type>(), false);
-  EXPECT_EQ(fp16_1.isa<ir::DenseTensorType>(), false);
   EXPECT_EQ(fp16_1.isa<ir::Type>(), true);
-  EXPECT_EQ(dense_tensor_1.isa<ir::DenseTensorType>(), true);
-
-  ir::DenseTensorType dense_tensor_cast_1 =
-      dense_tensor_1.dyn_cast<ir::DenseTensorType>();
-  EXPECT_EQ(dense_tensor_cast_1.isa<ir::DenseTensorType>(), true);
-  EXPECT_EQ(dense_tensor_cast_1.offset() == 0, 1);
-  const ir::DenseTensorType dense_tensor_cast_2 =
-      ir::dyn_cast<ir::DenseTensorType>(dense_tensor_1);
-  EXPECT_EQ(dense_tensor_cast_2.isa<ir::DenseTensorType>(), true);
-  EXPECT_EQ(dense_tensor_cast_2.offset() == 0, 1);
 }
 
 // Customize a parameterized TypeStorage IntegerTypeStorage.