diff --git a/paddle/fluid/distributed/fleet_executor/compute_interceptor.cc b/paddle/fluid/distributed/fleet_executor/compute_interceptor.cc
index 8da1ef87814deb..5e2be03108294c 100644
--- a/paddle/fluid/distributed/fleet_executor/compute_interceptor.cc
+++ b/paddle/fluid/distributed/fleet_executor/compute_interceptor.cc
@@ -176,7 +176,7 @@ bool ComputeInterceptor::IsInputReady() {
       flag = flag && (ready_size_map.at(i) != 0);
     }
     if (flag) {
-      if (scope_id_to_finish_flag.empty()) {
+      if (scope_id_to_finish_flag.empty()) {  // NOLINT
         cur_scope_id_ = i;
         return true;
       } else if (scope_id_to_finish_flag.find(i) !=
@@ -303,7 +303,7 @@ void ComputeInterceptor::RunOps() {
                           cur_scope_id_));
   }
 
-  if (!cores_.empty()) {
+  if (!cores_.empty()) {  // NOLINT
     cores_[cur_scope_id_]->Run(/*feed_names=*/{}, /*need_fetch=*/false);
   } else {
     for (auto op : node_->ops()) {
diff --git a/paddle/fluid/distributed/fleet_executor/dist_model.cc b/paddle/fluid/distributed/fleet_executor/dist_model.cc
index a1fd38295319ed..4c19069b33705b 100644
--- a/paddle/fluid/distributed/fleet_executor/dist_model.cc
+++ b/paddle/fluid/distributed/fleet_executor/dist_model.cc
@@ -215,7 +215,7 @@ bool DistModel::Init() {
 }
 
 bool DistModel::PreparePlace() {
-  if (config_.place == "GPU") {
+  if (config_.place == "GPU") {  // NOLINT
     place_ = paddle::platform::CUDAPlace(config_.device_id);
   } else if (config_.place == "CPU") {
     place_ = paddle::platform::CPUPlace();
diff --git a/paddle/fluid/eager/custom_operator/custom_operator_utils.cc b/paddle/fluid/eager/custom_operator/custom_operator_utils.cc
index b843e081c29bed..a9272053346a77 100644
--- a/paddle/fluid/eager/custom_operator/custom_operator_utils.cc
+++ b/paddle/fluid/eager/custom_operator/custom_operator_utils.cc
@@ -558,7 +558,7 @@ std::vector<std::vector<phi::DDim>> RunInferShapeFn(
     out_dims =
         RunInferShapeFunc(ctx, infer_shape_func, inputs, outputs, inplace_map);
   } else {
-    if (is_forward) {
+    if (is_forward) {  // NOLINT
       out_dims = RunDefaultInferShapeFunc(ctx, inputs, outputs, inplace_map);
     } else {
       out_dims =
@@ -592,7 +592,7 @@ std::vector<std::vector<phi::DataType>> RunInferDtypeFn(
     out_dtypes =
         RunInferDtypeFunc(ctx, infer_dtype_func, inputs, outputs, inplace_map);
   } else {
-    if (is_forward) {
+    if (is_forward) {  // NOLINT
       out_dtypes = RunDefaultInferDtypeFunc(ctx, inputs, outputs, inplace_map);
     } else {
       out_dtypes =
diff --git a/paddle/fluid/eager/grad_tensor_holder.cc b/paddle/fluid/eager/grad_tensor_holder.cc
index dac55f8f5462f1..47f41b5a4f93b7 100644
--- a/paddle/fluid/eager/grad_tensor_holder.cc
+++ b/paddle/fluid/eager/grad_tensor_holder.cc
@@ -79,7 +79,7 @@ void GradTensorHolder::CopyValueFromTensor(size_t slot_id,
     // Create new tensor->impl and fill it with 1.0
     if (t.defined()) {
       // Fill 1.0, use full to support complex, one_like don't support it.
-      if (t.is_dense_tensor()) {
+      if (t.is_dense_tensor()) {  // NOLINT
         buffer_[slot_id][rank] =
             paddle::experimental::full(t.shape(), 1, t.dtype(), t.place());
       } else if (t.is_sparse_csr_tensor() || t.is_sparse_coo_tensor()) {
diff --git a/paddle/fluid/framework/data_feed.cc b/paddle/fluid/framework/data_feed.cc
index 470b4962ba34c0..2d1b9842e82a31 100644
--- a/paddle/fluid/framework/data_feed.cc
+++ b/paddle/fluid/framework/data_feed.cc
@@ -1813,7 +1813,7 @@ int PaddleBoxDataFeed::Next() {
     this->batch_size_ = index;
     VLOG(3) << "pv_batch_size_=" << this->batch_size_
             << ", thread_id=" << thread_id_;
-    if (this->batch_size_ != 0) {
+    if (this->batch_size_ != 0) {  // NOLINT
       PutToFeedVec(pv_vec);
     } else {
       VLOG(3) << "finish reading, output_pv_channel_ size="
@@ -2113,7 +2113,7 @@ void SlotRecordInMemoryDataFeed::Init(const DataFeedDesc& data_feed_desc) {
   finish_init_ = true;
   input_type_ = data_feed_desc.input_type();
   size_t pos = pipe_command_.find(".so");
-  if (pos != std::string::npos) {
+  if (pos != std::string::npos) {  // NOLINT
     pos = pipe_command_.rfind('|');
     if (pos == std::string::npos) {
       so_parser_name_ = pipe_command_;
@@ -2129,7 +2129,7 @@ void SlotRecordInMemoryDataFeed::Init(const DataFeedDesc& data_feed_desc) {
 #if defined(PADDLE_WITH_GPU_GRAPH) && defined(PADDLE_WITH_HETERPS)
   gpu_graph_data_generator_.SetConfig(data_feed_desc);
 #endif
-  if (gpu_graph_mode_) {
+  if (gpu_graph_mode_) {  // NOLINT
     train_mode_ = true;
   } else {
     train_mode_ = data_feed_desc.graph_config().gpu_graph_training();
@@ -2780,7 +2780,7 @@ int SlotRecordInMemoryDataFeed::Next() {
     this->batch_size_ = batch.second;
     VLOG(3) << "batch_size_=" << this->batch_size_
             << ", thread_id=" << thread_id_;
-    if (this->batch_size_ != 0) {
+    if (this->batch_size_ != 0) {  // NOLINT
       PutToFeedVec(&records_[batch.first], this->batch_size_);
     } else {
       VLOG(3) << "finish reading for heterps, batch size zero, thread_id="
diff --git a/paddle/fluid/framework/data_set.cc b/paddle/fluid/framework/data_set.cc
index f31cba2f1ecbbe..4c15c26359aeb5 100644
--- a/paddle/fluid/framework/data_set.cc
+++ b/paddle/fluid/framework/data_set.cc
@@ -966,7 +966,7 @@ void DatasetImpl<T>::DynamicAdjustChannelNum(int channel_num,
     CHECK(output_channels_data_size == 0);  // NOLINT
     cur_channel = 1;
   }
-  if (cur_channel == 0) {
+  if (cur_channel == 0) {  // NOLINT
     origin_channels = &multi_output_channel_;
     other_channels = &multi_consume_channel_;
     origin_pv_channels = &multi_pv_output_;
@@ -1111,8 +1111,8 @@ void DatasetImpl<T>::CreateReaders() {
     if (input_pv_channel_ != nullptr) {
       readers_[i]->SetInputPvChannel(input_pv_channel_.get());
     }
-    if (cur_channel_ == 0 &&
-        static_cast<size_t>(channel_idx) < multi_output_channel_.size()) {
+    if (cur_channel_ == 0 && static_cast<size_t>(channel_idx) <
+                                 multi_output_channel_.size()) {  // NOLINT
       readers_[i]->SetOutputChannel(multi_output_channel_[channel_idx].get());
       readers_[i]->SetConsumeChannel(multi_consume_channel_[channel_idx].get());
       readers_[i]->SetOutputPvChannel(multi_pv_output_[channel_idx].get());
@@ -1722,7 +1722,7 @@ void MultiSlotDataset::PreprocessChannel(
     const std::set<std::string>& slots_to_replace,
     std::unordered_set<uint16_t>& index_slots) {  // NOLINT
   int out_channel_size = 0;
-  if (cur_channel_ == 0) {
+  if (cur_channel_ == 0) {  // NOLINT
     for (auto& item : multi_output_channel_) {
       out_channel_size += static_cast<int>(item->Size());
     }
@@ -1757,7 +1757,7 @@ void MultiSlotDataset::PreprocessChannel(
       input_channel_->ReadAll(slots_shuffle_original_data_);
     } else {
       CHECK(out_channel_size > 0);  // NOLINT
-      if (cur_channel_ == 0) {
+      if (cur_channel_ == 0) {      // NOLINT
         for (auto& item : multi_output_channel_) {
           std::vector<Record> vec_data;
           item->Close();
@@ -1792,7 +1792,7 @@ void MultiSlotDataset::PreprocessChannel(
   } else {
     // if already have original data for slots shuffle, clear channel
     input_channel_->Clear();
-    if (cur_channel_ == 0) {
+    if (cur_channel_ == 0) {  // NOLINT
       for (auto& item : multi_output_channel_) {
         if (!item) {
           continue;
@@ -1809,7 +1809,7 @@ void MultiSlotDataset::PreprocessChannel(
     }
   }
   int end_size = 0;
-  if (cur_channel_ == 0) {
+  if (cur_channel_ == 0) {  // NOLINT
     for (auto& item : multi_output_channel_) {
       if (!item) {
         continue;
diff --git a/paddle/fluid/framework/details/nan_inf_utils_detail.cc b/paddle/fluid/framework/details/nan_inf_utils_detail.cc
index 551a10f1ccacd1..d18cee16b19a64 100644
--- a/paddle/fluid/framework/details/nan_inf_utils_detail.cc
+++ b/paddle/fluid/framework/details/nan_inf_utils_detail.cc
@@ -264,7 +264,7 @@ void CheckOpHasNanOrInf(const framework::OperatorBase& op,
 
   if (IsSkipOp(op)) return;
 
-  if (op_var_nan_inf_white_list().count(op.Type()) == 0) {
+  if (op_var_nan_inf_white_list().count(op.Type()) == 0) {  // NOLINT
     // NOTE. vname may destruct in the end of this func.
     for (auto& vname : op.OutputVars(true)) {
       auto* var = exec_scope.FindVar(vname);
diff --git a/paddle/fluid/framework/dist_multi_trainer.cc b/paddle/fluid/framework/dist_multi_trainer.cc
index 6fd95267ef6ab3..119b6e569cef32 100644
--- a/paddle/fluid/framework/dist_multi_trainer.cc
+++ b/paddle/fluid/framework/dist_multi_trainer.cc
@@ -157,7 +157,7 @@ void DistMultiTrainer::Run() {
   std::vector<std::future<void>> wait_futures;
   CHECK_EQ(static_cast<int>(pool.size()), thread_num_);
   for (int i = 0; i < thread_num_; ++i) {
-    if (!debug_) {
+    if (!debug_) {  // NOLINT
       wait_futures.emplace_back(
           pool[i]->Run([this, i]() { workers_[i]->TrainFiles(); }));
     } else {
diff --git a/paddle/fluid/framework/executor.cc b/paddle/fluid/framework/executor.cc
index d935e9ea066bd3..fbc2565e755fa5 100644
--- a/paddle/fluid/framework/executor.cc
+++ b/paddle/fluid/framework/executor.cc
@@ -99,7 +99,7 @@ void Executor::CreateVariables(const ProgramDesc& pdesc,
   while (ancestor_scope->parent()) {
     ancestor_scope = ancestor_scope->parent();
   }
-  if (ancestor_scope != scope) {
+  if (ancestor_scope != scope) {  // NOLINT
     for (auto& var : global_block.AllVars()) {
       if (var->Name() == framework::kEmptyVarName) {
         continue;
diff --git a/paddle/fluid/framework/heter_section_worker.cc b/paddle/fluid/framework/heter_section_worker.cc
index 24be8e04d8d507..938a67712fdf8e 100644
--- a/paddle/fluid/framework/heter_section_worker.cc
+++ b/paddle/fluid/framework/heter_section_worker.cc
@@ -126,7 +126,7 @@ void HeterSectionWorker::Initialize(const TrainerDesc& desc) {
   bool is_first_stage = (pipeline_stage_ == 0);
   bool is_last_stage = (pipeline_stage_ + 1 == num_pipeline_stages_);
 
-  if (is_first_stage) {
+  if (is_first_stage) {  // NOLINT
     for (auto& op_desc : program_->Block(0).AllOps()) {
       auto op = std::move(OpRegistry::CreateOp(*op_desc));
       auto op_type = op->Type();
diff --git a/paddle/fluid/framework/infershape_utils.cc b/paddle/fluid/framework/infershape_utils.cc
index 88f0b496a8e4c6..f897a075353a73 100644
--- a/paddle/fluid/framework/infershape_utils.cc
+++ b/paddle/fluid/framework/infershape_utils.cc
@@ -658,7 +658,7 @@ CompatInferMetaContext BuildInferMetaContext(InferShapeContext* ctx,
         if (attr_ptr && !is_attr_var) {
           auto& attr = *attr_ptr;
           switch (AttrTypeID(attr)) {
-            case framework::proto::AttrType::INTS:
+            case framework::proto::AttrType::INTS:  // NOLINT
               infer_meta_context.EmplaceBackAttr(std::move(
                   phi::IntArray(PADDLE_GET_CONST(std::vector<int32_t>, attr))));
               break;
@@ -836,7 +836,7 @@ CompatInferMetaContext BuildInferMetaContext(InferShapeContext* ctx,
                       attr_names[i]));
               }
               break;
-            case phi::AttributeType::FLOAT32S:
+            case phi::AttributeType::FLOAT32S:  // NOLINT
               infer_meta_context.EmplaceBackAttr(
                   PADDLE_GET_CONST(std::vector<float>, attr));
               break;
diff --git a/paddle/fluid/framework/ir/coalesce_grad_tensor_pass.cc b/paddle/fluid/framework/ir/coalesce_grad_tensor_pass.cc
index 44cb004fec1729..966f4ea14967d9 100644
--- a/paddle/fluid/framework/ir/coalesce_grad_tensor_pass.cc
+++ b/paddle/fluid/framework/ir/coalesce_grad_tensor_pass.cc
@@ -134,7 +134,7 @@ class CoalesceGradTensorPass : public ir::Pass {
 
     auto &pinned_var_set =
         graph->GetOrInit<details::PinnedVars>(details::kPinnedVars);
-    if (IsUnifiedDtype(p_g_dense_grad, vars_info)) {
+    if (IsUnifiedDtype(p_g_dense_grad, vars_info)) {  // NOLINT
       RecordGradients(p_g_dense_grad, vars_info, &pinned_var_set);
       CoalesceTensors(vars_info, p_g_dense_grad, &result);
     } else {
diff --git a/paddle/fluid/framework/ir/generate_pass_tester.cc b/paddle/fluid/framework/ir/generate_pass_tester.cc
index 760e1e8ce4ef83..58a3741a924aab 100644
--- a/paddle/fluid/framework/ir/generate_pass_tester.cc
+++ b/paddle/fluid/framework/ir/generate_pass_tester.cc
@@ -25,7 +25,7 @@ REGISTER_GENERATE_PASS(generate_fc_fuse) {
       VLOG(3) << "exec lambda func.";
       auto mul = OP_(mul)({{"X", x}, {"Y", y}}).Out("Out");
       auto ewadd = OP_(elementwise_add)({{"X", mul}, {"Y", z}}).Out("Out");
-      if (with_relu) {
+      if (with_relu) {  // NOLINT
         return OP_(relu)({"X", ewadd}).Out("Out");
       } else {
         return ewadd;
diff --git a/paddle/fluid/framework/ir/identity_op_clean_pass.cc b/paddle/fluid/framework/ir/identity_op_clean_pass.cc
index ab9df0ae4abee7..55316c1b823105 100644
--- a/paddle/fluid/framework/ir/identity_op_clean_pass.cc
+++ b/paddle/fluid/framework/ir/identity_op_clean_pass.cc
@@ -70,7 +70,7 @@ FindUselessOpPattern::FindUselessOpPattern(PDPattern* pattern,
               auto in_dtype = x->Op()->GetAttrIfExists<int>("in_dtype");
               auto out_dtype = x->Op()->GetAttrIfExists<int>("out_dtype");
               return in_dtype == out_dtype;
-            } else if (op_type == "c_identity") {
+            } else if (op_type == "c_identity") {  // NOLINT
               return true;
             } else if (op_type == "assign") {
               const auto& in_name = x->Op()->Input("X")[0];
diff --git a/paddle/fluid/framework/ir/mkldnn/compute_propagate_scales_mkldnn_pass_tester.cc b/paddle/fluid/framework/ir/mkldnn/compute_propagate_scales_mkldnn_pass_tester.cc
index 7a811aae50e236..e5d270a7d57b61 100644
--- a/paddle/fluid/framework/ir/mkldnn/compute_propagate_scales_mkldnn_pass_tester.cc
+++ b/paddle/fluid/framework/ir/mkldnn/compute_propagate_scales_mkldnn_pass_tester.cc
@@ -161,7 +161,7 @@ class ComputePropagateScalesMkldnnPassTest : public testing::Test {
           begin(wh[i]),
           end(wh[i]),
           wh_tensor->mutable_data<float>(phi::CPUPlace()) + i * wh[0].size());
-    if (type == "gru") {
+    if (type == "gru") {  // NOLINT
       ComputeGruWeightScales(
           graph, &scope, wx_name, wh_name, &var_quant_scales);
     } else {
diff --git a/paddle/fluid/framework/ir/mkldnn/cpu_quantize_pass_tester.cc b/paddle/fluid/framework/ir/mkldnn/cpu_quantize_pass_tester.cc
index 5ffc2db337ada8..015bc6842788df 100644
--- a/paddle/fluid/framework/ir/mkldnn/cpu_quantize_pass_tester.cc
+++ b/paddle/fluid/framework/ir/mkldnn/cpu_quantize_pass_tester.cc
@@ -61,7 +61,7 @@ void SetOp(ProgramDesc* prog,
     op->SetOutput("Output", {outputs[0]});
   } else if (type == "pool2d" || type == "fused_transpose" ||
              type == "reshape2" || type == "nearest_interp" ||
-             type == "nearest_interp_v2") {
+             type == "nearest_interp_v2" || type == "dropout") {
     op->SetInput("X", {inputs[0]});
     op->SetOutput("Out", {outputs[0]});
   } else if (type == "slice") {
@@ -70,9 +70,6 @@ void SetOp(ProgramDesc* prog,
   } else if (type == "split") {
     op->SetInput("X", {inputs[0]});
     op->SetOutput("Out", {outputs});
-  } else if (type == "dropout") {
-    op->SetInput("X", {inputs[0]});
-    op->SetOutput("Out", {outputs[0]});
   } else if (type == "fc") {
     op->SetInput("Input", {inputs[0]});
     if (inputs.size() > 1) op->SetInput("W", {inputs[1]});
diff --git a/paddle/fluid/framework/ir/mkldnn/cpu_quantize_squash_pass_tester.cc b/paddle/fluid/framework/ir/mkldnn/cpu_quantize_squash_pass_tester.cc
index 5b1cd5fe87aed0..afbcf97ab1e29b 100644
--- a/paddle/fluid/framework/ir/mkldnn/cpu_quantize_squash_pass_tester.cc
+++ b/paddle/fluid/framework/ir/mkldnn/cpu_quantize_squash_pass_tester.cc
@@ -41,7 +41,7 @@ void SetOp(ProgramDesc* prog,
   if (type != "dropout" && type != "quantize" && type != "dequantize") {
     op->SetAttr("mkldnn_data_type", mkldnn_data_type);
   }
-  if (type == "pool2d") {
+  if (type == "pool2d") {  // NOLINT
     op->SetInput("X", {inputs[0]});
     op->SetOutput("Out", {outputs[0]});
     if (!scale.empty()) op->SetAttr("Scale_in", scale[0]);
diff --git a/paddle/fluid/framework/ir/mkldnn/int8_scale_calculation_mkldnn_pass_tester.cc b/paddle/fluid/framework/ir/mkldnn/int8_scale_calculation_mkldnn_pass_tester.cc
index fd6c7d20a95d2e..bd272635f15f1d 100644
--- a/paddle/fluid/framework/ir/mkldnn/int8_scale_calculation_mkldnn_pass_tester.cc
+++ b/paddle/fluid/framework/ir/mkldnn/int8_scale_calculation_mkldnn_pass_tester.cc
@@ -70,14 +70,7 @@ ProgramDesc BuildProgramDesc(bool convWithExistingBias,
     }
   }
 
-  if (convWithExistingBias) {
-    SetOp(&prog,
-          "conv2d",
-          "conv",
-          std::vector<std::string>({"c", "weights", "conv_bias"}),
-          std::vector<std::string>({"f"}),
-          scale_weights);
-  } else if (scale_weights.size() > 1) {
+  if (convWithExistingBias || scale_weights.size() > 1) {
     SetOp(&prog,
           "conv2d",
           "conv",
diff --git a/paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc b/paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
index 295ef57cfdfead..cc20f521808718 100644
--- a/paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
+++ b/paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
@@ -933,7 +933,7 @@ bool ReduceSSAGraphBuilder::DealWithSpecialOp(ir::Graph *result,
 
 void ReduceSSAGraphBuilder::InsertPostprocessOps(ir::Graph *result) const {
   if (UseGPU()) {
-    if (strategy_.fuse_broadcast_ops_ == true) {
+    if (strategy_.fuse_broadcast_ops_ == true) {  // NOLINT
       CreateFusedBroadcastOp(result, bcast_var_name_set_);
     } else {
       for (size_t dev_id = 0; dev_id < bcast_var_name_set_.size(); ++dev_id) {
@@ -1193,7 +1193,7 @@ int DistSSAGraphBuilder::CreateRPCOp(ir::Graph *result, ir::Node *node) const {
                                  node->Op()->Type()));
   // Create fetch_barrier op handle to enable output on all devices.
   // **NOTE** fetch_barrier should output variables list same as recv op does.
-  if (node->Op()->Type() == "fetch_barrier") {
+  if (node->Op()->Type() == "fetch_barrier") {  // NOLINT
     result->Get<GraphOps>(kGraphOps).emplace_back(
         new details::FetchBarrierOpHandle(
             result->CreateOpNode(node->Op()), local_scopes_, places_));
@@ -1354,7 +1354,7 @@ void DistSSAGraphBuilder::InsertPostprocessOps(ir::Graph *result) const {
         strategy_.reduce_ == details::BuildStrategy::ReduceStrategy::kReduce) {
       return;
     }
-    if (strategy_.fuse_broadcast_ops_ == true) {
+    if (strategy_.fuse_broadcast_ops_ == true) {  // NOLINT
       CreateFusedBroadcastOp(result, bcast_var_name_set_);
     } else {
       for (size_t dev_id = 0; dev_id < bcast_var_name_set_.size(); ++dev_id) {
diff --git a/paddle/fluid/framework/ir/transfer_layout_elim_pass.cc b/paddle/fluid/framework/ir/transfer_layout_elim_pass.cc
index 9ca4d4f482f085..c8cab82d950e0e 100644
--- a/paddle/fluid/framework/ir/transfer_layout_elim_pass.cc
+++ b/paddle/fluid/framework/ir/transfer_layout_elim_pass.cc
@@ -239,7 +239,7 @@ void TransferLayoutElimPass::ApplyImpl(ir::Graph *graph) const {
   FusePassBase::Init(pattern_name, graph);
 
   auto transfer_format = [&](std::string data_format) -> std::string {
-    if (data_format == "NCHW") {
+    if (data_format == "NCHW") {  // NOLINT
       return "NHWC";
     } else if (data_format == "NHWC") {
       return "NCHW";
diff --git a/paddle/fluid/framework/new_executor/garbage_collector/garbage_collector.cc b/paddle/fluid/framework/new_executor/garbage_collector/garbage_collector.cc
index 166853e2b18daf..0d73e2d3fede93 100644
--- a/paddle/fluid/framework/new_executor/garbage_collector/garbage_collector.cc
+++ b/paddle/fluid/framework/new_executor/garbage_collector/garbage_collector.cc
@@ -32,14 +32,14 @@ CreateInterpreterCoreGarbageCollector(
     const platform::Place& place,
     const std::vector<std::unique_ptr<InstructionBase>>& vec_instruction) {
   if (platform::is_gpu_place(place)) {
-    if (IsInterpretercoreFastGCEnabled()) {
+    if (IsInterpretercoreFastGCEnabled()) {  // NOLINT
       return std::unique_ptr<InterpreterCoreGarbageCollector>(
           new InterpreterCoreFastGarbageCollector());
     } else {
       return std::unique_ptr<InterpreterCoreGarbageCollector>(
           new InterpreterCoreEventGarbageCollector(vec_instruction));
     }
-  } else if (platform::is_xpu_place(place)) {
+  } else if (platform::is_xpu_place(place)) {  // NOLINT
     // Because there is no multi-stream on XPU device, fast GC can
     // be used.
     // Previously, XPU used no_event GC. But `Wait` in no_event GC
@@ -62,14 +62,14 @@ CreateInterpreterCoreGarbageCollector(
     const platform::Place& place,
     const std::vector<Instruction>& vec_instruction) {
   if (platform::is_gpu_place(place)) {
-    if (IsInterpretercoreFastGCEnabled()) {
+    if (IsInterpretercoreFastGCEnabled()) {  // NOLINT
       return std::unique_ptr<InterpreterCoreGarbageCollector>(
           new InterpreterCoreFastGarbageCollector());
     } else {
       return std::unique_ptr<InterpreterCoreGarbageCollector>(
           new InterpreterCoreEventGarbageCollector(vec_instruction));
     }
-  } else if (platform::is_xpu_place(place)) {
+  } else if (platform::is_xpu_place(place)) {  // NOLINT
     // Because there is no multi-stream on XPU device, fast GC can
     // be used.
     // Previously, XPU used no_event GC. But `Wait` in no_event GC
diff --git a/paddle/fluid/framework/new_executor/garbage_collector/no_event_garbage_collector.cc b/paddle/fluid/framework/new_executor/garbage_collector/no_event_garbage_collector.cc
index 3b7ebc18f36da1..d236e740679dd8 100644
--- a/paddle/fluid/framework/new_executor/garbage_collector/no_event_garbage_collector.cc
+++ b/paddle/fluid/framework/new_executor/garbage_collector/no_event_garbage_collector.cc
@@ -49,9 +49,10 @@ void InterpreterCoreNoEventGarbageCollector::Add(
 
   if (var->IsType<phi::DenseTensor>()) {
     Add(var->GetMutable<phi::DenseTensor>()->MoveMemoryHolder(), ctx);
-  } else if (var->IsType<
-                 operators::reader::
-                     OrderedMultiDeviceLoDTensorBlockingQueueHolder>()) {
+  } else if (
+      var->IsType<
+          operators::reader::
+              OrderedMultiDeviceLoDTensorBlockingQueueHolder>()) {  // NOLINT
     // TODO(xiongkun03) in old executor, this type of variable is not support
     // eager deletion. so we just leave it here ?
   } else if (var->IsType<LoDRankTable>()) {
diff --git a/paddle/fluid/framework/new_executor/new_executor_defs.cc b/paddle/fluid/framework/new_executor/new_executor_defs.cc
index bfa7542b65b75c..2a2393142dce37 100644
--- a/paddle/fluid/framework/new_executor/new_executor_defs.cc
+++ b/paddle/fluid/framework/new_executor/new_executor_defs.cc
@@ -96,7 +96,7 @@ void VariableScope::AddVar(const std::string& name,
     auto id = VarSize();
     name2id_[name] = static_cast<int>(id);
     vec_meta_info_.emplace_back(0, var_desc);
-    if (local_scope_ != nullptr) {
+    if (local_scope_ != nullptr) {  // NOLINT
       var_list_.push_back(local_scope_->FindVar(name));
     } else {
       var_list_.push_back(scope_->FindVar(name));
diff --git a/paddle/fluid/framework/new_executor/pir_interpreter.cc b/paddle/fluid/framework/new_executor/pir_interpreter.cc
index 76bb419742d376..60a69f70b74585 100644
--- a/paddle/fluid/framework/new_executor/pir_interpreter.cc
+++ b/paddle/fluid/framework/new_executor/pir_interpreter.cc
@@ -701,7 +701,7 @@ void PirInterpreter::BuildInstruction() {
         continue;
       }
     } else if (op.dialect()->name() == "pd_op") {
-      if (op.isa<paddle::dialect::IfOp>()) {
+      if (op.isa<paddle::dialect::IfOp>()) {  // NOLINT
         vec_instruction_base_.emplace_back(std::make_unique<IfInstruction>(
             op_idx++, place_, &op, value_exe_info_.get(), execution_config_));
         sub_blocks_.insert(
@@ -742,7 +742,7 @@ void PirInterpreter::BuildInstruction() {
       }
       VLOG(6) << "process " << op_name;
 
-      if (op.isa<paddle::dialect::LegacyKernelOp>()) {
+      if (op.isa<paddle::dialect::LegacyKernelOp>()) {  // NOLINT
         CREATE_INSTR(LegacyKernelInstruction);
       } else {
         CREATE_INSTR(PhiKernelInstruction);
diff --git a/paddle/fluid/framework/new_executor/standalone_executor.cc b/paddle/fluid/framework/new_executor/standalone_executor.cc
index 2bb0a7197774ee..74e09a15d62464 100644
--- a/paddle/fluid/framework/new_executor/standalone_executor.cc
+++ b/paddle/fluid/framework/new_executor/standalone_executor.cc
@@ -57,7 +57,7 @@ StandaloneExecutor::StandaloneExecutor(const platform::Place& place,
     const std::string& job_type = job->Type();
     std::shared_ptr<ProgramDesc> program = nullptr;
     std::shared_ptr<::pir::Program> ir_program = nullptr;
-    if (FLAGS_enable_pir_api || FLAGS_enable_pir_in_executor) {
+    if (FLAGS_enable_pir_api || FLAGS_enable_pir_in_executor) {  // NOLINT
       ir_program = plan_.IrProgram(job_type);
     } else {
       // NOTE (liuchenghao): std::make_shared will duplicate ProgramDesc object,
diff --git a/paddle/fluid/framework/operator.cc b/paddle/fluid/framework/operator.cc
index d731dd4b2086b4..edd7cf47037ceb 100644
--- a/paddle/fluid/framework/operator.cc
+++ b/paddle/fluid/framework/operator.cc
@@ -1754,7 +1754,8 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
   std::string phi_kernel_name;
   if (phi::KernelFactory::Instance().HasCompatiblePhiKernel(type_)) {
     if (kernel_signature_ == nullptr || phi_kernel_ == nullptr) {
-      if (phi::KernelFactory::Instance().HasStructuredKernel(type_)) {
+      if (phi::KernelFactory::Instance().HasStructuredKernel(
+              type_)) {  // NOLINT
         kernel_signature_ =
             std::make_unique<phi::KernelSignature>(type_.c_str());
       } else {
@@ -1989,7 +1990,7 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
                                        1,
                                        platform::EventRole::kInnerOp);
     if (need_prepare_data_) {
-      if (fallback_to_cpu) {
+      if (fallback_to_cpu) {  // NOLINT
         transfer_scope = PrepareData(scope,
                                      phi_cpu_kernel_key,
                                      &transfered_inplace_vars,
@@ -2278,7 +2279,7 @@ OpKernelType OperatorWithKernel::InnerGetExpectedKernelType(
 phi::KernelKey OperatorWithKernel::ChoosePhiKernel(
     const ExecutionContext& ctx) const {
   std::string phi_kernel_name;
-  if (phi::KernelFactory::Instance().HasStructuredKernel(type_)) {
+  if (phi::KernelFactory::Instance().HasStructuredKernel(type_)) {  // NOLINT
     kernel_signature_ = std::make_unique<phi::KernelSignature>(type_.c_str());
   } else {
     kernel_signature_ = std::make_unique<phi::KernelSignature>(
@@ -3104,7 +3105,7 @@ static void SetDnnAttrIntoDeviceContext(
       case proto::AttrType::STRING:
         one_dnn_ctx->SetDnnAttr(attr_name, PADDLE_GET_CONST(std::string, attr));
         break;
-      case proto::AttrType::INTS:
+      case proto::AttrType::INTS:  // NOLINT
         one_dnn_ctx->SetDnnAttr(attr_name,
                                 PADDLE_GET_CONST(std::vector<int>, attr));
         break;
@@ -3358,7 +3359,7 @@ void OperatorWithKernel::BuildPhiKernelContext(
       case phi::AttributeType::INT_ARRAY:
         if (attr_iter != Attrs().end()) {
           switch (AttrTypeID(attr_iter->second)) {
-            case proto::AttrType::INTS:
+            case proto::AttrType::INTS:  // NOLINT
               phi_kernel_context->EmplaceBackAttr(std::move(phi::IntArray(
                   PADDLE_GET_CONST(std::vector<int32_t>, attr_iter->second))));
               break;
@@ -3497,7 +3498,7 @@ void OperatorWithKernel::BuildPhiKernelContext(
             phi_kernel_context->EmplaceBackAttr(
                 PADDLE_GET_CONST(int64_t, attr_iter->second));
             break;
-          case phi::AttributeType::INT32S:
+          case phi::AttributeType::INT32S:  // NOLINT
             phi_kernel_context->EmplaceBackAttr(
                 PADDLE_GET_CONST(std::vector<int>, attr_iter->second));
             break;
@@ -3536,7 +3537,7 @@ void OperatorWithKernel::BuildPhiKernelContext(
                     attr_names[i]));
             }
             break;
-          case phi::AttributeType::FLOAT32S:
+          case phi::AttributeType::FLOAT32S:  // NOLINT
             phi_kernel_context->EmplaceBackAttr(
                 PADDLE_GET_CONST(std::vector<float>, attr_iter->second));
             break;
diff --git a/paddle/fluid/framework/section_worker.cc b/paddle/fluid/framework/section_worker.cc
index f88dbc409d1704..02f2bfb3c568d7 100644
--- a/paddle/fluid/framework/section_worker.cc
+++ b/paddle/fluid/framework/section_worker.cc
@@ -238,7 +238,7 @@ void SectionWorker::TrainFiles() {
 #endif
   }  // max_memory_size >= 0
 
-  if (schedule_mode_ == 0) {
+  if (schedule_mode_ == 0) {  // NOLINT
     RunFThenB(gc);
   } else {
     Run1F1B(gc);
diff --git a/paddle/fluid/imperative/amp_auto_cast.cc b/paddle/fluid/imperative/amp_auto_cast.cc
index 6637ddbe9ee670..ee8f855adf6304 100644
--- a/paddle/fluid/imperative/amp_auto_cast.cc
+++ b/paddle/fluid/imperative/amp_auto_cast.cc
@@ -187,7 +187,7 @@ AmpOperators::GetMutableUnsupportedOps(const phi::DataType& data_type) {
       true,
       phi::errors::InvalidArgument(
           "The data_type mismatch. It should be FLOAT16 or BFLOAT16."));
-  if (data_type == phi::DataType::FLOAT16) {
+  if (data_type == phi::DataType::FLOAT16) {  // NOLINT
     return unsupported_fp16_ops_;
   } else {
     return unsupported_bf16_ops_;
@@ -377,7 +377,8 @@ template <typename VarType>
 NameVarMap<VarType> AutoCastInputs(const std::string& op_type,
                                    const NameVarMap<VarType>& ins) {
   NameVarMap<VarType> new_ins(ins);
-  if (AmpOperators::Instance().GetMutableAllowOps()->count(op_type)) {
+  if (AmpOperators::Instance().GetMutableAllowOps()->count(
+          op_type)) {  // NOLINT
     for (auto& pair : new_ins) {
       // NOTE(zhiqiu): batch_norm and layer_norm support only input x is fp16.
       if ((op_type == "batch_norm" || op_type == "layer_norm" ||
diff --git a/paddle/fluid/imperative/gradient_accumulator.cc b/paddle/fluid/imperative/gradient_accumulator.cc
index f92755025bed70..8e4e7504eddaa6 100644
--- a/paddle/fluid/imperative/gradient_accumulator.cc
+++ b/paddle/fluid/imperative/gradient_accumulator.cc
@@ -518,7 +518,7 @@ void VariableWrapperAdd(std::shared_ptr<VariableWrapper> var,
 static platform::Place GetPlaceOfVar(
     const std::shared_ptr<VariableWrapper>& var) {
   platform::Place place;
-  if (var->Var().IsType<phi::DenseTensor>()) {
+  if (var->Var().IsType<phi::DenseTensor>()) {  // NOLINT
     place = var->Var().Get<phi::DenseTensor>().place();
   } else if (var->Var().IsType<phi::SelectedRows>()) {
     place = var->Var().Get<phi::SelectedRows>().place();
@@ -735,7 +735,7 @@ void SortedGradientAccumulator::SumGrad(std::shared_ptr<VariableWrapper> var,
       }
 
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
-      if (paddle::platform::is_gpu_place(place)) {
+      if (paddle::platform::is_gpu_place(place)) {  // NOLINT
         // sum selected rows firstly
         for (auto& var_info : tmp_grad_vars_) {
           if (!var_info.var->Var().IsType<phi::SelectedRows>()) {
diff --git a/paddle/fluid/imperative/layout_autotune.cc b/paddle/fluid/imperative/layout_autotune.cc
index 80e216563f07d4..2e0e5ea26da9a4 100644
--- a/paddle/fluid/imperative/layout_autotune.cc
+++ b/paddle/fluid/imperative/layout_autotune.cc
@@ -145,7 +145,7 @@ LayoutAutotuneGuard::LayoutAutotuneGuard(std::shared_ptr<Tracer> tracer,
 }
 
 LayoutAutotuneGuard::~LayoutAutotuneGuard() {
-  if (pre_layout_autotune_) {
+  if (pre_layout_autotune_) {  // NOLINT
     tracer_->EnableLayoutAutoTune();
   } else {
     tracer_->DisableLayoutAutoTune();
diff --git a/paddle/fluid/imperative/nccl_context.cc b/paddle/fluid/imperative/nccl_context.cc
index d70d40808f915d..3ed9b97bfc362d 100644
--- a/paddle/fluid/imperative/nccl_context.cc
+++ b/paddle/fluid/imperative/nccl_context.cc
@@ -67,7 +67,7 @@ void NCCLParallelContext::Init() {
   std::vector<ncclUniqueId> nccl_ids;
   nccl_ids.resize(strategy_.nrings_);
 
-  if (strategy_.local_rank_ == 0) {
+  if (strategy_.local_rank_ == 0) {  // NOLINT
     // generate the unique ncclid on the root worker
     for (auto &nccl_id : nccl_ids) {
       platform::dynload::ncclGetUniqueId(&nccl_id);
diff --git a/paddle/fluid/imperative/partial_grad_engine.cc b/paddle/fluid/imperative/partial_grad_engine.cc
index 3fd37d5ec36744..c107c9f80ef524 100644
--- a/paddle/fluid/imperative/partial_grad_engine.cc
+++ b/paddle/fluid/imperative/partial_grad_engine.cc
@@ -366,7 +366,7 @@ class GradientAccumulationInfo {
       if (!grad_var_) {
         grad_var_ = std::make_shared<VarBase>(true, mapped_grad_var_->Name());
         grad_var_->SetOverriddenStopGradient(false);
-        if (sort_gradient_) {
+        if (sort_gradient_) {  // NOLINT
           accumulator_ = std::make_unique<SortedGradientAccumulator>(
               grad_var_->SharedVar().get());
         } else {
diff --git a/paddle/fluid/imperative/prepared_operator.cc b/paddle/fluid/imperative/prepared_operator.cc
index 8129ea244f4895..a60c81a4c22d9e 100644
--- a/paddle/fluid/imperative/prepared_operator.cc
+++ b/paddle/fluid/imperative/prepared_operator.cc
@@ -660,7 +660,7 @@ void PreparedOp::Run(const NameVarMap<VarBase>& ins,
                      const NameVarMap<VarBase>& outs,
                      const framework::AttributeMap& attrs,
                      const framework::AttributeMap& default_attrs) {
-  if (run_phi_kernel_) {
+  if (run_phi_kernel_) {  // NOLINT
     PreparedOpRunPtImpl<VarBase>(op_,
                                  kernel_key_,
                                  arg_map_fn_,
@@ -692,7 +692,7 @@ void PreparedOp::Run(const NameVarMap<VariableWrapper>& ins,
                      const NameVarMap<VariableWrapper>& outs,
                      const framework::AttributeMap& attrs,
                      const framework::AttributeMap& default_attrs) {
-  if (run_phi_kernel_) {
+  if (run_phi_kernel_) {  // NOLINT
     PreparedOpRunPtImpl<VariableWrapper>(op_,
                                          kernel_key_,
                                          arg_map_fn_,
@@ -724,7 +724,7 @@ void PreparedOp::Run(const NameVarMap<egr::EagerVariable>& ins,
                      const NameVarMap<egr::EagerVariable>& outs,
                      const framework::AttributeMap& attrs,
                      const framework::AttributeMap& default_attrs) {
-  if (run_phi_kernel_) {
+  if (run_phi_kernel_) {  // NOLINT
     PreparedOpRunPtImpl<egr::EagerVariable>(op_,
                                             kernel_key_,
                                             arg_map_fn_,
diff --git a/paddle/fluid/imperative/reducer.cc b/paddle/fluid/imperative/reducer.cc
index 461c2d3ff4bb87..5b8dc28d03111e 100644
--- a/paddle/fluid/imperative/reducer.cc
+++ b/paddle/fluid/imperative/reducer.cc
@@ -227,7 +227,7 @@ void SplitTensorsWithType<platform::XPUDeviceContext>(
 
 void Group::ConcatTensors(const platform::DeviceContext &context) {
   auto place = context.GetPlace();
-  if (platform::is_gpu_place(place)) {
+  if (platform::is_gpu_place(place)) {  // NOLINT
 #if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
     ConcatTensorsWithType(static_cast<const phi::GPUContext &>(context),
                           dense_tensors_,
@@ -263,7 +263,7 @@ void Group::ConcatTensors(const platform::DeviceContext &context) {
 
 void Group::SplitTensors(const platform::DeviceContext &context) {
   auto place = context.GetPlace();
-  if (platform::is_gpu_place(place)) {
+  if (platform::is_gpu_place(place)) {  // NOLINT
 #if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
     SplitTensorsWithType(static_cast<const phi::GPUContext &>(context),
                          &dense_contents_,
diff --git a/paddle/fluid/imperative/var_helper.cc b/paddle/fluid/imperative/var_helper.cc
index bafea5a720d3a7..9561962935ffe4 100644
--- a/paddle/fluid/imperative/var_helper.cc
+++ b/paddle/fluid/imperative/var_helper.cc
@@ -50,7 +50,8 @@ void InitializeVariable(paddle::framework::Variable *var,
     var->GetMutable<phi::SelectedRows>();
   } else if (var_type == paddle::framework::proto::VarType::FEED_MINIBATCH) {
     var->GetMutable<paddle::framework::FeedList>();
-  } else if (var_type == paddle::framework::proto::VarType::FETCH_LIST) {
+  } else if (var_type ==
+             paddle::framework::proto::VarType::FETCH_LIST) {  // NOLINT
     var->GetMutable<paddle::framework::FetchList>();
   } else if (var_type == paddle::framework::proto::VarType::STEP_SCOPES) {
     var->GetMutable<std::vector<paddle::framework::Scope *>>();
diff --git a/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc b/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc
index ad95fe3091ce18..286fe961bed06e 100644
--- a/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc
+++ b/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc
@@ -271,7 +271,7 @@ void LiteSubgraphPass::SetUpEngine(
       Get<std::vector<std::string>>("nnadapter_model_cache_token");
 
   lite_api::TargetType target_type = TARGET(kX86);
-  if (use_gpu) {
+  if (use_gpu) {  // NOLINT
     target_type = TARGET(kCUDA);
   } else if (use_xpu) {
     target_type = TARGET(kXPU);
@@ -417,13 +417,11 @@ void LiteSubgraphPass::ApplyImpl(framework::ir::Graph* graph) const {
   auto& lite_ops_filter = Get<std::vector<std::string>>("lite_ops_filter");
 
   auto teller = [&lite_ops_filter](const Node* node) {
-    if (!node->IsOp() || !node->Op())
-      return false;
-    else if (node->Op()->Type() == "feed" || node->Op()->Type() == "fetch")
-      return false;
-    else if (std::find(lite_ops_filter.begin(),
-                       lite_ops_filter.end(),
-                       node->Op()->Type()) != lite_ops_filter.end())
+    if (!node->IsOp() || !node->Op() || node->Op()->Type() == "feed" ||
+        node->Op()->Type() == "fetch" ||
+        std::find(lite_ops_filter.begin(),
+                  lite_ops_filter.end(),
+                  node->Op()->Type()) != lite_ops_filter.end())
       return false;
     return inference::lite::OpTeller::Global().Tell(node->Op()->Type(),
                                                     *node->Op());
diff --git a/paddle/fluid/inference/analysis/passes/ir_graph_build_pass.cc b/paddle/fluid/inference/analysis/passes/ir_graph_build_pass.cc
index 8106dfbb9e6aae..ea97be8f90a606 100644
--- a/paddle/fluid/inference/analysis/passes/ir_graph_build_pass.cc
+++ b/paddle/fluid/inference/analysis/passes/ir_graph_build_pass.cc
@@ -121,7 +121,7 @@ std::unique_ptr<framework::ProgramDesc> IrGraphBuildPass::LoadModel(
     bool model_from_memory,
     bool skip_load_params) {
   framework::Executor exe(place);
-  if (!model_from_memory) {
+  if (!model_from_memory) {  // NOLINT
     return Load(&exe, scope, program_path, params_path, !skip_load_params);
   } else {
     return LoadFromMemory(&exe, scope, program_path, params_path);
diff --git a/paddle/fluid/inference/api/analysis_predictor.cc b/paddle/fluid/inference/api/analysis_predictor.cc
index 96faaae05d8c00..9e6c50443ec646 100644
--- a/paddle/fluid/inference/api/analysis_predictor.cc
+++ b/paddle/fluid/inference/api/analysis_predictor.cc
@@ -1170,7 +1170,7 @@ bool AnalysisPredictor::LoadConverterConfig(
       int64_t key = std::stoll(one_line[0]);
       for (size_t i = 1; i < one_line.size(); ++i) {
         int64_t val = std::stoll(one_line[i]);
-        if (ring_to_rank) {
+        if (ring_to_rank) {  // NOLINT
           if (ring_id_to_ranks->find(key) == ring_id_to_ranks->end()) {
             ring_id_to_ranks->insert({key, std::vector<int64_t>()});
           }
@@ -1310,7 +1310,7 @@ bool AnalysisPredictor::Run(const std::vector<PaddleTensor> &inputs,
     HookCollectShapeRangeInfo();
   }
 
-  if (config_.new_executor_enabled()) {
+  if (config_.new_executor_enabled()) {  // NOLINT
     executor_->RunInterpreterCore();
   } else {
     // Run the inference program
@@ -1383,7 +1383,7 @@ bool AnalysisPredictor::Run(const std::vector<paddle::Tensor> &inputs,
     HookCollectShapeRangeInfo();
   }
 
-  if (config_.new_executor_enabled()) {
+  if (config_.new_executor_enabled()) {  // NOLINT
     executor_->RunInterpreterCore();
   } else {
     // Run the inference program
@@ -1804,7 +1804,7 @@ void AnalysisPredictor::PrepareArgument() {
           if (deleted_passes.count(pass)) continue;
           pass_builder->AppendPass(pass);
         }
-      } else if (config_.use_xpu()) {
+      } else if (config_.use_xpu()) {  // NOLINT
         // All passes support fp16. Not reset pass_builder.
       } else if (config_.use_custom_device()) {
         // All passes support fp16. Not reset pass_builder.
@@ -1927,7 +1927,8 @@ void AnalysisPredictor::OptimizeInferenceProgram() {
 #else
   if (config_.mkldnn_enabled() ||
       (config_.tensorrt_engine_enabled() &&
-       config_.tensorrt_precision_mode_ == AnalysisConfig::Precision::kInt8)) {
+       config_.tensorrt_precision_mode_ ==
+           AnalysisConfig::Precision::kInt8)) {  // NOLINT
     argument_->PartiallyRelease();
   } else {
     argument_.reset(nullptr);
@@ -2221,7 +2222,7 @@ std::unique_ptr<ZeroCopyTensor> AnalysisPredictor::GetInputTensor(
     const std::string &name) {
   framework::Scope *scope = nullptr;
 #if defined(PADDLE_WITH_DISTRIBUTE) && defined(PADDLE_WITH_PSCORE)
-  if (config_.dist_config().use_dist_model()) {
+  if (config_.dist_config().use_dist_model()) {  // NOLINT
     scope = scope_.get();
   } else {
     scope = executor_->GetScope();
@@ -2272,7 +2273,7 @@ std::unique_ptr<ZeroCopyTensor> AnalysisPredictor::GetOutputTensor(
     const std::string &name) {
   framework::Scope *scope;  // NOLINT
 #if defined(PADDLE_WITH_DISTRIBUTE) && defined(PADDLE_WITH_PSCORE)
-  if (config_.dist_config().use_dist_model()) {
+  if (config_.dist_config().use_dist_model()) {  // NOLINT
     scope = scope_.get();
   } else {
     scope = executor_->GetScope();
@@ -2322,7 +2323,7 @@ std::unique_ptr<ZeroCopyTensor> AnalysisPredictor::GetOutputTensor(
 bool AnalysisPredictor::ZeroCopyRun(bool switch_stream) {
   inference::DisplayMemoryInfo(place_, "before run");
 #if defined(PADDLE_WITH_DISTRIBUTE) && defined(PADDLE_WITH_PSCORE)
-  if (config_.dist_config().use_dist_model()) {
+  if (config_.dist_config().use_dist_model()) {  // NOLINT
     VLOG(3) << "ZeroCopyRun will use the fleet executor.";
     fleet_exe_->Run(config_.dist_config().carrier_id());
     return true;
@@ -2381,7 +2382,7 @@ bool AnalysisPredictor::ZeroCopyRun(bool switch_stream) {
   }
 #endif
 
-  if (config_.new_executor_enabled()) {
+  if (config_.new_executor_enabled()) {  // NOLINT
     executor_->RunInterpreterCore({}, false, switch_stream);
   } else {
     executor_->Run();
@@ -2647,7 +2648,7 @@ void AnalysisPredictor::StatisticShapeRangeInfo() {
 bool AnalysisPredictor::LoadProgramDesc() {
   // Initialize the inference program
   std::string filename;
-  if (!config_.model_dir().empty()) {
+  if (!config_.model_dir().empty()) {  // NOLINT
     filename = config_.model_dir() + "/__model__";
   } else if (!config_.prog_file().empty()) {
     // All parameters are saved in a single file.
diff --git a/paddle/fluid/inference/api/api_impl.cc b/paddle/fluid/inference/api/api_impl.cc
index d886885edb5ba5..e00d5d129a19e4 100644
--- a/paddle/fluid/inference/api/api_impl.cc
+++ b/paddle/fluid/inference/api/api_impl.cc
@@ -101,7 +101,7 @@ bool NativePaddlePredictor::Init(
   executor_ = std::make_unique<paddle::framework::Executor>(place_);
 
   // Initialize the inference program
-  if (!config_.model_dir.empty()) {
+  if (!config_.model_dir.empty()) {  // NOLINT
     // Parameters are saved in separate files sited in
     // the specified `dirname`.
     inference_program_ = paddle::inference::Load(
@@ -286,7 +286,7 @@ bool NativePaddlePredictor::SetFeed(const std::vector<PaddleTensor> &inputs,
     }
     input.set_lod(lod);
     int idx = -1;
-    if (config_.specify_input_name) {
+    if (config_.specify_input_name) {  // NOLINT
       idx = static_cast<int>(feed_names_[inputs[i].name]);
     } else {
       idx = PADDLE_GET_CONST(int, feeds_[i]->GetAttr("col"));
diff --git a/paddle/fluid/inference/api/mkldnn_quantizer.cc b/paddle/fluid/inference/api/mkldnn_quantizer.cc
index 54a198b4e2f590..6ddcd7b37ba0bd 100644
--- a/paddle/fluid/inference/api/mkldnn_quantizer.cc
+++ b/paddle/fluid/inference/api/mkldnn_quantizer.cc
@@ -78,7 +78,7 @@ void AnalysisPredictor::MkldnnQuantizer::CalculateScalesForRNNWeights(
     check_var(wh_var, wh_name);
     phi::DenseTensor* wx_tensor = wx_var->GetMutable<phi::DenseTensor>();
     phi::DenseTensor* wh_tensor = wh_var->GetMutable<phi::DenseTensor>();
-    if (gru) {
+    if (gru) {  // NOLINT
       scales_[wx_name] = GetMaxChGRUScalingFactor(*wx_tensor, *wh_tensor);
     } else {
       scales_[wx_name] = GetMaxChLSTMScalingFactor(*wx_tensor, *wh_tensor);
@@ -215,6 +215,7 @@ void AnalysisPredictor::MkldnnQuantizer::CalculateSingleScale(
 
   switch (rule) {
     case ScaleAlgo::MAX:
+    case ScaleAlgo::KL:
       scales_[var_name] = GetMaxScalingFactor(var_tensor, is_unsigned);
       break;
     case ScaleAlgo::MAX_CH:
@@ -227,9 +228,6 @@ void AnalysisPredictor::MkldnnQuantizer::CalculateSingleScale(
                                                 is_unsigned,
                                                 /*is_transposed*/ true);
       break;
-    case ScaleAlgo::KL:
-      scales_[var_name] = GetKLScalingFactor(var_tensor, is_unsigned);
-      break;
     default:
       throw std::runtime_error(
           "MkldnnQuantizer: Unexpected ScaleAlgo specified.");
diff --git a/paddle/fluid/ir_adaptor/translator/op_translator.cc b/paddle/fluid/ir_adaptor/translator/op_translator.cc
index 9295e5c643e5f0..f6cf5a7c62d711 100644
--- a/paddle/fluid/ir_adaptor/translator/op_translator.cc
+++ b/paddle/fluid/ir_adaptor/translator/op_translator.cc
@@ -1857,7 +1857,7 @@ struct FillConstant2FullTranscriber : public OpTranscriber {
       }
     }
     switch (place_type) {
-      case -1:
+      case -1:  // NOLINT
         attribute_map["place"] = paddle::dialect::PlaceAttribute::get(
             ctx, phi::Place(phi::AllocationType::UNDEFINED));
         break;
diff --git a/paddle/fluid/jit/property.cc b/paddle/fluid/jit/property.cc
index 687468df83a3dc..37c426bb5401b0 100644
--- a/paddle/fluid/jit/property.cc
+++ b/paddle/fluid/jit/property.cc
@@ -99,7 +99,7 @@ std::unordered_map<std::string, std::shared_ptr<Variable>> Property::Values() {
         case ValueProto::STRING:
           *var->GetMutable<paddle::framework::String>() = GetString(n);
           break;
-        case ValueProto::FLOATS:
+        case ValueProto::FLOATS:  // NOLINT
           *var->GetMutable<std::vector<float>>() = GetFloats(n);
           break;
         case ValueProto::INTS:
diff --git a/paddle/fluid/operators/reader/buffered_reader.cc b/paddle/fluid/operators/reader/buffered_reader.cc
index b73ffe4319be78..cc5034c86f90fb 100644
--- a/paddle/fluid/operators/reader/buffered_reader.cc
+++ b/paddle/fluid/operators/reader/buffered_reader.cc
@@ -380,7 +380,7 @@ void BufferedReader::ReadNextImpl(paddle::framework::LoDTensorArray *out) {
     return;
   }
 
-  if (platform::is_gpu_place(place_)) {
+  if (platform::is_gpu_place(place_)) {  // NOLINT
     *out = std::move(cuda_buffer_[i]);
   } else if (platform::is_xpu_place(place_)) {
     *out = std::move(xpu_buffer_[i]);
diff --git a/paddle/fluid/pir/drr/src/ir_operation_factory.cc b/paddle/fluid/pir/drr/src/ir_operation_factory.cc
index f792ccbdaff922..61c12c281e1398 100644
--- a/paddle/fluid/pir/drr/src/ir_operation_factory.cc
+++ b/paddle/fluid/pir/drr/src/ir_operation_factory.cc
@@ -81,7 +81,7 @@ pir::Attribute CreateIrAttribute(const std::any& obj) {
         std::any_cast<phi::DataType>(obj));
   } else if (obj.type() == typeid(phi::Place)) {
     return IrAttrbuteCreator<phi::Place>()(std::any_cast<phi::Place>(obj));
-  } else if (obj.type() == typeid(std::vector<int32_t>)) {
+  } else if (obj.type() == typeid(std::vector<int32_t>)) {  // NOLINT
     return IrAttrbuteCreator<std::vector<int32_t>>()(
         std::any_cast<std::vector<int32_t>>(obj));
   } else if (obj.type() == typeid(std::vector<int64_t>)) {
diff --git a/paddle/fluid/platform/place.cc b/paddle/fluid/platform/place.cc
index 118ba7d6b782c6..df66cc63e39863 100644
--- a/paddle/fluid/platform/place.cc
+++ b/paddle/fluid/platform/place.cc
@@ -62,8 +62,6 @@ bool is_same_place(const Place &p1, const Place &p2) {
   if (places_are_same_class(p1, p2)) {
     if (is_cpu_place(p1) || is_cuda_pinned_place(p1)) {
       return true;
-    } else if (is_xpu_place(p1) || is_ipu_place(p1) || is_custom_place(p1)) {
-      return p1 == p2;
     } else {
       return p1 == p2;
     }
diff --git a/paddle/fluid/platform/profiler.cc b/paddle/fluid/platform/profiler.cc
index 816ae57ff4c068..2630b36d0e8ad9 100644
--- a/paddle/fluid/platform/profiler.cc
+++ b/paddle/fluid/platform/profiler.cc
@@ -200,8 +200,8 @@ RecordMemEvent::RecordMemEvent(const void *ptr,
             RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][3];
         RecordMemEvent::has_initialized["gpu"][place.GetDeviceId()] = true;
       } else {
-        current_allocated =
-            DEVICE_MEMORY_STAT_CURRENT_VALUE(Allocated, place.GetDeviceId());
+        current_allocated = DEVICE_MEMORY_STAT_CURRENT_VALUE(
+            Allocated, place.GetDeviceId());  // NOLINT
         peak_allocated =
             DEVICE_MEMORY_STAT_PEAK_VALUE(Allocated, place.GetDeviceId());
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][0] =
@@ -283,10 +283,10 @@ RecordMemEvent::RecordMemEvent(const void *ptr,
             RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][3];
         RecordMemEvent::has_initialized["gpu"][place.GetDeviceId()] = true;
       } else {
-        current_reserved =
-            DEVICE_MEMORY_STAT_CURRENT_VALUE(Reserved, place.GetDeviceId());
-        peak_reserved =
-            DEVICE_MEMORY_STAT_PEAK_VALUE(Reserved, place.GetDeviceId());
+        current_reserved = DEVICE_MEMORY_STAT_CURRENT_VALUE(
+            Reserved, place.GetDeviceId());  // NOLINT
+        peak_reserved = DEVICE_MEMORY_STAT_PEAK_VALUE(
+            Reserved, place.GetDeviceId());  // NOLINT
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][1] =
             current_reserved;
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][3] =
@@ -366,10 +366,10 @@ RecordMemEvent::RecordMemEvent(const void *ptr,
             RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][3];
         RecordMemEvent::has_initialized["gpu"][place.GetDeviceId()] = true;
       } else {
-        current_allocated =
-            DEVICE_MEMORY_STAT_CURRENT_VALUE(Allocated, place.GetDeviceId());
-        peak_allocated =
-            DEVICE_MEMORY_STAT_PEAK_VALUE(Allocated, place.GetDeviceId());
+        current_allocated = DEVICE_MEMORY_STAT_CURRENT_VALUE(
+            Allocated, place.GetDeviceId());  // NOLINT
+        peak_allocated = DEVICE_MEMORY_STAT_PEAK_VALUE(
+            Allocated, place.GetDeviceId());  // NOLINT
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][0] =
             current_allocated;
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][2] =
@@ -449,10 +449,10 @@ RecordMemEvent::RecordMemEvent(const void *ptr,
             RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][3];
         RecordMemEvent::has_initialized["gpu"][place.GetDeviceId()] = true;
       } else {
-        current_reserved =
-            DEVICE_MEMORY_STAT_CURRENT_VALUE(Reserved, place.GetDeviceId());
-        peak_reserved =
-            DEVICE_MEMORY_STAT_PEAK_VALUE(Reserved, place.GetDeviceId());
+        current_reserved = DEVICE_MEMORY_STAT_CURRENT_VALUE(
+            Reserved, place.GetDeviceId());  // NOLINT
+        peak_reserved = DEVICE_MEMORY_STAT_PEAK_VALUE(
+            Reserved, place.GetDeviceId());  // NOLINT
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][1] =
             current_reserved;
         RecordMemEvent::size_cache["gpu"][place.GetDeviceId()][3] =
diff --git a/paddle/fluid/pybind/eager.cc b/paddle/fluid/pybind/eager.cc
index 3cb3ccf964ec81..00b6ba994233f4 100644
--- a/paddle/fluid/pybind/eager.cc
+++ b/paddle/fluid/pybind/eager.cc
@@ -442,7 +442,7 @@ Placements ParsePlacementsArgs(
   Placements placements;
   const std::string& placements_key = "placements";
 
-  if (kw_order_map[placements_key] <= args_num) {
+  if (kw_order_map[placements_key] <= args_num) {  // NOLINT
     placements = CastPyArg2VectorOfPlacement(
         PyTuple_GET_ITEM(args, kw_order_map[placements_key] - 1),
         kw_order_map[placements_key] - 1);
diff --git a/paddle/fluid/pybind/eager_functions.cc b/paddle/fluid/pybind/eager_functions.cc
index 60472c25c71827..17d9d50d8ab26a 100644
--- a/paddle/fluid/pybind/eager_functions.cc
+++ b/paddle/fluid/pybind/eager_functions.cc
@@ -644,7 +644,7 @@ PyObject* eager_api_run_custom_op(PyObject* self,
     } else if (attr_type_str == "std::string") {
       ctx.EmplaceBackAttr(
           CastPyArg2AttrString(obj, attr_start_idx + i));  // NOLINT
-    } else if (attr_type_str == "std::vector<int>") {
+    } else if (attr_type_str == "std::vector<int>") {      // NOLINT
       ctx.EmplaceBackAttr(CastPyArg2VectorOfInt(obj, attr_start_idx + i));
     } else if (attr_type_str == "std::vector<float>") {
       ctx.EmplaceBackAttr(CastPyArg2VectorOfFloat(obj, attr_start_idx + i));
diff --git a/paddle/fluid/pybind/eager_math_op_patch.cc b/paddle/fluid/pybind/eager_math_op_patch.cc
index 21fd549cb0b2db..17b36e9237e78f 100644
--- a/paddle/fluid/pybind/eager_math_op_patch.cc
+++ b/paddle/fluid/pybind/eager_math_op_patch.cc
@@ -818,10 +818,10 @@ static PyObject* tensor__rdiv__method(TensorObject* self,
   bool has_other_double = false;
   if (PyFloat_Check(other_obj) || PyCheckInteger(other_obj) ||
       IsNumpyType(other_obj)) {
-    if (PyFloat_Check(other_obj)) {
+    if (PyFloat_Check(other_obj)) {  // NOLINT
       other_double = CastPyArg2Double(other_obj, "__rdiv__", 0);
       has_other_double = true;
-    } else if (PyCheckInteger(other_obj) || IsNumpyType(other_obj)) {
+    } else if (PyCheckInteger(other_obj) || IsNumpyType(other_obj)) {  // NOLINT
       other_double = CastPyArg2Double(other_obj, "__rdiv__", 0);
       has_other_double = true;
     }
diff --git a/paddle/fluid/pybind/eager_utils.cc b/paddle/fluid/pybind/eager_utils.cc
index d613c008b4958a..a2ce701f2aa952 100644
--- a/paddle/fluid/pybind/eager_utils.cc
+++ b/paddle/fluid/pybind/eager_utils.cc
@@ -647,7 +647,7 @@ std::vector<std::vector<size_t>> CastPyArg2VectorOfVectorOfSize_t(
 
 platform::Place CastPyArg2Place(PyObject* obj, ssize_t arg_pos) {
   platform::Place place;
-  if (PyObject_TypeCheck(obj, g_place_pytype)) {
+  if (PyObject_TypeCheck(obj, g_place_pytype)) {  // NOLINT
     place = ::pybind11::handle(obj).cast<platform::Place>();
   } else if (PyObject_TypeCheck(obj, g_cudaplace_pytype)) {
     place = ::pybind11::handle(obj).cast<platform::CUDAPlace>();
@@ -761,7 +761,8 @@ std::vector<phi::DenseTensor> CastPyArg2VectorOfTensorBase(PyObject* obj,
             i));
       }
     }
-  } else if (PyObject_TypeCheck(obj, g_framework_lodtensorarray_pytype)) {
+  } else if (PyObject_TypeCheck(obj,
+                                g_framework_lodtensorarray_pytype)) {  // NOLINT
     for (auto& tensor :
          (::pybind11::handle(obj).cast<framework::LoDTensorArray>())) {
       result.emplace_back(tensor);
@@ -788,7 +789,7 @@ using phi::distributed::Shard;
 Placements CastPyArg2VectorOfPlacement(PyObject* obj, ssize_t arg_pos) {
   Placements result;
   auto check_and_emplace = [&](PyObject* item, ssize_t i) {
-    if (PyObject_TypeCheck(item, g_placement_shard_pytype)) {
+    if (PyObject_TypeCheck(item, g_placement_shard_pytype)) {  // NOLINT
       result.emplace_back(
           std::make_shared<Shard>(::pybind11::handle(item).cast<Shard>()));
     } else if (PyObject_TypeCheck(item, g_placement_replicated_pytype)) {
diff --git a/paddle/fluid/pybind/parallel_executor.cc b/paddle/fluid/pybind/parallel_executor.cc
index 9060e158c9ed9a..1b567fb51ba1e1 100644
--- a/paddle/fluid/pybind/parallel_executor.cc
+++ b/paddle/fluid/pybind/parallel_executor.cc
@@ -931,7 +931,7 @@ void BindParallelExecutor(pybind11::module &m) {  // NOLINT
       .def_property(
           "memory_optimize",
           [](const BuildStrategy &self) -> py::object {
-            if (self.memory_optimize_) {
+            if (self.memory_optimize_) {  // NOLINT
               return py::cast(self.memory_optimize_.get());
             } else {
               return py::cast(nullptr);
diff --git a/paddle/fluid/pybind/pybind.cc b/paddle/fluid/pybind/pybind.cc
index f1d53f3f88750c..58ef7c4cc8bbbf 100644
--- a/paddle/fluid/pybind/pybind.cc
+++ b/paddle/fluid/pybind/pybind.cc
@@ -1240,7 +1240,7 @@ All parameter, weight, gradient are variables in Paddle.
           py::return_value_policy::reference)
       .def("get_bytes",
            [](Variable &self) {
-             if (self.IsType<String>()) {
+             if (self.IsType<String>()) {  // NOLINT
                return py::bytes(*(self.GetMutable<String>()));
              } else {
                return py::bytes(
@@ -2229,7 +2229,7 @@ All parameter, weight, gradient are variables in Paddle.
            const std::string &var_name,
            size_t index) -> py::object {
           auto &var = framework::GetFetchVariable(scope, var_name, index);
-          if (data_is_lod_tensor(var)) {
+          if (data_is_lod_tensor(var)) {  // NOLINT
             return py::cast(PADDLE_GET(phi::DenseTensor, var));
           } else {
             return py::cast(PADDLE_GET(LoDTensorArray, var));
diff --git a/paddle/phi/core/compat/convert_utils.cc b/paddle/phi/core/compat/convert_utils.cc
index d4c5de0dbe6dc9..37053cc0c09ec2 100644
--- a/paddle/phi/core/compat/convert_utils.cc
+++ b/paddle/phi/core/compat/convert_utils.cc
@@ -63,6 +63,7 @@ phi::Place TransToPhiPlace(const Backend& backend, bool set_device_id) {
       return phi::Place();
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
     case phi::Backend::GPU:
+    case phi::Backend::GPUDNN:
       return phi::GPUPlace(
           set_device_id ? phi::backends::gpu::GetCurrentDeviceId() : 0);
 #endif
@@ -70,11 +71,6 @@ phi::Place TransToPhiPlace(const Backend& backend, bool set_device_id) {
     case phi::Backend::ONEDNN:  // NOLINT
       return phi::CPUPlace();
 #endif
-#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
-    case phi::Backend::GPUDNN:
-      return phi::GPUPlace(
-          set_device_id ? phi::backends::gpu::GetCurrentDeviceId() : 0);
-#endif
 #if defined(PADDLE_WITH_XPU)
     case phi::Backend::XPU:
       return phi::XPUPlace(
diff --git a/paddle/phi/core/kernel_registry.cc b/paddle/phi/core/kernel_registry.cc
index fa9d531b6534d6..6ce1af187e9a3f 100644
--- a/paddle/phi/core/kernel_registry.cc
+++ b/paddle/phi/core/kernel_registry.cc
@@ -47,139 +47,159 @@ void SetKernelArgsDef(const std::vector<std::type_index>& args_type,
     ) {
 #endif
       // do nothing, skip context arg now
-    } else if (arg_type == std::type_index(typeid(const DenseTensor&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(const DenseTensor&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
     } else if (arg_type ==
-               std::type_index(typeid(const paddle::optional<DenseTensor>&))) {
+               std::type_index(
+                   typeid(const paddle::optional<DenseTensor>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
     } else if (arg_type ==
-               std::type_index(typeid(
-                   const paddle::optional<std::vector<const DenseTensor*>>&))) {
+               std::type_index(
+                   typeid(const paddle::optional<
+                          std::vector<const DenseTensor*>>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
     } else if (arg_type ==
-               std::type_index(typeid(const paddle::optional<SelectedRows>&))) {
+               std::type_index(
+                   typeid(const paddle::optional<SelectedRows>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(
-                               const std::vector<const DenseTensor*>&))) {
+    } else if (arg_type ==
+               std::type_index(
+                   typeid(const std::vector<const DenseTensor*>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
     } else if (arg_type ==
-               std::type_index(typeid(const phi::ExtendedTensor&))) {
+               std::type_index(typeid(const phi::ExtendedTensor&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(
-                               const std::vector<const ExtendedTensor*>&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(
+                   const std::vector<const ExtendedTensor*>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(
-                               const std::vector<const SelectedRows*>&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(
+                   const std::vector<const SelectedRows*>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
     } else if (arg_type ==
-               std::type_index(typeid(const std::vector<const TensorBase*>&))) {
+               std::type_index(
+                   typeid(const std::vector<const TensorBase*>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(
-                               const std::vector<const TensorArray*>&))) {
+    } else if (arg_type ==
+               std::type_index(
+                   typeid(const std::vector<const TensorArray*>&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(const SelectedRows&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(const SelectedRows&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(const StringTensor&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(const StringTensor&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(const SparseCooTensor&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(const SparseCooTensor&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(
-                               paddle::optional<const SparseCooTensor&>))) {
+    } else if (arg_type ==
+               std::type_index(typeid(
+                   paddle::optional<const SparseCooTensor&>))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(const SparseCsrTensor&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(const SparseCsrTensor&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(
-                               paddle::optional<const SparseCsrTensor&>))) {
+    } else if (arg_type ==
+               std::type_index(typeid(
+                   paddle::optional<const SparseCsrTensor&>))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(const TensorArray&))) {
+    } else if (arg_type ==
+               std::type_index(typeid(const TensorArray&))) {  // NOLINT
       args_def->AppendInput(default_key.backend(),
                             default_tensor_layout,
                             default_key.dtype(),
                             arg_type);
-    } else if (arg_type == std::type_index(typeid(DenseTensor*))) {
+    } else if (arg_type == std::type_index(typeid(DenseTensor*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(std::vector<DenseTensor*>))) {
+    } else if (arg_type ==
+               std::type_index(typeid(std::vector<DenseTensor*>))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(SelectedRows*))) {
+    } else if (arg_type == std::type_index(typeid(SelectedRows*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(TensorArray*))) {
+    } else if (arg_type == std::type_index(typeid(TensorArray*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(SparseCooTensor*))) {
+    } else if (arg_type ==
+               std::type_index(typeid(SparseCooTensor*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(SparseCsrTensor*))) {
+    } else if (arg_type ==
+               std::type_index(typeid(SparseCsrTensor*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(StringTensor*))) {
+    } else if (arg_type == std::type_index(typeid(StringTensor*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
                              arg_type);
-    } else if (arg_type == std::type_index(typeid(ExtendedTensor*))) {
+    } else if (arg_type ==
+               std::type_index(typeid(ExtendedTensor*))) {  // NOLINT
       args_def->AppendOutput(default_key.backend(),
                              default_tensor_layout,
                              default_key.dtype(),
diff --git a/paddle/phi/infermeta/unary.cc b/paddle/phi/infermeta/unary.cc
index 9bee9357e1ed13..392d25fbc2787e 100644
--- a/paddle/phi/infermeta/unary.cc
+++ b/paddle/phi/infermeta/unary.cc
@@ -236,7 +236,7 @@ void ArgMinMaxInferMeta(const MetaTensor& x,
   if (!config.is_runtime && axis.FromTensor()) {
     std::vector<int64_t> vec;
     if (flatten) {
-      if (keepdims) {
+      if (keepdims) {  // NOLINT
         vec = std::vector<int64_t>(x.dims().size(), -1);
       } else {
         vec = {};
@@ -307,7 +307,7 @@ void ArgMinMaxInferMeta(const MetaTensor& x,
 
   std::vector<int64_t> vec;
   if (flatten) {
-    if (keepdims) {
+    if (keepdims) {  // NOLINT
       vec = std::vector<int64_t>(x.dims().size(), 1);
     } else {
       vec = {};
@@ -3989,7 +3989,8 @@ void SplitInferMeta(const MetaTensor& x,
   if ((sections.FromTensor() && !config.is_runtime) || axis_value == -1 ||
       (axis_value >= 0 && x.dims().at(axis_value) <= 0)) {
     std::vector<phi::DDim> out_dims;
-    if ((sections.FromTensor() && !config.is_runtime) || axis_value == -1) {
+    if ((sections.FromTensor() && !config.is_runtime) ||
+        axis_value == -1) {  // NOLINT
       out_dims = std::vector<phi::DDim>(
           sections_data.size(),
           common::make_ddim(std::vector<int>(x.dims().size(), -1)));
@@ -4081,7 +4082,7 @@ void SplitWithNumInferMeta(const MetaTensor& x,
   // fill out dims with -1
   if (axis_value == -1 || (axis_value >= 0 && x.dims().at(axis_value) <= 0)) {
     std::vector<phi::DDim> out_dims;
-    if (axis_value == -1) {
+    if (axis_value == -1) {  // NOLINT
       out_dims = std::vector<phi::DDim>(
           num, common::make_ddim(std::vector<int>(x.dims().size(), -1)));
     } else {
@@ -5370,7 +5371,7 @@ void WeightQuantizeInferMeta(const MetaTensor& x,
   }
 
   std::vector<int64_t> dim_out;
-  if (algo == "weight_only_int8" || algo == "llm.int8") {
+  if (algo == "weight_only_int8" || algo == "llm.int8") {  // NOLINT
     dim_out = std::vector<int64_t>({x_dims[1], x_dims[0]});
   } else if (algo == "weight_only_int4") {
     dim_out = std::vector<int64_t>({x_dims[1] / 2, x_dims[0]});
diff --git a/paddle/phi/kernels/cpu/batch_norm_grad_kernel.cc b/paddle/phi/kernels/cpu/batch_norm_grad_kernel.cc
index 1bdf25dd4eb82d..e9c5ae6a39e4ab 100644
--- a/paddle/phi/kernels/cpu/batch_norm_grad_kernel.cc
+++ b/paddle/phi/kernels/cpu/batch_norm_grad_kernel.cc
@@ -611,7 +611,7 @@ void BatchNormDoubleGradKernel(
     EigenArrayMap<T> ddy_arr(
         ctx.template Alloc<T>(&transformed_ddy), C, sample_size);
     ddy_arr.setZero();
-    if (use_global_stats) {
+    if (use_global_stats) {  // NOLINT
       // math: ddy = r * ddx * inv_var + ddbias +
       //           ddscale * (x - mean) * inv_var
       if (ddX) {
diff --git a/paddle/phi/kernels/cpu/batch_norm_kernel.cc b/paddle/phi/kernels/cpu/batch_norm_kernel.cc
index 39d53fec10a9fa..f6d5e97dc7245a 100644
--- a/paddle/phi/kernels/cpu/batch_norm_kernel.cc
+++ b/paddle/phi/kernels/cpu/batch_norm_kernel.cc
@@ -159,7 +159,7 @@ void BatchNormKernel(const Context& ctx,
 
   // use SavedMean and SavedVariance to do normalize
   Eigen::Array<T, Eigen::Dynamic, 1> inv_std(C);
-  if (global_stats) {
+  if (global_stats) {  // NOLINT
     ConstEigenVectorArrayMap<T> var_arr(variance.data<T>(), C);
     inv_std = (var_arr + epsilon).sqrt().inverse();
   } else {
@@ -178,7 +178,7 @@ void BatchNormKernel(const Context& ctx,
   auto* Bias = bias.get_ptr();
   Eigen::Array<T, Eigen::Dynamic, 1> new_scale(C);
   Eigen::Array<T, Eigen::Dynamic, 1> new_bias(C);
-  if (Scale && Bias) {
+  if (Scale && Bias) {  // NOLINT
     ConstEigenVectorArrayMap<T> scale_arr(Scale->data<T>(), C);
     ConstEigenVectorArrayMap<T> bias_arr(Bias->data<T>(), C);
     new_scale = inv_std * scale_arr;
diff --git a/paddle/phi/kernels/cpu/elementwise_divide_kernel.cc b/paddle/phi/kernels/cpu/elementwise_divide_kernel.cc
index b7fdefe023e73d..ed80148344e1f3 100644
--- a/paddle/phi/kernels/cpu/elementwise_divide_kernel.cc
+++ b/paddle/phi/kernels/cpu/elementwise_divide_kernel.cc
@@ -35,7 +35,7 @@ void DivideKernel(const Context& dev_ctx,
   } else {
     auto x_dims = x.dims();
     auto y_dims = y.dims();
-    if (x_dims.size() >= y_dims.size()) {
+    if (x_dims.size() >= y_dims.size()) {  // NOLINT
       funcs::ElementwiseCompute<funcs::DivideFunctor<T>, T>(
           dev_ctx, x, y, funcs::DivideFunctor<T>(), out, -1);
     } else {
diff --git a/paddle/phi/kernels/cpu/rnn_grad_kernel.cc b/paddle/phi/kernels/cpu/rnn_grad_kernel.cc
index 3e0e4c7a3d7a5a..284f12381bd358 100644
--- a/paddle/phi/kernels/cpu/rnn_grad_kernel.cc
+++ b/paddle/phi/kernels/cpu/rnn_grad_kernel.cc
@@ -1311,7 +1311,7 @@ void RnnGradKernel(const Context& dev_ctx,
         pre_state_grad,
         weight_grad_list);
     // run gru
-  } else if (is_rnn_relu(mode)) {
+  } else if (is_rnn_relu(mode)) {  // NOLINT
     gate_num = 1;
     RnnGradFunc<SimpleRNNGradCell<T, funcs::ReluGradFunctor>,
                 SingleGradLayer,
diff --git a/paddle/phi/kernels/cpu/rnn_kernel.cc b/paddle/phi/kernels/cpu/rnn_kernel.cc
index a0035c6db4a75d..5b594089793c8c 100644
--- a/paddle/phi/kernels/cpu/rnn_kernel.cc
+++ b/paddle/phi/kernels/cpu/rnn_kernel.cc
@@ -868,7 +868,7 @@ void RnnKernel(const Context& dev_ctx,
         is_test,
         seed,
         reserve);
-  } else if (is_rnn_relu(mode)) {
+  } else if (is_rnn_relu(mode)) {  // NOLINT
     gate_num = 1;
     RnnFunc<SimpleRNNCell<T,
                           funcs::ReluCPUFunctor,
diff --git a/paddle/phi/kernels/funcs/sequence_pooling.cc b/paddle/phi/kernels/funcs/sequence_pooling.cc
index 004bef522ab16a..f4ee9c323366e6 100644
--- a/paddle/phi/kernels/funcs/sequence_pooling.cc
+++ b/paddle/phi/kernels/funcs/sequence_pooling.cc
@@ -417,7 +417,7 @@ class SequencePoolFunctor<phi::CPUContext, T> {
       int64_t h = static_cast<int64_t>(lod[i + 1] - lod[i]);
       auto in_e = EigenMatrix<T>::From(in_t, common::make_ddim({h, w}));
       auto out_e = EigenVector<T>::Flatten(out_t);
-      if (pooltype == "AVERAGE") {
+      if (pooltype == "AVERAGE") {  // NOLINT
         out_e.device(place) = in_e.mean(Eigen::array<int, 1>({{0}}));
       } else if (pooltype == "SQRT") {
         out_e.device(place) = in_e.sum(Eigen::array<int, 1>({{0}})) /
diff --git a/paddle/phi/kernels/legacy/cpu/elementwise_kernel.cc b/paddle/phi/kernels/legacy/cpu/elementwise_kernel.cc
index dafbf2889277d1..84ebbf04fee11c 100644
--- a/paddle/phi/kernels/legacy/cpu/elementwise_kernel.cc
+++ b/paddle/phi/kernels/legacy/cpu/elementwise_kernel.cc
@@ -55,7 +55,7 @@ void RemainderRawKernel(const Context& dev_ctx,
   dev_ctx.template Alloc<T>(out);
   auto x_dims = x.dims();
   auto y_dims = y.dims();
-  if (x_dims.size() >= y_dims.size()) {
+  if (x_dims.size() >= y_dims.size()) {  // NOLINT
     funcs::ElementwiseCompute<funcs::RemainderFunctor<T>, T>(
         dev_ctx, x, y, funcs::RemainderFunctor<T>(), out, axis);
   } else {
@@ -74,7 +74,7 @@ void FloorDivideRawKernel(const Context& dev_ctx,
   dev_ctx.template Alloc<T>(out);
   auto x_dims = x.dims();
   auto y_dims = y.dims();
-  if (x_dims.size() >= y_dims.size()) {
+  if (x_dims.size() >= y_dims.size()) {  // NOLINT
     funcs::ElementwiseCompute<funcs::FloorDivideFunctor<T>, T>(
         dev_ctx, x, y, funcs::FloorDivideFunctor<T>(), out, axis);
   } else {
diff --git a/test/cpp/fluid/framework/details/fused_broadcast_op_handle_test.cc b/test/cpp/fluid/framework/details/fused_broadcast_op_handle_test.cc
index 786b857a80dcca..aee187d77f4843 100644
--- a/test/cpp/fluid/framework/details/fused_broadcast_op_handle_test.cc
+++ b/test/cpp/fluid/framework/details/fused_broadcast_op_handle_test.cc
@@ -56,7 +56,7 @@ struct TestFusedBroadcastOpHandle : TestBroadcastOpHandle {
     // create op handle node
     nodes_.emplace_back(
         ir::CreateNodeForTest("fused_broadcast", ir::Node::Type::kOperation));
-    if (use_device_ == p::kCUDA) {
+    if (use_device_ == p::kCUDA) {  // NOLINT
 #if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
       op_handle_ = new FusedBroadcastOpHandle(
           nodes_.back().get(), local_scopes_, place_list_, nccl_ctxs_.get());
diff --git a/test/cpp/imperative/test_gradient_accmulator.cc b/test/cpp/imperative/test_gradient_accmulator.cc
index 58fb2c54df7741..303c3589534e43 100644
--- a/test/cpp/imperative/test_gradient_accmulator.cc
+++ b/test/cpp/imperative/test_gradient_accmulator.cc
@@ -379,7 +379,7 @@ static framework::Variable RandomSelectedRows(framework::DDim dims,
 
 static std::unique_ptr<GradientAccumulator> CreateAccumulator(
     const std::shared_ptr<VariableWrapper>& var, bool sort_gradient) {
-  if (sort_gradient) {
+  if (sort_gradient) {  // NOLINT
     return std::unique_ptr<GradientAccumulator>(
         new SortedGradientAccumulator(var.get()));
   } else {
@@ -403,7 +403,7 @@ static void TestGradientAccumulatorTestUnchangeInput(
   std::mt19937 engine(seed);
 
   auto create_var = [&](bool use_tensor) {
-    if (use_tensor) {
+    if (use_tensor) {  // NOLINT
       return RandomTensor<float>(dim, place);
     } else {
       return RandomSelectedRows<float>(dim, place, dist(engine));