add fp16 sub_graph preProcess

mindspore/lite/src/common/log_adapter.h

@@ -106,7 +106,7 @@ class LogWriter {
 
 }  // namespace mindspore
 
-#ifdef DEBUG
+#ifdef Debug
 #include <cassert>
 #define MS_ASSERT(f) assert(f)
 #else

mindspore/lite/src/executor.h

@@ -43,6 +43,5 @@ class Executor {
 
   int TransformTensorLayout(Tensor *tensor, schema::Format dst_format, Allocator *allocator = nullptr);
 };
-
 }  // namespace mindspore::lite
 #endif

mindspore/lite/src/lite_session.cc

@@ -162,7 +162,6 @@ void LiteSession::InitGraphInputMSTensors() {
 void LiteSession::InitGraphOutputTensors(const lite::Model *model) {
   MS_ASSERT(model != nullptr);
   MS_ASSERT(this->outputs_.empty());
-  MS_ASSERT(meta_graph != nullptr);
   auto graph_out_size = model->sub_graphs_.front()->output_indices_.size();
   for (size_t i = 0; i < graph_out_size; ++i) {
     auto out_tensor_idx = model->sub_graphs_.front()->output_indices_[i];
@@ -181,7 +180,7 @@ void LiteSession::InitGraphInputMap(const lite::Model *model) {
   for (auto in_node_index : graph_input_node_indexes) {
     auto in_node = model->all_nodes_[in_node_index];
     MS_ASSERT(in_node != nullptr);
-    MS_ASSERT(this->input_map_.find(in_node->name()->str()) == this->input_map_.end());
+    MS_ASSERT(this->input_map_.find(in_node->name_) == this->input_map_.end());
     auto in_size = in_node->input_indices_.size();
     for (size_t i = 0; i < in_size; ++i) {
       auto in_tensor_index = size_t(in_node->input_indices_[i]);
@@ -215,7 +214,6 @@ void LiteSession::InitGraphOutputNodeMap(const lite::Model *model) {
   for (auto out_node_index : graph_output_node_indexes) {
     auto out_node = model->all_nodes_[out_node_index];
     MS_ASSERT(out_node != nullptr);
-    MS_ASSERT(this->output_map_.find(out_node->name()->str()) == this->output_map_.end());
     auto out_size = out_node->output_indices_.size();
     for (size_t i = 0; i < out_size; ++i) {
       auto out_tensor_index = out_node->output_indices_[i];

mindspore/lite/src/ops/cast.cc

@@ -95,7 +95,6 @@ int Cast::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> output
   }
   output->SetFormat(input->GetFormat());
 
-  MS_ASSERT(cast_prim != nullptr);
   output->set_data_type(static_cast<TypeId>(GetDstT()));
   if (!GetInferFlag()) {
     return RET_OK;

mindspore/lite/src/ops/concat.cc

@@ -104,7 +104,6 @@ int Concat::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> outp
     return RET_OK;
   }
 
-  MS_ASSERT(concat_prim != nullptr);
   auto input0_shape = inputs_.at(0)->shape();
   auto axis = GetAxis() < 0 ? GetAxis() + input0_shape.size() : GetAxis();
   if (axis < 0 || axis >= input0_shape.size()) {

mindspore/lite/src/ops/conv2d_grad_filter.cc

@@ -197,7 +197,8 @@ int Conv2DGradFilter::InferShape(std::vector<Tensor *> inputs, std::vector<Tenso
 
   auto *in0 = inputs.at(0);
   auto *in = inputs.at(2);
-  MS_ASSERT(out != nullptr);
+  MS_ASSERT(in0 != nullptr);
+  MS_ASSERT(in != nullptr);
 
   std::vector<int> output_shape;
   int *out_shape = reinterpret_cast<int *>(in->MutableData());

mindspore/lite/src/ops/conv2d_grad_input.cc

@@ -200,7 +200,8 @@ int Conv2DGradInput::InferShape(std::vector<Tensor *> inputs, std::vector<Tensor
 
   auto *in0 = inputs.at(0);
   auto *in = inputs.at(2);
-  MS_ASSERT(out != nullptr);
+  MS_ASSERT(in0 != nullptr);
+  MS_ASSERT(in != nullptr);
 
   std::vector<int> output_shape;
   int *out_shape = reinterpret_cast<int *>(in->MutableData());

mindspore/lite/src/ops/detection_post_process.cc

@@ -169,8 +169,8 @@ int DetectionPostProcess::InferShape(std::vector<lite::Tensor *> inputs_, std::v
   const auto input_anchors_shape = anchors->shape();
   MS_ASSERT(input_scores_shape[2] >= GetNumClasses());
   MS_ASSERT(input_scores_shape[2] - GetNumClasses() <= 1);
-  MS_ASSERT(input_box_shape[1] = input_scores_shape[1]);
-  MS_ASSERT(input_box_shape[1] = input_anchors_shape[0]);
+  MS_ASSERT(input_box_shape[1] == input_scores_shape[1]);
+  MS_ASSERT(input_box_shape[1] == input_anchors_shape[0]);
 
   auto detected_boxes = outputs_.at(0);
   MS_ASSERT(detected_boxes != nullptr);

mindspore/lite/src/ops/gather.cc

@@ -113,7 +113,6 @@ int Gather::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> outp
     return RET_OK;
   }
 
-  MS_ASSERT(gather_prim != nullptr);
   int axis = GetAxis();
   int batch_dims = GetBatchDims();
   if (axis < 0) {

mindspore/lite/src/ops/group_conv2d_grad_input.cc

@@ -149,7 +149,8 @@ int GroupConv2DGradInput::InferShape(std::vector<Tensor *> inputs, std::vector<T
 
   auto *in0 = inputs.at(0);
   auto *in = inputs.at(2);
-  MS_ASSERT(out != nullptr);
+  MS_ASSERT(in0 != nullptr);
+  MS_ASSERT(in != nullptr);
 
   std::vector<int> output_shape;
   int *out_shape = reinterpret_cast<int *>(in->MutableData());

mindspore/lite/src/ops/hashtable_lookup.cc

@@ -46,7 +46,6 @@ int HashtableLookup::InferShape(std::vector<Tensor *> inputs_, std::vector<Tenso
   auto output = outputs_.at(0);
   auto hits = outputs_.at(1);
   MS_ASSERT(input != nullptr);
-  MS_ASSERT(keys != nullptr);
   MS_ASSERT(values != nullptr);
   MS_ASSERT(output != nullptr);
   MS_ASSERT(hits != nullptr);

mindspore/lite/src/ops/lsh_projection.cc

@@ -65,7 +65,7 @@ int LshProjection::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor
 
   if (inputs_.size() == kMultiNum) {
     MS_ASSERT(inputs_.at(2)->shape().size() == 1);
-    MS_ASSERT(inputs_.at(2)->DimensionSize(0) == in_value->DimensionSize(0));
+    MS_ASSERT(inputs_.at(2)->DimensionSize(0) == inputs_.at(1)->DimensionSize(0));
   }
 
   auto out_tensor = outputs_.front();

mindspore/lite/src/ops/lstm.cc

@@ -60,7 +60,7 @@ int Lstm::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> output
   auto input = inputs_.front();
   MS_ASSERT(input != nullptr);
   auto weight_i = inputs_[1];
-  MS_ASSERT(input0 != nullptr);
+  MS_ASSERT(input != nullptr);
   auto output = outputs_.front();
   MS_ASSERT(output != nullptr);
   for (int i = 0; i < kLstmOutputNum; i++) {

mindspore/lite/src/ops/pooling.cc

@@ -186,7 +186,6 @@ int Pooling::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> out
   int input_h = input->shape().at(1);
   int input_w = input->shape().at(2);
 
-  MS_ASSERT(pooling_prim != nullptr);
   auto window_h = GetWindowH();
   auto window_w = GetWindowW();
   if (GetGlobal()) {

mindspore/lite/src/ops/prior_box.cc

@@ -132,7 +132,7 @@ constexpr int kPriorBoxW = 1;
 constexpr int kPriorBoxC = 2;
 }  // namespace
 int PriorBox::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> outputs_) {
-  MS_ASSERT(param != nullptr);
+  MS_ASSERT(this->primitive_ != nullptr);
   auto input = inputs_.at(0);
   MS_ASSERT(input != nullptr);
   auto output = outputs_.at(0);
@@ -144,7 +144,6 @@ int PriorBox::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> ou
   }
   std::vector<float> different_aspect_ratios{1.0f};
   auto aspect_ratios = GetAspectRatios();
-  MS_ASSERT(aspect_ratios != nullptr);
   for (size_t i = 0; i < aspect_ratios.size(); i++) {
     float ratio = aspect_ratios[i];
     bool exist = std::any_of(different_aspect_ratios.begin(), different_aspect_ratios.end(),

mindspore/lite/src/ops/quant_dtype_cast.cc

@@ -59,7 +59,7 @@ int QuantDTypeCast::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor
   MS_ASSERT(input != nullptr);
   auto output = outputs_.front();
   MS_ASSERT(output != nullptr);
-  MS_ASSERT(input->data_type() == param->srcT);
+  MS_ASSERT(input->data_type() == this->GetSrcT());
   output->set_data_type(static_cast<TypeId>(GetDstT()));
   output->SetFormat(input->GetFormat());
   if (!GetInferFlag()) {

mindspore/lite/src/ops/range.cc

@@ -64,8 +64,6 @@ int Range::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> outpu
   auto output = outputs_.front();
   MS_ASSERT(output != nullptr);
 
-  MS_ASSERT(range_prim != nullptr);
-
   output->set_data_type(input->data_type());
   output->SetFormat(input->GetFormat());
   if (!GetInferFlag()) {

mindspore/lite/src/ops/reshape.cc

@@ -176,7 +176,6 @@ int Reshape::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> out
     return RET_OK;
   }
 
-  MS_ASSERT(reshape_prim != nullptr);
   std::vector<int> out_shape;
   if (inputs_.size() == kDoubleNum) {
     auto shape_tensor = inputs_.at(1);

mindspore/lite/src/ops/sparse_to_dense.cc

@@ -54,7 +54,6 @@ Registry SparseToDenseRegistry(schema::PrimitiveType_SparseToDense, SparseToDens
 
 int SparseToDense::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> outputs_) {
   MS_ASSERT(this->primitive_ != nullptr);
-  MS_ASSERT(output_shape != nullptr);
   auto output = outputs_.front();
   if (output == nullptr) {
     MS_LOG(ERROR) << "output null pointer dereferencing.";

mindspore/lite/src/ops/split.cc

@@ -104,7 +104,6 @@ int Split::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> outpu
   MS_ASSERT(this->primitive_ != nullptr);
   auto input = inputs_.front();
   MS_ASSERT(input != nullptr);
-  MS_ASSERT(spilt_prim != nullptr);
   if (inputs_.size() != kSplitInputNum) {
     MS_LOG(ERROR) << "inputs number is not equal to " << kSplitInputNum;
     return RET_ERROR;

mindspore/lite/src/ops/tile.cc

@@ -138,7 +138,6 @@ int Tile::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> output
     return RET_OK;
   }
 
-  MS_ASSERT(tile_prim != nullptr);
   std::vector<int> out_shape;
   std::vector<int> multiples = GetMultiples();
   const size_t in_dims = input->shape().size();

mindspore/lite/src/ops/topk.cc

@@ -71,7 +71,6 @@ int TopK::InferShape(std::vector<Tensor *> inputs_, std::vector<Tensor *> output
   if (!GetInferFlag()) {
     return RET_OK;
   }
-  MS_ASSERT(topk_prim != nullptr);
   auto out_shape = input->shape();
   out_shape[out_shape.size() - 1] = GetK();
   if (inputs_.size() == kDoubleNum && inputs_.at(1)->data_c() != nullptr) {

mindspore/lite/src/runtime/kernel/arm/fp16/convolution_fp16.cc

@@ -82,7 +82,7 @@ int ConvolutionFP16CPUKernel::InitWeightBias() {
       fp16_bias_data[i] = (float16_t)ori_bias[i];
     }
   } else {
-    MS_ASSERT(inputs_.size() == kInputSize1);
+    MS_ASSERT(in_tensors_.size() == kInputSize1);
   }
   return RET_OK;
 }

mindspore/lite/src/runtime/kernel/arm/fp16/convolution_winograd_fp16.cc

@@ -97,7 +97,7 @@ int ConvolutionWinogradFP16CPUKernel::InitWeightBias() {
       fp16_bias_data[i] = (float16_t)ori_bias[i];
     }
   } else {
-    MS_ASSERT(inputs_.size() == kInputSize1);
+    MS_ASSERT(in_tensors_.size() == kInputSize1);
   }
   return RET_OK;
 }

mindspore/lite/src/runtime/kernel/arm/fp16/deconvolution_winograd_fp16.cc

@@ -346,7 +346,7 @@ int DeConvWinogradFp16CPUKernel::InitDataParam() {
       fp16_bias_data[i] = (float16_t)src_bias[i];
     }
   } else {
-    MS_ASSERT(inputs_.size() == kInputSize1);
+    MS_ASSERT(in_tensors_.size() == kInputSize1);
   }
 
   return RET_OK;