fix_decoder_3-master

mindspore/lite/src/runtime/kernel/arm/base/tensorlist_fromtensor.cc

@@ -15,7 +15,7 @@
  */
 #include "include/errorcode.h"
 #include "src/kernel_registry.h"
-#include "src/runtime/kernel/arm/fp32/tensorlist_fromtensor_fp32.h"
+#include "src/runtime/kernel/arm/base/tensorlist_fromtensor.h"
 #include "src/runtime/runtime_api.h"
 
 using mindspore::kernel::KERNEL_ARCH::kCPU;
@@ -52,14 +52,7 @@ int TensorListFromTensorCPUKernel::IsCompatibleShape() {
   return RET_OK;
 }
 
-int TensorListFromTensorCPUKernel::Init() {
-#ifdef ENABLE_FP16
-  if (lite::IsSupportFloat16() && context_->IsCpuFloat16Enabled() && dtype_ == kNumberTypeFloat32) {
-    dtype_ = kNumberTypeFloat16;
-  }
-#endif
-  return RET_OK;
-}
+int TensorListFromTensorCPUKernel::Init() { return RET_OK; }
 
 int TensorListFromTensorCPUKernel::ReSize() { return RET_OK; }
 
@@ -71,6 +64,7 @@ int TensorListFromTensorCPUKernel::Run() {
     MS_LOG(ERROR) << "IsNotCompatibleShape!";
     return RET_ERROR;
   }
+  dtype_ = in_tensors_[0]->data_type();
   if (input0_->shape().size() == 0) {
     MS_LOG(ERROR) << "input0_->shape().size():" << input0_->shape().size() << " must be greater than 0";
   }
@@ -97,9 +91,10 @@ int TensorListFromTensorCPUKernel::Run() {
                     << " must be euqal to devision_dim0:" << devision_dim0;
       return RET_ERROR;
     }
-    auto out_data = out_ptr->MutableData();
+    auto out_data = out_ptr->data_c();
     MS_ASSERT(out_data != nullptr);
     memcpy(out_data, in_data, data_offset);
+    out_ptr->set_data_type(dtype_);
     in_data += data_offset;
   }
   output0->set_tensors_data_type(dtype_);

mindspore/lite/src/runtime/kernel/arm/base/tensorlist_getitem.cc

@@ -16,7 +16,7 @@
 #include "include/errorcode.h"
 #include "include/ms_tensor.h"
 #include "src/kernel_registry.h"
-#include "src/runtime/kernel/arm/fp32/tensorlist_getitem_fp32.h"
+#include "src/runtime/kernel/arm/base/tensorlist_getitem.h"
 #include "src/runtime/runtime_api.h"
 
 using mindspore::kernel::KERNEL_ARCH::kCPU;
@@ -28,16 +28,7 @@ using mindspore::schema::PrimitiveType_TensorListGetItem;
 
 namespace mindspore::kernel {
 
-int TensorListGetItemCPUKernel::Init() {
-  MS_ASSERT(in_tensors_.size() >= 2);
-  MS_ASSERT(in_tensors_.at(0) != nullptr);
-#ifdef ENABLE_FP16
-  if (lite::IsSupportFloat16() && context_->IsCpuFloat16Enabled() && dtype_ == kNumberTypeFloat32) {
-    dtype_ = kNumberTypeFloat16;
-  }
-#endif
-  return RET_OK;
-}
+int TensorListGetItemCPUKernel::Init() { return RET_OK; }
 
 int TensorListGetItemCPUKernel::Run() {
   MS_ASSERT(in_tensors_.size() >= 2);
@@ -48,10 +39,7 @@ int TensorListGetItemCPUKernel::Run() {
   if (input0->root_tensor() != nullptr) {
     input0 = reinterpret_cast<lite::TensorList *>(input0->root_tensor());
   }
-  if (dtype_ != input0->tensors_data_type()) {
-    MS_LOG(ERROR) << "op dtype: " << dtype_ << " is not equal in_tensor[0] dtype: " << input0->tensors_data_type();
-    return RET_ERROR;
-  }
+  dtype_ = input0->tensors_data_type();
   MS_ASSERT(in_tensors_.at(1)->data_c() != nullptr);
   index_ = reinterpret_cast<int *>(in_tensors_.at(1)->data_c())[0];
   int dim0 = input0->ElementsNum() - 1;

mindspore/lite/src/runtime/kernel/arm/base/tensorlist_reserve.cc

@@ -16,7 +16,7 @@
 #include <vector>
 #include "include/errorcode.h"
 #include "src/kernel_registry.h"
-#include "src/runtime/kernel/arm/fp32/tensorlist_reserve_fp32.h"
+#include "src/runtime/kernel/arm/base/tensorlist_reserve.h"
 
 using mindspore::kernel::KERNEL_ARCH::kCPU;
 using mindspore::lite::KernelRegistrar;
@@ -27,14 +27,7 @@ using mindspore::schema::PrimitiveType_TensorListReserve;
 
 namespace mindspore::kernel {
 
-int TensorListReserveCPUKernel::Init() {
-#ifdef ENABLE_FP16
-  if (lite::IsSupportFloat16() && context_->IsCpuFloat16Enabled() && element_dtype_ == kNumberTypeFloat32) {
-    element_dtype_ = kNumberTypeFloat16;
-  }
-#endif
-  return RET_OK;
-}
+int TensorListReserveCPUKernel::Init() { return RET_OK; }
 
 int TensorListReserveCPUKernel::Run() {
   auto input0 = in_tensors_.at(0);

mindspore/lite/src/runtime/kernel/arm/base/tensorlist_setitem.cc

@@ -16,7 +16,7 @@
 #include "include/errorcode.h"
 #include "include/ms_tensor.h"
 #include "src/kernel_registry.h"
-#include "src/runtime/kernel/arm/fp32/tensorlist_setitem_fp32.h"
+#include "src/runtime/kernel/arm/base/tensorlist_setitem.h"
 #include "src/runtime/runtime_api.h"
 
 using mindspore::kernel::KERNEL_ARCH::kCPU;
@@ -28,21 +28,9 @@ using mindspore::schema::PrimitiveType_TensorListSetItem;
 
 namespace mindspore::kernel {
 
-int TensorListSetItemCPUKernel::Init() {
-#ifdef ENABLE_FP16
-  if (lite::IsSupportFloat16() && context_->IsCpuFloat16Enabled() && dtype_ == kNumberTypeFloat32) {
-    dtype_ = kNumberTypeFloat16;
-  }
-#endif
-  return RET_OK;
-}
+int TensorListSetItemCPUKernel::Init() { return RET_OK; }
 
 int TensorListSetItemCPUKernel::CheckParam() {
-  if (dtype_ != kTypeUnknown && input0_->tensors_data_type() != kTypeUnknown &&
-      dtype_ != input0_->tensors_data_type()) {
-    MS_LOG(ERROR) << "op dtype:" << dtype_ << " is not equal in_tensors[0] dtype:" << input0_->tensors_data_type();
-    return RET_ERROR;
-  }
   if (in_tensors_[1]->data_type() != kNumberTypeInt && in_tensors_[1]->data_type() != kNumberTypeInt32) {
     MS_LOG(ERROR) << "in_tensors_[1]->data_type():" << in_tensors_[1]->data_type() << " must be int";
     return RET_ERROR;
@@ -70,7 +58,6 @@ int TensorListSetItemCPUKernel::IncrementOutputSize(int origin_size) {
 
 int TensorListSetItemCPUKernel::Run() {
   input0_ = reinterpret_cast<lite::TensorList *>(in_tensors_[0]);
-
   if (CheckParam() != RET_OK) {
     MS_LOG(ERROR) << "check param failed.";
     return RET_ERROR;

mindspore/lite/src/runtime/kernel/arm/base/tensorlist_setitem.h

@@ -28,8 +28,7 @@ class TensorListSetItemCPUKernel : public LiteKernel {
  public:
   TensorListSetItemCPUKernel(OpParameter *parameter, const std::vector<lite::Tensor *> &inputs,
                              const std::vector<lite::Tensor *> &outputs, const lite::InnerContext *ctx)
-      : LiteKernel(parameter, inputs, outputs, ctx),
-        dtype_(static_cast<TypeId>(reinterpret_cast<TensorListParameter *>(parameter)->element_dtype_)) {}
+      : LiteKernel(parameter, inputs, outputs, ctx) {}
   ~TensorListSetItemCPUKernel() = default;
 
   int Init() override;
@@ -43,7 +42,6 @@ class TensorListSetItemCPUKernel : public LiteKernel {
   lite::Tensor *input2_ = nullptr;
   lite::TensorList *output0_ = nullptr;
   int index_ = 0;
-  TypeId dtype_ = kTypeUnknown;
 };
 }  // namespace mindspore::kernel
 

mindspore/lite/src/runtime/kernel/arm/base/tensorlist_stack.cc

@@ -19,7 +19,7 @@
 #include "include/errorcode.h"
 #include "ir/dtype/type_id.h"
 #include "src/kernel_registry.h"
-#include "src/runtime/kernel/arm/fp32/tensorlist_stack_fp32.h"
+#include "src/runtime/kernel/arm/base/tensorlist_stack.h"
 
 using mindspore::kernel::KERNEL_ARCH::kCPU;
 using mindspore::lite::KernelRegistrar;
@@ -31,11 +31,6 @@ using mindspore::schema::PrimitiveType_TensorListStack;
 namespace mindspore::kernel {
 
 int TensorListStackCPUKernel::CheckParam() {
-  if (dtype_ != kTypeUnknown && input0_->tensors_data_type() != dtype_) {
-    MS_LOG(ERROR) << "in_tensors_[0].tensors_data_type:[" << input0_->tensors_data_type() << "] must be equal "
-                  << "param.data_type:[" << dtype_ << "]";
-    return RET_ERROR;
-  }
   if (num_element_ != -1 && input0_->ElementsNum() != num_element_) {
     MS_LOG(ERROR) << "in_tensors_[0].ElementsNum():[" << input0_->ElementsNum() << "] must be equal "
                   << "param.elements_num:[" << num_element_ << "]";
@@ -60,11 +55,6 @@ int TensorListStackCPUKernel::Init() {
   MS_ASSERT(input0_ != nullptr);
   output0_ = out_tensors_[0];
   MS_ASSERT(output0_ != nullptr);
-#ifdef ENABLE_FP16
-  if (lite::IsSupportFloat16() && context_->IsCpuFloat16Enabled() && dtype_ == kNumberTypeFloat32) {
-    dtype_ = kNumberTypeFloat16;
-  }
-#endif
   return RET_OK;
 }
 
@@ -146,18 +136,11 @@ int TensorListStackCPUKernel::MergeSubShape(const std::vector<int> &shape) {
 }
 
 int TensorListStackCPUKernel::Run() {
-  if (dtype_ == kTypeUnknown) {
-    dtype_ = input0_->tensors_data_type();
-#ifdef ENABLE_FP16
-    if (lite::IsSupportFloat16() && context_->IsCpuFloat16Enabled() && dtype_ == kNumberTypeFloat32) {
-      dtype_ = kNumberTypeFloat16;
-    }
-#endif
-  }
   if (CheckParam() != RET_OK) {
     MS_LOG(ERROR) << "CheckParam failed!";
     return RET_ERROR;
   }
+  dtype_ = input0_->tensors_data_type();
   if (output0_->ElementsNum() == 0) {
     return RET_OK;
   }

mindspore/lite/src/runtime/kernel/arm/fp16/arithmetic_self_fp16.cc

@@ -76,7 +76,11 @@ int ArithmeticSelfFp16CPUKernel::Run() {
   auto input_tensor = in_tensors_.at(0);
   auto output_tensor = out_tensors_.at(0);
 
-  input_fp16_ptr_ = reinterpret_cast<float16_t *>(input_tensor->data_c());
+  if (input_tensor->data_type() == kNumberTypeFloat32) {
+    input_fp16_ptr_ = ConvertInputFp32toFp16(input_tensor, context_);
+  } else {
+    input_fp16_ptr_ = reinterpret_cast<float16_t *>(input_tensor->data_c());
+  }
   output_fp16_ptr_ = reinterpret_cast<float16_t *>(output_tensor->data_c());
 
   auto ret = ParallelLaunch(this->context_->thread_pool_, ArithmeticSelfRun, this, op_parameter_->thread_num_);

mindspore/lite/src/scheduler.cc

@@ -199,7 +199,8 @@ int Scheduler::InferSubGraphShape(size_t subgraph_index, bool *infer_shape_inter
 }
 
 kernel::LiteKernel *Scheduler::FindBackendKernel(const std::vector<Tensor *> &in_tensors,
-                                                 const std::vector<Tensor *> &out_tensors, const Model::Node *node) {
+                                                 const std::vector<Tensor *> &out_tensors, const Model::Node *node,
+                                                 TypeId prefer_data_type) {
   kernel::LiteKernel *kernel = nullptr;
   TypeId data_type = GetFirstFp32Fp16OrInt8Type(in_tensors);
   OpParameter *op_parameter = op_parameters_[node->output_indices_.at(0)];
@@ -272,7 +273,8 @@ kernel::LiteKernel *Scheduler::FindBackendKernel(const std::vector<Tensor *> &in
     }
   }
 #endif
-  if (mindspore::lite::IsSupportFloat16() &&
+  if ((prefer_data_type == kNumberTypeFloat16 || prefer_data_type == kTypeUnknown) &&
+      mindspore::lite::IsSupportFloat16() &&
       ((context_->IsCpuFloat16Enabled() && data_type == kNumberTypeFloat32) || data_type == kNumberTypeFloat16)) {
     kernel::KernelKey fp16_cpu_desc{desc.arch, kNumberTypeFloat16, desc.type};
     auto tensor_origin_data_map =
@@ -301,15 +303,17 @@ kernel::LiteKernel *Scheduler::FindBackendKernel(const std::vector<Tensor *> &in
     MS_LOG(DEBUG) << "Get fp16 op failed, back to fp32 op.";
     desc.data_type = kNumberTypeFloat32;
   }
-  auto tensor_origin_data_map = DequantUtil::DequantTensor(op_parameter, in_tensors, desc.data_type, need_restore);
-  auto ret = KernelRegistry::GetInstance()->GetKernel(in_tensors, out_tensors, context_, desc, op_parameter, &kernel);
-  DequantUtil::RestoreTensorData(tensor_origin_data_map);
-  if (ret == RET_OK) {
-    return kernel;
-  } else if (ret == RET_ERROR) {
-    ret = InferNodeShape(node, &infer_shape_interrupt);
-    if (!(ret == RET_INFER_INVALID || ret == RET_OK)) {
-      MS_LOG(ERROR) << "Try repeat infer fail: " << node->name_;
+  if (prefer_data_type == kNumberTypeFloat32 || prefer_data_type == kTypeUnknown) {
+    auto tensor_origin_data_map = DequantUtil::DequantTensor(op_parameter, in_tensors, desc.data_type, need_restore);
+    auto ret = KernelRegistry::GetInstance()->GetKernel(in_tensors, out_tensors, context_, desc, op_parameter, &kernel);
+    DequantUtil::RestoreTensorData(tensor_origin_data_map);
+    if (ret == RET_OK) {
+      return kernel;
+    } else if (ret == RET_ERROR) {
+      ret = InferNodeShape(node, &infer_shape_interrupt);
+      if (!(ret == RET_INFER_INVALID || ret == RET_OK)) {
+        MS_LOG(ERROR) << "Try repeat infer fail: " << node->name_;
+      }
     }
   }
   return nullptr;
@@ -327,7 +331,7 @@ kernel::LiteKernel *Scheduler::SchedulePartialToKernel(const lite::Model::Node *
   std::vector<kernel::LiteKernel *> sub_kernels;
   std::vector<lite::Tensor *> in_tensors;
   std::vector<lite::Tensor *> out_tensors;
-  auto ret = ScheduleSubGraphToKernels(sub_graph_index, &sub_kernels, &in_tensors, &out_tensors);
+  auto ret = ScheduleSubGraphToKernels(sub_graph_index, &sub_kernels, &in_tensors, &out_tensors, kNumberTypeFloat32);
   if (ret != RET_OK) {
     MS_LOG(ERROR) << "Schedule partial failed, name: " << src_node->name_;
     return nullptr;
@@ -338,11 +342,11 @@ kernel::LiteKernel *Scheduler::SchedulePartialToKernel(const lite::Model::Node *
   return subgraph;
 }
 
-kernel::LiteKernel *Scheduler::ScheduleNodeToKernel(const lite::Model::Node *src_node) {
+kernel::LiteKernel *Scheduler::ScheduleNodeToKernel(const lite::Model::Node *src_node, TypeId prefer_data_type) {
   std::vector<Tensor *> inputs;
   std::vector<Tensor *> outputs;
   FindNodeInoutTensors(*src_node, &inputs, &outputs);
-  auto *kernel = this->FindBackendKernel(inputs, outputs, src_node);
+  auto *kernel = this->FindBackendKernel(inputs, outputs, src_node, prefer_data_type);
   if (kernel == nullptr) {
     MS_LOG(ERROR) << "FindBackendKernel return nullptr, name: " << src_node->name_
                   << ", type: " << PrimitiveTypeName(GetPrimitiveType(src_node->primitive_));
@@ -355,7 +359,7 @@ kernel::LiteKernel *Scheduler::ScheduleNodeToKernel(const lite::Model::Node *src
 
 int Scheduler::ScheduleSubGraphToKernels(size_t subgraph_index, std::vector<kernel::LiteKernel *> *dst_kernels,
                                          std::vector<lite::Tensor *> *in_tensors,
-                                         std::vector<lite::Tensor *> *out_tensors) {
+                                         std::vector<lite::Tensor *> *out_tensors, TypeId prefer_data_type) {
   MS_ASSERT(src_model_ != nullptr);
   MS_ASSERT(!src_model_->sub_graphs_.empty());
   MS_ASSERT(src_model_->sub_graphs_.size() > subgraph_index);
@@ -372,7 +376,7 @@ int Scheduler::ScheduleSubGraphToKernels(size_t subgraph_index, std::vector<kern
     if (IsPartialNode(primitive)) {  // sub_graph
       kernel = SchedulePartialToKernel(node);
     } else {  // kernel
-      kernel = ScheduleNodeToKernel(node);
+      kernel = ScheduleNodeToKernel(node, prefer_data_type);
     }
     if (kernel == nullptr) {
       MS_LOG(ERROR) << "FindBackendKernel return nullptr, name: " << node->name_

mindspore/lite/src/scheduler.h

@@ -57,14 +57,16 @@ class Scheduler {
 
   // schedule a node to kernel according to context and kernels registered
   kernel::LiteKernel *FindBackendKernel(const std::vector<Tensor *> &in_tensors,
-                                        const std::vector<Tensor *> &out_tensors, const Model::Node *node);
+                                        const std::vector<Tensor *> &out_tensors, const Model::Node *node,
+                                        TypeId prefer_data_type = kTypeUnknown);
   // schedule a partial node to a subgraph_kernel
   kernel::LiteKernel *SchedulePartialToKernel(const lite::Model::Node *src_node);
   // schedule a node to a kernel
-  kernel::LiteKernel *ScheduleNodeToKernel(const lite::Model::Node *src_node);
+  kernel::LiteKernel *ScheduleNodeToKernel(const lite::Model::Node *src_node, TypeId prefer_data_type = kTypeUnknown);
   // schedule a Model::SubGraph into a vector of kernel and subgraph_kernel
   int ScheduleSubGraphToKernels(size_t subgraph_index, std::vector<kernel::LiteKernel *> *dst_kernels,
-                                std::vector<lite::Tensor *> *in_tensors, std::vector<lite::Tensor *> *out_tensors);
+                                std::vector<lite::Tensor *> *in_tensors, std::vector<lite::Tensor *> *out_tensors,
+                                TypeId prefer_data_type = kTypeUnknown);
 
   // find in_kernels_ and out_kernels of kernel, sub_graph and nodes_ in sub_graph
   static void FindAllInoutKernels(const std::vector<kernel::LiteKernel *> &kernels);

mindspore/lite/src/sub_graph_kernel.cc

@@ -366,7 +366,8 @@ int CpuFp16SubGraph::PostProcess() {
         }
       }
     }
-    if (real_tensor->data_type() == kNumberTypeFloat16 && origin_input_data_.at(real_tensor) != nullptr) {
+    if (real_tensor->data_type() == kNumberTypeFloat16 &&
+        origin_input_data_.find(real_tensor) != origin_input_data_.end()) {
       auto origin_tensor_data = origin_input_data_.at(real_tensor);
       real_tensor->FreeData();
       MS_ASSERT(origin_tensor_data->data_ != nullptr);

mindspore/lite/test/models_tf_fp16.cfg

@@ -0,0 +1 @@
+decoder_step_201217_modified.pb 5

mindspore/lite/test/run_benchmark_nets.sh

@@ -214,6 +214,21 @@ function Run_Converter() {
         fi
     done < ${models_tflite_fp16_config}
 
+    while read line; do
+        fp16_line_info=${line}
+        if [[ $fp16_line_info == \#* ]]; then
+          continue
+        fi
+        model_name=`echo ${fp16_line_info}|awk -F ' ' '{print $1}'`
+        echo 'cp '${ms_models_path}'/'${model_name}'.ms' ${ms_models_path}'/'${model_name}'.fp16.ms'
+        cp ${ms_models_path}/${model_name}.ms ${ms_models_path}/${model_name}.fp16.ms
+        if [ $? = 0 ]; then
+            converter_result='converter fp16 '${model_name}' pass';echo ${converter_result} >> ${run_converter_result_file}
+        else
+            converter_result='converter fp16 '${model_name}' failed';echo ${converter_result} >> ${run_converter_result_file};return 1
+        fi
+    done < ${models_tf_fp16_config}
+
     # Convert tflite weightquant models:
     while read line; do
         weight_quant_line_info=${line}
@@ -1832,6 +1847,34 @@ function Run_arm64_fp16() {
             run_result='arm64_fp16: '${model_name}' failed'; echo ${run_result} >> ${run_benchmark_result_file}; return 1
         fi
     done < ${models_multiple_inputs_fp16_config}
+
+    # Run tf fp16 models
+    while read line; do
+        model_name_and_input_num=${line%;*}
+        length=${#model_name_and_input_num}
+        input_shapes=${line:length+1}
+        tf_line_info=${model_name_and_input_num}
+        if [[ $model_name == \#* ]]; then
+          continue
+        fi
+        model_name=`echo ${tf_line_info}|awk -F ' ' '{print $1}'`
+        input_num=`echo ${tf_line_info}|awk -F ' ' '{print $2}'`
+        input_files=''
+        for i in $(seq 1 $input_num)
+        do
+          input_files=$input_files'/data/local/tmp/input_output/input/'$model_name'.ms_'$i'.bin,'
+        done
+        echo ${model_name} >> "${run_arm64_fp16_log_file}"
+        echo 'cd  /data/local/tmp/benchmark_test' > adb_run_cmd.txt
+        echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/data/local/tmp/benchmark_test;./benchmark --inputShapes='${input_shapes}' --modelFile='${model_name}'.ms --inDataFile='${input_files}' --benchmarkDataFile=/data/local/tmp/input_output/output/'${model_name}'.ms.out --enableFp16=true' >> "${run_arm64_fp16_log_file}"
+        echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/data/local/tmp/benchmark_test;./benchmark --inputShapes='${input_shapes}' --modelFile='${model_name}'.ms --inDataFile='${input_files}' --benchmarkDataFile=/data/local/tmp/input_output/output/'${model_name}'.ms.out --enableFp16=true' >> adb_run_cmd.txt
+        adb -s ${device_id} shell < adb_run_cmd.txt >> "${run_arm64_fp16_log_file}"
+        if [ $? = 0 ]; then
+            run_result='arm64_fp16: '${model_name}' pass'; echo ${run_result} >> ${run_benchmark_result_file}
+        else
+            run_result='arm64_fp16: '${model_name}' failed'; echo ${run_result} >> ${run_benchmark_result_file}; return 1
+        fi
+    done < ${models_tf_fp16_config}
 }
 # Run on gpu platform:
 function Run_gpu() {
@@ -2069,6 +2112,7 @@ models_npu_config=${basepath}/models_npu.cfg
 models_compatibility_config=${basepath}/models_compatibility.cfg
 models_with_multiple_inputs_config=${basepath}/models_with_multiple_inputs.cfg
 models_for_process_only_config=${basepath}/models_for_process_only.cfg
+models_tf_fp16_config=${basepath}/models_tf_fp16.cfg
 
 ms_models_path=${basepath}/ms_models