!3873 fix quantdtypecast

Merge pull request !3873 from zhongligeng/master

mindspore/lite/src/ops/ops.cc

@@ -137,10 +137,8 @@ Primitive *Primitive::CreatePrimitive(schema::Primitive *primitive) {
       return new lite::SpaceToDepth(const_cast<schema::Primitive *>(primitive));
     case schema::PrimitiveType_SpaceToBatch:
       return new lite::SpaceToBatch(const_cast<schema::Primitive *>(primitive));
-    case schema::PrimitiveType_OnnxInt8Dequantize:
-      return new lite::Dequantize(const_cast<schema::Primitive *>(primitive));
-    case schema::PrimitiveType_OnnxInt8Quantize:
-      return new lite::Quantize(const_cast<schema::Primitive *>(primitive));
+    case schema::PrimitiveType_QuantDTypeCast:
+      return new lite::QuantDTypeCast(const_cast<schema::Primitive *>(primitive));
     default:
       break;
   }

mindspore/lite/src/ops/ops.h

@@ -691,17 +691,10 @@ class SpaceToDepth : public Primitive {
   int InferShape(std::vector<tensor::Tensor *> inputs, std::vector<tensor::Tensor *> outputs) override;
 };
 
-class Dequantize : public Primitive {
+class QuantDTypeCast : public Primitive {
  public:
-  explicit Dequantize(schema::Primitive *primitive) : Primitive(primitive) {}
-  const schema::OnnxInt8Dequantize *GetAttribute() const { return this->primitive->value_as_OnnxInt8Dequantize(); }
-  int InferShape(std::vector<tensor::Tensor *> inputs, std::vector<tensor::Tensor *> outputs) override;
-};
-
-class Quantize : public Primitive {
- public:
-  explicit Quantize(schema::Primitive *primitive) : Primitive(primitive) {}
-  const schema::OnnxInt8Quantize *GetAttribute() const { return this->primitive->value_as_OnnxInt8Quantize(); }
+  explicit QuantDTypeCast(schema::Primitive *primitive) : Primitive(primitive) {}
+  const schema::QuantDTypeCast *GetAttribute() const { return this->primitive->value_as_QuantDTypeCast(); }
   int InferShape(std::vector<tensor::Tensor *> inputs, std::vector<tensor::Tensor *> outputs) override;
 };
 }  // namespace lite

mindspore/lite/src/ops/dequantize.cc → mindspore/lite/src/ops/quant_dtype_cast.cc

@@ -20,15 +20,16 @@
 #include "src/ir/tensor.h"
 
 namespace mindspore::lite {
-int Dequantize::InferShape(std::vector<tensor::Tensor *> inputs_, std::vector<tensor::Tensor *> outputs_) {
+int QuantDTypeCast::InferShape(std::vector<tensor::Tensor *> inputs_, std::vector<tensor::Tensor *> outputs_) {
   MS_ASSERT(this->primitive != nullptr);
   auto input = inputs_.front();
   MS_ASSERT(input != nullptr);
   auto output = outputs_.front();
   MS_ASSERT(output != nullptr);
   output->set_shape(input->shape());
-  output->set_data_type(kNumberTypeFloat32);
+  auto param = primitive->value_as_QuantDTypeCast();
+  MS_ASSERT(input->data_type() == param->srcT);
+  output->set_data_type(static_cast<TypeId>(param->dstT()));
   return RET_OK;
 }
 }  // namespace mindspore::lite
-

mindspore/lite/src/ops/quantize.cc

@@ -1,34 +0,0 @@
-/**
- * Copyright 2019-2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "src/ops/ops.h"
-#include "include/errorcode.h"
-#include "utils/log_adapter.h"
-#include "src/ir/tensor.h"
-
-namespace mindspore::lite {
-int Quantize::InferShape(std::vector<tensor::Tensor *> inputs_, std::vector<tensor::Tensor *> outputs_) {
-  MS_ASSERT(this->primitive != nullptr);
-  auto input = inputs_.front();
-  MS_ASSERT(input != nullptr);
-  auto output = outputs_.front();
-  MS_ASSERT(output != nullptr);
-  output->set_shape(input->shape());
-  output->set_data_type(kNumberTypeInt8);
-  return RET_OK;
-}
-}  // namespace mindspore::lite
-

mindspore/lite/src/populate_parameter.cc

@@ -65,8 +65,7 @@
 #include "src/runtime/kernel/arm/base/prior_box.h"
 #include "src/runtime/kernel/arm/opclib/fp32/space_to_depth.h"
 #include "src/runtime/kernel/arm/opclib/fp32/space_to_batch.h"
-#include "src/runtime/kernel/arm/opclib/int8/dequantize.h"
-#include "src/runtime/kernel/arm/opclib/fp32/quantize.h"
+#include "src/runtime/kernel/arm/opclib/int8/quant_dtype_cast.h"
 
 namespace mindspore::kernel {
 OpParameter *PopulateFillParameter(const lite::Primitive *primitive) {
@@ -1030,24 +1029,17 @@ OpParameter *PopulateFlattenParameter(const lite::Primitive *primitive) {
   return reinterpret_cast<OpParameter *>(flatten_param);
 }
 
-OpParameter *PopulateDequantizeParameter(const lite::Primitive *primitive) {
-  DequantizeParameter *dequantize_parameter = new (std::nothrow) DequantizeParameter();
-  if (dequantize_parameter == nullptr) {
-    MS_LOG(ERROR) << "new DequantizeParameter fail!";
-    return nullptr;
-  }
-  dequantize_parameter->op_parameter_.type_ = primitive->Type();
-  return reinterpret_cast<OpParameter *>(dequantize_parameter);
-}
-
-OpParameter *PopulateQuantizeParameter(const lite::Primitive *primitive) {
-  QuantizeParameter *quantize_parameter = new (std::nothrow) QuantizeParameter();
-  if (quantize_parameter == nullptr) {
-    MS_LOG(ERROR) << "new QuantizeParameter fail!";
+OpParameter *PopulateQuantDTypeCastParameter(const lite::Primitive *primitive) {
+  QuantDTypeCastParameter *parameter = new (std::nothrow) QuantDTypeCastParameter();
+  if (parameter == nullptr) {
+    MS_LOG(ERROR) << "new QuantDTypeCastParameter fail!";
     return nullptr;
   }
-  quantize_parameter->op_parameter_.type_ = primitive->Type();
-  return reinterpret_cast<OpParameter *>(quantize_parameter);
+  parameter->op_parameter_.type_ = primitive->Type();
+  auto quant_dtype_cast_param = primitive->Value()->value_as_QuantDTypeCast();
+  parameter->srcT = quant_dtype_cast_param->srcT();
+  parameter->dstT = quant_dtype_cast_param->dstT();
+  return reinterpret_cast<OpParameter *>(parameter);
 }
 
 OpParameter *PopulateStridedSliceParameter(const lite::Primitive *primitive) {
@@ -1207,8 +1199,7 @@ PopulateParameterRegistry::PopulateParameterRegistry() {
   populate_parameter_funcs_[schema::PrimitiveType_Square] = PopulateSqueezeParameter;
   populate_parameter_funcs_[schema::PrimitiveType_Split] = PopulateSplitParameter;
   populate_parameter_funcs_[schema::PrimitiveType_PriorBox] = PopulatePriorBoxParameter;
-  populate_parameter_funcs_[schema::PrimitiveType_OnnxInt8Dequantize] = PopulateDequantizeParameter;
-  populate_parameter_funcs_[schema::PrimitiveType_OnnxInt8Quantize] = PopulateQuantizeParameter;
+  populate_parameter_funcs_[schema::PrimitiveType_QuantDTypeCast] = PopulateQuantDTypeCastParameter;
 }
 
 PopulateParameterRegistry *PopulateParameterRegistry::GetInstance() {

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

@@ -0,0 +1,148 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include "src/runtime/kernel/arm/base/quant_dtype_cast.h"
+#include <vector>
+#include "src/runtime/kernel/arm/opclib/int8/quant_dtype_cast.h"
+#include "src/runtime/runtime_api.h"
+#include "src/kernel_registry.h"
+#include "schema/model_generated.h"
+#include "include/errorcode.h"
+
+using mindspore::kernel::KERNEL_ARCH::kCPU;
+using mindspore::lite::KernelRegistrar;
+using mindspore::lite::RET_ERROR;
+using mindspore::lite::RET_OK;
+using mindspore::schema::PrimitiveType_QuantDTypeCast;
+
+namespace mindspore::kernel {
+namespace {
+constexpr int kQuantDTypeCastInputNum = 1;
+constexpr int kQuantDTypeCastOutputNum = 1;
+}  // namespace
+
+int QuantDTypeCastCPUKernel::Init() {
+  if (inputs_.size() != 1) {
+    MS_LOG(ERROR) << "inputs number should be 1, but " << inputs_.size() << " is given.";
+    return RET_ERROR;
+  }
+  if (outputs_.size() != 1) {
+    MS_LOG(ERROR) << "outputs number should be 1, but " << inputs_.size() << " is given.";
+    return RET_ERROR;
+  }
+  auto in_tensor = inputs_.front();
+  auto out_tensor = outputs_.front();
+  auto param = reinterpret_cast<QuantDTypeCastParameter *>(opParameter);
+  if (param->srcT == kNumberTypeFloat32 && param->dstT == kNumberTypeInt8) {
+    if (in_tensor->data_type() != kNumberTypeFloat32 || out_tensor->data_type() != kNumberTypeInt8) {
+      MS_LOG(ERROR) << "param data type and tensor data type do not match.";
+      return RET_ERROR;
+    }
+    inverse_ = false;
+  } else if (param->srcT == kNumberTypeInt8 && param->dstT == kNumberTypeFloat32) {
+    if (in_tensor->data_type() != kNumberTypeInt8 || out_tensor->data_type() != kNumberTypeFloat32) {
+      MS_LOG(ERROR) << "param data type and tensor data type do not match.";
+      return RET_ERROR;
+    }
+    inverse_ = true;
+  } else {
+    MS_LOG(ERROR) << "param data type not supported.";
+    return RET_ERROR;
+  }
+
+  num_unit_ = static_cast<int>(in_tensor->DataSize());
+  thread_n_num_ = MSMIN(thread_num_, num_unit_);
+  thread_n_stride_ = UP_DIV(num_unit_, thread_n_num_);
+
+  return RET_OK;
+}
+
+int QuantDTypeCastCPUKernel::ReSize() { return RET_OK; }
+
+int QuantDTypeCastCPUKernel::QuantDTypeCast(int task_id) {
+  int num_unit_thread = MSMIN(thread_n_stride_, num_unit_ - task_id * thread_n_stride_);
+  if (num_unit_thread <= 0) {
+    return RET_OK;
+  }
+  int thread_offset = task_id * thread_n_stride_;
+  auto quant_arg = inputs_.front()->GetQuantParams().front();
+  int ret;
+  if (inverse_) {
+    ret = DequantizeInt8(int8_ptr_ + thread_offset, float32_ptr_ + thread_offset, quant_arg.scale, quant_arg.zeroPoint,
+                         num_unit_thread);
+  } else {
+    ret = QuantizeToInt8(float32_ptr_ + thread_offset, int8_ptr_ + thread_offset, quant_arg.scale,
+                         quant_arg.zeroPoint, num_unit_thread);
+  }
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "QuantDTypeCast error task_id[" << task_id << "] error_code[" << ret << "]";
+    return RET_ERROR;
+  }
+  return RET_OK;
+}
+
+int QuantDTypeCastRun(int task_id, LiteParallelGroupEnv *penv, void *cdata) {
+  auto g_kernel = reinterpret_cast<QuantDTypeCastCPUKernel *>(cdata);
+  auto ret = g_kernel->QuantDTypeCast(task_id);
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "QuantDTypeCastRun error task_id[" << task_id << "] error_code[" << ret << "]";
+    return RET_ERROR;
+  }
+  return RET_OK;
+}
+
+int QuantDTypeCastCPUKernel::Run() {
+  if (inverse_) {
+    int8_ptr_ = reinterpret_cast<int8_t *>(inputs_[0]->Data());
+    float32_ptr_ = reinterpret_cast<float *>(outputs_[0]->Data());
+  } else {
+    float32_ptr_ = reinterpret_cast<float *>(inputs_[0]->Data());
+    int8_ptr_ = reinterpret_cast<int8_t *>(outputs_[0]->Data());
+  }
+
+  int ret = LiteBackendParallelLaunch(QuantDTypeCastRun, this, thread_n_num_);
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "Scale error error_code[" << ret << "]";
+    return RET_ERROR;
+  }
+
+  return RET_OK;
+}
+
+kernel::LiteKernel *CpuQuantDTypeCastFp32KernelCreator(const std::vector<lite::tensor::Tensor *> &inputs,
+                                                       const std::vector<lite::tensor::Tensor *> &outputs,
+                                                       OpParameter *opParameter, const lite::Context *ctx,
+                                                       const kernel::KernelKey &desc) {
+  if (opParameter == nullptr) {
+    MS_LOG(ERROR) << "Input opParameter is nullptr!";
+    return nullptr;
+  }
+  auto *kernel = new (std::nothrow) QuantDTypeCastCPUKernel(opParameter, inputs, outputs, ctx);
+  if (kernel == nullptr) {
+    MS_LOG(ERROR) << "new QuantDTypeCastCPUKernel fail!";
+    return nullptr;
+  }
+  auto ret = kernel->Init();
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "Init kernel failed! name: " << opParameter->name_ << ", type: "
+                  << schema::EnumNamePrimitiveType(static_cast<schema::PrimitiveType>(opParameter->type_));
+    delete kernel;
+    return nullptr;
+  }
+  return kernel;
+}
+
+REG_KERNEL(kCPU, kNumberTypeInt8, PrimitiveType_QuantDTypeCast, CpuQuantDTypeCastFp32KernelCreator)
+}  // namespace mindspore::kernel

mindspore/lite/src/runtime/kernel/arm/int8/dequantize.h → mindspore/lite/src/runtime/kernel/arm/base/quant_dtype_cast.h

@@ -14,33 +14,34 @@
  * limitations under the License.
  */
 
-#ifndef MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_INT8_DEQUANTIZE_H_
-#define MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_INT8_DEQUANTIZE_H_
+#ifndef MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_BASE_QUANTDTYPECAST_H_
+#define MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_BASE_QUANTDTYPECAST_H_
 
 #include <vector>
 #include "src/lite_kernel.h"
 
 namespace mindspore::kernel {
-class DequantizeCPUKernel : public LiteKernel {
+class QuantDTypeCastCPUKernel : public LiteKernel {
  public:
-  DequantizeCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
+  QuantDTypeCastCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
                       const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx)
       : LiteKernel(parameter, inputs, outputs), thread_num_(ctx->threadNum) {}
-  ~DequantizeCPUKernel() = default;
+  ~QuantDTypeCastCPUKernel() = default;
 
   int Init() override;
   int ReSize() override;
   int Run() override;
-  int Dequantize(int task_id);
+  int QuantDTypeCast(int task_id);
 
  private:
   int thread_num_;
   int thread_n_num_;
   int thread_n_stride_;
   int num_unit_;
-  int8_t *input_ptr_;
-  float *output_ptr_;
+  int8_t *int8_ptr_;
+  float *float32_ptr_;
+  bool inverse_;
 };
 }  // namespace mindspore::kernel
 
-#endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_INT8_DEQUANTIZE_H_
+#endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_BASE_QUANTDTYPECAST_H_

mindspore/lite/src/runtime/kernel/arm/fp32/quantize.cc

@@ -1,118 +0,0 @@
-/**
- * Copyright 2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "src/runtime/kernel/arm/fp32/quantize.h"
-#include <vector>
-#include "src/runtime/kernel/arm/opclib/fp32/quantize.h"
-#include "src/kernel_registry.h"
-#include "src/runtime/runtime_api.h"
-#include "schema/model_generated.h"
-#include "include/errorcode.h"
-
-using mindspore::kernel::KERNEL_ARCH::kCPU;
-using mindspore::lite::KernelRegistrar;
-using mindspore::lite::RET_ERROR;
-using mindspore::lite::RET_OK;
-using mindspore::schema::PrimitiveType_OnnxInt8Quantize;
-
-namespace mindspore::kernel {
-namespace {
-constexpr int kQuantizeInputNum = 1;
-constexpr int kQuantizeOutputNum = 1;
-}  // namespace
-
-int QuantizeCPUKernel::Init() {
-  if (inputs_.size() != 1) {
-    MS_LOG(ERROR) << "inputs number should be 1, but " << inputs_.size() << " is given.";
-    return RET_ERROR;
-  }
-  if (outputs_.size() != 1) {
-    MS_LOG(ERROR) << "outputs number should be 1, but " << inputs_.size() << " is given.";
-    return RET_ERROR;
-  }
-  auto in_tensor = inputs_.front();
-  num_unit_ = static_cast<int>(in_tensor->DataSize());
-  thread_n_num_ = MSMIN(thread_num_, num_unit_);
-  thread_n_stride_ = UP_DIV(num_unit_, thread_n_num_);
-
-  return RET_OK;
-}
-
-int QuantizeCPUKernel::ReSize() { return RET_OK; }
-
-int QuantizeCPUKernel::Quantize(int task_id) {
-  int num_unit_thread = MSMIN(thread_n_stride_, num_unit_ - task_id * thread_n_stride_);
-  if (num_unit_thread <= 0) {
-    return RET_OK;
-  }
-  int thread_offset = task_id * thread_n_stride_;
-  auto quant_arg = inputs_.front()->GetQuantParams().front();
-  int ret = QuantizeToInt8(input_ptr_ + thread_offset, output_ptr_ + thread_offset, quant_arg.scale,
-                           quant_arg.zeroPoint, num_unit_thread);
-
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "Quantize error task_id[" << task_id << "] error_code[" << ret << "]";
-    return RET_ERROR;
-  }
-  return RET_OK;
-}
-
-int QuantizeRun(int task_id, LiteParallelGroupEnv *penv, void *cdata) {
-  auto g_kernel = reinterpret_cast<QuantizeCPUKernel *>(cdata);
-  auto ret = g_kernel->Quantize(task_id);
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "QuantizeRun error task_id[" << task_id << "] error_code[" << ret << "]";
-    return RET_ERROR;
-  }
-  return RET_OK;
-}
-
-int QuantizeCPUKernel::Run() {
-  input_ptr_ = reinterpret_cast<float *>(inputs_[0]->Data());
-  output_ptr_ = reinterpret_cast<int8_t *>(outputs_[0]->Data());
-  int ret = LiteBackendParallelLaunch(QuantizeRun, this, thread_n_num_);
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "Scale error error_code[" << ret << "]";
-    return RET_ERROR;
-  }
-
-  return RET_OK;
-}
-
-kernel::LiteKernel *CpuQuantizeFp32KernelCreator(const std::vector<lite::tensor::Tensor *> &inputs,
-                                                 const std::vector<lite::tensor::Tensor *> &outputs,
-                                                 OpParameter *opParameter, const lite::Context *ctx,
-                                                 const kernel::KernelKey &desc) {
-  if (opParameter == nullptr) {
-    MS_LOG(ERROR) << "Input opParameter is nullptr!";
-    return nullptr;
-  }
-  auto *kernel = new (std::nothrow) QuantizeCPUKernel(opParameter, inputs, outputs, ctx);
-  if (kernel == nullptr) {
-    MS_LOG(ERROR) << "new QuantizeCPUKernel fail!";
-    return nullptr;
-  }
-  auto ret = kernel->Init();
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "Init kernel failed! name: " << opParameter->name_ << ", type: "
-                  << schema::EnumNamePrimitiveType(static_cast<schema::PrimitiveType>(opParameter->type_));
-    delete kernel;
-    return nullptr;
-  }
-  return kernel;
-}
-
-REG_KERNEL(kCPU, kNumberTypeFloat32, PrimitiveType_OnnxInt8Quantize, CpuQuantizeFp32KernelCreator)
-}  // namespace mindspore::kernel

mindspore/lite/src/runtime/kernel/arm/fp32/quantize.h

@@ -1,46 +0,0 @@
-/**
- * Copyright 2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#ifndef MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_FP32_QUANTIZE_H_
-#define MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_FP32_QUANTIZE_H_
-
-#include <vector>
-#include "src/lite_kernel.h"
-
-namespace mindspore::kernel {
-class QuantizeCPUKernel : public LiteKernel {
- public:
-  QuantizeCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
-                    const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx)
-      : LiteKernel(parameter, inputs, outputs), thread_num_(ctx->threadNum) {}
-  ~QuantizeCPUKernel() = default;
-
-  int Init() override;
-  int ReSize() override;
-  int Run() override;
-  int Quantize(int task_id);
-
- private:
-  int thread_num_;
-  int thread_n_num_;
-  int thread_n_stride_;
-  int num_unit_;
-  float *input_ptr_;
-  int8_t *output_ptr_;
-};
-}  // namespace mindspore::kernel
-
-#endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_FP32_QUANTIZE_H_

mindspore/lite/src/runtime/kernel/arm/fp32/space_to_depth.cc

@@ -16,11 +16,11 @@
 
 #include "src/runtime/kernel/arm/fp32/space_to_depth.h"
 #include <vector>
-#include "schema/ops_generated.h"
 #include "schema/model_generated.h"
 #include "src/kernel_registry.h"
 #include "src/runtime/kernel/arm/opclib/fp32/space_to_depth.h"
 #include "include/errorcode.h"
+#include "src/runtime/runtime_api.h"
 
 using mindspore::lite::KernelRegistrar;
 using mindspore::lite::RET_ERROR;
@@ -41,21 +41,48 @@ int SpaceToDepthCPUKernel::Init() {
     MS_LOG(ERROR) << "Input block_size should > 0!";
     return RET_PARAM_INVALID;
   }
+
+  num_unit_ = static_cast<int>(inputs_[0]->shape().at(kNHWC_H));
+  thread_h_num_ = MSMIN(thread_num_, num_unit_);
+  thread_h_stride_ = UP_DIV(num_unit_, thread_h_num_);
   return RET_OK;
 }
 
-int SpaceToDepthCPUKernel::Run() {
-  auto input = inputs_[0];
-  auto output = outputs_[0];
-  const float *input_data = static_cast<const float *>(input->Data());
-  float *output_data = static_cast<float *>(output->Data());
-  auto in_shape = input->shape();
-  auto out_shape = output->shape();
+int SpaceToDepthCPUKernel::SpaceToDepth(int task_id) {
+  int num_unit_thread = MSMIN(thread_h_stride_, num_unit_ - task_id * thread_h_stride_);
+  if (num_unit_thread <= 0) {
+    return RET_OK;
+  }
+  int thread_offset = task_id * thread_h_stride_;
+  auto in_shape = inputs_[0]->shape();
+  auto out_shape = outputs_[0]->shape();
   SpaceToDepthParameter *param = reinterpret_cast<SpaceToDepthParameter *>(opParameter);
-  if (input->GetFormat() == schema::Format_NHWC) {
-    auto ret = SpaceToDepthForNHWC(input_data, output_data, in_shape.data(), out_shape.data(), in_shape.size(),
-                                   param->block_size_);
-    return ret;
+  auto ret = SpaceToDepthForNHWC(input_ptr_, output_ptr_, in_shape.data(), out_shape.data(), in_shape.size(),
+                                 param->block_size_, thread_offset, thread_offset + num_unit_thread);
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "SpaceToDepth error task_id[" << task_id << "] error_code[" << ret << "]";
+    return RET_ERROR;
+  }
+  return RET_OK;
+}
+
+int SpaceToDepthRun(int task_id, LiteParallelGroupEnv *penv, void *cdata) {
+  auto g_kernel = reinterpret_cast<SpaceToDepthCPUKernel *>(cdata);
+  auto ret = g_kernel->SpaceToDepth(task_id);
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "SpaceToDepthRun error task_id[" << task_id << "] error_code[" << ret << "]";
+    return RET_ERROR;
+  }
+  return RET_OK;
+}
+
+int SpaceToDepthCPUKernel::Run() {
+  if (inputs_[0]->GetFormat() == schema::Format_NHWC) {
+    int ret = LiteBackendParallelLaunch(SpaceToDepthRun, this, thread_h_num_);
+    if (ret != RET_OK) {
+      MS_LOG(ERROR) << "SpaceToDepth error error_code[" << ret << "]";
+      return ret;
+    }
   } else {
     MS_LOG(ERROR) << "Only support NHWC now!";
     return RET_ERROR;
@@ -69,7 +96,7 @@ kernel::LiteKernel *CpuSpaceToDepthFp32KernelCreator(const std::vector<lite::ten
     MS_LOG(ERROR) << "Input opParameter is nullptr!";
     return nullptr;
   }
-  auto *kernel = new (std::nothrow) SpaceToDepthCPUKernel(opParameter, inputs, outputs);
+  auto *kernel = new (std::nothrow) SpaceToDepthCPUKernel(opParameter, inputs, outputs, ctx);
   if (kernel == nullptr) {
     MS_LOG(ERROR) << "new SpaceToDepthCPUKernel fail!";
     return nullptr;

mindspore/lite/src/runtime/kernel/arm/fp32/space_to_depth.h

@@ -24,13 +24,22 @@ namespace mindspore::kernel {
 class SpaceToDepthCPUKernel : public LiteKernel {
  public:
   SpaceToDepthCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
-                        const std::vector<lite::tensor::Tensor *> &outputs)
-      : LiteKernel(parameter, inputs, outputs) {}
+                        const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx)
+      : LiteKernel(parameter, inputs, outputs), thread_num_(ctx->threadNum) {}
   ~SpaceToDepthCPUKernel() = default;
 
+  int SpaceToDepth(int task_id);
   int Init() override;
   int ReSize() override { return 0; };
   int Run() override;
+
+ private:
+  int thread_num_;
+  int thread_h_stride_;
+  int thread_h_num_;
+  int num_unit_;
+  float *input_ptr_;
+  float *output_ptr_;
 };
 }  // namespace mindspore::kernel
 

mindspore/lite/src/runtime/kernel/arm/int8/dequantize.cc

@@ -1,118 +0,0 @@
-/**
- * Copyright 2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "src/runtime/kernel/arm/int8/dequantize.h"
-#include <vector>
-#include "src/runtime/kernel/arm/opclib/int8/dequantize.h"
-#include "src/runtime/runtime_api.h"
-#include "src/kernel_registry.h"
-#include "schema/model_generated.h"
-#include "include/errorcode.h"
-
-using mindspore::kernel::KERNEL_ARCH::kCPU;
-using mindspore::lite::KernelRegistrar;
-using mindspore::lite::RET_ERROR;
-using mindspore::lite::RET_OK;
-using mindspore::schema::PrimitiveType_OnnxInt8Dequantize;
-
-namespace mindspore::kernel {
-namespace {
-constexpr int kDequantizeInputNum = 1;
-constexpr int kDequantizeOutputNum = 1;
-}  // namespace
-
-int DequantizeCPUKernel::Init() {
-  if (inputs_.size() != 1) {
-    MS_LOG(ERROR) << "inputs number should be 1, but " << inputs_.size() << " is given.";
-    return RET_ERROR;
-  }
-  if (outputs_.size() != 1) {
-    MS_LOG(ERROR) << "outputs number should be 1, but " << inputs_.size() << " is given.";
-    return RET_ERROR;
-  }
-  auto in_tensor = inputs_.front();
-  num_unit_ = static_cast<int>(in_tensor->DataSize());
-  thread_n_num_ = MSMIN(thread_num_, num_unit_);
-  thread_n_stride_ = UP_DIV(num_unit_, thread_n_num_);
-
-  return RET_OK;
-}
-
-int DequantizeCPUKernel::ReSize() { return RET_OK; }
-
-int DequantizeCPUKernel::Dequantize(int task_id) {
-  int num_unit_thread = MSMIN(thread_n_stride_, num_unit_ - task_id * thread_n_stride_);
-  if (num_unit_thread <= 0) {
-    return RET_OK;
-  }
-  int thread_offset = task_id * thread_n_stride_;
-  auto quant_arg = inputs_.front()->GetQuantParams().front();
-  int ret = DequantizeInt8(input_ptr_ + thread_offset, output_ptr_ + thread_offset, quant_arg.scale,
-                           quant_arg.zeroPoint, num_unit_thread);
-
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "Dequantize error task_id[" << task_id << "] error_code[" << ret << "]";
-    return RET_ERROR;
-  }
-  return RET_OK;
-}
-
-int DequantizeRun(int task_id, LiteParallelGroupEnv *penv, void *cdata) {
-  auto g_kernel = reinterpret_cast<DequantizeCPUKernel *>(cdata);
-  auto ret = g_kernel->Dequantize(task_id);
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "DequantizeRun error task_id[" << task_id << "] error_code[" << ret << "]";
-    return RET_ERROR;
-  }
-  return RET_OK;
-}
-
-int DequantizeCPUKernel::Run() {
-  input_ptr_ = reinterpret_cast<int8_t *>(inputs_[0]->Data());
-  output_ptr_ = reinterpret_cast<float *>(outputs_[0]->Data());
-  int ret = LiteBackendParallelLaunch(DequantizeRun, this, thread_n_num_);
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "Scale error error_code[" << ret << "]";
-    return RET_ERROR;
-  }
-
-  return RET_OK;
-}
-
-kernel::LiteKernel *CpuDequantizeFp32KernelCreator(const std::vector<lite::tensor::Tensor *> &inputs,
-                                                   const std::vector<lite::tensor::Tensor *> &outputs,
-                                                   OpParameter *opParameter, const lite::Context *ctx,
-                                                   const kernel::KernelKey &desc) {
-  if (opParameter == nullptr) {
-    MS_LOG(ERROR) << "Input opParameter is nullptr!";
-    return nullptr;
-  }
-  auto *kernel = new (std::nothrow) DequantizeCPUKernel(opParameter, inputs, outputs, ctx);
-  if (kernel == nullptr) {
-    MS_LOG(ERROR) << "new DequantizeCPUKernel fail!";
-    return nullptr;
-  }
-  auto ret = kernel->Init();
-  if (ret != RET_OK) {
-    MS_LOG(ERROR) << "Init kernel failed! name: " << opParameter->name_ << ", type: "
-                  << schema::EnumNamePrimitiveType(static_cast<schema::PrimitiveType>(opParameter->type_));
-    delete kernel;
-    return nullptr;
-  }
-  return kernel;
-}
-
-REG_KERNEL(kCPU, kNumberTypeInt8, PrimitiveType_OnnxInt8Dequantize, CpuDequantizeFp32KernelCreator)
-}  // namespace mindspore::kernel

mindspore/lite/src/runtime/kernel/arm/opclib/fp32/quantize.h

@@ -1,28 +0,0 @@
-/**
- * Copyright 2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#ifndef MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_FP32_QUANTIZE_H_
-#define MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_FP32_QUANTIZE_H_
-
-#include "src/runtime/kernel/arm/opclib/op_base.h"
-
-struct QuantizeParameter {
-  OpParameter op_parameter_;
-};
-
-int QuantizeToInt8(float *real_values, int8_t *quant_values, float scale, int32_t zp, int size);
-
-#endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_FP32_QUANTIZE_H_

mindspore/lite/src/runtime/kernel/arm/opclib/fp32/space_to_depth.cc

@@ -19,13 +19,16 @@
 #include "src/runtime/kernel/arm/opclib/op_base.h"
 
 int SpaceToDepthForNHWC(const float *input, float *output, int *in_shape, int *out_shape, int shape_size,
-                        int block_size) {
+                        int block_size, int h_start, int h_end) {
   if (input == nullptr || output == nullptr) {
     return OPCLIB_NULL_PTR;
   }
   if (shape_size != C4NUM) {
     return OPCLIB_PARAM_INVALID;
   }
+  if (h_start < 0 || h_start >= h_end || h_end > out_shape[1]) {
+    return OPCLIB_PARAM_INVALID;
+  }
   int in_strides[C4NUM];
   ComputeStrides(in_shape, in_strides, shape_size);
   int out_strides[C4NUM];
@@ -33,7 +36,7 @@ int SpaceToDepthForNHWC(const float *input, float *output, int *in_shape, int *o
   for (int i = 0; i < out_shape[0]; ++i) {
     size_t in_offset_n = i * in_strides[0];
     size_t out_offset_n = i * out_strides[0];
-    for (int j = 0; j < out_shape[1]; ++j) {
+    for (int j = h_start; j < h_end; ++j) {
       size_t in_offset_h = in_offset_n + j * block_size * in_strides[1];
       size_t out_offset_h = out_offset_n + j * out_strides[1];
       for (int k = 0; k < out_shape[2]; ++k) {

mindspore/lite/src/runtime/kernel/arm/opclib/fp32/space_to_depth.h

@@ -23,5 +23,5 @@ struct SpaceToDepthParameter {
 };
 
 int SpaceToDepthForNHWC(const float *input, float *output, int *in_shape, int *out_shape, int shape_size,
-                        int block_size);
+                        int block_size, int h_start, int h_end);
 #endif  // MINDSPORE_LITE_SRC_BACKEND_ARM_OPCLIB_FP32_SPACE_TO_DEPTH_H_

mindspore/lite/src/runtime/kernel/arm/opclib/int8/dequantize.cc

@@ -1,29 +0,0 @@
-/**
- * Copyright 2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "src/runtime/kernel/arm/opclib/int8/dequantize.h"
-#include "src/runtime/kernel/arm/opclib/errorcode.h"
-
-int DequantizeInt8(int8_t *quant_values, float *real_values, float scale, int32_t zp, int size) {
-  if (quant_values == nullptr || real_values == nullptr) {
-    return OPCLIB_PARAM_INVALID;
-  }
-
-  for (int i = 0; i < size; ++i) {
-    real_values[i] = (quant_values[i] + zp) * scale;
-  }
-  return OPCLIB_OK;
-}

mindspore/lite/src/runtime/kernel/arm/opclib/fp32/quantize.cc → mindspore/lite/src/runtime/kernel/arm/opclib/int8/quant_dtype_cast.cc

@@ -14,9 +14,20 @@
  * limitations under the License.
  */
 
-#include "src/runtime/kernel/arm/opclib/fp32/quantize.h"
+#include "src/runtime/kernel/arm/opclib/int8/quant_dtype_cast.h"
 #include "src/runtime/kernel/arm/opclib/errorcode.h"
 
+int DequantizeInt8(int8_t *quant_values, float *real_values, float scale, int32_t zp, int size) {
+  if (quant_values == nullptr || real_values == nullptr) {
+    return OPCLIB_PARAM_INVALID;
+  }
+
+  for (int i = 0; i < size; ++i) {
+    real_values[i] = (quant_values[i] + zp) * scale;
+  }
+  return OPCLIB_OK;
+}
+
 int QuantizeToInt8(float *real_values, int8_t *quant_values, float scale, int32_t zp, int size) {
   if (quant_values == nullptr || real_values == nullptr) {
     return OPCLIB_PARAM_INVALID;

mindspore/lite/src/runtime/kernel/arm/opclib/int8/dequantize.h → mindspore/lite/src/runtime/kernel/arm/opclib/int8/quant_dtype_cast.h

@@ -14,15 +14,18 @@
  * limitations under the License.
  */
 
-#ifndef MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_INT8_DEQUANTIZE_H_
-#define MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_INT8_DEQUANTIZE_H_
+#ifndef MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_INT8_QUANTDTYPECAST_H_
+#define MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_INT8_QUANTDTYPECAST_H_
 
 #include "src/runtime/kernel/arm/opclib/op_base.h"
 
-struct DequantizeParameter {
+struct QuantDTypeCastParameter {
   OpParameter op_parameter_;
+  int32_t srcT;
+  int32_t dstT;
 };
 
 int DequantizeInt8(int8_t *quant_values, float *real_values, float scale, int32_t zp, int size);
+int QuantizeToInt8(float *real_values, int8_t *quant_values, float scale, int32_t zp, int size);
 
-#endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_INT8_DEQUANTIZE_H_
+#endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_OPCLIB_INT8_QUANTDTYPECAST_H_

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32/quantize_fp32_tests.cc

@@ -1,77 +0,0 @@
-/**
- * Copyright 2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include <iostream>
-#include <memory>
-#include "utils/log_adapter.h"
-#include "common/common_test.h"
-#include "mindspore/lite/src/runtime/kernel/arm/fp32/quantize.h"
-#include "mindspore/lite/src/runtime/kernel/arm/opclib/fp32/quantize.h"
-#include "mindspore/lite/src/kernel_registry.h"
-#include "mindspore/lite/src/lite_kernel.h"
-
-namespace mindspore {
-
-class QuantizeTestFp32 : public mindspore::Common {
- public:
-  QuantizeTestFp32() {}
-};
-
-TEST_F(QuantizeTestFp32, QuantizeTest1) {
-  const lite::tensor::QuantArg quant_arg = {0.3515625, -57};
-  QuantizeParameter param;
-  param.op_parameter_.type_ = schema::PrimitiveType_OnnxInt8Quantize;
-
-  std::vector<float> input = {1, 2, 5, 6, 10, -20, 3, 8, 18, 10, 3, 4, 11, 16, 15, 25};
-  std::vector<int> in_shape = {1, 4, 4, 1};
-  lite::tensor::Tensor input_tensor;
-  input_tensor.SetData(input.data());
-  input_tensor.set_shape(in_shape);
-  input_tensor.SetFormat(schema::Format_NHWC);
-  input_tensor.set_data_type(kNumberTypeFloat32);
-  input_tensor.AddQuantParam(quant_arg);
-  std::vector<lite::tensor::Tensor *> inputs_tensor;
-  inputs_tensor.emplace_back(&input_tensor);
-
-  const int out_size = 16;
-  int8_t expect_out[16] = {-54, -51, -43, -40, -29, -114, -48, -34, -6, -29, -48, -46, -26, -11, -14, 14};
-  std::vector<int8_t> output(16);
-  std::vector<int> out_shape = {1, 4, 4, 1};
-  lite::tensor::Tensor output_tensor;
-  output_tensor.SetData(output.data());
-  output_tensor.set_shape(out_shape);
-  output_tensor.SetFormat(schema::Format_NHWC);
-  output_tensor.set_data_type(kNumberTypeInt8);
-  std::vector<lite::tensor::Tensor *> outputs_tensor;
-  outputs_tensor.emplace_back(&output_tensor);
-
-  lite::Context ctx;
-  ctx.threadNum = 3;
-  kernel::KernelKey desc = {kernel::KERNEL_ARCH::kCPU, kNumberTypeFloat32, schema::PrimitiveType_OnnxInt8Quantize};
-  auto creator = lite::KernelRegistry::GetInstance()->GetCreator(desc);
-  ASSERT_NE(creator, nullptr);
-  kernel::LiteKernel *kernel =
-    creator(inputs_tensor, outputs_tensor, reinterpret_cast<OpParameter *>(&param), &ctx, desc);
-  ASSERT_NE(kernel, nullptr);
-  kernel->Run();
-
-  for (int i = 0; i < out_size; ++i) {
-    std::cout << output[i] << " ";
-  }
-  std::cout << "\n";
-  CompareOutputData(output.data(), expect_out, out_size, 0.000001);
-}
-
-}  // namespace mindspore

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32/space_to_depth_fp32_tests.cc

@@ -37,7 +37,9 @@ TEST_F(SpaceToDepthTestFp32, SpaceToDepthTest1) {
   float output[16];
   int in_shape[4] = {1, 4, 4, 1};
   int out_shape[4] = {1, 2, 2, 4};
-  SpaceToDepthForNHWC((const float *)input, output, in_shape, out_shape, 4, 2);
+  int h_start = 0;
+  int h_end = 2;
+  SpaceToDepthForNHWC((const float *)input, output, in_shape, out_shape, 4, 2, h_start, h_end);
   for (int i = 0; i < out_size; ++i) {
     std::cout << output[i] << " ";
   }
@@ -69,10 +71,11 @@ TEST_F(SpaceToDepthTestFp32, SpaceToDepthTest2) {
   outputs_tensor.emplace_back(&output_tensor);
 
   SpaceToDepthParameter op_param;
-  op_param.op_parameter_.type_ = schema::PrimitiveType_SpaceToBatch;
+  op_param.op_parameter_.type_ = schema::PrimitiveType_SpaceToDepth;
   op_param.block_size_ = 2;
 
   lite::Context ctx;
+  ctx.threadNum = 3;
   kernel::KernelKey desc = {kernel::KERNEL_ARCH::kCPU, kNumberTypeFloat32, schema::PrimitiveType_SpaceToDepth};
   auto creator = lite::KernelRegistry::GetInstance()->GetCreator(desc);
   ASSERT_NE(creator, nullptr);

mindspore/lite/test/ut/src/runtime/kernel/arm/int8/dequantize_int8_tests.cc → mindspore/lite/test/ut/src/runtime/kernel/arm/int8/quant_dtype_cast_tests.cc

@@ -17,28 +17,26 @@
 #include <memory>
 #include "utils/log_adapter.h"
 #include "common/common_test.h"
-#include "mindspore/lite/src/runtime/kernel/arm/int8/dequantize.h"
-#include "mindspore/lite/src/runtime/kernel/arm/opclib/int8/dequantize.h"
+#include "mindspore/lite/src/runtime/kernel/arm/base/quant_dtype_cast.h"
+#include "mindspore/lite/src/runtime/kernel/arm/opclib/int8/quant_dtype_cast.h"
 #include "mindspore/lite/src/kernel_registry.h"
 #include "mindspore/lite/src/lite_kernel.h"
 
 namespace mindspore {
 
-class DequantizeTestFp32 : public mindspore::Common {
+class QuantDTypeCastTestFp32 : public mindspore::Common {
  public:
-  DequantizeTestFp32() {}
+  QuantDTypeCastTestFp32() {}
 };
 
-TEST_F(DequantizeTestFp32, DequantizeTest1) {
+TEST_F(QuantDTypeCastTestFp32, QuantDTypeCastTest1) {
   const lite::tensor::QuantArg quant_arg{0.21176, 5};
-  // quant_arg.scale = 100.0;
-  // quant_arg.zeroPoint = 20;
-  DequantizeParameter param;
-  param.op_parameter_.type_ = schema::PrimitiveType_OnnxInt8Dequantize;
+  QuantDTypeCastParameter param;
+  param.srcT = kNumberTypeInt8;
+  param.dstT = kNumberTypeFloat32;
+  param.op_parameter_.type_ = schema::PrimitiveType_QuantDTypeCast;
 
   std::vector<int8_t> input = {10, 14, 29, 33, 52, 99, 19, 43, 90, 52, 19, 24, 57, 127, 76, 123};
-  // int8_t input0[] = {1, 2, 10};
-  // int32_t a = input0[0] + 2;
   std::vector<int> in_shape = {1, 4, 4, 1};
   lite::tensor::Tensor input_tensor;
   input_tensor.SetData(input.data());
@@ -59,13 +57,13 @@ TEST_F(DequantizeTestFp32, DequantizeTest1) {
   output_tensor.SetData(output.data());
   output_tensor.set_shape(out_shape);
   output_tensor.set_data_type(kNumberTypeFloat32);
-  output_tensor.SetFormat(schema::Format_NHWC);
+  // output_tensor.SetFormat(schema::Format_NHWC);
   std::vector<lite::tensor::Tensor *> outputs_tensor;
   outputs_tensor.emplace_back(&output_tensor);
 
   lite::Context ctx;
   ctx.threadNum = 3;
-  kernel::KernelKey desc = {kernel::KERNEL_ARCH::kCPU, kNumberTypeInt8, schema::PrimitiveType_OnnxInt8Dequantize};
+  kernel::KernelKey desc = {kernel::KERNEL_ARCH::kCPU, kNumberTypeInt8, schema::PrimitiveType_QuantDTypeCast};
   auto creator = lite::KernelRegistry::GetInstance()->GetCreator(desc);
   ASSERT_NE(creator, nullptr);
   kernel::LiteKernel *kernel =
@@ -80,4 +78,49 @@ TEST_F(DequantizeTestFp32, DequantizeTest1) {
   CompareOutputData(output.data(), expect_out, out_size, 0.000001);
 }
 
+TEST_F(QuantDTypeCastTestFp32, QuantDTypeCastTest2) {
+  const lite::tensor::QuantArg quant_arg = {0.3515625, -57};
+  QuantDTypeCastParameter param;
+  param.op_parameter_.type_ = schema::PrimitiveType_QuantDTypeCast;
+  param.dstT = kNumberTypeInt8;
+  param.srcT = kNumberTypeFloat32;
+  std::vector<float> input = {1, 2, 5, 6, 10, -20, 3, 8, 18, 10, 3, 4, 11, 16, 15, 25};
+  std::vector<int> in_shape = {1, 4, 4, 1};
+  lite::tensor::Tensor input_tensor;
+  input_tensor.SetData(input.data());
+  input_tensor.set_shape(in_shape);
+  // input_tensor.SetFormat(schema::Format_NHWC);
+  input_tensor.set_data_type(kNumberTypeFloat32);
+  input_tensor.AddQuantParam(quant_arg);
+  std::vector<lite::tensor::Tensor *> inputs_tensor;
+  inputs_tensor.emplace_back(&input_tensor);
+
+  const int out_size = 16;
+  int8_t expect_out[16] = {-54, -51, -43, -40, -29, -114, -48, -34, -6, -29, -48, -46, -26, -11, -14, 14};
+  std::vector<int8_t> output(16);
+  std::vector<int> out_shape = {1, 4, 4, 1};
+  lite::tensor::Tensor output_tensor;
+  output_tensor.SetData(output.data());
+  output_tensor.set_shape(out_shape);
+  output_tensor.SetFormat(schema::Format_NHWC);
+  output_tensor.set_data_type(kNumberTypeInt8);
+  std::vector<lite::tensor::Tensor *> outputs_tensor;
+  outputs_tensor.emplace_back(&output_tensor);
+
+  lite::Context ctx;
+  ctx.threadNum = 3;
+  kernel::KernelKey desc = {kernel::KERNEL_ARCH::kCPU, kNumberTypeInt8, schema::PrimitiveType_QuantDTypeCast};
+  auto creator = lite::KernelRegistry::GetInstance()->GetCreator(desc);
+  ASSERT_NE(creator, nullptr);
+  kernel::LiteKernel *kernel =
+    creator(inputs_tensor, outputs_tensor, reinterpret_cast<OpParameter *>(&param), &ctx, desc);
+  ASSERT_NE(kernel, nullptr);
+  kernel->Run();
+
+  for (int i = 0; i < out_size; ++i) {
+    std::cout << output[i] << " ";
+  }
+  std::cout << "\n";
+  CompareOutputData(output.data(), expect_out, out_size, 0.000001);
+}
 }  // namespace mindspore