pytorch/caffe2/core/test_utils.h

#ifndef CAFFE2_UTILS_TEST_UTILS_H_
#define CAFFE2_UTILS_TEST_UTILS_H_

#include "caffe2/core/tensor.h"
#include "caffe2/core/workspace.h"
#include "caffe2/utils/proto_utils.h"

#include <c10/macros/Macros.h>
#include <c10/util/irange.h>

#include <cmath>
#include <string>
#include <vector>

// Utilities that make it easier to write caffe2 C++ unit tests.
// These utils are designed to be concise and easy to use. They may sacrifice
// performance and should only be used in tests/non production code.
namespace caffe2 {
namespace testing {

// Asserts that the values of two tensors are the same.
TORCH_API void assertTensorEquals(
    const TensorCPU& tensor1,
    const TensorCPU& tensor2,
    float eps = 1e-6);

// Asserts that two float values are close within epsilon.
TORCH_API void assertNear(float value1, float value2, float epsilon);

// Asserts that the numeric values of a tensor is equal to a data vector.
template <typename T>
void assertTensorEquals(
    const TensorCPU& tensor,
    const std::vector<T>& data,
    float epsilon = 0.1f) {
  CAFFE_ENFORCE(tensor.IsType<T>());
  CAFFE_ENFORCE_EQ(tensor.numel(), data.size());
  for (const auto idx : c10::irange(tensor.numel())) {
    if (tensor.IsType<float>()) {
      assertNear(tensor.data<T>()[idx], data[idx], epsilon);
    } else {
      CAFFE_ENFORCE_EQ(tensor.data<T>()[idx], data[idx]);
    }
  }
}

// Assertion for tensor sizes and values.
template <typename T>
void assertTensor(
    const TensorCPU& tensor,
    const std::vector<int64_t>& sizes,
    const std::vector<T>& data,
    float epsilon = 0.1f) {
  CAFFE_ENFORCE_EQ(tensor.sizes(), sizes);
  assertTensorEquals(tensor, data, epsilon);
}

// Asserts a list of tensors presented in two workspaces are equal.
TORCH_API void assertTensorListEquals(
    const std::vector<std::string>& tensorNames,
    const Workspace& workspace1,
    const Workspace& workspace2);

// Read a tensor from the workspace.
TORCH_API const caffe2::Tensor& getTensor(
    const caffe2::Workspace& workspace,
    const std::string& name);

// Create a new tensor in the workspace.
TORCH_API caffe2::Tensor* createTensor(
    const std::string& name,
    caffe2::Workspace* workspace);

// Create a new operator in the net.
TORCH_API caffe2::OperatorDef* createOperator(
    const std::string& type,
    const std::vector<std::string>& inputs,
    const std::vector<std::string>& outputs,
    caffe2::NetDef* net);

// Fill a buffer with randomly generated numbers given range [min, max)
// T can only be float, double or long double
template <typename RealType = float>
void randomFill(
    RealType* data,
    size_t size,
    const double min = 0.0,
    const double max = 1.0) {
  std::mt19937 gen(42);
  std::uniform_real_distribution<RealType> dis(
      static_cast<RealType>(min), static_cast<RealType>(max));
  for (const auto i : c10::irange(size)) {
    data[i] = dis(gen);
  }
}

// Fill data from a vector to a tensor.
template <typename T>
void fillTensor(
    const std::vector<int64_t>& shape,
    const std::vector<T>& data,
    TensorCPU* tensor) {
  tensor->Resize(shape);
  CAFFE_ENFORCE_EQ(data.size(), tensor->numel());
  auto ptr = tensor->mutable_data<T>();
  for (int i = 0; i < tensor->numel(); ++i) {
    ptr[i] = data[i];
  }
}

// Create a tensor and fill data.
template <typename T>
caffe2::Tensor* createTensorAndFill(
    const std::string& name,
    const std::vector<int64_t>& shape,
    const std::vector<T>& data,
    Workspace* workspace) {
  auto* tensor = createTensor(name, workspace);
  fillTensor<T>(shape, data, tensor);
  return tensor;
}

template <typename T>
caffe2::Tensor createTensorAndFill(
    const std::vector<int64_t>& shape,
    const std::vector<T>& data) {
  Tensor tensor(caffe2::CPU);
  fillTensor<T>(shape, data, &tensor);
  return tensor;
}

// Fill a constant to a tensor.
template <typename T>
void constantFillTensor(
    const vector<int64_t>& shape,
    const T& data,
    TensorCPU* tensor) {
  tensor->Resize(shape);
  auto ptr = tensor->mutable_data<T>();
  for (int i = 0; i < tensor->numel(); ++i) {
    ptr[i] = data;
  }
}

// Create a tensor and fill a constant.
template <typename T>
caffe2::Tensor* createTensorAndConstantFill(
    const std::string& name,
    const std::vector<int64_t>& shape,
    const T& data,
    Workspace* workspace) {
  auto* tensor = createTensor(name, workspace);
  constantFillTensor<T>(shape, data, tensor);
  return tensor;
}

// Concise util class to mutate a net in a chaining fashion.
class TORCH_API NetMutator {
 public:
  // NOLINTNEXTLINE(clang-analyzer-optin.cplusplus.UninitializedObject)
  explicit NetMutator(caffe2::NetDef* net) : net_(net) {}

  NetMutator& newOp(
      const std::string& type,
      const std::vector<std::string>& inputs,
      const std::vector<std::string>& outputs);

  NetMutator& externalInputs(const std::vector<std::string>& externalInputs);

  NetMutator& externalOutputs(const std::vector<std::string>& externalOutputs);

  // Add argument to the last created op.
  template <typename T>
  NetMutator& addArgument(const std::string& name, const T& value) {
    CAFFE_ENFORCE(lastCreatedOp_ != nullptr);
    AddArgument(name, value, lastCreatedOp_);
    return *this;
  }

  // Set device name for the last created op.
  NetMutator& setDeviceOptionName(const std::string& name);

 private:
  caffe2::NetDef* net_;
  caffe2::OperatorDef* lastCreatedOp_;
};

// Concise util class to mutate a workspace in a chaining fashion.
class TORCH_API WorkspaceMutator {
 public:
  explicit WorkspaceMutator(caffe2::Workspace* workspace)
      : workspace_(workspace) {}

  // New tensor filled by a data vector.
  template <typename T>
  WorkspaceMutator& newTensor(
      const std::string& name,
      const std::vector<int64_t>& shape,
      const std::vector<T>& data) {
    createTensorAndFill<T>(name, shape, data, workspace_);
    return *this;
  }

  // New tensor filled by a constant.
  template <typename T>
  WorkspaceMutator& newTensorConst(
      const std::string& name,
      const std::vector<int64_t>& shape,
      const T& data) {
    createTensorAndConstantFill<T>(name, shape, data, workspace_);
    return *this;
  }

 private:
  caffe2::Workspace* workspace_;
};

} // namespace testing
} // namespace caffe2

#endif // CAFFE2_UTILS_TEST_UTILS_H_