pytorch/caffe2/core/test_utils.cc

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

#include "test_utils.h"

namespace {
template <typename T>
void assertTensorEqualsWithType(
    const caffe2::TensorCPU& tensor1,
    const caffe2::TensorCPU& tensor2,
    float /* unused */) {
  CAFFE_ENFORCE_EQ(tensor1.sizes(), tensor2.sizes());
  for (auto idx = 0; idx < tensor1.numel(); ++idx) {
    CAFFE_ENFORCE_EQ(
        tensor1.data<T>()[idx],
        tensor2.data<T>()[idx],
        "Mismatch at index ",
        idx);
  }
}

template <>
void assertTensorEqualsWithType<float>(
    const caffe2::TensorCPU& tensor1,
    const caffe2::TensorCPU& tensor2,
    float eps) {
  CAFFE_ENFORCE_EQ(tensor1.sizes(), tensor2.sizes());
  for (auto idx = 0; idx < tensor1.numel(); ++idx) {
    CAFFE_ENFORCE_LT(
        fabs(tensor1.data<float>()[idx] - tensor2.data<float>()[idx]),
        eps,
        "Mismatch at index ",
        idx,
        " exceeds threshold of ",
        eps);
  }
}
} // namespace

namespace caffe2 {
namespace testing {

// Asserts that two float values are close within epsilon.
void assertNear(float value1, float value2, float epsilon) {
  // These two enforces will give good debug messages.
  CAFFE_ENFORCE_LE(value1, value2 + epsilon);
  CAFFE_ENFORCE_GE(value1, value2 - epsilon);
}

void assertTensorEquals(
    const TensorCPU& tensor1,
    const TensorCPU& tensor2,
    float eps) {
  CAFFE_ENFORCE_EQ(tensor1.sizes(), tensor2.sizes());
  if (tensor1.IsType<float>()) {
    CAFFE_ENFORCE(tensor2.IsType<float>());
    assertTensorEqualsWithType<float>(tensor1, tensor2, eps);
  } else if (tensor1.IsType<int>()) {
    CAFFE_ENFORCE(tensor2.IsType<int>());
    assertTensorEqualsWithType<int>(tensor1, tensor2, eps);
  } else if (tensor1.IsType<int64_t>()) {
    CAFFE_ENFORCE(tensor2.IsType<int64_t>());
    assertTensorEqualsWithType<int64_t>(tensor1, tensor2, eps);
  }
  // Add more types if needed.
}

void assertTensorListEquals(
    const std::vector<std::string>& tensorNames,
    const Workspace& workspace1,
    const Workspace& workspace2) {
  for (const std::string& tensorName : tensorNames) {
    CAFFE_ENFORCE(workspace1.HasBlob(tensorName));
    CAFFE_ENFORCE(workspace2.HasBlob(tensorName));
    auto& tensor1 = getTensor(workspace1, tensorName);
    auto& tensor2 = getTensor(workspace2, tensorName);
    assertTensorEquals(tensor1, tensor2);
  }
}

const caffe2::Tensor& getTensor(
    const caffe2::Workspace& workspace,
    const std::string& name) {
  CAFFE_ENFORCE(workspace.HasBlob(name));
  return workspace.GetBlob(name)->Get<caffe2::Tensor>();
}

caffe2::Tensor* createTensor(
    const std::string& name,
    caffe2::Workspace* workspace) {
  return BlobGetMutableTensor(workspace->CreateBlob(name), caffe2::CPU);
}

caffe2::OperatorDef* createOperator(
    const std::string& type,
    const std::vector<std::string>& inputs,
    const std::vector<std::string>& outputs,
    caffe2::NetDef* net) {
  auto* op = net->add_op();
  op->set_type(type);
  for (const auto& in : inputs) {
    op->add_input(in);
  }
  for (const auto& out : outputs) {
    op->add_output(out);
  }
  return op;
}

NetMutator& NetMutator::newOp(
    const std::string& type,
    const std::vector<std::string>& inputs,
    const std::vector<std::string>& outputs) {
  lastCreatedOp_ = createOperator(type, inputs, outputs, net_);
  return *this;
}

NetMutator& NetMutator::externalInputs(
    const std::vector<std::string>& externalInputs) {
  for (auto& blob : externalInputs) {
    net_->add_external_input(blob);
  }
  return *this;
}

NetMutator& NetMutator::externalOutputs(
    const std::vector<std::string>& externalOutputs) {
  for (auto& blob : externalOutputs) {
    net_->add_external_output(blob);
  }
  return *this;
}

NetMutator& NetMutator::setDeviceOptionName(const std::string& name) {
  CAFFE_ENFORCE(lastCreatedOp_ != nullptr);
  lastCreatedOp_->mutable_device_option()->set_node_name(name);
  return *this;
}

} // namespace testing
} // namespace caffe2