// Get number of elements and fill data.
  int num_elems = TF_TensorElementCount(theta_tensor);
  vector<float> theta_data(num_elems);
  memcpy(theta_data.data(), TF_TensorData(theta_tensor),
         TF_TensorByteSize(theta_tensor));

  // Initialize space for the numerical gradient.
  vector<float> dtheta_approx(num_elems);

  // Get theta shape and store in theta_dims.

TFE_TensorHandle* TestScalarTensorHandle(TFE_Context* ctx, float value) {
  float data[] = {value};
  TF_Status* status = TF_NewStatus();
  TF_Tensor* t = TFE_AllocateHostTensor(ctx, TF_FLOAT, nullptr, 0, status);
  memcpy(TF_TensorData(t), &data[0], TF_TensorByteSize(t));
  TFE_TensorHandle* th = TFE_NewTensorHandleFromTensor(ctx, t, status);
  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TF_DeleteTensor(t);
  TF_DeleteStatus(status);

  // Check that a write to one tensor shows up in the other.
  *(static_cast<int64_t*>(TF_TensorData(a))) = 4;
  EXPECT_EQ(4, *(static_cast<int64_t*>(TF_TensorData(b))));
  *(static_cast<int64_t*>(TF_TensorData(b))) = 6;
  EXPECT_EQ(6, *(static_cast<int64_t*>(TF_TensorData(a))));

  TF_DeleteTensor(a);
  TF_DeleteTensor(b);
}

TEST(CAPI, TestFromProto) {

  ASSERT_EQ(TF_OK, TF_GetCode(status.get())) << TF_Message(status.get());
  TF_Tensor* result_tensor = TFE_TensorHandleResolve(result_t, status.get());
  float* result_value = static_cast<float*>(TF_TensorData(result_tensor));
  EXPECT_EQ(*result_value, 4.0);

  TF_DeleteTensor(result_tensor);
  TF_DeleteAbstractTensor(result);
  TF_DeleteOutputList(o);
  TF_DeleteExecutionContext(ctx);
}

// MatMul Test

  ASSERT_EQ(TF_GetCode(status), TF_OK) << TF_Message(status);
  TFE_DeleteTensorHandle(retvals[0]);
  float sum = 0;
  EXPECT_EQ(sizeof(sum), TF_TensorByteSize(t));
  memcpy(&sum, TF_TensorData(t), TF_TensorByteSize(t));
  TF_DeleteTensor(t);
  EXPECT_EQ(sum, 6.0);

  TFE_DeleteTensorHandle(h0);
  TFE_DeleteTensorHandle(h1);
  TFE_DeleteTensorHandle(h2);

  auto num_elem_numerical = TF_TensorElementCount(numerical_tensor);
  ASSERT_EQ(num_elem_numerical, num_grad);

  float* dnumerical = new float[num_elem_numerical]{0};
  memcpy(&dnumerical[0], TF_TensorData(numerical_tensor),
         TF_TensorByteSize(numerical_tensor));

  for (int j = 0; j < num_grad; j++) {
    ASSERT_NEAR(dnumerical[j], expected_grad[j], abs_error);
  }
  delete[] dnumerical;

  std::unique_ptr<float[]> actual_values(new float[expected_values.size()]);
  EXPECT_EQ(sizeof(float) * expected_values.size(), TF_TensorByteSize(t));
  memcpy(actual_values.get(), TF_TensorData(t), TF_TensorByteSize(t));
  TF_DeleteTensor(t);

  for (int i = 0; i < expected_values.size(); i++) {
    EXPECT_EQ(expected_values[i], actual_values[i])

                                                          TF_Status* status);

// Allocate and return a new Tensor on the host.
//
// The caller must set the Tensor values by writing them to the pointer returned
// by TF_TensorData with length TF_TensorByteSize.
TF_CAPI_EXPORT extern TF_Tensor* TFE_AllocateHostTensor(TFE_Context* ctx,
                                                        TF_DataType dtype,