NewInternalScope(&g->graph, &status->status, &g->refiner)
            .NewSubScope(child_scope_name);

    if (dx != nullptr) {
      std::vector<tensorflow::Output> dx_arg = OutputsFromTFOutputs(dx, ny);
      status->status =
          AddSymbolicGradients(scope, y_arg, x_arg, dx_arg, &dy_arg);
    } else {
      status->status = AddSymbolicGradients(scope, y_arg, x_arg, &dy_arg);
    }

  TF_Operation* y = Placeholder(graph_, s_, "y", TF_FLOAT);
  TF_Operation* xy = Mul(x, y, graph_, s_, "xy");
  TF_Output dxy_dx, dxy_dy;

  TF_Output outputs[1] = {{xy, 0}};
  TF_Output inputs[1] = {{x, 0}};
  TF_AddGradients(graph_, outputs, 1, inputs, 1, nullptr, s_, &dxy_dx);
  ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);

  inputs[0] = {y, 0};

     *
     *    dX = U / Y
     *    dY = -U*X / Y^2 = (X/Y) * -U / Y = -U*Z / Y
     *
     */

    AbstractTensorHandle* upstream_grad = grad_outputs[0];
    AbstractTensorHandle* Y = forward_inputs_[1];
    AbstractTensorHandle* Z = forward_outputs_[0];

    // Calculate dX =  U / Y
    std::string name = "Div_Grad_X";
    TF_RETURN_IF_ERROR(

    ASSERT_EQ(errors::OK, s.code()) << s.message();
    y.reset(y_raw);
  }

  float expected_dx[1] = {7.0f};
  ASSERT_NO_FATAL_FAILURE(CompareNumericalAndManualGradients(
      MulModel, ctx_.get(), {x.get(), y.get()}, 0, expected_dx, 1,
      UseFunction()));
}

#ifdef PLATFORM_GOOGLE
INSTANTIATE_TEST_SUITE_P(
    UnifiedCAPI, GradientCheckerTest,