TF_RETURN_IF_ERROR(
        RunAndMaybeSum(ctx, forward, theta_inputs, f_outputs, use_function));
    AbstractTensorHandlePtr fPlus(f_outputs[0]);

    // Get f(theta - eps):
    theta_inputs[input_index] = thetaMinus.get();
    TF_RETURN_IF_ERROR(
        RunAndMaybeSum(ctx, forward, theta_inputs, f_outputs, use_function));
    AbstractTensorHandlePtr fMinus(f_outputs[0]);

  }
  return outputs;
}

void TFOutputsFromOutputs(const std::vector<tensorflow::Output>& outputs,
                          TF_Output* tf_outputs) {
  for (int i = 0; i < outputs.size(); i++) {
    tf_outputs[i].oper = ToOperation(outputs[i].node());
    tf_outputs[i].index = outputs[i].index();
  }
}

//  noutputs - number of elements in `outputs` array
//  outputs - array of TF_Outputs that specify the outputs of the function.
//            If `noutputs` is zero (the function returns no outputs), `outputs`
//            can be null. `outputs` can contain the same tensor more than once.
//  output_names - The names of the function's outputs. `output_names` array

    inputs_.push_back(t->output_);
    *output = tensorflow::down_cast<TracingTensorHandle*>(outputs[0]);
    return absl::OkStatus();
  }

  absl::Status Finalize(OutputList* outputs, AbstractFunction** f) override {
    std::vector<TF_Output> graph_outputs;
    graph_outputs.reserve(outputs->outputs.size());
    for (auto* abstract_output : outputs->outputs) {
      GraphTensor* output = dyn_cast<GraphTensor>(abstract_output);

                            int noutputs, const TF_Output* outputs,
                            std::vector<OutputTensor>* output_tensors)
    TF_EXCLUSIVE_LOCKS_REQUIRED(fn_body->mu) {
  output_tensors->reserve(noutputs);
  for (int i = 0; i < noutputs; ++i) {
    Node* node = outputs[i].oper ? &outputs[i].oper->node : nullptr;
    int idx = outputs[i].index;
    TF_RETURN_WITH_CONTEXT_IF_ERROR(

  void TestGradientsSuccess(bool grad_inputs_provided) {
    TF_Output inputs[2];
    TF_Output outputs[1];
    TF_Output grad_outputs[2];
    TF_Output expected_grad_outputs[2];

    BuildSuccessGraph(inputs, outputs);
    BuildExpectedGraph(grad_inputs_provided, expected_grad_outputs);

    AddGradients(grad_inputs_provided, nullptr, inputs, 2, outputs, 1,
                 grad_outputs);

  VLOG(1) << "Running the dequeue op";
  TF_Output output{dequeue_op, 0};
  TF_Tensor* ret;
  TF_SessionRun(session, /*run_options*/ nullptr,
                // input related parameters
                /*inputs*/ nullptr, /*input_values*/ nullptr, /*ninputs*/ 0,
                // output related parameters
                /*outputs*/ &output, /*output_values*/ &ret,
                /*noutputs*/ 1,

  TF_Operation* feed2 = Placeholder(func_graph_, s_, "feed2");

  // Outputs of the while loop corresponding to the two inputs above
  // The first one will the function's output
  std::vector<TF_Output> outputs;

  // Add while loop to func_graph_
  {
    // The inputs to the while loop
    std::vector<TF_Output> inputs = {{feed1, 0}, {feed2, 0}};
    std::unique_ptr<TF_WhileParams> params(new TF_WhileParams(

  inputs_.clear();
  for (const auto& p : inputs) {
    inputs_.emplace_back(TF_Output{p.first, 0});
    input_values_.emplace_back(p.second);
  }
}

void CSession::SetOutputs(std::initializer_list<TF_Operation*> outputs) {
  ResetOutputValues();
  outputs_.clear();
  for (TF_Operation* o : outputs) {
    outputs_.emplace_back(TF_Output{o, 0});
  }
  output_values_.resize(outputs_.size());
}

    }
    if (TF_GetCode(s) != TF_OK) return;
  }
  std::vector<TFE_TensorHandle*> op_outputs(*num_outputs);
  TFE_Execute(op, op_outputs.data(), num_outputs, s);
  TFE_DeleteOp(op);
  if (TF_GetCode(s) != TF_OK) return;
  std::vector<TFE_TensorHandle*> unwrapped_outputs;
  unwrapped_outputs.reserve(op_outputs.size());
  for (auto* handle : op_outputs) {
    unwrapped_outputs.push_back(handle);
  }