Status RunAndMaybeSum(AbstractContext* ctx, Model forward,
                      absl::Span<AbstractTensorHandle* const> inputs,
                      absl::Span<AbstractTensorHandle*> outputs,
                      bool use_function) {
  AbstractTensorHandle* model_outputs[1];

  // Run the model.
  TF_RETURN_IF_ERROR(
      RunModel(forward, ctx, inputs, model_outputs, use_function));

#include "tensorflow/c/eager/c_api.h"
#include "tensorflow/c/eager/c_api_experimental.h"

namespace tensorflow {
namespace parallel_device {

// Allocate a parallel device named `device_name` which forwards operations to
// `underlying_devices`, maintaining "parallel tensors" with components placed
// on each underlying device.
//
// For example if `device_name` is
//   "/job:localhost/replica:0/task:0/device:CUSTOM:0"

                    CancellationManager& cancellation_manager,
                    absl::optional<int64_t> step_id = absl::nullopt);
  // Block until the previous `StartExecute` operation has executed. Forwards
  // the status from `TFE_Execute` and returns outputs if the status is OK.
  std::vector<TensorHandlePtr> Join(TF_Status* status);

  // Block until all Ops finished running on the thread.