// GradientTape which uses ones.
        Gradient* zero = input_tensors[target_index].ZerosLike();
        new_zeros.push_back(zero);
        in_grads.push_back(zero);
      } else {
        in_grads.push_back(nullptr);
      }
    } else {
      in_grads.push_back(current_grad->second);
    }
  }

  // Avoid infinite recursion. Whichever forward function we run, it'll end up

    for (int i = 0; i < noutputs; ++i) {
      output_names_vec.push_back(string(output_names[i]));
    }
  }

  // Process control output names.
  std::vector<string> control_output_names_vec;
  if (control_output_names) {
    control_output_names_vec.reserve(ncontrol_outputs);
    for (int i = 0; i < ncontrol_outputs; ++i) {
      control_output_names_vec.push_back(string(control_output_names[i]));
    }
  }

    components.reserve(underlying_devices_.size());
    for (int j = 0; j < underlying_devices_.size(); ++j) {
      components.push_back(std::move(per_device_output_tensors[j][i]));
    }
    if (expected_output_shapes[i].IsFullyDefined()) {
      per_device_outputs.push_back(ParallelTensor::FromTensorHandles(
          *this, std::move(components),
          absl::Span<const int64_t>(expected_output_shapes[i].dim_sizes()),

    plugin_status.reserve(files.size());
    for (int i = 0; i < files.size(); i++)
      plugin_status.push_back(TF_NewStatus());
    result = ops_->paths_exist(filesystem_.get(), translated_names.data(),
                               files.size(), plugin_status.data());
    for (int i = 0; i < files.size(); i++) {
      status->push_back(StatusFromTF_Status(plugin_status[i]));
      TF_DeleteStatus(plugin_status[i]);
    }
  }

  forward_op_->inputs.push_back(input);
  return absl::OkStatus();
}
Status AddInputList(AbstractOperation* op_,
                    absl::Span<AbstractTensorHandle* const> inputs,
                    ForwardOperation* forward_op_) {
  TF_RETURN_IF_ERROR(op_->AddInputList(inputs));
  for (auto input : inputs) {
    forward_op_->inputs.push_back(input);
  }
  return absl::OkStatus();
}

        lengths_vector.reserve(s_size);
        for (int i = 0; i < s_size; ++i) {
          const string& v = default_value.list().s(i);
          values_vector.push_back(v.data());
          lengths_vector.push_back(v.size());
        }
        TFE_OpSetAttrStringList(op, attr_name, values_vector.data(),
                                lengths_vector.data(), s_size);
      }

      // Int

    TF_Function* func = TF_FunctionImportFunctionDef(
        binary_proto_buf.data(), binary_proto_buf.size(), status);
    if (!status->status.ok()) return {};
    ret.push_back(UniqueFuncPtr(func, TF_DeleteFunction));
  }
  return ret;
}

TF_Tensor* TF_DequeueNamedTensor(TF_Session* session, int tensor_id,
                                 TF_Status* status) {
  assert(session);

    }
    auto* t = dyn_cast<GraphTensor>(outputs[0]);
    if (!t) {
      return tensorflow::errors::InvalidArgument(
          "Unable to cast input to GraphTensor");
    }
    inputs_.push_back(t->output_);
    *output = tensorflow::down_cast<TracingTensorHandle*>(outputs[0]);
    return absl::OkStatus();
  }

  Status Finalize(OutputList* outputs, AbstractFunction** f) override {

    TFE_TensorHandle* input_eager = TestScalarTensorHandle(eager_ctx, 2.0f);
    func_args.push_back(TF_CreateAbstractTensorFromEagerTensor(input_eager, s));
    ASSERT_EQ(TF_OK, TF_GetCode(s)) << TF_Message(s);
    input_eager = TestScalarTensorHandle(eager_ctx, 3.0f);
    func_args.push_back(TF_CreateAbstractTensorFromEagerTensor(input_eager, s));
    ASSERT_EQ(TF_OK, TF_GetCode(s)) << TF_Message(s);
  }

  return wrap(unwrap(o)->outputs[i]);
}
void TF_OutputListPushBack(TF_OutputList* o, TF_AbstractTensor* tensor,
                           TF_Status* s) {
  unwrap(o)->outputs.push_back(unwrap(tensor));
}

void TF_AbstractOpSetOpType(TF_AbstractOp* op, const char* const op_type,
                            TF_Status* s) {
  tsl::Set_TF_Status_from_Status(
      s, unwrap(op)->Reset(op_type,