// CPU tensor to GPU).
  // Setting a larger thread pool does not help with the Swift caller, as we use
  // a different TFE context for each thread of execution (for running graph
  // functions, and their send/recvs corountines).
  config.set_inter_op_parallelism_threads(1);

  TF_Buffer* ret = TF_NewBuffer();
  TF_CHECK_OK(MessageToBuffer(config, ret));
  return ret;
}

  TFE_OpSetAttrString(op, "tensor_name", "dummy", 5);
  TFE_OpSetAttrString(op, "send_device", send_device.c_str(),
                      send_device.size());
  TFE_OpSetAttrString(op, "recv_device", recv_device.c_str(),
                      recv_device.size());
  TFE_OpSetAttrInt(op, "send_device_incarnation", send_device_incarnation);

  return op;
}

  TFE_TensorHandle* h0_task0 = TestMatrixTensorHandle(ctx);
  TFE_TensorHandle* h1_task0 = TestMatrixTensorHandle(ctx);
  std::vector<TFE_TensorHandle*> handles_task0;
  if (heavy_load_on_streaming_rpc) {
    // Send 50 tensor copy requests to simulate that there have been some RPC
    // requests been enqueued.
    for (int i = 0; i < 50; ++i) {
      handles_task0.push_back(TestMatrixTensorHandle(ctx));
    }
  }