TF_DeviceList* device_list = list.get();

  CHECK_EQ(TF_OK, TF_GetCode(s)) << TF_Message(s);

  const int num_devices = TF_DeviceListCount(device_list);
  LOG(INFO) << "There are " << num_devices << " devices.";
  for (int i = 0; i < num_devices; ++i) {
    const char* device_name = TF_DeviceListName(device_list, i, s);
    CHECK_EQ(TF_OK, TF_GetCode(s)) << TF_Message(s);

                                                   size_t proto_len,
                                                   TF_Status* status);

// TODO(b/166642410): It would be nice, for custom devices and for other users,
// to have a non-string representation of devices (TF_Device) extracted from
// tensors/ops/etc. and usable in APIs like OpSetDevice/ResetOp/etc.

#define TFE_CUSTOM_DEVICE_VERSION 4

# pretend the path is a directory.
TFCI_ARTIFACT_FINAL_GCS_ENABLE=1
TFCI_ARTIFACT_FINAL_GCS_SA_PATH="${KOKORO_KEYSTORE_DIR}/73361_tensorflow_release_binary_uploader_service_account"
TFCI_ARTIFACT_FINAL_GCS_URI="gs://tensorflow/versions/"
TFCI_ARTIFACT_FINAL_PYPI_ARGS="--config-file=$KOKORO_KEYSTORE_DIR/73361_tensorflow_pypirc_using_global_api_token --repository pypi-warehouse"

  // Check if the remote task is alive.
  virtual Status CheckRemoteAlive(const std::string& remote_task_name,
                                  bool* is_alive) = 0;

  // Get pointer to the coordination service agent instance.
  virtual tsl::CoordinationServiceAgent* GetCoordinationServiceAgent() = 0;
};
}  // namespace tensorflow

  // in an initialized context.
  for (auto d = devices.begin(); d != devices.end();) {
    if (absl::StrContains(d->get()->name(), "CPU:0")) {
      d = devices.erase(d);
    } else {
      ++d;
    }
  }

  status->status = tensorflow::unwrap(ctx)->AddDevices(std::move(devices));
}

void TFE_InsertConfigKeyValue(TFE_Context* ctx, const char* key,

      TF_NewStatus(), TF_DeleteStatus);
  TF_DeviceList* devices = TFE_ContextListDevices(ctx, status.get());
  CHECK_EQ(TF_OK, TF_GetCode(status.get())) << TF_Message(status.get());

  const int num_devices = TF_DeviceListCount(devices);
  for (int i = 0; i < num_devices; ++i) {
    const string dev_type(TF_DeviceListType(devices, i, status.get()));

    *   Added support for cross-trainer data caching in tf.data service. This
        saves computation resources when concurrent training jobs train from the
        same dataset. See
        (https://www.tensorflow.org/api_docs/python/tf/data/experimental/service#sharing_tfdata_service_with_concurrent_trainers)
        for more details.
    *   Added `dataset_id` to `tf.data.experimental.service.register_dataset`.

  pip install twine==3.6.0
fi

# Scheduled nightly and release builds upload build artifacts (Pip packages,
# Libtensorflow archives) to GCS buckets. TFCI Mac VMs need to authenticate as
# a service account that has the right permissions to be able to do so.
set +x
if [[ -n "${GOOGLE_APPLICATION_CREDENTIALS:-}" ]]; then
  # Python 3.12 removed the module `imp` which is needed by gcloud CLI so we set

  }
  std::vector<std::unique_ptr<tensorflow::Device>> devices;
  status->status = tensorflow::DeviceFactory::AddDevices(
      opts->session_options.options, "/job:localhost/replica:0/task:0",
      &devices);
  if (!status->status.ok()) return nullptr;
  std::unique_ptr<tensorflow::DeviceMgr> device_mgr(
      new tensorflow::DynamicDeviceMgr(std::move(devices)));

        "//tensorflow/core/distributed_runtime/eager:eager_client",
        "//tensorflow/core/distributed_runtime/rpc:grpc_channel",
        "//tensorflow/core/distributed_runtime/rpc:grpc_server_lib",
        "//tensorflow/core/distributed_runtime/rpc:grpc_worker_cache",
        "//tensorflow/core/distributed_runtime/rpc:grpc_worker_service",