// Pack 3 variable handles into one TFE_TensorHandle.
  // When remote is false, function device is placed on task0. Handle types are
  // REMOTE, REMOTE, LOCAL on task0. When remote is true, function device is
  // placed on task1, Handle types are LOCAL, REMOTE, LOCAL on task1.
  int num_replicas = 3;
  std::vector<TFE_TensorHandle*> handles = {h0, h1, h2};
  TFE_TensorHandle* packed_handle =

{* ../../docs_src/background_tasks/tutorial001.py hl[6:9] *}

## Add the background task

Inside of your *path operation function*, pass your task function to the *background tasks* object with the method `.add_task()`:

{* ../../docs_src/background_tasks/tutorial001.py hl[14] *}

`.add_task()` receives as arguments:

* A task function to be run in the background (`write_notification`).

```Python hl_lines="6-9"
{!../../docs_src/background_tasks/tutorial001.py!}
```

## Добавление фоновой задачи

Внутри функции вызовите метод `.add_task()` у объекта *background tasks* и передайте ему функцию, которую хотите выполнить в фоне:

```Python hl_lines="14"
{!../../docs_src/background_tasks/tutorial001.py!}
```

`.add_task()` принимает следующие аргументы:

在FastAPI中仍然可以单独使用 `BackgroundTask`，但您必须在代码中创建对象，并返回包含它的Starlette `Response`。

更多细节查看 <a href="https://www.starlette.io/background/" class="external-link" target="_blank">Starlette's official docs for Background Tasks</a>.

## 告诫

如果您需要执行繁重的后台计算，并且不一定需要由同一进程运行（例如，您不需要共享内存、变量等），那么使用其他更大的工具（如 <a href="https://docs.celeryq.dev" class="external-link" target="_blank">Celery</a>）可能更好。

```Python hl_lines="6-9"
{!../../docs_src/background_tasks/tutorial001.py!}
```

## バックグラウンドタスクの追加

*path operations 関数* 内で、`.add_task()` メソッドを使用してタスク関数を *background tasks* オブジェクトに渡します。

```Python hl_lines="14"
{!../../docs_src/background_tasks/tutorial001.py!}
```

`.add_task()` は以下の引数を受け取ります:

* バックグラウンドで実行されるタスク関数 (`write_notification`)。
* タスク関数に順番に渡す必要のある引数の列 (`email`)。

  private static class FakeExecutor implements Executor {
    Queue<Runnable> tasks = Queues.newArrayDeque();

    @Override
    public void execute(Runnable command) {
      tasks.add(command);
    }

    boolean hasNext() {
      return !tasks.isEmpty();
    }

    void runNext() {
      assertTrue("expected at least one task to run", hasNext());
      tasks.remove().run();
    }

    void runAll() {

    public <T> Future<T> submit(Callable<T> task) {
      lastMethodCalled = "submit";
      assertThat(task).isInstanceOf(WrappedCallable.class);
      return inline.submit(task);
    }

    @Override
    public Future<?> submit(Runnable task) {
      lastMethodCalled = "submit";
      assertThat(task).isInstanceOf(WrappedRunnable.class);
      return inline.submit(task);
    }

    @Override

    public <T> Future<T> submit(Callable<T> task) {
      lastMethodCalled = "submit";
      assertThat(task).isInstanceOf(WrappedCallable.class);
      return inline.submit(task);
    }

    @Override
    public Future<?> submit(Runnable task) {
      lastMethodCalled = "submit";
      assertThat(task).isInstanceOf(WrappedRunnable.class);
      return inline.submit(task);
    }

    @Override

   * turn, and does not create any threads of its own.
   *
   * <p>After execution begins on a thread from the {@code delegate} {@link Executor}, tasks are
   * polled and executed from a task queue until there are no more tasks. The thread will not be
   * released until there are no more tasks to run.
   *
   * <p>If a task is submitted while a thread is executing tasks from the task queue, the thread

        // without declaring an explicit or implicit dependency. This can lead to incorrect results being produced, depending on what order the tasks are executed.
        tasks.named("publishGradleDistributionPublicationToRemoteRepository") {
            dependsOn("signLocalPublication")
        }
        tasks.named("publishLocalPublicationToLocalRepository") {