/// tip | Consejo

El `shutdown` ocurriría cuando estás **deteniendo** la aplicación.

Quizás necesites iniciar una nueva versión, o simplemente te cansaste de ejecutarla. 🤷

///

TEST(CAPI, Session_Min_GPU) {
  const string gpu_device = GPUDeviceName();
  // Skip this test if no GPU is available.
  if (gpu_device.empty()) return;

  RunMinTest(gpu_device, /*use_XLA=*/false);
}

TEST(CAPI, Session_Min_XLA_GPU) {
  const string gpu_device = GPUDeviceName();
  // Skip this test if no GPU is available.
  if (gpu_device.empty()) return;

  RunMinTest(gpu_device, /*use_XLA=*/true);
}

/// tip | Tipp

Das `shutdown` würde erfolgen, wenn Sie die Anwendung **stoppen**.

Möglicherweise müssen Sie eine neue Version starten, oder Sie haben es einfach satt, sie auszuführen. 🤷

///

// Sets XLA's auto jit mode according to the specified string, which is parsed
// as if passed in XLA_FLAGS. This has global effect.
TF_CAPI_EXPORT void TF_SetXlaAutoJitMode(const char* mode);

// Returns whether the single GPU or general XLA auto jit optimizations are
// enabled through MarkForCompilationPassFlags.
TF_CAPI_EXPORT unsigned char TF_GetXlaAutoJitEnabled();

// Sets XLA's minimum cluster size. This has global effect.

{!../../docs_src/events/tutorial003.py!}
```

在这里，我们在 `yield` 之前将（虚拟的）模型函数放入机器学习模型的字典中，以此模拟加载模型的耗时**启动**操作。这段代码将在应用程序**开始处理请求之前**执行，即**启动**期间。

然后，在 `yield` 之后，我们卸载模型。这段代码将会在应用程序**完成处理请求后**执行，即在**关闭**之前。这可以释放诸如内存或 GPU 之类的资源。

/// tip | 提示

**关闭**事件只会在你停止应用时触发。

可能你需要启动一个新版本，或者你只是你厌倦了运行它。 🤷

///

## 生命周期函数

首先要注意的是，我们定义了一个带有 `yield` 的异步函数。这与带有 `yield` 的依赖项非常相似。

```Python hl_lines="14-19"

    def "matches google-chrome-for-testing"() {
        def line = '3723579 /usr/bin/google-chrome-for-testing --allow-pre-commit-input --disable-background-networking --disable-client-side-phishing-detection --disable-default-apps --disable-gpu --disable-hang-monitor --disable-popup-blocking --disab'

        def projectDir = "/whatever"

        expect:
        (line =~ KillLeakingJavaProcesses.generateLeakingProcessKillPattern(projectDir)).find()
    }

  TF_AddInput(desc, {zero, 0});
  TF_AddInput(desc, {input, 0});
  TF_SetAttrInt(desc, "num_split", 3);
  TF_SetAttrType(desc, "T", TF_INT32);
  // Set device to CPU since there is no version of split for int32 on GPU
  // TODO(iga): Convert all these helpers and tests to use floats because
  // they are usually available on GPUs. After doing this, remove TF_SetDevice
  // call in c_api_function_test.cc
  TF_SetDevice(desc, "/cpu:0");

* Do not query the metadata server to find out if running on GCE.  Retry metadata server query for gcr if running on gce. ([#28871](https://github.com/kubernetes/kubernetes/pull/28871), [@vishh](https://github.com/vishh))
* Fix GPU resource validation ([#28743](https://github.com/kubernetes/kubernetes/pull/28743), [@therc](https://github.com/therc))

)

config_setting(
    name = "with_xla_support",
    define_values = {"with_xla_support": "true"},
    visibility = ["//visibility:public"],
)

# By default, XLA GPU is compiled into tensorflow when building with
# --config=cuda even when `with_xla_support` is false. The config setting
# here allows us to override the behavior if needed.
config_setting(
    name = "no_xla_deps_in_cuda",

    * Fix stop hook failure on kubernetes-worker charm

    * Fix handling of juju kubernetes-worker.restart-needed state

    * Fix nagios checks in charms

  * Enable GPU mode if GPU hardware detected ([#43467](https://github.com/kubernetes/kubernetes/pull/43467), [@tvansteenburgh](https://github.com/tvansteenburgh))