yield
    # Clean up the ML models and release the resources
    ml_models.clear()


app = FastAPI(lifespan=lifespan)


@app.get("/predict")
async def predict(x: float):
    result = ml_models["answer_to_everything"](x)
    return {"result": result}
```

**Note**: This is the recommended way going forward, instead of using `startup` and `shutdown` events.

    *   `Model.fit_generator`, `Model.evaluate_generator`,
        `Model.predict_generator`, `Model.train_on_batch`,
        `Model.test_on_batch`, and `Model.predict_on_batch` methods now respect
        the `run_eagerly` property, and will correctly run using `tf.function`
        by default. Note that `Model.fit_generator`, `Model.evaluate_generator`,
        and `Model.predict_generator` are deprecated endpoints. They are

        "//learning/lib/ami/simple_ml/...",
        "//learning/pathways/...",
        "//learning/serving/contrib/tfrt/mlir/canonical_ops/...",
        "//learning/serving/experimental/remote_predict/...",
        "//perftools/accelerators/xprof/convert/...",
        "//perftools/accelerators/xprof/integration_tests/...",
        "//smartass/brain/configure/...",
        "//tensorflow/...",

### Transition Status

//     Those cannot be well described when embedded.
//  3. Inconsistent validation.  Because the usages are different, the validation rules are different by usage, which makes it hard for users to predict what will happen.
//  4. The fields are both imprecise and overly precise.  Kind is not a precise mapping to a URL. This can produce ambiguity