update your imports accordingly, the old files will be removed in Release
    2.11.
*   `tf.keras.optimizers.experimental.Optimizer` will graduate in Release 2.11,
    which means `tf.keras.optimizers.Optimizer` will be an alias of
    `tf.keras.optimizers.experimental.Optimizer`. The current
    `tf.keras.optimizers.Optimizer` will continue to be supported as
    `tf.keras.optimizers.legacy.Optimizer`,

"elég", "ellen", "elő", "először", "előtt", "első", "én", "éppen", "ebben", "ehhez", "emilyen", "ennek", "erre", "ez", "ezt", "ezek", "ezen", "ezzel", "ezért", "és", "fel", "felé", "hanem", "hiszen", "hogy", "hogyan", "igen", "így", "illetve", "ill.", "ill", "ilyen", "ilyenkor", "ison", "ismét", "itt", "jó", "jól", "jobban", "kell", "kellett", "keresztül", "keressünk", "ki", "kívül", "között", "közül", "legalább", "lehet", "lehetett", "legyen", "lenne", "lenni", "lesz", "lett", "maga", "magát", "majd",...

"elég", "ellen", "elő", "először", "előtt", "első", "én", "éppen", "ebben", "ehhez", "emilyen", "ennek", "erre", "ez", "ezt", "ezek", "ezen", "ezzel", "ezért", "és", "fel", "felé", "hanem", "hiszen", "hogy", "hogyan", "igen", "így", "illetve", "ill.", "ill", "ilyen", "ilyenkor", "ison", "ismét", "itt", "jó", "jól", "jobban", "kell", "kellett", "keresztül", "keressünk", "ki", "kívül", "között", "közül", "legalább", "lehet", "lehetett", "legyen", "lenne", "lenni", "lesz", "lett", "maga", "magát", "majd",...

            V1Field["Pydantic v1 Model"]
        end
        subgraph V1["Pydantic v1 Model"]
            V2Field["Pydantic v2 Model"]
        end
    end

    style V2 fill:#f9fff3
    style V1 fill:#fff6f0
    style V1Field fill:#fff6f0
    style V2Field fill:#f9fff3
```

...but, you can have separated models using Pydantic v1 and v2 in the same app.

```mermaid
graph TB
    subgraph "✅ Supported"
        direction TB

numpy>=2.2.6 ; python_version == "3.14"
wheel ~= 0.41.2
h5py >= 3.11.0, < 3.15.0 ; python_version <= "3.13"
h5py ~= 3.15.1 ; python_version == "3.14"
lit ~= 17.0.2
opt_einsum == 3.3.0
astunparse == 1.6.3
dill == 0.3.7
astor == 0.7.1
typing_extensions ~= 4.14.1
gast == 0.4.0
termcolor == 2.3.0
wrapt == 1.16.0
tblib == 2.0.0
ml_dtypes >= 0.5.4, < 0.6.0
auditwheel >= 6.1.0
# Install tensorboard, and keras

**Note**: You can mix `def` and `async def` in your *path operation functions* as much as you need and define each one using the best option for you. FastAPI will do the right thing with them.

Anyway, in any of the cases above, FastAPI will still work asynchronously and be extremely fast.

But by following the steps above, it will be able to do some performance optimizations.

## Technical Details { #technical-details }

### Server Memory { #server-memory }

For example, if your code loads a Machine Learning model with **1 GB in size**, when you run one process with your API, it will consume at least 1 GB of RAM. And if you start **4 processes** (4 workers), each will consume 1 GB of RAM. So in total, your API will consume **4 GB of RAM**.

      Iterable<E> expectedElements,
      KnownOrder knownOrder,
      int startIndex) {
    // periodically we should manually try (steps * 3 / 2) here; all tests but
    // one should still pass (testVerifyGetsCalled()).
    stimuli = (Stimulus<E, ? super I>[]) new Stimulus<?, ?>[steps];
    checkArgument(elementsToInsertIterable.iterator().hasNext());
    elementsToInsert = Helpers.cycle(elementsToInsertIterable);

/// info

If you are using containers, for example with Docker or Kubernetes, I'll tell you more about that in the next chapter: [FastAPI in Containers - Docker](docker.md).

In particular, when running on **Kubernetes** you will probably **not** want to use workers and instead run **a single Uvicorn process per container**, but I'll tell you about it later in that chapter.

///

## Multiple Workers { #multiple-workers }

    * This will be used by the **automatic docs**.
    * It will also be used by automatic client code generation tools.
* **Serialize** the returned data to JSON using Pydantic, which is written in **Rust**, so it will be **much faster**.

But most importantly:

* It will **limit and filter** the output data to what is defined in the return type.