the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all
      other entities that control, are controlled by, or are under common
      control with that entity. For the purposes of this definition,
      "control" means (i) the power, direct or indirect, to cause the
      direction or management of such entity, whether by contract or

      the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all
      other entities that control, are controlled by, or are under common
      control with that entity. For the purposes of this definition,
      "control" means (i) the power, direct or indirect, to cause the
      direction or management of such entity, whether by contract or

      the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all
      other entities that control, are controlled by, or are under common
      control with that entity. For the purposes of this definition,
      "control" means (i) the power, direct or indirect, to cause the
      direction or management of such entity, whether by contract or

You can also return Pydantic models (you'll see more about that later).

There are many other objects and models that will be automatically converted to JSON (including ORMs, etc). Try using your favorite ones, it's highly probable that they are already supported.

## Recap { #recap }

* Import `FastAPI`.
* Create an `app` instance.

# Deployments Concepts { #deployments-concepts }

When deploying a **FastAPI** application, or actually, any type of web API, there are several concepts that you probably care about, and using them you can find the **most appropriate** way to **deploy your application**.

Some of the important concepts are:

* Security - HTTPS
* Running on startup
* Restarts
* Replication (the number of processes running)
* Memory
* Previous steps before starting

 * These always reflect the latest state of the multimap itself. When they support modification, the
 * changes are <i>write-through</i> (they automatically update the backing multimap). These view
 * collections are:
 *
 * <ul>
 *   <li>{@link #asMap}, mentioned above
 *   <li>{@link #keys}, {@link #keySet}, {@link #values}, {@link #entries}, which are similar to the
 *       corresponding view collections of {@link Map}

   * an IPv4 address or not.
   *
   * <p>IPv6 addresses are <b>coerced</b> to IPv4 addresses before being converted to integers.
   *
   * <p>As long as there are applications that assume that all IP addresses are IPv4 addresses and
   * can therefore be converted safely to integers (for whatever purpose) this function can be used

That's what we'll solve, let's load the model before the requests are handled, but only right before the application starts receiving requests, not while  the code is being loaded.

## Lifespan { #lifespan }

# Cuda Comment
cuda_comment: >
   From the template it looks like you are installing **TensorFlow** (TF) prebuilt binaries:
      * For TF-GPU - See point 1
      * For TF-CPU - See point 2
   -----------------------------------------------------------------------------------------------
   
   **1. Installing **TensorFlow-GPU** (TF) prebuilt binaries**
   
   
   Make sure you are using compatible TF and CUDA versions.

     * This filter can be considerably more efficient than other file filters
     * as the specified wildcard and attributes are passed to the server for
     * filtering there (although attributes are largely ignored by servers
     * they are filtered locally by the default accept method).
     *
     * @param wildcard the wildcard pattern for matching file names (e.g., "*.txt")