*       Thread acquires its first hold (and releases its last remaining hold).
 *   <li>Before the lock is acquired, the lock is checked against the current set of acquired
 *       locks---to each of the acquired locks, an edge from the soon-to-be-acquired lock is either
 *       verified or created.
 *   <li>If a new edge needs to be created, the outgoing edges of the acquired locks are traversed

   * the cache, but only the "generic" one is type-safe. That is, it will properly prevent you from
   * building caches whose key or value types are incompatible with the types accepted by the
   * weigher already provided; the {@code CacheBuilder} type cannot do this. For best results,
   * simply use the standard method-chaining idiom, as illustrated in the documentation at top,

   * given {@code Future}. If the given {@code Future} fails, the returned {@code Future} fails with
   * the same exception (and the function is not invoked).
   *
   * <p>More precisely, the returned {@code Future} takes its result from a {@code Future} produced
   * by applying the given {@code AsyncFunction} to the result of the original {@code Future}.
   * Example usage:
   *

  }

  /**
   * Returns an immutable multimap containing the specified entries. The returned multimap iterates
   * over keys in the order they were first encountered in the input, and the values for each key
   * are iterated in the order they were encountered. If two values for the same key are {@linkplain
   * Object#equals equal}, the first value encountered is used.
   *

///

### Declare a *path parameter*

Then create a *path parameter* with a type annotation using the enum class you created (`ModelName`):

```Python hl_lines="16"
{!../../docs_src/path_params/tutorial005.py!}
```

### Check the docs

Because the available values for the *path parameter* are predefined, the interactive docs can show them nicely:

<img src="/img/tutorial/path-params/image03.png">

}

// HPAScalingRules configures the scaling behavior for one direction.
// These Rules are applied after calculating DesiredReplicas from metrics for the HPA.
// They can limit the scaling velocity by specifying scaling policies.
// They can prevent flapping by specifying the stabilization window, so that the
// number of replicas is not set instantly, instead, the safest value from the stabilization
// window is chosen.

Actually, all (or most) of the web frameworks work in this same way.

You never call those functions directly. They are called by your framework (in this case, **FastAPI**).

/*
Copyright The Kubernetes Authors.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,

# GPU version of TensorFlow for Python 3.12, using the public TF bazel cache and
# a local build cache:
TFCI=py312,linux_x86_cuda,public_cache,disk_cache ci/official/wheel.sh

# First, set your TFCI variable to choose the environment settings.
#   TFCI is a comma-separated list of filenames from the envs directory, which
#   are all settings for the scripts. TF's CI jobs are all made of a combination

   * timeouts (due to a problem with the transport). When a stream times out we don't know whether
   * the problem impacts just one stream or the entire connection.
   *
   * To differentiate the two cases we ping the server when a stream times out. If the overall
   * connection is fine the ping will receive a pong; otherwise it won't.
   *