load("//tensorflow:tensorflow.default.bzl", "filegroup")
load("//tensorflow/core/platform:rules_cc.bzl", "cc_library")

# Experimental ops. These will eventually be replaced by machine-generated versions.
package(
    # copybara:uncomment default_applicable_licenses = ["//tensorflow:license"],
    licenses = ["notice"],
)

cc_library(
    name = "array_ops",
    srcs = [
        "array_ops.cc",
    ],
    hdrs = [
        "array_ops.h",

Voici un petit bout de cette histoire.

## Alternatives

Je crée des API avec des exigences complexes depuis plusieurs années (Machine Learning, systèmes distribués, jobs asynchrones, bases de données NoSQL, etc), en dirigeant plusieurs équipes de développeurs.

Dans ce cadre, j'ai dû étudier, tester et utiliser de nombreuses alternatives.

of these things:

    a) Accompany the work with the complete corresponding
    machine-readable source code for the Library including whatever
    changes were used in the work (which must be distributed under
    Sections 1 and 2 above); and, if the work is an executable linked
    with the Library, with the complete machine-readable "work that
    uses the Library", as object code and/or source code, so that the

pkg crypto/tls, type RecordHeaderError struct, Conn net.Conn
pkg debug/elf, const R_RISCV_32_PCREL = 57
pkg debug/elf, const R_RISCV_32_PCREL R_RISCV
pkg debug/pe, const IMAGE_FILE_MACHINE_ARMNT = 452
pkg debug/pe, const IMAGE_FILE_MACHINE_ARMNT ideal-int
pkg expvar, method (*Map) Delete(string)
pkg go/doc, const PreserveAST = 4
pkg go/doc, const PreserveAST Mode

  }

  companion object {
    /**
     * A network that resolves only one IP address per host. Use this when testing route selection
     * fallbacks to prevent the host machine's various IP addresses from interfering.
     */
    private val SINGLE_INET_ADDRESS_DNS =
      Dns { hostname ->
        val addresses = Dns.SYSTEM.lookup(hostname)
        listOf(addresses[0])
      }

In the code, a `Bucket` represents the main entrypoint of communication with the database.

This utility function will:

* Connect to a **Couchbase** cluster (that might be a single machine).
    * Set defaults for timeouts.
* Authenticate in the cluster.
* Get a `Bucket` instance.
    * Set defaults for timeouts.
* Return it.

```Python hl_lines="12-21"

build --linkopt="-lm"
build --linkopt="-Wl,--undefined-version"

# Prevent double-compilation of some TF code, ref. b/183279666 (internal)
# > TF's gen_api_init_files has a genrule to run the core TensorFlow code
# > on the host machine. If we don't have --distinct_host_configuration=false,
# > the core TensorFlow code will be built once for the host and once for the
# > target platform.

En utilisant le mode Docker Swarm, vous pouvez commencer par un "cluster" d'une seule machine (il peut même s'agir
d'un serveur à 5 USD/mois) et ensuite vous pouvez vous développer autant que vous le souhaitez en ajoutant d'autres serveurs.

build --define=tf_api_version=2 --action_env=TF2_BEHAVIOR=1

# Prevent double-compilation of some TF code, ref. b/183279666 (internal)
# > TF's gen_api_init_files has a genrule to run the core TensorFlow code
# > on the host machine. If we don't have --distinct_host_configuration=false,
# > the core TensorFlow code will be built once for the host and once for the
# > target platform.

```hl_lines="4"
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
```

Cette ligne montre l'URL par laquelle l'app est actuellement accessible, sur votre machine locale.

### Allez voir le résultat

Ouvrez votre navigateur à l'adresse <a href="http://127.0.0.1:8000" class="external-link" target="_blank">http://127.0.0.1:8000</a>.

Vous obtiendrez cette réponse JSON :