.. _sample_hello_axon:

Hello Axon
##########

.. contents::
   :local:
   :depth: 2

The Hello Axon sample demonstrates how to run neural model inference on the Axon NPU using the Axon NPU driver directly.
Use this sample if you are working directly with the Axon NPU compiler.

Requirements
************

The sample supports the following development kits:

.. table-from-sample-yaml::

Overview
********

The regression model runs on the Axon NPU and supports both synchronous and asynchronous inference modes.

This regression model uses Zephyr's `TensorFlow Lite for Microcontrollers: Hello World`_ sample.
The model's task is to replicate the sine function in the range from 0 to 2π.
The TensorFlow Lite file describing this model is processed by the Axon NPU Compiler to convert it into a format accepted by the Axon NPU.
The compilation output is saved in :file:`src/generated/nrf_axon_model_hello_axon_.h`.

Configuration
*************

|config|

Configuration options
=====================

|sample_kconfig|

.. options-from-kconfig::
   :show-type:

Building and running
********************

.. |sample path| replace:: :file:`samples/hello_axon`

.. include:: /includes/build_and_run.txt

Testing
=======

|test_sample|

#. |connect_kit|
#. |connect_terminal_kit|
#. Reset the development kit.
#. Observe the logging output.

Sample output
-------------

The following output is logged in the terminal:

   .. code-block:: console

      I: Hello Axon sample
      I: Initializing Axon NPU
      I: Running asynchronous inference
      I: prediction:  0.051,  ideal  0.072
      I: prediction:  0.847,  ideal  0.842
      I: prediction: -0.491,  ideal -0.500
      ...

Dependencies
************

This sample uses the following |EAI| libraries:

* :ref:`Axon NPU driver <axon_driver>`

It uses the following Zephyr libraries:

* `Logging`_