You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Added changelog and known issues placeholder
Applied changes based on the customer's feedback.
Quick start and setup sections have been merged
Signed-off-by: Anna Wojdylo <anna.wojdylo@nordicsemi.no>
Copy file name to clipboardExpand all lines: doc/quick_start.rst
+10-2Lines changed: 10 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,4 +1,5 @@
1
1
.. _quick_start:
2
+
.. _setting_up_environment:
2
3
3
4
Quick start guide
4
5
#################
@@ -7,15 +8,22 @@ Quick start guide
7
8
:local:
8
9
:depth: 2
9
10
10
-
11
-
This section includes quick start guides for running machine learning workloads on Nordic Semiconductor devices using nRF Edge AI.
11
+
This section includes quick start guides for setting up your environment and running machine learning workloads on Nordic Semiconductor devices using nRF Edge AI.
12
12
These pages cover common development paths, from simplified end‑to‑end solutions to lower‑level integrations with direct hardware control.
13
13
14
14
Select the guide that best matches your required level of control, tooling preferences, and target hardware.
15
15
Each guide walks through the required setup steps and shows how to deploy and run an edge AI application.
16
16
17
17
If you are unsure which approach fits your use case, see the :ref:`solution_comparison` page for a comparison of features, performance characteristics, and supported workflows.
18
18
19
+
Before you follow any of the guides in this section, complete :ref:`setup_sdk` to install |NCS| and the development toolchain.
The following guide explains how to use the Axon NPU driver API to run TensorFlow Lite models directly on the Axon NPU.
11
-
It is ideal if you require low‑level control over inference execution, memory usage, and system integration.
11
+
The following guide outlines the requirements that you need to fulfill before you start working with the Axon NPU using Axon API directly, without the abstraction layer of |EAILib| API.
12
+
It also explains how to use the Axon NPU driver API to run TensorFlow Lite models directly on the Axon NPU.
13
+
This solution is ideal if you require low‑level control over inference execution, memory usage, and system integration.
12
14
13
15
To follow this guide, you should be familiar with embedded systems development and C‑based APIs.
14
16
Compared to higher‑level frameworks such as |EAILib|, using the driver API requires more manual setup but enables finer control over performance and resource utilization.
15
17
16
18
After completing this guide, you will have compiled a TensorFlow Lite model for the Axon NPU and deployed a custom application that performs inference using the Axon driver API.
17
19
18
-
.. _quick_start_axon_driver_model_compilation:
20
+
.. _axon_requirements_hardware:
19
21
20
-
Model compilation
21
-
*****************
22
+
Hardware requirements
23
+
*********************
22
24
23
-
Before you can deploy to the Axon NPU, you need a TensorFlow Lite model compiled specifically for Axon hardware.
24
-
The :ref:`axon_npu_tflite_compiler` transforms your TFLite model into an optimized format that leverages the NPU's specialized hardware.
25
+
Axon NPU library is included as part of the :ref:`nrf_edgeai_lib` and provided as compiled binaries for Cortex-M33F architectures.
26
+
Axon NPU is currently available on the `nRF54LM20B`_ device.
25
27
26
-
.. rst-class:: numbered-step
28
+
.. _axon_requirements_software:
27
29
28
-
Set up the compiler
29
-
===================
30
+
Software requirements
31
+
*********************
32
+
33
+
To start working with the Axon NPU, complete the setup based on your use case:
34
+
35
+
* If you want to deploy models on the device, you just need to complete :ref:`setup_sdk` to install |NCS|, toolchain, and |EAILib|.
36
+
* If you want to prepare models for deployment, you only need to set up a Python environment to run the :ref:`axon_npu_tflite_compiler`.
37
+
Follow instructions in :ref:`axon_setup_compiler` to set up the environment.
38
+
39
+
.. _axon_setup_compiler:
40
+
41
+
Setting up Axon TFlite Compiler
42
+
===============================
43
+
44
+
Before you can run the :ref:`axon_npu_tflite_compiler`, you need to set up a Python environment with the required dependencies.
45
+
The executor of the compiler is compatible with Python ``3.11``.
46
+
47
+
You can set up the Python environment using one of the methods below.
48
+
49
+
.. tabs::
50
+
51
+
.. group-tab:: Python virtual environment
52
+
53
+
Using a virtual environment is strongly recommended to isolate dependencies.
54
+
You can use any virtual environment tool you prefer.
55
+
This section shows one example using Miniforge (Conda):
56
+
57
+
1. Install `Miniforge`_
58
+
59
+
#. Ensure the Conda :file:`scripts` directory is added to your system ``PATH``, for example, :file:`C:/Users/<user>/AppData/Local/miniforge3/Scripts`.
60
+
61
+
#. Create a new environment with the supported Python version:
62
+
63
+
.. code-block:: console
64
+
65
+
conda create -n <env_name> python=3.11
66
+
67
+
#. Activate the environment.
68
+
All installation and execution commands must be run from the activated environment:
69
+
70
+
.. code-block:: console
71
+
72
+
conda activate <env_name>
73
+
74
+
#. Install the required Python packages using the :file:`requirements.txt` file:
75
+
76
+
.. code-block:: console
77
+
78
+
cd tools/axon/compiler/scripts
79
+
pip install -r requirements.txt
80
+
81
+
.. note::
82
+
83
+
On macOS, you may encounter the ``ERROR: No matching distribution found for tensorflow==2.15.1`` error.
84
+
To fix it, install TensorFlow using Conda instead:
85
+
86
+
.. code-block:: console
87
+
88
+
conda install -c conda-forge tensorflow=2.15.1
89
+
90
+
Once you complete the setup, you can try running the compiler by following instructions in :ref:`axon_npu_tflite_compiler`.
91
+
92
+
.. group-tab:: Docker
93
+
94
+
Docker provides a fully isolated way to run the compiler without installing dependencies locally.
95
+
96
+
Before using Docker with the compiler, you must install it on the system.
97
+
You should also verify that Docker is working correctly by building and running a simple Docker container.
98
+
99
+
The following links provide introductory material and best-practice guidance for Docker:
100
+
101
+
* `A beginner's guide to Docker`_
102
+
* `Creating the Dockerfile`_
103
+
* `Intro Guide to Dockerfile Best Practices`_
104
+
105
+
Once you have installed and verified Docker, you can follow :ref:`axon_npu_tflite_compiler_docker` to build and run a Docker container for the Python compiler executor.
106
+
107
+
.. group-tab:: Podman
108
+
109
+
Podman is a daemonless alternative to Docker.
110
+
Follow the steps below to set up Podman and run the compiler in a Podman container:
111
+
112
+
1. Install Podman by following the `Podman installation guide`_.
113
+
#. Set up and run a `simple container with Podman <Setting up Podman container_>`_
114
+
115
+
Once you have installed and verified Podman, you can follow :ref:`axon_npu_tflite_compiler_podman` to build and run a Podman container for the Python compiler executor.
116
+
117
+
.. _quick_start_axon_driver_model_compilation:
118
+
119
+
Model compilation
120
+
*****************
30
121
31
-
First, set up the Axon compiler's environment on your development system.
32
-
Follow the :ref:`Executor and compiler setup <axon_setup_compiler>` instructions to install the necessary tools.
122
+
With the compiler environment ready, you can transform your TensorFlow Lite model into Axon-optimized code and verify that inference produces correct results.
0 commit comments