Skip to content

getting-started.md

Latest commit

 

History

History
235 lines (168 loc) · 9.13 KB

getting-started.md

File metadata and controls

235 lines (168 loc) · 9.13 KB
Raw

Getting Started with Intel® Edge Developer Kits

Welcome to Intel®'s edge computing ecosystem! This guide will help you get up and running quickly, regardless of your experience level.

Table of Contents

Installation Walkthrough

Step 1: Prepare Your System

Required:

  • Ubuntu* 24.04 LTS Desktop (Download)
  • Internet connection
  • Administrator (sudo) access

Recommended:

Step 2: Enable Hardware Features

For Intel® Arc™ Graphics:

  1. Access BIOS/UEFI settings
  2. Enable "Resizable BAR" or "Above 4G Decoding"
  3. Save and reboot

Why? This significantly improves GPU performance with large AI models.

Step 3: Run the Installer

Option A: One-Line Install (Recommended)

sudo bash -c "$(wget -qLO - https://raw.githubusercontent.com/open-edge-platform/edge-developer-kit-reference-scripts/refs/heads/main/main_installer.sh)"

Option B: Step-by-Step Install

# Download the repository
git clone https://github.com/open-edge-platform/edge-developer-kit-reference-scripts.git
cd edge-developer-kit-reference-scripts

# Run main installer
sudo ./main_installer.sh

What the installer sets up:

  • Kernel and HWE stack verification
  • Intel® GPU drivers, plus NPU drivers on Core Ultra platforms
  • OpenVINO™ runtime

Optional components

Docker, qmassa and DL Streamer are available but are not installed by default. When the installer runs on a terminal it asks which, if any, you want:

  • docker Docker Engine, with the Compose and Buildx plugins
  • qmassa Terminal Intel GPU usage monitor
  • dlstreamer Intel® DL Streamer, GStreamer elements for media analytics

Selecting nothing is a valid answer, and the installation continues without them.

To decide up front and skip the question, which is what scripted and CI runs need:

sudo ./main_installer.sh --with-extras=docker
sudo ./main_installer.sh --with-extras=docker,dlstreamer
sudo ./main_installer.sh --with-extras=all
sudo ./main_installer.sh --no-extras

A run with no terminal attached installs nothing unless --with-extras names a component, so unattended installs are unaffected.

Optional components are installed after the drivers, so a problem with any of them cannot block the driver setup. Anything you skip is listed at the end of the run, and you can add it later by running the installer again with the flag.

If you install Docker, your user is added to the docker group, so log out and back in, or run newgrp docker, before using docker without sudo. If HTTP_PROXY and HTTPS_PROXY are set in your environment, the installer configures Docker to use them for both the daemon and the CLI.

DL Streamer can also be installed on its own after setup, and DLSTREAMER_VERSION pins the release in either case:

sudo DLSTREAMER_VERSION=2026.1.0 ./main_installer.sh --with-extras=dlstreamer
sudo DLSTREAMER_VERSION=2026.1.0 ./dlstreamer_installer.sh

Step 4: Handle Reboot (if prompted)

Some driver installations require a reboot. If prompted:

  1. Reboot your system
  2. Return to the installation directory
  3. Run the post-reboot commands:
sudo ./openvino_installer.sh
sudo ./print_summary_table.sh

Step 5: Verify Installation

You should see output like this:

Installation is completed when you see this message:

========================================================================
Running Installation Summary

==================== System Installation Summary ====================
Item                      | Value
------------------------ -+-----------------------------------------
Kernel Version            | 6.14.0-27-generic
HWE Stack                 | Installed
Ubuntu Version            | Ubuntu 24.04.3 LTS
NPU Status                | Detected
NPU Package               | intel-level-zero-npu
NPU Version               | 1.19.0.20250707-16111289554
intel-driver-compiler-npu | 1.19.0.20250707-16111289554
intel-fw-npu              | 1.19.0.20250707-16111289554
intel-level-zero-npu      | 1.19.0.20250707-16111289554
level-zero                | 1.22.4
GPU Type                  | Intel
GPU Count                 | 4 Intel graphics device(s) detected
GPU Driver                | i915 (loaded)
GPU Device 1              | 00:02.0 VGA compatible controller: Intel Corporation Arrow Lake-U [Intel Graphics] (rev 06)
GPU Device 2              | 03:00.0 VGA compatible controller: Intel Corporation Device e20b
GPU Device 3              | 08:00.0 VGA compatible controller: Intel Corporation Device e20b
GPU Device 4              | 80:14.5 Non-VGA unclassified device: Intel Corporation Device 7f2f (rev 10)
------------------------ -+-----------------------------------------
Intel Graphics Packages   |
------------------------ -+-----------------------------------------
i965-va-driver:amd64      | 2.4.1+dfsg1-1build2
intel-gsc                 | 0.9.5-0ubuntu1~24.04~ppa1
intel-media-va-driver-non-free:amd64 | 25.3.1-0ubuntu1~24.04~ppa1
intel-opencl-icd          | 25.27.34303.9-1~24.04~ppa1
libegl-mesa0:amd64        | 25.0.7-0ubuntu0.24.04.1
------------------------ -+-----------------------------------------
Platform Status           | [✓] Platform is configured
=====================================================================

========================================================================
Installation completed: 2025-08-11 10:11:54
Log file saved: /var/log/intel-platform-installer.log
========================================================================

🎉 Congratulations! You're now ready to build amazing AI applications with Intel® hardware.

Unattended Installation

When the kernel needs upgrading, the default flow stops and asks you to reboot and run the installer again. DEVKIT_AUTO_INSTALL=1 does it in one pass instead:

sudo DEVKIT_AUTO_INSTALL=1 ./main_installer.sh

This restarts your system after a ten second countdown. Press Ctrl+C to cancel.

Driver packages are user-space and install fine on the old kernel, so only verification is deferred. It runs automatically on the next boot:

sudo tail -40 /var/log/intel-platform-installer.log

In the default flow, the installer instead ends with a summary of what needs a restart and why, shown only when one is actually required.

Optional: Install DL Streamer

Intel® Deep Learning Streamer adds GStreamer elements such as gvadetect, gvaclassify and gvawatermark, letting you build media analytics pipelines on CPU, GPU and NPU.

It is not part of the base install. Run it after the main installer has completed, since it relies on the GPU and NPU drivers already being in place:

sudo ./dlstreamer_installer.sh

To install a specific release instead of the latest:

sudo DLSTREAMER_VERSION=2026.1.0 ./dlstreamer_installer.sh

List what is available with apt show -a intel-dlstreamer.

About the environment variables

DL Streamer needs a set of GStreamer variables that the upstream guide asks each user to paste into their own ~/.bashrc. This installer instead writes /etc/profile.d/intel-dlstreamer.sh, which sources the environment script shipped by the package. The variables are then set for every user at login, with nothing to edit per user and no paths duplicated.

It applies at your next login. To use it in the shell you already have open:

source /etc/profile.d/intel-dlstreamer.sh

Verify

gst-inspect-1.0 gvadetect

You should see the element documentation rather than No such element. If you do not, log out and back in so the environment is applied.

Run a sample pipeline

Download a model, then run a detection pipeline against it:

/opt/intel/dlstreamer/samples/download_public_models.sh yolo11s

export MODEL=$MODELS_PATH/public/yolo11s/FP16/yolo11s.xml
gst-launch-1.0 videotestsrc num-buffers=100 ! video/x-raw,width=640,height=640 ! \
  videoconvert ! gvadetect model=$MODEL device=CPU ! gvafpscounter ! fakesink

MODELS_PATH defaults to ~/models and is set for you by the profile script. A frames-per-second figure followed by Got EOS means inference is working. Swap device=CPU for device=GPU to exercise the GPU path.

Note: Intel also ships /opt/intel/dlstreamer/scripts/hello_dlstreamer.sh. On systems with more than one GPU it fails with [: too many arguments, because its GPU detection does not quote the result of find /dev/dri/ -name "render*". The pipeline above avoids it.

Note: intel-dlstreamer depends on a specific OpenVINO™ version. If openvino_installer.sh has already installed a different one, the installer warns you and prints the commands to remove the conflicting packages.

Next Steps

Explore all samples — browse everything available here, share your projects and help others learn!