QAIRT SDK Installation

The QAIRT (Qualcomm® AI Runtime) SDK is a software package that integrates Qualcomm® AI software products, including Qualcomm® AI Engine Direct, Qualcomm® Neural Processing SDK, and Qualcomm® Genie. QAIRT provides developers with all the necessary tools for porting and deploying AI models on Qualcomm® hardware accelerators, along with the runtime for running models on CPU, GPU, and NPU.

Currently, the QAIRT SDK supports model porting for Pytorch, TensorFlow, TFLite, and ONNX frameworks.

Supported inference hardware accelerators: CPU, GPU, NPU.

QAIRT SDK Architecture

QAIRT SDK System Requirements

• 64-bit x86 processor
• Ubuntu 22.04
• Python 3.10

Download QAIRT SDK Package

info

Please use QAIRT SDK version 2.42.0.251225

X86 Linux PC

export QAIRT_VERSION=2.42.0.251225

• Download the 2.42.0.251225 QAIRT SDK from Qualcomm® Software Center

  X86 Linux PC / Qualcomm Device

  wget https://softwarecenter.qualcomm.com/api/download/software/sdks/Qualcomm_AI_Runtime_Community/All/${QAIRT_VERSION}/v${QAIRT_VERSION}.zip

• Extract the QAIRT SDK package

  X86 Linux PC / Qualcomm Device

  unzip v${QAIRT_VERSION}.zip
  cd qairt/${QAIRT_VERSION}

Configure QAIRT Environment

Create Python Environment

QAIRT requires a Python 3.10 environment. Users can create one using Anaconda.

tip

• For Anaconda installation, refer to: Conda Installation

• For creating a conda Python environment, refer to: Creating a Specific Python Version Environment

After installing Anaconda, create and activate a Python 3.10 environment using the terminal:

X86 Linux PC

conda create -n qairt python=3.10
conda activate qairt

Configure QAIRT Environment Variables

X86 Linux PC

source bin/envsetup.sh

Check Dependencies

• Check library dependencies

  X86 Linux PC

  sudo ${QAIRT_SDK_ROOT}/bin/check-linux-dependency.sh

  Checking for libncurses5
  Checking for libgl1
  Checking for libgtk-3-dev
  Checking for libasound2-dev
  Checking for libnss3
  Checking for git
  Checking for nodejs
  Checking for npm
  Checking for libgbm-dev
  Checking for desktop-file-utils
  Checking for python3-dev
  Checking for python3-pip
  Checking for python3-setuptools
  Checking for clang
  Checking for libc++-dev
  Checking for libc++abi-dev
  Checking for flatbuffers-compiler
  Checking for libflatbuffers-dev
  Checking for rename
  =============================================================
  All Dependency Packages Found
  Done!!

• Check Toolchain Dependencies

  X86 Linux PC

  ${QAIRT_SDK_ROOT}/bin/envcheck -c

  Checking Clang Environment
  --------------------------------------------------------------
  [INFO] Found clang++ at /usr/bin/clang++
  --------------------------------------------------------------

• Install Python Library Dependencies

  X86 Linux PC

  python3 "${QAIRT_SDK_ROOT}/bin/check-python-dependency"

Install AI Frameworks

Select the model framework libraries to install based on your business model format. It's not necessary to install all of them.

Package Name	Version
TensorFlow	2.10.1
TFLite	2.3.0
PyTorch	1.13.1
ONNX	1.16.1
ONNX Runtime	1.17.1
ONNX Simplifier	0.4.36

Complete SDK Documentation

The QAIRT SDK integrates Qualcomm® AI Engine Direct (QNN), Qualcomm® Neural Processing SDK (SNPE), and Qualcomm® Genie (Genie). The detailed documentation is available at the following locations:

Local Documentation

• qairt/2.42.0.251225/docs/QNN

• qairt/2.42.0.251225/docs/SNPE

• qairt/2.42.0.251225/docs/Genie

Online Documentation

• QNN

• SNPE

• Genie

ARM64 Docker Image

Radxa provides a Docker image for using the QAIRT SDK on the board. Alternatively, you can follow the instructions in this document to install it manually.

• Install Docker

  tip

  Refer to Docker Usage for Docker installation instructions

• Pull the Docker image

  QCS6490

  Device

  sudo docker pull radxazifeng278/qairt-npu-v68:v1.2

  • Run the Docker image

  Device

  sudo docker run --privileged -it -v /dev:/dev -v $(pwd):/workspace --name qairt-sdk radxazifeng278/qairt-npu-v68:v1.2 /bin/bash

  SC8280XP

  Device

  sudo docker pull radxazifeng278/qairt-npu-v68:v1.0

  • 运行 docker 镜像

  Device

  sudo docker run --privileged -it -v /dev:/dev -v $(pwd):/workspace --name qairt-sdk radxazifeng278/qairt-npu-v68:v1.0 /bin/bash

  QCS9075

  Device

  sudo docker pull radxazifeng278/qairt-npu-9075:v1.1

  • Run the Docker image

  Device

  sudo docker run --privileged -it -v /dev:/dev -v $(pwd):/workspace --name qairt-sdk radxazifeng278/qairt-npu-9075:v1.1 /bin/bash

  Note: In the radxazifeng278/qairt-npu-9075:v1.1 image, the QAIRT directory is located at /root/qairt. If you are unsure after entering the container, confirm it with:

  find / -name "qairt"