Why isn’t there a Linux installation script available?

[changjingzhi]

I noticed the installation script is only for Windows (ai-hub\1.Install_QAI_AppBuilder.bat). Is there a reason why no Linux script is provided? Are there any plans to add Linux support?

Thanks!

[quic-zhanweiw]

Thanks @changjingzhi for using QAI AppBuilder.
We're preparing Linux/Ubuntu version document and Python wheel, we will release the first version soon.

[quic-zhanweiw]

@changjingzhi
We've released some wheels based on QAIRT 2.37 for Linux/Ubuntu.
https://github.com/quic/ai-engine-direct-helper/releases/tag/v2.37.0

Here are some samples and build steps:
https://github.com/quic/ai-engine-direct-helper/tree/main/samples/linux

[janjongboom]

FYI #16 builds on Linux out of the box (tested with libraries in /usr/lib and Genie headers in /usr/include/Genie):

Install Genie (QNN libs already on this system):

# Grab Qualcomm® AI Runtime SDK - Community Edition (https://softwarecenter.qualcomm.com/catalog/item/Qualcomm_AI_Runtime_Community?osArch=Any&osType=All&version=2.35.0.250530)
wget https://apigwx-aws.qualcomm.com/qsc/public/v1/api/download/software/sdks/Qualcomm_AI_Runtime_Community/All/2.35.0.250530/v2.35.0.250530.zip
unzip -q v2.35.0.250530.zip
rm v2.35.0.250530.zip
cd qairt/2.35.0.250530/

# Link in library files / headers to /usr/lib/
sudo ln -s $PWD/lib/aarch64-oe-linux-gcc11.2/libGenie.so /usr/lib/libGenie.so
sudo ln -s $PWD/include/Genie/ /usr/include/Genie

# Add the genie files to your PATH
if grep -q 'export QNN_SDK_ROOT' ~/.bash_profile; then
    echo "QNN_SDK_ROOT already set in ~/.bash_profile, not overwriting..."
else
    echo "" >> ~/.bash_profile
    echo "# Begin AI Engine Direct SDK" >> ~/.bash_profile
    echo "export QNN_SDK_ROOT=/usr" >> ~/.bash_profile
    echo "export QAIRT_SDK_ROOT=/usr" >> ~/.bash_profile
    echo "export PATH=\$PATH:$PWD/bin/aarch64-oe-linux-gcc11.2/" >> ~/.bash_profile
    echo "# End AI Engine Direct SDK" >> ~/.bash_profile
    echo "" >> ~/.bash_profile
fi

# To use Genie in your current session
source ~/.bash_profile

Build app:

# Clone the repository (can switch back to upstream once https://github.com/quic/ai-engine-direct-helper/pull/16 is landed)
git clone https://github.com/edgeimpulse/ai-engine-direct-helper
git checkout linux-paths

# Create a new venv
python3 -m venv .venv
source .venv/bin/activate

# Build the wheel
pip3 install setuptools
python setup.py bdist_wheel
pip3 install ./dist/qai_appbuilder-*-linux_aarch64.whl

[quic-zhanweiw]

Thanks @janjongboom for sharing the detailed steps.

[Barleysack]

@janjongboom Good day, Is there any plan to support ai engine direct appbuilder for linux_x86_64?
Well currently linux support is only for amd64, arm64, and windows, but I think there should be need for linux_x86...

[quic-zhanweiw]

@Barleysack
Thank you for your interest in the QAI AppBuilder tool.
This tool is for running QNN model on QC NPU. I think there is no x86 device which has QC NPU, right? May you share the detailed use case?

[ZIFENG278]

You can find the Linux installation guide by this link: https://docs.radxa.com/en/dragon/q6a/app-dev/npu-dev/qai-appbuilder

[quic-zhanweiw]

Yes. This is QC Linux platform. And it has QC NPU in the platform.