# Setup Qualcomm AI Engine Direct

Source: [https://docs.qualcomm.com/doc/80-70015-15B/topic/qnn-setup.html](https://docs.qualcomm.com/doc/80-70015-15B/topic/qnn-setup.html)

## Setup SDK

To use the AI Engine Direct SDK, the following Ubuntu OS and Python versions are
            required:

- Ubuntu 22.04
- Python 3.10

Refer to [AI Engine Direct SDK Setup](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/setup.html) for detailed setup
            instructions.

Note: `${QNN_SDK_ROOT}`refers to the installed AI Engine Direct SDK root
                path.
- If downloaded via the direct download method, set
                            `${QNN_SDK_ROOT}` to the downloaded and unzipped SDK root
                        directory.
- If installed via QPM, the root directory path is
                            /opt/qcom/aistack/qairt/<version>. This
                        example uses version 2.26.2.240911, so the path is:
                            /opt/qcom/aistack/qairt/2.26.2.240911.

SDK setup steps are:

- Install [Linux platform dependencies](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/setup.html#linux-platform-dependencies)
- Install Python dependencies with
                    `${QNN_SDK_ROOT}/bin/check-python-dependency`
- Install Ubuntu packages with `sudo bash
                        ${QNN_SDK_ROOT}/bin/check-linux-dependency.sh`
- Install ML frameworks that are required by your models (PyTorch, TensorFlow, ONNX).
                Refer to Frameworks for supported versions. 

| Framework | Version |
    | --- | --- |
    | **TensorFlow** | v2.10.1 |
    | **TFLite** | v2.3.0 |
    | **PyTorch** | v1.13.1 |
    | **ONNX** | v1.12.0 |

        # Install tensorflow
        pip install tensorflow==2.10.1
                            
        # Install tflite
        pip install tflite==2.3.0
                            
        # Install PyTorch
        pip install torch==1.13.1+cpu torchvision==0.14.1+cpu torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cpu
                            
        # Install Onnx
        pip install onnx==1.12.0 onnxruntime==1.17.1 onnxsim==0.4.36Copy to clipboard

## Setup environment

Source the environment setup script provided by QNN SDK to ensure all necessary tools
                and libraries are available in `$PATH `for the workflow.

    source ${QNN_SDK_ROOT}/bin/envsetup.shCopy to clipboard

Last Published: Jan 21, 2026

[Previous Topic
Install AI Engine Direct](https://docs.qualcomm.com/bundle/publicresource/80-70015-15B/topics/qnn-download.md) [Next Topic
QNN model porting](https://docs.qualcomm.com/bundle/publicresource/80-70015-15B/topics/qnn-port-model.md)