PyTorch Environment Setup

PyTorch is an open-source deep learning framework developed by Meta (formerly Facebook), renowned for its flexibility and ease of use, and widely used in academic research and industrial applications.

1. System Information

2. Installing Dependencies

Installing nvpl

Download the installation file

xxxxxxxxxx
cd ~/Downloads wget https://developer.download.nvidia.com/compute/nvpl/25.11/local_installers/nvpl-local-repo-ubuntu2404-25.11_1.0-1_arm64.deb

Install the deb package and obtain keyrings.

xxxxxxxxxx
sudo dpkg -i nvpl-local-repo-ubuntu2404-25.11_1.0-1_arm64.deb

Following the prompts, run the command to copy keyrings and install nvpl.

xxxxxxxxxx
sudo cp /var/nvpl-local-repo-ubuntu2404-25.11/nvpl-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install nvpl

Install cudss

Download the cudss file package for CUDA 13. Here, we take version v0.7.1.4_cuda13 as an example.

xxxxxxxxxx
wget https://developer.download.nvidia.cn/compute/cudss/redist/libcudss/linux-aarch64/libcudss-linux-aarch64-0.7.1.4_cuda13-archive.tar.xz

After decompression, place the file in the /opt/nvidia/cudss directory.

xxxxxxxxxx
tar xvf libcudss-linux-aarch64-0.7.1.4_cuda13-archive.tar.xz
sudo mv libcudss-linux-aarch64-0.7.1.4_cuda13-archive /opt/nvidia/cudss

Note: The extracted directory is named libcudss-linux-aarch64-0.7.1.4_cuda13-archive; to simplify the file path, we have renamed it to cudss. The contents of the cudss directory are shown in the figure below:

Add environment variables to .bashrc:

xxxxxxxxxx
vim ~/.bashrc

Then, append the following lines to the end of the file, save, and exit.

xxxxxxxxxx
export CUDSS_DIR=/opt/nvidia/cudss
export LD_LIBRARY_PATH=$CUDSS_DIR/lib:$LD_LIBRARY_PATH

3. Installing PyTorch

You may choose either the offline or the online method to install PyTorch.

3.1. Offline Installation

Open a web browser, navigate manually to the download URL, and download the PyTorch .whl installation file to your Downloads directory.

Download URL: https://pypi.jetson-ai-lab.io/sbsa/cu130/+f/631/7b71e212e5966/torchaudio-2.10.0-cp312-cp312-linux_aarch64.whl#sha256=6317b71e212e59667eff088ba1d0631dc58b37a6fc28179c319bf359512e50f1

xxxxxxxxxx
cd ~/Downloads/

xxxxxxxxxx
sudo pip3 install torchaudio-2.10.0-cp312-cp312-linux_aarch64.whl --break-system-packages

3.2 Online Installation

xxxxxxxxxx
sudo pip3 install torch --break-system-packages --index-url https://pypi.jetson-ai-lab.io/sbsa/cu130

4. Verify Installation

xxxxxxxxxx
python3 -c "import torch; print(f'Torch: {torch.__version__}')"