Installing PyTorch for NVIDIA Jetson Orin Nano Developer Kit

A Simple Guide to Installing PyTorch on Jetson Orin Nano with latest JetPack version

Installing PyTorch on the NVIDIA Jetson platform can be a bit tricky, and you’d think the official docs would make it easier, right? While NVIDIA does provide instructions, they sometimes fall behind when it comes to updates for the latest JetPack OS or the newer Jetson Orin Nano. So, I decided to create this guide to help you get PyTorch up and running smoothly — no guesswork required!

If installing 24.06 PyTorch or later versions, cusparselt needs to be installed first. Check here to find out the compatible version.

1. Install cuSPARSELT

1.1. Create install_cusparselt.sh bash file with following commands.

#!/bin/bash

set -ex

# cuSPARSELt license: https://docs.nvidia.com/cuda/cusparselt/license.html
mkdir tmp_cusparselt && cd tmp_cusparselt

CUSPARSELT_NAME="libcusparse_lt-linux-aarch64-0.6.3.2-archive"
curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-aarch64/${CUSPARSELT_NAME}.tar.xz
tar xf ${CUSPARSELT_NAME}.tar.xz
cp -a ${CUSPARSELT_NAME}/include/* /usr/local/cuda/include/
cp -a ${CUSPARSELT_NAME}/lib/* /usr/local/cuda/lib64/…