Install and use PyTorch and torchvision

PyTorch is one of the most popular and easy-to-use deep learning frameworks in Python. It allows developers to design and train complex neural network models as intuitively and flexibly as writing ordinary Python code. Its clean API design and powerful GPU acceleration support make the development process — from research ideas to actual deployment — extremely efficient and convenient, which is why it is widely favored by developers.

NVIDIA provides packages specifically adapted for the Jetson series of devices. Their version dependencies are as follows:

PyTorch Version	NVIDIA Framework Container	NVIDIA Framework Wheel	JetPack Version
2.8.0a0+5228986c39	25.06	-	6.2
2.8.0a0+5228986c39	25.05	-	6.2
2.7.0a0+79aa17489c	25.04	-	6.2
2.7.0a0+7c8ec84dab	25.03	-	6.2
2.7.0a0+6c54963f75	25.02	-	6.2
2.6.0a0+ecf3bae40a	25.01	-	6.1
2.6.0a0+df5bbc09d1	24.12	-	6.1
2.6.0a0+df5bbc0	24.11	-	6.1
2.5.0a0+e000cf0ad9	24.10	-	6.1
2.5.0a0+b465a5843b	24.09	24.09	6.1
2.5.0a0+872d972e41	24.08	-	6.0
2.4.0a0+3bcc3cddb5	24.07	24.07	6.0
2.4.0a0+f70bd71a48	24.06	24.06	6.0
2.4.0a0+07cecf4168	24.05	24.05	6.0
2.3.0a0+6ddf5cf85e	24.04	24.04	6.0 Developer Preview
2.3.0a0+40ec155e58	24.03	24.03	6.0 Developer Preview
2.3.0a0+ebedce2	24.02	24.02	6.0 Developer Preview
2.2.0a0+81ea7a4	23.12, 24.01	23.12, 24.01	6.0 Developer Preview
2.2.0a0+6a974bec	23.11	23.11	6.0 Developer Preview
2.1.0a		23.06	5.1.x
2.0.0		23.05	5.1.x
2.0.0a0+fe05266f		23.04	5.1.x
2.0.0a0+8aa34602		23.03	5.1.x
1.14.0a0+44dac51c		23.02, 23.01	5.1.x
1.13.0a0+936e930		22.11	5.0.2
1.13.0a0+d0d6b1f		22.09, 22.10	5.0.2
1.13.0a0+08820cb	22.07	22.07	5.0.2
1.13.0a0+340c412	22.06	22.06	5.0.1
1.12.0a0+8a1a93a9	22.05	22.05	5.0
1.12.0a0+bd13bc66		22.04	5.0
1.12.0a0+2c916ef		22.03	5.0
1.11.0a0+bfe5ad28		22.01	4.6.1

The tutorial below uses JetPack 6.2.1 with CUDA 12.6 as an example

1. Install the torch packages

1.1 Download and install torch and torchvision

wget https://pypi.jetson-ai-lab.io/jp6/cu126/+f/62a/1beee9f2f1470/torch-2.8.0-cp310-cp310-linux_aarch64.whl 
wget https://pypi.jetson-ai-lab.io/jp6/cu126/+f/907/c4c1933789645/torchvision-0.23.0-cp310-cp310-linux_aarch64.whl
pip install torch-2.8.0-cp310-cp310-linux_aarch64.whl torchvision-0.23.0-cp310-cp310-linux_aarch64.whl -i https://pypi.tuna.tsinghua.edu.cn/simple

1.2 Check whether the installation is correct

Execute the following three statements with python

jetson@jetson-desktop:~$ python
Python 3.10.16 (main, Dec 11 2024, 16:18:56) [GCC 11.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> print(torch.__version__)
2.8.0
>>> print(torch.cuda.is_available())
True

Example from a JetPack 7.2.1 system, including a minimal CUDA tensor operation in addition to the version checks:

2. Run YOLO11

YOLO is a real-time object detection algorithm. It treats object detection as a single-stage regression problem by dividing an image into a grid and directly predicting bounding boxes and class probabilities, achieving high-speed and high-accuracy detection. Thanks to being open source, easy to use, and flexible to deploy, the YOLO family is widely used in fields such as autonomous driving, security surveillance, and industrial quality inspection.

2.1 Install miniconda

curl -L https://repo.anaconda.com/miniconda/Miniconda3-py310_25.3.1-1-Linux-aarch64.sh | bash
source ~/.bashrc
conda --version

2.2 Switch conda to a mirror source

conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/msys2/
conda config --set show_channel_urls yes

2.3 Create a conda environment

conda create -n jetson-ai python=3.10

2.4 Enter the conda environment

conda activate jetson-ai

2.5 Install torch and torchvision

wget https://pypi.jetson-ai-lab.io/jp6/cu126/+f/62a/1beee9f2f1470/torch-2.8.0-cp310-cp310-linux_aarch64.whl 
wget https://pypi.jetson-ai-lab.io/jp6/cu126/+f/907/c4c1933789645/torchvision-0.23.0-cp310-cp310-linux_aarch64.whl
pip install torch-2.8.0-cp310-cp310-linux_aarch64.whl torchvision-0.23.0-cp310-cp310-linux_aarch64.whl -i https://pypi.tuna.tsinghua.edu.cn/simple

2.6 Install ultralytics

pip install ultralytics -i https://pypi.tuna.tsinghua.edu.cn/simple

2.7 Run the camera video inference example

Connect a camera and run the following program in the environment created above.

import cv2
import time
from ultralytics import YOLO
from ultralytics import YOLOWorld

# Load the YOLO model
model = YOLO("yolo11s.pt")

# Open the video file
video_path = 0
cap = cv2.VideoCapture(video_path)

# Loop through the video frames
while cap.isOpened():
    
    # Read a frame from the video
    success, frame = cap.read()
    start = time.time()
    if success:
        # Run YOLO inference on the frame
        results = model(frame)
        inf_time = time.time() - start
        # Visualize the results on the frame
        annotated_frame = results[0].plot()      
        fps = 1.0 / inf_time if inf_time > 0 else 0
        # show FPS
        cv2.putText(annotated_frame, f"FPS: {fps:.2f}", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0,255,0), 2)
        cv2.imshow("YOLO Inference", annotated_frame)

        # Break the loop if 'q' is pressed
        if cv2.waitKey(1) & 0xFF == ord("q"):
            break
    else:
        # Break the loop if the end of the video is reached
        break

For more information, see Ultralytics YOLO11 - Ultralytics YOLO Docs