Install pytorch1.7 with python 3.8

[Argo_sa]

Hey everyone! I have a Jetson Nano developer board:

 - Jetpack:        UNKNOWN [L4T 32.7.4]                                                   
 - Board:                                                                                 
   * Type:           Nano (Developer Kit Version)
   * SOC Family:     tegra210     ID: 33
   * Module:         P3448-0000   Board: P3449-0000
   * Code Name:      porg
   * Cuda ARCH:      5.3
   * Serial Number:  1420121016413
   * Board ids:      3448
 - Libraries:                                                                         
   * CUDA:         10.2.300                                                           
   * OpenCV:       4.6.0  compiled CUDA: YES                                            
   * TensorRT:     8.2.1.8                                                       
   * VPI:          ii libnvvpi1 1.2.3 arm64 NVIDIA Vision Programming Interface library
   * VisionWorks:  1.6.0.501
   * Vulkan:       1.2.70
   * cuDNN:        8.2.1.32

I need to install ultralytics (yolov8), so python3.8 and pytorch 1.7 are required (at least). I’ve installed python3.8 with sudo apt install python3.8 and addded to .bashrc:

alias python3='python3.8'
alias pip3='python3.8 -m pip'

Next I tried to build pytorch from source, but got errors. May I get an instruction here how to install pytorch1.7, torchvision 0.8, ultralytics?

If it is needed, I can also share how I tried to build pytorch and what problems I saw (it was build but I could not import pytorch because something like this: OSError: /home/argo/pytorch/torch/lib/libtorch_global_deps.so: cannot open shared object file: No such file or directory)

[AastaLLL]

Hi,

We don’t have a PyTorch package for Python 3.8 on Nano.
Please build it from the source.

You can find the instructions below:
https://docs.nvidia.com/deeplearning/frameworks/install-pytorch-jetson-platform/index.html

Thanks.

[Argo_sa]

Thanks for the instructions, I will look into it. Also, may I know, do you have any plans on updating low end hardware like jetson nano? Orin nano is too expensive, it would be great to get a new affordable nano. The problem is not even in the hardware side, but support.

[AastaLLL]

Hi,

We provide a package for the default Python version.
For Ubuntu 18.04 used by Nano, it’s Python 3.6.

Thanks.