How to create virtual environment with OpenCv and Pytorch to execute your program on GPU

A98 · September 16, 2023, 12:33pm

This is my solution of this question.

My sequence of actions:

1.Anaconda installation:

Clicked on:

Archiconda3-0.2.3-Linux-aarch64.sh

It downloaded to /Download folder

In terminal:

gerardg@Nvidia:~$ cd Downloads

gerardg@Nvidia:~/Downloads$ bash Archiconda3-0.2.3-Linux-aarch64.sh

(I installed anaconda to create virtual environment)

JetPack installation:

1.sudo apt dist-upgrade

2.sudo apt update

3.sudo apt-get install nvidia-container

4.sudo apt-get install nvidia-jetpack
(As a result i intalled jetpack 4.6.1; if you have other version of jetpack you also need other pytorch version (most likely); in this situation my sequence of actions wont work)

https://docs.nvidia.com/jetson/jetpack/install-jetpack/index.html

(Next step couldn’t be completed without JetPack)

Pytorch installation:

1.conda create -n pytorch3.6 -c conda-forge python=3.6

2.conda activate pytorch3.6

3.pip install --upgrade pip

4.wget https://nvidia.box.com/shared/static/p57jwntv436lfrd78inwl7iml6p13fzh.whl -O torch-1.8.0-cp36-cp36m-linux_aarch64.whl

5 .sudo apt-get install python3-pip libopenblas-base libopenmpi-dev libomp-dev

6.sudo apt-get install libomp5 (: I installed libomp5 because I had errors, associated with requirement problems)

7.pip3 install Cython

pip3 install numpy torch-1.8.0-cp36-cp36m-linux_aarch64.whl

9.pip install Pillow (Optional: I installed Pillow to work with images)

10.pip install cmake (Optional: I installed cmake because I had errors, associated with requirement problems)

pip3 install numpy torch-1.8.0-cp36-cp36m-linux_aarch64.whl

12.Test:

1.python

2.import torch

3.print(torch.version)

4.print('CUDA available: ’ + str(torch.cuda.is_available()))

(If CUDA is available installation completed successfully)

python3 -m pip install opencv-python==4.6.0.66 –verbose

(I installed specific version to solve problem with infinite wheels building)

AastaLLL · September 18, 2023, 3:53am

Hi,

For PyTorch, please use our prebuilt that has enabled the CUDA support.

For OpenCV, please build it from the source.
Below is the script for your reference:

github.com

AastaNV/JEP/blob/master/script/install_opencv4.5.0_Jetson.sh

#!/bin/bash
#
# Copyright (c) 2020, NVIDIA CORPORATION.  All rights reserved.
#
# NVIDIA Corporation and its licensors retain all intellectual property
# and proprietary rights in and to this software, related documentation
# and any modifications thereto.  Any use, reproduction, disclosure or
# distribution of this software and related documentation without an express
# license agreement from NVIDIA Corporation is strictly prohibited.
#

version="4.5.0"
folder="workspace"

echo "** Remove other OpenCV first"
sudo sudo apt-get purge *libopencv*


echo "** Install requirement"
sudo apt-get update

This file has been truncated. show original

Thanks.