Friday, 15 January 2021

Using Python Virtual Environments

Different Python projects depend on different packages. To keep them separated and avoid polluting the global space with package installations we can use the concept of virtual environments. There are several ways to work with virtual environments in Python:

virtualenv
virtualenvwrapper
Conda (part of Anaconda)

virtualenv

Virtualenv — virtualenv 20.3.2.dev1+g7185e57 documentation

To use virtualenv we first need to install all necessary components:

$ sudo apt update
$ sudo apt install python3-dev python3-pip
$ sudo pip3 install -U virtualenv # system-wide install

Verification:

$ python3 --version
$ pip3 --version
$ virtualenv --version

or

$ virtualenv

usage: virtualenv [--version] [--with-traceback] [-v | -q] [--read-only-app-data] [--app-data APP_DATA] [--reset-app-data] [--upgrade-embed-wheels] [--discovery {builtin}] [-p py] [--try-first-with py_exe]

[--creator {builtin,cpython3-posix,venv}] [--seeder {app-data,pip}] [--no-seed] [--activators comma_sep_list] [--clear] [--no-vcs-ignore] [--system-site-packages] [--symlinks | --copies] [--no-download | --download]

[--extra-search-dir d [d ...]] [--pip version] [--setuptools version] [--wheel version] [--no-pip] [--no-setuptools] [--no-wheel] [--no-periodic-update] [--symlink-app-data] [--prompt prompt] [-h]

dest

virtualenv: error: the following arguments are required: dest

SystemExit: 2

$ virtualenv --help

[--extra-search-dir d [d ...]] [--pip version] [--setuptools version] [--wheel version] [--no-pip] [--no-setuptools] [--no-wheel] [--no-periodic-update] [--symlink-app-data] [--prompt prompt] [-h]

dest

optional arguments:

--version display the version of the virtualenv package and its location, then exit

--with-traceback on failure also display the stacktrace internals of virtualenv (default: False)

--read-only-app-data use app data folder in read-only mode (write operations will fail with error) (default: False)

--app-data APP_DATA a data folder used as cache by the virtualenv (default: /home/bojan/.local/share/virtualenv)

--reset-app-data start with empty app data folder (default: False)

--upgrade-embed-wheels trigger a manual update of the embedded wheels (default: False)

-h, --help show this help message and exit

verbosity:

verbosity = verbose - quiet, default INFO, mapping => CRITICAL=0, ERROR=1, WARNING=2, INFO=3, DEBUG=4, NOTSET=5

-v, --verbose increase verbosity (default: 2)

-q, --quiet decrease verbosity (default: 0)

discovery:

discover and provide a target interpreter

--discovery {builtin} interpreter discovery method (default: builtin)

-p py, --python py interpreter based on what to create environment (path/identifier) - by default use the interpreter where the tool is installed - first found wins (default: [])

--try-first-with py_exe try first these interpreters before starting the discovery (default: [])

creator:

options for creator builtin

--creator {builtin,cpython3-posix,venv}

create environment via (builtin = cpython3-posix) (default: builtin)

dest directory to create virtualenv at

--clear remove the destination directory if exist before starting (will overwrite files otherwise) (default: False)

--no-vcs-ignore don't create VCS ignore directive in the destination directory (default: False)

--system-site-packages give the virtual environment access to the system site-packages dir (default: False)

--symlinks try to use symlinks rather than copies, when symlinks are not the default for the platform (default: True)

--copies, --always-copy try to use copies rather than symlinks, even when symlinks are the default for the platform (default: False)

seeder:

options for seeder app-data

--seeder {app-data,pip} seed packages install method (default: app-data)

--no-seed, --without-pip do not install seed packages (default: False)

--no-download, --never-download

pass to disable download of the latest pip/setuptools/wheel from PyPI (default: True)

--download pass to enable download of the latest pip/setuptools/wheel from PyPI (default: False)

--extra-search-dir d [d ...] a path containing wheels to extend the internal wheel list (can be set 1+ times) (default: [])

--pip version version of pip to install as seed: embed, bundle or exact version (default: bundle)

--setuptools version version of setuptools to install as seed: embed, bundle or exact version (default: bundle)

--wheel version version of wheel to install as seed: embed, bundle or exact version (default: bundle)

--no-pip do not install pip (default: False)

--no-setuptools do not install setuptools (default: False)

--no-wheel do not install wheel (default: False)

--no-periodic-update disable the periodic (once every 14 days) update of the embedded wheels (default: False)

--symlink-app-data symlink the python packages from the app-data folder (requires seed pip>=19.3) (default: False)

activators:

options for activation scripts

--activators comma_sep_list activators to generate - default is all supported (default: bash,cshell,fish,powershell,python,xonsh)

--prompt prompt provides an alternative prompt prefix for this environment (default: None)

config file /home/bojan/.config/virtualenv/virtualenv.ini missing (change via env var VIRTUALENV_CONFIG_FILE)

Create virtual environment (we chose to use python3 as Python interpreter and venv as the directory where to store virtual environment):

$ virtualenv -p python3 ./venv_dir_name

$ virtualenv --system-site-packages -p python3 ./venv_dir_name

From virtualenv · PyPI:

If you build with virtualenv --system-site-packages ENV, your virtual environment will inherit packages from /usr/lib/python2.7/site-packages (or wherever your global site-packages directory is).

This can be used if you have control over the global site-packages directory, and you want to depend on the packages there. If you want isolation from the global system, do not use this flag. BK: This flag should be avoided as it defeats the entire point of virtualenvs (which is why --no-site-packages was made the default).

Example:

$ virtualenv -p python3 ./venv

created virtual environment CPython3.8.5.final.0-64 in 253ms

creator CPython3Posix(dest=/home/bojan/dev/github/python-demo/venv, clear=False, no_vcs_ignore=False, global=False)

seeder FromAppData(download=False, pip=bundle, setuptools=bundle, wheel=bundle, via=copy, app_data_dir=/home/bojan/.local/share/virtualenv)

added seed packages: pip==20.3.3, setuptools==51.1.2, wheel==0.36.2

activators BashActivator,CShellActivator,FishActivator,PowerShellActivator,PythonActivator,XonshActivator

I once experienced this error when tried to create a virtual environment:

$ virtualenv -p python3 ./venv

Traceback (most recent call last):

File "/usr/local/bin/virtualenv", line 5, in <module>

from virtualenv.__main__ import run_with_catch

File "/usr/local/lib/python3.8/dist-packages/virtualenv/__init__.py", line 3, in <module>

from .run import cli_run, session_via_cli

File "/usr/local/lib/python3.8/dist-packages/virtualenv/run/__init__.py", line 6, in <module>

from ..app_data import make_app_data

File "/usr/local/lib/python3.8/dist-packages/virtualenv/app_data/__init__.py", line 9, in <module>

from appdirs import user_data_dir

ModuleNotFoundError: No module named 'appdirs'

I then installed the missing module via pip but this showed another error:

$ sudo pip3 install appdirs

Collecting appdirs

Downloading appdirs-1.4.4-py2.py3-none-any.whl (9.6 kB)

ERROR: virtualenv 20.3.1 requires filelock<4,>=3.0.0, which is not installed.

Installing collected packages: appdirs

Successfully installed appdirs-1.4.4

I resolved this by reinstalling the virtualenv:

$ sudo pip3 install virtualenv

Requirement already satisfied: virtualenv in /usr/local/lib/python3.8/dist-packages (20.3.1)

Requirement already satisfied: appdirs<2,>=1.4.3 in /usr/local/lib/python3.8/dist-packages (from virtualenv) (1.4.4)

Requirement already satisfied: six<2,>=1.9.0 in /usr/lib/python3/dist-packages (from virtualenv) (1.14.0)

Requirement already satisfied: distlib<1,>=0.3.1 in /usr/local/lib/python3.8/dist-packages (from virtualenv) (0.3.1)

Collecting filelock<4,>=3.0.0

Downloading filelock-3.0.12-py3-none-any.whl (7.6 kB)

Installing collected packages: filelock

Successfully installed filelock-3.0.12

Once virtual environment is created, we'll have venv/bin/activate script created. Let's explore it:

$ cat venv/bin/activate
...

deactivate() {

...

}

...

VIRTUAL_ENV="/home/bojan/dev/github/tensorflow-demo/venv"
export VIRTUAL_ENV

_OLD_VIRTUAL_PATH="$PATH"
PATH="$VIRTUAL_ENV/bin:$PATH"
export PATH

...

pydoc () {

python -m pydoc "$@"

}

...

We can see that we'll now have a new environment variable set - VIRTUAL_ENV and also a deactivate commmand.

To activate virtual environment:

$ source ./venv/bin/activate

We can also use a dot command:

$ . ./venv/bin/activate

When virtual environment is active, your shell prompt is prefixed with its name in form: (venv).

(venv) $ which pip

/home/bojan/dev/github/python-demo/venv/bin/pip

(venv) $ which pip3

/home/bojan/dev/github/python-demo/venv/bin/pip3

If we've saved requirements.txt, we can install all dependencies now with on of the following commands (see Python Package Management with pip | My Public Notepad):

$ pip install -r requirements.txt

$ pip3 install -r requirements.txt

$ pip3 install -r requirements.txt --verbose

Installing packages using pip and virtual environments — Python Packaging User Guide

It is possible to install a single new package in the virtual environment:

$ pip install package_name

python - How to install a package inside virtualenv? - Stack Overflow

Example:

(venv) $ pip3 install psycopg2

Collecting psycopg2

Using cached psycopg2-2.8.6-cp38-cp38-linux_x86_64.whl

Installing collected packages: psycopg2

Successfully installed psycopg2-2.8.6

To verify installation path:

(venv) $ pip3 show psycopg2

Name: psycopg2

Version: 2.8.6

Summary: psycopg2 - Python-PostgreSQL Database Adapter

Home-page: https://psycopg.org/

Author: Federico Di Gregorio

Author-email: fog@initd.org

License: LGPL with exceptions

Location: /home/user/dev/my-app/venv/lib/python3.8/site-packages

Requires:

Required-by:

To check the version of the package installed in virtual environment:

(venv) $ pip3 list | grep psycopg

psycopg2 2.8.6

To exit virtual environment:

(venv) $ deactivate

How to rename virtual environment?

directory - How to rename a virtualenv in Python? - Stack Overflow

From inside active virtual environment:

$ pip freeze > requirements.txt
$ deactivate

From python - Pip freeze vs. pip list - Stack Overflow:

pip list shows ALL packages.

pip freeze shows packages YOU installed via pip (or pipenv if using that tool) command in a requirements format.

Also, be aware of `$ pip freeze > requirements.txt` considered harmful.

Then delete old environment directory. E.g.

$ rm -r ~/python-envs/tensorrt-5.1.5.1-tensorflow-1.14.0/

...and create a new one with correct name:

$ python3 -m virtualenv -p python3 ~/python-envs/tensorrt-5.1.5.1-tensorflow-1.15.0
Already using interpreter /usr/bin/python3
Using base prefix '/usr'
New python executable in /home/nvidia/python-envs/tensorrt-5.1.5.1-tensorflow-1.15.0/bin/python3
Also creating executable in /home/nvidia/python-envs/tensorrt-5.1.5.1-tensorflow-1.15.0/bin/python
Installing setuptools, pkg_resources, pip, wheel...done.

Activate it and install all packages from the saved list:

$ source ~/python-envs/tensorrt-5.1.5.1-tensorflow-1.15.0/bin/activate

$ pip install -r requirements.txt

$ pip3 install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v43 tensorflow-gpu==1.15.0+nv19.12

virtualenvwrapper

TBD...

Conda

Managing Python Environments with Conda | My Public Notepad

References

Setting up your python development environment (with pyenv, virtualenv, and virtualenvwrapper)
The Right Way to Use Virtual Environments
Pipenv & Virtual Environments
Create a virtual environment
Virtual Environments and Packages
Python Virtual Environments: A Primer

An Introduction To Venv

Troubleshooting virtualenvwrapper errors

It was my mistake that I did not track immediately the cause of the error that one day started appearing each time I'd open a new Terminal:

/usr/bin/python3: Error while finding module specification for 'virtualenvwrapper.hook_loader' (ModuleNotFoundError: No module named 'virtualenvwrapper')
virtualenvwrapper.sh: There was a problem running the initialization hooks.
If Python could not import the module virtualenvwrapper.hook_loader,
check that virtualenvwrapper has been installed for
VIRTUALENVWRAPPER_PYTHON=/usr/bin/python3 and that PATH is
set properly.

virtualenvwrapper is, as its name says, a wrapper around virtualenv.

virtualenvwrapper 5.0.1.dev2 — virtualenvwrapper 5.0.1.dev2 documentation

Bash script to source a Python virtual environment - Ask Ubuntu

When a new Terminal is opened, it starts a shell which sources the first of .bash_profile, .bash_login, .profile that exists and is readable. [bash - Why must I source .bashrc every time I open terminal for aliases to work? - Ask Different]

In my case .bash_profile and .bash_login don't exist but .profile does and it sources ~/.bashrc:

$ cat ~/.profile

# ~/.profile: executed by the command interpreter for login shells.

# This file is not read by bash(1), if ~/.bash_profile or ~/.bash_login

# exists.

# see /usr/share/doc/bash/examples/startup-files for examples.

# the files are located in the bash-doc package.

# the default umask is set in /etc/profile; for setting the umask

# for ssh logins, install and configure the libpam-umask package.

#umask 022

# if running bash

if [ -n "$BASH_VERSION" ]; then

# include .bashrc if it exists

if [ -f "$HOME/.bashrc" ]; then

. "$HOME/.bashrc"

# set PATH so it includes user's private bin if it exists

if [ -d "$HOME/bin" ] ; then

PATH="$HOME/bin:$PATH"

# set PATH so it includes user's private bin if it exists

if [ -d "$HOME/.local/bin" ] ; then

PATH="$HOME/.local/bin:$PATH"

So I looked at ~/.bashrc and I found that some time in past I changed it to include some additional environment variables:

$ cat ~/.bashrc
...

# B.Komazec added:
# virtualenv and virtualenvwrapper
export WORKON_HOME=$HOME/.virtualenvs
export VIRTUALENVWRAPPER_PYTHON=/usr/bin/python3
source /usr/local/bin/virtualenvwrapper.sh

...

Indeed, the error above was coming from virtualenvwrapper.sh, from function virtualenvwrapper_run_hook():

$ cat /usr/local/bin/virtualenvwrapper.sh

...

# Run the hooks

function virtualenvwrapper_run_hook {

typeset hook_script

typeset result

hook_script="$(virtualenvwrapper_tempfile ${1}-hook)" || return 1

# Use a subshell to run the python interpreter with hook_loader so

# we can change the working directory. This avoids having the

# Python 3 interpreter decide that its "prefix" is the virtualenv

# if we happen to be inside the virtualenv when we start.

( \

virtualenvwrapper_cd "$WORKON_HOME" &&

"$VIRTUALENVWRAPPER_PYTHON" -m 'virtualenvwrapper.hook_loader' \

${HOOK_VERBOSE_OPTION:-} --script "$hook_script" "$@" \

)

result=$?

if [ $result -eq 0 ]

then

if [ ! -f "$hook_script" ]

then

echo "ERROR: virtualenvwrapper_run_hook could not find temporary file $hook_script" 1>&2

command \rm -f "$hook_script"

return 2

# cat "$hook_script"

source "$hook_script"

elif [ "${1}" = "initialize" ]

then

cat - 1>&2 <<EOF

virtualenvwrapper.sh: There was a problem running the initialization hooks.

If Python could not import the module virtualenvwrapper.hook_loader,

check that virtualenvwrapper has been installed for

VIRTUALENVWRAPPER_PYTHON=$VIRTUALENVWRAPPER_PYTHON and that PATH is

set properly.

EOF

command \rm -f "$hook_script"

return $result

}

...

I checked which packages are installed for pip3:

$ pip3 list

...but among them were neither virtualenv nor virtualenvwrapper. So I installed them:

$ sudo pip3 install virtualenv virtualenvwrapper

Collecting virtualenv

Downloading virtualenv-20.3.1-py2.py3-none-any.whl (5.7 MB)

Collecting virtualenvwrapper

Downloading virtualenvwrapper-4.8.4.tar.gz (334 kB)

Requirement already satisfied: six<2,>=1.9.0 in /usr/lib/python3/dist-packages (from virtualenv) (1.14.0)

Requirement already satisfied: appdirs<2,>=1.4.3 in /usr/lib/python3/dist-packages (from virtualenv) (1.4.3)

Requirement already satisfied: filelock<4,>=3.0.0 in /usr/lib/python3/dist-packages (from virtualenv) (3.0.12)

Collecting distlib<1,>=0.3.1

Downloading distlib-0.3.1-py2.py3-none-any.whl (335 kB)

Collecting stevedore

Downloading stevedore-3.3.0-py3-none-any.whl (49 kB)

Collecting virtualenv-clone

Downloading virtualenv_clone-0.5.4-py2.py3-none-any.whl (6.6 kB)

Collecting pbr!=2.1.0,>=2.0.0

Using cached pbr-5.5.1-py2.py3-none-any.whl (106 kB)

Building wheels for collected packages: virtualenvwrapper

Building wheel for virtualenvwrapper (setup.py) ... done

Created wheel for virtualenvwrapper: filename=virtualenvwrapper-4.8.4-py2.py3-none-any.whl size=24833 sha256=0e77420c4c5bd24518d388768cf102170bd8b0243fd3052fdbec24b01f3bb59a

Stored in directory: /root/.cache/pip/wheels/47/15/3d/7a26eaf92e79f80a3df3ac5f8e0f0f5b7efdf24d313c594a44

Successfully built virtualenvwrapper

Installing collected packages: distlib, virtualenv, pbr, stevedore, virtualenv-clone, virtualenvwrapper

Attempting uninstall: distlib

Found existing installation: distlib 0.3.0

Not uninstalling distlib at /usr/lib/python3/dist-packages, outside environment /usr

Can't uninstall 'distlib'. No files were found to uninstall.

Successfully installed distlib-0.3.1 pbr-5.5.1 stevedore-3.3.0 virtualenv-20.3.1 virtualenv-clone-0.5.4 virtualenvwrapper-4.8.4

To verify that packages are installed indeed:

$ pip3 list

...

virtualenv 20.3.1

virtualenv-clone 0.5.4

virtualenvwrapper 4.8.4

...

I opened a new Terminal and voila - the error is gone!

For a reference, my Python paths are:

$ which python

/usr/bin/python

$ which python3

/usr/bin/python3

References

python - /usr/bin/python3: Error while finding spec for 'virtualenvwrapper.hook_loader' (: No module named 'virtualenvwrapper') - Stack Overflow

Sunday, 10 January 2021

Installing JetPack on Jetson TX2 from NVIDIA SDK Manager Docker container

For a very long time, if you wanted to flash your Jetson TX and install JetPack SDK you had to download and install NVIDIA JetPack and later, NVIDIA SDK Manager, on the Linux host computer first. Every package installation pollutes your host machine and also takes some disk space. To avoid this, NVIDA created a Docker image with SDK Manager so once JetPack is installed on Jetson, this Docker image can be deleted and your Linux host remains in the same state as before. This has been available since NVIDIA SDK Manager 1.4 (December 2020).

I had some of the older versions of JetPack installed on my Jetson TX2 and I wanted to install the most recent one (4.4.1 at the moment; 4.5 is announced for January 2021). One of the benefits I wanted to get is the upgrading to the next JetPack release via apt package management tool (this has been available since JetPack 4.4).

I want to share here my experience with the process of running NVIDIA SDK Manager Docker container and flashing the Jetson TX2 with it. I followed the official documentation about this process: Docker Images :: NVIDIA SDK Manager Documentation.

I logged in to NVIDIA Developer center and downloaded this Docker image from this URL: https://developer.nvidia.com/nvidia-sdk-manager-docker-image. The image came as an 942MB archive named sdkmanager-1.4.0.7363_docker.tar.gz.

$ docker load -i ./sdkmanager-1.4.0.7363_docker.tar.gz

805802706667: Loading layer 65.61MB/65.61MB

3fd9df553184: Loading layer 15.87kB/15.87kB

7a694df0ad6c: Loading layer 3.072kB/3.072kB

2f694c79b042: Loading layer 148.2MB/148.2MB

26765aed7e25: Loading layer 502.3kB/502.3kB

b398b8335e67: Loading layer 6.015MB/6.015MB

08b68150484c: Loading layer 1.135MB/1.135MB

31fbfacf550e: Loading layer 1.135MB/1.135MB

84979f95b15f: Loading layer 502.3kB/502.3kB

cd9205d2e1f9: Loading layer 2.075MB/2.075MB

414d8a00c66e: Loading layer 66.07MB/66.07MB

cc3271f36011: Loading layer 84.83MB/84.83MB

5072b3ebcb77: Loading layer 2.108MB/2.108MB

38be13d541b9: Loading layer 1.781MB/1.781MB

ff63398d24ea: Loading layer 99.17MB/99.17MB

7cb4d04a8659: Loading layer [==================================================>] 462.3MB/462.3MB

cd2c9f6e22b0: Loading layer [==================================================>] 2.048kB/2.048kB

65cd593db96f: Loading layer [==================================================>] 3.584kB/3.584kB

8900dbcf5626: Loading layer [==================================================>] 3.584kB/3.584kB

9ac46abadb31: Loading layer [==================================================>] 3.072kB/3.072kB

a20c9aaeb9c3: Loading layer [==================================================>] 417.3kB/417.3kB

cae1bf65143a: Loading layer [==================================================>] 3.584kB/3.584kB

Loaded image: sdkmanager:1.4.0.7363

As this is the latest version of this Docker image, I tagged it with the latest tag:

$ docker tag sdkmanager:1.4.0.7363 sdkmanager:latest

I made sure that the image is listed among other Docker images on my machine:

$ docker images

REPOSITORY TAG IMAGE ID CREATED SIZE

sdkmanager 1.4.0.7363 0e9d62e318ad 2 weeks ago 913MB

sdkmanager latest 0e9d62e318ad 2 weeks ago 913MB

...

SDK Manager executable (sdkmanager) is the entrypoint of this Docker image and I wanted to test the Docker image by running it with some simple CLI commands listed here.

$ docker run -it --rm sdkmanager --help

NVIDIA SDK MANAGER

NVIDIA SDK Manager is an all-in-one tool that bundles developer software and

provides an end-to-end development environment setup solution for NVIDIA

SDKs.

General Options

-h, --help Displays this usage guide.

--ver Output the version of the installed SDK Manager client

--settings Optional. Configure SDK Manager settings in the terminal.

--query interactive|noninteractive Prints all options available for the user. Must be executed with the --use or --offline settings

--showallversions Prints all available product versions for the user.

--logs Optional. Set this option to export the log files when the process is complete.

--exitonfinish Optional. Automatically exit from SDK Manager when the install/uninstall session is finished (skip user input). Intended for scripts/automation usage.

--user email_address Optional. Set the user email to login. Valid only for NVOnline login.

--password string Optional. Set the user login password. Valid only for NVOnline login.

--logintype devzone|nvonline Optional. Login with developer.nvidia.com or partners.nvidia.com account. Default is devzone.

--staylogin true|false Optional. Keep the user account logged-in for next running session.

--logout Logout user account from SDK Manager.

--offline Optional. Skip login to NVIDIA servers. Install SDK from pre downloaded location, used with --downloadfolder option.

--downloadfolder string Optional. Set the download folder for the SDK components. Used for downloading the files and for locating the SDK components when using --offline.

--archivedversions Optional. Display only archived versions.

--cli install|uninstall|downloadonly Mandatory. Set the requested action.

--sudopassword string Optional. Set the sudo password to skip the authentication prompt.

--datacollection enable|disable Optional. Set to enable or disalbe usage data collection.

Specific arguments for install/uninstall:

--product product_name Mandatory. Set the product name.

--version string Mandatory. Set the product version. Use --query to get available version values.

--targetos target_os Mandatory. Set the target hardware operating system.

--host Optional. Set if host side components need to be installed.

--target target_hardware Optional. Set the target hardware in use. Use hardware code name.

--flash all|a|b|ab|skip Optional. Set the flash operation mode, which of the Tegras should be flashed.

--additionalsdk additional_sdk_title Optional. Specify any additional SDK to install. Multiple entries are allowed.

--select section_or_group_title Optional. Specify section or group to installation list. Multiple entries are allowed.

--deselect section_or_group_title Optional. Specify section or group to exclude from installation list. Multiple entries are allowed.

--license accept Optional. Set this option to accept the terms and conditions of SDK license agreements.

--targetimagefolder string Optional. Set the host location of the target hardware image for flashing.

--responsefile string Optional. Set the response file path. Response file samples can be found in the product folder /opt/nvidia/sdkmanager.

Example

$ sdkmanager [--user user@user.com] [--query]

$ sdkmanager [--cli install|uninstall|downloadonly] [cli options] ...

$ sdkmanager [--settings]

$ sdkmanager [--help]

$ docker run -it --rm sdkmanager --ver

1.4.0.7363

I connected Jetson TX2 to my Ubuntu host via USB cable and put Jetson into forced recovery mode (as described here: Jetson_X2_Developer_Kit_User_Guide.pdf).

I checked that Jetson is listed among other USB devices:

$ lsusb

...

Bus 002 Device 004: ID 0955:7c18 NVIDIA Corp. APX

...

I then run a query command on SDK manager to get a list of available install options:

$ docker run -it --rm sdkmanager --query

To initiate login process open https://static-login.nvidia.com/service/default/pin?user_code=36223035 in a browser (can be done on a different machine) and login with your NVIDIA Developer account. SDK Manager will start once done.

? SDK Manager is waiting for you to complete login.

1) Generate a new login user code

2) Cancel login

Answer:

Waiting for user information from NVIDIA authentication server...

Retrieving user information...

Loading and processing available products...

Loading user information...

User information loaded successfully.

Loading server data...

Server data loaded successfully.

Available options are:

Jetson 4.4

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P2888-0001 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P2888-0004 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P2888-0006 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P2888-0060 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3668-0000 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3668-0001 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3310-1000 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3489-0080 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3489-0888 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3489-0000 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P2180-1000 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3448-0000 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3448-0002 --flash all --additionalsdk DeepStream

sdkmanager --cli install --logintype devzone --product Jetson --version 4.4 --targetos Linux --host --target P3448-0020 --flash all --additionalsdk DeepStream

Query completed.

I was not sure which target I should choose so I used table from NVIDIA Jetson Linux Developer Guide : Introduction | NVIDIA Docs to check the P-number of my Jetson TX2 and it was P3310-1000.

It was now the time to do the main part of the job. As a guide, I looked at the docker run command example used for flashing Jetson Nano (listed here: Docker Images :: NVIDIA SDK Manager Documentation) and modified it for Jetson TX2:

$ docker run -it --rm --privileged -v /dev/bus/usb:/dev/bus/usb/ --name JetPack_TX2_Devkit sdkmanager --cli install --logintype devzone --product Jetson --target P3310-1000 --targetos Linux --version 4.4.1 --flash all --license accept --staylogin true --datacollection enable --exitonfinish

To initiate login process open https://static-login.nvidia.com/service/default/pin?user_code=64552553 in a browser (can be done on a different machine) and login with your NVIDIA Developer account. SDK Manager will start once done.

? SDK Manager is waiting for you to complete login.

1) Generate a new login user code

2) Cancel login

Answer:

Waiting for user information from NVIDIA authentication server...

Retrieving user information...

Loading and processing available products...

Loading user information...

User information loaded successfully.

Loading server data...

Server data loaded successfully.

Session initialized...

Installation of this software is under the terms and conditions of the license agreements located in /opt/nvidia/sdkmanager/Eula/

===== INSTALLATION COMPLETED SUCCESSFULLY. =====

- Drivers for Jetson: Installed

- File System and OS: Installed

- Device Mode Host Setup in Flash: Installed

- Flash Jetson TX2: Installed

- Device Mode Host Setup in Target SDK: Installed

- DateTime Target Setup: Installed

- CUDA Toolkit for L4T: Installed

- cuDNN on Target: Installed

- TensorRT on Target: Installed

- OpenCV on Target: Installed

- VisionWorks on Target: Installed

- VPI on Target: Installed

- NVIDIA Container Runtime with Docker integration (Beta): Installed

- Multimedia API: Installed

===== Installation completed successfully - Total 14 components =====

===== 14 succeeded, 0 failed, 0 up-to-date, 0 skipped =====

Here are some screenshots of the SDK manager running from within Docker container which show the process of downloading the packages, flashing the OS and installing JetPack on Jetson board: