Anaconda and Docker Container

Hello,

  1. It seems that it is not possible to have anaconda on jetson agx xavier but it is possible to have it on nano and tx2, is that correct? if yes but why?
  2. Can I have docker container on agx xavier?
    Thank you

Hi,

The support is already added.
Based on this document, we can install anaconda via pip3 directly.

$ pip3 install anaconda-client
Defaulting to user installation because normal site-packages is not writeable
Collecting anaconda-client
  Downloading anaconda-client-1.2.2.tar.gz (64 kB)
     |████████████████████████████████| 64 kB 340 kB/s            
  Preparing metadata (setup.py) ... done
Collecting clyent
  Downloading clyent-1.2.1.tar.gz (20 kB)
  Preparing metadata (setup.py) ... done
Requirement already satisfied: requests>=2.0 in /usr/lib/python3/dist-packages (from anaconda-client) (2.18.4)
Requirement already satisfied: pyyaml in /usr/lib/python3/dist-packages (from anaconda-client) (3.12)
Requirement already satisfied: python-dateutil in /home/nvidia/.local/lib/python3.6/site-packages (from anaconda-client) (2.8.2)
Requirement already satisfied: pytz in /home/nvidia/.local/lib/python3.6/site-packages (from anaconda-client) (2021.3)
Requirement already satisfied: six>=1.5 in /home/nvidia/.local/lib/python3.6/site-packages (from python-dateutil->anaconda-client) (1.16.0)
Building wheels for collected packages: anaconda-client, clyent
  Building wheel for anaconda-client (setup.py) ... done
  Created wheel for anaconda-client: filename=anaconda_client-1.2.2-py3-none-any.whl size=75915 sha256=806129128d8884b949e4156664a32053a02ed9c90877dcd18e263c116b400bb6
  Stored in directory: /home/nvidia/.cache/pip/wheels/76/77/3d/2c52b4f0935097369ef8091741a2d2a5a347002047ab8b9a5c
  Building wheel for clyent (setup.py) ... done
  Created wheel for clyent: filename=clyent-1.2.1-py3-none-any.whl size=9902 sha256=6534bf2b6f19b146885b8aeb456aff66bbdb728e9e6e5ab3466bde4795ca3c8e
  Stored in directory: /home/nvidia/.cache/pip/wheels/a8/86/12/4bf8e399dc66805e687ae1a23276a5cc543e4a921d5c9d91c8
Successfully built anaconda-client clyent
Installing collected packages: clyent, anaconda-client
Successfully installed anaconda-client-1.2.2 clyent-1.2.1

Docker is pre-installed on the Jetson device.
You can find it with docker -h.

Thanks.

1 Like
  1. Last year when I first setup the agx xavier dev kit I searched and asked here in nvidia forums and I got the answer that it is not possible to install anaconda. Is this possibility new? or I got wrong info from the beginning?

Let me explain my problem, I hope you can help.
to run my several algorithms on agx xavier dev kit I need to have tensorflow, opencv, pytorch, mediapipe and intel realsense camera sdk.
I am using a single virtual env, I tried to install the mediapipe at last but I faced error regarding bazel. I wasn’t successful to get rid of it. Also I am getting ‘no matching distribution’ error, so I have to be careful about versions and the order of the installation.

  1. Is there any efficient installation order for these libraries? which ones should be installed first?

  2. I decided to flash the device but I have updated my laptop to ubuntu 20 recently and It is not possible to do the flashing issue by ubuntu 20. what should I do?

  3. if I’m not mistaken the mediapipe and the bazer will be installed on the system, not the environment or python, so I cannot fix the issue by creating another virtual env or conda env, is that correct?

  4. Is it possible to fix the issue by docker container? is working with it on jetson devices easy and straightforward? or it doesn’t worth for someone who is not familiar with it?

  5. would using another OS on a SD card fix the issue? I saw in forums that it makes more trouble, is that correct?

and finally:

  1. how can I save my current env on a SD cart to use it later again on agx xavier dev kit?

I really appreciate if you answer my questions.
Thank you

Hi,

1. Since anaconda is a third-party library, you will get better support from the library owner.
Based on our testing, it can be installed on Jetson with JetPack 4.6 environment now.

2. Based on the requirement, you can start with the l4t-ml container directly.

3. You can use sdkmanager docker to flash on a Ubuntu 20.04 host.
https://docs.nvidia.com/sdk-manager/docker-containers/index.html

4.no matching distribution” indicates that there is no prebuilt package for ARM.
But if a library is open-sourced (ex. mediapipe), you can build it from the source directly.

5. Yes, it is preinstalled via JePack.

Thanks.

1 Like

Hello,

  1. Nice update, thx. If I install anaconda on jetson device and separately use the containers you mentioned through the link in question 2, Does that make any trouble or conflict? can I have them both?
  2. very interesting! It makes things much easier. How to know the version of my JetPack-L4T ?
  3. I did build it from source but it didn’t work. Is there any container that have mediapipe installed in it already?

Thank you for the detailed answers, very helpful.

Hi,

1.
It should work.

2.
To get the OS version, you can run the following command:

For example:

$ apt-cache show nvidia-jetpack

Package: nvidia-jetpack
Version: 4.6-b197
Architecture: arm64
Maintainer: NVIDIA Corporation
Installed-Size: 194

3.
Sorry that we don’t have a container with mediapipe preinstalled.
But it should work with some customization for ARM.
Please find the previous discussion below:

Thanks.