Anaconda for Jetson Nano

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Hi,

I just bought an AMAZING Jetson Nano yesterday, and I realized that Jetson Nano runs on a AArch64 architecture. So the official version of anaconda is unavailable.

Until I found this: Archiconda, a distribution of Conda for 64 bit ARM platform.

If anyone needs this, plz review: Releases · Archiconda/build-tools · GitHub

I think that Archiconda should also be able to be installed on Jetson TX1/TX2 but I didn’t tried.

4 Likes
2

Hi,

Thanks for sharing it. I just install it.

3

@yannickdeicaza
I see a bash shell at Archiconda With just conda-forge

I have HDF5 tools installed.
Do you know of any requirements beyond that?
Could you please provide a procedure to install it?
Thank you.

To answer my own question, I went to this link:

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

1 Like
4

Do you know how to launch archiconda3?

1 Like
5

Ihave the same question

6

@phd.matamoros: Have a look at https://github.com/helmuthva/jetson/blob/master/workflow/deploy/ml-base/src/Dockerfile which shows how to install and use Archiconda on a Jetson (Nano) - in this case inside Docker.

See https://github.com/helmuthva/jetson for the overall project where this is used as part of the “ml-base” Docker Image including Jupyter notebooks in the “jupyter” image - see https://github.com/helmuthva/jetson/blob/master/workflow/deploy/jupyter/src/Dockerfile

The build instructions in the Dockerfiles can be easily replicated on the host in case you don’t want to use Docker.

7

@friskit.china: Thank you for sharing this info! I was not aware of archiconda’s existence and was racking my brain how to get anaconda installed on aarch64 - then i saw this post.

For me getting the archiconda3 work came down to adding the path to the directory of its installation to the environment variable PATH.

8

Unfortunately, this only works for systems that return uname -m as aarch64. Fails if it comes back arm71, like my Raspberry Pi on Raspbian.

9

I installed archiconda successfully on my Jetson TX1. However, when I try to run conda update I get an error:

nvidia@nvidia-desktop:~/archiconda3$ uname -m
aarch64
nvidia@nvidia-desktop:~/archiconda3$ conda update -n base --all
Solving environment: failed

NotWritableError: The current user does not have write permissions to a required path.
path: /home/nvidia/archiconda3/pkgs/urls.txt
uid: 1000
gid: 1000

If you feel that permissions on this path are set incorrectly, you can manually
change them by executing

$ sudo chown 1000:1000 /home/nvidia/archiconda3/pkgs/urls.txt

In general, it’s not advisable to use ‘sudo conda’.

I tried looking for the “urls.txt” file under the above path but it doesn’t seem to exist. Any ideas?

10

Hi,

This is a permission issue.
Have you try to execute the command with root to see if works?

<b>sudo</b> conda update -n base --all

Or update the permission of this folder:

/home/nvidia/archiconda3/pkgs/

Thanks.

1 Like
11

Dear Sir:

The big problem is that archiconda3 lacks of the installation environment of tensorflow-gpu. So it failed to install tensorflow-gpu. Do
you have any solutuion to the problem.

PachagesNotFoundError: The following packages are not available from the current channels

  • tensorflow-gpu

I am pleased to add the installation steps as follows.

Installation Steps

  1. create an environment:

conda create --name tf_gpu

  1. activate tf-gpu

$ activate tf-gpu

  1. install tensorflow-gpu

$ conda install tensorflow-gpu==2.0.0
or
$ conda install tensorflow-gpu

  1. Failed messages

PachagesNotFoundError: The following packages are not available from the current channels

  • tensorflow-gpu

…

Thanks in advance,

12

For those who are already familiar with installation and setup for anaconda or miniconda the steps for archiconda are similar. First download and install a version of the archiconda installer:

# use install path: $HOME/archiconda3
wget --quiet -O archiconda.sh https://github.com/Archiconda/build-tools/releases/download/0.2.3/Archiconda3-0.2.3-Linux-aarch64.sh && \
    sh archiconda.sh -b -p $HOME/archiconda3 && \
    rm archiconda.sh

Next we update the conda environment and add channels containing recipes for 64 bit Arm packages.

export PATH=$HOME/archiconda3/bin:$PATH
conda config --add channels gaiar && \
conda config --add channels conda-forge && \
conda config --add channels c4aarch64 && \
conda update -n base --all && \
conda install -y python=3.6.7 libiconv && \
conda install -y conda-build && \
conda install -y anaconda-client

A standard conda environment should now be available in $HOME/archiconda3/bin. To automatically edit your .bashrc and activate the base environment on login, use conda init bash or see $HOME/archiconda3/etc/profile.d/conda.sh

I hope it helps,

17 Likes