Anaconda/en: Difference between revisions
No edit summary |
(Updating to match new version of source page) |
||
Line 1: | Line 1: | ||
<languages /> | <languages /> | ||
[[Category:Software]] | [[Category:Software]] | ||
'''Attention:''' Alors que Conda fonctionne bien dans l'environnement d'un ordinateur personnel, il a tendance à créer plus de problèmes qu'il en résout dans le contexte d'une grappe. Pour cette raison, on encourage les usagers de se tourner vers des outils comme un environnement virtuel ou des paquets binaires (Python wheels), qui sont documentés sur la page [[Python]]. | |||
== Description == | == Description == | ||
Revision as of 15:52, 18 April 2018
Attention: Alors que Conda fonctionne bien dans l'environnement d'un ordinateur personnel, il a tendance à créer plus de problèmes qu'il en résout dans le contexte d'une grappe. Pour cette raison, on encourage les usagers de se tourner vers des outils comme un environnement virtuel ou des paquets binaires (Python wheels), qui sont documentés sur la page Python.
Description
Anaconda is an open source distribution of Python and R which aims to simplify the management and deployment of modules.
Installation
Python distributions installed on Compute Canada servers are compiled from the source available at the python.org site. Users are however free to install Anaconda in their own directory. The following instructions should simplify this job and thus avoid compatibility errors.
In order to limit the installation time and the amount of storage needed, we encourage the installation of Miniconda instead of Anaconda. Miniconda includes the conda package manager and Python. You can then use the conda command to install the software you need.
Home directory (user-based installation)
By default the installation is performed in your home directory to which only you have access. If you want to share your installation with other members of your group, use the instructions in the following section.
To install Miniconda with Python 2, execute the following command:
[name@server ~]$ eb Miniconda2-4.3.27.eb
For Miniconda with Python 3, execute the following command:
[name@server ~]$ eb Miniconda3-4.3.27.eb
Be patient, the installation of Miniconda can take several minutes.
Project directory (group-based installation)
To install Anaconda with Python 2, execute the following commands, replacing <project>
by your project identifier:
[name@server ~]$ eb --sticky-bit --set-gid-bit --prefix=$(readlink ~/projects/<project>) Miniconda2-4.3.27.eb
For Anaconda with Python 3, execute the following command, replacing <project>
by your group identifier:
[name@server ~]$ eb --sticky-bit --set-gid-bit --prefix=$(readlink ~/projects/<project>) Miniconda3-4.3.27.eb
Be patient, the installation of Miniconda can take several minutes.
The installation of Miniconda automatically creates a module file that you can load with the command of the same name. For the module command to find this file, you need to tell it where to look with the following command, replacing <project>
by your project identifier:
[name@server ~]$ module use ~/projects/<project>/modules/*/Core
If you want this module to be available every time you connect to the cluster, you can add the preceding command to the end of your .bashrc
file.
Other versions
To install any other version of Anaconda, we recommend that you contact technical support.
Usage
Load the Miniconda 2 module with
[name@server ~]$ module load miniconda2
or the Miniconda 3 module with
[name@server ~]$ module load miniconda3
conda
You can install Python modules in your Miniconda installation by using the command conda directly. For example, to install the theano module, use
[name@server ~]$ conda install theano