From radwiki
Jump to navigation Jump to search

Anaconda Is a software package that contains a python interpreter and an optimized set of common python modules. It also includes a useful management tool called conda that helps with package updates and setup of custom python environments.

Using it is optional, but recommended; by default the GPU cluster system has Python 3.6 and some basic packages. The anaconda interpreter is at least version 3.7 and the included numpy module uses the faster intel MLK library for linear algebra.

Installing packages

conda install <package>

If conda can not find a certain package, use pip to install the package into the current environment. To be safe ensure that all required conda packages are installed first, then use pip, after that conda won't be aware of the new pip-installed packages which may cause requirement collisions.

Usage Examples

Activate an environment

Depending on your configuration you may already have the default anaconda base environment active. To enter an existing environment use: conda activate <environment>

Setup a tensorflow environment and run a jupyter notebook

Exit anaconda environment