Installing Packages

Note: It is worth noting here again that the packages that can be installed do not have to be python packages. conda is language-agnostic and you can even create an environment without a python installation. Nevertheless, conda is mostly used to manage python environments.

Installing packages

The simplest way to install a package in the activated environment is simply

conda install <PACKAGE>

and to update

conda update <PACKAGE>

If you want to specify a different environment you can add the --name option

conda install --name test-env <PACKAGE>

Channel specification

Quite many packages come from specific channels. Most notably pytorch. Channels can be specified using the -c option.

PyTorch:

conda install -c pytorch pytorch

NVIDIA Rapids for Power9 Architecture (Traverse)

CHNL=https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda
conda create --name rapids-env --channel $CHNL cudf cuml

Installing packages using Pip

conda will also track installations performed through pip. I.e. if you, for example, installed scipy using pip

pip install scipy

conda is aware of this installation. You can even export the environment and include the pip dependencies:

conda env export

DISCLAIMER: conda env export --from-history does not include the pip dependencies. There is an open issue for conda on GitHub to include this in future releases (Github Issue).

Version constraints

Often, the package version has to be constraint by some range in version etc. The table below shows an overview on how to apply the version constraints

NOTE: You have to use quotation marks if you are using any of [> < | *] in the specification or if the specification contains a space.

Next: Creating a package