Conda

Conda is a python package manager used by a wide variety of projects.

Marlowe uses an optimized version of Conda called Mamba.

To load Mamba, run the following after logging in:

module load conda

Every conda command works with mamba. No code customization is needed.

Setting up Conda and pip

Conda works for the most part straight out of the box, but if you need to pip install packages, you will need to follow these steps:

1) Run module load conda

2) Run mamba init

3) Log out of Marlowe

4) Log back in to Marlowe

After you’ve done this, mamba will be added to your path and pip install will function properly.

Installing a Conda environment in a different location

Home directories have a quota of 15GB by default. This means that even small conda environments can run into quota issues rather quickly.

To combat this, we recommend setting up your conda environment in a separate directory. You can do this with the --prefix command.

For example:

conda create --prefix /projects/m223813/mycondadir numpy=1.21

The above will create a new conda environment in the /projects/m223813/mycondadir folder with numpy 1.21. Since it’s not being installed into your home directory, you don’t have to worry about the same 15GB quota as before!

You can also use the --prefix argument with conda env and other commands.