Overview

This page provides the basics of how to use Anaconda3 on the Compute2 platform through command line.

Load Modules

• Load the RIS THPC module environment first.

  $ module load ris

• Load the Anaconda3 module.

  $ module load anaconda3

Activating the base environment

• The (base) environment needs to be activated before using other conda commands

  $ source activate

• Once conda has been activated, users will need to start a job as work should not be run on the login nodes.

  (base) $ srun -A compute2-account -p general-interactive --pty /bin/bash

Setting Up a Conda Directory

• By default, conda will use $HOME/.conda directory as the source for installing. Given that $HOME directories are limited in space, it is highly recommended to use a Storage Platform for packages.

• The following conda variables can be set either before or while running an interactive job.

  (base) $ export CONDA_ENVS_PATH="/storageN/fs1/allocation_name/Active/conda/envs" (base) $ export CONDA_PKGS_DIRS="/storageN/fs1/allocation_name/Active/conda/pkgs"

• Once these are set, conda will use those directories as a source for environments and packages.

• These variables can also be set in the .bashrc file so that they are set upon login.

Creating a Conda Environment

• Make sure to be working in an interactive job and not on the login clients.

  (base) $ srun -A compute2-account -p general-interactive --pty /bin/bash

• While not a requirement, it is recommended that a .yml file is used when creating an environment.

• Create the .yml file. An example is below.

  name: example channels: - conda-forge - bioconda dependencies: - python=3.12 - pip

• Create the environment using the .yml file.

  (base) $ conda env create -f environment.yml

Activate the Environment

• Once the environment has been created, it can now be activated and used.

  (base) $ conda activate example

Deactivate the Environment

• Once work has been completed you can deactivate the environment if needed.

  (example) $ conda deactivate

• This should put you back in the (base) environment. You can deactivate this one as well.

  (base) $ conda deactivate

Using a Conda Environment in Jobs

• Users will need to make sure the ris and anaconda3 modules are loaded as part of the job.

• Users will need to make sure the Conda environment variables are set either in their .bashrc or as part of running a job.

• Users will need to make sure to activate the Conda environment as part of the job.

Interactive Job (srun)

Lines 5 and 6 are not required if you have these variables in your .bashrc file.

Batch Job

• In addition to the above, for a batch job, users will need to source the .bashrc file.

• Example job file: example.slurm.

  #!/bin/bash #SBATCH -A compute2-account #SBATCH -p general-cpu #SBATCH --output=example.log export CONDA_ENVS_DIRS="/storageN/fs1/allocation_name/Active/conda/envs" export CONDA_PKGS_DIRS="/storageN/fs1/allocation_name/Active/conda/pkgs" module load ris shared module load anaconda3 source activate conda activate example conda info python helloworld.py

Lines 7 and 8 are not required if you have these variables in your .bashrc file or have otherwise defined them in a .condarc.

• Example job submission

  sbatch example.slurm