Skip to content

Installing User-Level Software Packages

This document provides guidance for installing additional software packages, particularly Python libraries, within the REPACSS computing environment. Users are permitted to install software in their own workspace without requiring administrative privileges, provided installations are performed in designated directories such as $HOME or $WORK.

Installing Miniforge (Python 3.11)

Miniforge is a minimal Conda-based Python distribution available for local installation. To install Miniforge in your home directory:

  1. Download the installer: 

    wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh

  2. Run the installer: 

    bash Miniforge3-$(uname)-$(uname -m).sh
    When prompted, accept the license agreement and install to a directory under $HOME, for example $HOME/miniforge3.

  3. Initialize Conda: 

    source ~/miniforge3/etc/profile.d/conda.sh
conda init

  4. Restart your shell or manually activate Conda: 

    source ~/.bashrc    # or ~/.zshrc
conda activate base

Python Package Management

Miniforge provides Python 3.11 and supports isolated environments using conda.

Creating and Managing Environments

To create a new conda environment with selected packages:

conda create -n myenv numpy scipy matplotlib
conda activate myenv

To install more packages into an existing environment:

conda activate myenv
conda install pandas scikit-learn

To deactivate or remove an environment:

conda deactivate
conda remove --name myenv --all

Installing CUDA

You can install the CUDA Toolkit from either conda-forge or NVIDIA channels.
Replace x with the minor version you need (e.g., 12.2).

# Install from conda-forge
conda install cudatoolkit=12.x -c conda-forge

# OR install from NVIDIA channel
conda install cudatoolkit=12.x -c nvidia

Best Practices and Guidelines

  • All software should be installed to user-controlled directories such as $HOME or $WORK. Installation to system directories is not permitted.
  • Use conda environments or venv for managing dependencies and improving reproducibility.
  • For large-scale or portable workflows, consider encapsulating environments using containers (e.g., Apptainer/Singularity).
  • Maintain separate environments for distinct projects to prevent dependency conflicts.

Additional Resources

For further configuration or software access guidance, refer to the following documentation:

For additional support or to request software installation at the system level, please contact:

repacss.support@ttu.edu