Problem
- This page should probably be updated to replace
miniconda with miniforge as the recommended way of researchers using conda.
- Benefits: Miniforge uses
conda-forge as the default channel (which is also open source), and even includes Mamba for faster package resolution, while Miniconda can use Anaconda's default channels which have licensing restrictions.
Suggestion:
The current installation method would just need to be replaced with something like:
export CONDADIR=/nobackup/projects/<project>/$USER/aarch64 # Update this with your <project> code.
mkdir -p $CONDADIR
pushd $CONDADIR
# Download the latest miniforge installer for aarch64
wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-aarch64.sh
# Validate the file checksum (optional - check against releases page)
sha256sum Miniforge3-Linux-aarch64.sh
sh Miniforge3-Linux-aarch64.sh -b -p ./miniforge
source miniforge/etc/profile.d/conda.sh
conda update -n base conda -y
and similarly the .bashrc routine:
arch=$(uname -i) # Get the CPU architecture
if [[ $arch == "aarch64" ]]; then
# Set variables and source scripts for aarch64
export CONDADIR=/nobackup/projects/bddur53/$USER/aarch64 # Update this with your <project> code.
source $CONDADIR/miniforge/etc/profile.d/conda.sh
fi
Problem
minicondawith miniforge as the recommended way of researchers using conda.conda-forgeas the default channel (which is also open source), and even includes Mamba for faster package resolution, while Miniconda can use Anaconda's default channels which have licensing restrictions.Suggestion:
The current installation method would just need to be replaced with something like:
and similarly the
.bashrcroutine: