Miniconda on DARWIN

Miniconda is available through the VALET packages on Caviness clusters. To check the versions,

$ vpkg_versions miniconda

Available versions in package (* = default version):

[/opt/shared/valet/2.1/etc/miniconda.vpkg_yaml]
miniconda  A free, miniature installation of Anaconda Distribution
* 25.1.1.2 Miniconda py39 25.1.1-2 for Linux (x86_64)

To load it, use

$ vpkg_require miniconda/25.1.1.2
Adding package `miniconda/25.1.1.2` to your environment

You can create a new Conda environment with a specific package and in a specific path. For example, to create an environment in a specific path, such as your /lustre/workgroup directory,

[(my_workgroup:user)@login01 ~]$ conda create -p /lustre/workgroup/sw/myenv <my_package>

After creating the Conda environments, you can activate the local conda environment using,

[(my_workgroup:user)@login01 ~]$ conda activate /lustre/workgroup/sw/myenv

You can install the desired packages in the current environment, for example,

(/lustre/workgroup/sw/myenv)[(my_workgroup:user)@login01 ~]$ conda install <my_package>

To list packages in the currently active environment, use

(/lustre/workgroup/sw/myenv)[(my_workgroup:user)@login01 ~]$ conda list
# packages in environment at /lustre/workgroup/sw/myenv:
#
# Name                    Version                   Build  Channel
_libgcc_mutex             0.1                 conda_forge    conda-forge
_openmp_mutex             4.5                       2_gnu    conda-forge
absl-py                   2.3.1                    pypi_0    pypi
astunparse                1.6.3                    pypi_0    pypi
bzip2                     1.0.8                hda65f42_8    conda-forge
ca-certificates           2025.10.5            hbd8a1cb_0    conda-forge
certifi                   2025.10.5                pypi_0    pypi

To deactivate the environment, do:

(/lustre/workgroup/sw/myenv)[(my_workgroup:user)@login01 ~]$ conda deactivate
[(my_workgroup:user)@login01 ~]$

To remove an environment, run:

[(my_workgroup:user)@login01 ~]$ conda remove -p /lustre/workgroup/sw/myenv --all

To remove the unused packages and caches, do

[(my_workgroup:user)@login01 ~]$ conda clean --all

In the terminal window, to view the environments available to you, use

[(my_workgroup:user)@login01 ~]$ conda env list
# conda environments:
#
base                   /opt/shared/miniconda/25.1.1.2
                      */lustre/workgroup/sw/myenv

After activating the desired environment, export the current environment,

[(my_workgroup:user)@login01 ~]$ conda activate /lustre/workgroup/sw/myenv
(/lustre/workgroup/sw/myenv)[(my_workgroup:user)@login01 ~]$ conda export --file=myenv.yaml
name: /lustre/workgroup/sw/myenv
channels:
  - conda-forge
dependencies:
  - _libgcc_mutex=0.1=conda_forge
  - _openmp_mutex=4.5=2_gnu
  - bzip2=1.0.8=hda65f42_8
  - ca-certificates=2025.10.5=hbd8a1cb_0
  - cuda-version=11.8=h70ddcb2_3
  - cudatoolkit=11.8.0=h4ba93d1_13
  - cudnn=8.9.7.29=hbc23b4c_3 
...
...
 
prefix: /lustre/workgroup/sw/myenv  

Then you can duplicate the same environment with

(/lustre/workgroup/sw/myenv)[(my_workgroup:user)@login01 ~]$ conda deactivate
[(my_workgroup:user)@login01 ~]$ conda create -p /lustre/workgroup/sw/myenv_2 -f myenv.yml

Examples documented as recipes to be used for specific installations, including using a workgroup directory as well as creating VALET packages for these environments and job scripts setup for batch runs, but can also perhaps help others in solving similar installation dilemmas.

• Building a TensorFlow Python Virtual Environment