This is a short tutorial on how to set up a Conda environment and manage it on the HUJI computer cluster.

Initial Setup:

These four commands download the latest 64-bit version of the Linux installer of miniconda in the user repository, rename it to a shorter file name, silently install, and then delete the installer. Ideally it's better to replace to path with one in a lab folder since personal storage is very limited.

1mkdir -p ~/miniconda3
2wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
3bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
4rm ~/miniconda3/miniconda.sh

Finally, run the following which updates .zsrch and adds conda to PATH:

1~/miniconda3/bin/conda init zsh

And restart the shell (exec zsh).

Managing Environments:

For full documentation see https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html.

Create an environment:

1conda create -n <env_name> python=<python_version>
2
3OR
4
5# for custom path
6conda create -p <path/to/envname> python=<python_version>
7conda config --append envs_dirs <path/to>
8

Delete an environment:

1conda remove -n <env_name> --all

List environments:

1conda info --envs

Activate an environment (this is a prerequisite to all following commands):

1conda activate <env_name>

Deactivate an environment:

1conda deactivate

List packages in environment:

1conda list

Install package:

1conda install <package_name>
2
3OR
4
5pip install <package_name>

Running python:

1python <file_path.py>

Share an environment: 1. Export:

1conda env export > environment.yml

1. Import:

1conda env create -f environment.yml

Misc:

• For running Jupyter notebooks remotely with the kernel being a conda environment, activate the conda environment and run conda install ipykernel before creating the kernel. Other instruction for setting up IDE remote development are here. (In my experience VS code is easier to set up than PyCharm).
  • Note: Some VS code extensions need to be installed remotely (Jupyter, Python, PyLance)
  • Note: do not include large dataset folders or parent dataset within the remote VS code workspace or PyLance will crash.