“It worked in Jupyter but crashed on the cluster.” Almost always a dependency problem.
System Package Managers
apt (Debian/Ubuntu)
sudo apt update
sudo apt install htop python3-pip git vim
sudo apt remove htop
sudo apt upgrade
apt search "mesh generator"yum/dnf (Red Hat/CentOS/Fedora)
sudo dnf install htop
sudo dnf updatebrew (macOS)
brew install wget python@3.12
brew upgrade
brew list
brew search paraviewTip: On a new Linux server: sudo apt update && sudo apt upgrade -y first.
Python: Don’t pip install Globally
# BAD
pip install numpy pandas
# GOOD — use virtual environmentVirtual Environments (venv)
cd my-project/
python3 -m venv venv
source venv/bin/activate
pip install numpy pandas
pip list
pip freeze > requirements.txt
deactivateReproducing:
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txtKey takeaway: Every Python project gets its own virtual environment. Always. No exceptions.
conda / mamba
conda create -n myproject python=3.12
conda activate myproject
conda install numpy scipy matplotlib
conda install -c conda-forge fenics
conda env export > environment.yml
conda env create -f environment.yml
conda deactivate
conda env list
conda env remove -n myprojectTip: Use conda for compiled scientific packages. Use venv+pip for pure Python. conda-install first, pip-install second.
Checking What’s Installed
pip list
pip show numpy
conda list
which python
python --versionWarning: Always check which python — you might be using system Python instead of your venv.
Locking Dependencies
pip freeze > requirements.txt
conda env export > environment.ymlTry It Yourself
Create an isolated Python venv for a data project, install pandas and requests, freeze the requirements, deactivate, and verify that system Python doesn’t have pandas.
Hint
An ImportError is expected — that’s the point of isolation.
Quiz
Why should you avoid pip install without a virtual environment?
Answer
B) It installs globally, creating version conflicts between projects.