When deploying Python code for data science projects, you have several options:
- Regular Python scripts: You just deploy a bunch of Python scripts to the server and run them. This is the simplest form of deployment, but it requires a lot of manual preparation: you need to install all required packages, fill in configuration files, and so on. While those actions can be automated by using tools such as Ansible (https://www.ansible.com/), it's not recommended to use this form of deployment for anything but the simplest projects with no long-term maintainability goals.
- Python packages: Creating a Python package using a setup.py file is a much more convenient way to package Python code. Tools such as PyScaffold provide ready-to-use templates for Python packages, so you won't need to spend much time structuring your project. In the case of Python packages, Ansible...
