AGENTS.md

General instructions and guidelines for automatic code agents like Github Copilot for interpreting and generating code in this repository.

This repository contains course material for the Machine Learning Operations (MLOps) course at the Technical University of Denmark (DTU). The audience for this repository is therefore primarily computer science students that have some prior experience with programming in Python and machine learning, and is looking to learn the basics of MLOps. Below is a brief overview of course content as of right now:

Proper coding environments, code organization, good coding practices, code and data version control, reproducible and containerized environments, reproducible experiment management, debugging tools, code profiling, large scale collaborative experiment logging and monitoring, unit testing, continuous integration, continuous machine learning, cloud infrastructure, cloud based machine learning, distributed data loading and training, optimization methods for inference, local and cloud based deployment, monitoring of deployed applications.

Project structure

The course is setup as a mkdocs project. Importantly, the structure is flattened such that all course material is in the root of the repository. This means that:

• The course is divided into several sessions, named s1_*, s2_*, etc. Each session folder contains multiple *.md files which are different learning modules within that session. In addition the pages/ folder contains additional pages for the course website.

• For each session folder there is a subdir called exercise_files/ which contains the relevant exercise files (e.g. Jupyter Notebooks, Python scripts, data files, etc.) for all modules in that session.

• Figures for all pages are stored in the figures/ folder.

• An additional folder tools/ contains various utility scripts used for course management and should not be modified unless explicitly stated.

Development environment

• The project uses uv for dependency management, meaning that you should use uv run to execute commands
  • instead of python script.py, use uv run script.py
  • instead of pip install package, use uv add package.
    • If the dependency is is only needed for development, use uv add --dev package
    • If the dependency is only needed for optional features, use uv add --group extra package
  • instead of command ... use uv run command ...
• Additionally the project uses invoke for task management, so you can run tasks defined in tasks.py using uv run invoke <task_name>. Run uv run invoke --list to see all available tasks.
• Project uses pre-commit for git hooks. Make sure to run the pre-commit hooks after making changes to the codebase: uv run pre-commit

Material mkdocs specific instructions

The project uses the material for mkdocs theme for mkdocs. You can find the documentation for this theme here: https://squidfunk.github.io/mkdocs-material/reference/

That said here is some general guidelines for features that are used in this repository:

Admonitions

Admonitions are used to highlight important information in the course material, for example sometimes code snippets and solutions are inside admonitions. Example of admonitions used in the course material:

!!! example "Simple script"

    ```python linenums="1" title="sklearn_cloud_functions.py"
    --8<-- "s7_deployment/exercise_files/sklearn_cloud_functions.py"
    ```

And a collapsible admonition:

??? success "Solution"

    The line instructs Python to create an executable script named `train` that executes the `main` function within
    the `train.py` file, located in the `my_project` package.

Code snippets

As showcased in the admonition example above, code snippets are often included from external files using the --8<-- "path/to/file" syntax. This ensures that the code snippets in the documentation are always in sync with the actual code files.

Content tabs

Content tabs are used to group alternatives together, for example different ways to execute a command:

=== "Using uv"
    ```bash
    uv run python script.py
    ```
=== "Using python directly"
    ```bash
    python script.py
    ```