Ein GitHub-Actions-Workflow für ML

GitHub Actions runs your scripts on GitHub's servers whenever something happens in your repo, like a push, so you never run CI by hand again. ⚙️

A workflow is a YAML file inside .github/workflows. GitHub reads it and follows your instructions step by step on each trigger.

The on key says when to run. Here it fires on every push, so each commit kicks off your ML checks automatically.

A workflow holds one or more jobs. Each job runs on a fresh virtual machine, so your pipeline always starts from a clean, predictable state.

Inside a job, steps run in order from top to bottom. Each step is either a reusable action or a plain shell command you write.

The first step almost always uses the checkout action to copy your repo onto the runner. Without it, the machine has no code to test.

For an ML repo you set up Python next, choosing the exact version so the runner matches the environment your team trains in.

Then you install your pinned packages so the runner has the same libraries you use locally. This makes every CI run reproducible.

Now add the ML checks: a lint step to catch style issues, then pytest to run your data and model tests on the fresh runner.

A final step can train on a small sample to prove the pipeline works end to end, keeping CI fast while still exercising real code.

Push your YAML and open the Actions tab. GitHub shows a live log of each step, with a green check when your ML pipeline passes. ✅

Where must a GitHub Actions workflow file live to be picked up?

A workflow YAML in .github/workflows defines triggers, jobs, and ordered steps. For ML you check out code, set up Python, install deps, then lint, test, and train.