Skip to content

Available Templates

ExpOps ships with two built-in project templates. Use the --template flag with expops create to start from either one.

sklearn-basic

A minimal, self-contained project for getting started with ExpOps. All ML logic lives in custom @step-decorated Python functions.

Best for: Learning the platform, experimenting with custom model code.

expops create my-project --template sklearn-basic
expops run my-project --local

What's included:

  • configs/project_config.yaml — basic pipeline config with a local SQLite backend
  • src/models/model.py — custom train/evaluate steps using @step
  • src/charts/plot_metrics.py — a @chart() function for static PNG output
  • requirements.txt — sklearn and standard dependencies
  • data/train.csv — small example dataset

See the sklearn-basic template guide for a full walkthrough.

sklearn-component

A template built around the sklearn component library and matplotlib component library. All ML logic is driven by component: keys in the config — no custom Python step code required.

Best for: Standard sklearn workflows, quickly trying different estimators, learning the component system.

expops create my-project --template sklearn-component
expops run my-project --local

What's included:

  • configs/project_config.yaml — pipeline using sklearn.* and matplotlib.* components
  • src/plot_iterative_loss.js — Chart.js dynamic chart for MLP training loss
  • requirements.txt — expops-sklearn and expops-matplotlib
  • data/train.csv — small example dataset

To swap the estimator, change the class name in the component: string and update parameters as needed:

component: "sklearn.RandomForestClassifier.fit_predict"
parameters:
  n_estimators: 100

premier-league

A comprehensive example project that demonstrates the full platform: distributed execution, dynamic JS charts, multiple environments, and a GCP backend.

Best for: Understanding advanced configuration, reference for production-style setups.

expops create my-project --template premier-league

What's included:

  • Full cluster config (configs/compute_config.yaml) for SLURM/Dask execution
  • Multiple pipeline environments
  • Dynamic Chart.js visualisation (src/plot_metrics.js)
  • GCP backend configuration (Firestore + GCS)

See the Premier League template guide for details.