How cjdearaujo/auto-mpg works
This repository is an automated Auto MPG regression pipeline: a push triggers data validation, model training and baseline comparison, then application checks and conditional publication of a downloadable model package. The runtime code also provides a prediction interface for saved models. The source excerpts show data loading, training, and prediction behavior; workflow orchestration and packaging links below are based on the README, since the workflow file was not excerpted.
Main entry points: ml-pipeline.yml ML workflow · train.py Train/validation split · data.py Load and validate · predict.py Prediction interface
Components 13
CI workflow
- ML workflow
.github/workflows/ml-pipeline.yml - Model package
.github/workflows/ml-pipeline.yml - Actions artifact
.github/workflows/ml-pipeline.yml
Data validation
- Auto MPG data
data/mpg.csv - Load and validate
src/data.py
Training and evaluation
- Train/validation split
src/train.py - Mean baseline
src/train.py - Random forest pipeline
src/train.py - MAE quality gate
src/train.py - Metrics report
src/train.py
Prediction
- Saved model
src/train.py - Prediction interface
src/predict.py
Other components
- Push to main
Connections 17
- Push to main → ML workflow: triggers
README.md - ML workflow → Load and validate: runs validation
README.md - Auto MPG data → Load and validate: loads
src/data.py - ML workflow → Train/validation split: runs training
README.md - Train/validation split → Load and validate: loads validated data
src/train.py - Train/validation split → Mean baseline: splits data for
src/train.py - Train/validation split → Random forest pipeline: splits data for
src/train.py - Mean baseline → MAE quality gate: reports MAE to
src/train.py - Random forest pipeline → MAE quality gate: reports MAE to
src/train.py - Random forest pipeline → Saved model: saves
src/train.py - MAE quality gate → Metrics report: writes report
src/train.py - ML workflow → Prediction interface: packages script
README.md - Prediction interface → Saved model: loads for prediction
src/predict.py - ML workflow → Model package: builds after checks
README.md - Model package → Saved model: includes
README.md - Model package → Metrics report: includes
README.md - ML workflow → Actions artifact: publishes on success
README.md