How devendra-5470/mlops-118 works
This is a house-price prediction application with a batch data-preparation and model-training workflow, followed by an HTTP prediction service and a small web interface. Cleaning uploads versioned CSV data to S3; training reads a processed CSV, evaluates a random-forest model, and logs parameters, metrics, and the model to MLflow. The service loads the champion model, validates prediction inputs, and serves results. Source excerpts show these interactions; the README is empty, so broader capabilities are not documented.
Main entry points: serve.py FastAPI service · train.py Random forest training · data-clean-versioning.py CSV cleaning · index.html Prediction page
Components 13
Data preparation
- Raw house CSV
- CSV cleaning
data-clean-versioning.py - Processed data store
Model training
- Random forest training
train.py - MLflow tracking
- Champion model
Prediction service
- FastAPI service
serve.py - Feature validation
serve.py - Price inference
serve.py
Web interface
- Prediction page
static/index.html - Page styling
static/style.css
Other components
- Data operator
- Prediction client
Connections 12
- Data operator → Raw house CSV: provides
- CSV cleaning → Raw house CSV: reads
data-clean-versioning.py - CSV cleaning → Processed data store: uploads
data-clean-versioning.py - Random forest training → Processed data store: fetches
train.py - Random forest training → MLflow tracking: logs run
train.py - Random forest training → Champion model: logs model
train.py - FastAPI service → Champion model: loads champion
serve.py - Prediction client → FastAPI service: requests prediction
- FastAPI service → Feature validation: validates input
serve.py - FastAPI service → Price inference: invokes model
serve.py - FastAPI service → Prediction page: serves page
serve.py - FastAPI service → Page styling: mounts static files
serve.py