How sqb-codes/ku-cse-mlops works
This repository is a coursework collection whose substantial product is an Iris-classification MLOps project. Its workflow trains and evaluates a classifier, rejects models below an accuracy threshold, logs and registers accepted models, then serves predictions through a FastAPI application that loads the champion model. Training and serving both use MLflow; training also saves a local artifact. The graph reflects the sampled source and README. The API app instance is commented out, so deployment and process startup are not shown.
Main entry points: train.py Training run · main.py FastAPI application · iris_random_forest_model.joblib Local model artifact
Components 13
Model development
- Iris dataset
- Training run
mlops-first-pipeline/train.py - Metrics and gate
mlops-first-pipeline/train.py - Local model artifact
mlops-first-pipeline/models/iris_random_forest_model.joblib
Model registry
- MLflow tracking and registry
- Registered classifier
mlops-first-pipeline/train.py - Champion alias
mlops-first-pipeline/train.py
Prediction service
- FastAPI application
mlops-first-pipeline/main.py - Model and version loading
mlops-first-pipeline/main.py - Prediction endpoint
mlops-first-pipeline/main.py - Health endpoints
mlops-first-pipeline/main.py
Other components
- Project user
- API client
Connections 12
- Training run → Iris dataset: loads
mlops-first-pipeline/train.py - Training run → Metrics and gate: fits and evaluates
mlops-first-pipeline/train.py - Training run → Local model artifact: saves artifact
mlops-first-pipeline/train.py - Training run → MLflow tracking and registry: logs run and model
mlops-first-pipeline/train.py - Training run → Registered classifier: registers model
mlops-first-pipeline/train.py - Training run → Champion alias: sets alias
mlops-first-pipeline/train.py - FastAPI application → Model and version loading: starts lifespan
mlops-first-pipeline/main.py - Model and version loading → MLflow tracking and registry: loads champion and version
mlops-first-pipeline/main.py - API client → Prediction endpoint: submits features
mlops-first-pipeline/main.py - Prediction endpoint → FastAPI application: routes through
mlops-first-pipeline/main.py - FastAPI application → Model and version loading: uses loaded model
mlops-first-pipeline/main.py - API client → Health endpoints: checks status
mlops-first-pipeline/main.py