MLflow Demo

This project demonstrates a minimal workflow using MLflow integrated with DagsHub and Evidently. It covers experiment tracking, model registry, and drift detection in a single, reproducible pipeline.

Project Structure

mlflow-demo/
├── data/
├── notebooks/
│   └── exploratory.ipynb
├── src/
│   ├── data_processing.py
│   ├── model_training.py
│   ├── drift_detection.py
│   ├── model_testing.py
│   └── pipeline.py
├── environment.yml
├── main.py
└── README.md

Setup

Create the Conda environment and install dependencies:

conda env create -f environment.yml
conda activate mlflow

Running the Demo

Execute the full pipeline with:

python main.py

The script initializes MLflow tracking with DagsHub:

import dagshub

dagshub.init(repo_owner="yahiaehab10", repo_name="MLFlow_demo", mlflow=True)
mlflow.set_tracking_uri("https://dagshub.com/yahiaehab10/MLFlow_demo.mlflow")

Metrics, parameters and artifacts (including an Evidently drift report) are logged to the remote MLflow server. Open the DagsHub interface to explore runs and the model registry.

Reproducibility

All dependencies are captured in environment.yml. The dataset used is the built-in Iris dataset from scikit-learn for simplicity.

Drift Detection

src/drift_detection.py uses Evidently to generate an HTML data drift report comparing the training data to the evaluation data. The report is logged as an MLflow artifact for easy inspection.

Collaboration

Since the tracking URI points to DagsHub, everyone with access to the repository can view experiments remotely, facilitating collaboration and model lifecycle management.