ML-PROJET-WITH-DOCKER

This project is a complete Machine Learning pipeline that includes model training, batch predictions, and an interactive web app using Streamlit, DVC, and Docker. It also integrates with DagsHub for version control and experiment tracking.

---

Features

• Train a Gradient Boosting model on the Wine dataset
• Store and version data/models with DVC
• Run batch predictions on uploaded .csv files
• Use a clean and interactive Streamlit interface
• Fully reproducible thanks to Docker
• Use of a centralized configuration with config.yaml

---

📁 Project Structure

ML-PROJET-WITH-DOCKER/
│
├── app/
│   ├── training/
│   │   └── train_model.py              # Script to train and save the model
│   ├── model/
│   │   ├── load_model.py               # Load model from file
│   │   └── predictor.py                # Make predictions using the model
│   ├── streamlit_app.py                # Streamlit web application
│   ├── utils.py                        # Utility functions
│   └── dvc_utils.py                    # DVC utility integration
│
├── data/                               # Raw input data (tracked by DVC)
├── models/                             # Trained models (tracked by DVC)
│   └── gradiant_boosting_model.pkl     # Trained Gradient Boosting model
├── uploads/                            # Uploaded files via Streamlit
│
├── .dvc/                               # DVC internal files
├── .venv/                              # Python virtual environment (excluded)
├── .env                                # Environment variables
├── Dockerfile                          # Docker image definition
├── config.yaml                         # Centralized parameters and paths
├── requirements.txt                    # Python dependencies
├── dvc.yaml                            # DVC pipeline definition
├── dvc.lock                            # DVC stage locks for reproducibility
└── README.md                        

🐳 Docker Usage

🔧 Build the image

docker build -t ml-docker-app .

🔧 Run de Streamlit app

docker run -p 8501:8501 ml-docker-app

📦 DVC Pipeline

We use DVC to track data and model versions, and define reproducible pipelines.

Train model with DVC

dvc repro

➕ Add new stages (example)

dvc stage add -n train_model \
  -d app/training/train_model.py \
  -d config.yaml \
  -o models/gradiant_boosting_model.pkl \
  python3 app/training/train_model.py

DagsHub Integration

This project can be synced with DagsHub to manage:

✅ Data & model versioning

✅ Experiments tracking with MLflow

✅ Git/DVC repository hosting

https://dagshub.com/Herman-Motcheyo/ML-Project-Docker