ML Pipeline Assignment

This repository contains a GAN (Generative Adversarial Network) implementation trained on the MNIST dataset, with MLflow experiment tracking and a GitHub Actions CI pipeline.

Project Structure

assignment/
├── student_A/
│   ├── student_a_gan.py        # Original GAN script (Student A)
│   ├── requirements.txt        # Python dependencies for Student A
│   ├── Dockerfile              # Docker image for Student A's script
│   └── env.yaml                # Conda environment definition
├── Student_B/
│   ├── student_a_gan.py        # Reproducibility-improved GAN script
│   └── report.md               # Student B's MLOps/SRE analysis report
├── student_a_gan_mlflow.py     # Enhanced GAN with MLflow experiment tracking (5 runs)
├── docker-compose.yml          # Docker Compose configuration
├── requirements.txt            # Root-level Python dependencies for CI
└── .github/
    └── workflows/
        └── ml_pipeline.yml     # GitHub Actions CI pipeline

CI / CD Pipeline

The GitHub Actions workflow (ml_pipeline.yml) runs on every push to any branch except main. It:

1. Checks out the code
2. Sets up Python 3.10
3. Installs dependencies from requirements.txt
4. Runs flake8 linting (max line length: 120)
5. Uploads README.md as a GitHub Actions artifact named project-doc

Running Locally

Student A — Original Script

cd student_A
pip install -r requirements.txt
python student_a_gan.py

MLflow Experiment Tracking (5 runs)

pip install -r requirements.txt
python student_a_gan_mlflow.py

View MLflow UI

mlflow ui
# Open http://localhost:5000

Docker

docker-compose up --build

Dependencies

Package	Version
pandas	2.1.1
numpy	1.26.0
torch	2.1.0
mlflow	latest
torchvision	latest
flake8	latest