MNIST Classification with CI/CD Pipeline

This project implements a lightweight Convolutional Neural Network (CNN) for MNIST digit classification with a complete CI/CD pipeline using GitHub Actions. The model is designed to achieve >95% accuracy in a single epoch while maintaining less than 25,000 parameters.

Model Architecture

The model uses a compact but effective CNN architecture:

4 Convolutional layers with batch normalization and ReLU activation
Max pooling layers for dimensionality reduction
Single fully connected layer for classification
Batch normalization for training stability

Layer details:

Conv2d(1→16, 3×3) + BatchNorm + ReLU
Conv2d(16→32, 3×3) + BatchNorm + ReLU + MaxPool
Conv2d(32→32, 3×3) + BatchNorm + ReLU
Conv2d(32→16, 3×3) + BatchNorm + ReLU + MaxPool
MaxPool
Fully Connected (144→10)

Key features:

Input: 28×28 grayscale images
Output: 10 classes (digits 0-9)
Parameters: <25,000
Training accuracy: >95% in one epoch

Project Structure

.
├── .github
│ └── workflows
│ └── ml-pipeline.yml
├── model
│ ├── init.py
│ └── network.py
├── tests
│ ├── init.py
│ └── test_model.py
├── .gitignore
├── README.md
├── requirements.txt
└── train.py

Setup and Installation

Clone the repository:

bash
git clone <your-repo-url>
cd <repo-name>

Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate

Install dependencies:

bash pip install -r requirements.txt

Usage

Training

To train the model:

bash python train.py

The trained model will be saved in the models/ directory with a timestamp suffix.

Testing

To run the tests:

bash pytest tests/test_model.py -v

The tests verify:

Model architecture (input/output dimensions)
Parameter count (<25,000)
Training accuracy (>95%)

CI/CD Pipeline

The project includes a GitHub Actions workflow that automatically:

Sets up a Python environment
Installs dependencies
Runs all tests
Uploads the trained model as an artifact

The pipeline is triggered on every push to the repository.

Requirements

Python 3.8+
PyTorch
torchvision
pytest

Model Artifacts

Trained models are saved with timestamps in the format: mnist_model_YYYYMMDD_HHMMSS.pth

Contributing

Fork the repository
Create your feature branch (git checkout -b feature/amazing-feature)
Commit your changes (git commit -m 'Add some amazing feature')
Push to the branch (git push origin feature/amazing-feature)
Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

MNIST Dataset: LeCun et al.
PyTorch Documentation and Tutorials

Gokkulnath/GithubActionMLTestingDemo