DMMONA – Multi-Objective Meta-Optimizer with Neural Architecture Adaptation

Version: 1.0.0

Overview

DMMONA is a Python framework that optimizes ML model training on resource-constrained desktops. It continuously monitors system resources and uses a reinforcement learning (RL) meta controller to dynamically adjust hyperparameters, model architecture, and computation precision. This ensures efficient, adaptive, and stable training even on everyday hardware.

Key Features

• Real-Time Resource Monitoring: Tracks CPU and memory usage with psutil and forecasts availability via a moving average.
• RL-Based Meta Controller: A lightweight PyTorch model that generates adjustment signals from resource metrics.
• Dynamic Architecture Adaptation: Automatically prunes or expands network layers based on meta signals.
• Adaptive Precision Switching: Chooses the optimal precision mode (fp32, mixed, or quantized) according to current resources.
• Modular Design: Easily integrates into existing ML pipelines.
• Centralized Logging: Detailed logs are output to the console and saved to a log file.

Project Structure

DMMONA/
├── README.md                   # Project overview and usage instructions.
├── requirements.txt            # Python dependencies.
├── config/
│   └── config.yaml             # User-configurable parameters.
├── src/
│   ├── __init__.py             # Package initializer.
│   ├── main.py                 # Entry point: loads config and starts training.
│   ├── resource_monitor.py     # Monitors system resources and forecasts availability.
│   ├── meta_controller.py      # RL-based meta controller that outputs adjustment signals.
│   ├── architecture_adaptation.py  # Simulates dynamic model architecture adaptation.
│   ├── adaptive_precision.py       # Selects computation precision mode based on resources.
│   ├── training_scheduler.py       # Coordinates the training loop.
│   └── logger.py               # Sets up centralized logging.
├── tests/                      # Unit tests for each module.
│   ├── test_resource_monitor.py
│   ├── test_meta_controller.py
│   ├── test_architecture_adaptation.py
│   └── test_adaptive_precision.py
├── docs/
│   └── design_documentation.md # Detailed design documentation.
└── notebooks/
    └── exploration.ipynb       # Notebook for experiments and prototyping.

Installation

Prerequisites

• Python 3.7 or higher
• Git (optional, for cloning the repository)

Setup Instructions

1. Clone the Repository:

   git clone https://github.com/your-username/dmmona.git
   cd dmmona
   

2. Create & Activate a Virtual Environment:

   • Windows:

     python -m venv venv
     .\venv\Scripts\activate
     

   • macOS/Linux:

     python3 -m venv venv
     source venv/bin/activate
     

3. Install Dependencies:

   pip install -r requirements.txt
   

Usage

1. Configure the Project: Edit config/config.yaml to set your dataset paths, training parameters, resource limits, and adaptation thresholds.

2. Launch the Training Process:

   python src/main.py -c config/config.yaml
   

   The training loop will monitor system resources, adapt the model architecture, and adjust precision settings while logging progress.

3. Monitor Logs:

   • Real-time logs appear in the console.
   • Detailed logs are saved to dmmona.log (configured in src/logger.py).

Packaging & Distribution (Optional)

To package DMMONA as a Python package (for PyPI):

1. Build the Package:

   python setup.py sdist bdist_wheel
   

2. Upload with Twine:

   pip install twine
   twine upload dist/*
   

Users can then install DMMONA via:

pip install dmmona

Contributing

Contributions are welcome! To contribute:

1. Fork the Repository.
2. Clone Your Fork:

   git clone https://github.com/your-username/dmmona.git
   cd dmmona
   

3. Create a Branch:

   git checkout -b feature-or-bugfix-name
   

4. Commit & Push Your Changes:

   git add .
   git commit -m "Description of changes"
   git push origin feature-or-bugfix-name
   

5. Submit a Pull Request on GitHub.

License

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