Gaiatheia 🌱

Gaiatheia is a machine learning benchmarking tool that tracks carbon emissions using CodeCarbon and integrates with Prometheus + Grafana for real-time monitoring.

Features

• Tracks CO₂ emissions for ML models using CodeCarbon
• Pushes emissions data to Prometheus via Pushgateway
• Visualizes emissions in Grafana

---

Setup

1️ Clone the Repository

First, clone the repository:

git clone https://github.com/yourusername/gaiatheia.git
cd gaiatheia

---

2️ Install Dependencies

If you’re not using Docker, install the required Python dependencies:

pip install -r requirements.txt

Ensure CodeCarbon is installed:

pip install codecarbon

---

3️ Run the Machine Learning Models

Run the scripts to track the emissions. There are two ways to monitor the system:

• Export Data to CSV format
• Use Prometheus & Grafana to Track(currently in MVP phase)

There are 3 simple experiments attached to it, you can run those or you can use the Template.py file and follow the instruction to build and run your own model.

Run Experiment 1

python experiment_1.py

---

4️ Setup Prometheus, Pushgateway, and Grafana with Docker

To track and visualize emissions, use Docker Compose:

docker compose up

or if you are using podman

podman compose up

This will start:

• Prometheus (http://localhost:9090)
• Pushgateway (http://localhost:9091)
• Grafana (http://localhost:3000)

---

5️ View Emissions in Grafana(Currently in MVP)

1. Open Grafana: http://localhost:3000
2. Login with default credentials:

   Username: admin  
   Password: admin
   

3. Add Prometheus as a Data Source (http://prometheus:9090)
4. Import a new dashboard and use the metric:

   codecarbon_emissions_kg
   

---

Contributing

Want to improve Gaiatheia? Feel free to open an issue or submit a pull request!