A real-time NVIDIA GPU monitoring web application built with Flask and Chart.js.

• Real-time GPU utilization and memory usage graphs
• Dark/light mode toggle
• Detailed GPU statistics including:
  • GPU utilization percentage
  • Memory usage
  • Temperature
  • Running processes
• Color-coded indicators for quick status assessment
• Responsive design that works on desktop and mobile
• Auto-refreshing data every second

• NVIDIA GPU
• NVIDIA drivers and nvidia-smi tool
• Docker and Docker Compose
• NVIDIA Container Toolkit

1. Clone this repository:

git clone <repository-url>
cd gpu-monitor

2. Start the application using Docker Compose:

docker compose up --build

3. Open your web browser and navigate to:

http://localhost:5000

The application is containerized using Docker with NVIDIA GPU support. The configuration includes:

• Base image: nvidia/cuda:12.3.1-base-ubuntu22.04
• Python 3 environment
• Gunicorn WSGI server
• Automatic container restart policy
• GPU device passthrough

• /api/data - Historical GPU utilization and memory usage data
• /api/stats - Current detailed GPU statistics
• /api/processes - List of processes currently using the GPU

To run the application in development mode without Docker:

1. Install requirements:

pip install -r requirements.txt

2. Run the Flask application:

python app.py

MIT License