This project is a lightweight GPU monitor designed for real-time web-based viewing of GPU server status. It provides detailed information on GPU status and processes utilizing the GPU. The container names are displayed for processes running inside Docker containers, facilitating GPU monitoring for servers shared by multiple users.
Go to the folder where you want to store your project, and clone this repository:
git clone git@github.com:senli1073/Docker-GPU-Monitor.git
Python >= 3.7 is required.
Then install numpy and Flask:
pip install -r requirements.txt
Edit the copyright information, page update interval, and port the website in config.py。
You can choose method (a) or (b) to run this project:
(a) Run in dev-mode:
bash ./start_dev_app.sh
(b) Run in production-mode:
gunicorn --daemon -w 4 --access-logfile=./access.log -b 127.0.0.1:80 app:app
Fire up a browser and go to http://<server_ip_address>/
Copyright Sen Li, 2024. Licensed under an MIT license.