gpuview

What gpuview is

gpuview is a web dashboard for watching GPU health and usage across one or more servers. It displays temperature, memory, utilization, and users, and can combine statistics from multiple registered GPU hosts in a single dashboard. The view refreshes every three seconds, with interactive tooltips, temperature-based color coding, and pause and resume controls. gpuview also provides a REST API for statistics from the main host and aggregated statistics from registered hosts. It runs on top of gpustat, which it installs from PyPI, and requires Python 3.9 or higher. The dashboard listens by default on 0.0.0.0 at port 9988; command-line options let you configure the host and port. On Linux systems with systemd, it can run as a permanent background service, and installing that service requires sudo privileges. The project states that it has no user authentication and omits sensitive details such as usernames by default. The --safe-zone option enables reporting all details, including usernames; --exclude-self prevents other dashboards from receiving the current machine’s statistics. gpuview is free and MIT-licensed. Its README describes GPU monitoring for deep learning practitioners and notes that it can run on a non-GPU machine, such as a laptop, to monitor remote GPU servers.

Who it suits

gpuview suits deep learning practitioners who want a lightweight dashboard for GPU status across multiple servers. It can also suit someone using a non-GPU machine, such as a laptop, to monitor remote GPU hosts through a browser or REST API.

Limits

gpuview has no user authentication. By default, it omits sensitive details such as usernames; the --safe-zone option enables reporting all details, including usernames.

Getting started

1. Visit the gpuview GitHub project.
2. Use a system with Python 3.9 or higher.
3. Install gpuview, which installs gpustat from PyPI.
4. Run the dashboard and configure its host or port with command-line options if needed.
5. Register GPU hosts to aggregate their statistics.
6. Optionally configure a systemd service on Linux; installation requires sudo privileges.

Verdict

Choose gpuview if you want a free dashboard and API for monitoring GPU statistics across registered hosts. Look elsewhere if your setup requires user authentication or cannot accommodate Python 3.9 or higher.