This repository contains a simple GPU monitoring script for mutliple heterogeneous GPU servers.
When there's multiple heterogeneous GPU servers, it was quite inconvenient to ssh into each server and use nvidia-smi to see which GPUs were idle. This script simplifies the process by only displaying the utilization and memory information for all the GPUs on each server.
The output of the script is shown below.
- sshpass
- This script relies on
nvidia-smi
Run the script from the command line using bash:
bash monitor_gpus.sh