Detailed explanation of monitoring NVIDIA GPU usage under Linux

When using TensorFlow for deep learning, insufficient video memory often occurs, so we hope to be able to check the GPU usage at any time. If you have an Nvidia GPU, you can do it with just one line of command on the command line.

1. Display current GPU usage

Nvidia comes with a command line tool called nvidia-smi, which displays video memory usage:

$ nvidia-smi

Output:

2. Periodically output GPU usage

But sometimes we want to know not only the GPU usage at a fixed moment, but also its trend. In this case, we want to output it periodically, such as updating the display every 10 seconds. At this time, you need to use the watch command to periodically execute the nvidia-smi command.

Understand the functions of watch:

$ whatis watch
 watch(1) - execute a program periodically, showing output fullscreen

Function: Execute a command periodically and display the output.

The basic usage of watch is:

$ watch [options] command

The most commonly used parameter is -n, which specifies how many seconds to execute the command.

Monitor video memory: We set it to display the video memory status every 10 seconds:

$ watch -n 10 nvidia-smi

The display is as follows:

In this way, as long as the command line window is open, you can refresh it every ten seconds. Isn’t it convenient?

If we want to periodically perform other command line operations, we can simply replace the following nvidia-smi. So Cool !

The above is the full content of this article. I hope it will be helpful for everyone’s study. I also hope that everyone will support 123WORDPRESS.COM.