gom is a CLI tool that displays a human-readable table with GPU usage information. Think nvidia-smi, but minimalist and pretty.

It also shows per-container GPU usage information if Docker is installed.

Use the package manager pip to install gom.

gom show displays a table with GPU usage information.

gom watch displays a table with GPU usage information and updates it every second.

Compare the output of gom show and nvidia-smi. I hope you'll agree that gom produces more clear and helpful output (ex. it breaks usage down across the 4 running Docker containers), while nvidia-smi is long and complex (I couldn't even screenshot the whole thing).

You may need to install a different version of pynvml depending on your CUDA version.