nvitop

An interactive NVIDIA-GPU process viewer and beyond, the one-stop solution for GPU process management.

What is nvitop?

An interactive NVIDIA-GPU process viewer and beyond, the one-stop solution for GPU process management.

Overview of nvitop

What it is

nvitop is an interactive NVIDIA-GPU process viewer and resource monitor. It provides a colorful, informative interface that continuously updates device and process status. The tool is designed as a one-stop solution for GPU process management, offering both a monitor mode and a set of APIs for integration into other applications.

Key features

  • Informative and fancy output with colorized box drawing, showing more information than nvidia-smi.
  • Monitor mode with bar charts, history graphs, process sorting, filtering, signal sending, tree-view, environment variable screen, help screen, and mouse support.
  • Efficient: uses NVML Python bindings directly, supports sparse query with TTL cache, uses curses library, and gathers information asynchronously with multi-threading.
  • Portable: works on Linux and Windows, uses psutil for host process information, written in pure Python.
  • Integrable: provides low-level APIs for device, process, and host metrics, and ships a CUDA device selection tool nvisel.

Use cases

  • Real-time monitoring of NVIDIA GPU devices and processes in a terminal.
  • Managing GPU processes: filtering, sorting, sending signals, viewing environment variables.
  • Selecting CUDA-visible devices for deep learning researchers via the nvisel tool.
  • Building custom monitoring tools using the provided Python APIs.

Getting started

pip3 install --upgrade nvitop

After installation, run nvitop or python3 -m nvitop to start the interactive monitor. For a one-shot status query, use nvitop -1.

License and pricing

This project is open source under the Apache-2.0 license.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/xuehaipannvitop.svg)](https://yunodb.com/tools/xuehaipannvitop?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/xuehaipannvitop?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/xuehaipannvitop.svg" alt="Featured on YunoDB" height="20"></a>