nvgpu

What it is and what it does

nvgpu wraps nvidia-smi and NVIDIA's NVML library to provide a simpler interface for querying GPU state on multi-GPU machines. It solves the problem of selecting which GPU to use when multiple devices are available and frameworks would otherwise claim all memory, blocking other processes. The package offers both a command-line tool (nvgpu available, nvgpu list) and a Python API (nvgpu.available_gpus(), nvgpu.gpu_info()) to check memory usage, temperature, and running processes per device.

It also includes an optional distributed monitoring mode: agents on each node expose GPU status as JSON over REST, and a master node aggregates and displays the cluster status in a web dashboard. This is useful for coordinating GPU allocation across a cluster of machines. The package is marked Alpha and has been dormant since its latest release, so it may not be actively maintained or tested against recent driver and Python versions.

Use it for:

• Select an available GPU for a single training job before launching a framework to avoid memory conflicts.
• Monitor GPU utilization and temperature across a cluster of machines via a web dashboard for capacity planning.
• Query which GPUs are free and their memory state programmatically to implement custom job scheduling logic.
• Display a colored table of all GPUs, their users, and running processes for quick status checks on a multi-user system.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Queries and monitors NVIDIA GPUs on a machine, providing a CLI and Python API to check availability, memory usage, and running processes, plus optional multi-node monitoring via REST agents.

Yes, if you have NVIDIA GPUs and need to programmatically select available devices or monitor a cluster—but with caution. The package is dormant (last release 2023-03-30) and classifiers list Python 2 support, raising questions about compatibility with modern Python and driver versions. High install friction and no recent maintenance mean you should verify it works in your environment before relying on it in production.

Install

nvgpu on PyPI

pip install nvgpu

uv add nvgpu

poetry add nvgpu

Installing nvgpu

Before you install

High install friction: the package is distributed as a source tarball and has no runtime dependencies listed, suggesting it may require system-level NVIDIA tools to be present. Maintenance is dormant—last release was 2023-03-30, with no recent commits.

License in practice

MIT license is permissive and poses no restrictions on use, modification, or distribution in proprietary or open-source projects.

Quickstart

pip install nvgpu

import nvgpu
available = nvgpu.available_gpus()
info = nvgpu.gpu_info()

Requires nvidia-smi and NVIDIA GPU drivers to be installed on the system; CUDA_VISIBLE_DEVICES environment variable may need to be set for GPU selection to work.

Verify before relying

• Whether the package works with current NVIDIA driver versions and modern Python releases (classifiers list Python 2, which is end-of-life).
• Whether NVML Python bindings are bundled or must be installed separately as a system dependency.
• Current compatibility with Flask versions if using the web application agent feature.

Package facts

Evidence: nvgpu-0.10.0.tar.gz

Tags

More Monitoring packages

Further reading