---
id: pyrsmi
version: "1.1.0"
license: MIT License
license_treatment: permissive
maintenance: aging
---

# pyrsmi — Python System Management Library for AMD GPUs

License: permissive · Maintenance: aging · Downloads: 116.3K/mo

## What it is and what it does

pyrsmi is a Python wrapper around AMD's System Management Interface (SMI) library for monitoring and controlling AMD Instinct MI-series GPUs. It exposes functions to enumerate devices, query memory (total, used, reserved pages), measure utilization and power consumption, and retrieve device identifiers and metadata. Version 1.0+ migrated from the deprecated rocm-smi-lib to the modern amdsmi backend, maintaining backward-compatible APIs while adding access to PCIe metrics and extended telemetry.

The package is designed for system administrators, monitoring tools, and GPU-aware applications running on Linux with ROCm 6.0+. It requires explicit initialization and shutdown calls and returns -1 or empty strings on error. The optional amdsmi Python package is recommended for correct device naming and improved compatibility in containerized environments, though the core library functions work without it.

Use it for:

- Monitor GPU utilization, memory usage, and power consumption in real-time dashboards or logging systems.
- Enumerate and identify AMD Instinct GPUs in a system for resource allocation and scheduling.
- Track memory allocation patterns and reserved pages for debugging GPU memory issues.
- Build automated thermal or power management tools that query fan RPM and average power draw.
- Integrate GPU telemetry into cluster management or containerized workload orchestration systems.
- Retrieve device UUIDs and identifiers in formats compatible with ROCm, raw, or NVIDIA conventions.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Provides Python bindings to monitor and manage AMD Instinct MI-series GPUs via the AMD SMI Library, offering device enumeration, memory tracking, utilization metrics, and power monitoring.

Yes, if you have AMD Instinct MI-series GPUs on a ROCm 6.0+ Linux system and need Python-based GPU monitoring or management. The package has no runtime dependencies, permissive licensing, and active maintenance. The aging status reflects time since release rather than abandonment. Install friction is low. Verify that your ROCm installation includes libamd_smi.so and consider installing the optional amdsmi package for full device naming support.

## Install

pip install pyrsmi
uv add pyrsmi
poetry add pyrsmi

## Installing pyrsmi

Before you install: Low install friction with no runtime dependencies. Maintenance status is aging—last commit 2025-12-15 with 254 days since release—but the package is actively maintained on GitHub. Requires ROCm 6.0+ and the libamd_smi.so library to be present on the system.

License in practice: MIT License (permissive) allows unrestricted use, modification, and distribution in both open and closed-source projects with minimal obligations.

Quickstart:

pip install pyrsmi

from pyrsmi import rocml

rocml.smi_initialize()
num_gpus = rocml.smi_get_device_count()
for i in range(num_gpus):
    name = rocml.smi_get_device_name(i)
    util = rocml.smi_get_device_utilization(i)
    print(f"GPU {i}: {name}, Utilization: {util}%")
rocml.smi_shutdown()

Requires ROCm 6.0 or later with libamd_smi.so library installed; Linux OS only; AMD Instinct MI-series GPU required; Python 3.9 or later.

Verify before relying:

- Whether the amdsmi package is truly optional or required for core functionality beyond device naming.
- Performance improvement claims ('10-50x faster UUID retrieval') and backward compatibility scope.
- Thread-safety guarantees and behavior under concurrent access patterns.
- Support matrix for specific MI-series models beyond the tested MI350X, MI300X, MI250X, MI210.

## Package facts

- License: MIT License (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: aging
- Downloads: 116.3K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

AMD GPU monitoring python, rocm system management interface, AMD instinct MI series control, GPU memory utilization tracking, AMD SMI library bindings, GPU power and thermal monitoring, AMD compute device management, gpu-management, rocm, system-monitoring

[View on SkillFed](https://skillfed.io/packages/pyrsmi) · [View on PyPI](https://pypi.org/project/pyrsmi/)