furiosa-smi-py 2024.2.0

No project description provided

Furiosa System Management Interface Python Binding

Overview

Furiosa System Management Interface, is a programmatic interface for managing and monitoring FuriosaAI NPUs.

The interface provides the following API modules, each designed to offer distinct functionalities for managing and monitoring NPU devices. These modules enable developers to access essential hardware information, topology details, system-wide information, and performance metrics.

Each module provides the following features:

• Device module provides NPU device discovery and information including device specification and liveness.

• Topology module provides the topology status within a system including device-to-device link type and p2p accessibility.

• System module provides system-wide information of each NPU device, including firmware version and driver version.

• Performance module provides the device performance metrics including power consumption, temperature, and utilization.

Installation

Furiosa-smi-py is available on the Python Package Index (PyPI). We recommend installing furiosa-smi-py via pip:

pip install furiosa-smi-py

Once installed, you can import the furiosa_smi_py module:

import furiosa_smi_py

Usage

To get started with Furiosa-smi-py, simply import the furiosa_smi_py module and utilize its functions to interact with NPU devices. The package provides various methods to access the NPU device information and status. For more detailed documentation, check out the API documentation.

from furiosa_smi_py import list_devices

devices = list_devices() # Retrieve a list of NPU devices in the system.

for device in devices:
    device_info = device.device_info() # Acquire information about the NPU device.
    print("Device Info")
    print(f"\t\tDevice Arch: {device_info.arch()}")
    print(f"\t\tDevice Cores: {device_info.core_num()}")

    ... # You can use other APIs. Please refer to the documentation.

The expected output is as below.

Device Info
		Device Arch: Rngd
		Device Cores: 8

		...