NVIDIA GPU Module

System monitoring for NVIDIA GPU utilization and performance metrics.

Overview

This module provides GPU monitoring capabilities when NVIDIA hardware is available, useful for tracking resource usage during LLM inference and multimodal generation tasks.

Feature Flag

Enabled via Cargo feature:

[features]
nvidia = []

Functions

has_nvidia_gpu()

Check if NVIDIA GPU is available:

#![allow(unused)]
fn main() {
if nvidia::has_nvidia_gpu() {
    // GPU acceleration available
}
}

Returns true if nvidia-smi command succeeds.

get_gpu_utilization()

Get current GPU and memory utilization:

#![allow(unused)]
fn main() {
let util = nvidia::get_gpu_utilization()?;
let gpu_percent = util.get("gpu");      // GPU compute utilization %
let mem_percent = util.get("memory");   // GPU memory utilization %
}

get_system_metrics()

Get combined CPU and GPU metrics:

#![allow(unused)]
fn main() {
let metrics = nvidia::get_system_metrics()?;
println!("CPU: {}%", metrics.cpu_usage);
if let Some(gpu) = metrics.gpu_usage {
    println!("GPU: {}%", gpu);
}
}