rust_metrics

rust_metrics is an ML evaluation toolkit that brings Torchmetrics-style metrics to Rust. Each metric implements the same incremental Metric trait, so you can feed batched predictions over time and ask for the final score when ready.

Getting started

Add the crate to your project:

cargo add rust_metrics
# or enable the BERT-based similarity metric
cargo add rust_metrics --features text-bert

Evaluate batched predictions:

use rust_metrics::{BinaryAccuracy, BinaryAuroc, Metric};

let predictions = [0, 1, 1, 0];
let targets = [0, 1, 0, 0];

let mut accuracy = BinaryAccuracy::new();
accuracy.update((&predictions, &targets)).unwrap();
assert_eq!(accuracy.compute(), Some(0.75));

let scores = [0.9, 0.6, 0.1, 0.2];
let mut auroc = BinaryAuroc::new(0); // 0 => compute exact ROC AUC
auroc.update((&scores, &targets.map(|t| t as f64))).unwrap();
assert!(auroc.compute().unwrap() > 0.6);

Available metrics

Classification

• BinaryAccuracy, MulticlassAccuracy, MultilabelAccuracy
• BinaryPrecision, BinaryRecall
• BinaryHinge
• BinaryAuroc (exact or binned ROC AUC)

Regression

• MeanSquaredError
• MeanAbsoluteError
• R2Score

Text

• Bleu with optional smoothing and arbitrary n-gram depth
• EditDistance with sum or mean reduction
• SentenceEmbeddingSimilarity (requires the text-bert feature) backed by fastembed. This metric embeds each sentence pair with lightweight BERT embeddings and reports cosine similarity scores.

Feature flags