use-ml-metric

Metric name, value, and direction metadata primitives for RustUse.

Experimental

use-ml-metric is experimental while use-ml remains below 0.3.0.

Example

use use_ml_metric::{MlClassificationMetric, MlMetricDirection, MlMetricName, MlMetricValue};

let name = MlMetricName::new("accuracy")?;
let value = MlMetricValue::new(0.93)?;

assert_eq!(name.as_str(), "accuracy");
assert_eq!(value.value(), 0.93);
assert_eq!(MlClassificationMetric::Accuracy.direction(), MlMetricDirection::HigherIsBetter);
# Ok::<(), use_ml_metric::MlMetricError>(())

Scope

• Metric names, finite metric values, directions, aggregations, and metric categories.
• Generic classification, regression, ranking, clustering, and generation metric labels.
• Obvious metric-direction helpers.

Non-goals

• Complete metric calculation implementations in v0.1.
• LLM-specific metrics such as faithfulness, groundedness, toxicity, refusal quality, jailbreak resistance, instruction following, or answer relevance.

License

Licensed under either Apache-2.0 or MIT.