std.ml — AI/ML runtime_

import std.ml as ml

Evaluation metrics

std.ml exposes production evaluation helpers for model gates:

• metrics_classification(labels: int, predictions: int) -> string
• metrics_regression(expected: int, predicted: int) -> string
• metrics_ranking(relevance: int, scores: int, top_k: int) -> string
• metrics_generation(output: string, reference: string) -> string
• serving_metrics(latencies_ms: int, requests: int, errors: int) -> string
• evaluation_report(path: string, name: string, classification: string, regression: string, ranking: string, generation: string, serving: string) -> string

The metric functions return deterministic JSON payloads. evaluation_report writes a versioned machine-readable JSON report and a human-readable .txt companion report.

import std.ml as ml
import std.tensor as tensor

let labels = tensor.arange(0, 4, 1)
let predicted = tensor.arange(0, 4, 1)
let classification = ml.metrics_classification(labels, predicted)

---