rskit-bench — ML Benchmarking Framework
ML benchmarking framework: evaluators, metrics, reports, and visualization.
Features
Evaluator<L>trait +EvaluatorFuncclosure wrapper +FromProvideradapterBenchRunner— orchestrates evaluation with configurable concurrency- Classification metrics: accuracy, precision, recall, F1, AUC-ROC
- Multi-branch comparison for A/B model testing
- Reports: JSON, CSV, Markdown, JUnit, Vega-Lite visualizations
FileRunStoragefor persistent result storage and regression detection- ROC curves, confusion matrices, score distribution charts
BenchClock/FixedClockfor deterministic timestamps and durations in tests
Benchmark orchestration accepts injected clock and storage implementations. Production CLIs can choose SystemClock and FileRunStorage; tests and reproducible harnesses should inject FixedClock and an in-memory or tempdir storage implementation.
Usage
[]
= "0.2.0-alpha.5"
use ;
use AppResult;
use Arc;
let eval = new;
println!;
let runner = new
.register
.with_clock;
// Load samples and run with RunOptions