Available on crate feature
eval only.Expand description
Baseline storage for regression detection.
Provides persistence of evaluation metric snapshots and comparison against baselines to detect regressions.
§Example
ⓘ
use adk_eval::BaselineStore;
use std::collections::HashMap;
let store = BaselineStore::new(".eval-baseline.json");
// Save current metrics as baseline
let mut metrics = HashMap::new();
let mut case_metrics = HashMap::new();
case_metrics.insert("accuracy".to_string(), 0.95);
metrics.insert("case_1".to_string(), case_metrics);
store.save("my_eval_set", &metrics).unwrap();
// Check for regressions on a later run
let regressions = store.check_regressions(&metrics, 0.05).unwrap();
assert!(regressions.is_empty());Structs§
- Baseline
- Baseline file content containing metric snapshots.
- Baseline
Store - Manages baseline persistence and regression detection.
- Regression
- A regression detected between baseline and current run.