File size: 2,777 Bytes
715cc5a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 | from __future__ import annotations
import math
from typing import Any
def query_metrics(candidates: list[dict[str, Any]], gold_ids: set[str] | None, cutoffs: tuple[int, ...] = (5, 10, 30, 100)) -> dict[str, Any]:
if gold_ids is None:
return {
"has_gold": None,
**{f"gold_at_{k}": None for k in cutoffs},
"mrr": None,
"ndcg_at_10": None,
}
has_gold = bool(gold_ids)
ranks = [idx + 1 for idx, row in enumerate(candidates) if row["candidate_id"] in gold_ids]
first_rank = min(ranks) if ranks else None
metrics = {"has_gold": has_gold}
for cutoff in cutoffs:
metrics[f"gold_at_{cutoff}"] = bool(first_rank is not None and first_rank <= cutoff)
metrics["mrr"] = 1.0 / first_rank if first_rank else 0.0
dcg = 0.0
for idx, row in enumerate(candidates[:10], start=1):
if row["candidate_id"] in gold_ids:
dcg += 1.0 / math.log2(idx + 1)
ideal_hits = min(len(gold_ids), 10)
idcg = sum(1.0 / math.log2(idx + 1) for idx in range(1, ideal_hits + 1))
metrics["ndcg_at_10"] = dcg / idcg if idcg > 0 else 0.0
return metrics
def aggregate_metrics(dataset: str, split: str, method: str, rows: list[dict[str, Any]], notes: str) -> dict[str, Any]:
evaluable = [row for row in rows if row["metrics"].get("has_gold") is not None]
if not evaluable:
return {
"Dataset": dataset,
"Split": split,
"Method": method,
"R@5": "N/A",
"R@10": "N/A",
"R@30": "N/A",
"R@100": "N/A",
"MRR": "N/A",
"nDCG@10": "N/A",
"candidate_count_avg": round(sum(len(row["candidates"]) for row in rows) / max(1, len(rows)), 4),
"gold_coverage": "N/A",
"Notes": notes,
}
with_gold = [row for row in evaluable if row["metrics"].get("has_gold")]
denom = max(1, len(with_gold))
def avg_bool(key: str) -> float:
return sum(1 for row in with_gold if row["metrics"].get(key)) / denom
return {
"Dataset": dataset,
"Split": split,
"Method": method,
"R@5": round(avg_bool("gold_at_5"), 6),
"R@10": round(avg_bool("gold_at_10"), 6),
"R@30": round(avg_bool("gold_at_30"), 6),
"R@100": round(avg_bool("gold_at_100"), 6),
"MRR": round(sum(float(row["metrics"].get("mrr", 0.0)) for row in with_gold) / denom, 6),
"nDCG@10": round(sum(float(row["metrics"].get("ndcg_at_10", 0.0)) for row in with_gold) / denom, 6),
"candidate_count_avg": round(sum(len(row["candidates"]) for row in rows) / max(1, len(rows)), 4),
"gold_coverage": round(len(with_gold) / max(1, len(evaluable)), 6),
"Notes": notes,
}
|