Datasets:
File size: 4,313 Bytes
195f84f | 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 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 | #!/usr/bin/env python3
"""Validate the public SABRE-Prior release tree before upload."""
from __future__ import annotations
import json
from collections import Counter, defaultdict
from pathlib import Path
from typing import Any
EXPECTED = {
"context": {"questions": 400, "images": 200},
"texture": {"questions": 400, "images": 200},
"attribute": {"questions": 100, "images": 100},
"language": {"questions": 100, "images": 100},
}
PROBES = {
"context": {"base_source", "base_target", "edited_source", "edited_target"},
"texture": {"base_normal", "base_counterfactual", "edited_normal", "edited_counterfactual"},
}
def load_jsonl(path: Path) -> list[dict[str, Any]]:
return [json.loads(line) for line in path.read_text(encoding="utf-8").splitlines() if line.strip()]
def main() -> int:
root = Path(__file__).resolve().parent
all_ids: set[str] = set()
totals = Counter()
for subset, expected in EXPECTED.items():
subset_root = root / "data" / subset
metadata_path = subset_root / "metadata.jsonl"
rows = load_jsonl(metadata_path)
if len(rows) != expected["questions"]:
raise ValueError(f"{subset}: expected {expected['questions']} questions, found {len(rows)}")
ids = [str(row.get("id") or "") for row in rows]
if any(not item_id for item_id in ids) or len(ids) != len(set(ids)):
raise ValueError(f"{subset}: IDs must be present and unique")
overlap = all_ids.intersection(ids)
if overlap:
raise ValueError(f"ID appears in multiple subsets: {sorted(overlap)[0]}")
all_ids.update(ids)
referenced: set[Path] = set()
for row in rows:
if row.get("subset") != subset:
raise ValueError(f"{row['id']}: incorrect subset")
file_name = Path(str(row.get("file_name") or ""))
if file_name.is_absolute() or ".." in file_name.parts:
raise ValueError(f"{row['id']}: unsafe file_name")
image_path = subset_root / file_name
if not image_path.is_file():
raise FileNotFoundError(f"{row['id']}: {image_path}")
referenced.add(image_path.resolve())
image_files = {
path.resolve()
for path in (subset_root / "images").iterdir()
if path.is_file() and not path.name.startswith(".")
}
if len(image_files) != expected["images"]:
raise ValueError(f"{subset}: expected {expected['images']} images, found {len(image_files)}")
if image_files != referenced:
raise ValueError(f"{subset}: image directory and metadata references differ")
if subset in PROBES:
by_pair: dict[str, set[str]] = defaultdict(set)
for row in rows:
if str(row["answer"]).lower() not in {"yes", "no"}:
raise ValueError(f"{row['id']}: invalid yes/no answer")
by_pair[str(row.get("pair_id") or "")].add(str(row.get("probe") or ""))
if len(by_pair) != 100 or any(probes != PROBES[subset] for probes in by_pair.values()):
raise ValueError(f"{subset}: expected 100 complete four-probe pairs")
elif subset == "attribute":
if any(not str(row["answer"]).isdigit() for row in rows):
raise ValueError("attribute: every public answer must be an integer count")
elif subset == "language":
for row in rows:
options = row.get("options")
if not isinstance(options, dict) or set(options) != {"A", "B", "C", "D"}:
raise ValueError(f"{row['id']}: invalid options")
if row["answer"] not in options or row.get("answer_text") != options[row["answer"]]:
raise ValueError(f"{row['id']}: answer and answer_text disagree")
totals["questions"] += len(rows)
totals["images"] += len(image_files)
print(f"{subset}: {len(rows)} questions, {len(image_files)} images [ok]")
if totals != Counter({"questions": 1000, "images": 600}):
raise ValueError(f"Unexpected totals: {dict(totals)}")
print("SABRE-Prior: 1,000 questions, 600 images [valid]")
return 0
if __name__ == "__main__":
raise SystemExit(main())
|