Web4Style / scripts /prepare_release.py
Jerryxrc's picture
Initial dataset release
311c1fe verified
Raw History Blame Contribute Delete
2.09 kB
from __future__ import annotations
import json
from pathlib import Path
import pandas as pd
REPO_ROOT = Path(__file__).resolve().parents[2]
RELEASE_ROOT = Path(__file__).resolve().parents[1]
SOURCE_JSON = REPO_ROOT / "dataset_with_images.json"
OUTPUT_DIR = RELEASE_ROOT / "data"
PARQUET_PATH = OUTPUT_DIR / "train-00000-of-00001.parquet"
JSONL_PATH = OUTPUT_DIR / "train.jsonl"
def load_records() -> list[dict]:
with SOURCE_JSON.open("r", encoding="utf-8") as f:
raw_records = json.load(f)
records = []
for raw in raw_records:
record = {
"id": raw["id"].strip(),
"category": raw["category"].strip(),
"website": raw["website"].strip(),
"brief": raw["brief"].strip(),
"content": raw["content"].strip(),
"function": raw["function"].strip(),
"style": raw["style"].strip(),
}
records.append(record)
return sorted(records, key=lambda item: item["id"])
def validate_records(records: list[dict]) -> None:
ids = [record["id"] for record in records]
if len(ids) != len(set(ids)):
raise ValueError("Duplicate ids detected in release data.")
for record in records:
for key, value in record.items():
if not isinstance(value, str) or not value:
raise ValueError(f"Invalid value for {key!r} in record {record['id']}.")
def write_jsonl(records: list[dict]) -> None:
with JSONL_PATH.open("w", encoding="utf-8") as f:
for record in records:
f.write(json.dumps(record, ensure_ascii=False) + "\n")
def write_parquet(records: list[dict]) -> None:
frame = pd.DataFrame.from_records(records)
frame.to_parquet(PARQUET_PATH, index=False)
def main() -> None:
OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
records = load_records()
validate_records(records)
write_jsonl(records)
write_parquet(records)
print(f"Wrote {len(records)} records to {JSONL_PATH}")
print(f"Wrote {len(records)} records to {PARQUET_PATH}")
if __name__ == "__main__":
main()