Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
Polish
Size:
10K - 100K
License:
File size: 3,028 Bytes
aae3233 | 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 | """Validate DynaWord Parquet schema and Slayer binary pack invariants."""
from __future__ import annotations
import argparse
import json
from pathlib import Path
import struct
import pyarrow.parquet as pq
import tiktoken
ROOT = Path(__file__).resolve().parents[1]
SCHEMA = ["id", "text", "source", "added", "created", "token_count", "license", "author"]
SPLITS = {"train": "data/train.parquet", "valid": "data/validation.parquet", "test": "data/test.parquet"}
PACK_DIRS = {"train": "train", "valid": "valid", "test": "test"}
def validate(root: Path) -> dict:
cl100k = tiktoken.get_encoding("cl100k_base")
manifest = json.loads((root / "manifest.json").read_text(encoding="utf-8"))
result = {}
for split, relative in SPLITS.items():
parquet_files = [root / relative]
if not parquet_files[0].is_file():
raise ValueError(f"missing Parquet files for {split}")
rows = 0
tokens = 0
ids = set()
for parquet_path in parquet_files:
table = pq.read_table(parquet_path)
if table.column_names != SCHEMA:
raise ValueError(f"schema mismatch in {parquet_path}: {table.column_names}")
for row in table.to_pylist():
if row["id"] in ids:
raise ValueError(f"duplicate id in {split}: {row['id']}")
ids.add(row["id"])
measured = len(cl100k.encode_ordinary(row["text"]))
if row["token_count"] != measured:
raise ValueError(f"cl100k token mismatch for {row['id']}: {row['token_count']} != {measured}")
rows += 1
tokens += row["token_count"]
expected = manifest["splits"][split]
if rows != expected["documents"] or tokens != expected["tokens"]:
raise ValueError(f"Parquet manifest mismatch for {split}: {rows}/{tokens} vs {expected}")
pack_dir = root / "slayer_v6" / PACK_DIRS[split]
bin_name = "train.bin" if split == "train" else f"{split}.bin"
bin_path = pack_dir / bin_name
idx_path = pack_dir / "docs.idx"
raw_size = bin_path.stat().st_size
if raw_size % 2:
raise ValueError(f"binary token file is not uint16-sized: {bin_path}")
idx_raw = idx_path.read_bytes()
if len(idx_raw) != rows * 8:
raise ValueError(f"index count mismatch for {split}")
final_offset = struct.unpack_from("<Q", idx_raw, len(idx_raw) - 8)[0]
if final_offset * 2 != raw_size:
raise ValueError(f"final document offset mismatch for {split}")
result[split] = {"documents": rows, "parquet_tokens": tokens, "packed_tokens_including_eos": final_offset}
return result
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("--root", type=Path, default=ROOT / "data" / "cnc_dynaword")
args = parser.parse_args()
print(json.dumps(validate(args.root), ensure_ascii=False, indent=2))
if __name__ == "__main__":
main()
|