scripts/validate_evo2_dataset.py

#!/usr/bin/env python3
import argparse
import gzip
import hashlib
import sys
from pathlib import Path

REQUIRED_FASTA = [
    "chr20.fa",
    "chr21.fa",
    "chr22.fa",
    "chr20_21_22.fa",
    "chr20.fa.gz",
    "chr21.fa.gz",
    "chr22.fa.gz",
]

REQUIRED_PREPROCESSED_PREFIXES = [
    "chr20_21_22_uint8_distinct_byte-level_train",
    "chr20_21_22_uint8_distinct_byte-level_val",
    "chr20_21_22_uint8_distinct_byte-level_test",
]


def sha256(path: Path) -> str:
    h = hashlib.sha256()
    with path.open("rb") as f:
        for block in iter(lambda: f.read(1024 * 1024), b""):
            h.update(block)
    return h.hexdigest()


def read_sha_manifest(path: Path) -> dict[str, str]:
    result = {}
    for line in path.read_text().splitlines():
        if not line.strip():
            continue
        digest, filename = line.split(None, 1)
        result[filename.strip()] = digest
    return result


def read_sizes(path: Path) -> dict[str, int]:
    result = {}
    for line in path.read_text().splitlines():
        if not line.strip():
            continue
        filename, size = line.split("\t", 1)
        result[filename] = int(size)
    return result


def check_fasta_head(path: Path) -> bool:
    opener = gzip.open if path.suffix == ".gz" else open
    with opener(path, "rt", encoding="utf-8", errors="replace") as f:
        for line in f:
            line = line.strip()
            if line:
                return line.startswith(">")
    return False


def main() -> int:
    parser = argparse.ArgumentParser(description="Validate the OneScience Evo2 mini genome dataset.")
    parser.add_argument("--dataset-root", default="data_mini", help="Dataset root in the dataset repository.")
    parser.add_argument("--package-root", default=".", help="Dataset package root. Default: current directory.")
    parser.add_argument("--skip-sha256", action="store_true", help="Skip SHA256 checks.")
    args = parser.parse_args()

    root = Path(args.package_root).resolve()
    dataset_root = (root / args.dataset_root).resolve() if not Path(args.dataset_root).is_absolute() else Path(args.dataset_root)
    genome_root = dataset_root / "genome_data"
    preprocessed = genome_root / "preprocessed_data"
    errors: list[str] = []

    if not genome_root.is_dir():
        errors.append(f"missing genome_data directory: {genome_root}")

    for rel in REQUIRED_FASTA:
        path = genome_root / rel
        if not path.is_file():
            errors.append(f"missing FASTA file: genome_data/{rel}")
        else:
            try:
                if not check_fasta_head(path):
                    errors.append(f"FASTA header not found in first record: genome_data/{rel}")
            except Exception as exc:
                errors.append(f"cannot read FASTA file genome_data/{rel}: {exc}")

    for prefix in REQUIRED_PREPROCESSED_PREFIXES:
        bin_path = preprocessed / f"{prefix}.bin"
        idx_path = preprocessed / f"{prefix}.idx"
        if not bin_path.is_file():
            errors.append(f"missing preprocessed bin: {bin_path}")
        elif bin_path.stat().st_size <= 0:
            errors.append(f"empty preprocessed bin: {bin_path}")
        if not idx_path.is_file():
            errors.append(f"missing preprocessed idx: {idx_path}")
        elif idx_path.stat().st_size <= 0:
            errors.append(f"empty preprocessed idx: {idx_path}")

    size_manifest = root / "metadata" / "data_mini.files.tsv"
    if size_manifest.is_file():
        for rel, expected_size in read_sizes(size_manifest).items():
            path = dataset_root / rel
            if not path.is_file():
                errors.append(f"file listed in size manifest is missing: {rel}")
            elif path.stat().st_size != expected_size:
                errors.append(f"size mismatch for {rel}: {path.stat().st_size} != {expected_size}")
    else:
        errors.append("missing metadata/data_mini.files.tsv")

    if not args.skip_sha256:
        sha_manifest = root / "metadata" / "data_mini.sha256"
        if sha_manifest.is_file():
            for rel, expected_digest in read_sha_manifest(sha_manifest).items():
                path = dataset_root / rel
                if path.is_file() and sha256(path) != expected_digest:
                    errors.append(f"sha256 mismatch for {rel}")
        else:
            errors.append("missing metadata/data_mini.sha256")

    if errors:
        for error in errors:
            print(f"[FAIL] {error}", file=sys.stderr)
        return 1

    print("[OK] Evo2 mini dataset files, FASTA readability, preprocessed splits, sizes and hashes passed validation.")
    return 0


if __name__ == "__main__":
    raise SystemExit(main())