File size: 5,932 Bytes
483ae68
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
#!/usr/bin/env python3
"""Validate the standardized DeepMind Lagrangian Water dataset package."""

from __future__ import annotations

import argparse
import hashlib
import json
from pathlib import Path


EXPECTED_SPLITS = {
    "train": 1000,
    "valid": 30,
    "test": 30,
}


def fail(message: str) -> None:
    raise SystemExit(f"[FAIL] {message}")


def sha256_file(path: Path) -> str:
    digest = hashlib.sha256()
    with path.open("rb") as f:
        for chunk in iter(lambda: f.read(1024 * 1024), b""):
            digest.update(chunk)
    return digest.hexdigest()


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser()
    parser.add_argument("--dataset-root", default=".", help="dataset package root")
    parser.add_argument(
        "--verify-sha256",
        action="store_true",
        help="recompute and verify SHA256 values from files_sha256.jsonl",
    )
    parser.add_argument(
        "--skip-tfrecord-read",
        action="store_true",
        help="skip optional TensorFlow first-record readability check",
    )
    return parser.parse_args()


def load_metadata(data_dir: Path) -> dict:
    meta_path = data_dir / "metadata.json"
    if not meta_path.is_file():
        fail(f"missing metadata file: {meta_path}")
    metadata = json.loads(meta_path.read_text(encoding="utf-8"))
    if metadata.get("dim") != 2:
        fail(f"expected metadata dim=2, got {metadata.get('dim')}")
    if metadata.get("sequence_length") != 1000:
        fail(f"expected sequence_length=1000, got {metadata.get('sequence_length')}")
    for key in ["vel_mean", "vel_std", "acc_mean", "acc_std"]:
        value = metadata.get(key)
        if not isinstance(value, list) or len(value) != 2:
            fail(f"metadata {key} must be a length-2 list")
    return metadata


def check_files(dataset_root: Path) -> dict:
    data_dir = dataset_root / "data" / "Water"
    if not data_dir.is_dir():
        fail(f"missing data directory: {data_dir}")
    sizes = {}
    for split in EXPECTED_SPLITS:
        path = data_dir / f"{split}.tfrecord"
        if not path.is_file():
            fail(f"missing TFRecord split: {path}")
        size = path.stat().st_size
        if size <= 0:
            fail(f"empty TFRecord split: {path}")
        sizes[str(path.relative_to(dataset_root))] = size
    if not (data_dir / "metadata.json").is_file():
        fail("missing data/Water/metadata.json")
    sizes["data/Water/metadata.json"] = (data_dir / "metadata.json").stat().st_size
    return sizes


def verify_inventory(dataset_root: Path, sizes: dict, verify_sha256: bool) -> None:
    inventory_path = dataset_root / "files_sha256.jsonl"
    if not inventory_path.is_file():
        fail("missing files_sha256.jsonl")
    seen = {}
    for line in inventory_path.read_text(encoding="utf-8").splitlines():
        if not line.strip():
            continue
        item = json.loads(line)
        rel_path = item["path"]
        path = dataset_root / rel_path
        if not path.is_file():
            fail(f"inventory path missing on disk: {rel_path}")
        if path.stat().st_size != item["size"]:
            fail(f"size mismatch for {rel_path}")
        if sizes.get(rel_path) != item["size"]:
            fail(f"required file size mismatch for {rel_path}")
        if verify_sha256:
            actual = sha256_file(path)
            if actual != item["sha256"]:
                fail(f"sha256 mismatch for {rel_path}")
        seen[rel_path] = item
    missing = sorted(set(sizes) - set(seen))
    if missing:
        fail(f"inventory missing required files: {missing}")


def check_tfrecord_first_record(data_dir: Path, metadata: dict) -> None:
    try:
        import numpy as np
        import tensorflow.compat.v1 as tf
    except Exception as exc:  # pragma: no cover - depends on runtime env
        print(f"[WARN] TensorFlow first-record check skipped: {exc}")
        return

    feature_description = {"position": tf.io.VarLenFeature(tf.string)}
    context_features = {
        "key": tf.io.FixedLenFeature([], tf.int64, default_value=0),
        "particle_type": tf.io.VarLenFeature(tf.string),
    }
    expected_steps = metadata["sequence_length"] + 1
    dim = metadata["dim"]
    for split in EXPECTED_SPLITS:
        record_iter = iter(tf.data.TFRecordDataset(str(data_dir / f"{split}.tfrecord")).take(1))
        try:
            raw = next(record_iter)
        except StopIteration:
            fail(f"{split}.tfrecord contains no records")
        context, features = tf.io.parse_single_sequence_example(
            raw,
            context_features=context_features,
            sequence_features=feature_description,
        )
        position = np.frombuffer(features["position"].values[0].numpy(), dtype=np.float32)
        if position.size % (expected_steps * dim) != 0:
            fail(f"{split}.tfrecord first record position shape is incompatible with metadata")
        particle_type = np.frombuffer(context["particle_type"].values[0].numpy(), dtype=np.int64)
        num_particles = position.size // (expected_steps * dim)
        if particle_type.shape[0] != num_particles:
            fail(f"{split}.tfrecord particle_type length does not match position particles")
        print(
            f"[OK] {split}.tfrecord first record: "
            f"position_shape=({expected_steps}, {num_particles}, {dim}), "
            "position_dtype=float32, particle_type_dtype=int64"
        )


def main() -> None:
    args = parse_args()
    dataset_root = Path(args.dataset_root).resolve()
    data_dir = dataset_root / "data" / "Water"
    sizes = check_files(dataset_root)
    metadata = load_metadata(data_dir)
    verify_inventory(dataset_root, sizes, args.verify_sha256)
    if not args.skip_tfrecord_read:
        check_tfrecord_first_record(data_dir, metadata)
    print("[OK] Lagrangian Water dataset validation passed")


if __name__ == "__main__":
    main()