cylinder_flow / scripts /validate_cylinder_flow_dataset.py
Zhongning's picture
Upload folder using huggingface_hub
f0652e2 verified
Raw
History Blame Contribute Delete
7.09 kB
#!/usr/bin/env python3
"""Validate the standardized DeepMind CylinderFlow dataset package."""
from __future__ import annotations
import argparse
import hashlib
import json
from pathlib import Path
EXPECTED_SPLITS = ["train", "valid", "test"]
EXPECTED_REQUIRED = {
"data/cylinder_flow/meta.json",
"data/cylinder_flow/train.tfrecord",
"data/cylinder_flow/valid.tfrecord",
"data/cylinder_flow/test.tfrecord",
"data/cylinder_flow/stats/edge_stats.json",
"data/cylinder_flow/stats/node_stats.json",
}
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 check_files(dataset_root: Path) -> dict:
sizes = {}
for rel_path in sorted(EXPECTED_REQUIRED):
path = dataset_root / rel_path
if not path.is_file():
fail(f"missing required file: {rel_path}")
if path.stat().st_size <= 0:
fail(f"empty required file: {rel_path}")
sizes[rel_path] = path.stat().st_size
return sizes
def check_metadata(data_dir: Path) -> dict:
meta = json.loads((data_dir / "meta.json").read_text(encoding="utf-8"))
if meta.get("simulator") != "comsol":
fail("meta.json simulator must be comsol")
if meta.get("trajectory_length") != 600:
fail("meta.json trajectory_length must be 600")
expected_fields = ["cells", "mesh_pos", "node_type", "velocity", "pressure"]
if meta.get("field_names") != expected_fields:
fail("meta.json field_names mismatch")
expected_shapes = {
"cells": [1, -1, 3],
"mesh_pos": [1, -1, 2],
"node_type": [1, -1, 1],
"velocity": [600, -1, 2],
"pressure": [600, -1, 1],
}
expected_dtypes = {
"cells": "int32",
"mesh_pos": "float32",
"node_type": "int32",
"velocity": "float32",
"pressure": "float32",
}
for key in expected_fields:
feature = meta.get("features", {}).get(key)
if not feature:
fail(f"meta.json missing feature {key}")
if feature.get("shape") != expected_shapes[key]:
fail(f"meta.json feature {key} shape mismatch")
if feature.get("dtype") != expected_dtypes[key]:
fail(f"meta.json feature {key} dtype mismatch")
return meta
def check_stats(stats_dir: Path) -> None:
edge = json.loads((stats_dir / "edge_stats.json").read_text(encoding="utf-8"))
node = json.loads((stats_dir / "node_stats.json").read_text(encoding="utf-8"))
for key in ["edge_mean", "edge_std"]:
if not isinstance(edge.get(key), list) or len(edge[key]) != 3:
fail(f"edge_stats.json {key} must be length 3")
for key in ["velocity_mean", "velocity_std", "velocity_diff_mean", "velocity_diff_std"]:
if not isinstance(node.get(key), list) or len(node[key]) != 2:
fail(f"node_stats.json {key} must be length 2")
for key in ["pressure_mean", "pressure_std"]:
if not isinstance(node.get(key), list) or len(node[key]) != 1:
fail(f"node_stats.json {key} must be length 1")
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 rel_path in sizes and sizes[rel_path] != item["size"]:
fail(f"required file size mismatch for {rel_path}")
if verify_sha256 and sha256_file(path) != item["sha256"]:
fail(f"sha256 mismatch for {rel_path}")
seen[rel_path] = item
missing = sorted(EXPECTED_REQUIRED - set(seen))
if missing:
fail(f"inventory missing required files: {missing}")
def check_tfrecord_first_record(data_dir: Path, meta: dict) -> None:
try:
import numpy as np
import tensorflow.compat.v1 as tf
except Exception as exc: # pragma: no cover - optional runtime dependency
print(f"[WARN] TensorFlow first-record check skipped: {exc}")
return
feature_dict = {k: tf.io.VarLenFeature(tf.string) for k in meta["field_names"]}
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")
features = tf.io.parse_single_example(raw, feature_dict)
velocity = np.frombuffer(features["velocity"].values[0].numpy(), dtype=np.float32)
pressure = np.frombuffer(features["pressure"].values[0].numpy(), dtype=np.float32)
mesh_pos = np.frombuffer(features["mesh_pos"].values[0].numpy(), dtype=np.float32)
cells = np.frombuffer(features["cells"].values[0].numpy(), dtype=np.int32)
if velocity.size % (meta["trajectory_length"] * 2) != 0:
fail(f"{split}.tfrecord velocity shape incompatible with metadata")
nodes = velocity.size // (meta["trajectory_length"] * 2)
if pressure.size != meta["trajectory_length"] * nodes:
fail(f"{split}.tfrecord pressure node count mismatch")
if mesh_pos.size != nodes * 2:
fail(f"{split}.tfrecord mesh_pos node count mismatch")
if cells.size % 3 != 0:
fail(f"{split}.tfrecord cells are not triangular")
print(f"[OK] {split}.tfrecord first record: nodes={nodes}, cells={cells.size // 3}")
def main() -> None:
args = parse_args()
dataset_root = Path(args.dataset_root).resolve()
data_dir = dataset_root / "data" / "cylinder_flow"
sizes = check_files(dataset_root)
meta = check_metadata(data_dir)
check_stats(data_dir / "stats")
verify_inventory(dataset_root, sizes, args.verify_sha256)
if not args.skip_tfrecord_read:
check_tfrecord_first_record(data_dir, meta)
print("[OK] CylinderFlow dataset validation passed")
if __name__ == "__main__":
main()