File size: 4,279 Bytes
3f316da | 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 | #!/usr/bin/env python3
"""Audit LOB ordering and volume ranges in preprocessed NPZ windows."""
from __future__ import annotations
import argparse
import csv
import json
from pathlib import Path
import numpy as np
def audit_file(path: Path) -> dict:
data = np.load(path, allow_pickle=False)
x = data["x"]
bid_p = x[:, :, :10][:, :, ::-1]
ask_p = x[:, :, 10:20]
volume = x[:, :, 20:40]
spread_bad = bid_p[:, :, 0] >= ask_p[:, :, 0]
bid_bad = (bid_p[:, :, 1:] >= bid_p[:, :, :-1]).any(axis=-1)
ask_bad = (ask_p[:, :, 1:] <= ask_p[:, :, :-1]).any(axis=-1)
any_price_bad = spread_bad | bid_bad | ask_bad
return {
"stock": path.stem.replace("_pilot", ""),
"path": str(path),
"n_windows": int(x.shape[0]),
"seq_len": int(x.shape[1]),
"features": int(x.shape[2]),
"price_min": float(np.nanmin(x[:, :, :20])),
"price_max": float(np.nanmax(x[:, :, :20])),
"volume_min": float(np.nanmin(volume)),
"volume_max": float(np.nanmax(volume)),
"volume_negative_rate": float((volume < 0).mean()),
"target_spread_cross_rate": float(spread_bad.mean()),
"target_bid_order_bad_rate": float(bid_bad.mean()),
"target_ask_order_bad_rate": float(ask_bad.mean()),
"target_any_price_bad_rate": float(any_price_bad.mean()),
}
def write_csv(path: Path, rows: list[dict]) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
with path.open("w", newline="", encoding="utf-8") as fh:
writer = csv.DictWriter(fh, fieldnames=list(rows[0].keys()))
writer.writeheader()
writer.writerows(rows)
def write_md(path: Path, rows: list[dict], title: str) -> None:
lines = [f"# {title}", ""]
lines += [
"| stock | windows | normalized volume <0 rate* | target spread bad | target strict bid bad | target strict ask bad | target any price bad |",
"|---|---:|---:|---:|---:|---:|---:|",
]
for row in rows:
lines.append(
f"| `{row['stock']}` | {row['n_windows']} | {row['volume_negative_rate']:.4f} | "
f"{row['target_spread_cross_rate']:.4f} | {row['target_bid_order_bad_rate']:.4f} | "
f"{row['target_ask_order_bad_rate']:.4f} | {row['target_any_price_bad_rate']:.4f} |"
)
lines.extend(
[
"",
"\\* Negative normalized volume is expected after training-only affine "
"normalization and is not a raw-volume validity failure.",
]
)
path.write_text("\n".join(lines) + "\n", encoding="utf-8")
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--processed-dir", type=Path, default=Path("experiments/tbd_revision/outputs/mixed_windows"))
parser.add_argument(
"--stocks",
nargs="+",
help="Asset ids. If omitted, infer every asset from manifest.json.",
)
parser.add_argument("--out-name", default="mixed_windows_5stock_preprocessed_lob_validity")
parser.add_argument("--out-dir", type=Path, default=Path("experiments/tbd_revision/outputs/summary"))
return parser.parse_args()
def main() -> None:
args = parse_args()
if args.stocks:
stocks = args.stocks
else:
manifest_path = args.processed_dir / "manifest.json"
if manifest_path.exists():
stocks = sorted(
json.loads(manifest_path.read_text(encoding="utf-8")).keys()
)
else:
stocks = sorted(
path.stem.removesuffix("_pilot")
for path in args.processed_dir.glob("*_pilot.npz")
)
if not stocks:
raise SystemExit(f"No assets found under {args.processed_dir}")
rows = []
for stock in stocks:
path = args.processed_dir / f"{stock}_pilot.npz"
if not path.exists():
raise SystemExit(f"Missing NPZ for {stock}: {path}")
rows.append(audit_file(path))
csv_path = args.out_dir / f"{args.out_name}.csv"
md_path = args.out_dir / f"{args.out_name}.md"
write_csv(csv_path, rows)
write_md(md_path, rows, args.out_name)
print(f"wrote {csv_path}")
print(f"wrote {md_path}")
if __name__ == "__main__":
main()
|