File size: 5,977 Bytes
94a8511 1650142 94a8511 1650142 94a8511 2de4806 94a8511 | 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 | """Is a session fully registered? Answers before you discover it is not.
Adding a session touches six places, and a prose checklist gets skipped. Each
omission has a characteristic silent failure:
calibration epoch missing -> a wrong-epoch projection looks like a slightly
miscalibrated rig, not like a bug. It shipped
that way once: 35-73 px off.
episodes.jsonl missing -> world_offset_m has nothing to read, so poses in
a redefined frame are used uncorrected
world residual missing -> a consumer assumes zero error
bad_frames / segments -> dropouts train as if they were data
splits.json stale -> the new session is entirely in train, and the
probe set can draw start frames the model saw
not validated -> nothing above is checked
python scripts/check_session_ready.py --task motherboard
python scripts/check_session_ready.py --task motherboard --date 2026-05-19
"""
from __future__ import annotations
import argparse
import json
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
sys.path.insert(0, str(Path(__file__).resolve().parents[2]))
from react_paths import release_root # noqa: E402
RELEASE = release_root()
def main() -> int:
ap = argparse.ArgumentParser()
ap.add_argument("--task", default="motherboard")
ap.add_argument("--date", default=None, help="check one session; default all")
a = ap.parse_args()
root = RELEASE / a.task
rows, problems = [], []
def note(ok, item, detail):
rows.append((ok, item, detail))
if not ok:
problems.append(item)
# 1 — a calibration epoch is declared for the task, and it is the right one
try:
from twm.calib_epoch import calib_dir, check_epoch, epoch_of
d = calib_dir(a.task, date=a.date)
try:
check_epoch(a.task, a.date)
note(True, "calibration epoch declared", f"{epoch_of(a.task, a.date)} at {d.name}")
except ValueError as e:
note(False, "calibration epoch declared", str(e)[:90])
except Exception as e:
note(False, "calibration epoch declared", f"{type(e).__name__}: {e}")
# 2 — episodes.jsonl covers every episode on disk
ep_file = root / "episodes.jsonl"
on_disk = sorted({f"{p.parent.name}/{p.stem}"
for p in (root / "meta").glob("*/*.parquet")})
if a.date:
on_disk = [e for e in on_disk if e.startswith(a.date)]
listed = set()
if ep_file.is_file():
listed = {json.loads(l)["episode"]
for l in ep_file.read_text().splitlines() if l.strip()}
miss = [e for e in on_disk if e not in listed]
note(not miss and bool(on_disk), "episodes.jsonl covers every episode",
f"{len(on_disk)} on disk, {len(miss)} unlisted"
+ (f": {miss[:3]}" if miss else ""))
# 3 — each episode declares a world offset (even if zero)
try:
from twm.calib_epoch import world_offset_m
bad = []
for e in on_disk:
dt, ep = e.split("/")
try:
world_offset_m(a.task, dt, ep, up_axis="z")
except Exception as ex:
bad.append(f"{e}: {type(ex).__name__}")
note(not bad, "every episode declares a world offset",
f"{len(on_disk)} episodes readable" + (f"; {bad[:2]}" if bad else ""))
except Exception as e:
note(False, "every episode declares a world offset", str(e)[:80])
# 4 — the world residual is declared per date, so a consumer can bound error
from twm.calib_epoch import world_residual
dates = sorted({e.split("/")[0] for e in on_disk})
nores = [d_ for d_ in dates if not world_residual(a.task, d_)]
note(not nores, "every session publishes a world residual",
f"{len(dates)} sessions: " + ", ".join(
f"{d_}(tilt {world_residual(a.task, d_).get('tilt_deg')})" for d_ in dates)
+ (f"; missing {nores}" if nores else ""))
# 5 — curation artefacts exist and cover the episodes
for name in ("bad_frames.json", "segments.json"):
f = root / name
if not f.is_file():
note(False, f"{name} present", "missing"); continue
j = json.loads(f.read_text())
cov = set(j["episodes"]) if name == "bad_frames.json" else \
{s["source_episode"] for s in j["segments"]}
m = [e for e in on_disk if e not in cov]
note(not m, f"{name} covers every episode",
f"{len(cov)} covered, {len(m)} uncovered" + (f": {m[:3]}" if m else ""))
# 6 — splits.json exists and is current
sf = root / "splits.json"
if not sf.is_file():
note(False, "splits.json present", "missing — run scripts/build_splits.py")
else:
sp = json.loads(sf.read_text())
m = [e for e in on_disk if e not in sp["episodes"]]
st = sp["stats"]
note(not m, "splits.json covers every episode",
f"test {st['test_fraction']*100:.1f}% guard {st['guard_fraction']*100:.1f}%, "
f"{len(m)} uncovered" + (f": {m[:3]}" if m else "")
+ f"; rebuild with scripts/build_splits.py" if m else
f"test {st['test_fraction']*100:.1f}% guard {st['guard_fraction']*100:.1f}%, "
f"{st['n_test_intervals']} intervals")
w = max(len(x) for _, x, _ in rows)
print()
for ok, item, detail in rows:
print(f" [{'ok' if ok else 'TODO'}] {item:<{w}} {detail}")
print(f"\nsession readiness: {len(rows)} items, {len(problems)} incomplete")
if problems:
print("\nnext: see docs/USAGE.md section 8 — 'Adding a session or a task'")
else:
print("run scripts/validate_all.py, then scripts/test_frame_consistency.py")
return 1 if problems else 0
if __name__ == "__main__":
raise SystemExit(main())
|