File size: 4,761 Bytes
06c11b0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations

import h5py
import numpy as np
import pytest

from tests._shared.dataset_generation import DatasetCase

pytestmark = pytest.mark.dataset


def _make_case(env_id: str, base_seed: int) -> DatasetCase:
    return DatasetCase(
        env_id=env_id,
        episode=0,
        base_seed=base_seed,
        difficulty="easy",
        save_video=True,
        mode_tag="waypoint_phase_isolation",
    )


def _decode_h5_string(raw) -> str:
    if isinstance(raw, np.ndarray):
        raw = raw.flatten()[0]
    if isinstance(raw, (bytes, np.bytes_)):
        return raw.decode("utf-8")
    return str(raw)


def _collect_records(ep_group: h5py.Group) -> list[dict]:
    timestep_keys = sorted(
        (k for k in ep_group.keys() if k.startswith("timestep_")),
        key=lambda k: int(k.split("_")[1]),
    )
    out: list[dict] = []
    for key in timestep_keys:
        ts = ep_group[key]
        info = ts["info"]
        waypoint_action = np.asarray(ts["action"]["waypoint_action"][()]).flatten()
        out.append(
            {
                "timestep": int(key.split("_")[1]),
                "is_demo": bool(np.reshape(np.asarray(info["is_video_demo"][()]), -1)[0]),
                "subgoal": _decode_h5_string(info["simple_subgoal_online"][()]),
                "waypoint_action": waypoint_action,
                "is_finite_waypoint": bool(
                    waypoint_action.shape == (7,) and np.all(np.isfinite(waypoint_action))
                ),
            }
        )
    return out


def _find_demo_to_non_demo_boundary(records: list[dict]) -> int | None:
    for idx in range(1, len(records)):
        if records[idx - 1]["is_demo"] and not records[idx]["is_demo"]:
            return idx
    return None


def _last_finite_demo_waypoint(records: list[dict], boundary_idx: int) -> np.ndarray | None:
    for idx in range(boundary_idx - 1, -1, -1):
        row = records[idx]
        if not row["is_demo"]:
            continue
        if row["is_finite_waypoint"]:
            return np.asarray(row["waypoint_action"]).flatten()
    return None


def _unique_finite_waypoints(rows: list[dict]) -> list[np.ndarray]:
    uniques: list[np.ndarray] = []
    prev: np.ndarray | None = None
    for row in rows:
        if not row["is_finite_waypoint"]:
            continue
        wa = np.asarray(row["waypoint_action"]).flatten()
        if prev is None or not np.array_equal(wa, prev):
            uniques.append(wa.copy())
            prev = wa.copy()
    return uniques


@pytest.mark.parametrize(
    "env_id,base_seed,assert_first_segment_midpoint",
    [
        ("PatternLock", 15001, False),
        ("RouteStick", 16000, True),
    ],
)
def test_waypoint_isolation_across_demo_phase(
    env_id: str,
    base_seed: int,
    assert_first_segment_midpoint: bool,
    dataset_factory,
):
    generated = dataset_factory(_make_case(env_id, base_seed))

    with h5py.File(generated.raw_h5_path, "r") as h5f:
        records = _collect_records(h5f["episode_0"])

    assert records, f"{env_id}: episode_0 has no recorded timesteps."
    boundary_idx = _find_demo_to_non_demo_boundary(records)
    assert boundary_idx is not None, f"{env_id}: missing demo->non-demo boundary."

    first_non_demo = records[boundary_idx]
    assert not first_non_demo["is_demo"], (
        f"{env_id}: boundary row should be non-demo, got demo at timestep "
        f"{first_non_demo['timestep']}."
    )

    last_demo_waypoint = _last_finite_demo_waypoint(records, boundary_idx)
    if first_non_demo["is_finite_waypoint"] and last_demo_waypoint is not None:
        assert not np.array_equal(
            np.asarray(first_non_demo["waypoint_action"]).flatten(),
            last_demo_waypoint,
        ), (
            f"{env_id}: boundary non-demo step consumed demo-phase pending waypoint "
            f"(timestep {first_non_demo['timestep']})."
        )

    non_demo_rows = [row for row in records if not row["is_demo"]]
    assert non_demo_rows, f"{env_id}: no non-demo rows."
    all_non_demo_unique = _unique_finite_waypoints(non_demo_rows)
    assert all_non_demo_unique, f"{env_id}: non-demo phase has no finite waypoints."

    if assert_first_segment_midpoint:
        first_subgoal = non_demo_rows[0]["subgoal"]
        first_segment_rows: list[dict] = []
        for row in non_demo_rows:
            if row["subgoal"] != first_subgoal:
                break
            first_segment_rows.append(row)
        first_segment_unique = _unique_finite_waypoints(first_segment_rows)
        assert len(first_segment_unique) >= 2, (
            f"{env_id}: first non-demo subgoal segment lost midpoint waypoint; "
            f"expected >=2 unique finite waypoints, got {len(first_segment_unique)}."
        )