File size: 7,984 Bytes
0185029 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 | from __future__ import annotations
import hashlib
import sys
import tempfile
import unittest
from copy import deepcopy
from pathlib import Path
ORARL_ROOT = Path(__file__).resolve().parents[1]
if str(ORARL_ROOT) not in sys.path:
sys.path.insert(0, str(ORARL_ROOT))
from orarl.evaluation import ( # noqa: E402
EVALUATION_SCHEMA_VERSION,
EvaluationSchemaError,
annotation_path,
artifact_directory,
media_directory,
validate_dataset_id,
validate_evaluation_row,
validate_evaluation_rows,
validate_repository_path,
)
def _row(
sample_id: str = "sample-001",
video: str = "media/video_qa/videomme/videos/clip.mp4",
):
return {
"schema_version": EVALUATION_SCHEMA_VERSION,
"eval_task": "videomme",
"sample_id": sample_id,
"benchmark": "videomme",
"split": "test",
"problem": "What happens next?",
"answer": "B",
"images": [],
"videos": [video],
"problem_type": "video_qa_mc",
"source": "Video-MME",
"family": "video_qa",
"choices": ["A. Nothing", "B. A person enters"],
"subtitles": ["media/video_qa/videomme/subtitles/clip.srt"],
"preprocessed": {
"video_path": "artifacts/video_qa/videomme/preprocessed/clip.npz",
"settings": {"frames": 128},
},
"task_payload": {"duration": 12.5},
"metadata": {"question_id": "q-1"},
"evaluation": {"group": "short_video"},
}
class EvaluationSchemaTests(unittest.TestCase):
def test_accepts_training_core_and_eval_envelope(self) -> None:
row = _row()
self.assertIs(validate_evaluation_row(row), row)
self.assertEqual(
annotation_path("videomme", "test"),
"annotations/video_qa/videomme/test.jsonl",
)
def test_accepts_mask_backed_segmentation_without_text_answer(self) -> None:
row = _row(video="")
row.update(
{
"eval_task": "segmentation",
"benchmark": "segmentation",
"split": "mevis",
"answer": None,
"images": [],
"videos": ["media/segmentation/videos/mevis/clip.mp4"],
"problem_type": "segmentation",
"source": "mevis",
"family": "segmentation",
"task_payload": {
"segmentation_output": {
"frames": ["00000"],
"segmentation_rle": {
"00000": {"size": [2, 2], "counts": "13"}
},
}
},
}
)
row.pop("subtitles")
row.pop("preprocessed")
self.assertIs(validate_evaluation_row(row), row)
row["task_payload"] = {}
with self.assertRaisesRegex(
EvaluationSchemaError,
"answer must contain a nonempty oracle label",
):
validate_evaluation_row(row)
def test_video_qa_benchmarks_share_one_physical_family(self) -> None:
for benchmark in (
"videomme",
"videommev2",
"mvbench",
"mmvu",
"videoholmes",
"longvideobench",
"mlvu",
):
with self.subTest(benchmark=benchmark):
self.assertEqual(
annotation_path(benchmark, "test"),
f"annotations/video_qa/{benchmark}/test.jsonl",
)
self.assertEqual(
media_directory(benchmark, "videos"),
f"media/video_qa/{benchmark}/videos",
)
self.assertEqual(
artifact_directory(benchmark),
f"artifacts/video_qa/{benchmark}",
)
self.assertEqual(
annotation_path("spatial_grounding", "refcoco_val"),
"annotations/spatial_grounding/refcoco_val.jsonl",
)
def test_spatial_intelligence_benchmarks_share_one_physical_family(self) -> None:
for benchmark in ("vsi", "mmsi", "mindcube", "revsi"):
with self.subTest(benchmark=benchmark):
self.assertEqual(
annotation_path(benchmark, "test"),
f"annotations/spatial_intelligence/{benchmark}/test.jsonl",
)
self.assertEqual(
media_directory(benchmark, "videos"),
f"media/spatial_intelligence/{benchmark}/videos",
)
self.assertEqual(
artifact_directory(benchmark),
f"artifacts/spatial_intelligence/{benchmark}",
)
def test_repository_paths_are_strict_posix_relative_paths(self) -> None:
invalid = (
"/tmp/clip.mp4",
"https://example.invalid/clip.mp4",
r"media\videomme\videos\clip.mp4",
"media/videomme/videos/../clip.mp4",
"media/videomme/videos/./clip.mp4",
)
for path in invalid:
with self.subTest(path=path):
with self.assertRaises(ValueError):
validate_repository_path(path)
for path in invalid:
with self.subTest(row_path=path):
row = _row(video=path)
with self.assertRaises(EvaluationSchemaError):
validate_evaluation_row(row)
def test_requires_snake_case_dataset_identifiers_and_canonical_scope(self) -> None:
for identifier in ("VideoMME", "video-mme", "video__mme", "_videomme"):
with self.subTest(identifier=identifier):
with self.assertRaises(ValueError):
validate_dataset_id(identifier)
row = _row(video="media/other_benchmark/videos/clip.mp4")
with self.assertRaisesRegex(
EvaluationSchemaError,
"media/video_qa/videomme/videos",
):
validate_evaluation_row(row)
def test_rejects_duplicate_rows_and_asset_case_collisions(self) -> None:
with self.assertRaisesRegex(EvaluationSchemaError, "duplicate sample_id"):
validate_evaluation_rows([_row(), deepcopy(_row())])
second = _row(
sample_id="sample-002",
video="media/video_qa/videomme/videos/Clip.mp4",
)
with self.assertRaisesRegex(EvaluationSchemaError, "case collision"):
validate_evaluation_rows([_row(), second])
def test_can_verify_asset_existence_and_checksums(self) -> None:
with tempfile.TemporaryDirectory() as raw_tmp:
root = Path(raw_tmp)
assets = {
"media/video_qa/videomme/videos/clip.mp4": b"video",
"media/video_qa/videomme/subtitles/clip.srt": b"subtitle",
"artifacts/video_qa/videomme/preprocessed/clip.npz": b"artifact",
}
for relative, content in assets.items():
path = root / relative
path.parent.mkdir(parents=True, exist_ok=True)
path.write_bytes(content)
video_path = "media/video_qa/videomme/videos/clip.mp4"
checksum = hashlib.sha256(assets[video_path]).hexdigest()
validate_evaluation_row(
_row(),
repository_root=root,
checksums={video_path: checksum},
)
with self.assertRaisesRegex(EvaluationSchemaError, "checksum mismatch"):
validate_evaluation_row(
_row(),
repository_root=root,
checksums={video_path: "0" * 64},
)
(root / video_path).unlink()
with self.assertRaisesRegex(EvaluationSchemaError, "does not exist"):
validate_evaluation_row(_row(), repository_root=root)
if __name__ == "__main__":
unittest.main()
|