Datasets:
File size: 7,918 Bytes
0bf2389 | 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 | from __future__ import annotations
import hashlib
import json
import tempfile
import unittest
from pathlib import Path
from barbet_bgd.audit import StatsAccumulator, write_manifest, write_report
from barbet_bgd.io import (
iter_gzip_jsonl,
iter_jsonl_gz,
sha256_file,
write_gzip_jsonl_atomic,
write_jsonl_gz_atomic,
)
def _record(
record_id: str,
*,
split: str = "train",
source_text: str = "模型將於九月十五日上線。",
evidence: str = "src:w03",
context_bucket: str = "0.5K-4K",
intentional_simplified: bool = False,
) -> dict:
return {
"record_id": record_id,
"schema_version": "bgd.canonical.v1",
"primitive": "VERIFY",
"mixture_category": "VERIFICATION_QE_ERROR_LOCALIZATION",
"task_family": "ASR_QUALITY",
"domain": "meeting",
"language": ["zh-TW", "en"],
"locale": "zh-TW",
"split": split,
"source_group": "synthetic-test",
"task_spec": {"instruction": "判斷候選是否忠實，並引用證據。"},
"inputs": [
{
"id": "src",
"type": "reference_text",
"units": [
{"id": "src:w01", "text": "模型"},
{"id": "src:w03", "text": source_text},
],
},
{
"id": "hyp",
"type": "asr_hypothesis",
"units": [{"id": "hyp:w03", "text": "模型將於九月五日上線。"}],
},
],
"target": {
"decision": "FAIL",
"confidence_target": 0.98,
"errors": [
{
"type": "VALUE_CORRUPTION",
"source_evidence": [evidence],
"target_evidence": ["hyp:w03"],
}
],
"action": {"tool": "EDIT", "arguments": {"target": "hyp:w03"}},
},
"component_targets": {},
"presentation": {
"serializer": "JSON",
"serializer_version": "1",
},
"provenance": {"source_hash": "0" * 64},
"quality": {
"context_length_bucket": context_bucket,
"intentional_negative_simplified_hits": (
[{"unit_id": "src:w03", "hits": ["这"]}]
if intentional_simplified
else []
)
},
}
class AtomicGzipJsonlTests(unittest.TestCase):
def test_round_trip_digest_and_deterministic_bytes(self) -> None:
rows = [_record("bgd-001"), _record("bgd-002", split="validation")]
with tempfile.TemporaryDirectory() as tmp:
root = Path(tmp)
first = write_gzip_jsonl_atomic(root / "first.jsonl.gz", rows)
second = write_gzip_jsonl_atomic(root / "second.jsonl.gz", rows)
self.assertEqual(list(iter_gzip_jsonl(first.path)), rows)
self.assertEqual(first.records, 2)
self.assertEqual(first.record_count, 2)
self.assertEqual(first.sha256, sha256_file(first.path))
self.assertEqual(first.sha256, second.sha256)
self.assertEqual(
first.sha256,
hashlib.sha256(first.path.read_bytes()).hexdigest(),
)
self.assertFalse((root / "first.jsonl.gz.incomplete").exists())
alias = write_jsonl_gz_atomic(root / "alias.jsonl.gz", rows)
self.assertEqual(list(iter_jsonl_gz(alias.path)), rows)
def test_failed_write_preserves_published_file_and_cleans_incomplete(self) -> None:
with tempfile.TemporaryDirectory() as tmp:
path = Path(tmp) / "shard.jsonl.gz"
path.write_bytes(b"published")
def bad_rows():
yield _record("bgd-001")
yield {"not_json": object()}
with self.assertRaises(TypeError):
write_gzip_jsonl_atomic(path, bad_rows())
self.assertEqual(path.read_bytes(), b"published")
self.assertFalse(Path(f"{path}.incomplete").exists())
class StatsAccumulatorTests(unittest.TestCase):
def test_ingest_counts_distributions_and_integrity_findings(self) -> None:
stats = StatsAccumulator()
stats.ingest(_record("same-id"))
stats.ingest(
_record(
"same-id",
split="test",
source_text="这个模型将上线。",
evidence="src:missing",
context_bucket="4K-16K",
)
)
result = stats.to_dict()
self.assertEqual(result["total_records"], 2)
self.assertEqual(result["split"], {"test": 1, "train": 1})
self.assertEqual(result["primitive"], {"VERIFY": 2})
self.assertEqual(
result["mixture_category"], {"VERIFICATION_QE_ERROR_LOCALIZATION": 2}
)
self.assertEqual(result["task_family"], {"ASR_QUALITY": 2})
self.assertEqual(result["domain"], {"meeting": 2})
self.assertEqual(result["serializer"], {"JSON": 2})
self.assertEqual(result["decision"], {"FAIL": 2})
self.assertEqual(result["error_type"], {"VALUE_CORRUPTION": 2})
self.assertEqual(result["action"], {"EDIT": 2})
self.assertEqual(result["context_bucket"], {"0.5K-4K": 1, "4K-16K": 1})
self.assertEqual(result["primary_zh_tw_records"], 2)
self.assertEqual(result["primary_zh_tw_ratio"], 1.0)
self.assertEqual(result["simplified_only_hit_records"], 1)
self.assertEqual(result["duplicate_record_ids"], 1)
self.assertEqual(result["duplicate_record_id_occurrences"], 1)
self.assertEqual(result["evidence_reference_missing"], 1)
self.assertEqual(result["evidence_reference_missing_records"], 1)
def test_merge_detects_cross_shard_duplicates_and_annotations(self) -> None:
left = StatsAccumulator().ingest(_record("shared"))
right = StatsAccumulator().ingest(
_record(
"shared",
source_text="这个模型将上线。",
intentional_simplified=True,
)
)
merged = left.merge(right).to_dict()
self.assertEqual(merged["total_records"], 2)
self.assertEqual(merged["duplicate_record_ids"], 1)
self.assertEqual(merged["duplicate_record_id_occurrences"], 1)
self.assertEqual(merged["simplified_only_hit_records"], 1)
self.assertEqual(merged["intentional_negative_simplified_hit_records"], 1)
self.assertEqual(merged["unannotated_simplified_only_hit_records"], 0)
def test_manifest_and_report_are_atomic_json(self) -> None:
rows = [_record("bgd-001")]
with tempfile.TemporaryDirectory() as tmp:
root = Path(tmp)
shard = write_gzip_jsonl_atomic(root / "shard.jsonl.gz", rows)
stats = StatsAccumulator().ingest(rows[0])
manifest_path = write_manifest(
root / "manifest.json",
[shard],
stats=stats,
metadata={"dataset_name": "unit-test"},
)
report_path = write_report(root / "report.json", stats)
manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
report = json.loads(report_path.read_text(encoding="utf-8"))
self.assertEqual(manifest["totals"]["records"], 1)
self.assertEqual(manifest["shards"][0]["sha256"], shard.sha256)
self.assertEqual(manifest["metadata"]["dataset_name"], "unit-test")
self.assertTrue(report["checks"]["all_records_primary_zh_tw"])
self.assertTrue(report["checks"]["record_ids_unique"])
self.assertFalse(Path(f"{manifest_path}.incomplete").exists())
self.assertFalse(Path(f"{report_path}.incomplete").exists())
if __name__ == "__main__":
unittest.main()
|