ONNX
onnxruntime
onnx-mlir
quantization
fp32
File size: 12,681 Bytes
ed3aeeb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
#!/usr/bin/env python3
"""Validate, test, checksum, and manifest a completed OD06/OD07 COCO run."""

from __future__ import annotations

import argparse
import hashlib
import json
import os
import platform
import shlex
import subprocess
import tempfile
from pathlib import Path
from typing import Any


def sha256_file(path: Path) -> str:
    digest = hashlib.sha256()
    with path.open("rb") as handle:
        for block in iter(lambda: handle.read(1024 * 1024), b""):
            digest.update(block)
    return digest.hexdigest()


def atomic_json(path: Path, value: Any) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    with tempfile.NamedTemporaryFile("w", encoding="utf-8", dir=path.parent, delete=False) as handle:
        json.dump(value, handle, ensure_ascii=False, indent=2, sort_keys=True)
        handle.write("\n")
        temporary = Path(handle.name)
    os.replace(temporary, path)


def atomic_text(path: Path, value: str) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    with tempfile.NamedTemporaryFile("w", encoding="utf-8", dir=path.parent, delete=False) as handle:
        handle.write(value)
        temporary = Path(handle.name)
    os.replace(temporary, path)


def run_stage(name: str, command: list[str], root: Path, run_dir: Path) -> dict[str, Any]:
    command_path = run_dir / f"{name}.command.txt"
    stdout_path = run_dir / f"{name}.stdout.log"
    stderr_path = run_dir / f"{name}.stderr.log"
    exit_path = run_dir / f"{name}.exit_code.txt"
    atomic_text(command_path, shlex.join(command) + "\n")
    completed = subprocess.run(command, cwd=root, text=True, capture_output=True, check=False)
    atomic_text(stdout_path, completed.stdout)
    atomic_text(stderr_path, completed.stderr)
    atomic_text(exit_path, f"{completed.returncode}\n")
    return {
        "name": name,
        "status": "PASS" if completed.returncode == 0 else "FAIL",
        "exit_code": completed.returncode,
        "command": shlex.join(command),
        "command_log": str(command_path.relative_to(root)),
        "stdout_log": str(stdout_path.relative_to(root)),
        "stderr_log": str(stderr_path.relative_to(root)),
        "exit_code_log": str(exit_path.relative_to(root)),
        "stdout_sha256": sha256_file(stdout_path),
        "stderr_sha256": sha256_file(stderr_path),
    }


def collect_manifest_files(root: Path, config: dict[str, Any], result_dir: Path, run_dir: Path) -> list[Path]:
    excluded_names = {"artifact_manifest.json", "artifacts.sha256"}
    files = [path for path in result_dir.rglob("*") if path.is_file() and path.name not in excluded_names]
    files.extend(
        path
        for path in run_dir.parent.rglob("*")
        if path.is_file()
        and path.name not in {
            "checksum.stdout.log",
            "checksum.stderr.log",
            "checksum.exit_code.txt",
            "checksum.command.txt",
            "final_result.json",
        }
    )
    files.extend(
        root / path
        for path in (
            "scripts/stages/evaluate_od06_od07_coco.py",
            "scripts/stages/resume_od06_od07_coco_metrics.py",
            "scripts/validate_od06_od07_coco.py",
            "scripts/finalize_od06_od07_coco_evaluation.py",
            "tests/test_od06_od07_coco_evaluator.py",
            "tests/test_od06_od07_coco_results.py",
            "environment/quality/od_coco/requirements.in",
            "environment/quality/od_coco/requirements.lock",
            "research/evidence/detection/od06_od07_coco2017/sources/mediapipe_LICENSE_cb0902f",
            "research/evidence/detection/od06_od07_coco2017/sources/cocoapi_license_8c9bcc3.txt",
            "research/evidence/detection/od06_od07_coco2017/sources/mediapipe_object_detector_guide_20260812.html",
            "research/evidence/detection/od06_od07_coco2017/sources/coco_official_site_20260812.html",
            "research/evidence/detection/od06_od07_coco2017/method_audit.md",
        )
    )
    files.append(root / "configs/evaluation/object_detection/OD06_OD07_coco2017_quality_eval.json")
    files.extend(root / source["path"] for source in config["authoritative_sources"])
    files.extend(
        [
            root / config["dataset"]["annotation_path"],
            root / config["dataset"]["image_archive"]["path"],
            root / config["dataset"]["annotation_archive"]["path"],
            root / config["dataset"]["license"]["terms_path"],
            root / config["models"]["OD06"]["published_context"]["source_path"],
        ]
    )
    for model in config["models"].values():
        files.extend(root / variant["path"] for variant in model["variants"].values())
    unique = sorted(set(path.resolve() for path in files), key=lambda path: str(path.relative_to(root)))
    missing = [str(path) for path in unique if not path.is_file()]
    if missing:
        raise FileNotFoundError(f"manifest inputs missing: {missing}")
    return unique


def main() -> int:
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("--repo-root", required=True, type=Path)
    parser.add_argument("--config", required=True, type=Path)
    parser.add_argument("--result-dir", required=True, type=Path)
    parser.add_argument("--run-dir", required=True, type=Path)
    args = parser.parse_args()
    root = args.repo_root.resolve()
    config_path = args.config.resolve()
    result_dir = args.result_dir.resolve()
    run_dir = args.run_dir.resolve()
    config = json.loads(config_path.read_text(encoding="utf-8"))
    summary_path = result_dir / "quality_summary.json"
    if not summary_path.is_file():
        raise FileNotFoundError("quality_summary.json is required before finalization")
    if (result_dir / "artifact_manifest.json").exists() or (result_dir / "artifacts.sha256").exists():
        raise FileExistsError("refusing to overwrite existing artifact manifest")
    python = root / config["runtime"]["python"]
    stages: list[dict[str, Any]] = []
    stages.append(
        run_stage(
            "validate",
            [
                str(python),
                "scripts/validate_od06_od07_coco.py",
                "--repo-root",
                str(root),
                "--config",
                str(config_path),
                "--result-dir",
                str(result_dir),
                "--output",
                str(result_dir / "validation.json"),
            ],
            root,
            run_dir,
        )
    )
    stages.append(
        run_stage(
            "targeted_tests",
            [
                str(python),
                "-m",
                "pytest",
                "-q",
                "tests/test_od06_od07_coco_evaluator.py",
                "tests/test_od06_od07_coco_results.py",
            ],
            root,
            run_dir,
        )
    )
    stages.append(
        run_stage(
            "py_compile",
            [
                str(python),
                "-m",
                "py_compile",
                "scripts/stages/evaluate_od06_od07_coco.py",
                "scripts/stages/resume_od06_od07_coco_metrics.py",
                "scripts/validate_od06_od07_coco.py",
                "scripts/finalize_od06_od07_coco_evaluation.py",
            ],
            root,
            run_dir,
        )
    )
    summary = json.loads(summary_path.read_text(encoding="utf-8"))
    validation = json.loads((result_dir / "validation.json").read_text(encoding="utf-8")) if (result_dir / "validation.json").is_file() else {}
    preliminary_pass = (
        summary.get("status") == "PASS"
        and validation.get("status") == "PASS"
        and all(stage["status"] == "PASS" for stage in stages)
    )
    inference_attempt = run_dir.parent / "attempt_003"
    inference_exit_path = inference_attempt / "evaluate.exit_code.txt"
    inference_exit_code = int(inference_exit_path.read_text().strip()) if inference_exit_path.is_file() else None
    metrics_exit_path = run_dir / "metrics_only_02.exit_code.txt"
    metrics_exit_code = int(metrics_exit_path.read_text().strip()) if metrics_exit_path.is_file() else None
    full_inference_complete = all(
        summary["models"][model_id]["variants"][variant]["worker_summary"]["image_count"] == 5000
        for model_id in ("OD06", "OD07")
        for variant in ("fp32", "public_int8")
    )
    execution_manifest = {
        "schema_version": "1.0",
        "evaluation_id": config["evaluation_id"],
        "status": "PASS" if preliminary_pass and metrics_exit_code == 0 and full_inference_complete else "FAIL",
        "inference": {
            "process_exit_code": inference_exit_code,
            "process_exit_note": "The process failed only after all four inference bundles were atomically saved, during the first COCOeval compatibility attempt.",
            "command_log": str((inference_attempt / "evaluate.command.txt").relative_to(root)),
            "stdout_log": str((inference_attempt / "evaluate.stdout.log").relative_to(root)),
            "stderr_log": str((inference_attempt / "evaluate.stderr.log").relative_to(root)),
            "code_freeze": str((inference_attempt / "code_freeze.json").relative_to(root)),
            "full_coco_val2017_complete": full_inference_complete,
            "prediction_bundle_count": 4,
        },
        "metrics_resume": {
            "exit_code": metrics_exit_code,
            "command_log": str((run_dir / "metrics_only_02.command.txt").relative_to(root)),
            "stdout_log": str((run_dir / "metrics_only_02.stdout.log").relative_to(root)),
            "stderr_log": str((run_dir / "metrics_only_02.stderr.log").relative_to(root)),
            "code_freeze": str((run_dir / "metrics_code_freeze.json").relative_to(root)),
            "model_runtime_executed": False,
        },
        "post_evaluation_stages": stages,
        "result_status": summary.get("status"),
        "validation_status": validation.get("status"),
        "environment": {"python": platform.python_version(), "platform": platform.platform()},
        "policy": {
            "latency_measurement": False,
            "training_calibration_or_quantization": False,
            "model_conversion_or_modification": False,
            "mlir_allocator_or_codegen": False,
            "published_family_values_used_as_exact_threshold": False,
        },
    }
    atomic_json(run_dir / "execution_manifest.json", execution_manifest)
    if execution_manifest["status"] != "PASS":
        print(json.dumps({"status": "FAIL", "stages": stages}, sort_keys=True))
        return 1
    manifest_files = collect_manifest_files(root, config, result_dir, run_dir)
    records = [
        {
            "path": str(path.relative_to(root)),
            "bytes": path.stat().st_size,
            "sha256": sha256_file(path),
            "role": "input" if path.is_relative_to(root / "research/downloads") else "evidence_or_output",
        }
        for path in manifest_files
    ]
    artifact_manifest = {
        "schema_version": "1.0",
        "evaluation_id": config["evaluation_id"],
        "status": "PASS",
        "models": ["OD06", "OD07"],
        "variants": ["fp32", "public_int8"],
        "dataset": "COCO 2017 val2017, full 5,000 images",
        "validation_status": validation["status"],
        "files": records,
    }
    manifest_path = result_dir / "artifact_manifest.json"
    atomic_json(manifest_path, artifact_manifest)
    checksum_records = records + [
        {
            "path": str(manifest_path.relative_to(root)),
            "sha256": sha256_file(manifest_path),
        }
    ]
    checksum_path = result_dir / "artifacts.sha256"
    atomic_text(checksum_path, "".join(f"{row['sha256']}  {row['path']}\n" for row in checksum_records))
    checksum_stage = run_stage("checksum", ["sha256sum", "-c", str(checksum_path)], root, run_dir)
    final_status = "PASS" if checksum_stage["status"] == "PASS" else "FAIL"
    final_result = {
        "status": final_status,
        "result_status": summary["status"],
        "validation_status": validation["status"],
        "artifact_manifest": str(manifest_path.relative_to(root)),
        "artifact_manifest_sha256": sha256_file(manifest_path),
        "artifacts_sha256": str(checksum_path.relative_to(root)),
        "artifacts_sha256_sha256": sha256_file(checksum_path),
        "checksum_stage": checksum_stage,
    }
    atomic_json(run_dir / "final_result.json", final_result)
    print(json.dumps({"status": final_status, "validation": validation["status"], "files": len(records)}, sort_keys=True))
    return 0 if final_status == "PASS" else 1


if __name__ == "__main__":
    raise SystemExit(main())