File size: 7,682 Bytes
9d24374
 
 
9f9fbec
9d24374
 
 
 
ffa621b
9f9fbec
9d24374
ffa621b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9d24374
 
 
 
9f9fbec
 
ffa621b
9f9fbec
 
9d24374
 
9f9fbec
 
 
ffa621b
9f9fbec
 
 
 
 
 
ffa621b
9f9fbec
 
ffa621b
9f9fbec
 
ffa621b
 
 
 
 
9f9fbec
 
 
ffa621b
9d24374
ffa621b
9f9fbec
 
 
b3d11b8
ffa621b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b3d11b8
 
 
ffa621b
 
 
3a2b2e4
ffa621b
 
0536091
ffa621b
 
 
393bb89
9f9fbec
 
ffa621b
 
9f9fbec
ffa621b
9f9fbec
ffa621b
 
9f9fbec
ffa621b
9f9fbec
ffa621b
9f9fbec
 
ffa621b
 
 
9f9fbec
ffa621b
b3d11b8
ffa621b
b3d11b8
ffa621b
9f9fbec
 
 
 
ffa621b
 
 
 
 
 
 
9f9fbec
ffa621b
 
 
 
 
 
 
 
 
 
 
 
 
 
9f9fbec
 
 
ffa621b
 
 
 
 
 
 
 
9f9fbec
ffa621b
9f9fbec
ffa621b
9838759
9f9fbec
ffa621b
 
 
 
 
 
 
 
9f9fbec
ffa621b
 
 
 
 
 
 
 
9f9fbec
 
 
 
ffa621b
 
9f9fbec
 
ffa621b
9f9fbec
ffa621b
9d24374
ffa621b
 
 
 
 
 
9d24374
 
ffa621b
b3d11b8
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
from __future__ import annotations

import json
import subprocess
from collections import Counter
from pathlib import Path

ROOT = Path(__file__).resolve().parents[1]
MAX_FILE_SIZE_BYTES = 5_000_000

REQUIRED = [
    "README.md",
    "DESIGN.md",
    "app.py",
    "requirements.txt",
    "research_config.json",
    "featurelens/runtime.py",
    "featurelens/sae.py",
    "featurelens/interventions.py",
    "featurelens/stats.py",
    "featurelens/study.py",
    "experiments/run_all.py",
    "experiments/run_causal.py",
    "experiments/run_feature_sets.py",
    "experiments/analyze_stability.py",
    "experiments/analyze_study.py",
    "experiments/run_analysis_only.py",
    "data/prompts.jsonl",
    "data/causal_tasks.jsonl",
    "notebooks/README.md",
    "notebooks/FeatureLens_Offline_Study_Colab.ipynb",
    "notebooks/FeatureLens_Causal_Addendum_Colab.ipynb",
    "scripts/ui_smoke.py",
    "scripts/validate_artifacts.py",
    "artifacts/feature_catalog.csv",
    "artifacts/layer_metrics.csv",
    "artifacts/stability.csv",
    "artifacts/selection_stability.csv",
    "artifacts/causal_results_final_token.csv",
    "artifacts/causal_results_max_active.csv",
    "artifacts/causal_position_summary.csv",
    "artifacts/feature_set_results.csv",
    "artifacts/study_feature_summary.csv",
    "artifacts/study_summary.json",
    "artifacts/summary.json",
    "artifacts/report.md",
]


def load_jsonl(path: Path) -> list[dict]:
    return [
        json.loads(line)
        for line in path.read_text(encoding="utf-8").splitlines()
        if line.strip()
    ]


def repository_candidates() -> list[Path]:
    try:
        result = subprocess.run(
            ["git", "ls-files", "--cached", "--others", "--exclude-standard"],
            cwd=ROOT,
            capture_output=True,
            text=True,
            check=True,
        )
    except FileNotFoundError as exc:
        raise SystemExit("Git is required to run the FeatureLens release check.") from exc
    except subprocess.CalledProcessError as exc:
        raise SystemExit(
            f"Could not inspect repository files with Git: {exc.stderr.strip()}"
        ) from exc

    return [
        ROOT / rel
        for rel in result.stdout.splitlines()
        if rel.strip() and (ROOT / rel.strip()).is_file()
    ]


def check_required_files() -> None:
    missing = [name for name in REQUIRED if not (ROOT / name).exists()]
    if missing:
        raise SystemExit(f"Missing required files: {missing}")


def check_config(config: dict) -> None:
    expected = {
        "model_id": "Qwen/Qwen3-1.7B-Base",
        "layers": [4, 14, 26],
        "sae_width": 32768,
        "discovery_prompts": 224,
        "causal_tasks": 28,
        "feature_set_sizes": [1, 3, 5],
        "live_random_controls": 8,
        "offline_random_controls_default": 8,
        "offline_selection_resamples": 128,
        "offline_causal_position_policies": [
            "final_token",
            "max_feature_activation",
        ],
        "primary_offline_causal_position_policy": "max_feature_activation",
        "release_status": "final",
    }
    for key, value in expected.items():
        if config.get(key) != value:
            raise SystemExit(
                f"Unexpected {key}: {config.get(key)!r}. Expected {value!r}."
            )

    if "prompt-wide" not in str(config.get("offline_feature_pooling", "")):
        raise SystemExit("Offline SAE concept evidence must use prompt-wide pooling.")

    concepts = config.get("concepts", [])
    if "german_language" not in concepts or "french_language" in concepts:
        raise SystemExit("Controlled language concept must be german_language.")


def check_datasets(config: dict) -> tuple[list[dict], list[dict]]:
    prompts = load_jsonl(ROOT / "data" / "prompts.jsonl")
    causal = load_jsonl(ROOT / "data" / "causal_tasks.jsonl")

    if len(prompts) != config["discovery_prompts"]:
        raise SystemExit(
            f"Discovery prompt count mismatch: {len(prompts)} != "
            f"{config['discovery_prompts']}."
        )
    if len(causal) != config["causal_tasks"]:
        raise SystemExit(
            f"Causal task count mismatch: {len(causal)} != {config['causal_tasks']}."
        )

    concept_counts = Counter(row["concept"] for row in prompts)
    if set(concept_counts) != set(config["concepts"]):
        raise SystemExit("Discovery concepts do not match research_config.json.")
    if len(set(concept_counts.values())) != 1:
        raise SystemExit(f"Discovery concepts are not balanced: {dict(concept_counts)}")

    pair_counts = Counter(row["pair_id"] for row in prompts)
    if set(pair_counts.values()) != {2}:
        raise SystemExit("Every discovery paraphrase pair must contain exactly two prompts.")

    return prompts, causal


def check_study_summary() -> None:
    study = json.loads(
        (ROOT / "artifacts" / "study_summary.json").read_text(encoding="utf-8")
    )
    summary = json.loads(
        (ROOT / "artifacts" / "summary.json").read_text(encoding="utf-8")
    )

    if study.get("primary_causal_position_policy") != "max_feature_activation":
        raise SystemExit("Committed study must use max_feature_activation as primary policy.")
    if "causal task" not in str(study.get("causal_statistical_unit", "")).lower():
        raise SystemExit("Committed study must document causal-task-level inference.")
    if float(study.get("max_active_feature_coverage", 0.0)) <= float(
        study.get("final_token_feature_coverage", 0.0)
    ):
        raise SystemExit("Expected max-active coverage to exceed final-token coverage.")
    if float(study.get("max_active_target_specificity_ratio", 0.0)) <= 1.0:
        raise SystemExit("Committed max-active study specificity ratio is invalid.")

    headline = str(summary.get("headline", ""))
    if "0.962" not in headline or "2.33" not in headline:
        raise SystemExit("Committed summary.json does not contain the finalized measured headline.")


def check_readme() -> None:
    readme = (ROOT / "README.md").read_text(encoding="utf-8")
    required = [
        "0.962 held-out AUROC",
        "2.33×",
        "28.6%",
        "82.1%",
        "notebooks/FeatureLens_Offline_Study_Colab.ipynb",
        "artifacts/report.md",
    ]
    missing = [text for text in required if text not in readme]
    if missing:
        raise SystemExit(f"README.md missing finalized study content: {missing}")


def check_oversized_files() -> None:
    oversized: list[str] = []
    for path in repository_candidates():
        size = path.stat().st_size
        if size > MAX_FILE_SIZE_BYTES:
            oversized.append(
                f"{path.relative_to(ROOT)} ({size / 1_000_000:.1f} MB)"
            )

    if oversized:
        formatted = "\n  - ".join(oversized)
        raise SystemExit(
            "Repository contains unexpectedly large tracked/unignored candidates:\n"
            f"  - {formatted}\n\n"
            "Model weights, SAE checkpoints, activation dumps, virtual environments, "
            "and caches should not be committed."
        )


def main() -> None:
    check_required_files()

    config = json.loads((ROOT / "research_config.json").read_text(encoding="utf-8"))
    check_config(config)
    prompts, causal = check_datasets(config)
    check_study_summary()
    check_readme()
    check_oversized_files()

    print("FeatureLens release check: PASS")
    print(f"  discovery prompts: {len(prompts)}")
    print(f"  causal tasks: {len(causal)}")
    print(f"  layers: {config['layers']}")
    print("  committed offline study: complete")
    print("  release: 1.0.0")


if __name__ == "__main__":
    main()