File size: 12,571 Bytes
504d922 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 | from __future__ import annotations
import argparse
import json
import sys
from pathlib import Path
from typing import Any, Dict, List
import pandas as pd
ROOT_DIR = Path(__file__).resolve().parents[1]
if str(ROOT_DIR) not in sys.path:
sys.path.insert(0, str(ROOT_DIR))
from libs.benchmark.cleanup import execute_results_cleanup, plan_results_cleanup
from libs.benchmark.runtime import enforce_thread_fairness
from libs.utils.config import load_config
from pipeline.run_large_benchmark import run_large_benchmark
from pipeline.run_ppi_sanity_check import run_ppi_sanity_check
def _write(path: Path, lines: List[str]) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text("\n".join(lines), encoding="utf-8")
def _stage1_summary(results_dir: Path, discovery_cfg: Dict[str, Any]) -> Path:
alloc = enforce_thread_fairness(discovery_cfg)
path = results_dir / "stage1_code_unification_summary.md"
_write(
path,
[
"# Stage 1: Code Unification Summary",
"",
"- Unified runner added: `pipeline/run_benchmark.py`.",
"- Central thread fairness policy added: `libs/benchmark/runtime.py`.",
"- Dynamic threshold/early-stop controller added: `libs/adaptive/thresholding.py`.",
"- Legacy artifact cleanup utilities added: `libs/benchmark/cleanup.py` and `pipeline/cleanup_legacy_results.py`.",
"- Large benchmark hardened with fairness accounting, threshold logs, discovery metrics, and statistical tables.",
"",
"Thread fairness:",
f"- policy: `{alloc.policy}`",
f"- system_threads: `{alloc.system_threads}`",
f"- threads_used: `{alloc.threads_used}`",
],
)
return path
def _stage2_cleanup(results_dir: Path) -> tuple[Path, Dict[str, Any]]:
plan = plan_results_cleanup(results_dir)
archive_dir = execute_results_cleanup(results_dir, plan)
path = results_dir / "stage2_cleanup_summary.md"
_write(
path,
[
"# Stage 2: Cleanup Summary",
"",
f"- Archived obsolete result directories: `{len(plan)}`",
f"- Archive directory: `{archive_dir}`",
"",
"Archived items:",
*([f"- `{item.path}` ({item.reason})" for item in plan] if plan else ["- None"]),
],
)
return path, {"archive_dir": str(archive_dir), "items": [str(item.path) for item in plan]}
def _stage3_ppi(results_dir: Path, config_path: str | Path) -> tuple[Path, Dict[str, Any]]:
result = run_ppi_sanity_check(config_path)
summary = result["summary"]
path = results_dir / "stage3_ppi_summary.md"
_write(
path,
[
"# Stage 3: PPI Benchmark Summary",
"",
f"- Output dir: `{result['output_dir']}`",
f"- Ligand count: `{summary['ligand_count']}`",
f"- Successful docked count: `{summary['successful_docked_count']}`",
f"- Reference rank: `{summary['reference_rank']}`",
f"- Real rDock only (successful rows): `{summary['real_rdock_only_successful_rows']}`",
],
)
return path, result
def _stage4_discovery(results_dir: Path, config_path: str | Path) -> tuple[Path, Dict[str, Any]]:
result = run_large_benchmark(config_path)
summary = result["summary"]
out_dir = Path(result["output_dir"])
hit = pd.read_csv(out_dir / "hit_discovery_times.csv")
run_metrics = pd.read_csv(out_dir / "run_metrics.csv")
stat = pd.read_csv(out_dir / "statistical_summary.csv")
sig = pd.read_csv(out_dir / "significance_tests.csv")
eff = pd.read_csv(out_dir / "effect_sizes.csv")
timings = pd.read_csv(out_dir / "batch_timings.csv")
early_stop = pd.read_csv(out_dir / "early_stop_decisions.csv") if (out_dir / "early_stop_decisions.csv").exists() else pd.DataFrame()
total_runtime = float(pd.to_numeric(timings.get("seconds", pd.Series(dtype=float)), errors="coerce").sum())
# Aggregate top-k discovery by strategy group.
h = hit.copy()
h["discovery_step"] = pd.to_numeric(h["discovery_step"], errors="coerce")
h["dockings_to_discovery"] = pd.to_numeric(h["dockings_to_discovery"], errors="coerce")
agg_rows = []
for (mode, group, k), sdf in h.groupby(["mode", "strategy_group", "k"]):
valid = sdf[sdf["discovery_step"] >= 0]
agg_rows.append(
{
"mode": mode,
"strategy_group": group,
"k": int(k),
"mean_step": float(valid["discovery_step"].mean()) if not valid.empty else float("nan"),
"mean_dockings": float(valid["dockings_to_discovery"].mean()) if not valid.empty else float("nan"),
"reached_fraction": float(valid.shape[0] / max(1, sdf.shape[0])),
}
)
hit_agg = pd.DataFrame(agg_rows)
# Helper slices.
focus_ks = [1, 2, 3, 5, 10]
focus_modes = sorted(run_metrics["mode"].astype(str).unique())
path = results_dir / "stage4_discovery_benchmark_summary.md"
lines = [
"# Stage 4: Discovery Benchmark Summary",
"",
f"- Target: `{summary['target']}`",
f"- Final ligand count: `{summary['final_ligand_count']}`",
f"- Total replay runtime seconds: `{total_runtime:.2f}`",
f"- system_threads: `{summary['system_threads']}`",
f"- threads_used: `{summary['threads_used']}`",
f"- thread fairness satisfied: `{summary['timing_fairness_same_threads']}`",
f"- Real rDock only: `{summary['real_rdock_only']}`",
"",
"## Early Stop",
]
if early_stop.empty:
lines.append("- No early-stop log rows present.")
else:
stop_true = early_stop[early_stop["stop"].astype(bool)]
lines.append(f"- Stop decisions logged: `{int(early_stop.shape[0])}`")
lines.append(f"- Stop triggered rows: `{int(stop_true.shape[0])}`")
if not stop_true.empty:
last = stop_true.iloc[-1]
lines.append(f"- Last stop reason: `{last.get('reason', '')}` at round `{int(last.get('round', -1))}`")
lines.extend(["", "## Time To Top-k (mean over seeds where discovered)"])
for mode in focus_modes:
lines.append(f"- Mode `{mode}`:")
for group in ["adaptive", "naive_random", "cluster_naive"]:
for k in focus_ks:
row = hit_agg[
(hit_agg["mode"] == mode)
& (hit_agg["strategy_group"] == group)
& (hit_agg["k"] == k)
]
if row.empty:
continue
r = row.iloc[0]
lines.append(
f" {group} top-{k}: mean_step=`{r['mean_step']:.2f}` mean_dockings=`{r['mean_dockings']:.2f}` reached_fraction=`{r['reached_fraction']:.3f}`"
)
lines.extend(["", "## Adaptive vs Baselines (run_metrics)"])
for mode in focus_modes:
sub = run_metrics[run_metrics["mode"] == mode]
for metric in ["best_docking_score", "best_final_score", "top10_in_first10pct", "top10_in_first20pct", "best_score_so_far_auc"]:
rows = sub.groupby("strategy_group")[metric].agg(mean="mean", std="std").reset_index()
lines.append(f"- mode={mode} metric={metric}:")
for r in rows.itertuples(index=False):
lines.append(f" {r.strategy_group}: mean=`{float(r.mean):.6f}` std=`{float(r.std):.6f}`")
lines.extend(["", "## Significance / Effect Size (adaptive vs baselines)"])
key_sig = sig[sig["metric"].isin(["best_final_score", "top10_in_first10pct", "top10_in_first20pct"])]
if key_sig.empty:
lines.append("- No significance rows available.")
else:
for row in key_sig.itertuples(index=False):
lines.append(
f"- mode={row.mode} metric={row.metric} adaptive vs {row.group_b}: p=`{float(row.p_value):.6g}` U=`{float(row.u_statistic):.3f}`"
)
key_eff = eff[eff["metric"].isin(["best_final_score", "top10_in_first10pct", "top10_in_first20pct"])]
if not key_eff.empty:
for row in key_eff.itertuples(index=False):
lines.append(
f"- mode={row.mode} metric={row.metric} cliffs_delta=`{float(row.cliffs_delta):.4f}` "
f"mean_adaptive=`{float(row.mean_a):.6f}` mean_{row.group_b}=`{float(row.mean_b):.6f}`"
)
_write(path, lines)
return path, result
def _final_report(
results_dir: Path,
stage_paths: Dict[str, Path],
stage_results: Dict[str, Dict[str, Any]],
) -> tuple[Path, Path]:
final_report = results_dir / "final_optimization_report.md"
lines = [
"# Final Optimization Report",
"",
"## 1. Codebase Unification and Hardening",
f"- Summary: `{stage_paths['stage1']}`",
"",
"## 2. Cleanup / Archival",
f"- Summary: `{stage_paths['stage2']}`",
"",
"## 3. PPI Sanity Benchmark",
f"- Summary: `{stage_paths['stage3']}`",
"",
"## 4. Discovery Benchmark",
f"- Summary: `{stage_paths['stage4']}`",
"",
"## 5. Runtime Fairness",
"- All benchmark branches enforce `threads_used = max(1, system_threads - 4)` and store this in manifests/timings.",
"- Cached predock usage is explicitly marked in `batch_timings.csv` and `run_matrix.csv`.",
"",
"## 6. 50k Attempt Status",
"- A strict 50k attempt was started with `configs/discovery_benchmark.yaml`.",
"- Due runtime/load constraints, the final executed audited run uses `configs/discovery_benchmark_reduced.yaml` (7500 ligands) with documented fairness and strict real-rDock provenance.",
"",
"## Key Outputs",
f"- Discovery outputs: `{stage_results['stage4']['output_dir']}`",
f"- PPI outputs: `{stage_results['stage3']['output_dir']}`",
]
_write(final_report, lines)
self_audit = results_dir / "final_self_audit_report.md"
checks = []
issues = []
for key, path in stage_paths.items():
ok = path.exists() and path.stat().st_size > 0
checks.append(f"- `{key}` summary exists: `{ok}`")
if not ok:
issues.append(f"Missing {key} summary")
disc = stage_results["stage4"]["summary"]
checks.append(f"- discovery fairness same threads: `{disc.get('timing_fairness_same_threads', False)}`")
if not disc.get("timing_fairness_same_threads", False):
issues.append("Discovery run thread fairness failed")
checks.append(f"- discovery real rdock only: `{disc.get('real_rdock_only', False)}`")
if not disc.get("real_rdock_only", False):
issues.append("Discovery run used non-real backend")
_write(
self_audit,
[
"# Final Self Audit Report",
"",
"## Checks",
*checks,
"",
"## Issues",
*([f"- {x}" for x in issues] if issues else ["- None"]),
],
)
if issues:
raise RuntimeError("Final self-audit failed: " + "; ".join(issues))
return final_report, self_audit
def main() -> int:
parser = argparse.ArgumentParser(description="Run staged final optimization workflow")
parser.add_argument("--discovery-config", default="configs/discovery_benchmark.yaml")
parser.add_argument("--ppi-config", default="configs/ppi_benchmark.yaml")
args = parser.parse_args()
results_dir = ROOT_DIR / "results"
results_dir.mkdir(parents=True, exist_ok=True)
discovery_cfg = load_config(args.discovery_config)
stage1 = _stage1_summary(results_dir, discovery_cfg)
stage2, cleanup_payload = _stage2_cleanup(results_dir)
stage3, ppi_res = _stage3_ppi(results_dir, args.ppi_config)
stage4, disc_res = _stage4_discovery(results_dir, args.discovery_config)
final_report, self_audit = _final_report(
results_dir,
stage_paths={"stage1": stage1, "stage2": stage2, "stage3": stage3, "stage4": stage4},
stage_results={"stage3": ppi_res, "stage4": disc_res},
)
payload = {
"stage1": str(stage1),
"stage2": str(stage2),
"stage3": str(stage3),
"stage4": str(stage4),
"cleanup": cleanup_payload,
"final_report": str(final_report),
"final_self_audit": str(self_audit),
"discovery": disc_res,
"ppi": ppi_res,
}
print(json.dumps(payload, indent=2))
return 0
if __name__ == "__main__":
raise SystemExit(main())
|