| """Aggregate the scaled Hugging Face GPU Job results (downloaded from the bucket).""" |
| from __future__ import annotations |
|
|
| import argparse |
| import glob |
| import json |
| import os |
| from collections import defaultdict |
|
|
| import numpy as np |
|
|
| LABEL = {"stgfn": "ST-GFN (Ours)", "tb": "TB", "fm": "FM", "subtb": "SubTB", "db": "DB", |
| "eflownet": "EFlowNet", "stochastic_gfn": "Stochastic-GFN", "tb_rnd": "TB+RND", |
| "tb_novelty": "TB+Novelty", "tb_icm": "TB+ICM", "tb_cv": "TB+ControlVar"} |
| ORDER = ["stgfn", "tb", "fm", "subtb", "db", "eflownet", "stochastic_gfn", |
| "tb_rnd", "tb_novelty", "tb_icm", "tb_cv"] |
| METRICS = { |
| "hypergrid": [("modes_found", "Modes (of 256)", True), |
| ("coverage_pct", "Cov.%", True), |
| ("l1_to_target", "L1 to P*", False)], |
| "singlecell_proxy": [("target_corr", "Corr.", True), |
| ("l1_error", "L1", False), |
| ("mean_reward", "Mean R", True)], |
| } |
|
|
|
|
| def welch_p(a, b): |
| from math import erfc, sqrt |
| a, b = np.array(a, float), np.array(b, float) |
| if len(a) < 2 or len(b) < 2: |
| return None |
| se = np.sqrt(a.var(ddof=1) / len(a) + b.var(ddof=1) / len(b)) |
| if se == 0: |
| return None |
| return float(erfc(abs((a.mean() - b.mean()) / se) / sqrt(2))) |
|
|
|
|
| if __name__ == "__main__": |
| ap = argparse.ArgumentParser() |
| ap.add_argument("--dir", default="../outputs/scaled") |
| ap.add_argument("--save", default="../outputs/summary_scaled.json") |
| args = ap.parse_args() |
|
|
| data = defaultdict(lambda: defaultdict(list)) |
| for p in glob.glob(os.path.join(args.dir, "**", "*.json"), recursive=True): |
| if os.path.basename(p).startswith("summary"): |
| continue |
| try: |
| with open(p) as f: |
| r = json.load(f) |
| if "env" in r and "method" in r: |
| data[r["env"]][r["method"]].append(r) |
| except (json.JSONDecodeError, KeyError): |
| continue |
|
|
| summary = {} |
| for env, metrics in METRICS.items(): |
| if env not in data: |
| continue |
| cfg = data[env][list(data[env])[0]][0]["config"] |
| scale = (f"grid {cfg['grid_size']}x{cfg['grid_size']}, period {cfg['period']}" |
| if env == "hypergrid" else |
| f"{cfg['n_genes']} choose {cfg['k_genes']}") |
| print(f"\n{'='*88}\n{env.upper()} — SCALED on HF GPU Job ({scale}), " |
| f"{cfg['iters']} iters, mean ± 95% CI\n{'='*88}") |
| print(f"{'Method':22s} {'n':>2s} " + " ".join(f"{n:>18s}" for _, n, _ in metrics)) |
| summary[env] = {"scale": scale, "iters": cfg["iters"], "methods": {}} |
| base = {k: [r["final"].get(k) for r in data[env].get("stgfn", [])] |
| for k, _, _ in metrics} |
| for m in ORDER: |
| if m not in data[env]: |
| continue |
| runs, cells, entry = data[env][m], [], {} |
| for key, _, _ in metrics: |
| vals = [r["final"].get(key) for r in runs if r["final"].get(key) is not None] |
| if not vals: |
| cells.append(f"{'—':>18s}") |
| continue |
| mu = float(np.mean(vals)) |
| ci = float(1.96 * np.std(vals, ddof=1) / np.sqrt(len(vals))) if len(vals) > 1 else 0.0 |
| star = "" |
| if m != "stgfn": |
| bv = [x for x in base.get(key, []) if x is not None] |
| if len(bv) > 1 and len(vals) > 1: |
| p = welch_p(bv, vals) |
| entry[key + "_p_vs_stgfn"] = p |
| if p is not None and p < 0.05: |
| star = "*" |
| entry[key] = {"mean": mu, "ci95": ci, "values": vals} |
| cells.append(f"{mu:13.3f}±{ci:4.3f}{star}"[:18].rjust(18)) |
| summary[env]["methods"][m] = entry |
| print(f"{LABEL.get(m,m):22s} {len(runs):2d} " + " ".join(cells)) |
| print(" * = p<0.05 (Welch) vs ST-GFN") |
|
|
| with open(args.save, "w") as f: |
| json.dump(summary, f, indent=2) |
| print(f"\nwrote {args.save}") |
|
|
