scripts/train_bandit.py

"""Local-machine bandit training. Mirrors hf/train_on_hf.py but with smaller defaults."""

from __future__ import annotations

import argparse
import json
import sys
from pathlib import Path

sys.path.insert(0, str(Path(__file__).resolve().parent.parent / "hf"))
from train_on_hf import _gather_instances, _train_one_pass  # noqa: E402

from dash_jsp.bandit.linucb import LinUCBDispatcher
from dash_jsp.bandit.thompson import ThompsonDispatcher


def main() -> None:
    parser = argparse.ArgumentParser()
    parser.add_argument("--data-dir", default="data")
    parser.add_argument("--output-dir", default="models")
    parser.add_argument("--benchmarks", nargs="+", default=["lawrence"],
                        choices=["taillard", "lawrence", "dmu"])
    parser.add_argument("--epochs", type=int, default=2)
    parser.add_argument("--alpha", type=float, default=1.0)
    parser.add_argument("--seed", type=int, default=42)
    args = parser.parse_args()

    instances = _gather_instances(args.benchmarks, Path(args.data_dir))
    out_dir = Path(args.output_dir)
    out_dir.mkdir(parents=True, exist_ok=True)

    linucb = LinUCBDispatcher(alpha=args.alpha, rng_seed=args.seed)
    thompson = ThompsonDispatcher(rng_seed=args.seed)

    log = {"epochs": [], "benchmarks": args.benchmarks, "n_instances": len(instances)}
    for ep in range(args.epochs):
        lin = _train_one_pass(linucb, instances)
        ts = _train_one_pass(thompson, instances)
        log["epochs"].append({"epoch": ep, "linucb": lin, "thompson": ts})
        print(f"epoch {ep}: linucb pulls={lin['n_arm_pulls']} thompson pulls={ts['n_arm_pulls']}")

    linucb.save(str(out_dir / "bandit_linucb.npz"))
    thompson.save(str(out_dir / "bandit_thompson.npz"))
    (out_dir / "training_log.json").write_text(json.dumps(log, indent=2))
    print(f"Saved to {out_dir}/")


if __name__ == "__main__":
    main()