mirror of
https://github.com/priyanshujain/messageboardbench.git
synced 2026-10-02 11:07:07 +00:00
79 lines
2.8 KiB
Python
79 lines
2.8 KiB
Python
"""Preview or freeze a no-communication ImpossibleBench prompt-calibration plan.
|
|
|
|
This command never calls a model and never starts Docker. It records a design that a
|
|
separate runner may execute later with explicit authorization.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import argparse
|
|
import json
|
|
from pathlib import Path
|
|
|
|
from messageboardbench.prompt_calibration import (
|
|
DEFAULT_PARTITIONS,
|
|
VARIANTS_BY_ID,
|
|
TaskPartitions,
|
|
build_manifest,
|
|
write_manifest,
|
|
)
|
|
|
|
|
|
def parser() -> argparse.ArgumentParser:
|
|
p = argparse.ArgumentParser(description=__doc__)
|
|
p.add_argument("--dataset-revision", required=True, help="Pinned dataset commit/revision")
|
|
p.add_argument("--variants", nargs="+", choices=tuple(VARIANTS_BY_ID),
|
|
default=list(VARIANTS_BY_ID))
|
|
p.add_argument("--development-ids", nargs="+", default=list(DEFAULT_PARTITIONS.development))
|
|
p.add_argument("--validation-ids", nargs="+", default=list(DEFAULT_PARTITIONS.validation))
|
|
p.add_argument("--communication-holdout-ids", nargs="+",
|
|
default=list(DEFAULT_PARTITIONS.communication_holdout))
|
|
p.add_argument("--replicates", type=int, default=1)
|
|
p.add_argument("--seed", type=int, default=909)
|
|
p.add_argument("--model", default="openrouter/z-ai/glm-5.3-flash")
|
|
p.add_argument("--messages", type=int, default=90)
|
|
p.add_argument("--tokens", type=int, default=1_000_000)
|
|
p.add_argument("--sample-seconds", type=int, default=1_800)
|
|
p.add_argument("--temperature", type=float, default=1)
|
|
p.add_argument("--reasoning-effort",
|
|
choices=("none", "minimal", "low", "medium", "high", "xhigh"),
|
|
default="high")
|
|
p.add_argument("--out", type=Path, help="Fresh manifest path; omit for preview only")
|
|
return p
|
|
|
|
|
|
def configuration(args: argparse.Namespace) -> dict:
|
|
partitions = TaskPartitions(
|
|
tuple(args.development_ids),
|
|
tuple(args.validation_ids),
|
|
tuple(args.communication_holdout_ids),
|
|
)
|
|
return build_manifest(
|
|
partitions=partitions,
|
|
variant_ids=tuple(args.variants),
|
|
replicates=args.replicates,
|
|
seed=args.seed,
|
|
model=args.model,
|
|
dataset_revision=args.dataset_revision,
|
|
message_limit=args.messages,
|
|
token_limit=args.tokens,
|
|
time_limit=args.sample_seconds,
|
|
temperature=args.temperature,
|
|
reasoning_effort=args.reasoning_effort,
|
|
)
|
|
|
|
|
|
def main() -> None:
|
|
args = parser().parse_args()
|
|
manifest = configuration(args)
|
|
print(json.dumps(manifest, indent=2))
|
|
if args.out is None:
|
|
print("Preview only; no file, Docker container, or model request was created.")
|
|
return
|
|
write_manifest(args.out, manifest)
|
|
print(f"Wrote frozen calibration plan to {args.out}")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|