Datasets:
File size: 5,653 Bytes
4a2f66d | 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 | """``mcr`` command line entry point.
mcr list
mcr eval --model st_text --benchmark ssrb,multiconir,mcmr,merit --output runs/foo
"""
from __future__ import annotations
import argparse
import json
import os
import sys
from typing import Any, Dict, List
from .config import resolve_benchmark_cfg, resolve_model_cfg
from .cache import EmbeddingCache
from .data.registry import available_benchmarks, build_tasks
from .models.registry import build_encoder, build_reranker
from .pipeline import evaluate_task, evaluate_task_agentic
def _cmd_list(_: argparse.Namespace) -> int:
print("Available benchmarks:")
for name in available_benchmarks():
print(f" - {name}")
return 0
def _cmd_eval(args: argparse.Namespace) -> int:
model_cfg = resolve_model_cfg(args.model)
model_tag = model_cfg.get("tag") or model_cfg.get("model_name_or_path", "model")
agentic_cfg = None
if args.agentic:
agentic_cfg = resolve_model_cfg(args.agentic)
encoder = None
reranker = None
if agentic_cfg is None:
encoder = build_encoder(model_cfg)
if args.reranker:
reranker = build_reranker(resolve_model_cfg(args.reranker))
cache = EmbeddingCache(args.cache_dir, enabled=not args.no_cache)
os.makedirs(args.output, exist_ok=True)
summary: List[Dict[str, Any]] = []
for bench in args.benchmark.split(","):
bench = bench.strip()
if not bench:
continue
bcfg = resolve_benchmark_cfg(bench)
k_values = bcfg.get("k_values", [1, 5, 10, 20, 50, 100])
tasks = build_tasks(bcfg["name"], bcfg)
print(f"[{bench}] {len(tasks)} task(s)", flush=True)
for task in tasks:
print(f" -> {task.name}", flush=True)
if agentic_cfg is not None:
report = evaluate_task_agentic(
task,
agentic_cfg,
cache=cache,
k_values=k_values,
top_k=args.top_k,
retriever_kwargs=bcfg.get("retriever", {}),
max_queries=args.max_queries,
)
report["benchmark"] = bench
_dump(args.output, task.name, report)
row = {"benchmark": bench, "task": task.name}
row.update({m: report["agentic"]["metrics"].get(m) for m in report["official_metrics"]})
summary.append(row)
else:
report = evaluate_task(
task,
encoder,
cache=cache,
model_tag=str(model_tag),
k_values=k_values,
top_k=args.top_k,
batch_size=model_cfg.get("batch_size", 64),
retriever_kwargs=bcfg.get("retriever", {}),
reranker=reranker,
rerank_top_k=args.rerank_top_k,
max_queries=args.max_queries,
)
report["benchmark"] = bench
_dump(args.output, task.name, report)
row = {"benchmark": bench, "task": task.name}
row.update({m: report["dense"]["metrics"].get(m) for m in report["official_metrics"]})
summary.append(row)
with open(os.path.join(args.output, "summary.json"), "w", encoding="utf-8") as f:
json.dump(summary, f, indent=2, ensure_ascii=False)
_print_summary(summary)
return 0
def _dump(output: str, task_name: str, report: Dict[str, Any]) -> None:
safe = task_name.replace("/", "__")
with open(os.path.join(output, f"{safe}.json"), "w", encoding="utf-8") as f:
json.dump(report, f, indent=2, ensure_ascii=False)
def _print_summary(summary: List[Dict[str, Any]]) -> None:
if not summary:
print("(no results)")
return
cols = [k for k in summary[0] if k not in ("benchmark", "task")]
header = ["task"] + cols
print("\n" + " ".join(h.ljust(14) for h in header))
for row in summary:
cells = [str(row["task"])[:14].ljust(14)]
cells += [f"{row.get(c, 0) or 0:.4f}".ljust(14) for c in cols]
print(" ".join(cells))
def build_parser() -> argparse.ArgumentParser:
p = argparse.ArgumentParser(prog="mcr", description="Unified multi-condition retrieval eval")
sub = p.add_subparsers(dest="command", required=True)
sub.add_parser("list", help="list available benchmarks").set_defaults(func=_cmd_list)
e = sub.add_parser("eval", help="run evaluation")
e.add_argument("--model", required=True, help="model config name or yaml path")
e.add_argument("--benchmark", required=True, help="comma-separated benchmark names/paths")
e.add_argument("--output", required=True, help="output directory for reports")
e.add_argument("--reranker", default=None, help="optional reranker config name or path")
e.add_argument("--agentic", default=None, help="agentic search config name or yaml path")
e.add_argument("--cache-dir", default=".mcr_cache", help="embedding cache directory")
e.add_argument("--no-cache", action="store_true", help="disable embedding cache")
e.add_argument("--top-k", type=int, default=100)
e.add_argument("--rerank-top-k", type=int, default=100)
e.add_argument("--max-queries", type=int, default=None, help="limit queries (smoke tests)")
e.set_defaults(func=_cmd_eval)
return p
def main(argv: List[str] | None = None) -> int:
args = build_parser().parse_args(argv if argv is not None else sys.argv[1:])
return args.func(args)
if __name__ == "__main__":
raise SystemExit(main())
|