File size: 6,072 Bytes
406a8a0 | 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 | #!/usr/bin/env python3
"""Build the Hugging Face release layout for the current benchmark."""
from __future__ import annotations
import json
import shutil
from pathlib import Path
ROOT = Path(__file__).resolve().parents[5]
EVAL_STAGES = ROOT / "Kaggle" / "analyze_code" / "eval_stages"
HF_DIR = EVAL_STAGES / "huggingface_version"
QUESTION_SRC = EVAL_STAGES / "supplementmaterial" / "datasets"
ENV_DIR = (
EVAL_STAGES
/ "eval_environment"
/ "eval_run_260118_finalset_AGENT_swe-agent-gpt-5.2_20260331"
)
QUESTION_FILES = ("codabench.json", "codabench-hard.json")
SKIP_ROOT_FILES = {"result.txt", "task_description.txt"}
def community_sort_key(pair: tuple[str, str]) -> tuple[str, int]:
source_type, community = pair
return source_type, int(community.rsplit("_", 1)[1])
def load_questions() -> dict[str, list[dict]]:
return {
name: json.loads((QUESTION_SRC / name).read_text(encoding="utf-8"))
for name in QUESTION_FILES
}
def write_json(path: Path, data: object) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(
json.dumps(data, ensure_ascii=False, indent=2) + "\n",
encoding="utf-8",
)
def copy_tree_deref(src: Path, dst: Path) -> None:
"""Copy a file or directory while dereferencing symlinks."""
if src.is_dir():
shutil.copytree(src, dst, symlinks=False, dirs_exist_ok=True)
elif src.is_file():
dst.parent.mkdir(parents=True, exist_ok=True)
shutil.copy2(src, dst, follow_symlinks=True)
def build() -> dict:
questions_by_file = load_questions()
all_pairs = sorted(
{
(item["source_type"], item["community"])
for rows in questions_by_file.values()
for item in rows
},
key=community_sort_key,
)
community_map = {
f"{source_type}/{community}": {
"community_id": f"community_{idx}",
"source_type": source_type,
"source_community": community,
"source_env_dir": f"{source_type}_{community}",
"data_path": f"data/community_{idx}/full_community",
}
for idx, (source_type, community) in enumerate(all_pairs)
}
datasets_dir = HF_DIR / "datasets"
data_dir = HF_DIR / "data"
datasets_dir.mkdir(parents=True, exist_ok=True)
data_dir.mkdir(parents=True, exist_ok=True)
for name, rows in questions_by_file.items():
enriched = []
for item in rows:
key = f"{item['source_type']}/{item['community']}"
mapped = community_map[key]
enriched.append(
{
**item,
"release_community": mapped["community_id"],
"data_path": mapped["data_path"],
}
)
write_json(datasets_dir / name, enriched)
copied_roots = 0
copied_instances = 0
copied_dataset_entries = 0
missing_env_dirs: list[str] = []
for community_index, (source_key, mapped) in enumerate(community_map.items(), start=1):
env_comm_dir = ENV_DIR / mapped["source_env_dir"]
dst_full = HF_DIR / mapped["data_path"]
dst_full.mkdir(parents=True, exist_ok=True)
copied_entry_names: set[str] = set()
print(
f"[{community_index}/{len(community_map)}] {source_key} -> {mapped['community_id']}",
flush=True,
)
if not env_comm_dir.is_dir():
missing_env_dirs.append(str(env_comm_dir))
continue
instance_dirs = sorted(
(p for p in env_comm_dir.glob("instance_*") if p.is_dir()),
key=lambda p: int(p.name.rsplit("_", 1)[1]),
)
for instance_dir in instance_dirs:
full_dir = instance_dir / "full_community"
if not full_dir.is_dir():
continue
copied_instances += 1
for entry in full_dir.iterdir():
if entry.name in SKIP_ROOT_FILES:
continue
if entry.name in copied_entry_names:
continue
if (dst_full / entry.name).exists():
copied_entry_names.add(entry.name)
continue
copy_tree_deref(entry, dst_full / entry.name)
copied_entry_names.add(entry.name)
copied_dataset_entries += 1
copied_roots += 1
manifest = {
"name": "huggingface_release_current_benchmark",
"source_eval_environment": str(ENV_DIR),
"question_files": list(QUESTION_FILES),
"num_release_communities": len(community_map),
"num_full_questions": len(questions_by_file["codabench.json"]),
"num_hard_questions": len(questions_by_file["codabench-hard.json"]),
"copied_community_roots": copied_roots,
"copied_instance_full_community_dirs": copied_instances,
"copied_dataset_entries_before_dedup": copied_dataset_entries,
"missing_env_dirs": missing_env_dirs,
"community_map": community_map,
"copy_policy": "All data files are copied with symlinks dereferenced; root result.txt and task_description.txt are excluded.",
}
write_json(HF_DIR / "release_manifest.json", manifest)
readme = """# Hugging Face Benchmark Release
This directory contains the current benchmark release split into question JSON files and copied data files.
- `datasets/codabench.json`: full benchmark questions.
- `datasets/codabench-hard.json`: hard subset questions.
- `data/community_*/full_community`: data directories referenced by the JSON `data_path` field.
- `release_manifest.json`: source-to-release community mapping and copy statistics.
The data copy dereferences the symlinks produced by `stage1_setup`, so the release contains actual file contents rather than symlink placeholders.
"""
(HF_DIR / "README.md").write_text(readme, encoding="utf-8")
return manifest
if __name__ == "__main__":
result = build()
print(json.dumps(result, ensure_ascii=False, indent=2))
|