HCAI-Lab/w2-consensus-deepdive-unlearning-artifacts / social-data-attribution-w2 /scripts /query_data /generate_soc171_queries.py
| """Generate query JSONL files for SOC-171 (GSM8K + ARC, base + instruct).""" | |
| import logging | |
| from pathlib import Path | |
| from data_attribution.recipes.olmes.manifest_writer import ( | |
| FORMAT_BASE, | |
| FORMAT_INSTRUCT, | |
| write_olmes_manifest, | |
| ) | |
| logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s") | |
| log = logging.getLogger(__name__) | |
| BASE_HF_REPO = "HCAI-Lab/olmes-eval-olmo3-7b-base" | |
| INSTRUCT_HF_REPO = "HCAI-Lab/olmes-eval-olmo3-7b-instruct-base" | |
| SUBSETS = ("gsm8k", "arc_easy", "arc_challenge") | |
| VARIANTS = { | |
| "base": { | |
| "hf_repo": BASE_HF_REPO, | |
| "output_format": FORMAT_BASE, | |
| "file_prefix": "olmes_", | |
| }, | |
| "instruct_base": { | |
| "hf_repo": INSTRUCT_HF_REPO, | |
| "output_format": FORMAT_INSTRUCT, | |
| "file_prefix": "olmes_instruct_", | |
| }, | |
| } | |
| def main() -> None: | |
| for variant_name, cfg in VARIANTS.items(): | |
| output_dir = Path(f"queries/{variant_name}") | |
| output_dir.mkdir(parents=True, exist_ok=True) | |
| for subset in SUBSETS: | |
| output_path = output_dir / f"{cfg['file_prefix']}{subset}.jsonl" | |
| log.info("Generating %s", output_path) | |
| count = write_olmes_manifest( | |
| output_path, | |
| subset, | |
| hf_repo=cfg["hf_repo"], | |
| output_format=cfg["output_format"], | |
| ) | |
| log.info("Wrote %d rows to %s", count, output_path) | |
| log.info("Done. All query JSONL files generated.") | |
| if __name__ == "__main__": | |
| main() | |
Xet Storage Details
- Size:
- 1.52 kB
- Xet hash:
- 42f77051f15eb5010aaddd5fa526e7b03e9d8acfaef113434ac16a8177ad6040
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.