Spaces:
Running
Running
| import json | |
| import logging | |
| import os | |
| import time | |
| import pandas as pd | |
| from huggingface_hub import snapshot_download | |
| from src.envs import DATA_ARENA_PATH, DATA_PATH | |
| # Configure logging | |
| logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s") | |
| def time_diff_wrapper(func): | |
| def wrapper(*args, **kwargs): | |
| start_time = time.time() | |
| result = func(*args, **kwargs) | |
| end_time = time.time() | |
| diff = end_time - start_time | |
| logging.info("Time taken for %s: %s seconds", func.__name__, diff) | |
| return result | |
| return wrapper | |
| def download_dataset(repo_id, local_dir, repo_type="dataset", max_attempts=3, backoff_factor=1.5): | |
| """Download dataset with exponential backoff retries.""" | |
| attempt = 0 | |
| while attempt < max_attempts: | |
| try: | |
| logging.info("Downloading %s to %s", repo_id, local_dir) | |
| snapshot_download( | |
| repo_id=repo_id, | |
| local_dir=local_dir, | |
| repo_type=repo_type, | |
| tqdm_class=None, | |
| token=os.environ.get("HF_TOKEN_PRIVATE"), | |
| etag_timeout=30, | |
| max_workers=8, | |
| ) | |
| logging.info("Download successful") | |
| return | |
| except Exception as e: | |
| wait_time = backoff_factor**attempt | |
| logging.error("Error downloading %s: %s, retrying in %ss", repo_id, e, wait_time) | |
| time.sleep(wait_time) | |
| attempt += 1 | |
| logging.error("Failed to download %s after %s attempts", repo_id, max_attempts) | |
| def download_openbench(): | |
| """Downloads pre generated data""" | |
| os.makedirs(DATA_ARENA_PATH, exist_ok=True) | |
| # download prev autogenerated leaderboard files | |
| download_dataset("Vikhrmodels/arena-leaderboard-metainfo", DATA_PATH) | |
| # download answers of different models that we trust | |
| download_dataset("Vikhrmodels/openbench-eval", DATA_ARENA_PATH) | |
| def build_leadearboard_df(): | |
| # Retrieve the leaderboard DataFrame | |
| with open(f"{DATA_PATH}/leaderboard.json", "r", encoding="utf-8") as eval_file: | |
| leaderboard_df = pd.DataFrame.from_records(json.load(eval_file)) | |
| return leaderboard_df.copy() | |