Datasets:
Tasks:
Other
Languages:
English
Multilinguality:
monolingual
Size Categories:
100M<n<1B
ArXiv:
Tags:
License:
| import glob | |
| import json | |
| import os | |
| from collections import defaultdict | |
| _DATA_PATH = "data" | |
| data_split_sizes = defaultdict(dict) | |
| for stats in glob.glob(f"{_DATA_PATH}/*/stats.*.json"): | |
| folder_path = os.path.dirname(stats) | |
| task_name = folder_path.split("/")[-1] | |
| split_name = os.path.basename(stats).split(".")[1] | |
| if not os.path.exists(f"{folder_path}/COMPLETED"): | |
| continue | |
| with open(stats, "r") as f: | |
| split_stats = json.load(f) | |
| nb_examples = split_stats["examples"] | |
| if nb_examples > 0: | |
| data_split_sizes[task_name][split_name] = nb_examples | |
| with open("data_split_sizes.csv", "w", encoding="utf=8") as f: | |
| # The file is now merged into `tasks_splits_and_features.py` | |
| f.write("Data(sub)set|Number of examples per splits\n") | |
| for task_name in sorted(list(data_split_sizes.keys())): | |
| split_sizes_dict = json.dumps(data_split_sizes[task_name]) | |
| f.write(f"{task_name}|{split_sizes_dict}\n") | |