create_dataset.py

import json
import argparse
import random
from pathlib import Path
from tqdm import tqdm
import datasets
from huggingface_hub import HfApi, RepoCard
from transformers import HfArgumentParser

random.seed(0)

def generate_unique_multiplication_data(a_max, b_max, n_train, n_test):
    """Generate train and test datasets for each multiplication range ensuring no overlap."""
    datasets = {}
    
    for a in range(1, a_max + 1):
        for b in range(1, b_max + 1):
            all_pairs = [(x, y) for x in range(1, a + 1) for y in range(1, b + 1)]
            
            # Convert to list before sampling to avoid Python 3.9+ warning
            test_data = set(random.sample(list(all_pairs), min(n_test, len(all_pairs))))  
            train_data = set(random.sample(list(set(all_pairs) - test_data), min(n_train, len(set(all_pairs) - test_data))))

            datasets[f"{a}x{b}"] = {"train": list(train_data), "test": list(test_data)}

    return datasets

def save_to_jsonl(data, file_path):
    """Save dataset to JSONL format."""
    with open(file_path, "w") as f:
        for a, b in data:
            json.dump({"problem": f"What is {a} times {b}?", "answer": str(a * b)}, f)
            f.write("\n")

def prepare_datasets(output_dir):
    """Prepare train and test datasets ensuring no overlap for all 1x1 to 15x15 combinations."""
    output_dir = Path(output_dir)
    output_dir.mkdir(parents=True, exist_ok=True)

    all_datasets = generate_unique_multiplication_data(a_max=15, b_max=15, n_train=1000, n_test=100)

    train_files, test_files = [], []
    for name, data in all_datasets.items():
        train_file = output_dir / f"multiplication_train_{name}.jsonl"
        test_file = output_dir / f"multiplication_test_{name}.jsonl"

        save_to_jsonl(data["train"], train_file)
        save_to_jsonl(data["test"], test_file)

        train_files.append(train_file)
        test_files.append(test_file)

    print(f"\n✅ Datasets saved to {output_dir}")
    return train_files, test_files

def process_file(file_path):
    """Convert JSONL data into Hugging Face dataset format."""
    with open(file_path, "r") as f:
        data = [json.loads(line.strip()) for line in f if line.strip()]

    dataset = {
        "messages": [[
            {"role": "user", "content": item["problem"]},
            {"role": "assistant", "content": item["answer"]},
        ] for item in data],
        "ground_truth": [item["answer"] for item in data],
        "dataset": ["multiplication"] * len(data),
    }
    return datasets.Dataset.from_dict(dataset)

def push_to_huggingface(train_files, test_files, hf_entity):
    """Push datasets to Hugging Face Hub and print the dataset link."""
    api = HfApi()
    hf_entity = hf_entity or api.whoami()["name"]

    print("\n📤 Uploading datasets to Hugging Face...\n")

    for file in train_files + test_files:
        dataset = process_file(file)
        dataset_name = file.stem
        repo_id = f"{hf_entity}/{dataset_name}"
        hf_url = f"https://huggingface.co/datasets/{repo_id}"

        print(f"✅ Dataset uploaded: {dataset_name}")
        # print(f"🔗 Click to view: {hf_url}\n")  # 👈 PRINTS THE LINK

        dataset.push_to_hub(repo_id)

        api.upload_file(
            path_or_fileobj=__file__,
            path_in_repo="create_dataset.py",
            repo_type="dataset",
            repo_id=repo_id,
        )

        # Add RepoCard with Hugging Face link
        repo_card = RepoCard(
            content=f"""\
# Multiplication Dataset - {dataset_name}

This dataset contains multiplication problems for numbers up to 15x15.

## Dataset Format

- `messages`: User question and assistant answer.
- `ground_truth`: Correct multiplication result.
- `dataset`: "multiplication"

## Hugging Face Dataset Link
➡️ [View dataset on Hugging Face]({hf_url})
"""
        )
        repo_card.push_to_hub(repo_id, repo_type="dataset")

def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--output_dir", type=str, default="math_data", help="Output directory")
    parser.add_argument("--push_to_hub", action="store_true", help="Upload to Hugging Face")
    parser.add_argument("--hf_entity", type=str, default=None, help="Hugging Face entity")
    args = parser.parse_args()

    train_files, test_files = prepare_datasets(args.output_dir)

    if args.push_to_hub:
        push_to_huggingface(train_files, test_files, args.hf_entity)

if __name__ == "__main__":
    main()