generate_dataset_files.py

#!/usr/bin/env python3
import os
import json
import glob
from tqdm import tqdm

# Directory containing the PDF files
data_dir = "data"

# Get all PDF files
pdf_files = glob.glob(os.path.join(data_dir, "*.pdf"))

# Create dataset structure
dataset = {
    "file_name": [],
    "file_path": [],
    "file_size": [],
    "content_type": []
}

# Add information for each PDF file
for pdf_file in tqdm(pdf_files, desc="Processing PDF files"):
    file_name = os.path.basename(pdf_file)
    file_path = pdf_file
    file_size = os.path.getsize(pdf_file)
    
    dataset["file_name"].append(file_name)
    dataset["file_path"].append(file_path)
    dataset["file_size"].append(file_size)
    dataset["content_type"].append("application/pdf")

# Create the default directory if it doesn't exist
os.makedirs("default", exist_ok=True)

# Write the dataset to a JSON file
with open("default/train.json", "w") as f:
    json.dump(dataset, f, indent=2)

# Update dataset_dict.json with correct information
dataset_dict = {
    "default": {
        "splits": {
            "train": {
                "name": "train",
                "num_bytes": sum(dataset["file_size"]),
                "num_examples": len(dataset["file_name"]),
                "dataset_name": "jfk-files-2025"
            }
        },
        "download_checksums": {},
        "download_size": sum(dataset["file_size"]),
        "post_processing_size": None,
        "dataset_size": sum(dataset["file_size"]),
        "size_in_bytes": sum(dataset["file_size"])
    }
}

# Write the dataset_dict.json file
with open("dataset_dict.json", "w") as f:
    json.dump(dataset_dict, f, indent=2)

print(f"Successfully processed {len(dataset['file_name'])} PDF files")
print(f"Total dataset size: {sum(dataset['file_size']) / (1024*1024):.2f} MB")