convert_to_jsonl.py

"""
Convert HuggingFace dataset Volko76/french-classic-conversations to JSONL format
with messages structure including system prompt.
"""

from datasets import load_dataset
import json
import os

# Configuration
SYSTEM_MESSAGE = "You are a helpful assistant."
OUTPUT_FILE = "french_classic_conversations.jsonl"

def main():
    # Load the dataset from HuggingFace
    print("Loading dataset from HuggingFace...")
    dataset = load_dataset("Volko76/french-classic-conversations")
    print(f"Dataset loaded: {len(dataset['train'])} rows")
    print(f"Columns: {dataset['train'].column_names}")
    
    # Check the structure of the first row
    sample = dataset['train'][0]
    print("\nSample row structure:")
    print(str(sample)[:1000])
    
    # Convert to JSONL format with system message
    print(f"\nConverting to JSONL format...")
    
    with open(OUTPUT_FILE, 'w', encoding='utf-8') as f:
        for row in dataset['train']:
            # Get the conversations from the row - it might be a JSON string
            conversations = row['conversations']
            if isinstance(conversations, str):
                conversations = json.loads(conversations)
            
            # Create the messages list with system prompt first
            messages = [{"role": "system", "content": SYSTEM_MESSAGE}]
            
            # Add the conversation messages
            for msg in conversations:
                messages.append({
                    "role": msg['role'],
                    "content": msg['content']
                })
            
            # Write as JSONL
            json_line = json.dumps({"messages": messages}, ensure_ascii=False)
            f.write(json_line + '\n')
    
    print(f"Conversion complete! Output saved to: {OUTPUT_FILE}")
    
    # Verify the output - read and display first few lines
    print("\n" + "="*60)
    print("First 2 entries from the output file:\n")
    with open(OUTPUT_FILE, 'r', encoding='utf-8') as f:
        for i, line in enumerate(f):
            if i >= 2:
                break
            data = json.loads(line)
            print(f"Entry {i+1}:")
            print(json.dumps(data, indent=2, ensure_ascii=False)[:1000])
            print("\n" + "-"*40 + "\n")
    
    # Count total entries and file size
    with open(OUTPUT_FILE, 'r', encoding='utf-8') as f:
        total_lines = sum(1 for _ in f)
    
    file_size = os.path.getsize(OUTPUT_FILE) / (1024 * 1024)  # MB
    
    print("="*60)
    print(f"Total conversations: {total_lines}")
    print(f"File size: {file_size:.2f} MB")

if __name__ == "__main__":
    main()