examples/README.md

# NexForge Tokenizer Examples

This directory contains example scripts demonstrating advanced usage of the NexForge tokenizer.

## Quick Start

### Basic Tokenizer Creation

```python
from nexforgetokenizer import build_tokenizer

# Create a tokenizer with default settings
build_tokenizer(
    input_dir="path/to/your/files",
    output_path="custom_tokenizer.json",
    vocab_size=40000,
    min_frequency=2
)
```

### Example Scripts

1. **Basic Example** (`basic_usage.py`)
   - Simple tokenizer creation and usage
   - Basic encoding/decoding
   - Vocabulary inspection

2. **Advanced Usage** (`advanced_usage.py`)
   - Custom special tokens
   - Batch processing
   - Performance optimization
   - Error handling

## Running Examples

```bash
# Install in development mode
pip install -e .

# Run basic example
python examples/basic_usage.py

# Run advanced example
python examples/advanced_usage.py --input-dir ../Dataset --output my_tokenizer.json
```

## Example: Creating a Custom Tokenizer

```python
from nexforgetokenizer import build_tokenizer

# Create a tokenizer with custom settings
build_tokenizer(
    input_dir="../Dataset",
    output_path="my_tokenizer.json",
    vocab_size=30000,      # Smaller vocabulary for specific domain
    min_frequency=3,        # Only include tokens appearing at least 3 times
    max_files=1000,         # Limit number of files to process
    special_tokens=["[PAD]", "[UNK]", "[CLS]", "[SEP]", "[MASK]"]
)
```

## Best Practices

1. **For General Use**
   - Use default settings (40k vocab, min_freq=2)
   - Process all files in your dataset
   - Test with the built-in test suite

2. **For Specialized Domains**
   - Adjust vocabulary size based on domain complexity
   - Consider increasing min_frequency for smaller vocabularies
   - Test with domain-specific files

## Need Help?

- Check the [main README](../README.md) for basic usage
- Review the test cases in `Test_tokenizer/`
- Open an issue on GitHub for support

## License

MIT License - See [LICENSE](../LICENSE) for details.