A PyTorch implementation of a WaveNet-inspired character-level language model for name generation. This project demonstrates deep learning fundamentals through a clean, educational implementation that generates realistic names by learning character-level patterns.
- Custom Neural Network Implementation: Built from scratch with custom layers (Linear, BatchNorm, Embedding, etc.)
- WaveNet-Inspired Architecture: Hierarchical processing with consecutive flattening
- Character-Level Generation: Learns to generate names one character at a time
- Comprehensive Training Pipeline: Complete training, validation, and evaluation framework
- Modular Design: Clean, educational code structure perfect for learning
- Interactive Examples: Easy-to-use scripts for quick experimentation
After training, the model generates creative names like:
Generated Names:
─────────────────
1. Aria
2. Elena
3. Marcus
4. Zara
5. Kai
6. Luna
7. Davi
8. Nova
9. Leo
10. Maya
- Inspired by Andrej Karpathy's excellent neural network tutorials
- WaveNet architecture concepts from DeepMind's original paper
- Names dataset compiled from various public sources