LSTM, from first principles
A character-level language model implemented from scratch in NumPy.
Learning the mechanics
Implemented a character-level LSTM language model from scratch with NumPy to learn sequential patterns in text.
Model setup
- 204,800 characters of training text
- Sequence length: 50
- Hidden size: 128
Reusability
Exported model weights and vocabulary for inference and reuse.