A tiny 8.9M parameter causal language model with a custom Llama-style architecture and digit-aware BPE tokenizer. Enter a prompt and watch it continue the text!