# Data source
Downloaded via Andrej Karpathy's nanogpt repo from this [link](https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt)

# Data Format
- The entire dataset is split into train (90%) and test (10%).
- All rows are at most 1024 tokens, using the Llama 2 tokenizer.
- All rows are split cleanly so that sentences are whole and unbroken.