A pedogical implementation of a decoder-only transformer from scratch.
Install dependencies:
python -m venv diyt
source diyt/bin/activate
pip install -r requirements.txt
pip install -e .To be able to run jupyter notebooks:
pip install ipykernel
python -m ipykernel install --user --name diyt --display-name "Python (diyt)"Find data in data/harry_potter.txt
To train:
python src/diyt/pre_train.pySee notebook notebooks/playground.ipynb for generation.
TODO:
- Implement KV caching
