./nanogpt/prepare.py is provided to convert the dataset to the format used in nanoGPT

Use the huggingface transformer library to download the model checkpoint, or load it in a GPT2LMHeadModel from ./gpt2m/pytorch_model.bin

from transformers import GPT2LMHeadModel
model = GPT2LMHeadModel.from_pretrained("commaai/commavq-gpt2m")