The embedding model used is the default OpenAI API embedding, text-embedding-ada-002. GPT2 is only used during the tokenization process to calculate token lengths efficiently. I have updated my readme to reflect this information correctly.