LLama3 chat and train LLM from scratch This project aim is to use LLama-3.1-8B in order to predict masked words from Danish clinical text data.