MyLLM is a deep-learning personal project where I built a modern LLM (LF_LLM-269M) from the ground up. I focused on developing the core components required for pre-training an LLM, including writing the model-architecture code, handling large datasets, training the model efficiently, and evaluating its performance.
For more info on the model itself and how it was train see https://github.com/LF-Luis/MyLLM.
Model was trained on 2024_12_07-05_14_UTC