nanoGPT

nanoGPT is a streamlined repository for training and fine-tuning medium-sized Generative Pre-trained Transformers (GPTs), initially based on minGPT. Designed for simplicity and speed, it enables users to reproduce GPT-2 on a single node with minimal code overhead and includes straightforward mechanisms for training from scratch or fine-tuning using pre-trained models. Notably, it features an easy-to-read code structure, allowing modifications suited to individual needs, and facilitates quick setup for character-level text generation tasks.