LLM-from-Scratch

Building a GPT-style Large Language Model (LLM) from scratch in PyTorch step-by-step, featuring custom BPE tokenization, multi-head self-attention, and transformer architectures.

GitHub

Official source

Start locally with the CLI guide