After reading the Transformer paper and watching some explainer videos, many people still don’t really know what kind of code is actually running during GPT training. This project builds a modern LLM from scratch—attention mechanism, positional encoding, LayerNorm, residual connections, training loop, learning rate scheduling—each module is in its own file, and every single line of code comes with comments explaining the principles and intentions behind it. By the end, you won’t just know how to use libraries, but truly understand what’s happening at the matrix level. It’s perfect for beginners who want to dive deep into LLM principles, or developers with plenty of hands-on experience but gaps in foundational knowledge. It takes you from 'kind of understand' to 'understand every single line.'