Part 04 · LLM Internals

Transformer Architecture

Put the pieces together: attention, feed-forward layers, residuals and normalisation stacked into a decoder-only model, and where its billions of parameters actually live.

⏱ 1-2 hours · Station 27 of 92