LLMs
Transformer
A transformer is a sequence architecture built around attention and position-wise feed-forward blocks.
How it works
Attention exchanges information across positions, while feed-forward layers transform each position. Residual paths and normalization stabilize the stack.
A concrete example
Illustrative: 24 blocks process a sequence of 2,048 tokens at model width 2,048.
Do not confuse it with RNN
a transformer processes attention relationships directly; an RNN carries a recurrent state step by step.