Lesson 2: Three ways to arrange the same block
15 min read·13 Oct 2026
Module 6 · Lesson 2 · About 40 minutes · Needs: Lesson 1 (attention, the causal mask, encoder and decoder blocks)
The situation
Priya wants two things from the ticket "The vacuum stopped charging after 3 months." She wants it sorted into refund, shipping, warranty or other the moment it arrives, and she wants a draft reply for her agents. Dev has searched the Hugging Face Hub and come back with three names: BERT, GPT and T5. "Which one do we use?"
They are three arrangements of the block from Lesson 1, and each produces a different kind of output. This lesson runs one small model of each kind on that one ticket.