Transformer fundamentalsRapid Fire
Transformer Fundamentals 1
Practise explaining how attention, positional information and Transformer blocks contribute to a model's output.
Unlock with Full AccessThe round
This seven-question round moves through self-attention, multi-head attention and the components around them. You explain the purpose of the operations and how architecture choices affect which tokens a model can use. The questions connect these mechanisms to classification, generation and sequence-to-sequence tasks.
What you'll practise
Attention mechanics
Describe how queries, keys and values contribute to a representation.
Order and visibility
Explain how the model represents token order and which information attention may use.
Transformer blocks
Connect multiple heads, residual paths and normalisation to their roles in the architecture.
Get feedback on each answer when you finish the round.