Claude transformer_attention — practice questions
15 free MCQs with worked solutions. Tap any question for the answer + explanation, or practice them all in the app.
Practice Claude transformer_attention in the app →The Transformer architecture replaces recurrence and convolutions withScaled dot-product attention, per the paper, computes the output asThe encoder of the original Transformer uses how many identical layers?Multi-head attention uses how many heads in the base model?The decoder differs from the encoder by addingThe √d_k scaling in scaled dot-product attention exists toThe per-layer complexity of self-attention isWhat is the maximum path length in self-attention layers?Position-wise feed-forward networks in the Transformer are essentiallyThe chosen positional encoding scheme usesThe THREE distinct ways attention is used in the Transformer areMulti-head attention is preferred over a single d_model-dimensional head becauseWhy does the decoder need to mask future positions in self-attention?Residual connections and layer normalisation are applied asThe architectural property that lets self-attention parallelise where RNNs cannot is