Quiz: Applying Transformers for AI Text Summarization

In the T5 model, there is only an encoder stack, unlike the original transformer model, which has both an encoder and decoder stack.

The self-attention layer is order-independent in T5, unlike the original transformer model.

T5 uses relative position embeddings instead of adding arbitrary positions to the input.

Positional embeddings are shared and reevaluated through all the layers of the original transformer model, unlike the T5 model.