Attention and the Transformer

This practicum introduces self-attention, hidden-layer representations, and the key-value store paradigm. It uses these ideas to interpret a basic transformer, trace its forward pass in PyTorch, and compare encoder-decoder with sequential architectures.