A Transformer model built from scratch to perform basic arithmetic operations, implementing multi-head attention, feed-forward layers, and layer normalization from the Attention is All You Need paper.