Whenever you're ready, walk me through how you'd approach this.
Explain transformers.
Discuss the transformer architecture, self-attention, positional information, encoder and decoder blocks, training objectives, and why transformers differ from recurrent networks. Include a concise practical example showing how a transformer could be trained for sequence classification, and explain the main computational and modeling tradeoffs.