Chapter 6: Embeddings, the Forward Pass, and the Loss Function

Give tokens and positions learned vector identities, assemble a minimal forward pass to logits, and compute cross-entropy loss.