lesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blockslesson 3 of 5 · attention and transformer blocks
Complete one attention row: dot the query against every key (scaled by √dim), softmax into weights, mix the values.