Comment by pjio
2 years ago
Layman question: Does the ability to "understand" such orders arise from one of the many layers which transform the tokens or is it hidden in the parameters/weights or is this due to some special type of attention head?
2 years ago
Layman question: Does the ability to "understand" such orders arise from one of the many layers which transform the tokens or is it hidden in the parameters/weights or is this due to some special type of attention head?
No comments yet
Contribute on Hacker News ↗